Structural and Functional Variation in Outer Membrane Polysaccharide Export (OPX) Proteins from the Two Major Capsule Assembly Pathways Present in Escherichia coli

Capsules are protective layers of polysaccharides that surround the cell surface of many bacteria, including that of Escherichia coli isolates and Salmonella enterica serovar Typhi. Capsular polysaccharides (CPSs) are often essential for virulence because they facilitate evasion of host immune responses. The attenuation of unencapsulated mutants in animal models and the involvement of protein families with conserved features make the CPS export pathway a novel candidate for therapeutic strategies. However, appropriate “antivirulence” strategies require a fundamental understanding of the underpinning cellular processes. Investigating export proteins that are conserved across different biosynthesis strategies will give important insight into how CPS is transported to the cell surface.

ABSTRACT

Capsular polysaccharides (CPSs) are virulence factors for many important pathogens. In Escherichia coli, CPSs are synthesized via two distinct pathways, but both require proteins from the outer membrane polysaccharide export (OPX) family to complete CPS export from the periplasm to the cell surface. In this study, we compare the properties of the OPX proteins from the prototypical group 1 (Wzy-dependent) and group 2 (ABC transporter-dependent) pathways in E. coli K30 (Wza) and E. coli K2 (KpsD), respectively. In addition, we compare an OPX from Salmonella enterica serovar Typhi (VexA), which shares structural properties with Wza, while operating in an ABC transporter-dependent pathway. These proteins differ in distribution in the cell envelope and formation of stable multimers, but these properties do not align with acylation or the interfacing biosynthetic pathway. In E. coli K2, murein lipoprotein (Lpp) plays a role in peptidoglycan association of KpsD, and loss of this interaction correlates with impaired group 2 capsule production. VexA also depends on Lpp for peptidoglycan association, but CPS production is unaffected in an lpp mutant. In contrast, Wza and group 1 capsule production is unaffected by the absence of Lpp. These results point to complex structure-function relationships between different OPX proteins.

IMPORTANCE Capsules are protective layers of polysaccharides that surround the cell surface of many bacteria, including that of Escherichia coli isolates and Salmonella enterica serovar Typhi. Capsular polysaccharides (CPSs) are often essential for virulence because they facilitate evasion of host immune responses. The attenuation of unencapsulated mutants in animal models and the involvement of protein families with conserved features make the CPS export pathway a novel candidate for therapeutic strategies. However, appropriate “antivirulence” strategies require a fundamental understanding of the underpinning cellular processes. Investigating export proteins that are conserved across different biosynthesis strategies will give important insight into how CPS is transported to the cell surface.

INTRODUCTION

Many important Gram-negative human pathogens assemble complex polysaccharides on their cell surfaces. Capsular polysaccharides (CPSs) represent a prevalent class of surface glycans, and they form an extensive hydrophilic layer, called a capsule, surrounding the cell (1). CPSs can protect against environmental threats such as desiccation, promote biofilm formation, and frequently serve as virulence factors by preventing clearance by the host immune system (2, 3).

CPSs display substantial structural diversity (4), with Escherichia coli alone producing more than 80 structures that give rise to a corresponding number of serologically distinct K (capsular) antigens (1). Despite the structural variations, E. coli CPSs are synthesized via one of two assembly strategies, which are shared with other bacteria (1). In E. coli, these two pathways are represented by group 1 and group 2 capsules. Both strategies initiate CPS biosynthesis at the cytoplasmic face of the inner membrane, using nucleotide-activated sugars as donors to build on a lipid-anchored acceptor. In the prototypical group 1 Wzy-dependent pathway found in E. coli K30, individual CPS repeat units are built on undecaprenol diphosphate before being exported across the inner membrane (IM) by Wzx, a member of the MurJ flippase family (5). Once available in the periplasm, the lipid-linked repeat units are polymerized (by the pathway defining Wzy polymerase) into the full-length polymer (6). In the group 2 ABC-transporter pathway found in E. coli K1, K2, and K5, the entire CPS is synthesized in the cytoplasm on a glycolipid acceptor composed of a reducing terminal phosphatidylglycerol lipid linked to a short oligosaccharide of β-linked 3-deoxy-d-manno-octulosonic acid (Kdo residues), which is extended by addition of the repeat-unit region of the CPS (7). Based on bioinformatics information, this acceptor appears to be conserved in almost all examples of CPS biosynthesis using this pathway (8). The only current exception is found in Salmonella enterica serovar Typhi (S. Typhi) and some members of the Burkholderiales, in which the Kdo-containing glycolipid is replaced by a diacyl N-acetylhexosamine (diacyl-HexNAc) molecule (9). The completed CPS is then exported across the IM by an ABC transporter (10). Regardless of the overall assembly strategy, the terminal translocation of CPS from periplasm to the cell surface requires members of two broad families (11). An outer membrane polysaccharide export (OPX) protein is believed to form a channel across the outer membrane (OM), and a periplasmic adaptor protein, called a polysaccharide copolymerase (PCP) protein, is believed to link the inner membrane transport machinery to the outer membrane OPX protein, in a hypothetical model resembling drug efflux pumps (11). These proteins differ between the group 1 and 2 capsule producers but are conserved within E. coli isolates using each CPS assembly strategy. All available evidence indicates they function independently of the CPS repeat unit structure and, in some cases, that they have been exchanged between species in genetic complementation studies (1, 12).

