Lipopolysaccharides

The Cell Envelope of gram-negative bacteria is characterized by its outer membrane. The outer membrane is an asymmetric lipid bilayer, in which the inner leaflet contains phospholipids and the outer leaflet contains the unique amphiphilic glycolipid known as lipopolysaccharide (LPS). There are estimated to be ~106 LPS molecules per Escherichia coli cell. The distinctive structural features of LPS are crucial for the protective barrier properties of the outer membrane. In gram-negative sepsis, LPS molecules released from the bacterial surface stimulate macrophages and endothelial cells to overproduce cytokines and proinflammatory mediators, leading to the often fatal syndrome of septic shock. The involvement of LPS in this process is the reason that it is often referred to as endotoxin, and these biological effects have inspired a substantial part of LPS research. The complex structures of LPS molecules also provide fascinating research topics in the areas of synthesis and export of macromolecules, as well as in membrane biogenesis. The broad spectrum of LPS research is reflected in the activities of the International Endotoxin Society (http://www.kumc.edu/IES/).  

I. LIPOPOLYSACCHARIDE STRUCTURE Early structural analyses of LPSs were driven, in part, by the need to resolve the identity of the molecule responsible for the endotoxic effect. One of the key breakthroughs in early LPS research came from the establishment of techniques for the extraction and isolation of LPS by O. Westphal and O. Lüderitz in the 1940s. Although other methods have followed, the hot aqueous phenol method they developed remains one of the most common and valuable extraction procedures in current use. This early work led to the understanding that the endotoxic phenomenon is attributable to LPS, and equally importantly, the finding that LPS molecules with similar compositions are found in different gram-negative bacteria. More recent application of analytical techniques such as nuclear magnetic resonance spectroscopy and mass spectroscopy, has led to highly refined structures for LPS molecules from diverse bacteria. It is now clear that there are general structural features or themes that are highly conserved in LPSs from different sources, but that there is significant variation in the structural fine details. Extensive research has been performed on the LPS molecules of Salmonella enterica serovar Typhimurium and E. coli, and these LPSs form a basis for comparative analysis of other LPSs. For the purpose of discussion, the LPSs of E. coli and S. enterica sv. Typhimurium can be conveniently subdivided into three structural domains (S-LPS; Fig. 56.1). 

Lipid A is the hydrophobic part of the LPS molecule and is a major component of the outer leaflet of the outer membrane. Extending outward from lipid A is the branched and often phosphorylated oligosaccharide known as the core oligosaccharide (core OS). The O-antigen side chain polysaccharide (O antigen; O-PS) is a polymer of defined repeat units, attached to the core OS. The O-PS extends from the surface to form a protective layer. This complete tripartite LPS structure is known as “smooth LPS” (S-LPS), taking its name after the “smooth” colony morphology displayed by enteric bacteria that have the complete molecule on their cell surfaces. Mutants with defects in O-PS or core OS assembly produce truncated LPS molecules. For example, the widely used E. coli K-12 strains carry a defect in O-PS biosynthesis. The resulting colonies lack the smooth character, and the truncated LPS is, therefore, widely known as “rough LPS” (R-LPS) (Fig. 56.1). Preparations of LPS from bacteria that produce S-LPS contain a heterogeneous mixture of molecules and always have a variable amount of truncated R-LPS. Some bacteria, particularly mucosal pathogens naturally lack O-PS chains in their LPS. Their LPS contains oligosaccharide extensions attached to various points of a typical inner core OS, in a form of LPS known as lipooligosaccharide (LOS) (Fig. 56.1). The LPS molecules from different bacteria typically show closer structural relationships in the cell-proximal lipid Aand inner core OS regions and increasing diversity in the distal outer core OS and O-PS domains. The inner portions of the LPS molecule play important roles in establishing the essential barrier function of the outer membrane, and this likely places constraints on the extent of structural variation. The outer parts of the LPS molecule interact with environmental factors, such the host immune response. These selective pressures may have played a significant role in the diversification of outer LPS structures. 

