Introduction

Pseudomonas aeruginosa is a ubiquitous Gram-negative bacterium and an opportunistic pathogen that can infect compromised individuals suffering from various conditions including cancer, AIDS, severe burn wounds, post-operative wounds and cystic fibrosis1. Lipopolysaccharide (LPS) in the outer membrane of the bacterium is important for its persistence in the environment as well as its virulence in infection settings, affecting many bacterial and host phenotypes2.

LPS is anchored in the outer membrane by lipid A, which is linked through the core oligosaccharide to the O-antigen (O-Ag) capping motif2. P. aeruginosa PAO1 simultaneously synthesizes two forms of O-Ag. The first is a common polysaccharide antigen (CPA, A-band) composed of repeating D-rhamnose sugars. The second is composed of a repeating trisaccharide unit containing a proximal D-fucosamine sugar and two distal substituted dideoxy-mannuronic acid sugars termed O-specific antigen (OSA, B-band). B-band O-Ag is the immunodominant cell-surface antigen of P. aeruginosa, with differences in sugar subunit composition and/or linkage stereospecificity leading to the identification of 20 distinct immunological serotypes of the bacterium2.

B-band O-Ag biosynthesis is mediated by the Wzx/Wzy-dependent pathway3 (Supplementary Fig. S1), requiring the activities of a series of integral inner membrane (IM) proteins4,5 that have homologues involved in the biosynthesis of other cell-surface and secreted bacterial polysaccharides including enterobacterial common antigen (ECA), capsule polysaccharide (CPS) and exopolysaccharide (EPS)6. In P. aeruginosa PAO1, once the anionic trisaccharide B-band O-Ag repeat units have been synthesized in the cytoplasm on the lipid carrier undecaprenyl pyrophosphate (UndPP), they are translocated across the hydrophobic IM through the cationic lumen of Wzx to the periplasmic leaflet7,8,9 via a putative antiport mechanism10. Transported O-units are subsequently polymerized by Wzy11,12, with new O-units added at the reducing terminus of the growing chain13. The length of O-unit polymerization is regulated by polysaccharide co-polymerase (PCP) proteins14, with distinct modal lengths in P. aeruginosa PAO1 of 12–16 and 22–30 units (termed “long”) specified by Wzz115 and 40–50 units (termed “very long”) specified by Wzz216. Polymerized O-Ag is then ligated to lipid A-core oligosaccharide by WaaL to form a complete LPS molecule17,18, which is transported to the outer membrane via the Lpt pathway19.

Evidence to elucidate the mechanism of O-Ag polymerization was lacking until a recent investigation in which our group mapped the IM topologies of Wzx, Wzy and WaaL, revealing the presence of two large periplasmic loops (PL3 and PL5) in Wzy of P. aeruginosa PAO1 (WzyPa)4. Despite a high level of sequence homology between them, PL3 and PL5 were found to possess distinct cationic and anionic characteristics, respectively. In addition, RX10G motifs were localized to the middle of each PL and shown to contain functionally-important Arg residues that confer a combination of charge and stereochemistry-specific properties. These observations led us to propose a “catch-and-release” mechanism for WzyPa-mediated O-Ag polymerization (Supplementary Fig. S2), wherein both PL3 and PL5 were in essence binding the same substrate, i.e. an UndPP-linked O unit, via their respective RX10G motifs, but in different capacities. PL3 was suggested to be the recruitment arm for a new UndPP-linked O unit, while PL5 was proposed as the retention arm of the growing chain composed of single O units polymerized upon themselves. For WzyPa, the cationic properties of PL3 would contribute to the binding of an incoming anionic O unit, while the net anionic character of PL5 would be required to facilitate the more transitory interactions required at this site20. Both the identification of two PLs of comparable sequence and the presence of key Arg residues therein would provide support for this mechanism in other Wzy proteins.

Furthermore, the proteins in the Wzx/Wzy-dependent assembly pathway have long been proposed to form an IM complex3; however, only evidence from cross-complementation experiments exists to indirectly support these proposed interactions21. These findings were based on the ability of wzx from an Escherichia coli ECA pathway to complement a wzx deficiency in the O-Ag pathway in the same organism (based on relaxed substrate specificities between the two flippases). However, O-Ag biosynthesis could only be restored by the ECA wzx when the ECA wzy and wzz genes were deleted, suggesting preferred interaction of Wzx with the corresponding Wzy and Wzz from the same pathway21.

