Microbial exopolysaccharides, defined here as cell-adherent capsules or excreted, unattached or loosely attached slime, are used as additives in a wide range of food and cosmetic products, medical replacements in ocular and joint surgeries, and are essential components of microbial biofilms (Sutherland, 1998; 2001). Some exopolysaccharides are also important virulence determinants in bacterial or fungal invasive disease where, as capsules surrounding the microbial cell surfaces, they function as ligands for cell adhesion or modulators in the avoidance or inhibition of innate and acquired immunity (Vimr et al., 2004). Capsular polysaccharides conjugated to protein carriers conferring T-cell dependency are among the most effective anti-capsule vaccines in current medical use: Streptococcus spp., group b Haemophilus influenzae, and all groups except group B Neisseria meningitidis, with more anti-capsule vaccines in development (Robbins and Schneerson, 2004). Supporting these applied uses of exopolysaccharides are basic studies to understand the molecular details of capsule biosynthesis, which we define here as the processes of intracellular polysaccharide synthesis and export (secretion or translocation) to the external environment. In many cases, the linkage of polysaccharides to the outer membrane of Gram-negative bacteria or cytoplasmic membrane of Gram-positive organisms is unknown (Whitfield, 2006).

Escherichia coli is a model organism for investigating bacterial exopolysaccharide function and biosynthesis. Different types of E. coli capsules have been subdivided into one of four biosynthetic groups (Whitfield, 2006). After synthesis of the required nucleotide sugar(s) unique to a given polymeric structure, functionally diverse glycosyltransferases catalyse synthesis of homo- or heteropolysaccharides that are exported to the cell surface by ATP-binding cassette-like (ABC) transporters or Traffic ATPases, in the case of groups 2 and 3 capsules, or by Wzy-dependent machineries required for groups 1 and 4 export. In the case of groups 2 and (probably) 3 polysaccharides, some sugar chains are anchored to the cell membrane by esterification to phospholipid, further distinguishing them from the lipid A-linked somatic or O antigens of lipopolysaccharide. Mechanistic distinctions are also based on the genetic regulation and map positions of the capsule gene clusters. We suggest a fifth group of polysaccharides in which the glycosyltransferase (polymerase) simultaneously functions as the translocase. This system, where the polymerase is also the cytoplasmic membrane exporter, has been inferred for the biosynthesis of Gram-positive and -negative polyglucosamines and hyaluronans (Wang et al., 2004; Fluckiger et al., 2005; Tlapak-Simmons et al., 2005). Despite our understanding of groups 1 and 4 Wzy-dependent export systems, transport of group 2 capsular polysaccharides has been a more difficult process to study (Whitfield, 2006). This difficulty stems in part from problems in reconstituting the translocation process after subcellular fractionation (Vimr and Steenbergen, 1993).

Attempts to reconstitute polysialic acid (PSA) synthesis and export using inverted (inside-out) membrane vesicles and an in vitro protection assay indicated that only a relatively small and temporally constant fraction of the total PSA synthesized by this system was resistant to endo-N digestion, implying a failure of the system to export polysialic acid (Vimr and Steenbergen, 1993). Endo-N is a K1-specific bacteriophage-derived depolymerase that cleaves α2,8-linked sialic acid residues of PSA chains ≥ 7 monomers in length (Petter and Vimr, 1993). Our in vitro results suggested that some factor or physical property of the system was lost during subcellular fractionation which prevented efficient export of PSA despite continual polymer synthesis in the presence of exogenously added sugar nucleotide donor cytidine 5′-monophospho N-acetylneuraminic acid (CMP-Neu5Ac) and ATP, and thus continual sensitivity to endo-N, a homotrimer composed of 76 kDa subunits. Although the ‘flippase’ activity of Wzx found in groups 1 and 4 systems has been demonstrated in vitro (Rick et al., 2003), we are unaware of any successful reconstitution of group 2 polysaccharide export. We concluded from our in vitro protection studies that PSA biosynthesis involves essential protein–protein, protein–polysaccharide or protein–lipid interactions that are disrupted during fractionation. In the current study, we describe an in vivo investigation of PSA biosynthesis that provides the first direct evidence for a novel subcellular compartment, which we propose to designate the sialisome. Similarly privileged cytoplasmic compartments may exist in other group 2 polysaccharide biosynthetic systems.

