Bacterial cell surface polysaccharides confer resistance to external stress and promote survival in biotic and abiotic environments. Glycan assembly often occurs at the periplasmic leaflet of the inner membrane (IM) from undecaprenyl pyrophosphate (UndPP)-linked polysaccharide units via the Wzx/Wzy-dependent pathway. Wzx is an integral IM protein found in Gram-negative and Gram-positive bacteria that mediates IM translocation of UndPP-linked sugar repeats from the cytoplasmic to the periplasmic leaflet; interaction of Wzx with other assembly proteins is indirectly supported by genetic evidence. Topological mapping has indicated 12 α-helical transmembrane segments (TMS), with the number of charged TMS residues fluctuating based on the mapping method used. A novel Wzx tertiary structure model has been built, allowing for substrate-binding or energy-coupling roles to be proposed for functionally important charged and aromatic TMS residues. It has also led to a proposed antiport-like mechanism of Wzx function. Exquisite substrate specificity of Wzx proteins was recently revealed in distinguishing between UndPP-linked substrates with identical main-chain sugar repeats, but differing in the chemical composition of a terminal sugar side-branch cap. The objective of this review is to synthesize the most up-to-date knowledge concerning Wzx flippases and to provide perspective for future investigations in this burgeoning field.
Cell surface polysaccharides play an essential role in the ability of bacteria to survive and persist in the environment and in host interaction settings. These polysaccharides are pivotal for the modulation of diverse microbial phenotypes ranging from biofilm formation and motility, to the evasion of host defence mechanisms such as complement deposition and phagocytic killing (Bazaka et al., 2011; Lam et al., 2011).
Three distinct pathways are responsible for synthesis of the majority of these glycans: (i) Wzx (flippase)/Wzy (polymerase)-dependent (Whitfield, 2006), (ii) ABC transporter-dependent (Greenfield and Whitfield, 2012) and (iii) synthase-dependent (Keenleyside and Whitfield, 1996); however, as the latter two pathways are not the focus of this review, they will not be discussed further. The Wzx/Wzy-dependent assembly pathway is found in a wide range of Gram-negative and Gram-positive bacteria as it is responsible for the synthesis of numerous cell surface sugar polymers, including lipopolysaccharide (LPS) heteropolymeric O antigen (O-Ag), enterobacterial common antigen (ECA), exopolysaccharide (EPS) and capsular polysaccharides (Whitfield, 2006).
Wzx/Wzy-dependent polysaccharide assembly
Assembly via the Wzx/Wzy-dependent pathway begins in the cytoplasm with the synthesis of lipid-linked polysaccharide repeat units. The first sugar is added by an initiating glycosyltransferase to the polyisoprenoid lipid carrier molecule undecaprenyl phosphate (UndP, C55P) in the form of an activated nucleotide precursor at the inner leaflet of the inner membrane (IM), resulting in a pyrophosphate linkage with the carrier (undecaprenyl pyrophosphate, UndPP) (Price and Momany, 2005). Subsequent sugar additions by specific glycosyltransferases result in the synthesis of an individual UndPP-linked polysaccharide repeat unit (referred to as an ‘O unit’ for simplicity, regardless of the final product), after which a series of integral IM assembly proteins take over (Fig. 1). The UndPP-linked repeat units are translocated from the inner to the outer leaflet of the IM by the flippase Wzx (Fig. 2) (Liu et al., 1996; Burrows and Lam, 1999; Feldman et al., 1999; Marolda et al., 2004), where they are polymerized by Wzy (de Kievit et al., 1995; Woodward et al., 2010) via a putative catch-and-release mechanism (Islam et al., 2011). O-unit addition occurs at the reducing terminus of the growing chain (Robbins et al., 1967), the length of which is regulated by the polysaccharide co-polymerase Wzz, resulting in organism-specific preferred modal lengths (Bastin et al., 1993; Morona et al., 1995; Daniels et al., 2002; Woodward et al., 2010). This polymerized glycan is then anchored to lipid A-core oligosaccharide by WaaL in the case of O-Ag to form a mature LPS molecule (Abeyrathne et al., 2005; Abeyrathne and Lam, 2007a; Hug et al., 2010; Han et al., 2011; Ruan et al., 2011).
The proteins that participate in the Wzx/Wzy-dependent pathway have long been proposed to function as part of an IM complex (Fig. 1) (Whitfield, 1995). This hypothesis is indirectly supported by persuasive genetic evidence obtained for Escherichia coli K-12, which possesses separate wzx, wzy and wzz genes for the production of serotype O16 O-Ag and ECA (Marolda et al., 2006). As the O16 O-Ag and ECA biosynthesis pathways in E. coli K-12 both utilize GlcNAc as the proximal UndPP-linked sugar in their respective repeats, previous data (described below) suggested that the O16 and ECA Wzx proteins should be able to functionally substitute for each other (Feldman et al., 1999; Marolda et al., 2004). However, the ECA wzx was only able to fully complement an O16 wzx deficiency when the majority of the ECA gene cluster was deleted, resulting in removal of ECA wzy and wzz. In turn, provision of ECA wzy or wzz in trans reduced the ability of ECA wzx to substitute for O16 wzx. These are intriguing observations, and they suggest that the proteins encoded by the wzx, wzy and wzz genes of each respective synthesis pathway may preferentially interact with each other (Marolda et al., 2006). To date, no biochemical evidence has been obtained to demonstrate interactions between the various constituents of the same Wzx/Wzy-dependent pathway. However, several recent reports in the literature are changing the landscape of our understanding concerning the structure and function of Wzx/Wzy-dependent pathway proteins, particularly with respect to Wzx. Therefore, this review will summarize the established knowledge gleaned from past investigations as well as highlight recent progress made by various groups concerning the characterization of this ubiquitous flippase.
