Structural insights into PSL polysaccharide assembly and export to the cell surface via the Wzx/Wzy-dependent pathway Authors: Jiao Liu, Yanlong Han, Lanxin Wang, Luyan Z Ma, Dongchun Ni, Yihua Huang Published: 2026 ABSTRACT AbstractBacterial extracellular polysaccharides play a crucial role in mediating pathogen-host interactions and bacterial fitness via biofilm formation. The Wzx/Wzy-dependent pathway is the most prevalent and conserved strategy for polysaccharide biosynthesis. Psl (polysaccharide synthesis locus) is a key biofilm matrix polysaccharide in Pseudomonas aeruginosa PAO1, and its biosynthesis machinery is predicted to be a Wzx/Wzy-dependent biosynthesis system. Here, using Psl in PAO1 as a model strain, we determine the cryo-EM structures of the PslD-PslE complex. These structures reveal that PslD-PslE complex forms a continuous, protected conduit across the entire cell envelope. Further structural and functional analyses demonstrate that the polymerase PslJ, is likely localized in the membrane lumen formed by the octameric arrangement of PslE’s transmembrane helical pairs. We propose a mechanistic model in which Und-PP-linked pentasaccharide units of Psl access PslJ through side portals in the PslE octamer, shielding the polymerization and translocation processes from degradation by PlsG, a periplasm-localized endoglycosidase. The iterative addition of incoming repeat units to the reducing terminus of the growing polysaccharide chain is hypothesized to drive Psl export through the channel, a mechanism that may be conserved across the Wzx/Wzy-dependent polysaccharide biosynthesis pathways.Subject terms: Biofilms, Cryoelectron microscopy, Structural biology Introduction Many bacteria synthesize abundant long-chain polysaccharides, which can either form a capsule structure tightly associated with the cell surface or exist as exopolysaccharides predominantly secreted into the surrounding environment1–3. These polymers play pivotal roles in mediating pathogen-host interactions and enhancing bacterial fitness through biofilm formation3. Despite extensive diversity in bacterial polysaccharide composition and chain length, their assembly and export primarily rely on one of three conserved biosynthesis pathways: the Wzx/Wzy-dependent pathway, the ATP-binding cassette (ABC) transporter-dependent pathway, and the synthase-dependent pathway4. The Wzx/Wzy-dependent pathway is the most prevalent and conserved strategy for polysaccharide biosynthesis in both Gram-negative and Gram-positive bacteria2,5–7. For example, the opportunistic human pathogen P. aeruginosa produces three secreted extracellular polysaccharides: alginate, polysaccharide synthesis locus (Psl), and pellicle polysaccharide (Pel)8,9, among which Psl biosynthesis employs the Wzx/Wzy-dependent pathway (Fig. 1a). In this pathway, glycosyltransferases first assemble polysaccharide repeat units on the inner membrane (IM) inner leaflet using the polyisoprenoid lipid carrier undecaprenol diphosphate (Und-PP). The lipid-linked repeat units are then translocated to the periplasm by PslK (Wzx), followed by chain elongation mediated by PslJ (Wzy), which transfers the growing glycan from its Und-PP carrier to the nonreducing terminus of an incoming Und-PP-linked repeat unit. The narrow size distribution of the resulting polymers is regulated by the polysaccharide co-polymerase (PCP) PslE, which interacts physically with PslJ and further binds its cognate outer membrane (OM) polysaccharide export protein (OPX) PslD, coupling glycan chain growth with trans-envelope translocation10 (Fig. 1a). Psl biosynthesis in PAO1 requires 12 co-transcribed psl genes (pslA to pslL)11–13 (Fig. 1b), relying on four functionally conserved proteins (PslK/Wzx, PslJ/Wzy, PslD/Wza, PslE/Wzc) analogous to those governing Group 1 capsular polysaccharide biosynthesis in E. coli14. Notably, PslG, a periplasmic endoglycosidase that cleaves between Psl pentasaccharide units, is critical for Psl biosynthesis and biofilm formation in PAO115, yet exogenous PslG inhibits biofilm formation and disrupts pre-existing biofilms across multiple Pseudomonas species11,16,17. Structurally, the Psl oligosaccharide repeat unit is a neutral pentasaccharide consisting of L-rhamnose, D-glucose, and D-mannose at a 1:1:3 stoichiometric ratio (Fig. 1c); functionally, it acts as the primary scaffold matrix component in PAO1, forming fiber-like structures that enmesh bacterial communities and maintain biofilm structural integrity throughout development3,11,18 (Fig. 1d). The Wzx/Wzy-dependent Psl biosynthesis pathway. a Proposed model of Psl biosynthesis, illustrating the core components of the assembly and export machinery. b Schematic representation of the PAO1 psl operon (pslA-pslL). Among these genes, pslD and pslE encode export proteins, while pslK (encoding the Wzx flippase) and pslJ (encoding the Wzy polymerase) are the key genes that define the Wzx/Wzy-dependent biosynthesis pathway. c Chemical structure of the Psl pentose repeating unit, modeled via GLYCAM server with linkages: →4)-β-D-GlcNAc-(1 → 4)-β-D-GlcA-(1 → 4)-β-D-Glc-(1 → 4)-α-L-Fuc-(1 → 3)-β-D-Glc-(1 → . Predicted structural dimensions: the maximum length of Psl pentose repeating unit is 17.6 Å, and the maximum width of Psl is 9.3 Å. d Qualitative analysis of Psl in PAO1 biofilms via Psl matrix staining. Bacterial biofilms and Psl were stained with FM4-64 and HHA-FITC, respectively. Merged micrographs showed that Psl was attached to the cell surface of PAO1. Abbreviations: FM4-64 N-(3-triethylammoniumpropyl)−4-(6-(4-(diethylamino) phenyl)hexatrienyl)pyridinium dibromide, HHA-FITC Hippeastrum hybrid agglutinin (HHA)-fluorescein isothiocyanate (FITC), DIC Differential Interference Contrast. The result was repeated three times independently with similar results. The study of the OPX protein Wza19 and several PCP proteins20–24 (e.g., Wzc) is limited, either as stand-alone subunits or in complexes where regulation is proposed to be mediated by dynamic autophosphorylation. In this study, we report the cryo-EM structures of the PslD-PslE