The Power of Asymmetry: Architecture and Assembly of the Gram-Negative Outer Membrane Lipid Bilayer

Abstract

Determining the chemical composition of biological materials is paramount to the study of natural phenomena. Here, we describe the composition of model gram-negative outer membranes, focusing on the predominant assembly, an asymmetrical bilayer of lipid molecules. We also give an overview of lipid biosynthetic pathways and molecular mechanisms that organize this material into the outer membrane bilayer. An emphasis is placed on the potential of these pathways as targets for antibiotic development. We discuss deviations in composition, through bacterial cell surface remodeling, and alternative modalities to the asymmetric lipid bilayer. Outer membrane lipid alterations of current microbiological interest, such as lipid structures found in commensal bacteria, are emphasized. Additionally, outer membrane components could potentially be engineered to develop vaccine platforms. Observations related to composition and assembly of gram-negative outer membranes will continue to generate novel discoveries, broaden biotechnologies, and reveal profound mysteries to compel future research.

INTRODUCTION

A defining feature of diderm, gram-negative bacteria is a second membrane bilayer that surrounds the peptidoglycan layer (Figure 1). Unlike the inner bacterial membrane, this outer membrane is an asymmetrical lipid bilayer (66), with the periplasmic leaflet composed of glycerophospholipid (PL) and the surface-exposed outer leaflet consisting of lipopolysaccharide (LPS) (Figure 1). This unique membrane organization affords gram-negative organisms protection not only from large polar molecules restricted by a typical membrane bilayer but also from lipophilic compounds. Evidence for the asymmetrical organization of the outer membrane was reported by Kamio & Nikaido (66) in the mid-1970s. They demonstrated that PLs of intact bacteria are not susceptible to degradation by phospholipases or to chemical labeling by macromolecular reagents unable to cross the outer membrane. Overall this suggested PLs were confined to the inner leaflet of the outer membrane.

Figure 1.

The gram-negative cell envelope and major lipids of Escherichia coli. (a) The typical inner membrane and outer membrane bilayers are separated by the periplasmic compartment, which contains the peptidoglycan layer. The inner membrane is a bilayer of glycerophospholipids, whereas the outer membrane is an asymmetrical bilayer, with glycerophospholipids found in the inner periplasmic leaflet and lipopolysaccharide localized to the outer-surface-exposed leaflet. (b) The major lipids of E. coli K12 are 3-deoxy-d-manno-oct-2-ulosonic acid (Kdo)₂–lipid A, phosphatidylethanolamine, phosphatidylglycerol, and cardiolipin. The length and composition of fatty acyl chains of each major lipid type, as well as the number of molecules within a bacterial cell during exponential growth phase, are indicated.

Given that the outer membrane is essential for bacterial growth, protects the bacterium from environmental stresses (e.g., antibiotics), contains components that activate the innate immune system, and interfaces directly with the surrounding environment, it is of no surprise that it has been the subject of intense research. Over the last two decades, there has been enormous progress toward understanding outer membrane assembly and maintenance of bilayer asymmetry. One key question has always been how lipophilic components, assembled at the inner face of the cytoplasmic membrane, are guided through the crowded, aqueous environment of the periplasm for assembly into the outer membrane bilayer. Other areas of intense focus include biosynthesis and remodeling of key outer membrane lipid components, development of antimicrobials targeting outer membrane assembly, and manipulation of outer membrane components in biotech applications.

BIOSYNTHESIS OF A DIVERSE GLYCEROPHOSPHOLIPID REPERTOIRE

Eugene Kennedy and members of his laboratory combined molecular genetics and biochemical approaches to identify and characterize the genes responsible for PL metabolism in Escherichia coli and Saccharomyces cerevisiae over 40 years ago (41, 67, 103). E. coli PL biosynthesis is a paradigm for most prokaryotes (Figure 2), whereas the S. cerevisiae pathway generally typifies related fungal species. Both model organisms contain elements of more complex lipid anabolism observed in higher-order eukaryotes. Today, genomic information combined with unprecedented sensitivity in detection of lipid species by means of mass spectrometry has enabled identification of previously unknown genes involved in lipid metabolism, even in well-trekked model organisms. Exciting discoveries in E. coli include a third phosphatidylglycerol-3-phosphate (PGP) phosphatase and two novel cardiolipin synthases (Figure 2); in yeast and humans a new PGP phosphatase and a mitochondrial cytidine diphosphate diacylglycerol synthetase have been identified (80, 97, 125, 126, 152). Mutations in these eukaryotic genes lead to severe mitochondrial dysfunction, as a consequence of cardiolipin loss. However, despite the existence of three differentially regulated cardiolipin synthase homologs in E. coli, the biological role of cardiolipin in this model bacterium remains poorly understood.

Figure 2.