OPX proteins possess a conserved polysaccharide export sequence (PES) motif, but the representatives from the E. coli group 1 and group 2 CPS export machineries are otherwise quite different (11). The only OPX protein with a solved structure is Wza from the E. coli group 1 systems (13). Wza is an N-terminally acylated lipoprotein that oligomerizes to form an octameric structure associated with the outer membrane (Fig. 1). The octamer generates an outer membrane channel composed of amphipathic α-helices, and three ring-like domains enclose a large lumen extending more than 85 Å into the periplasm, where it interacts with the cognate PCP protein (Wzc) (14). In vivo cross-linking has trapped CPS export intermediates within the channel, demonstrating that CPS does pass through this outer membrane channel (15). The conserved polysaccharide export sequence (PES) motif is found near the N terminus in OPX protein primary sequences and is located in the D1 domain of Wza, the region in closest proximity to the IM and other parts of the assembly machinery (Fig. 1B). While the PES motif is conserved in KpsD (the OPX protein for group 2 capsules), this protein lacks the characteristic sequence required for N-terminal acylation and signal peptidase II cleavage (11). In addition, KpsD has been observed to form heat-stable dimers, but not the higher-order oligomers characterizing Wza (16). KpsD does localize to the outer membrane when coexpressed with its PCP partner (17) and there is evidence for interactions between the cognate pairs, but a considerable amount of KpsD is periplasmic in an E. coli K-12 strain possessing the K5 (group 2) capsule genes on the chromosome (17, 18). KpsD was initially described as a periplasmic protein, but it was expressed in E. coli in the absence of other CPS assembly machinery in the initial report (19). KpsD proteins from E. coli (558 residues) and Campylobacter jejuni (552 residues) are significantly larger than Wza (379 residues) (Fig. 1A) (11). However, OPX proteins from some group 2 systems are more similar to Wza. For example, VexA from Salmonella enterica serovar Typhi and CtrA from Neisseria meningitidis are acylated and possess sizes and predicted secondary structures similar to those of Wza (11). The export substrate for CtrA possesses the same terminal glycolipid as that of E. coli CPSs, while that for VexA possesses diacyl-HexNAc (7, 9).

FIG 1.

Solved and predicted structures of OPX proteins from various CPS biosynthesis pathways. (A) The secondary structures (11) were predicted by Jpred. For Wza, the secondary structure elements were determined from PDB identifier 2J58, as well as predicted from sequence, to illustrate accuracy of the predictions. The region encompassing the polysaccharide export sequence (PES) motif characteristic of OPX proteins is indicated above the secondary structures. α-Helices are indicated by blue rectangles, and β-sheets are indicated by yellow arrows. (B) Wza forms an octameric structure, in which C-terminal amphipathic α-helices generate an outer membrane channel (D4) (at the top of this structure) and three ring-like domains (D1 to D3) form a structure that encloses a large lumen and extends more than 85 Å into the periplasm (13).

The outer membrane lipoprotein Lpp is required for proper surface expression of CPS on the cell surface in E. coli K2 (group 2 capsule), and lpp mutants are susceptible to complement killing and clearance in a mouse bacteremia model (20). Lpp inserts into the periplasmic face of the outer membrane and can be covalently linked to peptidoglycan (PG). It is the most abundant outer membrane protein in E. coli and is essential for maintaining membrane integrity (reviewed in reference 21). The N-terminal acyl groups anchor trimers of Lpp in the outer membrane, while C-terminal lysine residues provide a site for covalent linkage to the stem peptides of PG. Reduced encapsulation of E. coli K2 lpp mutants was correlated with the loss of Lpp-mediated association between KpsD and PG, suggesting that this structural organization is essential for proper functioning of the CPS export machinery in E. coli isolates with group 2 capsules (20). The possible involvement of Lpp in group 1 capsule-assembly systems is unknown.

The apparent structural differences in E. coli OPX representatives raise questions about whether they function in the same way. The underlying hypothesis for this study was that lipoprotein OPXs would share key properties, distinguishing them from nonacylated representatives. Therefore, the objective was a comparative analysis of key properties, including an investigation of whether Lpp plays a universal role in the organization of OPX proteins in the cell envelope, regardless of assembly pathway or details of OPX structure. To that end, the influence of Lpp on CPS assembly was examined for the Wzy-dependent pathway of E. coli K30 (group 1 CPS), and the ABC transporter-dependent pathway of S. Typhi Vi CPS (group 2 variant). The inclusion of S. Typhi allowed comparison of a lipoprotein OPX, resembling group 1 prototypes, with the more distant KpsD homologs, operating within a conserved ABC transporter-dependent export strategy.

RESULTS

Molecular modeling highlights the differences between KpsD and the Wza and VexA lipoprotein OPX homologs.

