Investigation of core machinery for biosynthesis of Vi antigen capsular polysaccharides in Gram-negative bacteria

Abstract

Salmonella enterica serovar Typhi causes typhoid fever. It possesses a Vi antigen capsular polysaccharide coat that is important for virulence and is the basis of a current glycoconjugate vaccine. Vi antigen is also produced by environmental Bordetella isolates, while mammal-adapted Bordetella species (such as Bordetella bronchiseptica) produce a capsule of undetermined structure that cross-reacts with antibodies recognizing Vi antigen. The Vi antigen backbone is composed of poly-α-(1→4)-linked N-acetylgalactosaminuronic acid, modified with O-acetyl residues that are necessary for vaccine efficacy. Despite its biological and biotechnological importance, some central aspects of Vi antigen production are poorly understood. Here we demonstrate that TviE and TviD, two proteins encoded in the viaB (Vi antigen production) locus, interact and are the Vi antigen polymerase and O-acetyltransferase, respectively. Structural modeling and site-directed mutagenesis reveal that TviE is a GT4-family glycosyltransferase. While TviD has no identifiable homologs beyond Vi antigen systems in other bacteria, structural modeling suggests that it belongs to the large SGNH hydrolase family, which contains other O-acetyltransferases. Although TviD possesses an atypical catalytic triad, its O-acetyltransferase function was verified by antibody reactivity and ¹³C NMR data for tviD-mutant polysaccharide. The B. bronchiseptica genetic locus predicts a mode of synthesis distinct from classical S. enterica Vi antigen production, but which still involves TviD and TviE homologs that are both active in a reconstituted S. Typhi system. These findings provide new insight into Vi antigen production and foundational information for the glycoengineering of Vi antigen production in heterologous bacteria.

Many bacteria are surrounded by a hydrated capsule layer composed of high-molecular-weight capsular polysaccharide (CPS) (1). In many pathogens, capsules are crucial virulence determinants; they can maintain a hydrated surface layer, aid in biofilm formation, and protect against complement-mediated killing and opsonophagocytosis (1, 2, 3). CPS structures are highly variable in their sugar composition, linkages, and nonsugar modifications. Despite this diversity, the assembly of capsular layers in Gram-negative bacteria is predominantly accomplished via one of two strategies (1).

In one assembly strategy, CPS molecules are synthesized at the cytoplasm–cytoplasmic membrane interface and are then exported to the cell surface using a dedicated trans-envelope system. The export machinery consists of an ATP-binding cassette (ABC) transporter, a (predominantly) periplasmic protein belonging to polysaccharide copolymerase family 3 (PCP-3), and an outer membrane polysaccharide export (OPX) translocon. Escherichia coli “group 2” capsules provide the prototype for this pathway (1). Sequence data (sometimes supported by mutant phenotypes) suggest that this translocation strategy is shared by several human pathogens, including extraintestinal pathogenic E. coli, Salmonella enterica serovar Typhi, Bordetella pertussis, Haemophilus influenzae, Campylobacter jejuni, and Neisseria meningitidis. Most of these bacteria are believed to generate CPS with a characteristic glycolipid at the reducing terminus, based on possession of genes encoding two CMP-3-deoxy-d-manno-oct-2-ulosonic acid (CMP-Kdo)-dependent glycosyltransferases (GTs), KpsS and KpsC. The functions of KpsS and KpsC have been characterized biochemically and structurally in E. coli (4, 5, 6, 7). They build a short β-Kdo oligosaccharide on phosphatidylglycerol acceptor, and this molecule serves as a shared acceptor for CPS structure-specific GTs to polymerize repeat unit polymers by growth at the nonreducing terminus (5, 6, 8). The terminal glycolipid is thought to provide a signal recognized by the export machinery to facilitate translocation of the completed product to the cell surface (1).

Currently, one variation of this prototype is known. This is the Vi antigen CPS produced by typhoidal isolates of S. enterica serovar Typhi, Citrobacter freundii, and some members of the Burkholderiales such as the soil bacteria Bordetella petrii and members of the genera Achromobacter and Orrella (9, 10, 11, 12). These bacteria possess homologs of the standard export-translocation machinery but are unable to synthesize the phosphatidyl-β-Kdo oligosaccharide acceptor due to the lack of the kpsSC genes. Although the structure and function of Vi antigen have not been studied in soil Burkholderiales, its properties and importance in the ability of S. Typhi to cause typhoid fever are well established. S. Typhi Vi antigen CPS is composed of a linear polymer of α-(1→4)-linked N-acetylgalactosaminuronic acid (GalNAcA) that is nonstoichiometrically O-acetylated at C-3 (13, 14). Vi antigen production differentiates S. Typhi from nontyphoidal Salmonellae that cause gastroenteritis. The importance of Vi antigen in pathogenicity was demonstrated in a murine model of infection, where the LD₅₀ increased approximately 10,000-fold for S. Typhi variants lacking Vi antigen (15, 16). Vi antigen participates in the evasion of the host immune system by reducing deposition of complement component 3b on the bacterial surface and clearance of S. Typhi by macrophages (17, 18). Purified Vi antigen is currently used in some vaccine formulations for typhoid fever, with the O-acetyl groups being essential for protection (19, 20, 21).

The ten-gene viaB locus in S. Typhi encodes proteins required for transcriptional regulation of Vi production and biosynthesis (tvi genes) and translocation (vex genes) (9, 22, 23) (Fig. 1, A and B). The minimal set of genes required to produce Vi antigen is tviBCDE; transposon insertions in any of these genes are known to abolish Vi antigen production (22, 23). Production of the Vi-cytoplasmic sugar nucleotide precursor, UDP-GalNAcA, is catalyzed by TviBC (24) (Fig. 1B). Vi antigen possesses a diacyl-HexNAc reducing terminus whose structure has not been definitively resolved. VexE is a dedicated acyltransferase associated with proper acylation and is essential for efficient export and for retention of the CPS glycan at the cell surface (9). However, the enzymes responsible for Vi-polymerization and acetylation have not been clearly established. Resolving their functions is central to the long-term goal of unravelling the full assembly-export pathway for an influential CPS-assembly prototype and may enable new strategies for vaccine production by in vivo glycoengineering or by in vitro chemi-enzymatic methods. Here, we demonstrate that TviE is the sole GT required (and sufficient) for polymerization and TviD is the Vi antigen O-acetyltransferase.

Figure 1.

Vi antigen biosynthesis.A, genetic organization of the viaB locus from S. Typhi. B, model of Vi antigen assembly and export. In the cytoplasm, TviBC synthesizes UDP-GalNAcA, the nucleotide sugar precursor. The UDP-GlcNAc C-6 dehydrogenase (TviB) oxidizes UDP-GlcNAc to UDP-GlcNAcA, and this is converted to UDP-GalNAcA by the UDP-GlcNAcA epimerase (TviC). TviDE is hypothesized to be required for polymerization and O-acetylation, while VexE is a dedicated acyltransferase that modifies a HexNAc residue at the reducing terminus; the enzymatic origin of the HexNAc and details of the terminal structure are currently unknown. Once completed, the Vi polysaccharide is exported by an ABC transporter (VexBC), a polysaccharide copolymerase family-3 protein (VexD), and an outer membrane export protein (VexA). VexE-mediated modification is essential for Vi retention at the cell surface. C, Western immunoblots demonstrating differential reactivities of anti-Vi-OAc monoclonal antibody and anti-Vi polyclonal antibody, which recognizes the glycan backbone independent of O-acetyl groups. The samples are purified WT Vi and chemically de-O-acetylated Vi (left) as reactivity controls and proteinase K-digested whole-cell lysates of E. coli Top10 transformants harboring the entire viaB cluster or tviBCDE showing products made in recombinant strains with and without Vi export, respectively (right). The immunoblot is representative of three biological replicates.