Identification of Wzx as an O-unit flippase
The requirement for O-unit translocation across the IM was first observed by McGrath and Osborn (1991) as the processes of O-unit polymerization and ligation to lipid A-core oligosaccharide were localized to the periplasm, while initial O-unit synthesis occurred in the cytoplasm (McGrath and Osborn, 1991). The first evidence to support the role of Wzx (formerly RfbX) as an O-unit flippase capable of mediating this transbilayer O-unit transport was obtained by Liu and colleagues (1996) in a Salmonella enterica LT2 background deleted for its O-Ag synthesis cluster and supplemented in trans with that of E. coli Dysenteriae 1. By monitoring the uptake and incorporation of [14C]galactose as part of the E. coli Dysenteriae 1 tetrasaccharide O unit, a wzx::Tn mutant was found to be deficient in O-Ag production and to accumulate UndPP-linked radiolabelled O units (Liu et al., 1996). To determine the location of the accumulated O units in the wzx mutant, spheroplasts and everted membrane vesicles were created to analyse the periplasmic and cytoplasmic faces of the IM respectively. These membrane vesicles were examined via ELISA for the detection of O-unit epitopes using polyclonal antiserum against E. coli Dysenteriae 1. Higher quantities of O-unit epitopes were detected in the everted membrane vesicles for the wzx mutant as compared with the membranes prepared from the relevant control strains (Liu et al., 1996). In a separate study by our group, working with Pseudomonas aeruginosa PAO1, Burrows and Lam (1999) created the first targeted knockout mutants of wzx via insertion of a non-polar gentamicin resistance cassette in trans, followed by allelic replacement of the chromosomal copy with the wzx::Gm construct. This wzx mutant lacked heteropolymeric O-Ag (Burrows and Lam, 1999) (Fig. 2), an observation that is consistent with the results described by Liu and colleagues (1996). Targeted and random mutagenesis studies in other bacteria have led to the identification of wzx in many bacteria (Klena and Schnaitman, 1993; Macpherson et al., 1995; Marolda et al., 1999; 2004; 2006; Rick et al., 2003; Tabei et al., 2009; Hong et al., 2012; Wang et al., 2012), while annotation based on sequence similarity has taken place for countless others.
Additional evidence for Wzx-mediated lipid-linked glycan transport was provided by Rick and colleagues through the examination of the wzxE gene required for the synthesis of ECA (Rick et al., 2003), which is built from Fuc4NAc-ManNAcA-GlcNAc-P-P-Und (i.e. lipid III) precursors. These authors observed that uptake of a water-soluble nerol pyrophosphate (soluble UndPP analogue with a shorter fatty-acid moiety)-linked [3H]GlcNAc residue into everted membrane vesicles from E. coli K-12 was abrogated for a wzxE mutant, suggesting that functional WzxE was mediating translocation of the nerol-linked sugar moiety (Rick et al., 2003).
Phylogeny of Wzx flippases
Wzx proteins have been identified in Gram-negative and Gram-positive bacteria, as well as in archaea, and are classified as part of the polysaccharide transporter (PST) family of proteins. Along with the multidrug and toxin extrusion (MATE), oligosaccharidyl-lipid flippase (OLF) and mouse virulence factor (MVF) families, these four families constitute the multidrug/oligosaccharidyl-lipid/polysaccharide (MOP) exporter superfamily (Fig. S1) (Hvorup et al., 2003). Within the MOP superfamily, phylogenetic comparison between the constituent families indicates that the MATE, OLF and MVF families are more closely related to the PST family than to each other, indicating that the PST family is the main link within the superfamily (Fig. S1) and that all four families evolved from a single origin (Saier, 1994). In turn, it is likely that PST family proteins represent the most closely related descendents of the ancestral homologues that gave rise to the MOP exporter superfamily (Hvorup et al., 2003).
Intriguingly, despite the widespread occurrence of Wzx proteins in bacteria and archaea, the overall sequence identity between orthologues in different microorganisms is quite low; this characteristic is also maintained between different serotypes of the same species. Investigators have taken advantage of this trait to design molecular serotyping schemes involving PCR and sequencing protocols of respective wzx genes to rapidly identify and differentiate between strains of species such as Shigella flexneri (Sun et al., 2011), Cronobacter spp. (Jarvis et al., 2011), P. aeruginosa (Raymond et al., 2002), S. enterica (Muñoz et al., 2010), Streptococcus pneumoniae (Kong et al., 2005) and Shiga toxin-producing E. coli (DebRoy et al., 2011).
