Translocation of Group 1 Capsular Polysaccharide in Escherichia coli Serotype K30

The late steps in assembly of capsular polysaccharides (CPS) and their translocation to the bacterial cell surface are not well understood. The Wza protein was shown previously to be required for the formation of the prototype group 1 capsule structure on the surface of Escherichia coli serotype K30 (Drummelsmith, J., and Whitfield, C. (2000) EMBO J. 19, 57–66). Wza is a conserved outer membrane lipoprotein that forms multimers adopting a ringlike structure, and collective evidence suggests a role for these structures in the export of capsular polymer across the outer membrane. Wza was purified in the native form and with a C-terminal hexahistidine tag. WzaHis6 was acylated and functional in capsule assembly, although its efficiency was slightly reduced in comparison to the native Wza protein. Ordered two-dimensional crystals of WzaHis6 were obtained after reconstitution of purified multimers into lipids. Electron microscopy of negatively stained crystals and Fourier filtering revealed ringlike multimers with an average outer diameter of 8.84 nm and an average central cavity diameter of 2.28 nm. Single particle analysis yielded projection structures at an estimated resolution of 3 nm, favoring a structure for the WzaHis6 containing eight identical subunits. A derivative of Wza (Wza*) in which the original signal sequence was replaced with that from OmpF showed that the native acylated N terminus of Wza is critical for formation of normal multimeric structures and for their competence for CPS assembly, but not for targeting Wza to the outer membrane. In the presence of Wza*, CPS accumulated in the periplasm but was not detected on the cell surface. Chemical cross-linking of intact cells suggested formation of a transmembrane complex minimally containing Wza and the inner membrane tyrosine autokinase Wzc.

In Gram-negative bacteria, macromolecules destined for the cell surface or the extracellular environment must cross both the inner (IM) 1 and the outer membrane (OM). In protein export, where the processes are arguably best understood, secretion systems with varying complexity accomplish the translocation steps; all involve multi-enzyme complexes where an outer membrane protein (channel) is linked directly, or via helper proteins, to IM components.
Cell-associated capsular polysaccharides (CPS) and their secreted (cell-free) counterparts, exopolysaccharides (EPS), represent another type of macromolecule that must be transported across the cell envelope. The early steps in assembly of these polymers are reasonably well established. However, there is little understanding of the terminal steps, including the mechanism by which they cross the bacterial cell envelope and the machinery involved.
In Escherichia coli, more than 80 antigenically distinct capsular (K) polysaccharide structures are recognized. They are divided into four groups based on the organization of their genetic loci, polymerization mechanisms, and regulation (1). Group 1 and 2 capsules have received the most attention, and they involve biosynthetic mechanisms that are conserved in other bacteria. Translocation of group 1 and 2 CPS occurs at specific sites where the IM and OM are in close apposition (so called "Bayer junctions" or zones of adhesion) (2)(3)(4), providing indirect evidence for transmembrane complexes in capsule assembly. Translocation of group 2 capsules is proposed to involve coordinated synthesis and export (1,(5)(6)(7). In the current model, group 2 CPS is polymerized in the cytoplasm by a multi-enzyme membrane-bound complex and then modified with a phospholipid anchor. Next, the nascent CPS is exported across the IM by a member of the ABC-2 family of ATP-binding cassette transporters, before being transported across the OM by a currently unidentified channel protein. Our laboratory is interested in the biosynthesis of the E. coli K30 CPS, the established prototype for group 1 capsules. A model representing the current thinking about the assembly mechanism is presented in Fig. 1. Individual group 1 K-antigen repeat units are assembled on an undecaprenol phosphate carrier lipid at the cytoplasmic face of the IM and are then proposed to be transferred across the IM by a "flippase" Wzx. The majority of the repeating units serve as substrate for polymerization in a process minimally requiring the IM protein, Wzy, and resulting in high molecular weight CPS (8). However, a small fraction of the repeat units can also serve as substrate for the O-antigen ligase WaaL, ligating the repeating unit onto the lipopolysaccharide (LPS) lipid A-core. This LPS-linked form of the group 1 antigens is called K LPS (9). Formation of the high molecular weight capsular form of K antigen requires four additional highly conserved proteins (Wzi (formerly OrfX), Wza, Wzb, and Wzc), encoded by all group 1 capsule gene loci (i.e. independent of CPS structure) in E. coli and Klebsiella pneumoniae (10).