Systematic site-directed mutagenesis of 83 periplasmic and cytoplasmic residue positions spanning the length of the protein was performed to identify residues of functional importance for the polymerization of O-Ag, as assayed by the ability of a construct to restore B-band LPS biosynthesis in a wzy chromosomal mutant12. Importantly, the periplasmic amino acids shown to be important for WzyPa function remained confined to PL3 and PL5, with the former also exhibiting positive-charge dependence; moreover, functionally-essential Arg residues in the RX10G motifs of these loops were found to be conserved in putative Wzy proteins identified in a range of phylogenetically distinct bacteria. Unexpectedly, site-specific mutations in cytoplasmic loop (CL) 6 were found to disrupt WzyPa-mediated O-Ag polymerization in the periplasm. B-band O-Ag from cells of a wzy chromosomal knockout mutant complemented with these CL6 mutants displayed atypical chain-length modality skewed towards Wzz2-regulated “very long” polymer lengths (despite the presence of a functional Wzz1). These results suggest a widespread occurrence of a catch-and-release mechanism for Wzy function in bacteria and provide the first evidence for the interaction of full-length Wzy and Wzz in a homologously-expressed system.

Results

PL3 and PL5 are the important periplasmic sites of Wzy

Based on the experimentally-derived topology map of WzyPa4 (Fig. 1), an initial mutagenesis characterization study had been focused on charged residues within the RX10G tracts shared between PL3 and PL520 (Supplementary Fig. S3). In the current investigation, we set out to identify functionally important residues in positions outside of the WzyPa RX10G motifs, including the flanking regions of PL3 and PL5 as well as the other identified periplasmic and cytoplasmic loop domains. In total, 89 site-specific substitutions were introduced across 83 distinct loop-exposed positions in WzyPa, including all aromatic and charged residues within these domains (Fig. 1). The P. aeruginosa PAO1 wzy chromosomal mutant strain12 was supplied in trans with wild-type and mutant gene copies, followed by analysis of the respective B-band O-Ag phenotypes via Western immunoblot to examine the ability of a given mutant construct to complement the assembly deficiency. Despite the sizeable lengths and varied amino acid compositions of PL1 and PL2, no residues of functional importance were identified therein. Moreover, the long stretches of amino acids flanking the RX10G motifs in PL3 and PL5 (Supplementary Fig. S3) were found to lack functionally-important residues. An exception was the anionic residue D286 in PL5, for which a structurally-innocuous Ala mutation (Supplementary Fig. S4) resulted in a WzyPa variant that could only partially restore B-band O-Ag biosynthesis when used in the complementation assay. This functional defect could be fully complemented by the like-charge substitution D286E, resulting in a WzyPa variant that retained WT activity (Fig. 2).

Figure 1
figure 1

WzyPa topology map indicating residue substitution positions and complementation phenotypes.

Highlighted residue colours: orange, no defect in O-Ag polymerization; green, visible defect in O-Ag polymerization. Text in highlighted residues: black, characterized in the current investigation; white, previously characterized20. , residue A199 (site of cationic-to-anionic PL3 charge alterations). RX10G motifs in PL3 and PL5 have been indicated from R175–G186 and R290–G301, respectively.

Full size image
Figure 2
figure 2

Western immunoblot analysis of LPS from P. aeruginosa PAO1 Δwzy complemented with periplasmic domain WzyPa mutants.

B-band O-Ag was detected with monoclonal antibody MF15-459. *, “long” heteropolymeric O-Ag bands regulated by Wzz1; **, “very long” heteropolymeric O-Ag bands regulated by Wzz2.

Full size image

Effects of PL3 charge alterations are consistent with cationic charge-dependence of this loop

Through recruiting the newly-flipped anionic O-Ag subunits polymerized by WzyPa, the overall positive charge of PL3 was proposed to play an important role in the catch-and-release mechanism20 (Supplementary Fig. S2). To further examine the charge dependence of PL3, substitutions were introduced at the base of the loop in order to alter its overall charge properties (Table 1). Position 199 was selected for several reasons. Firstly, it is distant from the previously characterized PL3 RX10G functional site (Fig. 1)20. Secondly, non-native residues introduced at this position were not found to alter the predicted tertiary structure of the PL3 peptide (Supplementary Fig. S5). Thirdly, alteration of any Ala-specific stereochemistry by substitution with Ser did not affect function of the protein (Fig. 2). Amino acid substitutions A199D and A199E that altered PL3 charge from net-positive to net-negative (Table 1) diminished WzyPa activity (Fig. 2) whereas maintenance of net-positive PL3 charge via substitutions A199S and A199K (Table 1) did not compromise WzyPa-mediated polymerization (Fig. 2).

Table 1 Isoelectric point (pI) values for PL3 mutant peptides
Full size table

Putative homologues possess both PL3–PL5-like sequence homology and conserved amino acids essential for WzyPa function

To test the hypothesis that characteristics required for the catch-and-release concept could be found in other bacteria using Wzx/Wzy-dependent assembly systems, we initially performed standard BLAST-based protein database searches, but could not uncover any notable hits using the WzyPa amino acid sequence as a query. Consequently, we employed a jackhmmer search22 to exploit the increased accuracy and ability of probabilistic inference-based HMMER methods to detect homologues23,24 from other bacterial genera; this involves building and searching with a global sequence “profile” rather than relying on position-specific residue scores. This approach yielded 30 sequence hits (Supplementary Table S1) for putative homologues from a variety of bacterial families, classes and phyla (Supplementary Table S2).