Wza is an N-terminally acylated lipoprotein that oligomerizes to form an octameric structure that is embedded in the outer membrane (Fig. 1B) (13). VexA from S. Typhi (Fig. 1A) is also acylated and possesses a size and predicted secondary structure similar to those of Wza, despite being involved in an ABC transporter-dependent export pathway (11). This secondary structure is conserved in CtrA from Neisseria meningitidis. The export substrate for CtrA possesses the same terminal glycolipid (a putative export signal) as that of E. coli group 2 CPSs, while the substrate for VexA possesses diacyl-HexNAc (7, 9). The conserved PES motif is located in the periplasmic D1 region of Wza (Fig. 1B). While the PES motif sequence is also present in KpsD (the OPX protein for E. coli group 2 capsules), this protein lacks the characteristic sequence required for N-terminal acylation and signal peptidase II cleavage (11). In addition, KpsD (558 residues) is significantly larger than Wza (379 residues) (Fig. 1A; 8). Its closest homolog is encoded within the CPS assembly locus in Campylobacter jejuni (552 residues) (11).

To further illustrate the differences between Wza, VexA and KpsD, Phyre2 was used to create structural models of the monomers (Fig. 2A). VexA was modelled solely on Wza (PDB identifier 2J58) and, as expected from the secondary structure predictions, its model is very similar to Wza. The most notable difference is the shorter trans-OM amphipathic helix in VexA (18 compared to 32 residues in Wza). The N-terminal domain of KpsD (residues 14 to 324) could be aligned with Wza (Fig. 2B), and the Phyre model predicts 3 domains corresponding to D2, D3, and the PES-containing D1 domain in Wza (Fig. 2A). In contrast, the C-terminal domain of KpsD shares no significant sequence homology with Wza or VexA. It creates an additional domain in KpsD, with more structural complexity than the amphipathic α-helical D4 region of Wza.

FIG 2.

Structural models of group 2 OPX proteins. (A) Phyre2 was used to predict the structures of the OPX protein from S. Typhi VexA and from E. coli K2 KpsD. PES domains are shown in yellow. VexA was modeled entirely on the OPX from E. coli K30, Wza (PDB identifier 2J58), while the KpsD prediction included contributions from Wza and the β-grasp from the E. coli group 4 CPS protein GfcC (PDB identifier 3P42). The N-terminal part of KpsD aligns to Wza (PDB identifier 2J58) (B), and the C-terminal part aligns to GfcC (C). The alignments were made using Phyre2 server, and the figures were created using ESPript 3. The KpsD structural model in panel A was therefore constructed using residues 14 to 324 (numbering includes the leader sequence) from Wza (in blue) and residues 330 to 547 from GfcC (in green). Given the approach to modeling, the orientation of the additional β-grasp domain in KpsD relative to the rest of the protein cannot be confidently assigned. For comparison, the Wza D3 β-grasp is shown as an inset in orange beside the corresponding β-grasp domain in GfcC.

The KpsD C-terminal domain is predicted to contain a soluble ligand binding β-grasp (SLBB) domain. One SLBB clade is found in several proteins associated with polysaccharide export (including domains D2 and D3 of Wza), and they are proposed to interact with the polysaccharide (22). Although the C-terminal domain of KpsD shares no significant primary sequence similarity with any region of Wza, it does share similarity with another β-grasp-containing protein, GfcC (PDB identifier 3P42); 15% identity; 30% similarity (Fig. 2C). The corresponding region of KpsD was modeled on a GfcC template (Fig. 2A). GfcC has been implicated in the export of group 4 CPS in E. coli; (23). Group 4 capsules employ a Wzy-dependent polymerization process (like that of group 1 capsules) to create a glycan that is normally attached to lipopolysaccharide (LPS) as an O-antigenic polysaccharide. However, the gene products of a separate locus (including gfcC) facilitate the diversion the polymerization machinery to create and export the O-antigen glycan as an LPS-independent CPS, in a process involving components similar to the group 1 machinery (1). The precise role of GfcC in export is unclear; the gfc locus also encodes a Wza paralog (GfcE) (23), and group 1 CPS export in E. coli has no dependence on GfcC (unpublished data). GfcC is a member of the DUF1017 protein family and actually contains two β-grasp domains. These folds typically comprise a central mixed β-sheet with 4 or 5 strands and an α-helix on one face; in GfcC, one β-grasp domain diverges from the typical structure (24). The regions comprising the β-grasp domains of GfcC and Wza (D2 and D3) share <18% sequence identity, and this is reflected in differences in the organizational details of the folds (24). GfcC was purified as a soluble periplasmic protein and, in the crystal structure, the C-terminal amphipathic α-helix folded down onto the body of the structure. Whether this conformation changes if the protein is associated with the OM is unclear. However, it was noted that the helix is shorter (13 residues in the solved structure) than that in Wza (32 residues) and may be insufficient to fully span the OM. This same concern applies to the predicted KpsD amphipathic α-helix (17 residues).

Lpp has no major effect on the outer membrane localization of Wza or VexA.