The Kennedy pathway for glycerophospholipid biosynthesis in Escherichia coli. Eugene Kennedy and coworkers discovered the critical role of cytidine nucleotides as activating groups for subsequent phosphatidyl-transfer reactions in glycerophospholipid biosynthesis. Enzymes involved in these pathways are either intrinsic membrane proteins or membrane associated. Phosphatidic acid, de novo synthesized on the cytoplasmic surface of the inner membrane, is converted to cytidine diphosphate diacylglycerol (CDP-DAG) by CDP-DAG synthase (CdsA) in a reaction that utilizes cytidine triphosphate (CTP) with release of pyrophosphate. CDP-DAG functions as a donor of phosphatidyl moieties for the biosynthesis of phosphatidylserine and phosphatidylglycerol-3-phosphate (PGP) with release of cytidine monophosphate (CMP). The PS formed by phosphatidylserine synthase (PssA) is rapidly decarboxylated by phosphatidylserine decarboxylase (Psd) to generate the predominant glycerophospholipid (~70%) found in E. coli, phosphatidylethanolamine. In phosphatidylglycerol (PG) synthesis, PGP is quickly dephosphorylated by one of several inner membrane phosphatases (e.g., PgpA) to form PG. Lastly, cardiolipin is synthesized through the condensation of two PG molecules by cardiolipin synthase A, B, or C (ClsA, ClsB, or ClsC) (ClsA being the major cardiolipin synthase). Enzyme names are in blue, lipid species in brown, and cofactors in black.

Cardiolipin’s enigmatic biological function in bacteria is not exceptional. The physiological importance of bacterial PL headgroup heterogeneity remains ill defined; however, a few consistent themes emerge from various insightful reports. The topology of integral membrane proteins is affected by interaction with specific lipid partners whereas, perhaps not surprisingly, numerous peripheral membrane proteins associate preferentially with anionic PLs—such as phosphatidic acid, phosphatidylglycerol (PG), and cardiolipin—rather than zwitterionic phosphatidylethanolamine (PE) (12, 42, 84, 142, 153). Protein interactions with anionic PL have been implicated as necessary for SecA-dependent protein translocation and oriC-dependent DNA replication (20, 35, 54, 79, 148). It is also reported that PLs serve as precursors in the chemical modification of sugars, proteins, and even other lipids. In many organisms, the diacylglycerol moiety of PG is used in the posttranslational modification of lipoproteins, a chemical modification required for proper insertion of lipoproteins into the membrane (16, 59, 88, 130, 147). During processing, lipoproteins are also aminoacylated at an N-terminal cysteine residue, where PLs serve as substrate acyl chain donors (88, 116). Membrane-derived oligosaccharide, an enigmatic osmoregulatory component found in the periplasm of some gram-negative bacteria, can be modified with PG and PE headgroups (13, 63, 64, 86). The Kdo–lipid A domain and core oligosaccharide of LPS can also be modified with PL headgroups or acyl chains (11, 92, 99, 105). Bacteria with these membrane modifications show increased resistance to cationic antimicrobial peptides, especially polymyxins that bind Kdo–lipid A specifically, but also an ever-growing list of other phenotypes (92) (see “Biological Significance of Chemically Altered Kdo–Lipid A Domains”).

In efforts to structurally characterize bacterial membrane lipids, lipidomics across known bacterial taxa have uncovered a panoply of lipid species (an area of research covered in greater depth elsewhere; 118). This array of molecular composition somewhat undermines the universality of the E. coli paradigm. It has become increasingly evident that lipid diversity in nature is exquisite and complex. Study of this diversity, and how chemical composition informs biological function, will surely continue to yield unique insights into our natural world.

PL TRANSPORT—THE UNKNOWN OUTER MEMBRANE ASSEMBLY APPARATUS

PL transport to the outer membrane presents a unique challenge to gram-negative bacteria. Lipoproteins and β-barrel outer membrane proteins contain specific amino acid signal sequences for export to the outer membrane, whereas LPS is strictly compartmentalized to the outer membrane surface by dedicated transport machinery. On the contrary, PLs appear randomly distributed between the inner bilayer and the periplasmic outer membrane leaflet, with no major observable difference in PL composition. The process by which PLs are distributed between sites of synthesis on the inner membrane to the periplasmic leaflet of the outer membrane is unknown. Transport machinery exists for other outer membrane components, so an analogous mechanism likely exists to distribute the cellular PL pool between inner and outer membranes. The rate of spontaneous, passive transfer of PLs between separated membranes in vitro is not sufficient to support bacterial growth (40); thus, an active mechanism is likely needed to facilitate movement of PLs across the periplasmic compartment.

Tracking Anterograde PL Transport to the Outer Membrane

The most detailed account of PL transport from the inner to outer membrane was written by Donohue-Rolf & Schaechter in 1980 (40). To determine the in vivo translocation rate of newly synthesized PLs, they used pulse-chase radiolabeling of cellular PL pools to track and compare PLs between inner and outer bilayers. The results indicated distinct rates of PL transportation, with rapid transport of PG or cardiolipin and slower rates of PE transport to the outer membrane (Figure 1). Depletion of cellular ATP and inhibition of lipid or protein synthesis did not appear to alter PE translocation rates, but inhibitors that target proton motive force significantly reduced the PE translocation rate. The authors postulated that zones of adhesion, or Bayer bridges, between the outer membrane and inner membrane provide a route for phospholipid movement between membranes and are controlled by the proton motive force (8). However, whether these zones exist in vivo is controversial (70), and despite these works no essential anterograde transport mechanisms have been characterized.