Some bacteria, including Sphingomonas paucimobilis, Treponema maritime, and Borrelia burgorferi, have outer membranes that lack LPS molecules entirely. In some cases genome sequences lack key genes for lipid Asynthesis and therefore support compositional data. In the case of S. paucimobilis, glycosphingolipids probably serve to replace the lipid A, and the same may be true for the other examples. In those organisms that have a traditional LPS, it is generally thought that lipid A is essential for viability. However, the recent identification of a viable Neisseria meningitidis mutant with a defect in an essential step in the lipid A synthesis pathway challenges the universality of this assumption. A. Lipid A Free lipid Ais not found on the bacterial cell surface but mild-acid treatment of most isolated LPSs releases lipid A by hydrolysis of the labile ketosidic linkage between the core OS and lipid A. Structural analysis of lipid Ais hampered by its microheterogeneity, as well as its amphipathic properties. However, the lipid A components of LPSs from a variety of bacteria have now been resolved, revealing a family of structurally related glycolipids based on common architectural features. In enteric bacteria, the backbone of lipid A is formed by a disaccharide, comprising two glucosamine (GlcN) residues joined by a _1,6-linkage. The disaccharide backbone is phosphorylated at the 1 (reducing) and 4_ (nonreducing) positions and is acylated with ester and amide-linked 3-hydoxyl saturated fatty acids. In E. coli and Salmonella, the fatty acyl chains on the nonreducing GlcN residue are substituted by nonhydroxylated fatty acids, creating an asymmetric arrangement (Fig. 56.2). Microheterogeneity is evident in lipid Apreparations isolated from a given bacterium and several regulated modifications to the lipid Astructure are evident in cells grown under specific conditions. Work with Salmonella has established that the two-component environmental sensor comprising PhoP/PhoQ is required for transcription of genes essential for virulence in mice. The sensor appears to respond to microenvironments encountered by the bacterium in the phagolysosome, with the adaptive response optimizing the bacterial physiology for intracellular growth. Among the genes whose transcription is controlled by PhoP/PhoQ are pmrA/B, encoding an additional two-component regulatory system. Together, these sensory systems modulate expression of enzymes that are involved in 4-amino-4-deoxy-L-arabinose (Ara4N) synthesis and its addition to lipid A, as well, as the those required for addition of an extra palmitate to form a heptaacyl lipid A in S. enterica sv. Typhimurium (Fig. 56.2). These alterations can have a considerable impact on the structure and function of LPS molecules. For example, Salmonella heptaacyl LPS shows a decreased capacity to activate release of cytokines and proinflammatory mediators (see following) and may be important for longer term survival inside the host cell. 

Addition of Ara4N to lipid A is an important determinant of the resistance of the bacterium to polycationic antimicrobial polypeptides, and polycationic antibiotics, such as polymyxin. The effect of this modification may be a dampening of the negative charges in lipid A, inhibiting the initial binding of polycations and their eventual perturbation of the outer membrane. Interestingly, these changes can be induced by metavanadate in E. coli K- 12. In P. aeruginosa, hexaacyl lipid A modified with palmitate and Ara4N is synthesized in response to environmental cues from the cystic fibrosis lung environment. These changes confer resistance to cationic peptides and generate LPS with increased inflammatory properties. These may play a significant role in persistence of P. aeruginosa and tissue damage in the lungs of cystic fibrosis patients. In contrast to these regulated modifications, the lipid As of Proteus mirabilis and Burkholderia cepacia are constitutively modified by Ara4N and these organisms are resistant to polymyxin under all growth conditions. It is clear from these examples that LPS should be considered to be a dynamic, rather than static, molecule, whose structure and function can be modulated in response to cues from the host. 

Most lipid A variations between different species involve alterations in acylation or phosphorylation. For example, the acylation pattern is symmetrical in N. meningitidis, whereas Rhodobacter sphaeroides lipid A is distinguished by amide-linked 3-oxotetradecanoic acid and the presence of unsaturated fatty acids. The structure of the R. sphaeroides lipid A is of particular importance since it results in an LPS lacking the normal biological activities attributed to endotoxins. In the Rhizobiaceae, some lipid A molecules lack phosphate residues and the proximal glucosamine may undergo an oxidation step to form an aminogluconate residue. However, structural variations can be as extreme as the 2,3-diamino-2,3-dideoxy-D-glucosecontaining monosaccharide backbone structure of lipid A molecules from Pseudomonas diminuta and Rhodopseudomonas viridis. B. Core oligosaccharides For the purpose of discussion of structure–function relationships, the core OS is often divided into inner and outer core regions (S-LPS; Fig. 56.1). The inner core of most known LPSs comprises characteristic residues of 3-deoxy-D-manno-octulosonic acid (Kdo) and L-glycero-D-manno-heptose (Hep). This region is highly conserved in enteric bacteria (Fig. 56.3A). The Kdo residues can be nonstoichiometrically modified by other sugars, or by 2-aminoethyl phosphate (phosphorylethanolamine; PEtN). The main chain Hep residues are also nonstoichiometrically decorated with phosphate, pyrophosphorylethanolamine (PPEtN), or a side-branch Hep. In some nonenteric bacteria, the Kdo residue proximal to lipid Ais phosphorylated, or is replaced with the derivative D-glycero-D-talo-octulosonic acid (Ko). These differences may influence the lability of the linkage that is usually cleaved by mild-acid hydrolysis to release lipid A from the intact LPS molecule. In Klebsiella pneumoniae, Kdo is also present in the outer core OS. The outer core OS structure is more variable among different bacteria. 