Results

TviBCDE is sufficient for biosynthesis of O-acetylated Vi antigen in an E. coli host

TviBCDE is sufficient for production of intracellular Vi antigen in recombinant E. coli (22, 23), but the O-acetylation status of this product from these proteins was not investigated. This was examined using differential antibody reactivities. Two antigenic determinants have been described for Vi antigen: the O-acetyl present at C-3 and the N-acetyl and carboxyl moieties associated with the glycan backbone (21, 25). Monoclonal antibody P2B1G2/A9 recognizes an epitope dependent on the O-acetyl groups (25), and its reported specificity was confirmed by the loss of reactivity in Western immunoblots of purified Vi antigen, following chemical de-O-acetylation (Fig. 1C). Here, this antibody is designated “anti-Vi-OAc.” In contrast, the reactivity of a rabbit polyclonal antibody (“anti-Vi”) was unaffected by the removal of O-acetyl groups. In Western blots of proteinase K-digested whole-cell lysates, both antibodies react with a disperse range of Vi antigen chain lengths produced by E. coli Top10 harboring plasmids containing either the complete viaB locus (plasmid pWQ783) or tviBCDE alone (pWQ1045) (Fig. 1C). As expected, E. coli Top10 cells transformed with plasmid containing viaB produced a capsular layer reactive with both antibodies, which was revealed by immunofluorescence microscopy of fixed cells (Fig. S1). In contrast, E. coli Top10 cells containing tviBCDE accumulated Vi antigen in intracellular inclusion bodies, which have been previously reported in mutants with compromised export (9). Reactivity with both antibodies was retained, ruling out the possible involvement of periplasmic O-acetylation known in some glycoconjugate-assembly systems (26) (Fig. S1). These data establish that the necessary GT and O-acetyltransferase activities for Vi antigen production are encoded within the tviBCDE gene block and stimulated further investigation of TviD and TviE.

TviE is a UDP-GalNAcA-dependent glycosyltransferase

TviE is the only identifiable GT encoded by the viaB locus and, while similarity shared between TviE and GTs has been reported (22), the details have not been investigated. The C-terminus of TviE contains a putative GT domain belonging to the GT4 family in the CAZy database (27) and is annotated as such by the Conserved Domain Database (CDD) (28) (Fig. 2A). Well-studied examples of GT4 enzymes include: WaaG from E. coli (29), PimA from Mycobacterium smegmatis (30), WbdA from E. coli (31), and PglH from C. jejuni (32). These representatives possess a GT-B fold, act via a retaining glycosylation mechanism, and utilize nucleoside diphosphate (NDP)-activated sugars as donor substrates (33). To provide further insight into the putative TviE GT4 domain, and the organization of the enzyme, the sequence of TviE was submitted to Phyre2 (34). The top five models were all created using members of the GT4 family as templates. These templates (in ranked order) are: sucrose synthase-1 from Arabidopsis thaliana (AtSus1: PDB 3S29), sucrose synthase from Nitrosomonas europaea (Ss2: PDB 4RBN), sucrose phosphate synthase from Halothermothrix orenii (SpsA: PDB 2R60), d-inositol-3-phosphate GT from Corynebacterium glutamicum (MshA: PDB 3C4V), and α-mannosyltransferase from C. glutamicum (PimB: 3OKA). The sequences of the C-terminal GT4 modules align well (Fig. S2), and Phyre2 (34) predicts with 100% confidence that TviE and each of the templates are homologous. In addition to the identifiable GT4 modules, AtSus1, Ss2, and TviE all possess additional N-terminal sequences that vary in size and share little similarity.

Figure 2.

Activity and predicted structure of TviE.A, predicted domain organization of TviE from the CDD (28). B, TviE structural model generated by Phyre2 (34) using sucrose synthase-1 from A. thaliana (AtSus1: PDB 3S29) as a template. The active site with the EX₇E motif is highlighted with the Glu residues depicted as red sticks. C, sequence and secondary structure alignment of the EX₇E-containing region of TviE and AtSus1. The predicted secondary structure of TviE and the solved structure of AtSus1 are displayed above and below the alignment, respectively. Predicted α-helices and β-strands are depicted by helices and arrows, respectively. Proposed catalytic Glu residues of the EX₇E motif are indicated with a red triangle. The alignment was generated with ClustalW (73) and displayed using ESPript (74). A complete alignment is shown in Fig S3. D, Western immunoblots of proteinase K-digested whole-cell lysates from E. coli BL21(DE3) recombinants showing the absence of Vi antigen production in constructs with tviE deleted. The mutation is complemented using plasmids expressing tviE in trans, and activity is eliminated by mutating Glu residues in the EX₇E motif. Expression of TviE and its mutants was confirmed by probing protein lysates with anti-His₅ antibody. The control (viaB) is E. coli BL21(DE3) transformed with plasmid pWQ783 containing the complete viaB locus. All other samples were from E. coli BL21(DE3) transformants. Western immunoblots are representative of three biological replicates.

The highest ranked model for TviE is based on the structure of AtSus1, where the GT4 modules (TviE, residues 163–562; AtSus1, residues 277–776) share only 25% identity (Fig. S3). Despite the low primary sequence homology, the entire sequence of TviE was modeled and predicts a GT domain located between residues 163–562. This region of the structure predicts two Rossmann folds, which is characteristic of GT-B-fold enzymes, like those of the GT4 family (27) (Fig. 2B). GT4 enzymes possess active site EX₇E motifs, which are involved in binding nucleotide sugar donors (Fig. 2, B and C). In biochemically characterized representatives, the first Glu residue of the EX₇E motif interacts with the sugar moiety of the NDP-sugar donor, while the 2′ and 3′ hydroxyl groups of the ribose sugar form hydrogen bonds with the second Glu (30, 32, 35). The relative importance of either of the Glu residues varies in different GT4 enzymes (30, 36, 37, 38, 39). In some, both residues are essential for activity but, in others, one residue is expendable. The functional importance of these residues was investigated in TviE.

Vi antigen was expressed in E. coli BL21(DE3) from a plasmid containing the viaB locus (pWQ783) and reacted with both anti-Vi-OAc and anti-Vi antibodies (Fig. 2D). Deletion of tviE from plasmid-encoded viaB (viaB-ΔtviE; pWQ1042) eliminated reactivity of the corresponding proteinase K-digested whole-cell lysate with either antibody (Fig. 2D), consistent with the previously described phenotype of a tviE transposon-insertion mutant (22, 23). Overexpression of His₆-TviE restored Vi antigen production in E. coli BL21(DE3) expressing viaB-ΔtviE. To evaluate the importance of the EX₇E motif for the activity of TviE, His₆-TviE^E483A and His₆-TviE^E491A mutants were made. Neither was able to support Vi antigen synthesis, indicating both Glu residues of the EX₇E motif are essential in TviE (Fig. 2D). Whole-cell lysates were probed with anti-His₅ antibody to verify both mutant proteins were expressed (Fig. 2D) and subcellular localization determined that the majority of each protein localized to the soluble fraction (S100; 100,000g) (Fig. S4A). Furthermore, the melting temperatures of the mutant proteins in thermal shift assays fell within 2 °C of wild-type His₆-TviE (Fig. S4B), ruling out global changes in stability as an explanation for loss of activity of the mutant protein.