Conditionally lethal mutants of MsbA accumulate LPS precursors and PLs in the inner membrane, leading some to conjecture that MsbA serves as a general flippase for both lipid molecules (37, 38). Further characterization has clearly shown MsbA is responsible for flipping nascent LPS intermediates from the cytoplasmic to periplasmic leaflet of the inner membrane. The role of MsbA in general PL transport is less clear, complicated by the fact that accumulation of LPS precursors could affect an unknown, independent PL transport mechanism. Evidence in support of this idea came from a study in lipooligosaccharide (LOS)-deficient Neisseria meningitides that convincingly demonstrated that MsbA is not required for PL distribution to the outer membrane (128).

The study of proteins that alter wild-type ratios of PLs, specifically in the outer membrane compartment, could prove beneficial in identifying putative PL transport mechanisms. Dalebroux et al. observed that activation of the PhoPQ two-component virulence regulatory system in Salmonella enterica serovar Typhimurium increases relative concentrations of cardiolipin in the outer membrane (32, 33). This change in lipid content requires the membrane-tethered PbgA (YejM), which contains an integral membrane domain and a periplasmic cardiolipin-binding domain. PhoPQ activation in S. enterica promotes direct interaction of the PbgA periplasmic domain with the outer membrane, leading the authors to hypothesize that PbgA can serve as a cardiolipin transporter upon PhoPQ activation. Conservation of PbgA is limited even among proteobacteria; however, the existence of PbgA-like analogs cannot be ruled out. Broadened searches of genes responsible for perturbed PL content will continue to identify machinery bacteria employ to transit bulk PLs from the inner membrane to the outer membrane.

Outer Membrane PL Removal via Retrograde Transport

The Mla (maintenance of outer membrane lipid asymmetry) pathway is an intermembrane transport system proposed to function in the removal of accumulated PLs in the outer membrane by transporting them to the inner membrane (81). The proteins that make up the Mla system contain elements similar to those of the Lol lipoprotein transport apparatus (150) (Figure 3). Mla is made up of an ABC transporter (MlaFEDB), a periplasmic protein (MlaC), and a lipoprotein tethered to the outer membrane (MlaA). Deletion of any Mla component leads to outer membrane defects quantifiable by SDS-EDTA sensitivity. As reported by Malinverni & Silhavy (81), deletion of any one Mla component results in identical sensitivity to SDS-EDTA, indicative that they comprise a single pathway. Shigella flexneri with mutations in Mla homologs fails to form plaques after invasion of human host cells, a proxy for intracellular spread (19). Like E. coli mla mutants, these Shigella mla mutants are sensitive to SDS-EDTA.

Figure 3.

Comparison of the Mla and Lol periplasmic transport systems. (a) The proposed model of the retrograde Mla (maintenance of outer membrane lipid asymmetry) system that is responsible for the removal of mislocalized glycerophospholipids from the gram-negative outer membrane. It is unclear whether the Mla system removes excess glycerophospholipids (e.g., phosphatidylethanolamine) from the inner leaflet prior to migration to the cell surface or if the system removes glycerophospholipids directly from the outer leaflet. It has been proposed that OmpC functions in a complex with the lipoprotein MlaA to extract phospholipids from the outer leaflet (22). Once extracted, the targeted lipid is delivered to the soluble periplasmic transporter MlaC, which delivers the substrate to the MlaFEDB ABC transporter complex. Using energy from ATP hydrolysis, the targeted lipid is reinserted into the inner membrane; however, the fate of the lipid is unknown. The biochemical mechanisms of substrate recognition, membrane extraction, and membrane insertion remain to be elucidated. (b) Schematic representation of the Lol (localization of lipoproteins) transport system. The LolCDE ABC transporter complex utilizes ATP to direct the movement of mature (i.e., acylated) lipoproteins destined for the outer membrane by LolA. The LolA-lipoprotein complex transports target lipoproteins across the periplasm to LolB, which inserts lipoproteins into the outer membrane. It is unclear how surface-exposed lipoproteins are transported to the outer leaflet of the outer membrane.

Suppressor mutations that repair the defective outer membrane phenotype (SDS-EDTA sensitivity) arise in the promoter region of pldA, leading to overexpression of the encoded outer membrane phospholipase A (81). PldA has long been thought to contribute to outer membrane lipid asymmetry under certain growth conditions, by degrading PLs that have miscompartmentalized to the outer leaflet of the outer membrane (10, 36, 117). PldA mutants do not accumulate PLs in the outer membrane outer leaflet to the dramatic extent seen in Mla pathway mutants, indicating the necessity of Mla-based maintenance of outer membrane asymmetry. It is important to note, however, that no biochemical evidence directly supports Mla retrograde PL transport from the outer membrane to the inner membrane.

Direct physical interaction between the OM lipoprotein, MlaA, and the β-barrel protein OmpC has been observed, and it is speculated that OmpC may provide a β-barrel-based channel for MlaA to reach PLs in the outer leaflet (Figure 3) (22). This hypothetical protein-interaction model draws parallels to the LptDE plug-and-barrel system used to span the outer membrane bilayer in export of LPS to the bacterial surface (50, 56) (Figure 4). OmpC might instead function in interleaflet PL transport, shuttling outer leaflet PLs to the inner leaflet, where they become accessible to periplasmic MlaA. It is also possible that MlaA is deposited on the bacterial surface, or spans across the bacterial outer membrane, as has been suggested with other lipoproteins (28, 100). A recent mutation in MlaA was reported by Silhavy and colleagues, with a variety of interesting phenotypes that are bound to provide a more robust understanding of the Mla pathway upon further mechanistic examination (124). To determine the exact mechanism of MlaA, a more definitive elucidation of its outer membrane leaflet distribution is needed. Moreover, interaction characterization, of and between Mla proteins, will further improve our understanding as to how Mla controls outer membrane PL composition.