However, within a species, these variations are quite limited and some may have a single outer core OS type. For many years, this was thought to be the case for Salmonella but a second core OS structure has been reported relatively recently. E. coli has five known distinct outer core OS types; all contain five glycose residues but they differ in glycose content and organization (Fig. 56.3B). In addition to altering the antigenic epitopes and diagnostic bacteriophage receptors, variations in outer core OS structure give rise to altered sites for the attachment of O-PS.  C. O polysaccharides The O-PS is the most variable portion of the LPS molecule. A remarkable array of novel structures arises from alterations in constituent sugars, linkages, and both complete and partial substitutions with nonsugar residues. At their simplest, O-PSs are homopolysaccharides with a disaccharide repeat unit, consisting of a single monosaccharide component. Complex homopolysaccharides can result from larger repeat units defined by a specific sequence of glycosidic linkages. At their most complex, O-PSs can be heteropolysaccharides in which the repeat units contain several component sugars, together with nonsugar substituents, such as O-acetyl groups and amino acids. The structural diversity in O-PSs has been exploited in serological classification of isolates from a given bacterial species. In E. coli, there are approximately 170 distinct O serogroups. While the LPSs from most bacterial strains tend to have a single O-PS type, there is a growing number of bacteria where the lipid A-core serves as an acceptor for two or more different polymeric structures. The length of the O-PS attached to lipid A-core is heterogeneous but the distribution of chain lengths is both strain- and growth-condition dependent. This is best reflected in the patterns of LPS molecules revealed by SDS-PAGE analysis (Fig. 56.4). D. Lipooligosaccharides The LPS of mucosal pathogens such as Neisseria and Haemophilus spp. is smaller than S-LPS and takes a form that is known as lipooligosaccharide (LOS). These bacteria generally do not produce S-LPS. As an example, the LOS structure from N. gonorrhoeae is given in Fig. 56.3C. The oligosaccharide chains linked to the inner core OS provide distinct antigenic epitopes, and a common feature in these bacteria is the phenomenon of phase variation, where the LOS epitopes are differentially expressed. This gives rise to difficulty in arriving at precise LOS structures, unless phase-locked variants are available. Some Bordetella spp. can form both S-LPS and LOS. 

II. BIOSYNTHESIS AND ASSEMBLY OF LIPOPOLYSACCHARIDES In E. coli, the minimal LPS molecule required for viablity consists of Kdo2-lipid A(sometimes known as Re-LPS; Fig. 56.2, Panel B). Since E. coli can efficiently insert Re-LPS and larger LPS molecules on their cell surfaces, the later steps in LPS assembly pathways have generally been amenable to dissection by genetic approaches (e.g. structural analyses of mutant LPSs), supported by biochemical assays to establish enzymatic function. In contrast, the dependence on Re-LPS for viability has complicated studies on the earlier biosynthesis steps but, nevertheless, these have been systematically resolved by biochemical analyses. The assembly of LPS molecules is complicated by their tripartite structure. The lipid A-core portion is assembled at the cytoplasmic face of the plasma membrane and, once complete, it is transferred across the membrane to the periplasm. The long hydrophilic O-PS is assembled independently and exported to the periplasmic face of the plasma membrane prior to its ligation to preformed lipid A-core. In a major recent finding, the ABC (ATP-binding cassette)-transporter MsbA has been implicated in the export of both lipid A-core and phospholipids. The structure of the transporter has been solved at 4.5Å and preliminary biochemical data suggests its ATP-hydrolysis activity is activated by LPS substrate. 

Subsequent and, as yet, undefined steps translocate LPS to the cell surface and assemble the nascent molecules into the outer membrane. A. Synthesis of lipid A The pathway for biosynthesis of lipid A in E. coli (Fig. 56.2B) has been resolved primarily through the efforts of C. R. H. Raetz and co-workers and a detailed and historical perspective of this work is provided elsewhere. The acylated glucosamine derivatives that provide the halves of the lipid Abackbone are formed from UDP-N-acetylglucosamine (UDP-GlcNAc) by sequential cytoplasmic reactions catalyzed by an O-acyltransferase (LpxA), a deacetylase (LpxC) and an N-acyltransferase (LpxD). The product of these reactions, UDP-2,3-diacylGlcN, provides a direct precursor for the nonreducing diacylGlcN derivative of the lipid A backbone. The reducing diacylGlcN is generated by LpxH by cleavage of a second UDP-2,3- diacylGlcN molecule at the pyrophosphate bond, to form the monophosphoryl derivative. This product is known as lipid X and its fortuitous accumulation in mutants with certain defects in phospholipid metabolism provided an essential clue for unraveling the lipid A assembly pathway. The two halves of the lipid A backbone are then joined by the disaccharide synthase (LpxB). The 4_-kinase (LpxK) completes formation of the tetracyl-derivative known as lipid IVA. This product has been useful for investigations of the biological activities of LPS. Full acylation of E. coli lipid A requires the prior addition of two inner core Kdo residues. The Kdo transferase (WaaA; formerly, KdtA) from E. coli is bifunctional and simultaneously adds two Kdo residues, providing substrates for the lauryl (HtrB) and myristoyl (MsbB) acyltransferases. Unlike the early steps of lipid A synthesis, modification with Ara4N and addition of palmitate to form the heptaacyl lipid A occur after the lipid A has been exported across the inner membrane. Addition of Ara4N occurs in the periplasm. Palmitate addition is mediated by an outer membrane protein, PagP, that is located in the outer membrane and uses glycerophospholipids as the acyl donor. Homologs of PagP are found in E. coli, Yersinia pestis, and members of the genera Salmonella, Bordetella, and Legionella. The essential features of the lipid Aassembly pathway and its key enzymes appear to be conserved in different bacteria, although there are some subtle speciesspecific variations. There are clearly differences in the specificity of the acyltransferases and, in the family Rhizobium, phosphatases are required to generate the phosphate-free lipid A. 