Topological prediction algorithms used for analysis of Wzx proteins consistently predict the presence of 12–14 transmembrane segments (TMS), emphasizing their integral IM nature (Hvorup et al., 2003); as such, overexpression and purification of these membrane protein constructs for downstream characterization is difficult (Abeyrathne and Lam, 2007b). While the use of topological prediction algorithms is sufficient in almost 70% of instances to correctly predict the overall number of putative TMS, as well as the orientation of the protein in the IM, key positioning characteristics and domain details can be overlooked if these predictions are not substantiated by experimental data (Elofsson and Heijne, 2007). Consequently, topological mapping of integral IM proteins is crucial for elucidating the position and extent of exposed and membrane-embedded domains.
To date, only four studies have reported the development of a topological map of Wzx based on experimental data. These include the topological characterizations of Wzx proteins from P. aeruginosa PAO1 (WzxPa) (Islam et al., 2010), S. enterica sv. typhimurium group B (WzxSeB) (Cunneen and Reeves, 2008), E. coli O157:H7 (WzxEc) (Marolda et al., 2010; 2011) and Rhizobium leguminosarum bv. trifolii TA1 (PssL) (Mazur et al., 2005) (Fig. 3). The latter three reports followed a strategy whereby a consensus output from five different in silico topology prediction algorithms was first used to assign the location of putative TMS in each protein (Nilsson et al., 2002). The authors then prepared individual C-terminal PhoA and LacZ/GFP fusions to validate the proposed algorithm-based model. As these reporter proteins are only functional in specific subcellular compartments in the Gram-negative cell envelope, namely the periplasm (PhoA) (Manoil et al., 1990) or cytoplasm (LacZ, GFP) (Matthews, 2005; Aronson et al., 2011), analysis of their respective enzyme activities or fluorescence allowed for the localization of various periplasmic and cytoplasmic loop domains (Mazur et al., 2005; Cunneen and Reeves, 2008; Marolda et al., 2010). For characterizing WzxEc, Marolda and colleagues (2010) performed additional experiments in which a select number of amino acid residues from their initial topology model were also analysed via the substituted-cysteine accessibility method (SCAM) to determine their exposure to solvent (Bogdanov et al., 2005).
The topology of WzxPa was mapped using a different approach. Random and interval-scanning libraries were created for WzxPa truncations fused C-terminally with a unique PhoALacZα dual reporter capable of exhibiting either PhoA activity or LacZ activity (via α-complementation) depending on its localization; as such, periplasmic, cytoplasmic or TMS localization could be detected based on the ratio of PhoA:LacZ activities for a given truncation fusion (Alexeyev and Winkler, 1999). Numerous positions within WzxPa were screened by this method, with the locations of TMS directly constrained by in vitro data (Islam et al., 2010).
Topological consensus and differences
Each of the four Wzx topology maps reports 12 TMS, with both the N and the C termini of the proteins in the cytoplasm (Fig. 3). A large cytoplasmic loop between TMS6 and TMS7, separating the two halves of the protein, was proposed for WzxPa (Islam et al., 2010) and PssL (Mazur et al., 2005), but not for WzxEc (Marolda et al., 2011) or WzxSeB (Cunneen and Reeves, 2008), reflecting predicted topological differences in PST family proteins (Paulsen et al., 1997). One important distinction between the topology map of WzxPa and those of WzxSeB, WzxEc and PssL is the number of charged residues (Asp, Glu, His, Lys, Arg) observed within the TMS regions; 17 were shown for WzxPa (Islam et al., 2010), while 6, 4 and 6 were indicated for WzxSeB (Cunneen and Reeves, 2008), WzxEc (Marolda et al., 2011) and PssL (Mazur et al., 2005) respectively (Fig. 3). This difference is likely twofold in nature. The use of consensus in silico prediction algorithms (generally weighted towards hydrophobicity) to assign TMS positions may have artificially omitted charged residues from the membrane-spanning regions. This observation is consistent with the borders of the TMS for WzxSe, WzxEc and PssL, which appear to be directly constrained and delimited by the locations of charged residues (Fig. 3); as WzxPa was not subject to this initial in silico limitation, the locations of its charged amino acids were based solely on biochemical enzyme reporter data (Islam et al., 2010), which revealed an increased number of charged TMS residues in the WzxPa topology map (Fig. 3). Additionally, as discussed below, Wzx proteins have been shown to exhibit exquisite substrate specificity, and as such the compositions of their TMS should reflect the chemically diverse range of strain-specific O-unit substrates identified in various organisms.