Wzi is a monomeric ␤-barrel OM protein acting late (after translocation) in the assembly pathway. In the absence of Wzi, K30 polymer is still made and exported but much of it is secreted into the surrounding medium, instead of being elaborated into a coherent capsular structure on the cell surface. Wzi is therefore involved at some level in linking CPS to the cell surface, rather than functioning in polymerization or capsule translocation (11). Wzc is an IM protein with tyrosine autokinase activity, and the cytosolic Wzb enzyme is its cognate phosphatase (12). Deletion of either Wzc or Wzb abolishes the ability of the bacteria to make a K30 capsule but not K30 LPS (which serves as an independent measure of synthesis of K30 repeat units) (8,12). The extent of phosphorylation of Wzc also affects the amount of K30 CPS (13). Despite the information now available concerning the biochemical properties of Wzc proteins in E. coli K30 (8,12,13) and in E. coli K-12 (14 -16), the precise role(s) of Wzc in capsule assembly has yet to be resolved. Wza is a conserved OM lipoprotein that forms multimers with a low resolution ringlike structure identified in electron microscopy (17). The Wza protein is required for the formation of the group 1 capsule structure on the surface, but mutants deficient in Wza can still form K30 LPS (8,17). These properties collectively suggest a role for Wza in export of capsular polymer across the OM, but the absence of capsule or any accumulated high molecular weight periplasmic polymer in wza mutants make it difficult to assign a definitive role to this protein in the assembly pathway. An attractive explanation for the inability of wza and wzc mutants to form any high molecular weight polymer is that high level Wzy-dependent polymerization and translocation are linked processes in an enzyme complex; a defect in any component of the complex would impact preceding steps.
In this report we further characterize the structure and function of Wza in E. coli K30. The secondary structure of Wza was investigated, and electron crystallography revealed an octameric quaternary structure. A non-acylated Wza derivative (Wza*) was found to target to the OM but was unable to form stable multimers, leading to a capsule-assembly pathway defect in which CPS was retained in the periplasm. Chemical cross-linking was used to show an association between Wza and Wzc. The collective data provide the first direct evidence for the involvement of Wza in translocation and for a transmembrane complex necessary for coordinated synthesis and surface assembly of group 1 capsules.
To obtain a Wza derivative that is not acylated but is processed, the lipoprotein consensus signal sequence and N-terminal 5 amino acid residues of mature Wza were replaced by the corresponding sequence from OmpF of E. coli K-12. Two complementary oligonucleotides providing the replacement sequence (OmpFSS1, 5Ј-AATTCACCATGATG-AAGCGCAATATTCTGGCAGTGATCGTCCCTGCTCTGTTAGTAGCA-GGTACTGCAAACGCTGCAGAAATCTATAACG-3Ј; and OmpFSS2, 5Ј-TCGACGTTATAGATTTCTGCAGCGTTGCAGTACCTGCTACTAACA-GAGCAGGGACGATCACTGCCAGAATATTGCGCTTCATCATGGTG-3Ј) were annealed giving terminal EcoRI and SalI overhangs. The fragment was ligated into pBAD24 (20) digested with EcoRI and SalI to obtain pWQ301. The region of the wza open reading frame corresponding to amino acids 8 -359 of the mature Wza protein was amplified by PCR from E. coli E69 using primers wzaaa8SalI (5Ј-GTACGCGTCGA-CGGATTGAACAGTCTGCGTAAA-3Ј) and wzaHindIII (5Ј-CTCCCAA-GCTTTTAGTTTGGCCATCTCTCTTAATGTAT-3Ј); introduced SalI and HindIII sites underlined. The SalI-and HindIII-digested PCR product was ligated into SalI/HindIII-digested pWQ301. The resulting plasmid, pWQ302, expresses a chimeric mature Wza* protein comprising the N-terminal 5 amino acid residues of OmpF plus 2 additional Step B, the undecaprenol pyrophosphate-linked repeat units are flipped across the IM by a process involving Wzx.