complex and the PslD-PslEY524C+R595C complex, in which the two Cys mutations stabilized the PslE cytoplasmic tyrosine kinase-like domains. Actually, the PslD lacked an obvious C-terminal helix, and the PslE lacked the kinase active site as well as the C-terminal Y-cluster compared with Wza and Wzc, respectively. These structures reveal that the PslD-PslE complex forms a continuous and protected conduit spanning the periplasm and OM. Notably, the PslE octamer possesses two surface pores, likely allowing the growing Psl chain to enter the periplasm but preventing periplasm-localized PslG hydrolase from accessing the Psl within the lumen. Further structural and biochemical analyses demonstrate that the PslJ polymerase is located within the central membrane lumen formed by the octameric arrangement of the transmembrane helices (TMHs) of PslE, which is presumably stabilized by the C-terminal tail of PslE. Complementation experiments with PslE disulfide crosslinking mutants demonstrate that the assembly and export of Psl do not require the cyclic assembly and disassembly of the entire PslE20,21. Results Overall architecture of the PslD-PslE complex We co-expressed pslD and pslE from P. aeruginosa PAO1 in E. coli C43 cells and purified the resulting PslD-PslE complex by affinity chromatography using a Strep-tag II fused to the C-terminus of PslE. Gel filtration chromatography yielded a distinct elution peak (P1), with SDS-PAGE analysis confirming the presence of both proteins in an approximate 1:1 molar ratio (Supplementary Fig. 1a). Cryogenic electron microscopy (cryo-EM) data were collected for the PslD-PslE complex. Two-dimensional (2D) classification predominantly revealed side-views of the complex, each featuring two detergent-solvated transmembrane disks connected by a slit-like structure, although the cytoplasmic regions of PslE exhibited only weak, diffuse density (Supplementary Fig. 1b). We determined the cryo-EM structure of the eightfold symmetric (C8) complex at 3.06 Å resolution, but as anticipated, the TM helices and cytoplasmic domains of PslE lacked discernible density after 3-D reconstruction (Fig. 2a and Supplementary Fig. 1c–f). The cytoplasmic domain of PslE diverges from that of its E. coli K30 orthologue Wzc in two key aspects: (1) it lacks the catalytic lysine residue conserved within the Walker box motif; and (2) its C-terminal tail does not harbor multiple tyrosine residues20,21 (Supplementary Fig. 2a,b). To stabilize the cytoplasmic domains of octameric PslE, we engineered an inter-subunit disulfide bond between residues Y524C and R595C, based on the octameric PslE structure predicted by AlphaFold3 (Supplementary Fig. 3a). Complementation assays demonstrated that the PslEY524C+R595C mutant fully rescued wild-type PslE function in Psl biosynthesis and biofilm formation, confirming its in vivo functionality (Supplementary Fig. 3b). As expected, gel filtration chromatography of the purified PslD-PslEY524C+R595C complex showed a stable assembly and non-reducing SDS-PAGE analysis verified the formation of inter-subunit disulfide bonds between PslE monomers (Supplementary Fig. 4a). Cryo-EM data collection for the stabilized complex yielded an overall 3.03 Å resolution map (Supplementary Fig. 4b). Following iterative local refinement, cryo-EM densities corresponding to the TMHs and cytoplasmic domains of octameric PslE were visualized. The cytoplasmic region achieved a local resolution of 4.06 Å (Fig. 2a; Supplementary Fig. 4f, g), a resolution sufficient to facilitate the model building of the full-length PslD-PslEY524C+R595C complex (Fig. 2b, right). Overall structure of the PslD-PslE complex. a Cryo-EM density maps of wild-type PslD-PslE (left) and disulfide-stabilized PslD-PslEY524C+R595C mutant (right) complexes. The mutant complex spans 370 Å in length and 108 Å in width, defining a ~ 230 Å periplasmic thickness. PslD (green) and PslE (purple) densities are color-coded. b Cartoon representation and domain architecture of the PslD-PslEY524C+R595C octamer, with 16 chains colored individually and key structural features labeled. A close-up view of the engineered inter-subunit disulfide bond within the PslE octamer is presented. Despite the absence of discernible density in TMHs and the cytoplasmic domain of PslE within the wild-type PslD-PslE complex, the PslD-PslEY524C+R595C complex exhibits similar overall architecture to the wild-type complex, with a root-mean-square deviation (r.m.s.d.) of 2.8 Å over 4707 aligned Cα atoms. The PslD-PslEY524C+R595C complex spans the entire cell envelope, measuring 370 Å in length and reaching a maximum width of 108 Å (Fig. 2a). Within the complex, PslD assembles into an octameric structure localized to the OM, with its periplasmic domains extending 74 Å into the periplasmic space. Conversely, PslE forms an octamer embedded in the IM, with its C-terminal tyrosine kinase-like domain situated in the cytoplasm. The PslE octamer extends 266 Å in length, and its periplasmic region protrudes ~156 Å into the periplasm (Fig. 2a). Collectively, these dimensions suggest the periplasmic space housing the PslD-PslE complex spans approximately 230 Å in thickness, and the structure of the PslD-PslEY524C+R595C complex represents a trans-envelope conduit that is responsible for Psl polysaccharide secretion (Fig. 2b). PslD represents an OM β-barrel protein composed of parallel β-strands As an OPX family member, PslD mediates Psl translocation across the OM25–27. Analogous to the prototypical OPX protein Wza, PslD is a lipoprotein that harbors a signal peptidase II cleavage site between alanine 15 (Ala15) and cysteine 16 (Cys16). Post-translational lipid modification of Cys16 enables the anchoring of PslD to the OM, where PslD assembles into a stable octameric complex that spans the OM bilayer (Fig. 3a). However, the PslD octamer exhibits several distinct features compared to Wza. First, each TM segment of PslD (residues 221-254) contributes a β-strand (240-254) to form the β-barrel in the PslD octamer (Fig. 3a), whereas the transmembrane domain of Wza forms a canonical