Figure 4.

Biosynthesis and transport of lipopolysaccharide (LPS) in Escherichia coli. Biosynthesis of the canonical Kdo₂–lipid A substructure of LPS is required for the growth of most gram-negative bacteria. It is synthesized via the nine-step Raetz pathway. All reactions in the pathway are catalyzed by a single enzyme (red) and occur at the interface between the cytosol and the inner membrane. Acyl-ACP (acyl carrier protein) is the preferred donor for each acylation event, and each acyltransferase utilizes an active site hydrocarbon ruler providing acyl chain specificity. Although Kdo is technically part of the core oligosaccharide, Kdo transfer is required for lipid A biosynthesis because the final two steps, catalyzed by LpxL and LpxM, require the presence of covalently attached Kdo. The remaining core oligosaccharide is extended at the cytoplasmic face of the inner membrane, requiring various glycosyl transferases (not shown). MsbA, the core Kdo–lipid A domain flippase, functions as an ABC transporter, moving core Kdo–lipid A domains to the periplasmic face of the inner membrane. For simplicity, the addition of O-antigen that typically occurs in the periplasm is not shown. The intermembrane translocation of mature LPS is carried out by the Lpt (LPS transport) system, which forms an envelope-spanning translocation machine. LPS is removed from the inner membrane via the ABC transporter LptBFG and delivered to LptC. Along with the periplasmic domain of LptC, soluble LptA forms a periplasmic bridge with the N-terminal domain of the outer membrane protein LptD. LptE sits within the β barrel formed by the C-terminal domain of LptD and promotes passage to the bacterial surface.

GRAM-NEGATIVE BIOSYNTHESIS OF ENDOTOXIC KDO–LIPID A DOMAINS

The genes and molecular intermediates along the near-completely conserved biosynthetic pathway for Kdo–lipid A assembly in gram-negative bacteria have been characterized over a span of 30 years (Figure 4). Pathway enzymes have been comprehensively characterized through crystallography and classical enzymology. Nine enzymes are required for canonical (Kdo)₂–lipid A biosynthesis in E. coli (105) and are considered essential for growth in all gram-negative bacteria, with the exception of the late-step acyltransferases LpxL and LpxM and their various homologs (60, 133, 139). Although not all organisms contain high-identity homologs to each gene in the E. coli pathway, each enzymatic step is universally conserved by organism-specific functional analogs. For example, UDP-2,3-diacyl-GlcN hydrolase activity (Figure 4) is catalyzed by LpxH in most gamma- and betaproteobacteria, whereas LpxI performs this role in alphaproteobacteria and other bacterial groups (145). LpxG, a third enzyme of distant homology to both LpxH and LpxI and capable of UDP-2,3-diacyl-GlcN hydrolase activity, was recently identified in Chlamydia trachomatis (151). A very similar distribution pattern is seen in lipid A late acyltransferases, where LpxJ substitutes for LpxM in many organisms (109) (Figure 4). Departures from the canonical biosynthetic structure include variations in acyl chain length and degree of saturation, important for homeoviscous adaptation, as controlled by selective hydrocarbon ruler domains of LpxA or LpxD and their functional equivalents (4, 7, 18, 104). Nuanced selectivity in a number of Kdo sugars, and modifications thereof (e.g., phosphorylation or hydroxylation), represent other organism-specific variations to the canonical biosynthetic route (23, 144).

Compounds that inhibit early steps in Kdo–lipid A domain synthesis are candidates for broad-spectrum antibacterials. However, impediments to the successful in vivo implementation of Raetz pathway inhibitors are problematic in antibacterial development in general, e.g., narrow-spectrum antibacterial activity, insolubility, and off-target protein binding. Compound screens have successfully identified competitive inhibitors of early-step acyltransferases LpxA and LpxD (65), and a sulfonyl piperazine inhibitor of LpxH was recently discovered (90). These compounds are effective against specific gram-negative taxa. Given that LpxC catalyzes the committed step for Kdo–lipid A biosynthesis, and has no homology to any protein domain in humans, LpxC inhibitors remain the most promising broad-spectrum target (2, 78). For many years CHIR-090 has been explored as a powerful inhibitor of many bacterial LpxC homologs (5, 6, 74, 85). A fine-tuned understanding of the interaction between CHIR-090 and LpxC has led to the design of dozens of novel compounds with differing specificity and increased potency to various LpxC homologs (17, 74–77, 82, 131). Compounds that inhibit the synthesis or transfer of Kdo to tetraacylated lipid A have also been described; however, their in vivo efficacy is unclear (9, 24–26, 101, 149).