Structural homology modelling
The use of existing X-ray crystal structures for homology modelling is a powerful tool to obtain 3D structural data for related proteins for which a biophysical structure has yet to be solved (Ravna and Sylte, 2012); this is particularly relevant for membrane proteins, for which there is a dearth of available structures despite their frequent occurrence in nature (White, 2009). The X-ray crystal structure of the MATE family protein NorM from Vibrio cholerae O1 El Tor (NorMVc, PDB ID: 3MKT) was recently determined by He and colleagues (2010), representing the first published high-resolution structure of a MOP exporter superfamily protein. Consistent with other MATE family members (Omote et al., 2006), NorM is an integral IM protein that functions to remove drugs and other toxic compounds from the cytoplasm of Gram-negative bacteria (Morita et al., 2000; Braibant et al., 2002; Burse et al., 2004). Recently, Islam and colleagues were able to take advantage of the sequence homology, hydrophobicity profile similarity, phylogenetic relatedness and fold recognition suitability of NorMVc and WzxPa to construct a 3D structural homology model of the latter (Fig. 4). This Wzx protein model was subjected to genetic, biochemical and bioinformatic validation methods, yielding the first tertiary structural context for a Wzx protein from which functional insights could be gleaned (Islam et al., 2012).
The WzxPa structure model reveals 12 TMS, with the protein displaying twofold rotational symmetry (Fig. 4) (Islam et al., 2012), consistent with the NorMVc template (He et al., 2010); this also results in the current conformation of the WzxPa structure reflecting the closed apo-form state of the NorMVc structure in which the protein has not undergone conformational transition to become suitable for substrate binding on its cytoplasmic face. The majority of the protein consists of α-helical TMS, with only minimal periplasmic and cytoplasmic loop regions. The only exception is the large cytoplasmic loop connecting TMS6 and TMS7; this appears to be mainly a disordered peptide linking the two halves of WzxPa. Additionally, two portals opening into the periplasmic leaflet of the IM were identified, flanked by TMS1 and TMS8 in the front, and by TMS2 and TMS7 in the rear.
Islam and colleagues (2012) were able to measure the interior volume of the WzxPa structure model using the molecular structure analysis tool HOLLOW to fill any interior cavities within the protein with dummy atoms (Ho and Gruswitz, 2008). Through this exercise, a lumenal space was predicted. Substantial positive charge in the lumen was observed when this region of WzxPa was displayed by overlaying the electrostatic potential contributed by the side chains lining the lumen wall (Fig. 5). This is noteworthy, as WzxPa mediates translocation (Burrows and Lam, 1999) of a negatively charged trisaccharide O-unit substrate containing a proximal UndPP-linked d-fucosamine moiety followed by two dideoxy-mannuronic acid sugars that are N- or N-acetyl-substituted (Knirel et al., 2006). Furthermore, the front periplasmic portal of WzxPa was also found to be lined with positive charge (Figs 5 and 6), thus providing a likely site of lateral exit into the periplasmic leaflet of the IM for the flipped UndPP-linked trisaccharide in P. aeruginosa PAO1.
Critical residues in Wzx
Site-directed mutagenesis is the gold standard for probing the functional importance of various amino acid residues within the context of the translated protein product; this is accomplished through the use of a mutant gene construct to complement a deficiency in the native gene copy and restore wild-type phenotype. In respective wzx chromosomal mutants, this approach revealed the functional importance of several charged amino acid residues in WzxPa (Arg59, Glu61, Arg146, Asp269, Lys272 and Asp359) (Islam et al., 2012) and WzxEc (Asp85, Arg298, Asp326 and Lys419) (Marolda et al., 2010), as well as several aromatic residues in the former (Tyr60 and Phe139); each of these residues could be complemented by the restoration of similar charge or aromatic character.
Tyr60 and Phe139 are internally located within the WzxPa structure (Fig. 5), which may indicate a role in substrate binding; this would be consistent with the finding that aromatic residues are commonly located in the binding pockets of structures for carbohydrate- and sugar-binding proteins (Malik and Ahmad, 2007; Elumalai et al., 2010). Residues Arg59, Lys272, Glu61 and Asp269 are also internally located (Fig. 5), with the former two cationic residues positioned to possibly interact with the anionic substrate, and the latter two anionic residues potentially taking part in charge repulsion events during conformational changes to ‘push’ the substrate through the lumen. Alternatively, these anionic residues may be involved in energy transduction (discussed below). It is interesting to note that each of the substitutions made to date resulting in compromised WzxPa function map to the cytoplasmic half of the protein; this may indicate that the essential events for O-unit flipping take place during the initial substrate loading and binding that would occur at the cytoplasmic face of the IM, while exit from the flippase in the periplasmic leaflet of the IM is a non-specific event (Islam et al., 2012).
Pairwise amino acid sequence comparison between WzxEc and WzxPa does not indicate direct alignment between the various residues of functional importance (Fig. S2). However, when compared qualitatively within the context of the WzxEc topology map (Marolda et al., 2011), Asp85 in WzxEc may be analogous to Glu61 in WzxPa, as both are localized towards the C-terminus of TMS2 in the respective proteins. Additionally, Asp326 in WzxEc is located in a similar overall region to Asp269 in WzxPa; the latter is localized near the base of TMS8 in the WzxPa structure while the former has been assigned to a cytoplasmic loop, seven residues downstream of TMS8 in the WzxEc topology map. Neither Arg298 nor Lys419 in WzxEc (predicted to be periplasmic and transmembrane respectively) corresponds to any other functionally important residues in WzxPa. Both Asp85 and Lys419 are located outside of the middle region of WzxEc that was topologically mapped through in vitro characterization (Marolda et al., 2010; 2011).
