Module 4.7.3

Structure and Assembly of Escherichia coli Capsules

CHRIS WHITFIELD

[SECTION EDITOR: LYNN SILVER]

Posted January 2, 2009

Department of Molecular and Cellular Biology, University of Guelph, Guelph, Ontario N1G 2W1, Canada

Mailing address: Department of Molecular and Cellular Biology, University of Guelph, Guelph, Ontario N1G 2W1, Canada. Phone: (519) 824-4120 (ext. 53361), Fax: (519) 837-3273, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it .

The surface of Escherichia coli is a complex array of proteins and polysaccharides that give rise to major cellular antigens. Some of these have been used for more than 60 years in serological tests to distinguish isolates (reviewed in references 87 and 89). The principle polysaccharide antigens are the O antigens, derived from the variable polysaccharide side chains of lipopolysaccharide (LPS), and the K antigens (from the German word kapsel), representing different capsular polysaccharides. The O and K antigens are identified by serological typing in whole-cell agglutination tests using defined antisera. Capsules form an enveloping structural layer on the surfaces of bacteria (see Fig. 1), and in serological procedures, K antigens mask the underlying O antigen. However, the relationships between O and K antigens are sometimes complicated. This masking effect of the K antigens reflects the role of capsules in the cell; capsules generally protect cells against host defenses or physical environmental stresses, such as desiccation.

Fig. 1Capsule of E. coli O9a:K30. This group 1 capsule was preserved by a freeze-substitution process (45).

Source: Adapted from the AnnualReviewofBiochemistry (146) with the permission of the publisher.

The capsules of E. coli are formed from structurally diverse repeat unit polysaccharides (see Fig. 2). In the repeat units, variations in the monomers and linkages, as well as the replacement of glycose residues with other residues, give rise to more than 80 recognized K antigen serotypes. E. coli capsules have been classified into four groups (1 to 4) based on characteristics including the genetic organization and regulation of the relevant loci and the mechanisms of biosynthesis and assembly (reviewed in reference 146) (Table 1). From a solely biosynthetic perspective, there are only two assembly pathways, defined by the group 1 and 2 prototypes. Groups 3 and 4 are essentially genetic variants of the group 2 and 1 systems, respectively. Homologs of the characteristic proteins in these biosynthesis pathways are widespread in gram-negative bacterial species, and some are also involved in conserved steps of capsule assembly in gram-positive bacteria. Consequently, E. coli provides essential prototypes for capsule and extracellular polysaccharide assembly in other bacteria. The first capsule gene cluster from any bacterium to be cloned and expressed was described by Silver and colleagues in 1981 (124) and was from E. coli K1 (group 2). Their report helped lay the foundation for what is now extensive literature on the molecular biology and biochemistry of capsule assembly in bacteria.

Fig. 2Structures of the repeat units of representative capsule-forming polysaccharides. Group 1 capsules always contain a hexose (glucose or galactose) residue in the main chain. Negative charges are provided by uronic acids, and the polymers are frequently decorated by (nonstoichiometric) O-acetyl or ketal substituents. They share structural features with colanic acid. The polysaccharides forming group 4 capsules may have either O or K antigen status, depending on whether an additional LPS-linked O antigen is also present. These polysaccharides all contain a main-chain acetamido sugar (GlcNAc or GalNAc). Group 2 and 3 capsules are structurally diverse, with no conserved themes in their repeating units. Like group 1 capsules, they can be modified by O-acetyl groups, and in some cases, like that of serotype K1, the O acetylation is subject to on-off form variation (denoted by *). Nonstoichiometric modification of the repeat units by amino acid residues is a feature of both groups 3 and 4. Qui4NMal, 4-(2-carboxyacetamido)-4,6-dideoxyglucose; Col, colitose (3,6-dideoxygalactose); Gro, glycerol; Rit, ribitol.

Source: Adapted from the AnnualReviewofBiochemistry (146) with the permission of the publisher.

Table 1Classification of E. coli capsules

BIOSYNTHESIS AND ASSEMBLY OF CAPSULES BELONGING TO GROUPS 1 AND 4 AND COLANIC ACID

Structure and Organization on the Cell Surface

Capsules belonging to groups 1 and 4 are most frequently found in enterohemorrhagic, enteropathogenic, and enterotoxigenic E. coli isolates causing intestinal diseases. In these bacteria, the K antigen is present on the cell surface in two forms. One is linked to LPS and the other is LPS independent; the two forms draw on the same pool of biosynthetic intermediates, but the end product is translocated to the cell surface via one of two independent pathways. In simple terms, these systems can be viewed as having a typical O antigen biosynthesis pathway that is augmented with an additional LPS-independent capsule translocation process. The high-molecular-mass (>100-kDa) capsular polysaccharide gives rise to the capsule structure that extends 200 to 300 nm from the cell surface (Fig. 1). It is not clear how group 1 or 4 capsular polysaccharides are linked to the cell surface, but the association is sufficiently robust that most of the capsular polysaccharide sediments with the cells in centrifugation. It has been suggested that non-covalent interactions between glycan strands, and perhaps between the capsule and other surface molecules, generate a higher-order group 1 capsule structure (53) and that this structure may play an important role in cell association. Alternatively, there is a specific anchor molecule that has gone undetected. The phenomenon of mixed LPS-linked and LPS-independent polysaccharides has led inevitably to confusion in serological designations. This confusion is compounded by the fact that bacteria with group 1 and 4 K antigens often produce another LPS-linked polymer. Thus, depending on the isolates, some of these capsular polysaccharides have been identified in serological terms as O antigens and others as K antigens. The important point to note here is that the serological entity known as the K antigen and the physical structure that is the capsule are not necessarily synonymous.

The lipid A-linked form of group 1 and 4 K antigens is termed K_LPS. The polysaccharide chain of K_LPS in isolates with group 4 capsules is long (typically >15 repeat units), and the K_LPS shows the classical modal cluster of preferred chain lengths seen upon sodium dodecyl sulfate-polyacrylamide gel electrophoresis of LPS (2, 43). This property is conferred by the activity of the O antigen chain length regulator, the Wzz protein (formerly known as Cld or Rol) (reviewed in references 100 and 147). In contrast, group 1 capsule producers have K_LPS with short K antigen oligosaccharides, in which very few molecules have more than one or two repeat units (40, 69). These isolates lack a chromosomal wzz gene, but the introduction of a plasmid-carried heterologous wzz gene generates modality in the K_LPS (28). There is one reliable structural distinction between group 1 and 4 K antigen repeat unit structures; all group 4 representatives contain an acetamido sugar (usually GlcNAc) in the repeat unit, while all group 1 examples lack GlcNAc but contain a hexose (Glc or Gal) (Fig. 2).

An additional polysaccharide, called colanic acid, is a close structural and biosynthetic relative of group 1 capsules (Fig. 2). Colanic acid is widespread in E. coli isolates that produce K antigens belonging to groups 2, 3, and 4, but the expression of group 1 K antigens and that of colanic acid are mutually exclusive (56, 58). Most of the colanic acid polymer is secreted into the growth medium (in contrast to capsular K antigens), and it was originally described as a slime polysaccharide, or M (mucoid) antigen (42). It is unknown whether this difference is imparted by features of the assembly system or by altered properties of the polymer and the surfaces with which it associates (104). In E. coli K-12, colanic acid repeat units are also found in a short LPS-linked form resembling group 1 K_LPS (78). This result is actually a little surprising since K-12 lineages carry wzz⁺ and may be expected to produce a modal distribution in the K_LPS-like unit. Colanic acid is generally not produced under typical growth conditions (rich medium at 37°C) because it is subject to complex regulation as part of the Rcs regulon (70) (see below). As may be expected, then, colanic acid is not a virulence determinant.