Also, in Pseudomonas aeruginosa and N. meningitidis, the sequential dependence of the terminal steps in lipid Aassembly (i.e. addition of Kdo to lipid IVA prior to full acylation) is not conserved. Some bacteria, for example, H. influenzae, have a monofunctional Kdo transferase, and the Chlamydia WaaA-homolog is trifunctional. Due to the requirement for at least a minimal LPS in most bacteria, only conditional mutants of lipid A or Kdo synthesis have been isolated from E. coli and several other pathogens, although N. meningitidis provides a notable exception. The typically essential nature of lipid A has been exploited in attempts to generate therapeutic antibacterial compounds, targeted against lipid A biosynthetic enzymes and CMP-Kdo synthesis. Structural analogs of Kdo proved to be very effective inhibitors of CMP-Kdo synthetase in vitro, but their ultimate failure was due to the inability of most bacteria to take up the inhibitor. In the laboratory, this could be circumvented by synthesis of a dipeptide-linked prodrug which exploited the oligopeptide permease transporter for uptake. Unfortunately, this provides mutation of the transporter as a very simple route to drug resistance. Another attractive candidate for inhibitor development is the deacetylase reaction catalyzed by LpxC. Synthetic LpxC inhibitors have proven effective in animal challenge models for E. coli infections but not against P. aeruginosa and, again, this is likely a reflection of uptake problems. Nevertheless, such approaches show promise. B. Synthesis of core oligosaccharide and LOS structures The completed Kdo2-lipid A provides an acceptor for glycosyltransferases that act sequentially to assemble the core OS, as well as those that add the oligosaccharide chains of LOSs. These enzymes are peripheral membrane proteins that act at the cytoplasmic face of the inner membrane. In E. coli and Salmonella, the structural genes for the core OS glycosyltransferases map together with the Kdo transferase gene (waaA), and genes required for modification of the Heptoseregion of the inner core OS (waaPYQ). 

These genes form the three separate operons in the chromosomal waa-region. Direct data for biochemical activities of individual core OS glycosyltransferase enzymes is unavailable in many cases. For example, studies of heptosyltransferases have generally been limited by the unavailability of the activated precursor, ADPHep, but the recent elucidation of the pathway for ADP-Hep biosynthesis will significantly help research in this area. Assignments of other glycosyltransferases have primarily been made by approaches where specific genes are individually mutated and the resulting LPS structure is resolved by chemical analysis. Synthesis of the core OS backbone can be carried to completion in mutants lacking the modifications that decorate the Hep-region. As a result, the precise timing of these modifications in the overall synthesis pathway is unknown. C. Synthesis of O-polysaccharides Despite the diversity in O-PS structures, only three mechanisms are known for the formation of O-PS (Fig. 56.5). O-PS synthesis begins at the cytoplasmic face of the inner membrane with activated precursors (sugar nucleotides; NDP-sugars) and the process terminates with a nascent O-PS at the periplasmic face. The ligation to lipid A-core then follows. The different pathways for assembly of O antigens vary in the components required for polymerization, in the cellular location of the polymerization reaction, and in the manner in which material is exported across the inner membrane. A carrier lipid, undecaprenyl phosphate (und-P), is involved in all three O-PS assembly pathways. The involvement of a carrier lipid scaffold may ensure fidelity in O-repeat unit structure, or simply provide an acceptor compatible with the membrane environment. The same three mechanisms are identified in the biosynthesis of capsular polysaccharides in gram-negative and gram-positive bacteria. In this respect, the primary distinction between the O-PSs and capsular polysaccharides is that the O-PSs are attached to lipid A-core. 