Multiple sequence alignment (MSA) output from the jackhmmer search was used to generate sequence logos to visually depict amino acid consensus and frequency at each position25. This analysis revealed that residues R176 (PL3), R290 and R291 (PL5) were highly conserved (Fig. 3). These three Arg residues were previously shown to be functionally-essential for WzyPa and could not be functionally-substituted with Lys20. Together, these data indicated the specific importance of the Arg guanidinium functional group at each of these positions20. Given the conserved positioning of these essential residues, the positions of PL3 and PL5 sequence termini from WzyPa in the MSA were used as constraints to identify analogous loop sequences in predicted homologues. As seen with PL3 and PL5 from WzyPa20 (Supplementary Fig. S3), the analogous sequences for the various predicted homologues displayed sequence homology to each other (Supplementary Fig. S6). These data support the widespread occurrence of traits required for a catch-and-release mechanism20 of O-Ag polymerization in other bacteria.

Figure 3
figure 3

Periplasmic and cytoplasmic loop sequence logos of residue conservation between WzyPa and aligned sequences.

The MSA used to generate the sequence logos was obtained from the results of the jackhmmer search (Table S1). Logo letter height is proportional to its frequency at a given position in the MSA, with the most common letter displayed on top25. Logo residue colours: blue, positively-charged (Arg, Lys, His); red, negatively-charged (Asp, Glu); green, aromatic (Tyr, Phe, Trp); magenta, polar (Gln, Ser, Thr). Amino acid identities and numbers along the x-axes correspond to the native WzyPa sequence, with “(−)” representing a gap in the MSA at specific positions of the WzyPa primary structure. X-axis legend: black, screened for functional importance via site-directed mutagenesis and LPS phenotype complementation; grey, has not been screened for functional importance; *, residue with demonstrated functional defect. Conservation of WzyPa amino acid sequences are displayed for (a) PL1, (b) PL2, (c) PL3, (d) PL5, (e) CL2 and (f) CL6.

Full size image

Cytoplasmic amino acid substitutions affect periplasmic O-Ag modality determination

Extensive amino acid substitutions were generated for the cytoplasmic domains of WzyPa, with particular emphasis on CL2 and CL6 (Fig. 1). Mutations in CL2 did not display any deleterious effects on WzyPa function. In contrast, multiple site-specific substitutions in CL6 were found to affect the B-band O-Ag phenotype (Fig. 4). Mutation N380A and R385A caused decreases in the amount of O-Ag synthesized. More importantly, B-band O-Ag prepared from a P. aeruginosa PAO1 wzy chromosomal mutant complemented with the N380A and R385A constructs showed a change in the preferred modal length of the polymer. In both cases, high-molecular weight LPS bands corresponding to lengths regulated by the Wzz2 PCP continued to be made; however, a slight reduction (R385A) or outright loss (N380A) of low-molecular weight LPS bands attributed to Wzz1-mediated regulation was reproducibly observed (Fig. 4).

Figure 4
figure 4

Western immunoblot analysis of LPS from P. aeruginosa PAO1 Δwzy complemented with cytoplasmic domain WzyPa mutants.

B-band O-Ag was detected with monoclonal antibody MF15-459. O-Ag from the wzz1-wzz2 double mutant specifically complemented with constructs for either wzz1 or wzz2 was analyzed to provide a reference for the respective O-Ag modalities conferred by each protein. The horizontal line has been drawn under the most prevalent Wzz1-specified modality as a reference for shifts in modality caused by mutations in WzyPa. All displayed samples were resolved via SDS-PAGE and Western immunoblotted at the same time. Sample lanes from distinct blots are separated by white space. Samples adjoining the vertical black line were separated by one lane on the same blot. *, “long” heteropolymeric O-Ag bands regulated by Wzz1; **, “very long” heteropolymeric O-Ag bands regulated by Wzz2.

Full size image

To ensure that this shifted-modality phenotype was not simply a result of low gene dosage from the uninduced background expression levels typically used for complementation, the effect of these mutations was compared in the presence and absence of expression induction via addition of 0.1% L-arabinose. For both N380A and R385A, increased expression of the respective WzyPa mutant construct resulted in the production of increased amounts of heteropolymeric O-Ag (Supplementary Fig. S7). The initial detection of a minor O-Ag modality shift with the R385A variant was greatly facilitated by the reduced amount of O-Ag produced in the uninduced samples (Fig. 4); as such, overexpression of the construct and the concomitant synthesis of more B-band O-Ag masked any slight modality shifts. For N380A, the drastic nature of the modality shift was reproducible despite the increased expression levels of the construct and the higher amounts of B-band O-Ag produced. Of particular note was the highest-molecular weight band visible in both the uninduced and induced N380A lanes, which became even more prominent in the latter treatment and was not detected in any of the other samples analyzed (Supplementary Fig. S7). Together, these data indicate an effect of Wzy mutations on the downstream LPS assembly process of chain-length regulation by Wzz proteins.