Fig. 1 Genetic organization of the group 2 capsular polysialic acid gene cluster in E. coli K1. Biosynthetic genes (not drawn to scale) are shown as open boxes with transcriptional directions indicated by the horizontal arrows above. The chromosomal positions (more ...)

Fig. 2 Three models of group 2 polysaccharide biosynthesis. Known or suspected functions of kps and neu gene products are defined in the legend to Fig. 1 and text. Gold ribbons represent PSA with red squares indicating phospholipids esterified at the reducing (more ...)

Model 2 indicates a two-step process in which synthesis occurs independently of export, and where phospholipid modification of the reducing terminus may trigger activation of or signal PSA entry into the export apparatus. This hypothesis is consistent with the accumulation of full-length polysaccharides in group 2 mutants with region 1 or 3 export defects, potentially explaining how a conserved export apparatus transports structurally distinct polysaccharides by recognizing a common chemical modification or secondary structural feature (reviewed in Vimr et al., 2004). Because the products of the region 1 kpsC and kpsS genes are homologous with certain enzymes of lipid metabolism, and because mutations in these genes were reported to cause accumulation of intracellular group 2 polysaccharides lacking the reducing terminal phospholipid modification (Bronner et al., 1993; Frosch and Muller, 1993), model 2 seemed to be supported. However, the experimental observations supporting model 2 are inconsistent with the phenotypes of E. coli K1 kpsC or kpsS mutants, which accumulate lipidated PSA, suggesting that lipidation is insufficient for group 2 capsule export (Cieslewicz and Vimr, 1996). The explanation for the experimental discrepancies is unclear, but suggests that neither KpsS nor KpsC is required for lipid addition, calling model 2 into question. This conclusion is supported by a recent report that group B meningococcal kpsS (lipB or ctrC) and kpsC (lipA or ctrD) mutants also accumulate lipidated, intracellular PSA (Tzeng et al., 2005). The reason why previous studies did not detect polysaccharide lipidation in kpsC or kpsS mutants in the E. coli K5 or meningococcal systems may have been failure to control for the heterologous or non-stoichiometric expression systems used in these studies, or relatively low sensitivity of the detection methods. In contrast, studies detecting lipidation of PSA in E. coli K1 and group B meningococcal kpsS and kpsC mutants were carried out with strains containing single-copy capsule gene clusters and enzymatic or structural methods that unambiguously detected phospholipid attachment.

Model 3 is a hybrid of models 1 and 2 and predicts that while polysaccharide synthesis occurs in the absence of transport, only those polymers synthesized in the presence of an intact biosynthetic complex are exported. In the extreme case, assembly of the functional translocase would occur only when triggered by nascent group 2 polysaccharide. Model 3 suggests intimate contact between synthetic and export components of group 2 biosynthetic complexes, consistent with the reduced polymerase activities observed for all regions 1, 3 and neuE mutants (reviewed in Vimr et al., 2004). This phenotype indicates that polymerase catalytic efficiency is affected by the operation or structure of the export apparatus, which is qualitatively similar to the feedback mechanism inferred for group 1 exopolysaccharides.

Fig. 3 PSA biosynthesis is protected from endo-N in vivo. Approximately 75 μl of K1-specific phage K1F (1 × 1010 pfu ml−1 phage stock solution) was streaked (arrow) down the middle of the plate with a sterile wooden stick, and bacteria (more ...)

