Lipopolysaccharide is a major immunogenic structure for the pathogen Yersinia pseudotuberculosis, which contains the O-specific polysaccharide (OPS) that is presented on the cell surface. The OPS contains many repeats of the oligosaccharide O-unit and exhibits a preferred modal chain length that has been shown to be crucial for cell protection in Yersinia. It is well established that the Wzz protein determines the preferred chain length of the OPS, and in its absence, the polymerization of O units by the Wzy polymerase is uncontrolled. However, for Y. pseudotuberculosis, a wzz mutation has never been described. In this study, we examine the effect of Wzz loss in Y. pseudotuberculosis serotype O:2a and compare the lipopolysaccharide chain-length profile to that of Escherichia coli serotype O111. In the absence of Wzz, the lipopolysaccharides of the two species showed significant differences in Wzy polymerization. Yersinia pseudotuberculosis O:2a exhibited only OPS with very short chain lengths, which is atypical of wzz-mutant phenotypes that have been observed for other species. We hypothesise that the Wzy polymerase of Y. pseudotuberculosis O:2a has a unique default activity in the absence of the Wzz, revealing the requirement of Wzz to drive O-unit polymerization to greater lengths.
O-specific polysaccharides (OPS, also known as O-antigen) are highly polymorphic cell surface structures exclusive to Gram-negative bacteria. OPS structures exist as variable-length polymers of repeating oligosaccharide units (O units) that form a major constituent of the lipopolysaccharide (LPS) anchored in the outer membrane. For many Gram-negative bacteria, including the pathogen Yersinia pseudotuberculosis, the LPS is highly immunogenic and has been shown to be an important virulence determinant (Porat et al., 1995; Mecsas et al., 2001; Ho et al., 2008). In Y. pseudotuberculosis, the majority of the genes required for OPS synthesis and processing are clustered in an operon flanked by hemH and gsk (Reeves et al., 2003). There are currently 21 defined OPS serotypes in Y. pseudotuberculosis, and for many, we have the OPS gene cluster sequences and O-unit structures (Reeves et al., 2003; Cunneen et al., 2009; De Castro et al., 2011, 2012; Kenyon et al., 2011).
There are several mechanisms by which the OPS can be synthesized and exported (Reeves & Cunneen, 2009), although only the Wzy-dependent pathway has been observed in Y. pseudotuberculosis (Reeves et al., 2003). In all pathways, construction of the OPS begins in the cytoplasm with the synthesis of activated nucleotide-linked sugar precursors. In the Wzy-dependent pathway, the O units are built up on the inner-membrane lipid carrier, undecaprenyl phosphate (UndP), by sequential transfer of sugars from the precursors added by glycosyltransferase enzymes. Four inner-membrane proteins then process the O units to form the OPS before export to the cell surface. This involves a Wzx translocase that ‘flips’ O units to the periplasmic face of the membrane before polymerization into the OPS chain by the Wzy polymerase. The chain-length determinant, Wzz, imposes a modal chain-length distribution on the OPS polymer before ligation to the lipid A-core of the LPS by the WaaL O-antigen ligase (Daniels et al., 2002; Valvano et al., 2011; Han et al., 2012). In Y. pseudotuberculosis, the wzx, wzy and wzz genes are always in the OPS gene cluster (Reeves & Cunneen, 2009), whereas waaL is located in the outer-core oligosaccharide gene cluster.
The individual functions of the Wzx, Wzy, Wzz and WaaL proteins involved in the OPS processing mechanism are supported by extensive experimental data (Robbins et al., 1967; Kanegasaki & Wright, 1970; Liu et al., 1996; Raetz & Whitfield, 2002; Whitfield & Larue, 2008). Recent studies have suggested that these proteins interact (Tocilj et al., 2008), although there is currently no direct evidence for these interactions or the formation of a protein complex (Carter et al., 2009). However, polymerization in vitro to give a modal chain-length distribution has been shown to only require the presence of Wzy and Wzz (Woodward et al., 2010). The nature of Wzz control on the elongation of the OPS polymer is still unknown, but several models have been proposed. The first proposed model suggests that Wzz may restrict WaaL ligation in a time-dependent manner or until a suitable polymer length has been achieved (Bastin et al., 1993). Alternatively, Morona et al. (1995) suggest that Wzz and WaaL form a complex in which Wzz acts as a chaperone to set a specific ratio of WaaL to Wzy. Moreover, it has been suggested that Wzz may form a scaffold for the recruitment of Wzy proteins (Tocilj et al., 2008; and reviewed in Whitfield & Larue, 2008) or alternatively interact with UndPP-O units to regulate their reception by Wzy (Larue et al., 2009).