FINAL ASSEMBLY AND TRANSPORT OF LPS TO THE OUTER MEMBRANE BILAYER

Core oligosaccharide is assembled stepwise on nascently synthesized Kdo–lipid A domains by specific glycosyltransferases that utilize activated sugar nucleotides. Upon completion MsbA translocates core Kdo–lipid A from the inner leaflet to the outer leaflet of the inner membrane, in an ATP-dependent manner (Figure 4). As translocation from the inner to outer leaflet appears essential, MsbA is a potential target for antimicrobial development. Characterization of MsbA in vitro also suggests it may serve as a capable multidrug efflux pump (43, 107, 146). To our knowledge no compounds that specifically block MsbA in vivo have been reported.

In the periplasmic space, O-antigen is ligated onto core Kdo–lipid A to form fully mature LPS (not shown in Figure 4; reviewed in Reference 145). In E. coli, transport of LPS onto the surface of the outer membrane requires a seven-protein complex (Figure 4). Inner membrane ABC transporter LptBFG is required for LPS extraction from the periplasmic leaflet (57, 89, 110). This protein complex interfaces with an LPS-binding protein, LptC, to transfer LPS from the periplasmic leaflet of the inner membrane to a periplasmic transporter, LptA (89, 96, 132). LptA forms a periplasm-spanning filament bridging the inner to the outer membrane (21, 51). At the outer membrane, LptA transfers LPS to the third and final LPS-binding protein, LptE, wherein LptD and LptE form a plug-and-barrel structure for final deposition of LPS into the outer leaflet of the outer membrane (50, 56). How large LPS structures are threaded through the LptDE complex is an interesting and unanswered question. This physical constraint is also seen with outer membrane β-barrel assembly machinery (BAM) that helps properly fold outer membrane β-barrel proteins. Substrate β-barrel proteins are often larger than the β-barrel component of BAM complexes. It has been suggested that LptDE, BAM, and related complexes contain a gated lateral opening to allow intramembranous transit of bulkier substrate molecules, as supported by experimental evidence (39, 58, 94). Given that components of the Lpt machinery are on the bacterial surface, and appear essential for growth, LPS transport is a viable target in the development of new antibacterial agents. Compounds that specifically target LptD in Pseudomonas aeruginosa (122) or LptB of E. coli have already been identified (114, 134, 143).

BIOLOGICAL SIGNIFICANCE OF CHEMICALLY ALTERED KDO–LIPID A DOMAINS

Most bacteria modify the basic Kdo–lipid A structure to optimize outer membrane integrity for survival in a given niche, and pathogens remodel their outer membrane in response to environmental cues so as to enhance virulence. In general, Kdo–lipid A modifications involve changes in the number or composition of acyl chains, phosphate groups, or any of various covalently attached functional groups (92, 105). The Kdo–lipid A domain from S. enterica serovar Typhimurium serves as an exemplary model of the full range of modifications one organism may contain (Figure 5).

Figure 5.

Chemical modifications of the Kdo₂–lipid A domain of Salmonella enterica. The Kdo₂–lipid A domain of Salmonella spp. can be highly modified by highly regulated enzymatic machinery (reviewed in 92, 145). Arrows indicate either the addition or the removal of acyl chains. Numbers indicate positions on the disaccharide portion of lipid A. Addition of free-amine-containing residues, 4-amino-4-deoxy-l-aminoarabinose (l-4-aminoarabinose) and phosphoethanolamine, or the fatty acid palmitate promotes resistance to antimicrobial peptides. The removal of acyl chains is associated with reduction of endotoxicity and TLR4/MD-2 activation.

Chemical modifications to LPS typically involve smaller chemical alterations of less than 200 Da. Larger chemical substitutions have been discovered (>500 Da), such as the covalent attachment of hopanoids to very long chain fatty acids of the Kdo–lipid A domain of Bradyrhizobium LPS (Figure 6) (115). Biophysical experiments in liposomes show that hopanoid-linked lipid A helps stabilize model asymmetric bilayers, whereas in nature hopanoid modification is necessary for the favorable environmental association with the bacteria’s leguminous symbiont. Interestingly, strains of Bradyrhizobium that were deficient for general hopanoid biosynthesis were less able to maintain healthy symbiosis with host Aeschynomene evenia legumes (115). A better understanding of how hopanoid–lipid A affects symbiosis could inform agricultural bioengineers interested in reducing our dependence on the environmentally and economically unsustainable use of chemical fertilizers.

Figure 6.

Kdo–lipid A domains of Bacteroides thetaiotaomicron, Helicobacter pylori, and Bradyrhizobium japonicum. The major Kdo–lipid A species presented on the surface of each organism is shown. Numbers indicate the length of acyl chains. The red box around the hopanoid hydroxyl group indicates the attachment site to indicated acyl chains on the Kdo–lipid A domain of the organism.