The N-terminal 276 residues of AtSus1 form a predominantly α-helical structural element required for cellular targeting. This region also binds plant host-derived peptides, which are involved in root nodule formation in legumes. This part of the structure is positioned away from the GT4 domain (35). Residues 1–36 of TviE align with residues 118–167 of AtSus1, which correspond to the end of the cellular targeting domain region (residues 11–127) and the first part of the peptide-binding region (residues 157–276) (Fig. S3) and could only be modeled with low confidence. The remaining sequence of TviE preceding the GT4 module (residues 37–162) was modeled with high confidence and is therefore predicted to be predominantly α-helical, like AtSus1 (35) (Figs. 2B and S3). Since the N-terminal part of TviE possesses no informative motifs, its functional importance in Vi antigen synthesis was investigated by removing residues 1–162. Removal of the entire N-terminal domain (His₆-TviE^163–578) rendered the protein inactive based on its inability to complement viaB-ΔtviE defect in E. coli BL21(DE3) and restore antibody reactivity in proteinase K-digested whole-cell lysates (Fig. S5C). However, cellular fractionation showed that the His₆-TviE^163−578 protein was predominantly found as inclusion bodies and within cellular debris (P12; 12,000g), with a small amount associated with the membrane fraction (P100; 100,000g) (Fig. S5D). A longer truncation mutant (His₆-TviE^50–578) removing only two short α-helices, with the endpoint located in a loop, was made in an attempt to minimally disrupt secondary structure elements. His₆-TviE^50–578 was present in the soluble fraction (S100) and the thermal shift profile did not deviate from the wild-type protein (Fig. S5, D and E). However, His₆-TviE^50–578 was unable to restore Vi antigen synthesis (Fig. S5C), suggesting that the N-terminal α-helical structure is essential for function in Vi antigen synthesis.

TviD is a novel O-acetyltransferase

Functions have been assigned to all of the viaB genes, except for tviD, making tviD a candidate for the gene encoding the missing Vi antigen O-acetyltransferase. A transposon insertion in tviD abolishes Vi antigen production in E. coli harboring the mutated viaB locus (23); however, detection of Vi antigen in this previous study relied solely on the anti-Vi-OAc antibody. Thus, it was unclear whether the phenotype reflected loss of polymer production or loss of the O-acetyl epitope. To distinguish these possibilities, the tviD gene was deleted from the tviABCDE-containing plasmid (creating pWQ1046; tviABCE). Proteinase K-digested whole-cell lysates from E. coli Top10 cells harboring tviABCE showed a reactivity profile (Fig. 3A) resembling chemically de-O-acetylated Vi antigen (Fig. 1C). These data are consistent with a model in which TviD is the O-acetyltransferase, and its activity is not required for polymerization of the Vi antigen backbone. Overexpression of His₆-TviD (pWQ1052) restored production of O-acetylated Vi antigen in E. coli Top10 containing tviABCE, thus attributing the Vi antigen phenotype solely to the tviD mutation (Fig. 3A).

Figure 3.

TviD is required for Vi antigen O-acetylation.A, Western immunoblots of proteinase K-digested lysates from E. coli Top10 harboring a plasmid encoding TviABCDE produces intracellular (see Fig. S1 - TviBCDE) Vi antigen reacting with both antibodies. Deletion of tviD (TviABCE) eliminates reactivity with the anti-Vi-OAc monoclonal antibody only, and this is restored by introducing a plasmid expressing His₆-TviD in trans. Western immunoblots are representative of three biological replicates. B, Parts of ¹H,¹³C HSQC spectra of purified Vi antigen from the samples shown in panel A, demonstrating correlations of methyl groups of N- and O-acetyl groups.

To definitively confirm that the absence of tviD only affected O-acetyl content, Vi antigen was purified from E. coli Top10 cells expressing TviABCDE, TviABCE, and TviABCE[His₆-TviD]. The ¹³C NMR spectra of Vi antigens from cells expressing TviABCDE and TviABCE[His₆-TviD] were comparable to published data (9) (Fig. 3B). In contrast, the ¹³C NMR spectrum from cells expressing only TviABCE lacked the diagnostic peak (21.2 ppm) corresponding to O-acetyl groups (Fig. 3B), like chemically de-O-acetylated Vi antigen (9). This data provided the foundation to further explore the activity of TviD.

TviD is separated into two domains. The N-terminal part of the protein (∼320 residues) has no identifiable homologs of known function in BLAST (40) searches (other than TviD proteins encoded by other viaB loci), nor does it contain any informative motifs. In contrast, the C-terminal part of the protein contains two distinct putative tetratricopeptide repeat (TPR) regions with four and five repeats, respectively (Fig. 4A). These repeats were predicted with 99.6% probability by TPRpred (41, 42). The TPR-containing region comprises over 60% of TviD and complicates modeling of the N-terminal domain, which was presumed to form the O-acetyltransferase. To produce a model offering insight into the putative catalytic domain, the TPR region of the protein was excluded. The N-terminal sequence (residues 1–322) was submitted to RaptorX (43), and the highest-scoring models were created using members of the SGNH hydrolase family from Enterococcus faecalis (PDB 1YZF), Veillonella parvula (PDB 4RW0), a predicted protein from the metagenome of Lake Arreo, Spain (PDB 5JD3), Alicyclobacillus acidocaldarius (PDB 3RJT), and Geobacillus stearothermophilus (Axe2: PDB 3W7V) (Fig. S6). None of the templates are known O-acetyltransferases, but the SGNH hydrolase family is comprised of a large group of esterases, lipases, and transferases and does include some established O-acetyltransferases (44, 45, 46). As the name suggests, SGNH proteins typically possess a conserved Ser, Gly, Asn, and His in their active sites (44).

Figure 4.

Activity and predicted structure of TviD.A, Predicted domain organization of TviD with putative C-terminal TPR domain based on TPRpred (41, 42). B, Structural model of the N-terminal part of TviD (residues 1–322). The model was generated by RaptorX (43) using an uncharacterized acyl hydrolase/lipase from E. faecalis (PDB 1YZF) as a template. Residues 277–322 form an extended loop and were removed here to highlight the putative active site (inset). Candidate catalytic residues are depicted as red sticks. C, sequence and secondary structure alignment of the DXXH motif-containing region of TviD^1–322 and 1YZF. The predicted secondary structure of TviD^1–322 and the solved structure of 1YZF are displayed above and below the alignment, respectively. Alpha-helices and strict β-turns are represented by helices and TT, respectively. The conserved Asp and His of the DXXH motif are indicated with a red triangle. The alignment was generated with ClustalW (73) and displayed with ESPript (74). Alignment of the complete proteins is shown in Fig. S7. D, Western immunoblots of proteinase K-digested whole-cell lysates from E. coli Top10 recombinants showing the absence of O-acetylation of Vi antigen in constructs with tviD deleted. The mutation is complemented using plasmids expressing tviD in trans, and activity is eliminated by mutating Cys18 or His253. Expression of TviD and its mutants was confirmed by probing protein lysates with anti-His₅ antibody. The control (viaB) and other constructs were transformed into E. coli Top10. Western immunoblots are representative of three biological replicates.

The top-ranked model was produced using the uncharacterized acyl hydrolase/lipase from E. faecalis (PDB 1YZF) as the template (Fig. 4B). The model has a p value of 5.15e⁻³, indicating a model of good quality, despite limited sequence identity (∼18%) shared between 1YZF and TviD^1–322 (Fig. S7). 1YZF possesses a canonical α/β hydrolase fold, conserved among members of the SGNH hydrolase family, and this is predicted in the TviD^1–322 model (Fig. 4B). Residues 277–322 form a long loop, which extends away from the rest of the structure and does not align well with the sequence of 1YZF (Fig. S7), so these residues were excluded in the model to focus on the putative active site (Fig. 4B). SGNH hydrolases typically possess four consensus sequence blocks with the conserved Ser-His-Asp catalytic triad housed in blocks I and V (44, 47). Block I contains the nucleophilic Ser in the GDS motif. Based on the alignment results (Fig. S6), TviD does not possess a typical block I and instead has a conserved Gly followed by Ser-Cys, which creates a gap in the alignment. The residue representing the nucleophile is not apparent from the sequence alignment alone, but Ser17 and Cys18 are both candidates based on proximity in the model. Although Cys is an atypical nucleophile, Cys18 has the correct orientation in the model to act as a nucleophile; it extends into the active site and is found on a loop adjacent to Asp250 and His253 (Fig. 4B). Sequence alignments suggest that Asp250 and His253 make up the conserved DXXH motif (block V) (Figs. 4C and S6).