The novelty of the results from complementation assays with N380A and R385A stems from their cytoplasmic localization in the WzyPa topological map; therefore, we sought to provide additional evidence for the cytoplasmic localization of the two residues. Since we have identified a number of putative WzyPa homologues, we used the MSA data as input for the FILM3 tool26 to generate a de novo tertiary structure model for WzyPa. This approach has recently been demonstrated to utilize correlated rates of amino acid mutations from a MSA to infer residue-residue contacts and successfully generate a representative tertiary structure for a membrane protein sequence of interest10,26,27. Due to the minimal number of input sequences available in our MSA, the lowest-energy WzyPa model generated had an ambiguous TM-score of 0.42 (as >0.5 indicates a fair model with probably correct fold, while <0.4 indicates a poor model with probably incorrect fold). As such, we were unable to confidently report on the overall TMS packing arrangement in the de novo model. However, two important observations were made. Firstly, even with a limited number of input sequences in the MSA, both N380 and R385 were still localized to the same C-terminal CL flanked by helical TMS bundles (Supplementary Fig. S8). Secondly, these data are consistent with the experimentally-derived WzyPa topology map (Fig. 1), providing firm support for the cytoplasmic localization of N380 and R385.

Substitutions N380A and R385A destabilize the C-terminus of WzyPa

To interpret the N380A- and R385A-mediated changes in O-Ag modality, we hypothesized that the conformation and/or stability of the C-terminal region of WzyPa could become disrupted through altered TMS packing caused by the mutations. To explore this further, the potential TMS packing arrangements for wild type (WT) and mutant WzyPa variants were analyzed using TMhit to predict inter-TMS interactions28. As TMhit is a support vector machine classification approach, the over-fitting of data was minimized29. When compared to WT WzyPa, the N380A variant displayed an altered number of total contacts for TMS1, 7, 10, 12 and 14, with reintroduction of a similarly polar side chain (N380Q) able to restore the sums to WT levels (Supplementary Fig. S9). In addition to the change in overall contacts per TMS compared to WT, the N380A mutation also altered the profile for which TMS is predicted to interact with another, a trait that was not observed with the N380Q mutation (Fig. 5). Conversely, changes in inter-TMS interaction profiles (Fig. 5) and the sum of contacts per TMS (Supplementary Fig. S9) were not observed with the R385A substitution, suggesting a more innocuous WzyPa structural alteration caused by substitutions at this position. The notion that mutation R385A caused less TMS perturbation than N380A was supported by (i) the higher amounts of B-band O-Ag produced by the former compared to the latter (Supplementary Fig. S7), as well as (ii) only a slight shift of the LPS bands to longer modal lengths for R385A versus the complete loss of shorter-length polymers corresponding to Wzz1-regulated chain lengths for N380A (Fig. 4). As Wzz2-mediated regulation was unaffected (particularly in the latter substitution), this indicates that Wzz1 and Wzz2 do not engage with WzyPa in an identical manner. Moreover, these data support an interaction between WzyPa and Wzz1 that would be largely mediated by the C-terminal TMS of the former.

Figure 5
figure 5

Inter-TMS contact web predictions for WT and mutant WzyPa.

Predicted contacts were obtained through analysis of WT and mutant WzyPa amino acid sequences with TMhit28, constrained by established TMS boundaries (Fig. 1). The WzyPa TMS have been labelled (1–14). Lines connecting TMS are indicative of at least one predicted pair of interacting residues between the TMS involved. The outputs for WT and substituted WzyPa variants were overlaid. Colour scheme: purple, interactions conserved between WT and mutants; red, interactions exclusive to WT; cyan, interactions exclusive to a mutant. WT comparison to N380A yielded distinct interaction profiles, while comparison to N380Q, R385A, or R385K did not reveal any differences.

Full size image

In addition, we analyzed the propensities of the various WzyPa mutant constructs to insert into the membrane. Constructs encoding WT WzyPa as well as functionally-compromised mutant variants identified in this investigation (A199D, A199E, D286A, N380A, R385A) were first overexpressed with a C-terminal GFP-His8 tag in the P. aeruginosa wzy chromosomal knockout background. Membranes from these samples were isolated and resolved via SDS-PAGE, followed by in-gel fluorescence analysis. Compared to the WT construct, WzyPa variants A199D, A199E and D286A were maintained at equivalent levels in the membrane as detected via C-terminal GFP fusion fluorescence (Fig. 6A), indicating comparable folding and stability of the complete fusion protein. Conversely, mutants N380A and R385A displayed significantly less fluorescence from the encoded C-terminal GFP tag (Fig. 6A).

Figure 6
figure 6

GFP fluorescence data.