α-helical barrel (Fig. 3b, c). Second, unlike most β-barrel OM proteins formed via anti-parallel β-strand interactions, the TM β-barrel of PslD is stabilized by parallel β-strand associations, representing a rare structural paradigm in bacterial OM proteins (Fig. 3c, left)28,29. Third, Wza contains three distinct domains (D1–D3) that form three stacked rings (R1-R3) in the periplasm (Fig. 3b), while PslD lacks the middle D2 domain, retaining only D1 and D3 to form two periplasmic rings (R1 and R3) (Fig. 3a). D1 domain is also termed as polysaccharide export sequence (PES) motif (Pfam PF02563), which mediates interactions with its cognate PCP protein, and is conserved in OPX proteins1,10. Fourth, PslD exhibits an expanded secretion pore compared to Wza. The isolated Wza structure appears to adopt a closed conformation with a constriction region in the secretion channel (9.6 Å in diameter) (Fig. 3d). In contrast, the PslD octamer in both the PslD-PslE complex and the PslD-PslEY524C+R595C complex features a wider constriction region (13.2 Å in diameter) (Fig. 3e), large enough to permit export of Psl in a stretched conformation. Structural comparison of PslD and Wza. a Octameric structure of PslD (green) and its monomeric protomer (rainbow-colored), highlighting the transmembrane (TM) segments composed of β-strands, as well as the two periplasmic domains (D1 and D3). The residue Cys16 in the N-terminal of PslD labeled is likely lipidation. b Octameric structure of Wza (cyan) and its monomeric protomer (rainbow-colored), showing the transmembrane (TM) segments composed of α-helices, along with the three periplasmic domains (D1, D2, and D3). c Contrast of the TM domains between PslD (residues M222-N254) (left) and Wza (residues L357-R254) (right): PslD contains a TM β-barrel assembled from parallel β-strands, whereas Wza forms an α-helical barrel. d Inner surface property of the Wza octamer (left), displaying a negatively charged lumen and a constriction diameter of ~9.6 Å. e Inner surface property of the PslD octamer, showing a hydrophilic lumen and a constriction diameter of ~13.2 Å. Hole sizes were measured using MOLE 2.048. Structure and surface features of PslE suggest that Psl polymerization and translocation occur in the lumen of PslE octamer Based on secondary structure predictions, PCPs can be categorized into three subfamilies: PCP1, PCP2, and PCP330. Both PCP1 and PCP2 subfamilies are involved in the Wzx/Wzy-dependent polysaccharide biosynthesis pathway. As a member of the PCP2 subfamily, PslE consists of three distinct structural regions: the periplasmic region (residues 39–455), the transmembrane region (19–38 and 457–482), and the cytoplasmic region (6–18 and 485–662) (Fig. 4a). The cytoplasmic region includes a short N-terminal loop (residues 6–18), the C-terminal tyrosine kinase-like domain (485–642), and the C-terminal tail (645–662). The tyrosine kinase-like domain in PslE has central β-sheets flanked by two bundles of α-helices, but it lacks the catalytic Lys residue found in similar sites of other bacterial tyrosine kinase domains such as Wzc, Etk, and CapB21,31 (Supplementary Fig. 2a). Additionally, unlike Wzc, whose tyrosine-rich C-terminal tail (residues Y705, Y708, Y710, Y711, Y713, and Y715 of Wzc) mediates inter-subunit interactions, PslE’s C-terminal tail contains only a single tyrosine residue and exhibits weak cryo-EM density at the IM-cytoplasm interface, near the cytoplasmic side of transmembrane helix 2 (TMH2) of the adjacent PslE protomer (Fig. 4b and Supplementary Fig. 2b). Each PslE monomer comprises two TMHs, which form an X-like arrangement similar to Wzc and Wzz20,22. The eight TMH pairs create oval-shaped portals (24 Å wide by 35 Å high) open to the IM (Fig. 4b). The periplasmic region is subdivided into three structural motifs. Motif 1 is formed by three noncontiguous stretches of the protein sequences (residues 41–103, 146–201, and 438–452), and comprises four β strands, four α helices, and a long loop forming the constriction region with 10 Å in diameter in the PslE octamer lumen (Fig. 4c, d). Motif 2 (residues 104–145) forms an intra-subunit pore, featured by a long α helix (112–139) that runs parallel to the IM (Fig. 4c). Unlike Wzc’s Motif 2, which folds into a carbohydrate-binding domain (CBD) found in the ABC transporter Wzm-Wzt32, PslE’s Motif 2 lacks this functional architecture. Motif 3 (residues 202–436) is composed of a long two-helical bundle flanked by two anti-parallel β strands (283–294 and 344–355) (Fig. 4c). In the PslE octamer, these strands assemble to form a 16-stranded β-barrel in the periplasm, absent in structure-available PCP proteins, such as Wzc and KpsE. (Fig. 4b) (Supplementary Fig. 7). Structural and surface features of PslE. a Domain architecture of the PslE monomer, colored in a rainbow gradient (N-terminus to C-terminus). b Cartoon representation of the PslE octamer, illustrating the membrane-exposed lateral portals with dimensions of 24 × 35 Å. A single protomer is colored in a rainbow gradient to highlight the C-terminal tails localized at the interface between the cytoplasm and inner membrane (IM), as well as the periplasmic 16-stranded β-barrel domain. c Periplasmic region of PslE, displaying conserved Motifs 1–3 and the constriction-loop embedded within Motif 2. d Inner surface characteristics of the PslE octamer, showing a hydrophilic lumen and a ~ 10 Å constriction site. e Molecular surface representations of the intra-subunit and inter-subunit pores of PslE, with the intra-subunit pore measuring 16.3 × 15.5 Å and the inter-subunit pore measuring 7.5 Å × 11.1 Å (the red dotted line represents the distance between the centers of the two atoms, and distances are measured using PyMOL). The PslE octamer harbors two surface pore types: One is an intra-subunit pore that is formed by the Motif 2 (Fig. 4c, e). This surface pore has an approximate dimension of 18.3 × 15.5 Å, located at the base of the periplasmic conduit. The other is an inter-subunit surface pore formed by neighboring protomers. This inter-subunit surface pore has