To investigate the role of these putative catalytic residues in His₆-TviD, site-directed mutagenesis was used to convert each residue to Ala, and the mutant variants were tested for their ability to restore O-acetylation in E. coli Top10 transformed with viaB-ΔtviD (pWQ1041). His₆-TviD^C18A and His₆-TviD^H253A were unable to restore O-acetylation based on reactivity with the anti-Vi-OAc antibody (Fig. 4D). In contrast, His₆-TviD^D250A restored activity to the viaB-ΔtviD mutant (Fig. 4D). The alternative nucleophile candidate, Ser17, also proved to be unimportant for activity (data not shown). To verify that the lack of O-acetyltransferase activity was not due to altered protein expression, lysates were probed in Western immunoblots with anti-His₅ antibody (Fig. 4D). In further analysis, the majority of His₆-TviD and mutant variants localized to the membrane (P100) (Figs. 5B and S8A). Checking protein folding for the full-length protein was compromised by its membrane association so, instead, folding was assessed using soluble truncated forms containing only residues 1–322, encompassing the functional catalytic domain (see below). Protein thermal melt curves confirmed that all of the truncated variants possessed similar melting temperatures (Fig. S8B). Based on these results, we propose that TviD catalysis is dependent on either an atypical Cys-His-Asp triad or a Cys-His dyad.

Figure 5.

Importance of the TPR-containing domain of TviD for subcellular localization.A, Predicted domain organization with TPR regions predicted by TPRpred (41, 42). B, Cellular distribution of TviD and C-terminal deletion constructs expressed in E. coli Top10. Protein was detected by Western immunoblots probed with anti-His₅ antibody. S12 and P12 are soluble and pelleted fractions, respectively, from centrifugation at 12,000g, while S100 and P100 are the corresponding fractions obtained after centrifugation of the S12 fraction at 100,000g. The pH of the resuspension buffer used for the analysis of all proteins was 7.4. C, Western immunoblots of proteinase K-digested whole-cell lysates from E. coli Top10 recombinants showing that O-acetylated Vi antigen is produced by expression of TviD, or its C-terminally-truncated derivatives, in trans to complement the tviD deletion. Expression of TviD and the truncation mutants was confirmed by probing protein lysates with anti-His₅ antibody. The control (viaB) and other constructs were transformed into E. coli Top10. D, Immunofluorescence microscopy of E. coli Top10 (negative control) and a transformant containing a plasmid carrying the viaB locus (positive control). His₆-TviD and His₆-TviD^1–322 were expressed in the tviD-deletion. Cells were probed with anti-Vi-OAc and anti-Vi antigen primary antibodies and fluorescent secondary antibodies for Vi antigen detection. All transformants produced surface-expressed Vi antigen. Scale bars represent 7 μm. All experiments are representative of three biological replicates.

The TPR domain of TviD is not essential for O-acetyltransferase activity

The C-terminal TPR domain of TviD contains nine putative TPR motifs; four located between residues 323 and 461 and five between residues 614 and 809 (Fig. 5A). To investigate the role of each TPR region, three sequential C-terminal truncations were made, beginning by removing the five TPRs (His₆-TviD^1–613) and then extending to include the intervening region between the two sets of TPRs (His₆-TviD^1–461) and finally the entire TPR region (His₆-TviD^1–322). To avoid disruption of α-helices, the end points for these truncations were located between predicted secondary structure elements. Cellular fractionation and Western immunoblots with anti-His₅ antibody revealed that His₆-TviD^1–613 localized to the membrane fraction (P100) like His₆-TviD (Fig. 5B). Two putative membrane-associated amphipathic helices were predicted at the C-terminus (residues 528–545 and 731–748) by HELIQUEST (48). In contrast, the majority of His₆-TviD^1–461 and His₆-TviD^1–322 were in the soluble fraction (S100) (Fig. 5B). These data implicate the C-terminus of TviD (residues 614–831) as being essential for membrane association.

To determine whether the truncated proteins retain O-acetyltransferase activity, E. coli Top10 was transformed with a plasmid carrying viaB-ΔtviD and another encoding either His₆-TviD^1–322, His₆-TviD^1–461, or His₆-TviD^1–613. Each of the truncation derivatives restored Vi antigen O-acetylation (Fig. 5C). Therefore, the TPRs are not essential for O-acetylation under these (plasmid expression) conditions. Immunofluorescence was performed to verify that the Vi antigen made in the presence of truncated TviD was still retained at the cell surface, to rule out the possibility of any unanticipated effects of the deletions on polymer export. As expected, viaB-ΔtviD cells only reacted with the anti-Vi antibody (Fig. 5D). Overexpression of His₆-TviD or His₆-TviD^1–322 restored surface assembly of O-acetylated Vi antigen, which reacted with both Vi antigen antibodies (Fig. 5D). These results established that the O-acetyltransferase catalytic domain resides between residues 1 and 322 and that the TPR structure is not essential for polymer biosynthesis, O-acetylation, or polysaccharide export under these conditions.

TviD and TviE interact via the TPR domain of TviD

TPR domains frequently mediate protein–protein interactions within complexes (49). TviE has an N-terminal domain of unknown function, leading to the hypothesis that these noncatalytic elements might participate in interactions between these proteins. To pursue this, a plasmid (pWQ1059) was created to coexpress His₆-TviD and TviE-FLAG. This construct maintains the native arrangement of the genes, which overlap by one nucleotide (Fig. 6A). Since TviD is membrane-associated, membranes from E. coli Top10[pWQ1059] were collected and solubilized with lauryldimethylamine oxide (LDAO). The solubilized protein fraction was isolated, and potential His₆-TviD/TviE-FLAG interactions were probed using a copurification strategy involving Ni-NTA chromatography. Western immunoblots using epitope-specific antibodies were used to detect His₆-TviD and TviE-FLAG (Fig. 6B). The analysis was complicated by the proteins becoming substantially more susceptible to proteolytic degradation after detergent solubilization (compare to Figs. S4A and 5B for samples without detergent solubilization), but interaction data were obtained. The expected molecular mass of His₆-TviD is 96.8 kDa and the anti-His₅ detects two prominent Ni-NTA-bound proteins (eluted with imidazole) with apparent sizes of 93.1 kDa and 80.0 kDa, indicating some proteolytic cleavage in the C-terminal region. TviE-FLAG has a calculated size of 66.0 kDa, and the anti-FLAG antibody reveals an apparently intact protein (apparent size 65.7 kDa), together with a smaller C-terminal fragment (48 kDa). Although TviE is predominantly soluble when expressed alone (Fig. S4A), a portion of the protein and its degradation product was retained by Ni-NTA in the presence of membrane-associated His₆-TviD (Fig. 6B). When the two proteins were subjected to Ni-NTA chromatography in the absence of their partner, His₆-TviD (and its fragment) bound to the resin, while TviE-FLAG (and its fragment) eluted in the unbound fraction, as expected (Fig. S9, A and B), ruling out nonspecific interactions of TviE-FLAG with the resin. These results support protein–protein interactions and a complex containing TviD and TviE (Fig. 6B).

Figure 6.

TviD and TviE interact.A, Organization of the expression construct used for analyzing His₆-TviD/TviE-FLAG interactions, which was expressed in E. coli Top10. B, LDAO was used to solubilize TviD from the membrane fraction, and the detergent extract was loaded onto an Ni-NTA column. The unbound flow-through (FT), imidazole washes (0 mM and 20 mM), and elution fractions (E1–3) are shown. Samples from each fraction were separated by SDS-PAGE, and proteins were visualized using Coomassie stain and Western immunoblotting with anti-His₅ and anti-FLAG antibodies. The elution fractions contained both His₆-TviD and TviE-FLAG. C, Replication of the experiment with a plasmid expressing His₆-TviD^1–322E-FLAG eliminated the interaction. The resuspension buffer used for both experiments had a pH of 7.4 and the washes contained 0 and 20 mM imidazole. Copurification results are representative of three biological replicates.