Overview of Genetics, Biosynthesis, and Assembly

Capsular polysaccharides belonging to groups 1 and 4 are synthesized by a blockwise polymerization process known as the Wzy-dependent pathway. Although the identification of the various components was not possible before the advent of cloning-sequencing strategies, the essential features of the Wzy-dependent pathway were actually first described in association with the LPS O antigens of Salmonella enterica serovar Typhimurium more than three decades ago. Briefly, polysaccharide biosynthesis is initiated by the reversible transfer of a glycose-1-P residue from an activated (UDP-linked) precursor to a carrier lipid, undecaprenol-P (Und-P). Successive glycosyl transfer reactions at the cytoplasmic face of the inner membrane yield Und-PP-linked repeat units. These molecules provide the polymerization substrates used by the putative polymerase (Wzy) to form the nascent polymer. The polymerization reaction is localized to the periplasmic face of the inner membrane, and the Und-PP-linked repeat units must therefore be exported across the inner membrane. The export process requires the Wzx protein. The early steps in these pathways in both O antigen and capsular systems have been reviewed extensively elsewhere (100, 146).

The genetic loci directing the production of group 1 capsules and colanic acid are similar in overall gene content and organization (Fig. 3). They are arranged as a large transcriptional unit separated into two regions by a stem-loop transcriptional attenuator (103). Downstream of the attenuator are genes required for the formation of the repeat unit structure and polymerization, including the characteristic wzx and wzy genes. This part of the locus is sufficient for K_LPS formation and has essentially the same gene content as a locus for the expression of a Wzy-dependent LPS O antigen (reviewed in reference 100). Upstream of the attenuator are genes required for polymerization control and for the functioning of the dedicated translocation pathway for the capsular polysaccharide (wza, wzb, and wzc). The first gene in the group 1 locus is wzi, and the colanic acid operon is distinguished by the absence of this gene.

Fig. 3Genetic loci responsible for the expression of the group 1 and 4 capsules and colanic acid production (146). The group 1 (K30) locus begins with four genes (wzi, wza, wzb, and wzc) required for the high-level polymerization of the capsular polysaccharide. wbaP is the first in a series of genes whose products are involved in repeat unit biosynthesis; these include genes encoding the characteristic PHPT (orange), Wzy (red), and Wzx (purple) proteins. The colanic acid biosynthesis genes show an overall similar organization, although wzi is missing and the locus is transcriptionally regulated by Rcs proteins. The genes encoding the group 4 capsule are located downstream of the colanic acid locus and essentially represent an O antigen biosynthesis cluster (in a position identical to that of the O16 locus in E. coli K-12). The production of the group 4 capsule occurs when these genes are supplemented by unlinked wza-wzb and wzc genes in the 22-min locus. The ymcDCBA genes are also involved in group 4 assembly, and Rcs-regulated homologs of these genes (yjbEFGH) are located near a pgi-linked locus, but it is not yet known whether these genes participate in the biosynthesis of colanic acid or group 1 capsules. Several housekeeping genes are shown as reference points (green), and the relative positions of these genes and the various loci suggest extensive recombination and rearrangement in this region of the chromosome. Points of regulation by RfaH and/or Rcs proteins and the stem-loop attenuator are identified by vertical black arrows.

Source: Adapted from the AnnualReviewofBiochemistry (146) with the permission of the publisher.

The transcription of the group 1 locus is driven from a constitutive promoter upstream of wzi (103) and involves an RfaH-dependent antitermination mechanism. This regulatory feature is seen in a range of genetic loci in E. coli, particularly in large operons involved in the assembly of cell surface components, including capsules and O antigens (50). The RfaH protein converts RNA polymerase to a form that favors transcript elongation (reviewed in references 6 and 116), effectively overriding stem-loop attenuators such as the one found within the group 1 capsule and colanic acid loci (103). RfaH is recruited to the transcription machinery by an 8-nucleotide ops (operon polarity suppressor) sequence located at the 5' end of the transcript. In the cps system, RfaH-mediated antitermination ensures that the transcription of the polysaccharide biosynthesis genes occurs. It may also result in high levels of the Wza and Wzc proteins relative to those of biosynthetic enzymes, consistent with the fact that Wza and Wzc exist as multimers (see below).

The regulation of colanic acid production is complex because transcription is controlled by Rcs (regulation of capsule synthesis) proteins, a system found in those members of the Enterobacteriaceae that are pathogens or commensals of the gut (34). While this system was first identified in the context of colanic acid regulation, there is a growing appreciation of the broader significance of the system in the cellular physiology of E. coli. The Rcs transcriptome is composed of more than 150 (up- or downregulated) genes implicated in surface remodeling in response to a change in lifestyle. Approximately 50% of the Rcs regulon genes encode proteins that are targeted to the envelope or are involved in envelope modifications (such as colanic acid formation) (37, 48). These proteins include cell envelope proteins induced by shock and osmotic stress conditions and others associated with swarming behavior and biofilm formation. Colanic acid production is also a feature of mature biofilms, suggesting complex temporal regulation (29). The core Rcs system involves two sensor kinase proteins, RcsC and RcsD (YojN), and a response regulator, RcsB. Upon the receipt of the appropriate signal, the phosphorelay proceeds from the autophosphorylation of the hybrid sensor kinase RcsC to RcsB via the RcsC receiver domain (111) and a histidine-containing phosphotransfer (Hpt) domain of RcsD (110). The Rcs regulon is divided on the basis of events that are RcsA dependent or RcsA independent. The colanic acid system requires the Lon protease-sensitive RcsA protein to form the functional regulator, an RcsA-RcsB heterodimer that binds to 14-bp RcsAB box sequences (145). A final component is RcsF, a predicted outer membrane lipoprotein (71). As determined from genetic epistasis data, RcsF operates upstream of RcsC and may serve to communicate information concerning physicochemical changes in the cell envelope to RcsC, but how it does this is open to speculation. The Rcs system is integrated into other cellular regulatory circuits, and the components are increasingly complex. From the perspective of biosynthesis and assembly, colanic acid can be viewed as a widespread and highly regulated form of group 1 K antigen.

In contrast to the group 1 capsule and colanic acid loci, the locus for group 4 capsules contains only the genes required for the synthesis and polymerization of the repeat units (Fig. 3). However, it cannot function alone because the gene encoding the initiating WecA enzyme is located elsewhere, as part of the enterobacterial common antigen biosynthesis locus found in most Enterobacteriaceae (77, 106). Group 1 capsules do not require WecA, but the synthesis of all E. coli O antigens studied to date is initiated by WecA (1). Together with WecA, the gene products encoded by the group 4 capsule locus are sufficient for K_LPS formation, and the modality of the resulting K_LPS is imparted by the product of the adjacent wzz gene. The translocation pathway for the LPS-free group 4 capsule is encoded by genes located outside the main biosynthesis cluster (Fig. 3), in a seven-gene operon located near appA on the E. coli K-12 chromosome (54, 97). We have termed this operon the 22-min locus due to its position on the E. coli K-12 linkage map (33). This locus encodes homologs of Wza, Wzb, and Wzc, as well as four additional open reading frames (ymcDCBA) that are all required for the formation of the group 4 capsule. The 22-min locus is missing in uropathogenic E. coli (UPEC), and its promoter is inactive in E. coli K-12 (97). ymcDCBA genes are expendable for group 1 capsules (A. Reid and C. Whitfield, unpublished data), although the activities of the proteins encoded by the adjacent wza, wzb, and wzc homologs in the 22-min locus can be detected when the copies in the group 1 capsule locus are mutated (33, 148). While this finding suggests a possible distinction between the group 1 and 4 systems, the situation is complicated by the existence of ymDCBA homologs (designated yjbEFGH) in a four-gene locus located adjacent to pgi (36). Interestingly, the yjbEFGH genes are part of the Rcs regulon and are preceded by an RcsAB box.