Flippase substrate specificity
Role of the UndPP-linked sugar
The structures of O units linked to UndPP are very diverse and directly contribute to the countless antigenic variations presented on the cell surfaces of different bacteria (Knirel, 2011). This is consistent with the high amino acid sequence variability for Wzx in different systems. The first set of data from examining the range of possible substrates translocated by Wzx was provided by Feldman and colleagues (1999) through the use of E. coli K-12 mutants deficient in their abilities to produce the complete pentasaccharide O unit corresponding to the O16 serotype of this species. These authors probed Western blots using a lectin with high specificity for terminal GlcNAc moieties, revealing that mono-, tri- and tetrasaccharide intermediates could all be ligated to lipid A-core oligosaccharide. As this ligation is a periplasmic process, this suggested that a complete O unit was not essential for WzxEcO16-mediated translocation (Feldman et al., 1999). Subsequently, Marolda and colleagues demonstrated that several wzx genes from Shigella and E. coli could be used to rescue E. coli O16 O-Ag biosynthesis in a reconstituted system for a ΔwzxEcO16 mutant when supplied in trans (Marolda et al., 2004). These wzx genes were all from strains in which the composition of the O unit varied, except for the presence of GlcNAc as the proximal UndPP-linked sugar. Complementation was not attained using wzx genes from P. aeruginosa PAO1 and S. enterica group B, two bacterial species that use a different UndPP-linked initiating sugar (Marolda et al., 2004). Taken together, these investigations suggested that Wzx substrate specificity was dependent on the identity of the proximal UndPP-linked sugar. However, new data (described below) indicate that the non-native high-level expression of plasmid-encoded wzx genes may have masked the preference of Wzx for its full O-unit substrate.
Recognition of side-branch modifications
Two recent independent investigations demonstrated that Wzx substrate specificity is not solely dependent on the recognition of the UndPP-linked sugar. The first involved characterization of the plant pathogens Pantoea stewartii and Erwinia amylovora (Wang et al., 2012), the etiologic agents of Stewart's wilt disease in maize (Roper, 2011) and fire blight disease in apple trees (Oh and Beer, 2005) respectively. Each of these Gram-negative bacterial species produce EPS polymers (from UndPP-linked repeats) termed stewartan and amylovoran that are required for virulence (Langlotz et al., 2011). The UndPP-linked main-chain trisaccharide structures of the stewartan and amylovoran repeats are almost identical, as are the terminal side branches; the key difference is in the differentially regulated capping of the side branches, with primarily a glucose residue present in the former and a ketal pyruvate in the latter (Fig. 9). Although uncommon, capping with the reciprocal motif can also occur for both glycans (Nimtz et al., 1996a,b; Wang et al., 2012). Both P. stewartii and E. amylovora were found to possess two wzx genes, with the first being present at the 3′ end of related primary EPS gene clusters (containing the pyruvylation machinery), termed wzx1 and amsL1 respectively. The second is localized elsewhere in the chromosome (in conjunction with a terminal glucosyltransferase), annotated as wzx2 and amsL2 respectively (Wang et al., 2012).
Insertional disruption of amsL1 in E. amylovora abrogated the production of amylovoran, a phenotype that could be restored through complementation in trans. Overexpression of the terminal glucosyltransferase (from P. stewartii) in the amsL1 mutant of E. amylovora also resulted in restoration of EPS production to wild-type levels, suggesting that another Wzx-like protein was specifically mediating translocation of non-native gluco-amylovoran repeat units (Fig. 7), with the likeliest candidate being AmsL2 (Wang et al., 2012). Conversely, analogous disruption of wzx1 in P. stewartii had no effect on the production of stewartan (Wang et al., 2012), while disruption of wzx2 resulted in a complete loss of its production (Carlier et al., 2009). However, overexpression of the terminal pyruvylation machinery (from E. amylovora) in the wzx1 mutant background of P. stewartii resulted in a 75% decrease in the amount of terminally glucosylated stewartan, likely due to an accumulation of pyruvylated repeat units (Fig. 7) that were unable to be transported in the absence of Wzx1. In this same pyruvylation scheme, generation of a double mutant defective in wzx1 and the terminal glucosyltransferase resulted in a complete loss of stewartan biosynthesis. Alternatively, creation of a double mutant defective in wzx2 and the terminal glucosyltransferase gene resulted in significant production of non-native pyruvylated stewartan (Fig. 7). Taken together, these data indicate that despite the identical nature of the UndPP-linked sugar for each EPS repeat unit, Wzx1/AmsL1 possesses specificity for EPS repeat units with a terminally pyruvylated side branch, while Wzx2/AmsL2 specifically translocates EPS repeat units that have a terminally glucosylated side branch (Wang et al., 2012).