As shown in Fig. 4B, EV36 had the same efficiency of K1F plating as its uninduced (Fig. 4C) or induced (Fig. 4D) derivatives harbouring pRep4 and pEndo-N. As expected, acapsular mutant EV93 (kpsC) did not support phage infection (Fig. 4A) because it lacks surface capsular polysaccharide, although it accumulates intracellular, unexported PSA in centrally located compartments termed lacunae to indicate their electron transparency (Cieslewicz and Vimr, 1996). Although plaque size differed under the three infection conditions (Fig. 4B–D), the results indicated that capsule was produced independently of the amount of intracellular endo-N, from none (Fig. 4B) to approximately 2% of the intracellular protein concentration of induced cells harbouring pRep4 and pEndo-N (Fig. 4D). We suggest that recombinant endo-N released by phage lysis accounts for the morphological plaque differences by altering cell-surface capsule or PSA released by disrupted cells. However, note that plaque size in the induced strain (Fig. 4D) was at least twice that of wild type (Fig. 4B) despite a reduced host growth rate caused by enzyme overproduction, indicating that even a massive accumulation of intracellular endo-N did not inhibit capsule biosynthesis. This is in contrast to in vitro results in which endo-N was shown to effectively degrade nascent PSA (Steenbergen et al., 1992; Vimr and Steenbergen, 1993).

Fig. 4 Effects of recombinant endo-N overproduction on K1-specific phage sensitivity. Bacteria were grown to early stationary phase with appropriate drugs and IPTG added where indicated to a final concentration of 1 mM to induce endo-N production when cells (more ...)

To demonstrate that both basal and induced endo-N productions by EV36 transformed with pRep4 and pEndo-N were sufficient for PSA degradation, we carried out radial immunodiffusion against anti-PSA antibody recognizing PSAs ≥ 8–10 sialic acid residues. If endo-N is present, PSA is degraded to oligomers < 8 residues in length that do not react with antibody. As shown in Fig. 5, extracts of both wild type (well 1) and mutant strain EV93 (well 2) contained PSA, as evident by the precipitin halos surrounding the respective wells. Note that strain EV93 PSA is derived from unexported, intracellular PSA whereas the wild-type antigen comes from capsular polysaccharide released during cell disruption (Cieslewicz and Vimr, 1996). In contrast, PSA from EV36 was degraded in the presence of basal or induced amounts of endo-N, resulting in the absence of precipitin halos (Fig. 5, wells 3 and 4 respectively). To directly demonstrate the susceptibility of intracellular PSA to endo-N, extracts of EV93 were mixed 1:1 with those of uninduced (Fig. 5, well 5) or induced (Fig. 5, well 6) extracts from EV36 harbouring pRep4 and pEndo-N. The absence of PSA antigen under either condition indicates that the expression of even a basal amount of endo-N is sufficient to depolymerize intracellular PSA.

Fig. 5 Production of endo-N in uninduced and induced EV36 harbouring pRep4 and pEndo-N. Bacteria (20 ml) were grown to early stationary phase with appropriate drugs and pelleted by centrifugation. The pellets were washed in phosphate-buffered saline (PBS), re-suspended (more ...)