Biosynthesis and Polymerization of Repeat Units

Group 1 and 4 capsule biosynthesis pathways are initiated by different enzymes, although the essential features of the reactions are similar. Group 4 capsules are initiated by WecA, an enzyme belonging to a family of polyisoprenyl-phosphate N-acetylhexosamine-1-phosphate transferases (PNPT) that includes representatives in both prokaryotes and eukaryotes (140). WecA catalyzes the reversible transfer of GlcNAc-1-P to Und-P. In group 1 capsular polysaccharide and colanic acid systems, the first residue is transferred to Und-P by members of the polyisoprenyl-phosphate hexose-1-phosphate transferase (PHPT) family (140). For example, colanic acid requires a Glc-1-P transferase (WcaJ) (135), and the prototype K30 antigen begins with Gal-1-P transferase (WbaP) (32). Given the essential requirement for a lipid intermediate in group 1 and 4 capsular polysaccharide biosynthesis, it is perhaps not too surprising that PNPT and PHPT enzymes are integral membrane proteins (65, 140). In contrast to the initiation proteins, the glycosyltransferases that complete the repeat units have no transmembrane domains (TMDs). It is unknown whether their location at the membrane reflects association with membrane phospholipids or interactions with other proteins. In an intriguing hypothesis, Und has also been implicated as a scaffold for organizing glycosyltransferases (150).

Wzx proteins are integral membrane proteins involved in the export (flipping) of Und-PP-linked repeat units across the inner membrane in O antigen biosynthesis (35, 67). They belong to a family of putative polysaccharide-specific transport (PST) proteins (92). Assembly systems for group 1 capsular polysaccharides contain PST-1 proteins, whereas PST-2 proteins are found in O antigen and group 4 capsule systems. This phylogeny identifies a correlation between a specific group of the PST protein family and either a PNPT or a PHPT protein. This correlation is consistent with genetic data suggesting that Wzx is specific for the initial sugar in the lipid-linked repeat unit, perhaps via specific protein-protein interactions involving Wzx and PNPT/PHPT (73). This scenario would create a coordinated process connecting the formation of the Und-PP–repeat unit complex and its release into the export pathway (140). However, it is still not clear whether Wzx proteins are sufficient by themselves for flippase activity or if they require the participation of additional components. In principle, Und-P could also play a role in a complicated export process by helping create a local environment with biophysical properties that promote flipping (149). Membrane vesicles containing a Wzx homolog have been shown to mediate the transmembrane flipping of a water-soluble isoprenyl-PP-GlcNAc derivative (105), and the proteoliposome approach holds promise for resolving the mechanism if a Wzx homolog can be purified in sufficient quantities for functional reconstitution.

Wzy proteins have long been known to be required for polymerization, yet their catalytic activity per se has not been demonstrated. Given the complex lipid-linked substrates and the involvement of a low-copy-number integral membrane protein, definitively resolving this issue is a challenging problem. Wzy homologs are involved in a chain extension reaction occurring in the periplasm (75). They catalyze the transfer of a nascent Und-PP-linked glycan to an incoming (exported) Und-PP–repeat unit complex, essentially extending the polymer at the reducing terminus one repeat unit at a time (reviewed in reference 100). Recent genetic data from analyses of O antigen biosynthesis strongly suggest that Wzy and Wzx form a complex that imparts a level of specificity to the Und-PP–repeat unit export system (72). In serotypes K30 (group 1) (32) and K40 (group 4) (2), wzy mutants lack capsules and add only a single K antigen repeat unit to the lipid A core in K_LPS. The phenotypes of wzy mutants with respect to these systems provided the first direct evidence that the two forms of K antigen (i.e., capsular antigen and K_LPS) share common repeat unit donors and polymerization machinery.

Once polymerization is complete, the capsular polysaccharide must be released from Und-PP into the translocation pathway. While the release of O antigens and K_LPS begins with the WaaL-mediated ligation reaction, the corresponding step in capsular polysaccharide biosynthesis has not been identified. Sequence data have identified shared motifs located in the periplasmic loops of the Wzy and WaaL (O antigen ligase) proteins (117). The similarities in these proteins are consistent with the fact that both proteins manipulate Und-PP-linked intermediates in their respective proposed activities. It is feasible that the release of the capsular K antigen from Und-PP into the translocation pathway is a side reaction of Wzy, in effect an abortive polymerization reaction step depending on altered protein-protein interactions in a complex multiprotein assembly system.

The Wzc Tyrosine Autokinase Is Essential for the Formation of High-Molecular-Weight Capsular Polymers

The biosynthesis of colanic acid and capsular polysaccharides in groups 1 and 4 requires the activity of a polysaccharide copolymerase 2a (PCP-2a) protein (83) known as Wzc. Wzc proteins are tyrosine autokinases and have been identified in many polysaccharide biosynthesis systems in gram-negative and gram-positive bacteria. Although Wzc and the O antigen chain length regulator Wzz (PCP-1) show some similarity in membrane topology, Wzc homologs have a larger periplasmic domain and possess a unique C-terminal cytoplasmic kinase domain (31). E. coli K30 mutants with wzc defects are unable to make detectable amounts of capsular K antigen but are still able to polymerize K_LPS (90, 148), suggesting that Wzc represents a branch point in assembly and is the first committed component in capsular K antigen formation. Assessing the contribution of Wzc in group 1 capsule assembly is complicated by the fact that most isolates have an additional paralog, known as Etk, encoded by the 22-min locus (54, 97). Etk contributes at a low level to the biosynthesis of the group 1 K30 capsular polysaccharide, and its activity is clearly seen only when wzc is mutated (148). In colanic acid biosynthesis, Etk activity goes unnoticed (143) because the operon in which the Etk gene resides is not transcribed in E. coli K-12 (54). In contrast, Etk provides the primary source of the tyrosine autokinase in group 4 capsule biosynthesis (97).

The kinase domain of Wzc (and Etk) contains essential ATP binding motifs (Walker A and B), as well as a tyrosine-rich region located at the extreme C terminus of the protein. The phosphorylation of the C-terminal domain is essential for the formation of group 1 capsular polysaccharides (90, 148). Wzc exists as a tetramer with extensive periplasmic contacts between the monomers (21). The formation of an oligomer provides a structural model to explain the observed transphosphorylation occurring between monomers (148). In vitro, the colanic acid system Wzc protein undergoes intraphosphorylation outside the kinase domain (residue Y569) as a prerequisite for efficient transphosphorylation (46). The structure of the kinase domain of Etk (the Wzc homolog) has recently been reported (64b). The solved structure identifies a kinase mechanism that is radically different to eukaryotic kinases and shares more resemblance to ATPases, supporting the observation of robust Wzc ATPase activity in vitro (128). The kinase domain can be modeled convincingly into each monomer in the cryoelectron microscopy structure of the Wzc tetramer (64b). It has been reported that the phosphorylation of Y569 may modulate the capacity of Wzc for ATP hydrolysis and transphosphorylation (128), and the solved structure affords a model for how this process may occur (64a). How these processes play out in vivo is more complicated since Y569 is not essential for the formation of colanic acid polysaccharide (86) or K30 capsular polysaccharide (90) in whole cells. Despite the conserved sequence of Wzc proteins in the group 1 and colanic acid systems, the behaviors of these proteins are surprisingly different. In K30 antigen biosynthesis, phosphorylated Wzc is essential for capsule formation (90, 148). In contrast, phosphorylated Wzc inhibits colanic acid production (86, 143). It is intriguing that the colanic acid and serotype K30 homologs do seem to function in similar ways when examined in the same biosynthetic context (i.e., genetic complementation of a wzc mutation in the group 1 K30 capsule system). This observation suggests that other specific protein partners may have an impact on the activities of Wzc homologs (104).