Provocative studies reveal the dynamic relationship between surface features of human commensals and host immunity, especially within the gastrointestinal tract, which contains more than 100 trillion bacteria. Of prime interest are preventive mechanisms that limit overstimulation of proinflammatory receptor-mediated pathways by resident gastrointestinal commensals. One such mechanism involves only sensing bacteria that invade past the luminal epithelial barrier, where resident professional immune cells are most abundant. Expression of TLR4/MD-2, the pattern-recognition receptor that specifically binds to and is activated by Kdo–lipid A, is localized to the basolateral surface of associated enterocytes (136), limiting unnecessary inflammatory responses to bulk commensals in the gastrointestinal lumen. Similarly, Paneth cells residing in intestinal crypts direct antimicrobial responses to TLR4/MD-2 activation (135). Some commensals appear to disguise their surface; for example, Bacteroides thetaiotaomicron expresses a lipid A phosphatase (LpxF). Dephosphorylation simultaneously provides resistance against host luminal antimicrobial peptides and produces a lipid A molecule that weakly stimulates TLR4/MD-2 (31) (Figure 6). Similar chemical remodeling of the cell surface may be a general commensal strategy, given that a modified lipid A species, dephosphorylated and underacylated, is displayed on the surface of the obligate, gastric bacterium Helicobacter pylori (30) (Figure 6). Further understanding of microbial-host homeostasis in the gastrointestinal tract will inform treatment options for inflammatory bowel syndromes due to diet, pathogen attack, or chemotherapeutic/antibiotic protocols. Other human niches prone to disease caused by perturbations in microbiome populations (such as the mouth, skin, and respiratory tract) also warrant deeper investigation (27, 44).

GRAM-NEGATIVE BACTERIA WITHOUT ENDOTOXIN

Biosynthesis of the Kdo–lipid A domain and subsequent transport of LPS/LOS onto the bacterial surface were thought to be essential for survival of gram-negative bacteria (14, 15, 110, 119–121). The first challenge to the notion of essentiality was mounted in 1998, when lpxA in N. meningitidis was inactivated by directed mutagenesis to produce a viable LOS-deficient mutant (123). Seven years later, Peng et al. (98) engineered an lpxA-deficient strain of Moraxella catarrhalis, also devoid of endotoxin. Without employing directed mutagenesis, Moffatt et al. (87) isolated LOS-deficient Acinetobacter baumannii using high concentrations (>10 μg/mL) of colistin (polymyxin E). Alterations in the composition and compartmentalization of PLs and outer membrane lipoproteins appear to be emergent properties of LOS-deficient bacteria, but more sophisticated comparative analyses are needed. In addition to these organisms, the obligate intracellular pathogen C. trachomatis appears to survive inactivation of the Kdo–lipid A biosynthetic pathway when treated with an LpxC inhibitor (93). Loss of LOS correlates with an inability of the bacterium to transition into the infectious stage of the chlamydial life cycle. It is worth noting that all these pathogens synthesize a smaller LOS glycoform that lacks the O-antigen domain found in canonical LPS structures.

Whereas some organisms survive inactivation of Kdo–lipid A biosynthetic genes, reports indicate other gram-negative bacteria naturally lack LOS/LPS altogether. Sphingomonas paucimobilis does not appear to produce LOS/LPS but instead produces glycosphingolipids, which are typically found in eukaryotes (68). Glycosphingolipid compartmentalization to the outer membrane was found on fractionation of S. paucimobilis membranes, suggesting that these glycolipids could substitute as divergent chemical analogs to LOS/LPS (69). Pathogenic spirochetes like Borrelia burgdorferi and Treponema pallidum also lack the necessary genes for Kdo–lipid A biosynthesis (48, 49). Work to characterize the surface of these pathogens, which naturally lack outer membrane β-barrel proteins, has shown an enrichment of lipoproteins in the outer membrane, similar to what transcriptomic data suggest for A. baumannii (29). It appears that LPS/LOS is not essential for some gram-negative bacteria wherein diverse, organism-specific compensatory mechanisms are required for survival.

LPS-deficient gram-negative bacteria have enormous potential for biotechnology applications, including (a) endotoxin-free production of recombinant proteins or antibiotics, (b) bioengineering of more effective probiotics and (c) production of enhanced vaccines.

ENGINEERING OUTER MEMBRANE COMPONENTS FOR APPLIED TECHNOLOGIES

Detoxified LPS—A Classic Adjuvant Refurbished for Modern Vaccines

Because of their immunostimulatory properties, Kdo–lipid A domains have been evaluated as potential vaccine adjuvants; however, inherent toxicity of the canonical hexaacylated bis-phosphorylated structure precludes its use in commercial vaccines without prior detoxification. This barrier was first overcome by Ribi and coworkers, who subjected S. enterica serovar Minnesota R95 to successive acid and base hydrolysis to detoxify LPS, a process that yields monophosphorylated lipid A (MPL) (102). The current clinical grade MPL adjuvant approved and licensed by the US Food and Drug Administration is a chemically heterogeneous mixture of several lipid A species (53). The primary lipid A species in the MPL adjuvant, 3-O-deacyl-4′-monophosphoryl A, induces a robust adaptive immune response with bias toward a less inflammatory TLR4-mediated signaling pathway, rendering the adjuvant mixture 100-fold less toxic than wild-type LPS, and is the first adjuvant used that is capable of activating effector T cell responses (62, 83, 95, 111, 129).