An alternative argument for the results shown in Figs 3 and 4 is that endo-N is inactive when expressed intracellularly. To show that endo-N is active in vivo, we investigated PSA stability by transmission electron microscopy (TEM) of intact and thin-sectioned cells. We had previously shown by TEM that E. coli K1 export mutants accumulate lipidated PSA in intracellular lacunae unbounded by a membrane and displaying morphologies unique to a particular export defect (Cieslewicz and Vimr, 1996; 1997), indicating that E. coli normally lacks depolymerase(s) for turnover of the PSA that accumulates in these export mutants. In the case of strain EV93, the unexported PSA tends to accumulate in centrally located lacunae evident in thin sections (Cieslewicz and Vimr, 1996), a result that was confirmed independently for this study (Fig. 6D). The PSA in these accretion bodies bear terminal phospholipids that we assume result in the intracellular micelles formed by otherwise highly soluble and electrostatically charged polysaccharides. Lacunae thus represent intracellular organelles resulting from synthesis of full-length, lipidated PSA that either does not enter the translocation pathway or is aborted during biosynthesis. When whole mounts of negatively stained EV93 were examined by TEM, lacunae were apparent as bubble-like organelles in the intact cells (Fig. 6A and B). In contrast, lacunae were absent in EV93 transformed with pEndo-N with or without IPTG induction (Fig. 6C), indicating that basal endo-N expression was sufficient to degrade intracellular PSA. Similarly, lacunae were not detected when EV93 harbouring pEndo-N was examined by TEM of thin-sectioned samples (Fig. 6E). Taken together, the results of the intracellular endo-N protection assay demonstrate in vivo depolymerase activity against intracellular PSA, supporting the conclusion that wild-type PSA synthesis and export occur within an environment inaccessible to endo-N. Note that this operational definition of the sialisome does not preclude the concept of a kinetic compartment, in which the synthesis of PSA is so rapid that endo-N, even at induced concentrations, does not have time to degrade nascent PSA. However, because even one intra-chain clip would be expected to disrupt biosynthesis, we think it is more likely that the sialisome represents the functional biosynthetic complex that has been previously shown to involve multiple protein–protein interactions (McNulty et al., 2006). More important, our results tend to exclude model 2 by favouring the directed coupling of PSA biosynthesis to its export shown in model 3 (Fig. 2).

Fig. 6 Susceptibility of intracellular PSA to endo-N digestion. A and B. TEM of PSA lacunae (arrows) in E. coli K1 mutant EV93 whole mounts. C. TEM of strain EV93 transformed with pEndo-N. D. TEM of thin-sectioned strain EV93. E. TEM of thin-sectioned strain (more ...)

Fig. 7 Complementation of neuE and neuS mutants. Strains EV725 (neuE) and EV136 (neuS) were transformed with the indicated plasmids and sensitivity to phage K1F was assayed by cross-streak and plaque assays. S (sensitive) and R (resistant) indicate relative (more ...)

Fig. 8 Overproduction of NeuS and membrane anchor function of the NeuE PIRS domain. A. Maxi-cell analysis of neuS expression in strain HB101 harbouring the indicated plasmids. Lane 1, pUC18; lane 2, pSX90; lane 3, pSX91; lane 4, pSX92. Asterisk indicates the (more ...)

NeuE has been shown to include a putative C-terminal polyisoprenoid recognition signal (PIRS) that could play a role in polyprenol binding during PSA synthesis or translocation (Fig. 8B) (Steenbergen et al., 1992; Zhou & Troy, 2003, 2005). However, NeuS activity is not eliminated in a neuE null mutant, indicating that PIRS is not obligatory for polymerization (Vimr et al., 2004), a conclusion supported by the analyses shown in Figs 7 and 8A. To determine if the NeuE PIRS domain functions as a membrane anchor, and thus a possible structural role consistent with the export defect of a neuE mutant (Vimr et al., 2004), we carried out localization experiments with maltose binding protein (MBP) fused to the domain.

Depending on the programme used, neuE is predicted to encode a polypeptide with one to five membrane-spanning regions (Table 1). Each of the nine programmes identified the putative C-terminal PIRS domain while four of them uniquely identified PIRS as the only membrane-spanning region (Table 1). To confirm that PIRS functions as membrane anchor, we fused the BglII-BamHI fragment from pSX92 to the MBP polylinker in plasmid pMal-c2, generating plasmid pSX101 expressing an N-terminal MBP lacking its leader peptide sequence fused to the PIRS domain and the last 25 amino acid residues of NeuE. As shown in Fig. 8C (lanes 4 and 5), an overproduced polypeptide of the expected molecular weight was localized to the membrane fraction after differential centrifugation and denaturing polyacrylamide gel electrophoresis, indicating that PIRS anchored a soluble polypeptide to what we presume is the inner membrane. No overproduced polypeptide was detected in the soluble fraction (Fig. 8C, lanes 6–9), indicating efficient localization of the putative fusion to the membrane. However, the PIRS anchor is unlikely to span the membrane because nine of the 25 amino acid residues following it to the C-terminus are positively charged and therefore unlikely to cross into the periplasm. On the basis of secondary structural predictions (MacVector 9.0.2), we further suggest that the PIRS domain is unlikely to be an alpha helix, and more likely to be an amphipathic beta sheet reminiscent of the proposal for a similar anchoring domain in hyaluronan synthase (Heldermon et al., 2001). Thus, while the exact function of NeuE is still unclear, our results indicate that it is not obligatory for PSA synthesis, although necessary for export. We have not yet tested whether deletion of this domain affects the export function.