Given the differences in predicted structures between KpsD and Wza/VexA, we performed a comparative investigation of their properties and the possible impact of Lpp defects. Constructing the lpp deletion in E. coli E69 (serotype K30) to generate CWG1380 was straightforward due to the single copy of the gene. In contrast, S. Typhi is complicated by possessing two adjacent copies of lpp genes; the lppA gene product is the homolog corresponding to the E. coli Lpp protein. Although transcriptomics data indicate that lppB is not expressed in Salmonella enterica serovar Typhimurium under typical laboratory growth conditions (25), and lpp-deficiency has no reported effect on envelope integrity in S. Typhimurium, both genes are required for wild-type virulence characteristics (26, 27). Since the expression situation in S. enterica Typhi is uncertain, lppA and lppB were both deleted in S. Typhi to generate CWG1381. IM and OM were separated using a sucrose gradient, and OPX content was assessed by Western immunoblotting (Fig. 3). In these experiments, YidC and OmpA were also probed as validated markers of IM and OM, respectively. YidC is a component of the protein export machinery (28), and OmpA is an outer membrane β-barrel protein (29).

FIG 3.

Membrane distribution of OPX proteins and the response to an lpp defect. Shown are Western immunoblots of IM and OM fractions from sucrose gradients, probed with the antibodies indicated. (A) Wza, (B) VexA, and (C) KpsD. Wza is associated predominantly with the OM fraction but, in contrast, both KpsD and VexA are distributed between the IM and OM fractions. Wza and KpsD exist predominantly as multimers in the OM, while a large pool of VexA monomer is found in the IM. There are no major shifts in profiles of any of the OPX proteins in lpp deletion backgrounds. OmpA and YidC serve as markers for OM and IM, and Lpp was examined to confirm the mutation. The aberrant migration (and “waviness”) of some bands corresponds to situations where the proteins comigrate with LPS molecules (data not shown).

As expected from previous studies, Wza was detected predominantly in the OM fraction, where it exists as multimers when samples are solubilized in SDS-PAGE buffer at 25°C (Fig. 3A), and the multimers dissociated into monomers when samples were heated at 100°C (30, 31). Only trace amounts of Wza were observed in the IM fraction. This material is also multimeric at the lower solubilization temperature, suggesting the presence of small amounts of OM contaminating the IM fraction, rather than reflecting protein translocation intermediates, which would be monomeric (Fig. 3A). Consistent with this conclusion, trace amounts of OmpA and Lpp (in the parent strains) were also detected in the IM fraction. No substantial changes were observed in the overall distribution (between the IM and OM fractions) of Wza and the OmpA and YidC marker proteins in E. coli E69 and its corresponding Δlpp:cat mutant. The only difference was a reproducible slight increase in the small amount of Wza and OmpA in the IM fraction from the mutant, presumably reflecting minor differences in OM and IM separation.

KpsD was initially described as a periplasmic protein, but in that study, it was expressed in E. coli K-12 in the absence of other CPS-assembly machinery (19). A subsequent investigation revealed that KpsD was distributed between the IM and OM membrane when coexpressed with its PCP partner (KpsE) (17). To confirm that the same is true in a wild-type strain, under the same conditions used to examine Wza, KpsD distribution was examined in E. coli CFT073 (a serotype K2 isolate [20]). Previously, we reported that the amount of KpsD in the outer membrane was unaffected in an lpp mutant background (20), but the IM/OM comparison was not performed. KpsD was found in the IM and OM fractions, but the multimeric form was confined to the OM fraction (Fig. 3C). KpsD was also detected in the IM fraction, consistent with published observations (17), but the amount was small relative to OM and consisted only of monomers. It has been reported that KpsD forms SDS-stable dimers (16), but it is unclear whether these reflect a final format or a more stable substructure of something larger. The oligomers seen in Fig. 3C migrate with an apparent molecular mass of 140 to 180 kDa on an SDS-PAGE gel, and the differences (compared to the reported dimers of 120 kDa [16]) could be due to either different experimental conditions or variations between homologs from different strain backgrounds. The distribution of KpsD was unaltered by the ΔlppAB::cat mutation.

To determine whether the differences between Wza and KpsD distributions correlated with acylation status or were a feature associated with the overall export machinery, VexA was examined (Fig. 3B). As indicated, VexA resembles Wza in terms of structure but is part of an export system similar to the one involving KpsD. VexA was distributed between IM and OM fractions, like KpsD. Only small amounts of SDS-stable VexA oligomers were evident, and these were confined to the OM; most of the VexA migrated as a monomer. The apparent molecular mass of the immunoreactive material (140 to 180 kDa) was smaller than that of the Wza octamers, and they were completely dissociated at 100°C. Given the apparent limited stability of VexA multimers in the OM, there is no way to know whether the monomers detected in the same fraction were originally multimeric, but the profiles suggest a substantial pool of monomeric intermediates in the IM. The various control proteins were distributed as expected, so the VexA results were not due to any gross effects on IM and OM separation. The distribution of VexA was unaltered by the ΔlppAB::cat mutation.

Taken together, the OPX proteins from the E. coli group 1 and 2 capsule prototypes both form multimers, independent of the export strategy or acylation status. However, the presence/absence of a substantial IM OPX pool does not correlate with either of those features.

Absence of a substantial periplasmic pool of acylated OPX representatives.