an approximate dimension of 7.5 Å×11.1 Å, positioned midway of the periplasm conduit formed by the PslD-PslE complex (Fig. 4e). Given that the largest width of Psl is 9.3 Å (Fig. 1c), both pore types are sufficiently sized to allow the non-reducing terminus of Psl to enter the periplasm. PslD stabilizes both the conformation and the octameric state of PslE via its PES motif OPX proteins possess a conserved PES motif26,27. This is the only region of primary sequence similarity shared by OPX proteins. In both the PslD-PslEY524C+R595C complex and the PslD-PslE complex, residues in the D1 domain (or the PES motif) of PslD directly interacts with residues in Motif 3 of PslE, and several key interaction sites were identified (Fig. 5a). Residues Pro118PslD and Ile119PslD each contribute a hydrogen bond to the interaction via their main chain atoms, while Arg121PslD makes a couple of hydrogen bonds to Thr317PslE and Met315PslE, and contributes a salt-bridge with residue Asp320PslE. Moreover, Tyr120PslD and Tyr318PslE participate in π-π interactions with each other. However, the PslDR121A mutant does not appear to affect its interaction with PslE (Supplementary Fig. 5). In stark contrast, disrupting the π-π interaction between Tyr120PslD and Tyr318PslE, either by employing the PslDY120A mutant or the PslEY318A mutant, weakens the PslD-PslE interaction and reduces Psl production and bacterial attachment in PAO1 (Fig. 5b, c). Interestingly, while the PES motifs adopt a similar fold in both PslD and Wza, sequence alignment of the PES motifs in various OPX proteins from different bacterial strains reveals that none of the PslE-interacting residues in PslD are conserved (Supplementary Fig. 6a, b). This finding may imply that OPX proteins are likely to interact specifically with their cognate PCP proteins, and such interactions exhibit a high degree of specificity. Close-up view and functional characterization of PslD-PslE interactions. a Close-up view of the interaction interface between PslD and PslE, highlighting key contact residues involved in complex formation. b Relative Psl polysaccharide production in P. aeruginosa PAO1 wild-type strain, ΔpslE mutant, ΔpslE mutant complemented with wild-type pslE or the pslEY318A mutant allele; ΔpslD mutant, ΔpslD mutant complemented with wild-type pslD or the pslDY120A mutant allele. Psl levels were normalized to those of the wild-type PAO1 strain. Top panel: Anti-Psl immunoblot analysis of Psl production (mean ± SD, n = 3 independent biological replicates). Bottom panel: Initial bacterial attachment assay measured by crystal violet (CV) staining, with the corresponding OD560 absorbance values indicated below each column. c Mutational analysis of PslD-PslE complex stability by Strep-tagged PslE pulldown assay. Lanes 1-2: Wild-type PslD-PslE complex incubated at 25 °C (Lane 1) or 95 °C (Lane 2); Lanes 3−4: PslDY120A mutant in complex with wild-type PslE; Lanes 5-6: Wild-type PslD in complex with PslEY318A mutant. The result was repeated in triplicate independently, and representative results are presented. d Size-exclusion chromatography (SEC) profiles comparing the PslD-PslE complex (orange curve) and isolated PslE protein (blue curve), demonstrating the formation of a higher-molecular-weight complex between PslD and PslE. e Cryo-EM analysis of the wild-type Wza-Wzc complex. SDS-PAGE analysis of the purified Wza-WzcWT (left). Representative cryo-EM micrographs and 2D classification results (right) showing the structural homogeneity and characteristic features of the Wza-Wzc complex. The result was repeated in triplicate independently, and representative results are presented. Importantly, PslD is crucial for maintaining the conformation and the octameric oligomerization of PslE. Gel filtration chromatography of isolated PslE reveals that it exhibits different oligomeric states (Fig. 5d). In contrast, in the presence of PslD, both the PslD-PslE complex and the PslD-PslEY524C+R595C complex are eluted as a sharp peak containing PslD-PslE complex in size-exclusion chromatography. This finding is further confirmed by the cryo-EM structures of the octameric PslD-PslE complex and PslD-PslEY524C+R595C complex. Consistent with this, the affinity-purified wild-type Wza-Wzc complex was found to form stable assemblies, as revealed by cryo-EM analysis of the purified sample (Fig. 5e). Furthermore, in the PslD-PslE complex, Motif 3 within the periplasmic domain of PslE adopts an extended conformation, extending approximately 120 Å into the periplasm. By contrast, in the structures of the isolated WzcK540M octamer and KpsE octamer, the corresponding Motif 3 is folded downward and displays distinct conformations20,33 (Supplementary Fig. 7). Based on these observations, we conclude that OPX proteins play a crucial role in stabilizing both the conformation and octameric state of their cognate PCP proteins via the conserved PES motif. PslJ is situated within the central membrane lumen formed by the octameric TMH pairs of PslE, a configuration presumably stabilized by PslE’s C-terminal tail Previous studies have shown that Wzy resides in the central membrane lumen created by the octameric arrangement of TMH pairs of Wzz, a member of the PCP1 subfamily34. To determine whether PslJ directly binds to the PslD-PslE complex, we co-expressed PslJ with the PslD-PslE complex in P. aeruginosa PAO1 using the pHERD20T vector. Apparently, the PslD-PslE complex can pull down PslJ, indicating that the PslD-PslE complex directly binds PslJ (Supplementary Fig. 8a). To investigate how PslJ binds to the PslD-PslE complex, we prepared a cryo-EM sample and collected cryo-EM data for the PslD-PslE-PslJ ternary complex (Supplementary Fig. 8b–g). The 3-D reconstruction of the entire complex revealed octameric PslD-PslE complex and additional densities surrounded by the TMHs of PslE octamer (Fig. 6a). After local refinement, we observed low-resolution map densities enclosed by the TMHs of PslE octamer (Fig. 6a). Since we could not build an atomic model for PslJ at this resolution, we performed 3-D structural prediction