In the K30 (group 1) system, Wzc proteins isolated from growing bacteria represent a heterogeneous population containing zero to three phosphate residues (22). Within the last 17 residues, there are seven tyrosines that can be phosphorylated, but no single tyrosine residue is essential, and none is sufficient for function (90). The phosphorylation state of Wzc is modulated both by its autokinase activity and by its dephosphorylation by a cognate phosphotyrosine protein phosphatase, Wzb. Wzb is essential for the production of both group 1 capsular polysaccharide (90) and colanic acid (143). The collective data on the influence of Wzc and Wzb on group 1 capsule formation have been interpreted by a model in which Wzc must cycle between the phosphorylated and nonphosphorylated state (90, 148). Alternatively, the overall biosynthesis and assembly process involves discrete reactions that require either phosphorylated or nonphosphorylated forms of Wzc (86). One subtle effect of the multiple potential phosphorylation states is an influence on the size heterogeneity of the polymer forms (86, 90). It is tempting to speculate that Wzc and Wzz have similar functions in polymer biosynthesis that would justify their inclusion as different branches in the putative PCP family (83), but if this is the case, why does only Wzc have the kinase domain? The answer may lie in additional interactions involving Wzc and outer membrane components in the assembly system (see below).

To add to the complexity, it is now apparent that Wzc can phosphorylate other proteins involved in sugar nucleotide biosynthesis, such as UDP-Glc dehydrogenase in E. coli K-12 (47). This has direct impact on colanic acid biosynthesis but is only one part of a complex regulatory system that also influences the modification of LPS in a process that alters susceptibility to cationic polypeptides (64a). In a related capsule-assembly system in a gram-positive bacterium (Streptococcus thermophilus), the corresponding kinase has also been implicated in the phosphorylation of EpsE, the initiating PHPT protein (82). Notably, the results from preliminary studies of serotype K30 show that wzc mutations result in dramatic reduction in the WbaP activity (A. Reid, A. Cockburn, and C. Whitfield, unpublished results), suggesting that the acapsular phenotype of the wzc mutant may be due to dramatically reduced flow in the polymer biosynthesis reactions. Whether this reduction is due to the phosphorylation of WbaP or Wzc’s performance of a structural role in coordinating the organization of an efficient multienzyme complex including WbaP remains to be seen.

Wza Forms a Multimeric Structure, the Polysaccharide Translocon, in the Outer Membrane

The final stage in capsule assembly involves the translocation of large hydrophilic capsular polysaccharides (with sizes of 10⁵ to 10⁶ Da) across the outer membrane. Achieving this translocation without compromising the integrity of the outer membrane barrier is a considerable challenge. The key component in this process is a capsule translocon formed by sodium dodecyl sulfate-stable oligomers of the outer membrane lipoprotein Wza (10, 30, 33, 84). Wza is a representative of a family of proteins designated outer membrane auxiliary (OMA) proteins, whose members are associated with capsular polysaccharide and extracellular polysaccharide biosynthesis loci in a variety of different gram-negative bacteria (92). E. coli K30 mutants lacking Wza are unable to assemble a capsule on the cell surface (33). Electron microscopy analysis of purified Wza incorporated into two-dimensional proteolipid crystals revealed ring-like structures in which each ring comprises eight monomers (84). Subsequent single-particle analysis of cryonegatively stained multimers (10) established that Wza is not a conventional outer membrane channel but, instead, forms a complex significantly larger than the current estimates of the thickness of the outer membrane.

A detailed structural understanding of the remarkable Wza complex has been possible because of the high-resolution X-ray crystal structure (30) (Fig. 4). The Wza translocon comprises four ring domains (R1 to R4), with each of the eight monomers contributing to each domain. R1 creates a filled concave face at the periplasmic base of the structure that opens into a cavity with an internal diameter of >25 Å that continues through R2. The folds in R2 and R3 are duplicated, but R3 has a substantially larger central cavity (diameter, 105 Å). The final domain is an amphipathic helix, and the eight C-terminal helices combine to form a novel α-helical barrel that spans the outer membrane (22). The acylated N termini of the monomers are wrapped around the top of R3, in a perfect position to be intercalated into the inner leaflet of the outer membrane. The internal diameter is reduced from ~30 Å at the interface with R3 to 17 Å at the open end of the structure. Wza is the first identified outer membrane channel protein that is not formed from a β-barrel. The overall structure has been described as an "amphora" with an open "neck" formed by R4.

Fig. 4Structure of the Wza-Wzc complex, a putative translocation system for group 1 and 4 capsules and for colanic acid. Panel A shows the crystal structure of the Wza octamer (30). In the ribbon diagram on the left, each of the monomers is identified in a different color and the space-filling shape representing the central cavity is highlighted in light brown. The four rings are clearly evident. When the side chains are filled in, in the diagram on the right, the amphora structure of the complex is revealed, with an open neck and a closed bottom (oxygen, red; nitrogen, blue; selenium and sulfur, gray). Lipid chains are identified as black spheres. Panel B shows the three-dimensional structure of the Wza-Wzc complex reconstructed by electron microscopy analysis of cryonegatively stained single particles. The images on the left are a surface-rendered representation. On the right, slices through the structure illustrate the central cavity (22). In Panel C, the crystal structure of Wza (purple) and the electron microscopy structure of the Wzc tetramer (yellow) are superimposed onto the overall structure of the Wza-Wzc complex (22), identified as a green wire frame. The orange regions on the Wzc structure denote the C termini containing the kinase domains; these were identified by Ni-nitrilotriacetic acid-gold labeling of an N-terminal hexahistidine tag. This image highlights the broadening of R4 (PES domain) that occurs upon interaction with Wzc. OM, outer membrane.

Source: The crystal structure of the Wza octamer is adapted from Nature (30), and the illustration of the central cavity and the structure of the Wza-Wzc complex are adapted from ProceedingsoftheNationalAcademyofSciencesoftheUnitedStatesofAmerica (22) with the permission of the publishers.

The residues lining the internal surface of the structure are predominantly polar, and there is little conservation among Wza homologs from different bacteria, suggesting passive involvement in export. This idea is consistent with essentially identical Wza proteins’ being involved in the export of different group 1 K antigens (102) and the ability of the colanic acid system Wza protein to translocate group 1 capsular polysaccharide in genetic complementation experiments (104).

The crystal structure is closed at the periplasmic face (30). If Wza oligomers form the efflux channel, other factors must come into play to gate access to the periplasmic part of the Wza oligomer in vivo. One logical approach would involve a conformational change directed by protein-protein interactions. Cross-linking experiments initially identified interactions between Wza and Wzc (84), and the association of these proteins was definitively established by the visualization of Wza-Wzc complexes in cryonegatively stained electron microscopy images (22). Wzc-Wza forms a remarkable contiguous complex from the cytoplasm to the cell surface (Fig. 4). In the heterooligomeric complex, the solid periplasmic domain seen in isolated Wzc tetramers (21) is opened up and R1, representing the closed face of Wza, is broadened. A conserved sequence referred to as the polysaccharide biosynthesis-export (PES) motif (Pfam02563) is shared by members of the OMA protein family and is located in R1 (30), and the residues of this sequence are candidates for structural or interfacial elements. Conformational changes in the complex create a portal connecting the periplasm to the central cavity, thus offering a potential route for the nascent polymer to move from the polymerization site at the periplasmic face of the inner membrane to the exterior of the cell. However, it is not clear how the nascent polymer may access the central cavity in the first place. It may do so by being produced in situ within the complex, or more likely, it may enter via the sides of the complex within the periplasm (22). The phosphorylation status of Wzc does not seem to affect complex formation, but changes in phosphorylation may influence the conformation of the channel (perhaps modulating the opening and closing of the complex). The biggest challenge remains to prove that the polymer really does pass through the identified conduit.