The most prevalent pathways for O-PS synthesis are distinguished by the involvement (or not) of the putative “O-PS polymerase” enzyme, Wzy. The “Wzy-dependent” system (Fig. 56.5A) is the classical pathway first described in S. enterica serogroups A, B, D, and E. However, sequence and biochemical data shows the key enzymes are shared by other bacteria. In the working model for this pathway, und-PPlinked O-repeat units are assembled by glycosyltransferase enzymes at the cytoplasmic face of the plasma membrane. These reactions have been known since work in the 1960s by H. Nikaido, M. J. Osborn, P. Robbins, A. Wright, and others. The initial transferase is an integral membrane protein that transfers sugar- 1-phosphate to the und-P acceptor. This is followed by sequential sugar transfers, catalyzed by additional peripheral glycosyltransferases, to form an und-PPlinked repeat unit. The polymerization reaction occurs at the periplasmic face of the membrane and utilizes und-PP-linked O-PS-repeat units as the substrate. The individual und-PP-linked O-PS-repeat units must, therefore, be exported across the inner membrane prior to polymerization, and preliminary biochemical analyses suggest that the likely candidate for this process is a multiple membrane-spanning protein, Wzx (formerly, RfbX). Polymerization of the O-PS repeat units minimally involves the putative polymerase (Wzy; formerly, Rfc) and the O-PS chain length regulator (Wzz, formerly Rol or Cld). A wzy mutant is unable to polymerize O-PS and its LPS comprises a single O-repeat unit attached to lipid A-core. In contrast, a wzz mutant makes S-LPS but loses the characteristic modal distribution of O-PS-chain lengths evident in SDS-PAGE analysis (e.g. Fig. 56.4). The O-PSs synthesized by this pathway are all heteropolymers and often have branched repeating unit structures. In the “ABC-transporter-dependent” pathway, the O-PS is synthesized exclusively inside the cytoplasm and once complete, it is exported to the periplasm via an ABC-transporter, belonging to the ABC-2 family (Fig. 56.5B). This mechanism is, so far, confined to O-PSs with relatively linear structures and several are homopolymers. As with the Wzy-dependent pathway, synthesis is initiated at the cytoplasmic face of the inner membrane by an integral membrane glycosyltransferase enzyme, to form an und-PP-sugar. 

In fact, in E. coli the UDP-GlcNAc:undecaprenylphosphate GlcNAc-1-phosphate transferase (WecA) can initiate for either pathway. However, in the ABC-2- transporter-dependent pathway the initiating transferase acts once per O-PS chain. Additional peripheral glycosyltransferases then act sequentially and processively to elongate the und-PP-linked intermediate at the nonreducing terminus to form a fully polymerized und-PP-O-PS. The specificities of glycosyl transferases dictate the repeat-unit structure of the product. Only one und-PP acceptor is used per polymer and there is no equivalent of the polymerase (Wzy) or chain-length regulator (Wzz) enzymes. Several O-PS formed by the ABC-2-transporter-dependent pathways terminate at their non-reducing end with novel residues that are not part of the repeat-unit structure (examples include methyl groups and Kdo). These residues may act as signals for chain-termination, for initiation of export, or both. The ABC-2 family transporter comprises a transmembrane (Wzm) protein and an ATP-binding (Wzt) component. The involvement of an ABC-transporter precludes the involvement of Wzx, since there is no requirement for export of individual und-PP-O-units. It remains unknown whether the O-PS retains its attachment to und-PP during export, whether it is removed from the lipid carrier for export, or, alternatively, if an alternative carrier molecule exists. 

The “synthase-dependent” pathway for O-PS biosynthesis is, so far, confined to the homopolymeric O antigen (factor 54) of S. enterica serovar Borreze. The model for this pathway (Fig. 56.5C) proposes that the initiating glycosyltransferase (WecA in this case) forms an und-PP-sugar acceptor that is elongated by a single multifunctional synthase enzyme, in a manner analogous to eukaryotic chitin and cellulose synthases and hyaluronan synthases from eukaryotes and bacteria. There is no dedicated ABC-2 transporter or Wzx homolog in the O:54 system and all experimental evidence points to the synthase having dual transferaseexport functions. It is not known whether this system requires the nascent polymer be linked to und-PP throughout the polymerization and export processes. In most bacteria, the genes dedicated to O-PS biosynthesis are clustered on the chromosome. The O-PS gene clusters contain a predictable spectrum of genes encoding novel sugar nucleotide synthetases, glycosyltransferases, and the characteristic enzymes such as Wzx-Wzy-Wzz, or an ABC-2 transporter, or synthase. As a result, sequence data can give an accurate first evaluation of the O-PS biosynthesis pathway involved. As might be expected from the range of OPS structures, the O-PS biosynthesis genetic loci are highly polymorphic. Genetic recombination within and between bacterial species has played a significant role in the diversification of O-PS. In some bacteria (e.g. Salmonella spp., Shigella spp., and P. aeruginosa), phage-encoded genes provide additional determinants of O serotype specificity. These modifications are best characterized for Wzy-dependent systems in Salmonella, Shigella, and P. aeruginosa. Examples include changes in the linkage specificity of the Wzy polymerase, and acetylation or glucosylation of the polymer and all of these modifications appear to occur in the periplasm. A large number of LPS biosynthesis genes have been identified in bacteria with different lifestyles. For a current listing of known genes see the Bacterial Polysaccharide Gene Database maintained by P. Reeves’ laboratory (www.microbio.usyd.edu. au/BPGD/default.htm). D. Terminal reactions in LPS assembly Once complete, the nascent O-PS chain must be linked (ligated) to lipid A-core. All available evidence points to a periplasmic location for this reaction. Since lipid A-core is formed in the cytoplasm, it too must be exported to the ligation site at the periplasmic face of the inner membrane. An ABC-transporter, known as MsbA, appears to be responsible for this step. The mechanism underlying the ligase reaction itself remains unknown. 