Step C, the repeat units are polymerized in a Wzy-dependent reaction.
Step D, Wzc function is essential for high level polymerization. To be active in capsule assembly, Wzc must undergo autophosphorylation, followed by transphosphorylation between proteins in an oligomeric form.
Step E, dephosphorylation of Wzc by the Wzb phosphatase is also crucial for capsule synthesis.
Step F, export of polymer to the surface requires the OM Wza multimeric complex, perhaps playing the role of an export channel.
Step G, Wzi is required for efficient assembly of the capsular layer residues (derived from the introduced SalI site) and Wza beginning at amino acid residue 8. To obtain a Wza* derivative containing a Cterminal His 6 tag, plasmid pWQ302 was digested with NcoI and EcoRI and the 750-bp fragment was ligated to the 4.9-kb NcoI-EcoRI fragment of pWQ300 to give pWQ303.
A C-terminal truncation derivative of Wza (Wza ⌬348 -359 ) was also constructed, in which the last 12 amino acid residues were deleted. Truncation of wza was achieved by PCR amplification of wza from E. coli E69 using primer pairs JD125 (17) and JN20 5Ј-AACTGCAGT-TACATATCATGAACACCTGATATTG-3Ј (PstI site is underlined, stop codon is in bold). The PCR fragment was cloned into pBAD24 using the PstI and EcoRI restriction sites to give pWQ304.
Plasmid pWQ305 expresses an N-terminal hexahistidine-tagged Wzc derivative under the control of the arabinose-inducible pBAD promoter. To construct the plasmid, an NcoI-SalI fragment encoding Wzc His 6 was isolated from the pET-based plasmid pWQ141 (12) and cloned into pBAD24 (20) to generate pWQ305. In some experiments, native Wzc was used. This was expressed from plasmid pWQ130, a pBAD18-Km derivative described previously (12).
To maintain the plasmids expressing Wza and Wzc His 6 together in E. coli LE392, the origin and antibiotic resistance cassette of the native Wza-expressing plasmid, pWQ126, was exchanged. Plasmid pWQ126 was digested with ClaI and ScaI, and the 3.3-kbp fragment containing araC and P BAD wza was ligated to pACYC184 previously digested with EcoRV and ClaI. This resulted in plasmid pWQ306 with a chloramphenicol resistance marker and Wza expression under control of the P BAD promoter.
The nucleotide sequences of all PCR-amplified products were verified using the service at the University of Guelph Molecular Supercenter.
Protein Purification-Proteins were overexpressed by growing cells containing the arabinose-inducible pBAD24 derivatives to an A 600 of 0.5 and inducing for 3 h at 37°C following addition of 0.006% (final concentration) L-arabinose. Cells were harvested, resuspended in 20 mM sodium phosphate buffer, pH 7.0, and lysed by two passages through a French pressure cell. Cell envelopes were isolated in a pellet following ultracentrifugation for 1 h at 100,000 ϫ g, and the IM was solubilized in 20 mM sodium phosphate buffer, pH 7.0, containing 2% N-laurylsarcosinate for 1 h at room temperature. The insoluble OM pellet collected after centrifugation at 100,000 ϫ g for 1 h was resuspended in 50 mM sodium phosphate buffer, pH 7.0, containing 0.5% SB 3-14 and 150 mM NaCl (for the native Wza protein), or 20 mM sodium phosphate buffer, pH 7.0, containing 0.5% SB 3-14 and 500 mM NaCl (for His 6 -tagged derivatives). After solubilization overnight at room temperature, the remaining insoluble material was removed by centrifugation at 100,000 ϫ g for 1 h.