Table 1 Computer-assisted analysis of NeuE TM regions.

Controls with anti-MBP confirmed that the overproduced fusion polypeptide included the expected MBP epitope(s). Thus, the pMal-c2 vector alone synthesized the expected 53 kDa polypeptide (MBP-β-gal-α), whereas fusion of an almost full-length copy of neuE (pSX104) synthesized the expected 65 kDa polypeptide resulting from truncation of the lacZα peptide of the vector by cloning the EcoRI-HindIII fragment from pSX90 into the pMal-c2 polylinker (Fig. 8D, lanes 3 and 1 respectively). The polyclonal anti-MBP antibody also detects lower-molecular-weight bands probably arising from proteolysis, including prominent bands near the 43 kDa size of native MBP, and a band common to strains independent of IPTG induction (Fig. 8D, lanes 1–5), representing an E. coli antigen recognized by the polyclonal antibody preparation. Induction of the strain expressing the malE-BglII-BamHI fragment from pSX90 (pSX101), which codes for the PIRS domain and last 25 amino acid residues of NeuE, stably produced the expected 45 kDa fusion polypeptide (Fig. 8D, lane 2) that was observed in the membrane fraction of induced cells (Fig. 8C, lanes 4 and 5). The results of this Western blot analysis confirm the identity of the overproduced membrane polypeptide as MBP fused to the NeuE PIRS domain. We conclude that this domain functions as NeuE membrane anchor, as predicted by the results in Table 1.

To develop an in vivo approach for investigating PSA synthesis and export, we used a two-hybrid bacterial system for the detection of both homo- and heterotypic protein–protein interactions. Daines and Silver (2000) previously used a Lex-based system to identify an interaction between NeuD and NeuB, implicating NeuD as a stabilizer or activator of the synthase (Fig. 2), a function recently supported by protein engineering of the homologous group B streptococcal NeuD (Lewis et al., 2006). This conclusion is consistent with the function of NeuD as a monomeric sialic acid transacetylase (Steenbergen et al., 2006), in which case direct interaction with NeuB might ensure efficient acetylation during the synthase-catalysed condensation of ManNAc with PEP to produce sialic acid. As an alternative to the Lex system, which requires the use of special strains to prevent possibly conflicting homodimeric interactions, we tested the applicability of the Bordetella pertussis adenylate cyclase system (Karimova et al., 1998; 2000). This system depends on reconstitution of an active adenylate cyclase and subsequent transcriptional amplification of cAMP-dependent promoters, such as lac or mal expressed in an E. coli cya host. Because the signalling cascade is intracellular, we did not expect to observe interactions between extracellular components of the PSA biosynthetic machinery, as supported by the lack of interactions between KpsD and all prey tested (Table S1). In contrast, both homo- and heterotypic interactions were detected between a variety of other kps and soluble or membrane-bound neu gene products.