In a separate study by the Reeves group, Hong and colleagues arrived at a similar conclusion when they investigated the substrate specificity of S. enterica Wzx flippases (Hong et al., 2012). Serogroups B, D1, D2 and E of S. enterica synthesize UndPP-linked O units containing identical main-chain trisaccharide sugars, with the terminal d-mannose residue either α- (groups B and D1) or β-linked (groups D2 and E) to the penultimate l-rhamnose. Certain O units also contain a terminal side-branch modification of an abequose (group B) or tyvelose (groups D1 and D2) dideoxyhexose sugar, neither of which is present in the O unit of group E (Fig. 8). The wzx genes from groups D1 and D2 are analogous (99% identical sequence), but different from that of either group B or group E, suggesting a lack of preference for the α/β linkage of the terminal d-mannose in groups D1 and D2 (Hong et al., 2012). Hong and colleagues analysed chromosomal wzx mutants for S. enterica groups B, D2 and E and performed cross-complementation for a given serotype with the wzx gene from another. To eliminate the potentially confounding effects of overexpression artefacts resulting from complementation in trans (described above), the exogenous wzx genes were introduced into the respective chromosomes via allelic replacement at the native site using the Lambda Red recombination approach (Datsenko and Wanner, 2000). This methodology ensured wild type-level regulation and expression for each introduced wzx gene (Hong et al., 2012). Through this meticulous experimental design, the authors demonstrated that the Wzx flippases for S. enterica groups B, D2 and E were specific for the presence (groups B and D2) or absence (group E) of the terminal dideoxyhexose side-branch decoration of the O unit, despite the identical chemical structure of the UndPP-linked sugar as well as the identity of the proximal main-chain residues (Fig. 8). Wzx flippases from groups B and D2 were also found to translocate O units containing either an abequose or tyvelose equally well, indicating that specificity was due to the physical presence of either side branch, and not to its specific chemical structure (Hong et al., 2012).
The recent characterization of WzxPa by the Lam laboratory provides the first tertiary structural context in which to interpret the substrate specificity of an O-unit flippase (Islam et al., 2012), and led to the discovery of a cationic lumen in WzxPa (Fig. 5). It is logical that such a characteristic would have evolved to specifically accommodate the translocation of the two negatively charged terminal sugars of the P. aeruginosa PAO1 O unit (Islam et al., 2012), particularly since the initial UndPP-linked d-fucosamine moiety is uncharged (Knirel et al., 2006). The presence of a cationic lumen to accommodate anionic mannuronic acid polymer transport has also been illustrated in the X-ray crystal structure of the outer membrane beta-barrel AlgE required for the secretion of alginate (Whitney et al., 2011). Additionally, the lack of conservation for several functionally important residues between WzxPa and WzxEc (Fig. S2) is not unexpected, particularly since the respective O-unit substrates are quite distinct (Knirel et al., 2006; Stenutz et al., 2006). The knowledge derived from these structural data, combined with the elegant genetic characterizations described above, indicates that determining the underpinnings of Wzx substrate specificity will be a complex process likely requiring a combination of genetic, biochemical, biophysical and bioinformatic techniques.
Energetics of Wzx-mediated flipping
In the report by Rick and colleagues (2003) of WzxE-mediated uptake of soluble nerol pyrophosphate-linked GlcNAc into everted membrane vesicles (described above), the authors claimed that the process was not affected by the addition of exogenous energy sources such as ATP, NADH and d-lactate, or uncoupling agents such as the H+ ionophores carbonyl cyanide m-chlorophenylhydrazone (CCCP) and carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP). Unfortunately, the relevant data that would lead to such a conclusion were not presented in that study. Instead, the authors postulated that WzxE-mediated translocation occurred through an equilibrium process via facilitated diffusion; in this manner, the translocation of the native UndPP-linked repeat from the cytoplasmic to the periplasmic leaflets of the IM could be driven by its utilization during the periplasmic polymerization of ECA by the respective Wzy protein of the ECA biosynthesis pathway (Rick et al., 2003).
However, the use of a non-native substrate in this in vitro assay may not have been representative of in vivo conditions, as Rick and colleagues did not observe an effect on nerol-linked soluble substrate uptake by the everted membrane vesicles upon disruption of the native E. coli O16 wzx gene (Rick et al., 2003). These data are in conflict with the results of Feldman and colleagues in which WzxO16 was implicated in the successful flipping of UndPP-GlcNAc across the IM, resulting in its ligation to lipid A-core in vivo (Feldman et al., 1999). Furthermore, the results of Rick and colleagues (2003) would suggest that the WzxE protein allowed for the transport of the lipid acyl chain (bound to the GlcNAc residue) through the lumen of the protein; this characteristic is at odds with the situation in vivo in which the lipid carrier may play an important role in flipping (discussed below).