For purification of His 6 -tagged proteins, the supernatant was loaded onto a 5-ml HiTrap chelating HP column (Amersham Biosciences) equilibrated with 20 mM sodium phosphate buffer, pH 7.0, containing 0.05% SB 3-14 and 500 mM NaCl. His 6 -tagged proteins were eluted with a step gradient using the same buffer containing 1 M imidazole. Detergent exchange was achieved on an anion-exchange column (5-ml Econo-Pac High Q; Bio-Rad) washed with at least 100 ml of 20 mM Tris-HCl, pH 8.0, containing 50 mM NaCl and either 0.008% DDM (n-dodecyl ␤-D-maltoside; Sigma) or 0.3% octyl-polyoxyethylene (Bachem). Protein was eluted with 1 M NaCl in the same buffer.
Wza was purified by applying the OM extract to the anion-exchange column equilibrated with 50 mM sodium phosphate buffer, pH 7.0, containing 150 mM NaCl and 0.05% SB 3-14. Wza-containing fractions were eluted with 50 mM sodium phosphate buffer, pH 7.0, containing 250 mM NaCl and 0.05% SB 3-14 and reloaded to the anion exchange column equilibrated with 20 mM Tris-HCl, pH 9.5, containing 50 mM NaCl and 0.008% DDM. After washing with 100 ml of the same buffer, Wza was eluted with a gradient of 0.05-1 M NaCl. Fractions containing Wza were applied to a hydroxyapatite column (5-ml Econo-PacCHT-II, Bio-Rad) equilibrated with 50 mM sodium phosphate buffer, pH 7.0, containing 0.008% DDM and the protein was eluted with a linear gradient of 0.05-1 M sodium phosphate buffer, pH 7.0, containing 0.008% DDM. To concentrate Wza, the eluted protein was rebound to the anion exchange column equilibrated with 20 mM Tris-HCl, pH 7.5, containing 0.008% DDM. Elution was performed using the same buffer containing 1 M NaCl.
Between the different purification steps, buffer was exchanged by using a HiPrep 26/10 desalting column (Amersham Biosciences) or by dialysis. All chromatography steps were performed on a AKTA Explorer 100 system (Amersham Biosciences).
Separation of Inner and Outer Membranes-Whole membrane fractions obtained as described above were resuspended in TE (10 mM Tris-HCl, pH 8.0, 1 mM EDTA) and fractionated on a discontinuous sucrose gradient (72%/49%/26% sucrose in TE) to separate the inner and outer membranes (21). The gradient was centrifuged at 100,000 ϫ g for 18 h in a swinging bucket rotor. The IM and OM fractions generated visible bands in the gradients that were collected individually. Western immunoblotting with anti-Wzc polyclonal serum (12) was used to verify purity of the OM; Wzc is an exclusively IM protein, and the OM fractions showed no reactivity.
In Vivo Cross-linking-Cross-linking experiments were performed essentially as described elsewhere (22). Briefly, E. coli CWG281 (pWQ130-pWQ300), CWG281(pWQ130-pWQ126), LE392(pWQ305-pWQ306), and LE392(pWQ306) were grown to mid-logarithmic phase (A 600 ϭ 0.5). To induce expression of proteins, arabinose was added to a final concentration of 0.02% and the cultures were incubated for an additional 3 h. Cultures were adjusted to give A 600 ϭ 1, washed, and concentrated 10-fold in 20 mM sodium phosphate buffer, pH 7.2, containing 150 mM NaCl. The chemical cross-linker dithiobissuccinimidyl propionate (DSP, Pierce) was added to a final concentration of 1 mM, and the mixture was incubated for 20 min at room temperature before the reaction was quenched with 100 mM Tris-HCl, pH 7.5 (final concentration). The treated cells were washed once in 20 mM Tris-HCl, pH 7.5, and pellets were frozen overnight.