A major periplasmic pool of monomeric KpsD was reported previously (17, 19). This was confirmed in E. coli CFT073 using a standard osmotic shock procedure (Fig. 4C). In contrast, there was no substantial amount of either Wza or VexA in the periplasm in the wild-type strains (Fig. 4AB). Under the conditions used, maltose-binding protein (MalE; an authentic periplasmic protein [32]) was fully released, but Lpp was confined to the spheroplasts, as expected. RNA polymerase was also included as a control to confirm that the cytoplasmic membrane remained intact. In the corresponding Δlpp::cat mutant, there was a small but reproducible increase in the amount of Wza in the periplasmic fraction, but the profiles for KpsD and VexA were indistinguishable in the parent and mutant strains. This result differed from published work with an E. coli K-12 hybrid strain, in which the periplasmic fraction of KpsD was absent in the corresponding lpp mutant (17). Collectively, these results suggest that acylated OPX proteins lack a periplasmic form and that Lpp is not an essential determinant of this feature in these wild-type strains.

FIG 4.

Effect of Lpp on periplasmic OPX content. Shown are Western immunoblots of (A) Wza, (B) VexA, and (C) KpsD. Cells were subjected to osmotic shock, and the periplasmic (P) and spheroplast (S) fractions were separated by centrifugation. Periplasmic maltose-binding protein (MalE; MBP) served as a control for release. Cytoplasmic RNA polymerase (RNAPol) served as a control for spheroplast lysis. Loading was calculated to show the protein derived from a set number of cells, reflected in the whole-cell lysate (L). Deletion of lpp only affected distribution in the case of Wza.

Lpp impacts the PG association of OPX proteins associated with ABC transporter-dependent export strategies.

Previous work indicated that loss of PG association of the KpsD OPX protein and reduced group 2 capsule production correlated with absence of PG-associated Lpp (20). The extent of Lpp-dependent PG association of Wza and VexA was examined in the same way, using the SDS-soluble and SDS-insoluble fractions from detergent-solubilized membranes. The SDS-insoluble fraction contains PG and any associated proteins, including Lpp (33). Wza was equally distributed between the SDS-soluble and SDS-insoluble fractions, and the distribution was unchanged in the absence of Lpp (Fig. 5A). In contrast, PG-associated VexA was severely diminished in the absence of Lpp, and this was largely reversed by complementing the ΔlppAB::cat mutation with a plasmid carrying lppAB (Fig. 5B). The dependence on Lpp for PG association is therefore reminiscent of E. coli KpsD, and distinct from that of Wza.

FIG 5.

Lpp differentially affects the PG association of OPX proteins. Membranes containing (A) Wza and (B) VexA were solubilized in SDS and separated by centrifugation into SDS-soluble (membrane) and SDS-insoluble (PG-associated) fractions, which were then analyzed by Western blotting. Lpp, which has a form covalently linked to PG, served as the control. Only the VexA profile showed any requirement for Lpp.

Influence of Lpp on Wza- and VexA-dependent capsule assembly.

In E. coli CFT073 with a classical group 2 capsule-assembly system, mutation of lpp leads to a substantial reduction in the amount of CPS produced and impairs virulence (20). In contrast, the Western immunoblot profiles of CPS in E. coli E69 was indistinguishable with or without Lpp (Fig. 6A). The overall amount of Vi CPS was similar in S. Typhi wild type and mutant, but there was a modest change in the chain length distribution; the mutant lost some larger material, which was restored by complementation (Fig. 6A). This method reports total CPS production but cannot distinguish between CPS on the cell surface and material accumulating in the cell due to an export defect. However, no obvious differences were found in the quantity of CPS antigen measured by whole-cell enzyme-linked immunosorbent assay (ELISA), which directly measures surface-associated glycan (Fig. 6B).

FIG 6.

lpp deletion had no effect on Wza- and VexA-dependent CPS assembly. (A) Total CPS in whole-cell lysates was analyzed by Western blotting. (B) Cell surface CPS was quantified by whole-cell ELISA, and CPS-deficient mutants were included as controls for background signal.

DISCUSSION

OPX proteins participate in the final steps in the export of CPSs, facilitating their passage across the OM, but the details of their function may vary depending on the overall export strategy. The most detailed information is available for the E. coli group 1 assembly pathway. Wza engages the cognate PCP protein (Wzc), but it is still unknown how these proteins interface with Wzx, which exports undecaprenol diphosphate-linked repeat units across the IM, or with the periplasmic domain of the CPS polymerase (Wzy) (1, 14). CPS has been detected in transit within the Wza lumen, validating the export conduit to the surface (15). In contrast, KpsD and VexA participate in export complexes containing an ABC transporter and an adaptor protein (KpsE and VexB, respectively) (1, 8). Until recently, the working model for CPS export in an ABC transporter-based system invoked a protected continuous conduit from the cytoplasm to the cell surface to translocate complete CPS molecules. This proposal has been challenged by the discovery that a periplasmic glycanase can degrade CPS export intermediates in S. enterica serovar Typhi (34). The step(s) involved in the transit of periplasmically accessible intermediates between the ABC transporter and the OPX translocon therefore requires further investigation.