To examine the potential role of the TPR domain of TviD in interactions, the codons for residues Pro323 and Val324 were each replaced with stop codons to truncate TviD before the C-terminal TPR domain (Fig. 6A). In the absence of the TPR domain, His₆-TviD^1–322 is soluble (Fig. 5B) (like His₆-TviE; Fig. S4A), therefore, the Ni-NTA copurification strategy was performed using the S100 fraction. Under these conditions, His₆-TviD^1–322 (expected molecular mass of 38.4 kDa) has minimal degradation, since the C-terminal proteolytically susceptible domain is absent. Most of the protein was bound to the Ni-NTA resin and subsequently eluted with imidazole (Fig. 6C). In the absence of LDAO, TviE-FLAG does not undergo N-terminal degradation but is no longer retained on the Ni-NTA resin. These data indicate that the TPR domain of TviD is required for its ability to interact with TviE.

Bordetella species possess TviDE homologs and produce a CPS related to Vi antigen

Loci resembling viaB from S. Typhi are found in soil-inhabiting species of members of the Burkholderiaceae (including some Bordetella species) and the VexE protein from A. denitrificans can replace the S. Typhi homolog (9). In contrast, human and animal adapted Bordetella sp., such as Bordetella bronchiseptica and B. pertussis, possess CPS (cap) loci sharing 99% identity (Fig. 7A) possessing the characteristic kpsMTED genes for ABC transporter-dependent systems. However, these loci lack vexE and instead contain kpsSC genes suggesting a conventional assembly system producing a glycan with a terminal phosphatidylglycerol-linked β-Kdo oligosaccharide, like the classical E. coli “group 2” prototype (1). The precise glycan structures governed by the cap loci are unknown (50), but they encode homologs of TviBCDE (Fig. 7A), offering a possible explanation for published reports of B. pertussis cells reacting with polyclonal mouse antibodies raised against Vi antigen (51). Bordetella possess many virulence factors, which fall under the control of the Bvg two-component system that senses and reacts to the surrounding environment by regulating the transcription of hundreds of genes (52). B. bronchiseptica whole-cell lysates were examined from the virulent (Bvg⁺) and avirulent (Bvg⁻) phases (52), since basal Vi antibody reactivity (and cap gene expression) was previously reported in Bvg⁺, with a substantial upregulation observed in Bvg⁻ cells (51). In our hands, reactivity with anti-Vi-OAc antibody was indistinguishable between samples from the two Bvg phases, while the Bvg⁺ sample reacted more strongly than the Bvg⁻ sample with the polyclonal anti-Vi antibody (Fig. 7B). The reason for the difference in Bvg⁻ phase reactivity, compared with published work, is unknown, but it was not pursued because it was not central to the goal of examining the activities of TviDE. As expected, deletion of the cap locus abolished reactivity with either antibody (Fig. 7B). These data indicate that the Vi-like similarities in the cap-encoded CPS of mammalian adapted Bordetella extend to both the glycan backbone and its O-acetylation.

Figure 7.

Production of a Vi-related polysaccharide by B. bronchiseptica and participation of TviD and TviE from B. bronchiseptica with S. Typhi proteins in Vi antigen synthesis.A, comparison of the genetic organization of the CPS loci from S. Typhi and B. bronchiseptica. B, B. bronchiseptica WT and cap locus deletion (Δcap) cells were collected from Bvg⁺ and Bvg⁻ phase bacteria. Lysates were analyzed via Western immunoblots probed with anti-Vi-OAc and anti-Vi antibodies. E. coli Top10 harboring a plasmid containing viaB provides the positive control. C, cellular distribution of His₆-TviD_Bb and His₆-TviE_Bb, expressed in E. coli Top10 and BL21(DE3), respectively. Protein was detected by Western immunoblots probed with anti-His₅ antibody. S12 and P12 are soluble and pellet fractions, respectively, from centrifugation at 12,000g, while S100 and P100 are the corresponding fractions obtained after centrifugation of the S12 fraction at 100,000g. The pH of the resuspension buffers used to analyze His₆-TviD_Bb and His₆-TviE_Bb was 6.7 and 7.7, respectively. D, Western immunoblots of proteinase K-digested whole-cell lysates showing the complementation of the tviD and tviE-deletions by expressing His₆-TviD_Bb and His₆-TviE_Bb, respectively. Expression of the proteins was monitored by probing protein lysates with anti-His₅ antibody. Western immunoblots in (C) and (D) are representative of three biological replicates.

TviD and TviE from B. bronchiseptica (referred to as TviD_Bb and TviE_Bb) share 23% and 32% identity with the homologs from S. Typhi, respectively, and domain organizations are preserved. TviD_Bb is proposed to have an N-terminal O-acetyltransferase domain, followed by a shorter TPR domain compared with TviD from S. Typhi, with only three repeats (Fig. S10A). The N-terminus of TviD_Bb (residues 1–274) was modeled using RaptorX (43) and an SGNH hydrolase from the metagenome of Lake Arreo, Spain (PDB 5JD3) as a template (Fig. S10B). This template was also used to produce the third highest scoring model of TviD from S. Typhi using RaptorX (43). The residues determined to be catalytically important in TviD from S. Typhi (Cys18 and His253) are conserved in TviD_Bb, along with Asp250 in the DXXH motif (Fig. S10, B and C). The TviE_Bb sequence predicts a C-terminal GT domain belonging to the GT4 family (Fig. S11A). A model of TviE_Bb, created using Phyre2 (34) and AtSus1 (PDB 3S29) as the template, resembles TviE from S. Typhi but possesses a larger, more complex, N-terminal domain (Fig. S11B). The catalytic site, including the important EX₇E motif, is also conserved (Fig. S11, B and C).

To examine the properties of the B. bronchiseptica homologs, cellular fractionation was performed with E. coli Top10 and BL21(DE3) cells expressing His₆-TviD_Bb and His₆-TviE_Bb, respectively. Samples from each fraction were analyzed by Western immunoblots with anti-His₅ antibodies and indicated similar cellular distribution to the S. Typhi homologs; His₆-TviD_Bb localized to the membrane (P100) and His₆-TviE_Bb was found in the soluble (S100) and membrane (P100) fractions (Fig. 7C). Since the early steps of Vi antigen biosynthesis differ in S. Typhi and B. bronchiseptica, the capacity of His₆-TviD_Bb and His₆-TviE_Bb to productively participate in the Vi antigen biosynthesis machinery from S. Typhi was investigated by complementation experiments using E. coli Top10 and BL21(DE3) transformed with viaB-ΔtviD and viaB-ΔtviE, respectively. Immunoblots of the corresponding proteinase K-digested whole-cell lysates demonstrated that the B. bronchiseptica proteins were active in these backgrounds (Fig. 7D), although the amount of Vi antibody reactive material was reduced compared with that produced by the native proteins (compare to Figs. 2D and 4D). No conclusions were drawn about the relative amounts due to variables beyond protein structures, including different host and vector combinations.

Discussion

The results described here show that TviD and TviE are an O-acetyltransferase belonging to the SGNH hydrolase family and a GT4 family GT for the biosynthesis of Vi antigen in S. Typhi and soil isolates of the Burkholderiaceae that include some Bordetella species, as well as Vi-related CPS (whose structure has not been elucidated) in mammalian-adapted Bordetella species. When combined with the enzymes for precursor formation, S. Typhi TviDE is sufficient for synthesis of O-acetylated Vi antigen.