Purification of Hexahistidine-tagged Protein Complexes-Pellets of cross-linked whole cells were thawed in 20 mM Tris-HCl, pH 7.5, and lysed by two passages through a French pressure cell, and cell envelopes were isolated as described above. The cell envelope pellet containing cross-linked complexes was then solubilized for 16 h at room temperature in solubilization buffer (10 mM Tris-HCl, pH 8.0, containing 100 mM sodium phosphate, 300 mM NaCl, 8 M urea, 0.5% SDS, 0.25% n-dodecyl ␤-D-maltoside). Insoluble material was removed by centrifugation for 1 h at 100,000 ϫ g and the supernatant mixed with Ni-NTAagarose (Qiagen) to bind cross-linked complexes. In experiments involving Wzc His 6 , binding of residual non-cross-linked Wzc His 6 presented a potential problem. To avoid this complication, the relevant cell envelope pellet was first solubilized in 20 mM Tris-HCl, pH 8.0, containing 2% N-laurylsarcosinate. After incubation for 1 h at room temperature and centrifugation at 100,000 ϫ g for 1 h, the supernatant containing solubilized IM (23), including non-cross-linked Wzc His 6 , was discarded. The residual pellet enriched in OM was then solubilized in n-dodecyl ␤-D-maltoside as described above. Binding to Ni-NTA-agarose was allowed to proceed for 16 h at room temperature on a rotating mixer. The resin was then washed extensively with 10 mM Tris-HCl, pH 6.3, containing 100 mM sodium phosphate, 8 M urea, and 0.5% SDS followed by a second wash step using 20 mM Tris-HCl, pH 8.0, containing 250 mM NaCl, 4 M urea, 0.5% SDS, and 5 mM imidazole. Bound proteins were eluted from the resin with PAGE gel loading buffer (NuPAGE, Invitrogen, containing no reducing agents) supplemented with 0.4 M imidazole and 50 mM Tris-HCl, pH 6.8 (final concentration), by incubating first at room temperature for 10 min, followed by 100°C for 5 min. The samples were then split, and half received dithiothreitol (DTT) to give a final concentration of 40 mM, to cleave the cross-links. Samples with or without DTT were incubated for 10 min at 100°C prior to separation on an 8% PAGE gel.
Affinity Purification of Anti-Wza Antiserum-OM extracts containing overexpressed Wza His 6 were solubilized overnight in 20 mM HEPES, pH 7.4, containing 300 mM NaCl and 0.5% SDS. The supernatant obtained after ultracentrifugation was mixed with Ni-NTA-agarose (Qiagen), incubated for 3 h at room temperature, and then transferred into an Econo-Pac column. The column was extensively washed in sequence with 10 mM Tris-HCl, pH 6.3, containing 0.1% SDS and 500 mM NaCl followed by 10 mM Tris-HCl, pH 8, containing 0.1% SDS and 5 mM imidazole. Finally, the column was equilibrated with 10 bed volumes of 20 mM Tris-HCl, pH 8.0, containing 28 mM NaCl. To remove serum proteins, the Wza K30 polyclonal antiserum (8) was first purified by chromatography on an Econo-Pac Serum IgG purification column (Bio-Rad). Fractions containing Wza antibodies were eluted with 20 mM Tris-HCl, pH 8.0, containing 28 mM NaCl and loaded onto the equilibrated Wza His 6 /Ni-NTA-agarose column. After sequentially washing with 10 mM Tris-HCl, pH 7.5, and 10 mM Tris-HCl, pH 7.5, containing 500 mM NaCl, the Wza-specific antibodies were eluted with 4 M MgCl 2 . Elution fractions were separately dialyzed against 20 mM Tris-HCl, pH 8, containing 28 mM NaCl and tested for specificity in Western immunoblot analysis.