The potential biosynthesis-assembly coordination offered by Wza-Wzc association in a complex is evident in mutant phenotypes and has led to a working model (Fig. 5). One may predict that a wza mutant would produce polymers accumulating in the periplasm, but these mutants actually produce no high-molecular-weight material and have a phenotype similar to that of wzc mutants (33, 84). This outcome is consistent with a feedback process with Wzc as the critical element coupling the biosynthesis and translocation events. An acylation-defective Wza derivative forms an unstable complex, presumably reflecting a failure to interact properly with the outer membrane (84). Periplasmic polymers do accumulate, consistent with a normal engagement with Wzc but an inability to complete the translocation process.

Fig. 5Hypothetical complex for the formation of group 1 capsules. Und-PP-linked repeat units are synthesized by the PHPT protein, WbaP, and additional glycosyltransferases. Wzx exports (flips) the lipid-linked repeat units across the inner membrane. Wzy catalyzes polymerization. The tyrosine autokinase (Wzc) and its cognate phosphatase (Wzb) form a tetramer and undergoes phosphorylation-transphosphorylation-dephosphorylation events. The resulting protein occupies a central position in the overall process, by controlling the high-level polymerization of the polymer and forming, together with octamers of the outer membrane translocon protein (Wza), an envelope-spanning complex for the translocation of polymers to the cell surface. Details of each biosynthetic step are described in the text. It is uncertain whether other housekeeping functions also participate in the overall process. Additional cell envelope proteins may be required, such as the YmcDCBA/YbjEFGH paralogs. The core components of the complex are required for group 4 capsule and colanic acid biosynthesis and assembly.

Source: The diagram of the loci is adapted from the AnnualReviewofBiochemistry (146) with the permission of the publisher.

The Wza-Wzc complex overcomes the translocation challenges resulting from the complexity of the cell envelope. Several decades ago, Manfred Bayer described regions where the inner and outer membranes come into contact (Bayer junctions, or zones of adhesion). In one of his experimental systems, these controversial regions corresponded to sites where new group 1 capsular polysaccharide emerged following the activation of a conditional mutation (8, 9). It is tempting to speculate that the capsule export sites identified by Bayer are composed of Wza-Wzc complexes. How these complexes are assembled and how they traverse the peptidoglycan layer are important questions that remain unanswered.

Other Proteins Acting Late in the Assembly Process

Most of the effort in unraveling the biosynthesis-assembly functions has focused on genes that are dedicated to biosynthesis and assembly and, particularly, genes within the main genetic loci. The functions of some proteins associated with these processes are still unknown. For example, only group 1 systems have Wzi, an outer membrane β-barrel protein (101). Wzi seems to have an impact on the surface association of K antigen in some way, but its precise role remains to be discovered. The potential involvement of YmcDCBA and/or YjbEFGH (36, 54, 97) adds to the number of periplasmic-outer membrane components. YmcDCBA proteins are certainly needed for group 4 capsules, and it seems unlikely given the overall conservation of Wza, Wzb, and Wzc that they (or their homologs) play no role in group 1 K antigen and colanic acid production. Based on the results of database searches, YmcD (YjbE) is predicted to be a small and unusually threonine (or serine)-rich exported protein. YmcC (YjbF) and YmcA (YjbH) are both predicted to be outer membrane lipoproteins. YmcA is a predicted β-barrel structure. The crystal structure of YmcC has been deposited in the Protein Data Bank (identification no. 2IN5) and reveals a β-strand-rich fold with two chains in the asymmetric unit. YmcB (YmcG) is a predicted periplasmic protein. Homologs of YmcCBA are widespread according to BLAST search results. Whether these proteins support the integrity of the Wza-Wzc complex or provide additional (currently unknown) activities awaits investigation. It is also important to consider that there may be other important players, including genes encoding normal cell envelope housekeeping functions.

ASSEMBLY OF CAPSULES BELONGING TO GROUPS 2 AND 3

Structure and Organization on the Cell Surface

Group 2 and 3 capsules are found in clinically prevalent isolates causing extraintestinal infections, such as neonatal meningitis E. coli and UPEC isolates, and are coexpressed with a wide range of O antigens. The group 2 and 3 capsules are distinct from those in groups 1 and 4 in both repeat unit structure (Fig. 2) and their linkage to the cell surface. As purified from cultures, group 2 capsular polysaccharides exist in two forms; both are high-molecular-weight polymers, and 20 to 50% of the chains have a diacylglycerophosphate moiety at the reducing terminus (55). There is some debate concerning the detailed chemistry of the lipid linkage. In the original characterization of the lipid terminus, using the polysialic acid serotype K92 capsular polysaccharide, the data were consistent with a reducing terminal Neu5Ac residue in the polysaccharide being linked directly to the lipid (44). However, in serotypes K12 and K82, the lipid termini are proposed to be linked to the polysaccharide chains via a linker-specific 3-deoxy-d-manno-octulosonic acid (Kdo) residue (118). While there are some biosynthetic data supporting the Kdo linkage (see below), definitive structural analyses have not been reported. It remains a mystery why the Kdo residue would be found as a linker in only a subset of group 2 capsular polysaccharides, given the overall conservation of their biosynthesis and assembly.

Nonlipidated group 2 polysaccharides are assumed to be the result of the lability of the phosphodiester linkage, but the impact of this feature on the extent and integrity of the capsule’s surface association and, ultimately, its influence on E. coli biology is not yet understood. Recent data suggest that the properties of the outer membrane are also important in the surface association of group 2 capsules. Mutants with certain defects in LPS core oligosaccharide biosynthesis are unable to efficiently retain the capsule at the cell surface and release capsular polysaccharide into the culture medium (136). Whether this phenotype results from the release of an acylated polymer from the surface or the enhanced cleavage of the phosphodiester linkage has not been determined.

Overview of Genetics, Biosynthesis, and Assembly

The chromosomal genes encoding the functions necessary for the biosynthesis and export of group 2 and 3 capsules are traditionally located in a locus called kps, positioned near serA (141). Detailed information on the K1 and K5 (group 2) loci is available. However, in at least one case (that of serotype K15) (119), the locus is located elsewhere, on a pathogenicity island. In the biosynthesis of group 2 capsular polysaccharides, all of the steps up to (and probably including) the termination of chain elongation are completed at the cytoplasm-inner membrane interface. The characteristic element of the assembly pathway is the ATP binding cassette (ABC) transporter, which exports the nascent polymer across the inner membrane. The polymer then becomes available to a translocation pathway that completes its transfer to the cell surface.

Group 2 capsule loci have a characteristic three-region format in which two blocks of conserved genes flank a serospecific central domain (Fig. 6) (7, 146). The conserved regions (1 and 3) encode proteins required for the export of the polysaccharide across the cell envelope. Region 3 encodes KpsMT, the components of the characteristic ABC transporter. While the role of KpsMT is well-established, the same is not true of some of the region 1 gene products, KpsFEDUCS. Some region 1-encoded proteins are required for the polymer to leave the cytoplasm, while others are clearly involved in completing the process of polymer translocation across the periplasm and outer membrane. The conservation in the region 1 and 3 gene products is reflective of roles that are independent of the polymer structure. Consequently, these proteins can be exchanged between serotypes with no impact on function (109). In contrast, genes within the central domain (region 2) encode proteins involved in polymer biosynthesis. They include enzymes required for any sugar nucleotide donors that are unique to the K antigen (note that some donors have additional cellular functions and are typically made by conserved housekeeping enzymes) and glycosyltransferases that build the polysaccharide from the activated donors. The variations in the size and complexity of this region reflect the range of different group 2 capsule structures (Fig. 2), and recombination within this region can account for serotype variation.