for the PslE octamer and PslJ using AlphaFold-Multimer (Fig. 6b). PslJ was predicted to be a membrane protein with 13 TM helices, situated within the octameric arrangement of the TMH pairs of PslE (Fig. 6b). In this predicted model, the first two TMHs of PslJ are in proximity to one pair of TMHs of PslE octamer, with the N-terminus of PslJ near the C-terminal tail of PslE (Fig. 6b). PslE octamer TM lumen harbors polymerase PslJ. a Cryo-EM density map of the PslD-PslE-PslJ complex, displaying IM-region density. Local refinement highlights potential PslJ density within the transmembrane (TM) lumen of PslE. b AlphaFold3-predicted structure of the PslE-PslJ complex, where PslJ (colored blue) is seated within the TM lumen of the PslE octamer. c SDS-PAGE and immunoblot analysis of the purified PslD-PslE-PslJ complex. PslJ was fused with a C-terminal His-tag, and PslE was fused with a C-terminal Strep-tag; the gel demonstrates the formation of inter-subunit disulfide bonds (Lane 1: in the presence of a reducing agent; Lane 2: in the absence of a reducing agent). Corresponding Western blots were probed with anti-His (mouse) and anti-Strep (rabbit) antibodies. d SDS-PAGE and immunoblot analysis of the wild-type PslD-PslE-PslJ complex and the PslD-PslE-PslJΔTM1–2 complex (Lanes 1-2). Expression levels of PslJ and PslJΔTM1–2 were verified by immunoblotting (right panel). e Relative Psl production in wild-type PAO1 overexpressing either full-length PslJ or the TM1-2 peptide from PslJ, normalized to the parental PAO1 strain (mean ± SD, n = 3 independent biological replicates). The bottom panel shows the initial attachment assay; crystal violet (CV) staining readings (OD) for the attachment assay are displayed below each corresponding column. f SDS-PAGE and UV analysis of the PslE-TM1-2-GFP complex, indicating pull-down of the complex by PslD-PslE. g SDS-PAGE analysis of the wild-type PslD-PslE-PslJ complex and the PslD-PslEΔ644-662-PslJ complex (Lanes 1-2). h Psl production in the ΔpslE strain complemented with either wild-type pslE or pslEΔ644-662 (mean ± SD, n = 3 independent biological replicates). To verify our structural predictions, we introduced Cys residues into both PslJ and PslE, aiming to form inter-subunit disulfide bonds if the two residues are close enough (Fig. 6b). As shown in Fig. 6c, co-expression of PslJS2C with PslD-PslEY658C resulted in purified complexes containing PslJ-PslE disulfide crosslinks, which disappeared upon the addition of β-ME on SDS-PAGE. The PslJ-PslE crosslinks were further confirmed by western blots using anti-His-tag and anti-Strep-tag antibodies. Moreover, based on the predicted structures, the first two TMHs of PslJ (TM1-2PslJ) are in close proximity to one pair of TMHs of PslE. To verify this, we deleted the first two TMHs of PslJ (residues 2–43), the PslD-PslE complex failed to pull down PslJΔTM1–2 despite its expression in P. aeruginosa PAO1 (Fig. 6d). In line with this, over-expression of TM1-2PslJ in P. aeruginosa PAO1 abolished Psl production (Fig. 6e) and the PslD-PslE was able to pull down the TM1-2PslJ peptide when its N-terminus was fused with a GFP protein (Fig. 6f). This inhibitory effect likely results from competitive binding at the PslE-PslJ interaction interface by TM1-2PslJ peptides. Similarly, deletion of the C-terminal tail of PslE resulted in PslD-PslEΔ644-662, which decreased interaction with PslJ (Fig. 6g). Consistent with this, complementation experiments also showed that the PslEΔ644-662 mutant could not rescue Psl production in P. aeruginosa PAO1 (Fig. 6h). Taken together, these results indicate that PslJ may be situated in the membrane lumen formed by the TMH pairs of the PslE octamer, a configuration that presumably stabilized PslE’s C-terminal tail. Discussion In this study, we present two cryo-EM structures of the PslD-PslE complex and the PslD-PslEY524C+R595C complex, where the two Cys mutations stabilized the cytoplasmic tyrosine kinase-like domains of PslE. These structures reveal that the PslD-PslE complex forms a continuous conduit spanning the periplasm and OM, consistent with the reported cryo-EM structures of the Wza-Wzc complex35,36. The length of the overall structure of PslD-PslE is about ~370 Å, which is similar to the length of Wza-Wzc ~360 Å. Furthermore, these structures clearly define the periplasmic thickness as ~230 Å and demonstrate that the OPX protein stabilizes the conformation of its cognate PCP protein. However, several distinct differences exist between the two systems with respect to their complex architectures and regulatory mechanisms. First, in the recently reported cryo-EM structures of the Wza-Wzc complex, the assembly displays dynamic behavior: notably, one Wzc octamer associates with two Wza translocons, and the overall conformations of the PCP protein Wzc differ substantially among distinct Wza-Wzc complexes37. These variations suggest that the Wza-Wzc system is either more dynamically flexible than the PslD-PslE system or subject to more complex regulation by additional accessory factors. Second, previous studies have established that colanic acid biosynthesis (mediated by the Wza-Wzc system) depends on a cyclic phosphorylation-dephosphorylation mechanism that governs the assembly and disassembly of the Wzc octamer20,21. In contrast, our complementation assays using PslE disulfide crosslinking mutants reveal that Psl assembly and export do not require cyclic assembly-disassembly of the full PslE complex. This observation further supports that the Wza-Wzc system is under more complex regulatory control than the PslD-PslE system. In fact, our biochemical data confirm that PslE lacks tyrosine kinase activity, which is consistent with the absence of a Wzb-like phosphatase (a key component of the Wza-Wzc system) in PAO1. Based on our structural and functional analyses of the PslD-PslE complex, we propose a working model for Psl biosynthesis (Fig. 7). Proposed model for Wzx/Wzy-dependent Psl assembly and export. Schematic illustration of Psl biogenesis in P. aeruginosa PAO1, incorporating three key processes: (1) Polymerization-export coupling: PslJ (blue) associates with PslE (purple) within the TM lumen to mediate concurrent Psl chain elongation and translocation; (2) Redirection of aberrant export: PslG (orange) cleaves pentasaccharide linkages to maintain unobstructed flow through the secretion conduit; (3) Channel assembly: PslD (green, OM) and PslE (purple, IM) form a contiguous secretion channel that connects to the bacterial cell surface. PslJ, the Wzy polymerase, is hypothesized to reside within the membrane lumen formed by the octameric PslE’s TMH pairs. Although we did not resolve the atomic-resolution structure of the PslJ-PslD-PslE ternary complex, multiple lines of evidence support this claim. Structurally, our low-resolution cryo-EM map of the PslJ-PslD-PslE complex shows additional density in the membrane lumen of the PslE octamer. The cryo-EM structure of the PslD-PslE complex demonstrates that this membrane lumen is sufficiently spacious to accommodate a single PslJ molecule, which is further supported by AlphaFold-Multimer modeling. Biochemically, inter-subunit disulfide crosslinking between PslJ and PslE, combined with the inhibitory effects of TM1-2PslJ peptide overexpression on Psl production, corroborates their physical interaction. As an IM-resident polymerase, PslJ catalyzes the polymerization of Und-PP-linked pentasaccharide units. The resulting long-chain polysaccharides must be protected from degradation by PslG, a highly active periplasmic endoglycosidase that specifically cleaves Psl pentasaccharide linkages16,17. Our hypothesis also aligns with recent studies on WzyE and WzzE, which suggest that WzyE may be seated within the membrane lumen formed by TMH pairs of WzzE octamer34. The Und-PP-linked pentasaccharide units of Psl are proposed to access PslJ through side portals in the PslE octamer. These portals, delineated by neighboring TMH pairs and the α helix in Motif 2, form oval-shaped channels (24 Å wide × 35 Å high), large enough to permit Und-PP-pentasaccharide entry. The entry mechanism may resemble the ATP-independent LPS uptake by the ABC transporter LptB₂FG37,38. PslG likely acts as a “failsafe” enzyme, removing aberrant Psl that misdirects into the periplasm via two types of surface pores in the PslE octamer: an intra-subunit and an inter-subunit pore, both of which are sufficiently large to permit awry entry of the growing Psl chain into periplasm. In this scenario, PslG may cleave pentasaccharide units from the growing Psl that is exposed to the periplasm, redirecting the growing chain into the periplasmic lumen of the PslD-PslE complex. In addition, our structural and functional data further suggest that Psl assembly and export do not rely on cyclic assembly/disassembly of the entire PslE octamer. Instead, the periplasmic domains of PslE and Wzc both form stable complexes with their respective cognate OPX proteins in the presence of OPX, which is different from the recent structural studies of the Wza-Wzc system in E.coli, the dynamic assembly of Wzc is regulated by the phosphatase Wzb and exists in a monomeric or octameric state when phosphorylated36. Moreover, inter-subunit disulfide-crosslinked PslE mutants (within the tyrosine kinase-like domains or within the periplasmic domains) remain fully functional in Psl biogenesis and biofilm formation (Supplementary Fig. 3b and Supplementary Fig. 9a, b). While local conformational changes in the TMHs and the tyrosine kinase domains of PCP proteins are likely required during polysaccharide synthesis, the overall stability of the complex should be maintained in the process of polysaccharide assembly and export, and the addition of new repeat units to the reducing terminus of the growing chain may drive polysaccharide export through the channel in a coupled process34. In summary, our study presents the cryo-EM structure of an OPX-PCP complex, reconciles prior findings, and provides insights into the mechanisms of Psl assembly and export in P. aeruginosa PAO1. These findings may not only advance our understanding of the Wzx/Wzy-dependent polysaccharide biosynthesis pathway but also offer valuable clues for designing PslJ-derived peptides to target biofilm formation in P. aeruginosa species. Methods Protein expression and purification Two genes, pslD and pslE (including their signal peptide-encoding sequences), were amplified from Pseudomonas aeruginosa (PAO1)39 genomic DNA (ATCC) via PCR. Each gene was individually cloned into the pQLink40 vector using BamHI and PacI restriction sites, with a C-terminal Strep-tag appended to PslE. A summary of the strain and plasmid configurations is provided in Supplementary Table 1-2. The two vectors were then sequentially combined using a ligation-independent cloning method. The genes wza and wzc were also amplified from Escherichia coli K-12 via PCR and cloned into the pQLink vector, with a C-terminal His-tag appended to Wzc. All these constructions were transformed into E. coli strain C43(DE3) for protein expression and cultured overnight in the presence of ampicillin (100 μg/ml). All bacterial cultures were initially grown at 37 °C for 8 h with shaking at 200 rpm, followed by induction with 0.2 mM isopropyl-β-D-thiogalactopyranoside (IPTG) at 20 °C for 12 h. Cells were harvested by centrifugation at 1500 ×g for 15 min. As PslJ does not express in E. coli, to obtain the PslD-PslE-PslJ ternary complex, a fragment containing pslD, pslE, and pslJ, three genes from P. aeruginosa PAO1 was subcloned into pHERD20T41 vector containing an araBAD (arabinose) promoter with a C-terminal His-tag appended to PslJ and a C-terminal Strep-tag appended to PslE. The constructed plasmid was then transformed into P. aeruginosa PAO1 strain and cultured overnight in the presence of carbenicillin (100 μg/ml) for protein expression. All bacterial cultures were initially grown at 37 °C for 8 h with shaking at 200 rpm, followed by induction with 0.1% L-arabinose at 20 °C for 16 h. All purification steps were performed at 4 °C or on ice unless otherwise specified. Cell pellets obtained from low-speed centrifugation