In many species, deletion or disruption of wzz typically results in the uncontrolled polymerization of O units by Wzy-producing nonmodal chain lengths ranging from short to very long (Batchelor et al., 1991; Bastin et al., 1993; Morona et al., 1995). This is widely accepted as the default behaviour of Wzy in the absence of Wzz and has been observed in many species including Yersinia enterocolitica (Bengoechea et al., 2002), E. coli (Bastin et al., 1991; Kalynych et al., 2011), Shigella flexneri (Carter et al., 2009), Pseudomonas aeruginosa (Daniels et al., 2002) and several Salmonella enterica serovars (Brown et al., 1991; Hoare et al., 2006; Larue et al., 2009).
In the case of Y. pseudotuberculosis, an early study involving cosmid cloning to isolate and sequence the O:2a OPS gene cluster after transfer to an E. coli host produced a suspected wzz mutant that had an atypical OPS phenotype (Kessler et al., 1991). The primary construct, pPR981, that included the O:2a gene cluster, was reduced in size using BamHI digestion to yield a smaller, secondary construct named pPR1197. Both plasmids were expressed in E. coli SΦ874, a K12 mutant host with native OPS genes deleted, and were assessed for expression of the O:2a OPS structure. The strain with pPR981 had LPS with O-antigen chain-length distribution comparable with that of the wild type, while the strain with pPR1197 produced only very short OPS chain lengths with no modality. Preliminary complementation data indicated that the wzz gene was truncated as a result of the BamHI digestion, but this was not further investigated.
In this article, we show that pPR1197 does indeed have a wzz mutation and that the atypical phenotype is due to the Wzy polymerase having different properties in comparison with those that have been studied previously.
Materials and methods
Bacterial strains and growth conditions
All strains and plasmids used in this study are listed in Table 1. Yersinia pseudotuberculosis strains were routinely grown at 25 °C in nutrient broth (sodium chloride 5 g L−1, yeast extract 5 g L−1, and bacteriological peptone 10 g L−1), with agarose added to 15 L−1 for nutrient agar. Escherichia coli strains were similarly grown at 37 °C. For strains that harboured a selectable antibiotic-resistant marker, the growth media was supplemented with 25 μg mL−1 ampicillin, 25 μg mL−1 chloramphenicol or 25 μg mL−1 kanamycin where appropriate.
pTrc99A with Y. pseudotuberculosis O:2a wzy inserted at NcoI and BamHI sites. AmpR
pTrc99A with Y. pseudotuberculosis O:2a wzz inserted at NcoI and KpnI sites. AmpR
Cloning and overexpression of constructs
The Y. pseudotuberculosis O:2a wzz gene was cloned into pTrc99A expression vector (Amann et al., 1988; obtained from Promega). The gene sequence was amplified from chromosomal DNA using high-fidelity PCR as described previously (De Castro et al., 2012), with primers #6547 and #6550 that included NcoI (5′) and KpnI (3′) restriction sites (Supporting Information, Table S1). All restriction enzymes were obtained from New England Biolabs (NEB) and were used according to the manufacturer's instructions. Cloning was achieved by standard procedures (Sambrook et al., 1989), using E. coli K12 JM109 for selection of positive constructs. Transformations were carried out by electroporation using a Bio-Rad gene pulser, with the following conditions: 25μFD, 25 kV, 200 Ohms. Immediately following electroporation, samples were recovered in nutrient broth for 1 h at 37 °C. To induce the expression of pTrc99A gene inserts, cells were supplemented with 1 mM isopropyl β-D-thiogalactopyranoside when at an optical density at 600 nm (OD600 nm) of 0.4 (c. 2.5 h after subculturing), prior to cell harvest at OD600 nm 0.7 after c. 3 h of growth.
Allelic gene replacement
Deletion of the chromosomal Y. pseudotuberculosis O:2a wzz gene was achieved by allelic gene replacement using homologous recombination (Derbise et al., 2003). Briefly, O:2a cells were transformed with the temperature-sensitive plasmid, pKD46, then grown at 30 °C to OD600 nm 0.4 before induction with 1 mM L-arabinose. Cells were harvested at OD600 nm 0.7 and transformed with a gene cassette consisting of a chloramphenicol resistance gene flanked by 42 bp of sequence homologous to sequence flanking the chromosomal wzz gene. The gene cassette was PCR-amplified from pKD3 using primers #6569 and #6570 (Table S1). Cells were recovered in NB at 25 °C for 3 h, and positive transformants were selected and confirmed by PCR and sequencing.