The starting hypothesis for this study was that the lipoprotein status and predicted structural similarity of Wza and VexA would confer shared properties distinct from those of the nonacylated KpsD. This turned out to be incorrect, and the structure-function relationships are more complicated. Wza exists primarily as a remarkably stable OM octamer. Oligomerization is crucial for the function of Wza because it creates an enclosed lumen and OM channel composed of amphipathic α-helices large enough (17-Å diameter) to accommodate the glycan substrate (13). KpsD also appears to be mostly oligomeric but possibly in a form larger than that of the reported dimers (16). Preliminary evidence is provided here for oligomerization of VexA, but the authentic in vivo oligomeric states of OPX proteins other than Wza cannot be concluded from the currently available data. The shared (predicted) secondary structures of Wza and VexA lead to a VexA model structure that closely resembles that of Wza; this includes the C-terminal amphipathic helices (D4) that create the OM channel in Wza (13). The implications of the reduced size of the predicted VexA helix (18 versus 32 residues in Wza) is currently unknown. In integral IM proteins, a length of 20 to 30 residues is considered sufficient for an α-helix to span the membrane (35), but there are few OM-spanning α-helices available to generate a precise range for the OM; Wza provided the first example (13). The model of KpsD suggests an additional mixed α/β β-grasp domain, sitting on top of a structure resembling Wza D1 to D3. This raises questions about its capacity for channel formation. The difficulty in assigning that function to the C-terminal α-helix has been discussed elsewhere in the context of GfcC, but the possible membrane association of that protein has not been examined. Although it has been reported that the C terminus of KpsD (detected via a C-terminal hexahistidine epitope) is not surface exposed, fluorescence-activated cell sorter (FACS) data with anti-KpsD antibodies did demonstrate exposure of KpsD epitopes at the cell surface (16). It is likely that the true organization of KpsD will only be resolved by a structure for the protein in a lipid environment.

Only the nonacylated KpsD has the presence of a substantial periplasmic pool of the protein. This could be due to more stable insertion into the OM or to differences in the translocation methods for these proteins. In Gram-negative bacteria, the Lol pathway is responsible for trafficking lipoproteins, such as Wza, VexA, and Lpp, to the OM (36). The Lol machinery consists of an IM complex (LolCDE) that extracts lipoproteins from the IM, a periplasmic chaperone (LolA) that delivers the targeted lipoprotein to the OM, and an OM component (LolB) that completes the insertion of the lipoprotein into the OM. Lol-dependent trafficking of Wza and VexA may preclude a significant periplasmic pool for these OPX proteins. A portion of the pool of all 3 proteins is associated with PG. The CPS export machinery must pass through the PG layer without compromising the essential integrity of the cell wall. The complexity of these systems and the importance of maintaining the PG barrier has recently been highlighted in the context of LPS translocation, where conditional depletion of the translocation machinery immediately activates a PG-editing enzyme complex (37). The nature of the interaction between OPX proteins and PG is unknown, but the noncovalent interaction of the β-barrel OM protein OmpA with PG has been studied in detail and residues have been identified that are important for PG association in OmpA and other PG-associated proteins (29, 38). In wild-type E. coli, OmpA monomers and dimers interact with PG, but only interactions with dimers are retained in the absence of Lpp (39). Molecular dynamics simulations predict that Lpp lifts the PG to position it closer to the OM to facilitate the interaction (39). The flexible C-terminal domain of OmpA contains a conserved aspartate and arginine pair that have been shown to interact with the diaminopimelate (DAP) of PG (29), but there is no significant sequence similarity shared by OPX proteins and the OmpA PG-binding regions, so the residues (and their location) required for binding are unknown. Interestingly, the association of Wza with PG is unaffected by the absence of Lpp; although the lpp mutant does show a reproducible slight increase in the amount of periplasmic Wza, the basis for this is uncertain. In contrast, both KpsD and VexA are severely affected by lpp defects (20). Although this feature correlates with the involvement of an ABC transporter in the export pathway, Lpp-mediated PG association of KpsD and VexA does not affect CPS assembly in the same way. The absence of Lpp only leads to a substantial decrease in assembled capsule in the KpsD-dependent system and this results in a severe impact on survival of the bacterium in the host (20). In contrast, there is no impact on CPS assembly and export in E. coli K30 and only a subtle change in CPS size distribution in S. Typhi. This lack of known envelope integrity defects in S. Typhimurium lacking Lpp (26), and the wild-type CPS phenotypes in E. coli K30 and S. Typhi, support the proposal that the CPS phenotype in E. coli CFT073 ΔlppAB::cat is not an indirect consequence of a general envelope perturbations (40). It is conceivable that the acylation of Wza and VexA in S. enterica overcomes the critical importance of Lpp and PG association in KpsD-dependent systems.

The results presented here reveal complex structural relationships and cell envelope interactions between OPX proteins associated with the two main capsule assembly strategies. They highlight the different approaches Gram-negative bacteria have evolved to deal with the challenge of translocating high-molecular-weight glycans that are critical for survival in the host without compromising the barrier properties of the OM. The different properties identified here can only be addressed by further structural biology investigations, but the systems described here provide the necessary prototypes to address these questions.

MATERIALS AND METHODS

Bacterial strains, mutant construction, and growth conditions.