Figure S1 shows a schematic diagram for one representative experiment testing interactions between a KpsC bait and a variety of prey plasmids expressed in BTH101, with some prey fusions tested in either of two orientations relative to the B. pertussis cya fusion partners. Note that both homo- and heterotypic interactions were detected, suggesting that KpsC has multiple binding domains. Table 2 shows the data for the positively and selected negatively interacting pairs in the EV727 and EV729 backgrounds. Note that NeuS and KpsC interacted strongly, but only when NeuS was the bait (Table 2). This orientation dependency is likely to reflect steric constraints imposed by the B. pertussis Cya domains, T25 or T18 (Karimova et al., 1998; 2000). Note that all plasmid fusions were sequenced to verify that translation would be in frame. Although we have not confirmed the positive interactions with antibody pull-down experiments, an immunological study in the E. coli K5 system indicated that KpsC-K5 was required for membrane localization of the heparosan (K5 antigen) synthase (Rigg et al., 1998). However, previous results established that NeuS does not require any other kps or neu gene product for membrane association (Steenbergen et al., 1992), suggesting an alternative function for KpsC in group 2 capsule biosynthesis. Surprisingly, we could not detect an interaction between NeuS and NeuE (Table 2) despite the report that such an interaction might be necessary for polymerization (Andreishcheva and Vann, 2006). However, other interactions between KpsC and itself or KpsE were detected (Table 2), suggesting that KpsC plays a role in connecting PSA export to the polymerase, NeuS, adding support to model 3 (Fig. 2).

Table 2 Two-hybrid analysis of selected pair constructs in EV727 and EV729, capsule-negative and capsule-positive, respectively, E. coli cya derivatives.

Vann and colleagues previously detected an in vitro requirement for KpsC in NeuS-catalysed polymerization (Andreishcheva and Vann, 2006). Furthermore, the estimated size of the radiation-sensitive polymerase complex was 78 kDa larger than the predicted molecular weight of the NeuS monomer (approximately 48 kDa), supporting interaction(s) between NeuS and one or more molecular species totalling less than 48 kDa (Vionnet et al., 2006). Note that NeuS did not interact with itself (Table S1), suggesting that the polymerase might function as a monomer as part of a larger complex. Also note that assaying the polymerase after irradiation only detects the size of the complex necessary for polymerization in vitro, indicating that the functional size of the biosynthetic complex (synthesis and export) could be much larger than suggested by enzyme assay. When taken together, the results of our two-hybrid analysis strongly support a direct interaction between KpsC and NeuS. Many other potentially interacting pairs of kps or neu gene products tested in the two-hybrid system yielded negative results (Table S1), suggesting no or only weak interactions that may have been below the limits of detection. We reasoned that potentially weak interactions might be distinguishable from true negatives if a more sensitive reporter assay was employed.

Fig. 9 Interacting C-terminal bait deletions of KpsC with full-length KpsC or KpsE prey. The indicated control (pKT25 and pSX750) and deletion bait plasmids were coexpressed with KpsC (pSX753) or KpsE (pSX758) prey plasmids. Beta-galactosidase (Beta-gal) activities (more ...)

Roberts and colleagues (Rigg et al., 1998) pointed out the apparent kpsC gene duplication event that gave rise to homologous N- (Domain A, residues 60–270) and C- (Domain B, residues 387–598) terminal domains with 39.2% similarity over 210 amino acid residues and 58% overall nucleotide identity in the E. coli K5 biosynthetic system. Our results indicate that Domain A functions homotypically while Domain B, at least in part, is necessary for heterotypic interaction with KpsE. Interestingly, KpsC-K5 was released by osmotic shock, leading to the speculation that this presumably soluble protein is part of an adhesion complex connecting polysaccharide synthesis and export, potentially involving a KpsE domain (Rigg et al., 1998). Although chemical cross-linking and co-purification with histidine-tagged KpsS-K5 identified apparent interactions between it and KpsM, KpsT, KpsE and KpsD, KpsS-K5 also interacted with other polypeptides whose relevance to capsule biosynthesis is unclear (McNulty et al., 2006), suggesting that chemical cross-linking might produce artefactual results. In contrast, we could not detect interactions between KpsS-K1 and any bait–prey combination tested (Table S1), further suggesting that chemical cross-linking might yield artificial associations or that relevant KpsS interactions are not detectable with the current two-hybrid system. Note that most of our results were obtained with full-length kps and neu gene fusions. However, the orientation effects on NeuS–KpsC interactions, for example, indicate probable steric constraints interfering with some fusions (Table 2). Therefore, we plan to test shorter fusions of negatively interacting pairs as well as those with demonstrated interactions to more precisely define binding domains and, ultimately, identify specific interacting peptides that should define the complete network of homo- and heterotypic interactions in group 2 capsule systems. This approach should lead to a better understanding of capsule biosynthesis while helping to identify new therapeutic targets in a wide range of encapsulated pathogens expressing group 2 or group 2-like capsules.