The waaL gene product, often referred to as “ligase,” is currently the only known protein that is essential for ligation. Its assignment as the ligase is based only on mutant phenotypes and there is no supporting mechanistic information. Interestingly, the ligase from E. coli K-12 will ligate structurally distinct O-PSs formed by any of the three assembly pathways, indicating that the form in which nascent O-PS is presented for ligation is conserved. Perhaps the most interesting open questions in LPS assembly surround the process by which the completed LPS is translocated through the periplasm and then inserted into the outer membrane. In most bacteria, the translocation pathway must have relaxed specificity, since it efficiently transfers a range of LPS molecules varying from S-LPS to Re-LPS to the cell surface. Recent data has implicated an outer membrane protein, Omp85, in LPS export. There is also preliminary data that implicates the Tol system (a multiprotein complex involved in the translocation of group A colicins and filamentous bacteriophages) in surface expression of O-PS but a detailed understanding of the process is not yet available. 

III. FUNCTIONS AND BIOLOGICAL ACTIVITIES OF LIPOPOLYSACCHARIDES A. Lipopolysaccharides and outer membrane stability From the construction of precise mutants with LPS defects it is well established that the minimal LPS molecule required for survival of E. coli consists of Kdo2- lipid A (Re-LPS). Although E. coli can assemble an outer membrane from Re-LPS, the barrier function of the resulting outer membrane is compromised. Outer membrane integrity depends on structural elements in both lipid A and core OS. The role of Ara4N-modified lipid A in resistance to polycations has been discussed above. In E. coli and Salmonella, the phosphorylated Hep-region of the core OS is crucial for outer membrane stability by facilitating the cross-linking of adjacent LPS molecules by divalent cations or polyamines and by its interaction with positively charged groups on proteins. The inability to synthesize or incorporate Hep, or the loss of phosphoryl derivatives alone (i.e. a waaP mutant), gives rise to significant compositional and structural changes in the outer membrane. In E. coli, these mutants are known as “deep-rough” and their perturbed outer membrane structure leads to pleiotropic phenotypes, most notably hypersensitivity to hydrophobic compounds, such as detergents, dyes, and some antibiotics. Salmonella waaP mutants are avirulent and in P. aeruginosa, waaP is an essential gene. For such bacteria, assembly and phosphorylation of the core OS Hep-region may therefore provide further avenues for therapeutic intervention. In most free-living bacteria, a negatively charged core OS is the important element in terms of outer membrane stability. However, phosphorylation is clearly not the only way to achieve a robust outer membrane, as some bacteria lack phosphorylation of the heptose region. For example, in the case of Klebsiella pneumoniae, galacturonic acid and Kdo residues in the core OS provide the only source of negative charges. Alimited number of wild-type gram-negative bacteria are viable without LPS. In the case of S. paucimobilis, no “typical” LPS is present in the outer membrane but, instead, the bacterium produces a glycosphingolipid– a modified ceramide derivative containing glucuronic acid and an attached trisaccharide. This lipid functionally replaces LPS in the formation of a stable outer membrane. In the case of N. meningitidis, mutants lacking the typical LOS-form are viable only if capsular polysaccharide is present and it is conceivable that the lipid anchor for this class of capsule partially replaces lipid A. 