The involvement of TviE as the sole GT required for Vi antigen biosynthesis indicates that this enzyme is a single catalytic site polymerase. There is precedent for GT4 family enzymes being able to add more than one glycosyl residues in the activity of PglH, a single active site processive polymerase from C. jejuni (32, 37). PglH adds precisely three GalNAc residues in biosynthesis of a heptasaccharide used for N-linked protein glycosylation, and it contains a molecular ruler to ensure fidelity (32). There is no comparable structure in TviE, as might be expected for an enzyme that generates much larger products. Polymerizing GTs may act on a glycose acceptor synthesized by one or more other GTs, as would be the case in a CPS with a terminal phosphatidylglycerol-linked β-Kdo oligosaccharide, which is implicated by the cap locus in B. bronchiseptica. In S. Typhi, the initiation reaction for Vi synthesis and the origin of the terminal diacyl-HexNAc have not yet been resolved, but proper acylation of the terminus by VexE is not required for polymerization (9). One attractive hypothesis is that S. Typhi TviE initiates polymerization using a sugar nucleotide acceptor. This strategy is used by cellulose synthase (BcsA), which has been studied in depth in Rhodobacter sphaeroides (53). Biosynthesis of cellulose occurs without the involvement of lipid-linked intermediates. The cellulose polymer is extended one glucose molecule at a time by BcsA (a GT2 enzyme). Growth of cellulose occurs at the nonreducing terminus, with the newly added glucose residue forming the acceptor for subsequent reactions (53). Another example is provided by the Class I bacterial hyaluronan synthase (HAS) from Streptococcus equisimilis (54). This enzyme initiates synthesis of chitin (oligo-GlcNAc) oligomers using UDP-GlcNAc as the initial acceptor and extends these oligosaccharides with the [3)-β-GlcNAc-(1→4)-β-GlcA-(1→] hyaluronan backbone (55). The mechanism by which this occurs has not been elucidated but addition occurs at the reducing terminus. The direction of Vi antigen growth has not been determined in S. Typhi, but the assembly system in which TviE_Bb participates seems to make extension at the nonreducing end (building on the β-Kdo oligosaccharide) more likely. Our working hypothesis is that synthesis of Vi antigen begins on UDP-HexNAc prior to the sequential transfer of GalNAcA residues (from their UDP-linked precursor) to the nonreducing terminus. VexE would then acylate the reducing terminal GlcNAc residue during or postpolymerization.

TviE requires an EX₇E motif for function like other GT4 proteins. Substitution of either Glu residue with Ala rendered TviE inactive without impacting protein solubility or stability. Like other GT4 enzymes, such as AtSus1, Ss2, SpsA, MshA, and PimB, TviE possesses basic residues within the active site to aid in the stabilization of the nucleotide sugar (35, 56, 57, 58, 59). Arg406 and Lys411 are conserved among these enzymes and are involved in coordinating the pyrophosphate oxygens in the active site (Fig. S2).

Despite the lack of sequence homology to characterized O-acetyltransferases, TviD possesses an O-acetyltransferase domain belonging to the SGNH hydrolase family. Other members of this family, such as CsaC (MynC/SacC) and CssE (OatC), are involved in O-acetylation of CPS from N. meningitidis serogroups A and C, respectively (60, 61). Based on structural predictions, an SGNH hydrolase-like fold with an α/β/α core is found in the N-terminal part of TviD (Fig. 4B). Typically, Ser, Gly, Asn, His, and Asp/Glu are located within the catalytic core in conserved blocks, but as the SGNH family has grown, it is evident that some enzymes may lack these canonical residues (44). Most SGNH hydrolases possess a catalytic triad consisting of a Ser (nucleophile) from block I and a His and Asp (sometimes Glu) from block V (Fig. S6). An oxyanion hole is typically formed by the NH of the catalytic Ser from block I, the NH of Gly from block II, and the amide of Asn from block III (44). These residues hydrogen bond with the negatively charged intermediate, stabilizing the transition state (44). Multiple sequence alignments highlight the lack of sequence similarity shared by TviD and other representatives in regions of blocks II (NXSXXGXT) and III (GXND), and they align poorly in block I due to the presence in TviD of Gly-Ser-Cys instead of the typical Gly-Asp-Ser motif. Block V (DXXH) is conserved (Fig. S6). These features are conserved in TviD_Bb (Fig. S10C). Although atypical, the Cys in block I and the His of block V are required for O-acetylation of Vi antigen by TviD (Fig. 4D). There is precedent for this type of arrangement provided by dienelactone hydrolase (Cys-His-Asp catalytic triad), and cysteine nucleophiles are commonly used by cysteine proteases in either a catalytic triad or dyad (47, 62, 63). O-acetylation was retained when the Asp of block V was replaced with Ala indicating that it is not essential for the activity of TviD (Fig. 4D). It is plausible that TviD has a Cys-His dyad. Asp is typically required to activate His, allowing it to deprotonate the nucleophile, but as Cys has a lower pK_a than Ser, the Asp may not be essential (44). An extracellular lipase from Streptomyces rimosus, a member of the SGNH hydrolase family, is known to possess a Ser-His catalytic dyad, although the mechanism has not been elucidated (64). Since blocks II and III are not conserved in TviD, it is challenging to identify the other residues, which compose the oxyanion hole in this enzyme, in the absence of a solved structure.

The C-terminal TPR domain of TviD is not essential for O-acetylation of Vi antigen but is required for membrane association and interaction with TviE. The function of the N-terminal α-helical domain of TviE offers a candidate for interaction, but this cannot be assessed because deletions of even a small part of this domain render TviE inactive in vivo. This part of the protein appears to be located away from the GT active site and may also be involved in protein–protein interactions. Unfortunately, attempts to get more insight into the stoichiometry of the complex by size-exclusion chromatography were unsuccessful. The His₆-TviDE-FLAG complex was found in a large single peak, calculated to be >600 kDa by interpolation from a standard curve (Fig. S12, A and B). The micelle size of LDAO is 17–22 kDa (65, 66), and monomeric His₆-TviD and TviE-FLAG are 96.8 kDa and 66.0 kDa, respectively, potentially suggesting a complex comprised of multiple copies of each protein, with higher levels of TviD:TviE (Fig. S12C). However, the elution profiles of individual His₆-TviD (Fig. S13A) and His₆-TviE (Fig. S13B) also showed large peaks of >600 kDa, so the contribution of aggregation cannot be excluded.

Although the removal of the TPR domains of TviD eliminates recruitment of TviE to the membrane and interactions between the proteins, this is not essential for in vivo synthesis of O-acetylated Vi antigen, under conditions where the truncated TviD protein is overexpressed. Nevertheless, evidence does exist for a functional complex coupling synthesis and export in the cell. During synthesis, Vi antigen is protected from cytoplasmic expression of a Vi antigen depolymerase enzyme (11). This would allow efficient integration of synthesis and O-acetylation. Notably, the VexE acyltransferase also contains a TPR domain (10 repeats) at the N-terminus of the protein (9, 41, 42), suggesting that it could also participate in a larger complex. The importance of TPR motifs is established for some other exported polysaccharides in Gram-negative bacteria; examples include alginate, cellulose, poly-β-(1,6)-N-acetyl-d-glucosamine (PNAG), and Pel (67). In these examples, the TPR-containing enzymes reside in the periplasm and act as a scaffold, coordinating various interactions. In the PNAG, Pel, and cellulose systems, the TPR-containing proteins have an additional domain, which facilitates polymer export across the outer membrane. In alginate, outer membrane export is completed by a separate protein, which interacts with the TPR-containing protein. The TPR domain has also been shown to interact with polymer-modifying enzymes (67). These data support the hypothesis that the TPR domain may guide some polymers to the export protein during synthesis (67). We cannot exclude further roles like this for the TPR motifs beyond protein–protein interactions in Vi antigen production.

The cap loci of B. pertussis and B. bronchiseptica are 99% identical, and studies on B. pertussis determined that CPS genes are not upregulated during the Bvg⁺ phase of the bacteria (51). However, B. pertussis lacking kpsT (the gene that encodes the nucleotide-binding domain of the ABC transporter) was drastically attenuated in vivo due to the downregulation of virulence genes, so the precise role played by CPS is uncertain (68). Moderate attenuation occurred when tviC was deleted, suggesting that only the CPS export machinery plays a critical role in virulence (68). The structure of the glycan product directed by the cap locus is unknown, but antibody reactivity shown here and reported previously (51), together with the functional studies of the TviDE homologs from B. bronchiseptica, supports the presence of a glycan resembling (or sharing epitopes with) Vi antigen. The involvement of the TviE homologs in two different polymerization contexts is interesting (and perhaps surprising) and merits further investigation. Unraveling the mechanistic underpinnings will depend on establishing in vitro systems and (potentially) structures of the enzymes. In addition to genes correlated with Vi antigen synthesis in S. Typhi and ABC transporter-dependent CPS export, the cap loci also contain genes encoding two putative GTs (arbitrarily designated GTX and GTY here) (Fig. 7A). By analogy to proposals for related systems in other bacteria (1), at least one GT is anticipated to be required to transition from the β-Kdo oligosaccharide (formed by KpsSC) to the repeat-unit domain of CPS, which is expected to be polymerized and O-acetylated by TviE and TviD, respectively. One of the unassigned GTs might participate in modifying a Vi antigen backbone. This seems likely with the Vi-epitopes still available but can only be resolved with a fully elucidated carbohydrate structure. The information described here provides a foundation for such investigations.