The wild-type bacterial strains used in this study were E. coli E69 serotype O9a:K30:H12 (group 1 capsule prototype; supplied by F. Ørskov [41]) and S. enterica serovar Typhi Ty2 trp, cys, ΔaroC1019 (Vi-antigen producer; K. E. Sanderson, Salmonella Genetic Stock Centre, University of Calgary, Canada). Mutants lacking surface CPS expression were used as controls: E69 wza_{22 min}::aadA wza_K30::aacC1 (CWG281 [15]), and S. Typhi ΔvexC (CWG1235 [9]). lpp deletion mutants in E69 (CWG1380) and S. Typhi (CWG1381) were constructed using the λ-Red recombineering system (42); the lpp genes were replaced by the chloramphenicol resistance gene, cat. Primers used for constructing the mutants are listed in Table 1 and were obtained from Sigma-Aldrich. pKD3 was used as a template for the nonpolar cat cassettes (42). pSIM6, the temperature-sensitive λ-Red expression vector, was utilized for recombination (43). Mutants were confirmed by PCR amplification of the region surrounding the gene and sequencing of the PCR products (AAC Genomics Facility, University of Guelph). The plasmid containing S. Typhi lppAB was constructed by cloning these genes in tandem from the S. Typhi genome into the pBAD24 vector. Cultures were grown at 37°C in LB medium unless otherwise stated. 2,3-Dihydroxybenzoic acid (100 μg/ml) and chloramphenicol (15 μg/ml) were added as appropriate. The S. Typhi lppAB mutant exhibited an extended lag phase when cells were subcultured from stationary culture, but this was not seen when cells were subcultured from a culture in the log phase. Once in the exponential phase, indistinguishable growth rates were observed for the parent and for the mutant cultured in either condition. All samples were taken from mid-exponential-phase cultures.

TABLE 1.

Primer sequences for construction of lpp mutations^a

Primer	Template	Sequence	Description
CS-P2F	E69/pKD3	AATACTTGTAACGCTACATGGAGATTAACTCAATCTAGAGGGTATTAATAgtgtaggctggagctgcttc	λ-Red recombination primer for lpp deletion in E. coli E69
CS-P2R	E69/pKD3	ACAAAAAAAATGGCGCACAATGTGCGCCATTTTTCACTTCACAGGTACTAcatatgaatatcctccttag	λ-Red recombination primer for lpp deletion in E. coli E69
CS-P9F	S. Typhi/pKD3	CGGACAAAAAAATGGCGCACGATGTGCGCCATTTTATATCATGCGTCAAAgtgtaggctggagctgcttc	λ-Red recombination primer for lppBA deletion in S. Typhi
CS-P9R	S. Typhi/pKD3	AATACTTGTAACGCTACATGGAGATTAACTCAATCTAGAGGGTATTAATAcatatgaatatcctccttag	λ-Red recombination primer for lppBA deletion in S. Typhi

^aUppercase nucleotides indicate homology to the genomic sequence, while lowercase nucleotides indicate homology to the plasmid sequence.

Release of periplasmic contents.

Periplasmic contents were released as described elsewhere (34). Briefly, cells were harvested by centrifugation (5,000 × g for 10 min) at late-log phase and 10 optical density at 600 nm (OD₆₀₀) unit equivalents were resuspended in 0.7 ml 100 mM Tris-HCl (pH 8.2) containing 500 mM sucrose. After incubation on ice for 5 min, lysozyme was added to 100 μg/ml, and EDTA (pH 8) was added to 1 mM. The cell suspension was incubated for another 20 min on ice, and MgSO₄ was added to 20 mM. The resulting spheroplasts were collected by centrifugation and resuspended in the same volume of water, and the supernatant containing the periplasmic fraction was removed. Spheroplast and periplasmic extract samples were combined with SDS-PAGE sample buffer, heated at 100°C, and examined by SDS-PAGE and Western immunoblotting. To ensure full release of the periplasm, the known periplasmic protein maltose binding protein (MalE; MBP) was used as a control. The cytoplasmic protein RNA polymerase was used as a control for potential spheroplast lysis. For E. coli, cultures were grown in 0.2% maltose to induce expression, and for S. Typhi, the pMAL-p2 plasmid (NEB) was used for MalE expression.

Membrane separation by sucrose gradient centrifugation.

Inner and outer membranes were separated by sucrose density centrifugation, as described previously (44). Cells were harvested by centrifugation at late-log phase and resuspended in 20 ml 5 mM EDTA (pH 8). Cells were then lysed using a French press, and the cell debris were removed by centrifugation. Membranes were collected by centrifugation at 100,000 × g for an hour, resuspended in 3 ml 5 mM EDTA (pH 8), and layered onto a 2-step gradient of 0.3 ml 65% sucrose and 1 ml 25% sucrose in 5 mM EDTA. The gradients were centrifuged at 117,000 × g in a Beckman MLS-50 swinging bucket rotor for 3 h to remove any residual soluble proteins, and the bottom 0.7 ml, containing the membranes, was collected. This membrane fraction was mixed with 1 ml 5 mM EDTA and loaded onto a discontinuous sucrose step gradient composed of 0.5 ml 65% sucrose, 1 ml 55% sucrose, and 2 ml each of 50%, 45%, 40%, and 35% sucrose in 5 mM EDTA. Gradients were centrifuged at 221,000 × g in a Beckman SW-41 swinging bucket rotor for 16 h, and 0.4 ml fractions were collected. Samples were combined with SDS-PAGE sample buffer and examined by SDS-PAGE and Western immunoblotting. Antibodies against YidC and OmpA were used as validated markers for IM and OM, respectively. Fractions containing IM or OM were pooled and combined with SDS-PAGE sample buffer at either 100°C or 25°C, and OPX localization was analyzed by SDS-PAGE and Western immunoblotting.