Production of the MPL adjuvant relies on costly, harsh chemical treatments. Alternatives to LPS detoxification protocols include engineering bacterial strains to produce MPL in vivo, or Kdo–lipid A domains with similarly desirable adjuvant properties (3, 34, 91). A combinatorial recombinant library of 61 E. coli strains was recently generated to produce homogenous mixtures of Kdo–lipid A domains, which display a gamut of TLR4/MD-2 agonist functions, as well as graded effects on immune system models in mice (91). Libraries of this nature have tremendous potential in the design of rational adjuvants to improve vaccines, not only for humans but also for species throughout the animal kingdom, particularly livestock. To improve on rationally designed libraries, metagenomic profiling and biochemical characterization of commensal organisms should help uncover additional mechanisms to produce LPS variants with a range of adjuvant potential (i.e., immunostimulatory but not toxic).

Design of Enhanced Vaccine Platforms—Outer Membrane Vesicles

Current barriers to the practical use of naturally derived outer membrane vesicles (OMVs) in humans (see sidebar about Designer Outer Membrane Vesicles; Figure 7) will be overcome by insights into mechanisms that drive OMV formation, improved commercial production pipelines, and clever application of bioengineering techniques derived from discoveries in basic science (1, 55, 72, 73, 113). A general mechanism that contributes to OMV production in gram-negative bacteria, regulated by Mla components, may have been recently identified (108). Undesirable reactogenicity of naturally derived OMVs is another important obstacle to overcome. The commercial OMV-based meningococcal group B vaccine (Bexsero) requires LPS detoxification, a procedure that can limit yields of presented antigenic molecules (45, 61, 127, 137, 138). Reformulation of this vaccine, and indeed future vaccines, could be improved through recombinant preparation of OMVs with less-toxic LPS mixtures, a strategy that has shown promise when applied to existing Neisseria vaccines (137, 138). Design principles employed in the aforementioned E. coli–based adjuvant detection library could also prove useful in developing desirable OMV detoxification schemes in other target OMV-producing organisms.

DESIGNER OUTER MEMBRANE VESICLES.

Outer membrane vesicles (OMVs) are spherical, nonreplicating nanostructures (20–250 nm) naturally produced in low numbers by gram-negative bacteria. They are formed through blebbing of the outer membrane and periplasmic compartments. Because OMVs are small and extracellular, ultracentrifugation protocols allow for easy purification of OMVs from associated source bacteria. There has been renewed interest in developing OMVs as vaccine platforms because they naturally contain, or can be designed to package, microorganism-associated molecular patterns (MAMPs) that are recognized by the innate immune system (e.g., lipopolysaccharide). This could make it possible to formulate vaccines with a combinatorial display of adjuvant and antigen. In addition to nonpathogenicity, one benefit of OMVs is the near native presentation of antigens normally found on pathogen surfaces. Equally important is the potential to improve presentation of antigens not normally associated with membranes, through presentation on host-cell membrane vesicles that double as delivery vehicle and adjuvant. Protective OMV vaccines have been in use since the 1980s, and many others have been investigated. However, their use in vaccine formulations has been limited because of hurdles discussed in this manuscript. Certain bacterial strains are well known for preferentially enriching OMVs with specific proteins; there is also a precedent for engineering multivalent OMV vaccines.

Figure 7.

Engineering outer membrane vesicles (OMVs) for vaccines. Numerous lipid A–modifying enzymes have been characterized and could be used to engineer the lipid A domain of lipopolysaccharide to serve as an adjuvant. Furthermore, both proteins and carbohydrates can be presented as antigens on the bacterial surface.

Outer Membrane Vesicles as Effective Glycoconjugate Vaccines

The abundant distribution of glycans on the surface of pathogens has always made carbohydrates an attractive vaccine target. Yet pure carbohydrates tend to be poorly immunogenic because they elicit a T cell–independent immune response. Protein-linked glycoconjugate vaccines have historically been used to enhance immunogenicity of glycan targets (47). Capsular polysaccharides and the LPS O-antigen are two such epitopes initially recognized by the host immune system (106, 145). Technology has now been developed to covalently display any of these glycan domains heterogeneously onto the core Kdo–lipid A domain using the WaaL ligase responsible for native O-antigen addition (140) (Figure 7). With this and similar technologies it is feasible that glycan antigen targets from pathogenic eukaryotes, viruses, or even cancer cells can be conjugated to and surface displayed on OMVs. Modulating the adjuvanticity of associated Kdo–lipid A further improves upon this promising design strategy.

Outer Membrane Vesicles Engineered to Present Proteinaceous Vaccine Targets

Multiple methods can engineer bacterial OMVs to package recombinant proteins (Figure 7). In E. coli and similar organisms, soluble proteins can be sorted to the bacterial periplasm by including an N-terminal signal sequence, for either Sec translocase or the twin-arginine translocation apparatus. This would enable recombinant proteins to be packaged in the lumen of subsequently formed OMVs. A desired epitope can also be conjugated to the surface of an OMV via fusion to a membrane-anchored protein (46), or through hybridization to an autotransporter domain (112).

Many groups have demonstrated the potential of OMV platforms for protein antigen delivery. Yersinia enterocolitica OMVs containing chimeric fusions of green fluorescent protein (GFP), which is usually not immunoreactive in humans, to the β-barrel outer membrane protein Ail generate an immune response (71). A similar strategy was developed in E. coli, where researchers fused GFP to enterobacterial ClyA hemolysin, a toxin that doubles as a potential adjuvant (52, 141). BALB/c mice vaccinated with the ClyA-GFP OMV formulation developed robust humoral immunity and an impressive effector T cell response to ClyA toxin and GFP. These reports substantiate the potential of OMV delivery as a universal protein vaccine platform, with an adjustable degree of antigen valency and the opportunity to include improved adjuvant Kdo–lipid A domains or other adjuvant biomolecules.