Fig. 6Genetic loci responsible for the expression of the group 2 and 3 capsules (146). The group 2 kps locus comprises serotype-specific region 2 flanked by two regions (1 and 3) conserved across group 2 serotypes. The two genes in region 3 encode the ABC transporter. Within region 1, four genes (kpsCDES) are required to various degrees in polymer export and the additional two genes (kpsFU) encode proteins involved in CMP-Kdo biosynthesis. Region 2 genes encode enzymes for polymer biosynthesis, and the complexity of this region corresponds to the repeat unit structures (Fig. 2). While most of the loci occupy the same area on the chromosome (near serA), the genes for K15 are found on a pathogenicity island. Group 3 loci contain region 3 genes and some (but not all) region 1 genes, and the loci show extensive evidence of rearrangement. Genes marked by asterisks are incomplete, perhaps as a result of past recombination-rearrangement events (146).

Source: The diagram of the loci is adapted from the AnnualReviewofBiochemistry (146) with the permission of the publisher.

A key property of group 2 capsules is that they are produced only at temperatures above 20°C. This property is apparently conserved in examples like K15 that are not found at the classical serA-linked location (39). In the K1 and K5 loci, thermoregulation is mediated via promoters located upstream of both regions 1 and 3 (19, 115, 126). A promoter(s) upstream of kpsM generates a large transcript encompassing regions 3 and 2 in a process that requires the antitermination activity of RfaH (134). A σ⁷⁰-dependent promoter upstream of kpsF generates a transcript encompassing the region 1 genes, but it is not clear whether there are additional (minor) promoters for this region (19, 126). Regulatory proteins including BipA, H-NS, integration host factor, and SlyA have all been implicated in a complex multifactorial control system for kps (23, 115). It remains to be seen how these various elements interact in the overall process and what cues the regulators respond to in vivo. One example of the potential complexity is the downregulation of kps when type I fimbriae in UPEC isolates engage their receptors (120).

In at least one case, genes outside kps also influence the repeat unit structure. The K1 antigen is subject to form variation dictated by the presence or absence of O acetylation (Fig. 2) (88). O acetylation results in significant alteration in the antigenic epitopes detected in the polysaccharide; the acetylated form is not recognized by K1-specific diagnostic antiserum. The O-acetyltransferase (NeuO) is part of a lysogenic bacteriophage-like element that is unlinked to kps. The on-off modulation of O acetylation occurs via slipped-strand DNA mispairing in vivo in a region within neuO (27, 60). Given this mispairing, coupled to the nonstoichiometric addition of O-acetyl residues, the potential for structural variations in this system is substantial (59).

In contrast to the wealth of data available for group 2 capsules, information on group 3 loci is largely confined to the findings of comparative studies. At least some group 3 loci are located near serA, and the bulk of the data suggest that they encode a series of conserved functions reflecting a common mechanism of assembly. However, the highly conserved format characteristic of group 2 kps loci is not evident (Fig. 6) (7, 146), suggesting extensive recombination and gene shuffling within the loci (20, 96). One highly important consequence is the loss of thermoregulation, a feature distinguishing group 2 and 3 capsules.

Biosynthesis of Group 2 (and 3) Capsular Polysaccharides

Chain Elongation.

Group 2 capsular polysaccharides are synthesized in the cytoplasm and are then exported to the periplasm by the ABC transporter. In vivo, these processes are likely to be coordinated. For the purpose of discussion, the process of biosynthesis is conceptually separated into three phases: initiation, chain extension, and termination (linked to export). While there exists a significant amount of information concerning chain extension for some serotypes, the processes of initiation and termination are still unclear. To date, there has been no direct study of the biosynthesis of group 3 capsules, but due to the conserved genetic functions, it is reasonable to assume that they will follow models established for group 2 capsules.

Initial studies of group 2 capsular polysaccharide biosynthesis exploited membrane preparations as a source of enzyme to extend either endogenous or exogenous acceptors. Considerable progress in understanding chain extension has been possible because the purified enzymes are able to extend synthetic acceptors in vitro, without any requirement for membranes or other proteins. The polysialyltransferases (NeuS enzymes) responsible for the formation of polysialic acid in serotypes K1 and K92 represent some of the best-characterized examples. In serotype K1, NeuS is the only enzyme required for chain extension (132); it transfers Neu5Ac residues to the nonreducing terminus of the growing chain in an α-2,8 linkage (112). The corresponding NeuS homolog in serotype K92 is responsible for the formation of alternating α-2,8 and α-2,9 linkages (74). The two E. coli NeuS enzymes show 82% identity, and some important conserved residues and motifs have been identified by sequence comparisons and biochemical approaches (41, 131).

The polymerization of the backbone of the K4 antigen may be anticipated to be more complex than that of the K1 and K92 antigens due to the requirement for two different sugar nucleotide donors (Fig. 2). However, the chain extension is achieved by a single enzyme (the chondroitin synthase KfoC), containing both β-GalNAc and β-GlcA transferase activities (66, 85). In contrast, the comparable heparosan synthase activity in serotype K5 is the product of two enzymes, KfiA (α-GlcNAc) and KfiC (β-GlcA ) (51). A single bifunctional heparosan synthase from Pasteurella multocida serotype D has domains with sequences similar to those of both KfiA and KfiC (26), but the impact (if any) of a one- versus two-polypeptide format on chain extension is unknown. The origin of the β-Fru residues in the K4 repeating unit (Fig. 2) is currently unknown.

One controversial question has concerned the exact mechanism of action of these enzymes—are they truly processive? Processive glycosyltransferases are expected to produce polymers predominantly with high molecular weights through rapid successive glycosyl transfer reactions. In contrast, nonprocessive activity would involve the release of the growing chain at each step and potentially result in additional products with shorter chain lengths. Again, most information has come from the polysialyltransferases. When the enzymes are examined in vitro, the available evidence suggests that NeuS operates in a nonprocessive manner (41, 144). However, in vivo data for K1 are consistent with a processive model (131). A nonprocessive mechanism raises particularly interesting questions for K92, in which the enzyme generates a product with alternating α-2,8 and α-2,9 linkages (144), or for the K4 and K5 systems, in which two different sugar nucleotide donors and potentially two catalytic polypeptides are involved. It remains to be established whether the different interpretations of in vivo and in vitro activities reflect the use of analytical methods with various levels of sensitivity. Alternatively, additional proteins and the (unidentified) endogenous acceptor in the group 2 capsule system may convert an inherently nonprocessive enzyme to one that is processive. At one level, these are detailed issues of fundamental interest to those interested in the structure-function of glycosyltransferases. However, the precise mechanism does have a bearing on the biology of these organisms because enzymes like NeuS must rapidly elongate polymers to form high-molecular-weight products that effectively cover the cell surface in a protective capsule coat. The in vivo products show a relatively narrow size distribution (maximum length, 160 to 230 residues) (98). Intuitively, a chain length control mechanism seems important, and control over the processivity of NeuS (or comparable glycosyltransferases) would provide one possible approach. Whether the various serotypes all follow the same model remains to be seen. Regardless of the exact extension mechanism, the process must end at some point, and the events (and components) leading to termination are unknown. The ABC transporter may itself play an active role in termination. In this context, it would be very informative to establish whether export occurs only after chain termination or if it can begin while extension is in progress (see below).

Several group 2 capsules have structures in which the main chain is modified by side chains, such as β-Fru in K4, or nonstoichiometric substitutions, such as O acetylation in serotype K1 (Fig. 2). In both cases, the main chain extension is independent of these substitutions. In K4, chain extension cannot occur on fructosylated acceptors (66). In K1, the O acetylation process mediated by NeuO is form variable (27). In vitro, the O-acetylase can utilize large acceptors with a minimum of 14 repeat units (12, 49), and in vivo, it modifies the polymer that accumulates in an export-deficient mutant (27). The collective data therefore identify a postpolymerization but preexport activity. The phase variation of O acetylation results from slipped-strand mispairing in a series of tandem repeats within neuO. These mispairs give rise to translational variants with different numbers of a heptad repeat close to the N terminus. Interestingly, the length of the repeat domains influences the kinetic properties of the enzyme (12), so the situation is potentially more complicated than a simple on-off process. A small amount of additional O acetylation (2 to 4%) is derived from a separate (minor) pathway involving a bifunctional catalytic protein, NeuD. The NeuD activity is offset by that of an O-esterase domain in the bifunctional NeuA protein (129). In contrast to those of the NeuO pathway, the substrates for NeuD are Neu5Ac monomers. The precise role of this additional O acetylation pathway in the biology of encapsulated E. coli remains unknown.