The role of UndPP
Undecaprenyl pyrophosphate is a highly hydrophobic molecule (Hartley and Imperiali, 2012); therefore, it is unlikely that its hydrophobic 55-carbon acyl chain would actually enter the lumen of the flippase. This notion is perfectly illustrated for the proposed structure of WzxPa in which the lumen of the protein is highly charged (Fig. 5) (Islam et al., 2012), making this environment rather inhospitable for the passage of the hydrophobic Und lipid tail. It is conceivable that at most, the sugar substrate as well as the proximal pyrophosphate components would be suited for translocation through the protein, with the Und lipid component remaining in the membrane. Based on the structure of WzxPa (Fig. 4), Islam and colleagues have speculated that the partial entry of UndPP from the cytoplasmic leaflet of the IM could take place between TMS1 and TMS9 in such a manner that the C55 tail remains embedded between the acyl chains of the periplasmic and cytoplasmic IM leaflets; once substrate flipping has occurred, this would also allow for the exit of the sugar repeat from between the same helices, which correspond to the front periplasmic portal of the protein (Figs 5 and 6) (Islam et al., 2012).
The (putative) TMS of certain glycosyltransferases and E. coli capsular polysaccharide biosynthesis enzymes have been found to contain so-called polyisoprenyl recognition sequences (PIRS) (Hartley and Imperiali, 2012). Within model membranes, these are tracts of TMS amino acids that have demonstrated selective binding to substrates linked to polyprenyl phosphate such as UndP, resulting in localized perturbation of membrane structure and the accumulation of potential energy (Zhou and Troy, 2003; 2005). While the sequence consensus for the limited number of PIRS that have been identified is poorly defined, one commonality appears to be the presence of a Pro residue in the TMS. A less conserved feature is the occurrence of multiple Phe/Tyr residues along the length of the PIRS (Zhou and Troy, 2003; 2005). From the WzxPa structure model, TMS8 (NH2-GDSAGWFALTLKIMGAPISLLAASVLDVFKEQAAR-COOH) and TMS9 (NH2-REFGNCRGIFLKTFRLLAVLALPPFIIFWFIG-COOH) are the two likeliest candidate TMS to have a role in any potential UndPP-mediated interactions, and both would qualitatively satisfy the loosely defined requirements described above for PIRS (the latter more so than the former). However, there is currently no evidence for any role for PIRS in Wzx-mediated UndPP-linked glycan flipping, and we explicitly state that the hypothetical scenario described above is purely conjecture and speculation that is meant to provoke new thoughts and to stimulate further work to elucidate the interplay between UndPP and Wzx.
Secondary active transport is a ubiquitous method for transport across membranes in all cells. This process involves the coupling of the electrochemical potential of one solute with membrane translocation of another. One such secondary active mechanism is antiport, in which two different solutes are transported across the membrane in opposite directions (Forrest et al., 2011). Antiport mechanisms have been experimentally demonstrated for various bacterial members of the MATE protein family, involving the uptake of either Na+ or H+ from the periplasm into the cytoplasm, coupled with the efflux of drugs and other toxic compounds from the cytoplasm into the periplasm (Kuroda and Tsuchiya, 2009). Consistent with these observations, and in addition to the native structure described earlier, He and colleagues were also able to co-crystallize the MATE protein NorMVc with a Rb+ ion (PDB ID: 3MKU) (He et al., 2010), which is often used in X-ray crystallography as a higher-density structural analogue of Na+. This Rb+ co-crystal has led to the proposal of an alternating-access antiport model (Forrest et al., 2011) for NorMVc in which Na+ would first bind at a conserved site in the outward-facing structure of the protein. This would induce conformational changes within NorMVc leading to an inward-facing state able to bind substrate in the cytoplasm and there release the bound cation. Finally, substrate binding would cause conformational changes returning NorMVc to the outward-facing conformation, resulting in substrate export to the periplasm and allowing binding of another cation to restart the transport cycle (He et al., 2010).
In keeping with this initial model for NorMVc efflux, as well as the putative similarities in functional mechanisms between MATE family and PST family proteins (Hvorup et al., 2003), and based on the WzxPa tertiary structure model, Islam and colleagues proposed that the flipping of UndPP-linked O units occurs through an analogous antiport-like mechanism (Islam et al., 2012). Briefly, this would entail periplasmic binding of a cation by WzxPa in an outward-facing conformation (Fig. 9A and B), followed by structural changes to an inward-facing state in which UndPP-linked O-unit substrate could bind from the inner leaflet of the IM with concurrent cation release in the cytoplasm (Fig. 9C). Upon O-unit binding, WzxPa would revert to an outward-facing conformation (Fig. 9D), allowing for release of the O unit into the periplasmic leaflet of the IM, exiting laterally from the Wzx lumen via the front periplasmic portal (Fig. 9E) (Islam et al., 2012). In addition to the structural data already described, this proposed model for WzxPa function is further reinforced by the essential nature of amino acid residues Glu61, Asp269 and Asp359 based on site-directed mutagenesis and complementation experiments. These carboxylate-containing residues in WzxPa may be analogous to the evolutionarily conserved and functionally important NorMVc residues Asp36, Glu255 and Asp371. These particular amino acid residues may bind the antiported ion in the outward-facing state, and be shared by organic cationic substrates in the inward-facing state (van Veen, 2010).