Experimental procedures

Bacterial strains and growth conditions

The strains used in this study were E. coli Top10 (F⁻, mcrA, Δ(mrr-hsdRMS-mcrBC), φ80, lacZΔM15, ΔlacX74, deoR, nupG, recA1, araD139, Δ(ara-leu)7697, galU, galK, rpsL(Str^r), endA1) (Invitrogen), E. coli BL21(DE3) (F⁻ ompT gal dcm lon hsdS_B(r_B⁻m_B⁻) λ(DE3[lacI lacUV5-T7p07 ind1 sam7 nin5]) [malB⁺]_K-12(λ^S)) (Invitrogen), CWG1238 (S. Typhi H251.1 ΔwaaG::kan; Km^r) (9), and B. bronchiseptica RB50 (50). E. coli and S. Typhi were routinely grown at 37 °C in lysogeny broth (LB) supplemented with the appropriate antibiotic (100 μg mL⁻¹ ampicillin, 34 μg mL⁻¹ chloramphenicol, 50 μg mL⁻¹ kanamycin, or 10 μg mL⁻¹ tetracycline) and 100 μg mL⁻¹ 2,3-dihydroxybenzoic acid when necessary (CWG1238). B. bronchiseptica RB50 was grown at 37 °C in LB (Bvg⁺ phase) or LB supplemented with 50 mM MgSO₄ (Bvg⁻ phase). In B. bronchiseptica, the capsule locus deletion mutant (Δcap) was constructed by replacing the entire locus except the 3′ half of kpsS and tviD with a kanamycin resistance cassette, using an approach as described previously (69).

General DNA methods

The plasmids used in this study are described in Table S1. Plasmid DNA was purified using the PureLink Quick (Invitrogen) or GeneJET (Thermo Fisher) Plasmid Miniprep kit. DNA fragments amplified by PCR or obtained from restriction endonuclease digests were purified using PureLink (Invitrogen) or GeneJET (Thermo Fisher) PCR purification kits. Restriction endonucleases, T4 DNA ligase, and Gibson Assembly mix (NEB) were used according to the manufacturer's instructions. Amino acid insertions and substitutions were generated using the QuikChange protocol (Agilent) for site-directed mutagenesis with KOD Hot Start DNA polymerase (EMD Millipore). The Q5 site-directed mutagenesis protocol (NEB) was followed for the generation of TviD^1–322 using PaCeR polymerase (GeneBio Systems, Inc.). Gene deletions were generated in pWQ783 using inverse PCR with either KOD Hot Start (EMD Millipore) or Phusion High-fidelity (NEB) DNA polymerase. Primers for inverse PCR were synthesized with a 5′-phosphate to facilitate the ligation of the linear product. Custom oligonucleotide primers used in this study were purchased from Sigma and Integrated DNA Technologies (Table S2). All constructs were confirmed via DNA sequencing at the Advanced Analysis Center at the University of Guelph.

Protein expression, cellular fractionation, and purification

Overnight bacterial cultures were diluted 1:50 into LB supplemented with the appropriate antibiotic. Prior to inducing protein expression, cultures were grown to an OD₆₀₀ of 0.5. Protein expression in E. coli Top10 containing pBAD plasmids was induced using L-arabinose (0.02% [w/v]) or isopropyl-β-D-thiogalactopyranoside (0.5 mM) for E. coli BL21(DE3) containing pET constructs. Cultures were transferred to 18 °C and growth continued for an additional 16 h. Cells were then collected by centrifugation at 5000g for 20 min at 4 °C and resuspended in 50 ml of resuspension buffer, containing one cOmplete Mini EDTA-free protease inhibitor tablet (Roche Applied Science). The resuspension buffer (50 mM Tris, 250 mM NaCl) was used with different pH values depending on the protein being analyzed, and these are indicated in relevant figure legends. The cells were lysed by passage through an EmulsiFlex-C3 (Avestin) at 15,000 psi. Unbroken cells and large debris (P12 fraction) were removed by centrifugation at 12,000g for 20 min at 4 °C, and the supernatant (S12 fraction) was separated into membranes (P100) and soluble fractions (S100) by centrifugation at 100,000g for 1 h at 4 °C. To extract membrane-associated His₆-TviD, the membrane pellet (P100) was resuspended in 20 ml of resuspension buffer containing 1% (v/v) LDAO (Sigma) and incubated overnight with rocking at 4 °C. Insoluble material was removed by centrifugation at 100,000g for 1 h at 4 °C. The supernatant, containing solubilized His₆-TviD, was used in the subsequent purification steps with the buffers containing 0.1% (v/v) LDAO. To purify soluble (or solubilized) proteins, samples were loaded on an IMAC gravity column containing 2 ml of Ni-NTA agarose resin (Qiagen). The matrix was washed successively with ten column volumes of each wash buffer (specified in the relevant figure legends), before the protein was eluted with five column volumes of resuspension buffer containing 300 mM imidazole. The eluted protein was buffer exchanged using a PD-10 desalting column (GE Healthcare) into imidazole-free resuspension buffer with 150 mM NaCl (containing 0.1% [v/v] LDAO for detergent-solubilized proteins). The protein was then concentrated using a Vivaspin6 10 kDa, 30 kDa, or 50 kDa MWCO column (Sartorius), depending on the size of the protein. The folding state of purified proteins was assessed by using the Protein Thermal Shift Dye kit (Thermo Fisher) in combination with a StepOnePlus Real-Time PCR System (ThermoFisher), according to the manufacturer's instructions. To examine the cellular distribution of epitope-tagged proteins, samples of the lysate, S12, P12, S100, and P100 fractions were separated by SDS-PAGE and subjected to Western immunoblotting.

Analysis of potential TviDE complexes

To examine potential interactions between TviD and TviE, His₆-TviDE-FLAG or His₆-TviD^1–322E-FLAG was expressed in E. coli Top10 transformed with plasmids pWQ1059 or pWQ1060, respectively, and P100 and S100 fractions were prepared as described above. His₆-TviD^1–322E-FLAG is present in the soluble fraction (S100), while His₆-TviDE-FLAG is membrane-associated and was solubilized with LDAO. To purify putative protein complexes, the column matrix was washed consecutively with 20 column volumes of each wash buffer prior to protein elution. Samples from each step in the process were collected for analysis using SDS-PAGE and Western immunoblotting.

For gel-filtration chromatography, His₆-TviD, His₆-TviE (pWQ1065), and His₆-TviDE-FLAG eluted from Ni-NTA columns were buffer exchanged using a PD-10 desalting column (GE Healthcare) into 50 mM Tris, 150 mM NaCl, pH 7.4, 0.1% (v/v) LDAO. The samples were then concentrated to approximately 5 mg mL⁻¹ using either a Vivaspin6 30 kDa (His₆-TviE and His₆-TviDE-FLAG) or 50 kDa (His₆-TviD) MWCO column (Satorius). Samples were analyzed using an ÄKTA Pure FLPC equipped with a Superdex 200 Increase 10/300 Gl column (GE Healthcare). Gel-filtration chromatography was carried out in the same buffer as the loaded sample, and the absorbance was monitored at 280 nm during elution. Fractions containing His₆-TviDE-FLAG were collected and concentrated using a Vivaspin6 30 kDa MWCO column (Satorius), and 5 μg samples were examined by SDS-PAGE and Western immunoblotting.