LPS extraction, visualization and densitometric analysis
A 10-ml overnight culture was diluted 1 : 100 and grown to OD600 nm 0.7. Cells were obtained by centrifugation and were resuspended in 200 μL of TAE buffer (Tris base, acetic acid and EDTA) and 400 μL LPS lysis buffer (100 mM SDS, 50 mM Tris, 0.128 mM NaCl). The samples were mixed with an equal amount of phenol/chloroform (1 : 1), then heated at 65 °C for 15 min. LPS was extracted from the aqueous phase by adding an equal amount of ice-cold ethanol and retrieving the precipitate by centrifugation. Nucleic acid was removed by digesting with DNase (0.6 μg μL−1) and RNase (0.6 μg μL−1) at 37 °C for 45 min, before protein removal by proteinase K digestion (1.2 μg μL−1) at 56 °C for 1 h. LPS was precipitated from the aqueous phase once more and resuspended in distilled water. Visualization of OPS was achieved by SDS-PAGE on a 16% acrylamide gel with silver nitrate staining as described (Tsai & Frasch, 1982). Densitometry of LPS samples was conducted using imagej software provided by the National Institute of Health (NIH). Density plots were constructed in Microsoft Excel using the data obtained from imagej analysis.
Characterization of the OPS gene cluster in pPR1197
A single BamHI restriction site was identified in the Y. pseudotuberculosis O:2a gene cluster sequence (GenBank accession AF461770). The site is located within the wzz gene, 828-bp downstream of the wzz start codon. The BamHI digestion of pPR981 to create the derivative pPR1197 would have resulted in a 324-bp truncation of wzz. Given the location of wzz at the 3′ end of the OPS gene cluster (Fig. 1), this would not cause polar effects in the operon, so the atypical chain-length distribution of pPR1197 would be only due to the inactivation of the wzz gene.
Differences in OPS polymerization in the absence of wzz
We compared the wzz-negative phenotype of Y. pseudotuberculosis O:2a (pPR1197) with that of E. coli O111. This was achieved using E. coli K-12 SΦ874 carrying plasmid pPR2105, which includes the entire E. coli O111 gene cluster but lacks the wzz gene (strain P4657), based on the same cosmid as for pPR981 and pPR1197 (Stevenson et al., 2008). The pPR2105 cosmid was also expressed in E. coli K12, JM109, which contains a copy of wzz on the chromosome to complement this missing factor in pPR2105.
The LPS profiles were visualized by SDS-PAGE (Fig. 2) and compared using densitometric analysis (Fig. 3). For SΦ874 carrying pPR981 (Yps O:2a wzz+) and JM109 with pPR2105 (Ec O111 wzz+), the SDS-PAGE profiles showed LPS with modal, long-chain OPS with 20–30 O units. The LPS profile of SΦ874 carrying pPR2105 (Ec O111 wzz−) was typical of other E. coli wzz mutants. However, the profile of SΦ874 with pPR1197 (Yps O:2a wzz−) had LPS with very short chain lengths, confirming the phenotype reported by Kessler et al. (1991), which we now know is due to the inactivation of the wzz gene. Polymers > 3 O units in length were only weakly detected for this strain, with a significantly higher proportion of nonpolymerized single O units than SΦ874 carrying pPR2105 (Ec O111 wzz−).
Restoration of wzz expression
The wzz-negative strains, SΦ874 with pPR1197 (Yps O:2a wzz−) and SΦ874 with pPR2105 (Ec O111 wzz−), were complemented with the Y. pseudotuberculosis O:2a wzz gene, which restored long-chain modality in both strains (Fig. 4). When pPR1197 (Yps O:2a wzz−) was expressed in E. coli JM101 that contains a chromosomal wzz gene, long-chain OPS modality is similarly restored although there is a minor difference in chain-length preference (Fig. 4). Thus, both wzzO:2a and wzzO111 confer modal chain length on both O antigens, showing that there is no major difference in the function of the two wzz genes.