However, there are other changes in outer membrane phospholipids and defects in expression of surface lipoproteins that may also help maintain viability. It is currently unclear whether this phenomenon extends beyond N. meningitidis. Perhaps the smallest wild-type LPS structure consists of only lipid A and a Kdo-trisaccharide, and is produced by members of the genus Chlamydia. Presumably, the intracellular growth environment for this organism, together with other features of the cell envelope, facilitate survival in the absence of a more complex LPS structure. B. O-polysaccharides as a protective barrier Molecular modeling of LPS structure and its organization in the outer membrane predict that the O-PS forms a significant layer on the cell surface. The O-PS partially lies flat on the cell surface, where the crossover of multiple chains forms a “felt-like” network. Since the O-PS is flexible, it can extend a significant distance from the surface of the outer membrane. It is, then, not surprising that many properties attributed to the O-PS are protective. In particular, long-chain O-PS is often essential for resistance to complementmediated serum killing and, therefore, represents a major virulence factor in many gram-negative bacteria. The serum proteins in the complement pathway interact to form a membrane attack complex (MAC) that can integrate into lipid bilayers to produce pores, leading to cell death. The MAC can be formed via a “classical” pathway, where surface antigen–antibody complexes initiate MAC formation, or through the “alternative” pathway, where complement component C3b interacts directly with the cell surface in the absence of antibody to facilitate MAC formation. In gram-negative bacteria with S-LPS, resistance to the alternative pathway does not result from defects in C3b deposition. Instead, C3b is preferentially deposited on the longest O-PS chains, and the resulting MAC is unable to insert into the outer membrane. In addition to O-PS chain length, complement-resistance can also be influenced by the extent of coverage of the available lipid A core with O-PS. As is often the case, there are exceptions to such generalizations. For example, there are some E. coli strains with S-LPS that are serum-sensitive unless an additional capsular polysaccharide layer is present. Although R-LPS variants of E. coli and Salmonella are almost invariably serumsensitive, other bacteria (including many with LOS) use alternate strategies to achieve resistance. 

The bactericidal/permeability inducing protein (BPI) is an antibacterial product found in polymorphonuclear leukocyte-rich inflammatory exudates. BPI binds LPS and may play a role in the clearance of circulating LPS but it also exhibits antimicrobial activity in the presence of serum. Resistance to BPImediated killing is also dependent on long chain O-PS. C. Lipopolysaccharide and gram-negative sepsis One potential outcome of gram-negative infections is septic shock, a syndrome manifested by hypotension, coagulopathy, and organ failure. In the United States, gram-negative sepsis accounts for 50,000–100,000 deaths each year. Septic shock results from the liberation of LPS from the bacterial cell surface, a phenomenon that naturally ensues from the growth and proliferation of bacteria. Tissue damage is not a result of direct interaction between the host tissues and an LPS “toxin,” but instead results from unregulated host production of cytokines and inflammatory mediators (including tumor necrosis factor (TNF-_), and a variety of interleukins) by over-stimulated macrophages and endothelial cells. Under normal circumstances, and at regulated levels, these components have beneficial effects and lead to moderate fever, general stimulation of the immune system, and microbial killing. In sepsis, however, their overproduction leads to tissue and vascular damage and the symptoms of sepsis. Since free LPS is required to initiate the process, treatment with antibiotics and the ensuing bacterial lysis may actually exacerbate the problem. The last decade has seen significant advances in our understanding of the manner in which LPS interacts with animal cells and stimulates their production of mediator molecules. It was suspected for some time that lipid A was the component responsible for those biological activities that LPS exhibited in sepsis. Definitive proof came from the observations that some partial LPS structures and chemically synthesized lipid A derivatives display the same biological activities as the complete molecule. Importantly, other partial structures are not only biologically inactive but can act as antagonists of LPS molecules that are active. Well-studied antagonist LPS molecules include the precursor lipid IVA and the naturally occurring R. sphaeroides lipid A molecule. These structures have directed the synthesis of potent synthetic LPS antagonists that are able to negate the effects of challenge with biologically active LPS. Circulating LPS molecules naturally form micellar aggregates and a variety of host LPS binding proteins are important in mobilizing LPS monomers from such complexes. These include BPI (see previous discussion) and LPS-binding protein (LBP), a 60 kDa acutephase protein produced by hepatocytes. One role of these proteins is to clear and detoxify LPS. For example, LBP is known to transfer LPS to high-density lipoprotein fractions. However, LBP is also a crucial component of the signaling pathway through which animal cells are stimulated to produce cytokines and inflammatory mediators. 