Initiation Reactions.

Current data indicate that the NeuS polysialyltransferases from E. coli are unable to initiate de novo synthesis of polysialic acid (74, 131, 132, 133). The nature of the endogenous acceptor is pivotal in understanding the initiation process. Early studies with serotype K1 identified compounds with properties consistent with Und-P-Neu5Ac in membrane preparations incubated with CMP-Neu5Ac (137), suggesting a parallel with the group 1 capsule biosynthesis process. Whether these compounds form a Neu5Ac donor or an acceptor for chain extension has not been resolved, and longer chain intermediates linked to Und-P were not identified. Subsequent attempts to isolate Und-P-linked intermediates from the K5 biosynthesis system were unsuccessful (38). This result has led to an ongoing debate concerning whether Und-linked intermediates are involved at all in group 2 capsular polysaccharide biosynthesis or whether the biosynthesis of glycans in the different serotypes utilizes different initiation processes. The resolution of the situation will require the identification of the endogenous acceptor that is used in representative systems.

If Und-P is not the endogenous acceptor, what are the other candidates? An obvious possibility is diacylglycerophosphate (or diacylglycerophosphate-Kdo), since this is present at the reducing terminus of polymers that accumulate in some export mutants (19). There is precedence for alternative lipid acceptors of this type in microbial glycobiology. Examples include phosphatidylinositol-containing mannosides in mycobacteria (11) and diacylglycerolphosphate in lipomannan and glucolipid biosynthesis in Micrococcus luteus (91) and Mycoplasma pneumoniae (61), respectively. Along similar lines, the processive Streptococcus pneumoniae type 3 synthase uses a phosphatidylglycerol acceptor when expressed in E. coli (17). The isolation and full structural characterization of biosynthetic intermediates in the group 2 capsular polysaccharide system are warranted and will inevitably shed light on the overall process. However, such approaches are hampered by the small amounts of material that are made in membrane preparations.

An indirect strategy that has been employed to get at the question of the initial acceptor is to examine the minimal gene set required for de novo polysaccharide biosynthesis in E. coli. In the case of polysialic acid, membranes from bacteria expressing NeuE and NeuS, together with the region 1 products KpsCS, can form significant amounts of de novo polymer (3, 4). The requirement for NeuE and KpsCS may reflect their role in the synthesis of the endogenous acceptor, or they may be required as structural proteins to allow NeuS to utilize an existing acceptor (or perhaps they play both roles?). NeuS adds Neu5Ac residues to a polysialyl (or oligosialyl) acceptor. It is not unreasonable to predict that the transfer of the first Neu5Ac residue to the unknown endogenous acceptor may require a different initiating sialyltransferase. While the minimal gene set may implicate NeuE as a candidate, mutants lacking neuE can still form polysialic acid (18). Furthermore, neuE defects are overcome by the overexpression of neuS, suggesting that NeuE may be more important for stabilizing the NeuS polysialyltransferase or organizing it into a functional biosynthetic complex (130). Given the overall conservation of the group 2 genetic loci and the retention of region 1 homologs in groups 2 and 3, the elucidation of the initiation strategy for one serotype may be informative for all.

ABC Transporter-Mediated Export across the Inner Membrane

The two gene products from region 3 form the characteristic ABC transporter. ABC transporters use the energy of ATP binding and hydrolysis to transport molecules across membranes. They are evolutionarily conserved throughout nature, and they import and export an array of molecules from small solutes to, in the current context, large polysaccharides. The core ABC transporter consists of four domains: two TMDs and two nucleotide binding domains (NBDs) that contain the conserved sequence motifs involved in ATP binding and hydrolysis (reviewed in references 24 and 52). These are arranged in different formats depending on the precise transporter. In the capsule system, the transporter comprises two polypeptides each of KpsM (TMD) and KpsT (NBD) (reviewed in references 14 and 125).

The critical role of the ABC transporter is clearly evident from phenotypic analyses of mutants deficient in KpsT (13, 63, 94). These mutants are acapsular, but they retain the ability to synthesize polymers that accumulate at the interface of the cytoplasm and the inner membrane. The precise nature of the exported substrate is still controversial; does it carry the diacylglycerophosphate(-Kdo) modification or not? Opposing results obtained by different methods have been reported for K1 (19) and K5 (63). Both studies relied on the properties of lipidated molecules rather than full structural determination, because of the limited amounts of material that can be purified. Several other K1 mutants also accumulate lipidated polymers (19) (see below), suggesting that the modification in this system occurs before transport.

Crystal structures are available for a number of isolated NBDs, as well as four intact transporters, and these, together with the findings of biochemical and biophysical investigations, have contributed significantly to generating the working models of ABC transporter function (52). Perhaps the most relevant ABC transporter structure for capsule export is the Sav1866 drug efflux transporter from Staphylococcus aureus (25). As in other ABC transporters, there is a clear translocation pathway in the structure but the transporter adopts conformations in which the opening of the cavity is either inward or outward facing, to facilitate an alternating access and release process. Conformational changes at the NBDs, induced by the binding and hydrolysis of ATP, are transmitted to the TMDs (reviewed in reference 52). While these models address the role of ATP turnover in the conformational state of ABC transporters, it is still unclear how a large capsular polysaccharide is introduced into the channel and exported. Does the process export the nascent polymer in a single export step or does the polymer thread through in sequential steps? How is the polymer recognized to initiate the export process and activate ATP binding hydrolysis at the NBDs? The conserved sequences of KpsM and KpsT (93, 127), coupled to their ability to cross complement ABC transporter mutations in different serotypes (109), argues strongly that they do not recognize the repeat unit structure per se. That conclusion does not preclude the recognition of the polymers via some general physical property, since all are anionic polymers. It is conceivable that the diacylglycerophosphate(-Kdo) modification (or another unidentified lipid acceptor; see above) represents a conserved export signal, as well as providing an anchor for the polymer at the cell surface. An interesting facet of such a model is that export can potentially begin before the chain extension process is completed.

Assembly Functions Encoded within Region 1

The region 1 gene products also play roles in the export of capsular polymers, but they appear to act at two different stages in the process, influencing export across the inner membrane as well as translocation across the periplasm and outer membrane. As in the case of group 1 capsule assembly, evidence has been reported for a transenvelope export complex, so that a shared capsule-defective phenotype (or phenotypes) may result from defects in any protein in the complex.

KpsCS are cytosolic proteins that have been implicated in the addition of the diacylglycerophosphate modification (108), and the absence of these proteins leads to the cytoplasmic accumulation of polymers (15, 19). Unlike kpsMT mutants, in which the polymer accumulates at the interface of the cytoplasm and the inner membrane, the polymer in kpsCS mutants accumulates in electron transparent cytoplasmic domains that have been called lacunae (19). Like those regarding ABC transporter mutants, reports concerning the nature of the accumulated polymer have been contradictory. In serotype K1, the polymer is apparently lipidated (19), whereas in the corresponding mutants of serotype K5, the polymer behaves as if the modification is absent (15). Mutations in the well-conserved kpsCS homologs in Neisseria meningitidis serogroup B (lipBA) provide an informative comparison. They result in the accumulation of intracellular polysialic acid modified with a lipid terminus whose structure corresponds to that of the native surface-expressed polymer (139). Thus, in the polysialic acid systems, the role of KpsCS seems to be confined to export. Whether the result seen with K5 reflects the potentially different chemistry of the lipid terminus (i.e., the Kdo linker) from that of the rest of the polysialic acid awaits further research.