(a) Densitometry analysis of monomeric Wzy-GFP-His8 fluorescence from membrane samples isolated from P. aeruginosa PAO1 Δwzy expressing WT and mutant constructs. (b) Densitometry analysis of cleaved GFP-His8 fluorescence from membrane-free lysates of P. aeruginosa PAO1 Δwzy expressing WT and mutant constructs. All samples were analyzed in triplicate (n = 3) and displayed ± standard error. Statistical significance (p ≤ 0.05) was calculated using the Student's t-test. *, statistically significant reduction compared to WT. **, statistically significant increase compared to WT. Panel “a”: WT–N380A (p < 0.0001); WT–R385A (p = 0.0083). Panel “b”: WT–N380A (p = 0.0256); WT–R385A (p = 0.0068).

Full size image

However, despite the reduced amount of GFP fluorescence in the membrane for N380A and R385A, both mutants were still able to polymerize O-Ag subunits (Fig. 4, Supplementary Fig. S7), albeit at lower efficiency compared to the WT, similar to variants A199D, A199E and D286A (Fig. 2). As WzyPa is an integral IM protein that acts on membrane-embedded substrates, this would indicate that the functional “core” of the protein was still membrane embedded for mutants N380A and R385A. Moreover, the drastic shift in preferred O-Ag modality was still maintained even with induced overexpression of the construct (Supplementary Fig. S7). For these reasons, the C-terminal GFP tag may have become more prone to cleavage due to the disrupted TMS packing described above. Consequently, we compared the levels of GFP fluorescence in the clarified lysates from above (lacking the membrane fraction). Lysates from cells expressing the R385A construct displayed significantly more GFP fluorescence than the WT (Fig. 6b), consistent with an intact but more unstable C-terminus prone to cleavage in this mutant construct. Conversely, N380A displayed less GFP fluorescence than the WT (Fig. 6b). We propose that the N380A substitution could result in misfolding of the C-terminus, possibly affecting lateral exit of the remaining domains into the IM from the Sec translocon. This would affect the ability of GFP to properly fold and/or reduce the overall amount of fusion protein being translated. These results for N380A and R385A support the interpretation that the C-terminus (particularly TMS 13 and 14) of WzyPa is disrupted in each of these two mutant variants.

Discussion

Prior to our characterization of WzyPa4, topological mapping of this class of proteins via experimentation had only been carried out on the O-Ag polymerase from Shigella flexneri (WzySf)30, as well as on the putative EPS polymerase PssT from Rhizobium leguminosarum bv. trifolii31. In both of these earlier studies, only one sizeable PL domain had been proposed30,31 for WzySf and PssT, whereas two major PL domains were identified in WzyPa4 (Fig. 1); this major difference was likely a result of the methodologies used in the former two investigations compared to that in the latter32. Consequently, modifications to the PssT topology map have recently been proposed to bring it more in line with that of WzyPa33.

Identification of sequence homology between PL3 and PL5 in WzyPa has led to the discovery of RX10G amino acid tracts in both loops that were shown to contain essential Arg residues20 (Supplementary Fig. S1). The R290 of MagA (proposed CPS polymerase) from Klebsiella pneumoniae was also present in the middle of a large predicted loop and found to be important for function; however, despite the use of three in silico topology prediction algorithms in this study, a consensus number of TMS for MagA was not obtained34 and as such the periplasmic or cytoplasmic localization of this residue cannot be inferred32. As Arg is one of the two most commonly observed amino acids in the binding sites of sugar- and carbohydrate-binding proteins35,36, its importance in PL3 and PL5 supports a role in O-Ag-binding for both these loops within the context of a catch-and-release mechanism of Wzy polymerase function20. Through the use of more sensitive and accurate bioinformatic search techniques that facilitate detection of remote homologues22,23,24, we have now identified conserved Arg residues in a diverse phylogenetic range of bacteria (Fig. 3). Interestingly, these Arg residues are in positions that correspond to each of the functionally-essential R176, R290 and R291 side chains in WzyPa20. Moreover, the putative loops in the homologues (equivalent to PL3 and PL5 in WzyPa) in which these Arg residues reside also share considerable sequence homology with one-another in the same protein (Supplementary Fig. S4). The characteristic properties of these periplasmic loops represent two key requirements of the catch-and-release mechanism: (i) the existence of essential Arg residues in both loops to mediate binding of UndPP-linked sugar repeats and (ii) the presence of two loops displaying comparable amino acid sequences, as one is for recruitment of a single UndPP-linked O-unit, while the other is for retention of the growing chain, which is composed of the same O-unit simply extended upon itself20.