were resuspended in buffer A (20 mM HEPES, pH 7.8, 150 mM NaCl, 1 mM PMSF, 10% glycerol). Cell suspensions were lysed using an ultrasonic apparatus at 95% power for 3 min. Cell membranes were collected by centrifugation at 39,000 ×g for 60 min, resuspended in buffer A supplemented with 1% n-dodecyl-β-D-maltoside (DDM, Anatrace), and incubated for 1 h with agitation. Insoluble material was removed by centrifugation at 39,000 ×g for 50 min, and the supernatant was incubated with Strep-Tactin Beads 4FF (Smart-Lifesciences) for 1 h. The beads were washed with 10 column volumes of buffer B (buffer A supplemented with 0.3% glyco-diosgenin [GDN]). The PslD-PslE complex was eluted with buffer C (buffer A plus 5 mM D-desthiobiotin [1169249, Leyan] and 0.01% GDN) and concentrated to 1 ml using a 100-kDa Millipore centrifugal filter (Merck Millipore). The concentrated protein was further purified by size-exclusion chromatography on a Superose 6 Increase 10/300 GL column equilibrated with buffer D (20 mM HEPES, pH 7.8, 150 mM NaCl, 0.006% GDN). Peak fractions were collected, concentrated to 5–6 mg/ml using a 100-kDa filter, and analyzed by 15% SDS-PAGE with a pre-stained protein marker. Cryo-EM sample preparation and data acquisition Cryo-EM grids were prepared using a Vitrobot Mark IV (Thermo Fisher Scientific) under the following conditions: Quantifoil 1.2/1.3 Cu 300-mesh grids were glow-discharged for 90 s under an Ar-O₂ atmosphere. A 3.6-μL aliquot of the concentrated protein sample (5–6 mg/mL) was applied to the glow-discharged grids, allowed to adsorb for 10 seconds, and blotted for 5 s at 4 °C and 100% humidity. Grids were vitrified by rapid plunge-freezing into liquid nitrogen-cooled ethane. Data were acquired using a Titan Krios transmission electron microscope (Thermo Fisher Scientific) operated at 300 kV, equipped with a K3 direct electron detector (Gatan). Movies were collected using EPU software at a calibrated magnification of 105,000× in super-resolution mode, resulting in a physical pixel size of 1.07 Å. The defocus range was set from −1.2 to −2.5 μm. Each movie stack comprised 32 frames with a total accumulated electron dose of 60 e⁻/Ų. Cryo-EM data processing A total of 6,487, 4,787, and 5,614 cryo-EM movies were collected for PslD-PslE (apo), PslD-PslEY524C+R595C, and PslD-PslE-PslJ complexes, respectively. Data processing was performed using cryoSPARC v4.3.042. Movie stacks were motion-corrected using Patch Motion, followed by contrast transfer function (CTF) estimation with Patch CTF. Micrographs exhibiting a resolution better than 6 Å and astigmatism <500 Å were manually selected for further analysis. Particles were manually picked with a 520 Å diameter template, extracted with a 640-pixel box size, and binned 4-fold to 4.28 Å/pixel. Following several rounds of 2D classification to remove contaminants, the remaining particles were re-extracted at the original pixel size (1.07 Å) and subjected to two rounds of ab initio reconstruction into three classes. The class exhibiting canonical transmembrane helix densities was used as an initial reference for heterogeneous refinement of all binned particles. The best class from heterogeneous refinement was re-extracted without binning and processed through additional ab initio reconstruction and non-uniform refinement steps. Final maps were improved via iterative local refinement and 3D classification without alignment using a soft mask, yielding 3D reconstructions at 3.03 and 2.89 Å resolutions for PslD-PslE and PslD-PslEY524C+R595C, respectively. Model building and refinement Initial models for the PslD-PslE complex were generated by rigid-body fitting of AlphaFold3-predicted subunit structures into the EM density using UCSF Chimera v1.16. Atomic models were manually built in COOT v0.9.843 using the cryo-EM density as a guide, with the apo structure serving as a template for all subsequent models. For PslD-PslEY524C+R595C, the model was fitted into the corresponding density map using Chimera and manually adjusted in COOT. Ligand models and geometric restraints were generated using the elBOW module in PHENIX v1.20.144. Final models were refined against respective maps using PHENIX Real_Space_Refine with secondary structure and geometry restraints. Model validation was performed using PHENIX Comprehensive Validation (cryo-EM), assessing clash scores, MolProbity scores, and Ramachandran plot statistics. A summary of the statistics for data collection and structure refinement is provided in Supplementary Table 3. Structural figures were prepared using PyMOL v2.5.045or UCSF ChimeraX v1.646. Polysaccharide extraction and immunoblotting protocol PAO1 strain was cultured in Jensen’s medium supplemented with 2% arabinose and 100 μg/mL carbenicillin at 37 °C overnight with shaking. For crude polysaccharide extraction, approximately eight OD equivalents (volume [mL] = 8/culture OD600) of the overnight culture (~24 h) were resuspended in 100 μL of 0.5 M EDTA, boiled at 100 °C for 10 min, and centrifuged. The supernatant was collected and treated with proteinase K (final concentration 0.5 mg/mL) at 60 °C for 60 min, followed by enzyme inactivation at 80 °C for 30 min. For immunoblotting, 3.5 μL of the polysaccharide extract was spotted onto a nitrocellulose membrane and air-dried. The membrane was blocked with 10% skim milk in TBST (20 mM Tris-HCl, 150 mM NaCl, 0.1% Tween 20, pH 7.6), then probed with α-Psl antibody (1:25,000 dilution in TBST) for 60 min at 25 °C with agitation. After washing, a goat anti-rabbit IgG secondary antibody (1:10,000 dilution in TBST, GE Healthcare) was applied for 45 min at 25 °C. The membrane was washed again and developed using SuperSignal West Dura Extended Duration Substrate (Pierce) according to the manufacturer’s protocol, with chemiluminescence detection performed using enhanced reagents. All the data are processed by ImageJ. Attachment assay Test strains were cultured overnight in LB medium at 37 °C, then diluted 1:100 in Jensen’s medium (100 μg/mL carbenicillin, 2% L-arabinose) and shaken at 37 °C until reaching an OD600 