KpsF and KpsU are homologs of KdsD (formerly known as YrbH) and KdsB, respectively. These are enzymes involved in the biosynthesis of the activated sugar nucleotide CMP-Kdo (57, 79, 80, 114). KdsDB are present in all E. coli isolates because of the essential requirement for Kdo in the biosynthesis of LPS. At a simple level, the additional capacity offered by these enzymes may relate to the involvement of Kdo in the lipid terminus. Given the redundancy in these enzymes, it is then not surprising that KpsF and KpsU mutations do not eliminate capsule formation. However, there is a reduction in the amount of capsule in a K1 kpsF mutant, and lipidated polymer accumulates (18, 19). This phenotype is hard to explain. Comparison with the meningococcal system is particularly informative here because some meningococcal isolates can survive without LPS, so mutants lacking CMP-Kdo synthesis (unlike the corresponding E. coli mutants) are viable. In the absence of CMP-Kdo, capsule expression is reduced in several different serotypes, regardless of the whether the capsular polysaccharide repeat unit contains Neu5Ac or not (138). Thus, it appears that Kdo (or the enzymes synthesizing CMP-Kdo) may be involved somewhere in the assembly pathway independently of the structure of the reducing terminal modification. Along these lines, it has been speculated that the influence of KpsF may have little to do with its activity in CMP-Kdo synthesis but may instead be related to additional regulatory roles of the protein (129). Alternatively, these proteins have an unidentified indirect effect on group 2 capsule biosynthesis. The significance of the absence of genes encoding KpsFU proteins in the genetic loci for group 3 capsule expression has not yet been resolved.

KpsCSFU proteins all contribute in some way to an efficient enzyme complex for group 2 capsule assembly. Membranes prepared from serotype K5 mutants affected in the corresponding genes show reduced K5 polysaccharide biosynthesis activity (15, 16), and it is now known that KpsCST proteins are all required for efficiently targeting the polymer biosynthesis (Kfi) proteins to the membrane (107). In serotype K1, KpsS is required for optimal endogenous polysialyltransferase activity in vitro (142) and KpsS is required for the attachment of NeuS to the membranes (133). Recent studies have established multiple interactions between KpsC and other cytosolic components of the K1 biosynthesis machinery (130). Cross-linking approaches in the K5 system identified KpsS to be associated with KpsMT (76), and KpsS is required for the targeting of enzymes for biosynthesis (see above) as well as export (76, 107). The available evidence is therefore consistent with a multiprotein biosynthesis-export complex. In elegant in vivo experiments with E. coli K1, it was established that nascent polysialic acid remains in a form that is inaccessible to an active neuraminidase, suggesting that the enzyme complex provides a protected compartment (the sialisome), presumably reflecting the biosynthesis-export complex (130). Thus, the intracellular lacunae of unexported polymers in kpsCS mutants may reflect the deposition of material whose synthesis cannot be efficiently coordinated with export via the ABC transporter. Clearly, synthesis can occur in the absence of export, but the key question is whether export can occur only in the context of coordinated active synthesis. If so, the sialisome concept provides a structural environment necessary to achieve the required coordination.

KpsDE proteins represent the last steps in the assembly process, as mutations in the relevant genes result in the accumulation of polymers in the periplasm (15, 95, 123). Like the ABC transporter, these gene products are functionally exchangeable between serotypes, indicating that they do not recognize specific structural features of the polysaccharide in their export substrate. KpsD is thought to be the OMA polysaccharide translocon for group 2 capsular polysaccharide, i.e., the counterpart of Wza in group 1 capsule systems. Although no structural information on KpsD is yet available, it shows some local sequence similarity to Wza (92), particularly in areas including the PES motif (30). Wza and KpsD do differ in key respects; KpsD is not a lipoprotein, and KpsD does not form stable multimers larger than a dimer (76).

KpsD requires KpsE for its proper assembly in the outer membrane (5, 76). In the absence of KpsE, KpsD localizes to the periplasm (123). KpsE is a member of the inner membrane periplasmic auxiliary protein 2 (MPA-2) family (92). While the bulk of the protein is periplasmic, KpsE forms oligomers associated with the inner membrane via an N-terminal TMD and a C-terminal amphipathic helix (5, 99, 113). The current working model for group 2 capsule assembly suggests that KpsE plays a critical role in bridging the periplasm as part of a multienzyme complex. Analogies have been made (5) between KpsE and membrane fusion proteins that recruit the TolC efflux channels to ABC transporters and other pumps involved in drug efflux. There are now considerable structure-function data available for some of these other systems. For example, in the AcrA-AcrB-TolC efflux pump, the AcrA adaptor protein is required to stabilize relatively weak interactions between the AcrB pump and TolC (68). AcrB acts like a periplasmic pump in which conformational changes pull material from the inner membrane or periplasm into the TolC channel (121, 122). A flexible hairpin domain in AcrA is implicated in assisting the conformational changes in TolC that are required to open the export channel (68, 81). Similar systems operate in type I protein secretion, in which the recruitment of TolC occurs when the appropriate ABC transporter is charged with the secreted protein substrate (reviewed in reference 62).

Cross-linking approaches suggested that there exists a complex that includes not just the biosynthetic enzymes but also the ABC transporter (KpsMT) and KpsDE (76). Interestingly, KpsDE proteins were found to be located at the poles of cells, although this exclusive localization property is not seen when either protein is expressed in isolation. In contrast, KpsS seems to have a natural tendency to target to the poles. The sites of the initial surface appearance of group 2 capsular polysaccharide can be monitored by microscopy using immunolabeled cells, following a shift from a nonpermissive (20°C) to a permissive (37°C) temperature. The appearance of a new polymer detected by immunofluorescence corresponds to the polar localization of key proteins. However, this result differs from those of electron microscopy analyses in which the newly synthesized polymer appeared around the cell periphery and seemed to coincide to sites where the inner and outer membranes come into apposition (i.e., Bayer junctions) (64). It remains to be determined whether the apparently contradictory observations result from different preparation methods or other fundamental differences. The details of location aside, the evidence for a multienzyme complex (Fig. 7) is now compelling.

Fig. 7Hypothetical complex for the formation of group 2 capsules. Polymer formation is initiated on an unknown endogenous acceptor (potentially diacylglycerophosphate or diacylglycerophosphate-Kdo) and is extended by a glycosyltransferase function(s) that adds residues to the nonreducing terminus of the chain. It is conceivable that the polymer is synthesized on a different acceptor and then transferred to diacylglycerophosphate-Kdo prior to export. The polymer is exported via the ABC transporter (KpsMT). KpsSC proteins are essential for this process, along with other proteins (e.g., KpsFU proteins also have some influence), but the details of their involvements are unknown. The orientation of the polymeric substrate during export (i.e., is the lipid terminus exported first or last?) has not been established, and biosynthesis and export may be temporally coupled. Translocation across the periplasm and outer membrane requires KpsE and KpsD, which provide putative membrane fusion (adaptor) protein and OMA protein functions.

CONCLUSIONS

Substantial progress has been made in understanding capsule biosynthesis and assembly over the past few years. Broad functions have been assigned to most of the dedicated proteins via detailed genetic approaches and some biochemical analyses. However, precise mechanisms for key proteins have yet to be established; this will happen only through combined approaches involving biochemistry and structural biology. The ultimate question is to understand how the dedicated and (often unidentified) housekeeping components come together to form a complex molecular machine that can coordinate rapid polymer biosynthesis and export. Given that the processes are dynamic, a full understanding of the processes will require the capture of the complex in different conformational states, preferably with the polymer in transit, and this goal represents a major challenge.

ACKNOWLEDGMENTS

I recognize generous support from the Canada Research Chairs program and from the Canadian Institutes of Health Research. The contributions of many lab members, past and present, are gratefully acknowledged.