Deletion of wzz in a Y. pseudotuberculosis O:2a strain
To confirm that the 3′ truncation in pPR1197 results in complete disruption of Wzz activity, the wzz gene was deleted from the wild-type Y. pseudotuberculosis O:2a strain by a modified method of allelic replacement (Derbise et al., 2003) to generate strain M2890. SDS-PAGE analysis shows that the polymer lengths of SΦ874 with pPR981 (Yps O:2a wzz+) and also pPR1197 (Yps O:2a wzz−) are comparable with those of the Y. pseudotuberculosis O:2a wild-type and wzz-mutant strains, respectively (Fig. 5).
Goldman & Hunt (1990) were the first to model O-antigen chain-length variation and used the premise that ligation of an O-antigen polymer to lipid A-core and the extension of that polymer by the addition of another O-unit are both irreversible events. They pointed out that if the probabilities of these alternative events were not influenced by chain length, then the frequency of LPS molecules of various lengths would decrease steadily as chain length increased. However, what was observed when S. enterica LT2 LPS was resolved by SDS-PAGE into bands based on the number of O units was a distribution with 2 peaks. The first peak was of LPS with single O-unit chains, followed by decreasing frequencies as chain lengths increased to 4 or 5 O units, before a rise to the second peak at 31–33 O units. They also noted that the observed distributions could be obtained only if the polymerization rate was affected by chain length with ligation being constant. The discovery of chain-length determination by Wzz in E. coli O75 and O111, S. enterica LT2 and Shigellla flexneri O2a (Bastin et al., 1991; Batchelor et al., 1991; Brown et al., 1991; Macpherson et al., 1991) clarified the situation, as in situations where Wzz was absent, the frequency decreased steadily as chain length increased. This was as Goldman & Hunt (1990) had predicted if neither rate was affected by chain length, showing that the deviation from the predicted distribution was due to Wzz. Bastin et al. (1993) undertook a quantitative study and showed that in a wzz mutant, the proportion of O units ligated was 0.065 for lengths between 4 and 37 O units, but was higher for 1–3 O units. Bands were not resolvable after 37 O units, but much longer chains were present. In effect, they had determined the relative rates of ligation and extension to be 6.5 : 93.5 when Wzy and WaaL are in competition without the involvement of other known factors.
In this article, we compare the Wzy polymerase activities of Y. pseudotuberculosis O:2a and E. coli O111 in the absence of Wzz, and find that they are very different. Only silver staining was used, so the precision of radioactive labelling, as in the example of S. enterica LT2, is not available. However, the rate of O-unit addition to extending polymers in the Y. pseudotuberculosis wzz mutant clearly decreases very rapidly as chain lengths increase, whereby the longest visible polymer is 9 O units. When Wzz function is restored, O-unit polymerization is substantially increased, demonstrating the requirement for Wzz to drive O-unit polymerization in Y. pseudotuberculosis O:2a. The E. coli O111 strain used for comparison has similar patterns to those observed for S. enterica LT2 in the presence and absence of Wzz (Bastin et al., 1993; Fig. 3).
When Wzz is absent in S. enterica LT2, 6.5% of O units are ligated regardless of chain length, but in the presence of Wzz that rapidly decreases to about 1.5%, only to rise to 40% after the switch to favouring ligation (Bastin et al., 1993). A similar pattern is observed with E. coli O111, but for O:2a, the rate of ligation in the absence of Wzz is well over 50%, and no chains reach the modal lengths of the wild type. In this case, Wzz plays no role at all in preventing formation of O-antigen chains longer than the range of the modal chain length, whereas this is very clear for O111 (compare lanes 4 and 5 in Fig. 2).
This study is the first to accurately characterize a wzz mutation in Y. pseudotuberculosis, revealing an unusually low default level of Wzy polymerization. This atypical characteristic provides new insight into the function and relationship between OPS processing proteins in Y. pseudotuberculosis O:2a, which differs from those established for other studied species. Elucidation of the biochemical mechanisms of OPS processing enzymes is needed to shed further light on this effect.
Wzz and related proteins are in the protein superfamily known as polysaccharide copolymerases (PCPs), with the Wzz of O antigens being in the PCP1a subgroup (Morona et al., 2009). The term polysaccharide copolymerase is particularly appropriate for the Y. pseudotuberculosis O:2a Wzz protein, as the Wzy polymerase is relatively inefficient in the absence of Wzz.
JK was supported by an Australian Postgraduate Award (APA).