The observation that functionally-essential periplasmic residues are localized almost exclusively to the RX10G motifs in WzyPa (Fig. 1) is in line with the overall lack of sequence conservation in these regions (Fig. 3). By extension, evolutionary retention of the catalytic Arg residues, within the context of organism-specific PL3 and PL5 flanking sequences, is consistent with expected differences in substrate specificity between bacteria conferred by their respective two PL domains; these would have evolved to handle particular chemical structures of organism-specific O-Ag repeat units. At present, a major limitation to substrate specificity characterization in these putative homologues is that the O-Ag compositions and/or structures from the respective bacterial strains have yet to be elucidated. Despite extensive searching, we were only able to find an O-Ag chemical structure37 for one of the strains (Photorhabdus luminescens subsp. laumondii TT01) identified in the list of non-redundant WzyPa sequence hits; unfortunately, no chemical or topological similarities could be found between this particular O-Ag structure and that of the B-band O-Ag of P. aeruginosa PAO1.

To handle the net-negative charge of the P. aeruginosa PAO1 B-band O-Ag subunit, the interior of the respective Wzx flippase has been found to be cationic in nature8. It is logical that the periplasmic WzyPa domain responsible for recruitment of the newly-flipped O-unit would also display this characteristic, for without this initial “catching” step, subsequent polymerization would not occur20. The positive-charge dependence of PL3 that we have revealed in this investigation (Supplementary Fig. S3) further strengthens the proposed role of this periplasmic domain in initial binding of UndPP-linked O-unit by the polymerase. The partial loss of WzyPa function observed for the D286A substitution is consistent with a decrease in net-negative charge affecting the role of PL5 in the proposed catch-and-release mechanism, as it would alter the affinity of the retention arm for the growing O-Ag chain (bound via the PL5 RX10G motif). However, a catalytic role cannot be ruled out for this residue, as Asp at this position was identified in a subset of the aligned putative Wzy homologues. In the other sequence hits, a conserved Ser residue was present at this position. Either residue can be commonly found in the active site of glycosyltransferases and may be involved in proton abstraction from the donor substrate to catalyze the formation of a new glycosidic bond38,39.

The presence of functionally-important amino acids on the cytoplasmic face of the protein was unexpected given the periplasmic nature of O-Ag polymerization and the lack of requirement for cytoplasmic factors such as ATP as an energy source for the protein11. It is for this reason that the identification of cytoplasmic WzyPa residues (in the same loop) that affect O-Ag polymerization is particularly intriguing. Consistent with their importance, both N380 and R385 were found to be conserved amongst the other identified putative Wzy proteins (Fig. 3). The novelty of observing a reduction in the amount of B-band O-Ag produced (for N380A and R385A) was only surpassed by the surprising finding that the remaining O-Ag was primarily composed of higher-molecular-weight polysaccharide chains displaying Wzz2-regulated modal lengths (Fig. 4). Despite the same level of Wzz1 expression as when all other WzyPa mutant constructs were screened, the chain-length regulator had lost its ability to regulate the polymerization of shorter O-Ag chain lengths by WzyPa. This would support different interactions coordinating functions of Wzy and Wzz2 compared to Wzy and Wzz1.

While using the same precursor substrates (UndPP-linked O units) and the same enzyme (Wzy), Wzz1 and Wzz2 display genetic and functional diversity. Typically, wzz1 in P. aeruginosa and its homologues in other bacteria that produce long O-Ag chains are found in the respective O-Ag biosynthesis gene clusters along with their cognate Wzy proteins. Conversely, wzz2 and its homologues that produce very-long O-Ag chains are located elsewhere in the chromosome or on plasmids16,40,41,42. Moreover, long and very-long O-Ag chains in the same bacterium have been shown to be play different roles in infection, virulence and persistence40,43,44,45.

It is important to note that in order for the bacterial cell to elaborate and display any form of polymeric B-band O-Ag on its surface, regardless of chain length, a functional Wzy protein must be present in the inner membrane12. Consistent with this, we detected its reaction products with B-band specific monoclonal antibodies46. As presented in a previous investigation20, mutations that affect function of Wzy may simply reduce the functionality of the protein, resulting in an overall reduction in the amount of O-Ag polymerized. Such is the case for the various A199D and A199E substitutions (Fig. 2), as well as the N380A and R385A variants (Fig. 4) characterized herein. However, in most instances of a partial reduction in WzyPa functionality from specific residue substitutions, decreased enzyme efficiency resulted in an overall shift in bands towards the lower molecular weight Wzz1-regulated bands. The only exceptions identified thus far are the N380A and R385A mutants from the same cytoplasmic loop that, despite producing decreased amounts of O-Ag, resulted in chain modality that was shifted away from Wzz1-regulated lengths (and conversely towards higher lengths regulated by Wzz2). For these very reasons, if the N380A and R385A mutations significantly contributed to reduced steady-state levels of the proteins in the membrane, O-Ag bands from these samples should still have been shifted downwards towards Wzz1-regulated lengths as there would simply not have been enough “functional” Wzy present to sustain polymerization to very long Wzz2-regulated lengths. However, this was not observed. Furthermore, considering that the modality shift away from Wzz1-regulated lengths was still maintained following induction of expression in particular for the N380A samples, this provides further evidence that steady-state levels of these mutant proteins did not play a significant role in the observed shifted-modality phenotypes.