The most important finding of this study is that PSA biosynthesis (synthesis and export) occurs within an intracellular compartment operationally defined as inaccessible to endo-N, a specific depolymerase previously used to investigate PSA in both bacterial and mammalian systems (Vimr et al., 1984). Our approach provides a novel in vivo protection assay whereby PSA was detected by K1-specific phage infectivity, electron microscopy and immunodiffusion assessing the processes of translocation, intracellular accumulation of unexported PSA and polymerization respectively. Although we do not know the minimum number of PSA chains necessary for productive lytic phage infection, the number must be no less than 5–10% of the normally present 50 000 chains, as below this minimum number only small plaques are produced because of diminished phage binding (Vimr, 1992; Cieslewicz and Vimr, 1997). Because plaque size was not diminished by endo-N overproduction (Fig. 4), we conclude that more than 5000 PSA chains are synthesized and exported to the outer surface of each cell during capsule biosynthesis in the presence of intracellular endo-N. This conclusion was supported by identical precipitin halos surrounding wild type or cells expressing endo-N (data not shown), which is a reflection of the equivalent amounts of capsular polysaccharide being synthesized and exported during cell growth (Steenbergen and Vimr, 1990). Our previous characterization of the export defects in mutants with regions 1 or 3 defects facilitated detection of endo-N activity in vivo. Indeed, our data indicate that there could be two sialisomes: one endo-N-accessible formed by intracellular accumulation of unexported PSA, and the endo-N-inaccessible compartment representing the authentic biosynthetic environment. These results strongly point to model 3 as a dynamic representation of group 2 polysaccharide biosynthesis. (Fig. 2)

We have carried out the first extensive two-hybrid analysis of capsule biosynthesis in any species. The results confirm and extend previous biochemical approaches involving protein cross-linking and immunological detection (Arrecubieta et al., 2001; McNulty et al., 2006), subcellular fractionation and irradiation (Andreishcheva and Vann, 2006; Vionnet et al., 2006), and complementation analyses (Fig. 7). For example, when the biochemical results are combined with our analysis of KpsC interactions, it is clear that KpsC functions in several homo- and heterotypic interactions important to capsule synthesis and export. The N-terminal region necessary for KpsC homotypic interaction has been identified, while a separate C-terminal KpsC domain was shown to mediate heterotypic interaction with KpsE. In contrast, there was no evidence for NeuS oligomerization, although this polypeptide was shown to interact with both KpsC and KpsE, supporting the concept of a dynamic biosynthetic apparatus that couples polymerization to export, in further support of model 3 (Fig. 2). This conclusion implies that the signalling event(s) involved in directed coupling of PSA synthesis and KpsC linking the polymerase to the periplasmic connector, KpsE, could mediate export. Experiments are in progress mapping the NeuS domain that interacts with KpsC. Our results suggest that all groups 2 and 3 polymerases will share similar interactive domains. Finally, despite our inability to detect an interaction between the polymerase and NeuE, suppression of a neuE defect by NeuS overproduction supports the conclusion that NeuE is not obligatory for capsule biosynthesis, although it is likely to function in some as yet undefined structural role that is necessary, under the normal single-copy circumstance, for PSA export.

Research was supported by NIH Grant AI042015. We thank Richard P. Silver, Daniel Ladant and Mark D. Gonzalez for providing bacterial strains, plasmids and helpful discussions. We are grateful to Lou Ann Miller for electron microscopy and Kerry Helms for photographic assistance.

This material is available as part of the online article from: http://www.blackwell-synergy.com/doi/abs/10.1111/j.1365-2958.2008.06231.x (This link will take you to the article abstract).

Please note: Blackwell Publishing is not responsible for the content or functionality of any supplementary materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.