The investigations discussed in this review have begun to shed light on our overall understanding of O-unit flippases, but they have only scratched the surface. Further investigations are warranted to address a multitude of questions that will require expertise in a variety of disciplines (Table 1). One of the central tenets of the Wzx/Wzy-dependent pathway is that Wzx interacts with other members of the assembly pathway. Biochemical and/or biophysical data indicating interactions between various assembly pathway constituents would undoubtedly consolidate the overall concept of the Wzx/Wzy-dependent pathway; however, the intrinsically low expression levels of these proteins is a hurdle that must be overcome towards this aim.
Table 1. Key questions for future research on the structure and function of Wzx flippases
Does Wzx interact with other proteins in the Wzx/Wzy-dependent pathway?
At the moment, only indirect genetic evidence suggests that Wzx may interact with Wzy and Wzz. Biochemical (e.g. pull down) and/or biophysical (e.g. surface plasmon resonance) data would be ideal for confirming heterotypic interactions.
How is expression of Wzx regulated?
The modulation of Wzx expression has never been explored; it is not known if the levels of Wzx are maintained at a constant level or if they purposely fluctuate over time.
How does the sugar substrate interact with Wzx?
Functionally important residues in Wzx have been identified, but whether or not they interact directly with the respective sugar substrates is unknown.
What are the roles of various functionally important amino acids for Wzx?
While substrate interaction would be a natural conclusion, functionally important residues may alternatively be involved in energy transduction or specific tertiary structure constraints required for flipping.
Can Wzx occupy multiple structural conformations?
The dynamic nature of Wzx is unknown, i.e. does the protein simply form a static conduit for passage of substrate, or are conformational changes required for mediate substrate translocation?
What powers Wzx-mediated transport?
The role of passive diffusion vs. secondary-active transport (antiport) is an important question that needs to be resolved as it will fundamentally alter the method by which UndPP-linked transbilayer transport is accomplished.
Can flippase activity be demonstrated in vitro for Wzx?
Wzx is the only protein among the constituents of the Wzx/Wzy-dependent pathway for which a robust flippase assay has yet to be developed. In vitro demonstration of UndPP-linked substrate flipping is an essential step to elucidating Wzx function.
Does the UndPP moiety of the flipped substrate play a role in translocation?
The manner in which Wzx proteins are able to lumenally transport sugar substrates is unknown, particularly as they are linked to UndPP, which in all likelihood remains in the lipid domain of the IM. As such, Wzx must be able to partially open in order to accommodate the presence of the UndPP lipid tail during translocation, otherwise the sugar substrate would never be able to enter/exit the Wzx lumen.
Additional tertiary structural data for Wzx proteins would be invaluable in characterizing flippase function. In particular, a substrate-bound conformational form would allow for comparison with apo-form structures/models; this would allow for comparisons between the TMS in each form, providing a basis for proposing dynamic TMS rearrangements required for the translocation cycle. These data would also allow for contextual interpretation for the roles of various functionally important amino acid residues that have already been identified.
Furthermore, Wzx is the only constituent of the Wzx/Wzy-dependent pathway for which a robust in vitro functional assay has not yet been developed. Biophysical studies to examine the gating and/or flipping activity of the purified protein would directly substantiate the proposed flippase function for Wzx. However, incorporation of UndPP-linked substrate into a reaction scheme would be quite challenging due to its inherent hydrophobicity. As purification of various UndPP-linked glycans from in vivo sources can be difficult, the chemical synthesis of the polysaccharide repeat units would be an alternative approach. A caveat to this approach is the complexity of the sugar structures in certain O units, for instance those of P. aeruginosa, which contain rare sugars. This is another hurdle that must be overcome.
Despite the various challenges that have been highlighted in this review, research into the structure, function and interactions of Wzx proteins is highly rewarding and provides the ideal platform for cross-disciplinary investigations involving genetics, molecular biology, biochemistry, biophysics and bioinformatics. No doubt there are many more exciting discoveries to be made for such widespread and physiologically important, yet poorly understood, membrane proteins.
Work in the laboratory of J. S. L. is supported by operating grants from Cystic Fibrosis Canada (CFC) and the Canadian Institutes of Health Research (CIHR) (Grant MOP-14687). S. T. I. is the recipient of a CIHR Frederick Banting and Charles Best Canada Graduate Scholarship doctoral award, a CIHR Michael Smith Foreign Study award, a CFC doctoral studentship and an Ontario Graduate Scholarship in Science and Technology. J. S. L. holds a Canada Research Chair in Cystic Fibrosis and Microbial Glycobiology. This article is dedicated to the memory and legacy of Dr J. William (Bill) Costerton, a cherished mentor, instructor and renowned pioneer recognized as ‘The Father of Biofilms’; he opened our eyes to appreciate biofilms as an important concept in microbiology and fundamentally changed our understanding of bacterial survival and persistence in nature.