The central pathway by which cells recognize low concentrations of LPS (or bacterial envelope fragments containing LPS) requires the participation of a receptor protein CD14. CD14-deficient cell lines, such as 70Z/3 pre-B lymphocytes, are less sensitive to LPS (i.e. responsive to nanomolar rather than picomolar levels), unless transfected with CD14. Consistent with these results, CD14-knockout mice have been shown to be 10,000-fold less sensitive to LPS in vivo. In myeloid cell lines, CD14 occurs as a 55 kDa glycosylphosphatidylinositol (GPI) anchored glycoprotein, attached to the membrane (mCD14 in Fig. 56.6). However, a variety of nonmyeloid (endothelial and epithelial) cells are also responsive to LPS, via a soluble form of CD14 (sCD14). Both sCD14 and mCD14 can bind LPS to form a complex, but the kinetics of binding are slow. LBP serves to overcome this rate-limiting step by delivering LPS to mCD14 or sCD14 (Fig. 56.6) and blood taken from an LBPknockout mouse shows a 1000-fold reduction in the ability to respond to LPS. At high LPS concentrations, CD14-independent stimulation is evident in CD14- deficient mice. For some time it was accepted that an accessory coreceptor protein was required to interact with CD14, to facilitate internalization of the lipid A signal. CD14 itself lacks a transmembrane domain to facilitate intracellular signaling, and CD14 alone is not able to discriminate between agonist and antagonist lipid A molecules; the antagonist molecules do not appear to operate by blocking the ability of agonist LPS to bind to CD14. Amajor development in this field was the identification of Toll-like receptor 4 (TLR4) as the lipid A-signaling coreceptor in animal cells. TLR4 recognizes CD14–lipid A complexes (or perhaps in some circumstances lipid Aalone) and acts in the initiation of the signal transduction events that lead to cytokine induction. Interestingly, human TLR4 is able to discriminate between modified hexaacyl lipid A from P. aeruginosa and its pentaacyl form. Transmission of proinflammatory signals occurs only with the biologically active hexaacyl form. It is not yet clear which (if any) additional proteins participate in the transmembrane delivery of the lipid A signal but the soluble protein MD-2 is implicated in the process. 

Once the relevant signal is transmitted across the animal cell membrane, a cascade of events leads to the release and overproduction of cytokines. The components of the latter stages of the response pathway are now beginning to be identified but they are complicated by an involvement of a multicomponent cascade. The cascade involves rapid protein phosphorylation events, and isoforms of the p38 mitogen-activated protein (MAP) kinase family play a central role. LPS antagonists block phosphorylation of p38. p38 is itself activated by the MAP kinase kinases. Downstream, p38 has a number of substrates including the myocyte enhancer factor (MEF2) family of transcription activators. The transcription factor NF-_B also plays an important regulatory role in the proinflammatory response, as well as in the development of LPS tolerance. A variety of therapeutic approaches have been designed to interfere with specific steps in the process leading to septic shock. The numbers of mediators involved complicates strategies based on blocking cytokines themselves. Neutralizing an individual mediator would not be expected to be an effective therapy, as appears to be the case for antibodyneutralized TNF-_. Significant efforts have been directed to neutralizing the LPS signaling molecule by administering antibodies but, to date, attempts to develop therapies based on anti-endotoxin monoclonal antibodies have been disappointing. Two commercial monoclonal antibodies (E5 and HA1A) recognizing lipid A have been the subject of clinical trials but, unfortunately, these provided no compelling evidence for the protective capacity of the antibodies. However, there are antibodies that recognize the core OS of E. coli and Salmonella and that do show promise both in vitro and in animal models, suggesting alternate immunotherapeutic strategies. LPS neutralization could also be achieved by using proteins that bind LPS and both LBP and BPI are being pursued in this respect. Approaches that attempt to block LPS receptor pathways with synthetic LPS antagonists are proving effective in animal models. Equally promising is the application of anti- CD14 and anti-LBP monoclonal antibodies that block the formation of CD14:LPS complexes and protect against LPS exposure in animal models. D. Molecular mimicry in LPS and LOS As detailed structures become available for the various components in heterogeneous LPS and LOS preparations from different bacteria, it is apparent that several successful pathogens employ a strategy of molecular mimicry. In the case of Helicobacter pylori LOS, glycoforms have been described with structures resembling blood group antigens, including the Lewis determinants. These structures are implicated in, evasion of host immune response, autoimmune responses and in adhesion, and colonization of the bacterium. In Campylobacter jejuni, several LOS glycoforms contain ganglioside mimics and phase variation is common. These virulence determinants may again provide a mechanism of avoiding the host immune response but they are also suggested as a possible causative agent of autoimmune responses in the development of Guillain–Barré syndrome. E. Biotechnological applications involving LPS The incredible number of LPS structures provide an extensive range of oligosaccharide and polysaccharide structures with novel biological properties. These may be of value for therapeutic or other commercial applications. In one novel example, a recombinant E. coli strain was constructed in which the LPS core OS provided a scaffold for expression of the globotriose receptor for Shiga toxins. The E. coli strain efficiently adsorbs and neutralizes the toxin, affording a therapeutic approach for treating infections whose pathogenesis involves these and related toxins. Structures of glycosyltransferases, including a LOS galactosyltransferase, are now being solved at high resolution by crystallographic methods. Ultimately, this will provide general insight into the detailed structure–function relationships among glycosyltransferases, and open the possibility for engineering enzymes with novel specificities for practical applications. For information regarding known glycosyltransferases see the CAZY (Carbohydrate Active Enzymes) website maintained by B. Henrissat’s laboratory (http://afmb.cnrs-mrs.fr/CAZY/index.html).