It is more plausible that the effect on B-band O-Ag biosynthesis caused by substitutions N380A and R385A is a result of tertiary structure disruptions in the C-terminal half of WzyPa, as supported by the complementary lines of evidence provided in this investigation (Figs. 3,4,5,6, Supplementary Figs. S7–S9). At a minimum, Wzy and Wzz are believed to be localized within close proximity to one another14. Since the observed effect of these mutations on WzyPa polymerization was to alter the modality conferred by Wzz proteins, this would suggest that a specific interaction between Wzy and Wzz1 was being affected. This is further supported by the observation that the function of Wzz2 was unaffected, as a general disruption in proximity should have affected regulation by both Wzz1 and Wzz2. Additional evidence is provided by the in vitro reconstitution of O-Ag polymerization and chain length regulation using purified Wzy and Wzz, respectively11; as the overall structure of Wzz has been evolutionarily selected to extend vertically into the periplasm (and away from the active site of Wzy)14, it is unlikely that Wzz molecules oriented randomly with respect to Wzy would be able to carry out their native function. As such, the occurrence of a direct interaction between Wzy and Wzz is the likeliest scenario. However, further work is warranted to confirm direct physical interactions between these two proteins.

While a N-terminally-truncated Wzy was still able to polymerize O-Ag to Wzz-specified modal lengths11, C-terminal polymerase truncations were found to affect PCP-regulated EPS lengths31. Such C-terminal truncations were found to affect proposed polymerase–PCP interactions arising from exogenous construct overexpression in a bacterial two-hybrid screen33, though co-immunoprecipitation of Wzy and Wzz could not be achieved when the target constructs were expressed at chromosomal levels in a native host47.

Recently, the TMS of Wzz were shown by our group to play a role in the mechanism of seroconversion of P. aeruginosa PAO1 from serotype O5 to serotype O16. Previously, the inhibitor of α-polymerase (Iap) peptide (encoded by the lysogenic D3 bacteriophage) had been found to abrogate synthesis of α-linked B-band O-Ag by WzyPa, while allowing a phage-encoded Wzy to polymerize the same substrate O-Ag subunits with a β-linkage48,49. The propensity of Iap to insert into the IM via a single TMS was subsequently demonstrated, with its inhibitory effect on WzyPa shown to be dependent on the presence/absence of Wzz1 and Wzz2. The amino acid sequence of the hydrophobic Iap peptide displayed sequence homology with a region in the first TMS of both Wzz proteins, suggesting that Iap mediates shutdown of WzyPa function through mimicry of Wzz TMS. This was further supported by the observation that Wzz overexpression was able to overcome the inhibitory effects of Iap on WzyPa50.

Though high-resolution crystallographic data is not yet available for Wzy proteins, interactions of the polymerase with associated Wzz oligomers are likely mediated by a combination of exposed and membrane-embedded domains between the two proteins. While detailed 3D structural data from X-ray crystallography has been obtained for the periplasmic domains of Wzz proteins51,52, the steric properties conferred by the two TMS of Wzz remain poorly understood. Cryo-electron microscopic analysis of 2D crystals of full-length PCP proteins suggests similar quaternary structures53,54; however, the role of TMS in heterotypic interactions of these PCPs remains unknown. As such, TMS domains remain an important avenue of future investigation towards better understanding the interplay between Wzy and Wzz. Together, the added depth of understanding regarding Wzy function we have uncovered as well as the support for its interaction with Wzz identified in this investigation have served to further elucidate the underpinnings of the elongation and length-regulation steps of the Wzx/Wzy-dependent pathway.

Methods

DNA manipulations

QuikChange (Agilent) site-directed mutagenesis was carried out on the previously-created pHERD26T-wzy-GFP plasmid; this fusion encodes WzyPa fused to a C-terminal GFP-His8 moiety that does not affect polymerase function4,20. The wzz1 and wzz2 genes were PCR amplified from P. aeruginosa PAO1 genomic DNA and cloned in-frame with the downstream mCherry coding sequence in the pRTL2 vector between the EcoRI and NheI restriction sites to create pRTL2-Wzz1-mCherry and pRTL2-Wzz2-mCherry. Fusion gene sequences were PCR amplified and subcloned in the arabinose-inducible pHERD28T backbone55 between the EcoRI and XbaI restriction sites to create pHERD28T-Wzz1-mCherry and pHERD28T-Wzz2-mCherry. Plasmids were maintained in cells through supplementation of lysogeny broth and agar plates with tetracycline (pHERD26T: 15 μg/mL for Escherichia coli, 90 μg/mL for P. aeruginosa) or trimethoprim (pHERD28T: 100 μg/mL for E. coli, 250 μg/mL for P. aeruginosa). Plasmid DNA was isolated with a GenElute (Sigma) or a QIAprep Spin (Qiagen) miniprep kit. Mutagenesis PCR reactions were carried out using KOD Hot Start DNA polymerase (Novagen). The sequences of oligonucleotide primers containing mutagenesis mismatches (Sigma) have been provided in the Supplementary Information (Supplementary Table S3).