Examination of PG-associated OPX proteins.

PG-associated proteins were examined as described previously (20). Cells were harvested by centrifugation at mid-log phase, resuspended in 20 ml 10 mM Tris-HCl (pH 8), and lysed using a French press, prior to centrifugation to remove unbroken cells and large cell debris. Membranes were collected from the resulting cell-free lysate by centrifugation at 100,000 × g for an hour and resuspended in 6 ml 10 mM Tris-HCl (pH 8), containing 2% SDS and cOmplete mini protease inhibitor cocktail. The samples were then centrifuged again at 100,000 × g for an hour, and the supernatant containing solubilized cell envelope proteins (i.e., the SDS-soluble fraction) was removed. The pellet (i.e., the SDS-insoluble fraction), containing PG and any associated proteins, was washed once in water and resuspended in 0. 3 ml 50 mM Tris-HCl (pH 8) containing 0.4 M NaCl, 5 mM EDTA, and cOmplete mini protease inhibitor cocktail. Samples of the SDS-soluble and SDS-insoluble fractions were combined with SDS-PAGE sample buffer and examined by SDS-PAGE and Western immunoblotting.

Analysis of proteins by SDS-PAGE and immunoblotting.

Protein samples were separated by SDS-PAGE with 12% resolving gels and transferred to membranes using the Thermo Scientific Pierce Electrophoretic Blotting system. Nitrocellulose membranes were blocked in Tris-buffered saline with Tween 20 (TBS-T) containing 5% skim milk for 2 h and then incubated overnight at 4°C with one of the following primary antibodies: mouse anti-MBP (1:5,000; NEB), mouse anti-RNA polymerase (1:1,000; Santa Cruz Biotechnology), rabbit anti-OmpA (1:50,000; Antibody Research Corp), rabbit anti-YidC (1:20,000) (20), rabbit anti-Lpp (1:20,000) (20), rabbit anti-Wza (1:1,000) (30), or rabbit anti-VexA (1:3,000). The anti-VexA antibodies were produced by immunization of New Zealand White rabbits with intramuscular injections of purified His₆-VexA mixed with Freund’s incomplete adjuvant (Animal Care Services, University of Guelph). Following incubation with primary antibodies, membranes were washed with TBS-T and incubated with alkaline phosphatase (AP)-conjugated goat anti-rabbit antibody (1:5,000; Cedarlane Laboratories) or AP-conjugated goat anti-mouse antibody (1:3,000; Jackson ImmunoResearch) for 2 h. After another wash step, blots were developed using nitroblue tetrazolium and 5-bromo-4-chloro-3′indolyphosphate p-toluidine salt substrate.

Detection and quantitation of CPS.

CPS profiles were examined in whole-cell lysates. Overnight cultures were diluted to an OD₆₀₀ of 0.02 and grown to the mid-log phase. One OD unit equivalent of cells was collected by centrifugation and resuspended in 70% ethanol. Following a 20 min incubation, cells were resuspended in 100% ethanol and incubated another 20 min. The supernatant was removed and the pellets were air dried. Pellets were then suspended in 10 μg/ml each DNase I and RNase A, 100 μg/ml lysozyme, and 10 mM MgCl₂, and incubated 1 h at room temperature. Samples were then mixed with equal volumes of SDS-PAGE sample buffer and incubated at 37°C for a minimum of 2 h before separation by SDS-PAGE (with 10% resolving gels), followed by Western immunoblotting. Samples were transferred using the Thermo Scientific Pierce electrophoretic blotting system to a Biodyne B nylon membrane. The membranes were blocked in TBS-T containing 5% skim milk for 2 h and then incubated overnight at 4°C with one of the following primary antibodies: rabbit anti-K30 (1:1,000) (45) or rabbit anti-Vi antigen (1:250; BD). Membranes were washed with TBS-T and incubated with AP-conjugated goat anti-rabbit antibody (1:5,000) for 2 h. After another wash step, blots were developed using nitroblue tetrazolium and 5-bromo-4-chloro-3′indolyphosphate p-toluidine salt substrate.

CPS quantitation was performed using a whole-cell ELISA as described previously (20). Cells were harvested at the mid-log phase by centrifugation, resuspended in phosphate-buffered saline (PBS) at 0.2 OD/ml, and incubated on a Nunc Maxisor 96-well plate at 4°C overnight. Wells were incubated with blocking buffer (5% bovine serum albumin [BSA]-PBS) for 1 h at 25°C, followed by rabbit anti-K30 (1:100) (45) or rabbit anti-Vi (1:100; BD) antibodies for 2 h at 25°C. After three washes with PBS, the wells were incubated with AP-conjugated goat anti-rabbit antibody (1:2,000) for 1 h at 25°C. The wells were washed again with PBS, and a detection solution containing p-nitrophenyl phosphate was added and incubated for 30 min before reading the absorbance at 406 nm. Control tests were performed to ensure that readings were taken within a linear range and saturated. A control reaction mixture containing no primary antibody was used to subtract background signal from the sample reads. Capsule-null mutants were included as negative controls.