SUMMARY POINTS.

1. The gram-negative outer membrane is a complex macromolecular assembly typically composed of lipoproteins, β-barrel outer membrane proteins, extracellular glycans, and an asymmetric lipid bilayer. PLs are compartmentalized to the periplasmic leaflet and LPS to the external leaflet of the outer membrane lipid bilayer.

2. Transport proteins assemble final outer membrane architecture from material built on or within the inner membrane. The separate translocation machinery for lipoproteins, β-barrel outer membrane proteins, or LPS are best understood, wherein a few minor components require further elucidation or identification.

3. LPS is considered essential for viability of gram-negative bacteria, with few reported exceptions. As such, LPS biosynthetic enzymes and transport proteins are valid targets for developing antibacterials, where protein structures of these components will inform discovery of novel compounds.

4. Remodeling of LPS chemical structures promotes bacterial survival in a given environmental or host niche. The type of LPS modifications and how they are regulated across gram-negative bacteria are highly variable.

5. Novel biotechnological applications manipulating bacterial surface architecture, e.g., OMV-based vaccines or probiotic therapies, will require more sophisticated methods for determining the structure of outer membrane biomolecules and new tools for understanding assembly dynamics.

FUTURE ISSUES.

1. How is LPS synthesis machinery organized and distributed within the bacterial cell? Given the essentiality and complexity of the Raetz pathway, one could predict that an “endotoxome” exists to ensure optimal flux of key intermediates along the biosynthetic pathway.

2. Why is Kdo–lipid A essential in nearly all gram-negative organisms? Understanding why endotoxin is dispensable in some organisms, such as A. baumannii, will provide insight into the general requirement of Kdo–lipid A and possibly allow for engineering of LPS-deficient E. coli.

3. What molecules make up the orphan PL transport pathway? Given the similarity to transport protein machinery involved with other outer membrane components, molecules dedicated to the flip of newly synthesized PLs from the cytosolic to periplasmic leaflet of the inner membrane are yet to be identified, as are translocation components that shuttle PLs across the periplasm to the outer membrane.

4. Why are there multiple genes for the enzymes (Cls and Pgp proteins) required for PG and cardiolipin synthesis? Does this indicate compartmentalization of PL synthesis, or differential regulation? Are these isoenzymes important for different stages of bacterial growth?

5. Given the unfavorable energy barrier to the movement of large biomolecules across a membrane bilayer, how are so-called surface-displayed lipoproteins flipped from the periplasmic to the surface leaflet of the outer membrane?

6. A more sophisticated, and dynamic, understanding of organisms with complex patterns of Kdo–lipid A domain modification will further inform how these alterations affect natural processes such as general bacterial physiology, pathogenesis, and other symbioses.

Glossary

Glycerophospholipid (PL)

a membrane lipid with a glycerol-3-phosphate backbone with fatty acyl chains esterified to positions 1 and 2

Lipopolysaccharide (LPS)

a glycolipid localized to the outer leaflet of the outer membrane comprising lipid A, core-oligosaccharide, and O-antigen domains

PL headgroup

the chemical moiety bound to the phosphate group at position 3 on the glycerol backbone of phosphatidic acid

Lipoprotein

a membrane protein that may be anchored to the inner or outer membrane by a lipid moiety

Kdo

3-deoxy-d-manno-oct-2-ulosonic acid

Lipid A

the hydrophobic anchor of lipopolysaccharide

Core oligosaccharide

an oligosaccharide attached to a lipid A domain, with the inner core Kdo sugar as the attachment site

MsbA

an integral membrane transporter that flips lipid A–core oligosaccharide to the periplasmic face of the inner membrane

Lipooligosaccharide (LOS)

a form of LPS with an extended core oligosaccharide, but lacking O-antigen

Two-component regulatory system

a system composed of a sensor histidine kinase that receives input stimuli and then phosphorylates a response regulator

PbgA

PhoPQ-barrier gene protein A

ABC (ATP-binding cassette) transporter

ubiquitous integral membrane proteins that actively transport ligands across membranes utilizing energy generated by ATP hydrolysis

Homeoviscous adaptation

maintenance of membrane fluidity through compositional adaptation of membrane lipids

O-antigen

a long-chain polysaccharide attached to the lipid A core oligosaccharide that provides a major cellular antigen

TLR4/MD-2 (Toll-like receptor 4/myeloid differentiation factor 2)

functions as a pattern-recognition receptor recognizing LPS; initiates a robust signal cascade in mammals

Pattern-recognition receptor

receptor of the innate immune system that bind MAMPs of infecting pathogens leading to an inflammatory response

Endotoxin

a synonym of LPS; introduced in the nineteenth century to describe a heat-stable toxin associated with gram-negative bacteria

Adjuvant

a vaccine component that increases the immune response to an antigen

Capsular polysaccharides

a thick layer of polysaccharides that surrounds a bacterial cell