Complementation analysis

Each mutant construct was assayed for its ability to complement a chromosomal deficiency and restore B-band O-Ag biosynthesis in a P. aeruginosa PAO1 wzy::GmR insertional mutant12. To reflect the intrinsically-low expression levels of Wzy12,56,57, cells expressing the mutant constructs were grown overnight with shaking at 37°C to stationary phase without expression induction20. Wzz1/2 constructs were expressed overnight in a P. aeruginosa PAO1 wzz1::GmR/Δwzz2 double-mutant background at 37°C16. Background expression from pHERD28T-Wzz1-mCherry was sufficient to restore Wzz1-specified modal lengths, while overnight induction with 0.1% L-arabinose was required for pHERD28T-Wzz2-mCherry-mediated restoration of Wzz2-specific B-band O-Ag modality.

LPS was prepared from complemented P. aeruginosa PAO1 mutant strains via a method modified from the Hitchcock and Brown protocol58. In brief, overnight cultures were equilibrated to 1 mL at an OD600 of 0.45. Cells were sedimented in a microfuge (12 000 × g, 1 min), followed by aspiration of the supernatant. Each pellet was washed in 1 mL of 0.9% NaCl, followed by sedimentation as described above. Washed pellets were resuspended in 250 μL lysis buffer (2% SDS, 4% β-mercaptoethanol, 10% glycerol, 1 M Tris HCl [pH 6.8], 0.002% bromophenol blue) and boiled for 30 min. Each sample was treated with Proteinase K (120 μg/mL) and incubated at 55°C overnight. Samples (5 μL) were resolved via SDS-PAGE (150 V, 1 h) and transferred to a nitrocellulose membrane via Western blotting (180 mA, 45 min). Blots were blocked (1 h) with 3% bovine serum albumin (BSA) in phosphate-buffered saline (PBS), followed by washing with PBS (5 min). Membranes were incubated overnight with anti-B-band O-Ag mouse monoclonal antibody MF15-459, followed by three washes in PBS. Blots were incubated (1 h) with goat anti-mouse secondary antibody conjugated to alkaline phosphatase (Jackson Immunoresearch) (1:2000, 3% BSA in PBS), followed by three washes in PBS. Blots were developed in Buffer A (100 mM Tris-HCl [pH 9.5], 100 mM NaCl, 5 mM MgCl2) supplemented with 0.0165% 5-bromo-4-chloro-3-indoyl phosphate (BCIP) and 0.033% nitro blue tetrazolium (NBT). All steps were carried out at room temperature.

Fluorescence analyses

The quantitation of membrane localization for different Wzy-GFP-His8 mutant constructs was carried out as previously detailed20. In brief, overnight cultures of P. aeruginosa PAO1 wzy::GmR induced for construct expression were sedimented via centrifugation (12 000 × g, 30 min), followed by lysis using BPER-II protein extraction reagent (Thermo) with 2× Complete protease inhibitor lacking EDTA (Roche). Crude debris and inclusion bodies were removed via sedimentation (12 000 × g, 30 min). This supernatant was sedimented in an ultracentrifuge to isolate the membrane fraction (120 000 × g, 1.25 h), followed by resuspension of the membrane pellet directly in Laemmli sample buffer. Samples were heated at 37°C (30 min) and resolved on a 12% SDS-PAGE gel. Resolved gels were subsequently scanned for in-gel GFP fluorescence using a Typhoon 9410 imager (GE Healthcare). Densitometry analysis of full-length Wzy-GFP-His8 in-gel fluorescence was performed using ImageJ software on three biological replicates. All statistical analyses were performed using GraphPad Prism software.

To measure the fluorescence of GFP released from lysed cells (i.e. in the soluble fraction), the clarified supernatant (5 mL) following sedimentation in the ultracentrifuge (described above) was obtained. Clarified supernatant was mixed at a 1:1 ratio with 2× Laemmli sample buffer and incubated at 37°C (30 min). Samples (20 μL) were resolved on a 12% SDS-PAGE gel. Resolved gels were scanned as described above using a Typhoon imager. Densitometry analysis of cleaved GFP-His8 in-gel fluorescence was performed as described above.

Bioinformatic analyses

Detection of remote homologues was accomplished by performing a jackhmmer23 search against the UNIREF100 data bank60 using three iterations with an E-value inclusion threshold of 10−6, resulting in a final MSA of 30 aligned sequences. Based on the MSA, amino acid sequence logos were generated using WebLogo (http://weblogo.berkeley.edu/logo.cgi)61. TMhit analysis (L/2 threshold) was performed online (http://bio-cluster.iis.sinica.edu.tw/TMhit/)28, with topological restraints introduced based on the published TMS boundaries for the WzyPa topology map4. Isoelectric point values for WT and mutant loop sequences were characterized via the ExPASy portal.