SDS-PAGE and Western immunoblotting

Samples consisted of purified polysaccharides (see below) or whole-cell lysates. To analyze Vi antigen in lysates, cells corresponding to 1 optical density (OD₆₀₀) unit-equivalent were collected by centrifugation, and whole-cell lysates were prepared by proteinase K-digestion (70). To detect protein expression, lysates were prepared in the absence of proteinase K. Samples were routinely heated at 100 °C for 10 min prior to SDS-PAGE. However, detection of membrane-associated TviD was an exception, where samples were prepared following the protocol described previously (71) and incubated at 37 °C for 30 min. All samples were separated using SDS-PAGE with 10% resolving gels. For detection of Vi antigen by Western immunoblotting, samples were transferred to positively charged nylon membranes (Biodyne B; Pall) at 250 mA for 1 h. Proteins were transferred to nitrocellulose membranes (Protran; GE Healthcare) using the Power Blotter Semi-dry transfer system (Thermo Fisher). The antibodies used to detect Vi antigens were murine monoclonal anti-Vi-OAc antibody P2B1G2/A9 ((25); diluted 1:300) or polyclonal anti-Vi antibody (BD Difco; diluted 1:350). The secondary antibodies were alkaline phosphatase-conjugated goat anti-mouse (Cedarlane; diluted 1:3000) or anti-rabbit secondary antibodies (Cedarlane; diluted 1:3000). Immunoblots were developed using 5-bromo-4-chloro-3-indoyl phosphate and nitroblue tetrazolium (Roche Applied Science). Epitope-tagged proteins were detected using either murine monoclonal anti-FLAG (Sigma; diluted 1:1000), murine monoclonal anti-pentahistidine (His₅) (Qiagen; diluted 1:2000), and the secondary antibody was horseradish-peroxidase-conjugated goat anti-mouse (Cedarlane; diluted 1:3000). Detection was accomplished using the HRP-substrate Luminata Crescendo (EMD Millipore). Molecular weight determination was performed using Image Lab 6.0.1 software (BioRad).

Purification of and NMR analysis of Vi antigen

The method for purifying native exported CPS from E. coli CWQ1238 has been described elsewhere (9). Large-scale purification of internal Vi antigen was achieved using a modification of the published method (9). Twelve-liter LB cultures, supplemented with the appropriate antibiotic(s), were inoculated (1:100) from overnight cultures of E. coli Top10 transformed with plasmids encoding TviABCDE (pWQ1043), TviABCE (pWQ1046), or TviABCE[His₆-TviD] (pWQ1052). The first two of these were grown at 37 °C for 16 h, while the latter was grown following the induction protocol described above for expressing the recombinant proteins. Cells were collected by centrifugation (5000g for 20 min) and lyophilized. The dried cell pellet was resuspended in 50 ml of 50 mM Tris buffer (pH 8.0) containing 5 mM EDTA and 2 mg mL⁻¹ lysozyme (Sigma). The mixture was stirred for 16 h at 4 °C and then diluted to a final volume of 100 ml by adding 50 mM Tris buffer (pH 8.0) containing 5 mM EDTA, 10 mM MgCl₂, and 1 μg mL⁻¹ each of DNAse, and RNAse (Roche Applied Science). The digestion of DNA and RNA was proceeded with stirring for 1 h at 37 °C. An equal volume of 90% aqueous phenol was added to the preparation, and the mixture was stirred for 1 h at 70 °C. The mixture was then cooled on ice for 20 min to bring the temperature below 15 °C, and phase separation was achieved by centrifugation at 10,000g for 20 min. The aqueous phase was collected, placed in a 3.5 kDa MWCO dialysis membrane (Spectrum Labs), and dialyzed against water for 2 days. Finally, the sample was centrifuged at 100,000g for 16 h, and the supernatant was retained and lyophilized.

NMR analysis was performed on intracellular Vi antigen, and the spectra were compared with native and chemically de-O-acetylated native CPS obtained by base treatment with ammonium hydroxide, as described (9). The inherent viscosity of native Vi antigen results in spectra with broad peaks in NMR spectra, so data were collected for a modified polysaccharide obtained by solvolysis with trifluoroacetic acid (TFA). Seventy-five milligram samples of purified polysaccharide were resuspended in 500 μl of 99% TFA and incubated at 100 °C for 2 h. After solvolysis, TFA was evaporated using nitrogen gas. The dried material was resuspended in water and lyophilized twice before resuspending in 1 ml of water. The sample was then loaded onto a Sephadex G-50 superfine column (2.5 × 75 cm) eluted with 1% (v/v) acetic acid and 0.4% (v/v) pyridine with a flow rate of 0.6 ml min⁻¹, and the elution was monitored using a Smartline 2300 refractive index detector (Knauer). The modified polysaccharide eluted immediately after the void volume of the column was collected, concentrated using a rotary evaporator, and lyophilized. The samples were deuterium-exchanged by freeze-drying twice with 99.9% D₂O, resuspended in 250 μl of 99.9% D₂O, and placed in a 5-mm Shigemi NMR microtube. ¹H, ¹³C NMR, and two-dimensional ¹H, ¹³C heteronuclear single-quantum coherence (HSQC) spectra were obtained at 25 °C using a Bruker Avance III 600-MHz spectrometer equipped with a 5 mm TCI cryoprobe, located in the NMR Centre of the Advanced Analysis Centre at the University of Guelph. Sodium 3-trimethylsilylpropanoate-2,2,3,3-d4 was used as an internal chemical shift reference (δ_H = 0 ppm, δ_C = −1.6 ppm).

Immunofluorescence microscopy

Immunofluorescence microscopy was performed on fixed cells to examine the localization of Vi antigen, following published protocols (71, 72). Overnight cultures of E. coli Top10 transformants were diluted 1:50 in 5 ml of LB supplemented with the appropriate antibiotics and L-arabinose, where induction of recombinant proteins was required. Cultures were grown at 37 °C until an OD₆₀₀ of 0.5 was reached, at which time 1 OD₆₀₀ unit-equivalent of cells was collected by centrifugation (5000g for 10 min). The cells were resuspended in 1 ml of 5% (v/v) formaldehyde and incubated at room temperature for 1 h. The fixed cells were collected by centrifugation and washed twice with 1 ml of phosphate-buffered saline (PBS), prior to resuspension in 50 μl of PBS. Glass microscope slides were coated with poly-L-lysine, and 10 μl samples were added to each well. The slides were incubated at room temperature for 10 min. Permeabilization of cells was used to visualize intracellular Vi antigen and was accomplished by adding 10 μl of 0.5 mg mL⁻¹ lysozyme (in 25 mM Tris, 10 mM EDTA, pH 8) and 0.1% Triton X-100 (in PBS), separately, with 15 min incubation for each. The wells were then blocked with 1% (w/v; in PBS) bovine serum albumin (BSA) for 15 min. Following blocking, the slides were treated with either murine anti-Vi-OAc antibody P2B1G2/A9 (25) or rabbit anti-Vi antibody (BD Difco) diluted 1:100 in 1% BSA. After incubation at room temperature for 30 min, the unbound primary antibody was removed by washing three times. The secondary fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse (Sigma) or anti-rabbit (Sigma) antibodies were diluted 1:250 in 1% BSA, added to the wells, and the slides were incubated for 30 min in the dark at room temperature. Finally, the slides were then washed and mounted in Vectashield (Vector Laboratories). Images were obtained using an inverted Leica DMi8 microscope connected to a Quorum Diskovery Spinning Disk system (Quorum Technologies Inc.) and processed using Volocity software (version 6.3; PerkinElmer) in the University of Guelph Advanced Analysis Centre.

Data availability

The authors declare that all data are contained within the manuscript and supporting information.

Supporting information

This article contains supporting information (9, 28, 34, 41, 42, 43, 73, 74, 75).

Conflicts of interest

The authors declare that they have conflicts of interest with the contents of this article.

Edited by Gerald Hart