of approximately 1.0. Subsequently, 100 μL aliquots of the bacterial suspension were transferred into 96-well PVC microtiter dish (BD Falcon) and statically incubated at 30 °C for 40 min. The non-adherent and loosely attached cells were then removed by washing with sterile water. Adherent cells firmly attached to the well surfaces were subsequently stained with 0.1% crystal violet for visualization. Complementation experiments The pslE-knockout and pslD-knockout PAO1 strains were kindly provided by Professor Luyan Z Ma, as described by Byrd, MS et al.47. The in vivo functional analysis of each mutant construct was performed by evaluating their ability to complement pslE deficiency in a P. aeruginosa PAO1 pslE chromosomal knockout mutant. The PAO1 ΔpslE strain harbors an arabinose-inducible promoter driving endogenous PslE expression. This strain was transformed with pHERD20T-pslE plasmids encoding mutant pslE variants, selected with carbenicillin (100 μg/mL), and plated on LB-agar containing 0.2% L-arabinose. Plates were incubated at 37 °C for 12 h. Single colonies were inoculated into 5 mL LB broth supplemented with 0.2% arabinose and 100 μg/mL carbenicillin. When cultures reached an OD₆₀₀ of 1.0, cells were harvested by centrifugation, washed twice with LB, and resuspended to a uniform starting OD₆₀₀ of 2.5 and samples were prepared for Psl extraction as described above. Each pslE and pslD mutant complementation assay was repeated in triplicate, and representative results are presented. Western blot analysis To compare the relative expression levels of wild-type pslJ and its mutants, PAO1 cells were transformed with plasmids encoding either wild-type or mutant pslJ constructs (all harboring a C-terminal His tag). Transformed cells were plated on LB agar supplemented with 100 μg/mL carbenicillin and 0.2 mg/mL L-arabinose, then incubated at 37 °C for 12 h. Colonies were harvested, resuspended in 1× PBS, and adjusted to an OD₆₀₀ of 1.0. Cell lysates (30 μL) were denatured at 95 °C for 10 min prior to SDS-PAGE analysis. Following electrophoresis, proteins were transferred to a PVDF membrane and blocked with TBST buffer (20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.1% Tween 20) containing 5–10% skim milk for 1 h. The membrane was probed with an anti-His antibody (1:3,000 dilution, AB102-01, Tiangen Biotech; validation available on manufacturer’s website) for 1 h at room temperature, washed thrice with TBST, and incubated with a horseradish peroxidase-conjugated secondary antibody (1:5,000 dilution, DW101, Trans) for 1 h at room temperature. Protein bands were visualized using an enhanced chemiluminescence kit (EasySee Western Blot Kit, Trans). Lectin staining protocol Psl matrix trails produced by bacteria were stained with HHA (Hippeastrum hybrid agglutinin) lectins at final concentrations of 100–200 μg/mL. For lectins and FM464 double staining, the lectins-stained surface-attached bacterial cells or biofilm was stained with FM4-64. Biofilms were first labeled with HHA-FITC (fluorescein isothiocyanate), followed by counterstaining with FM4-64 (1 μM final concentration, Molecular Probes, Invitrogen) for 10 min. Imaging was performed using a confocal laser scanning microscope (CLSM) with fluorescein isothiocyanate (FITC) and tetramethylrhodamine isothiocyanate (TRITC) optical filters. Overnight cultures were diluted to an OD₆₀₀ of 0.8, and 0.5 mL of the diluted culture was inoculated into a cover glass chamber. Surface-attached bacteria were allowed to adhere for 1 h at room temperature before lectin staining. Stained bacterial cells were visualized by CLSM, with all samples maintained in PBS buffer during image acquisition. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting Summary Transparent Peer Review file Source data Source Data Acknowledgements The authors would like to thank the staffs from Center for Biological Imaging (CBI), Institute of Biophysics, Chinese Academy of Sciences (CAS), and Professor Yan Zhao for help on cryo-EM data collection. Author contributions Y.H. supervised the project. J.L., Y.L.H. and L.X.W. performed all the experiments, cryo-EM sample preparation, data collection, processing, and model building: J.L. and Y.L.H., Structure analysis; D.C.N., Lectin staining analysis: L.X.W. and L.Y.M. All the authors wrote the manuscript. Peer review Peer review information Nature Communications thanks Bert van den Berg, and the other, anonymous, reviewer for their contribution to the peer review of this work. A peer review file is available. Funding This work is supported by grants from National Key Research and Development Project of the Ministry of Science and Technology, China (2024YFC3406201 to Y.H.), the Major project of Guangzhou National Laboratory (GZNL2024A01006), Scientific Foundation for Youth Scholars of Shenzhen University (000008021104 to D.C.N.) and the International Partnership Program of the Chinese Academy of Sciences Grant (079GJHZ2023029MI to L.M.). Data availability The cryo-EM maps of PslD-PslE, PslD-PslEY524C+ R595C, PslD-PslE-PslJ and their corresponding atomic coordinates, have been deposited in the Electron Microscopy Data Bank and the PDB under accession codes EMD-64802 and 9V6L, EMD-64810 and 9V75, EMD-66623 and 9X6P respectively. Source data are provided with this paper. Competing interests The authors declare that they have no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Jiao Liu, Yanlong Han. Contributor Information Dongchun Ni, Email: dongchunni@szu.edu.cn. Yihua Huang, Email: yihuahuang@ibp.ac.cn. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-74497-9. References Associated Data Supplementary Materials Supplementary Information Reporting Summary Transparent Peer Review file Source Data Data Availability Statement The cryo-EM maps of PslD-PslE, PslD-PslEY524C+ R595C, PslD-PslE-PslJ and their corresponding atomic coordinates, have been deposited in the Electron Microscopy Data Bank and the PDB under accession codes EMD-64802 and 9V6L, EMD-64810 and 9V75, EMD-66623 and 9X6P respectively. Source data are provided with this paper.