Purified Wza and Wza His 6 proteins were subjected to mass spectrometry analysis at the University of Guelph biological mass spectrometry facility. N-terminal sequencing of purified Wza* protein was performed by Edman degradation at the University of British Columbia, Biotechnology Laboratory-Nucleic Acids Protein Service facility. Circular dichroism (CD) spectra of purified Wza (0.5 mg/ml in 20 mM Tris-HCl, pH 7.5, containing 80 mM NaCl and 0.008% DDM) and Wza His 6 (1 mg/ml in 20 mM Tris-HCl, pH 7.5, containing 0.3% octyl-polyoxyethylene) were performed at the Scottish circular dichroism facility at Glasgow (Scotland, United Kingdom).
Two-dimensional Crystallization of Wza His 6 -Two-dimensional crystallization of Wza His 6 was carried out according to the method of Levy et al. (25). Briefly, 1 g of purified Wza His 6 in 20 mM Tris-HCl, pH 8.0, containing 625 mM NaCl, 0.4% SB 3-14, and 0.2 g of E. coli total lipids (Avanti Polar Lipids) was incubated under a functionalized lipid monolayer (1:1 E. coli total lipids/1,2-dioleoyl-sn-glycero-3-[N-(5-amino-1carboxypentyl)iminodiacetic acid)succinyl] (nickel salt) (Avanti Polar Lipids). Binding of the hexahistidine tags to the nickel ions presented by these functionalized lipids leads to concentration of Wza His 6 at the monolayer surface. Detergent was then removed by the addition of polystyrene BioBeads® (Bio-Rad). It is proposed that this results in the replacement of detergent molecules with lipids from the aqueous phase, reconstituting the protein within a bilayer matrix (25,26). A range of incubation times and temperatures was evaluated. The best planar arrays were obtained after 24 h of incubation at room temperature.
Electron Microscopy and Processing of Two-dimensional Crystal Images-The two-dimensional crystals were transferred to fenestrated cellulose acetate butyrate-coated copper grids and stained with 2% (w/v) uranyl acetate. Electron microscopy was performed at the University of Guelph Natural Sciences and Engineering Council of Canada Regional Scanning and Transmission Electron Microscopy facility using a LEO912AB transmission electron microscope (LEO GmbH, Oberkochen, Germany) at an accelerating voltage of 100 kV and nominal magnifications of either ϫ20,000 or ϫ40,000. Digital image data were collected using an EsiVision CCD-BM/1k SSCCD camera. Imaging was routinely performed under low electron dose conditions. Images were analyzed by either Fourier filtering or single-particle analysis (1591 individual particles) with the IMAGIC-V electron image processing system (27,28).
Examination of Capsular Phenotype by Electron Microscopy-The different Wza variants were expressed in the wza mutant E. coli CWG281. Overnight cultures were diluted and plated on Luria-Bertani agar plates containing gentamicin, ampicillin, and 0.02% arabinose and incubated for 18 h at 37°C. Alternatively, the cultures were subcultured, grown to an A 600 ϭ 0.5 and expression of the Wza derivatives were induced by growth for 18 h following addition of 0.02% arabinose. Capsule formation was examined by electron microscopy of thin sections. Cells were stained with cationized ferritin (11), or thin sections was immunolabeled using a monoclonal anti-K30 antibody (29) and a gold-conjugated anti-mouse IgG (Sigma). Preparation of thin sections and examination by EM were done at the University of Guelph Natural Sciences and Engineering Research Council of Canada Regional Scanning and Transmission Electron Microscopy facility.
Cell-surface Polysaccharide Analysis-Cell surface polysaccharides were prepared by the proteinase K digestion method of Hitchcock and Brown (30), separated on 8% polyacrylamide SDS gels and subjected to Western immunoblot analysis as described elsewhere (8). K30 polysaccharide was detected with a rabbit anti-K30 polyclonal antiserum and an alkaline phosphatase-conjugated goat anti-rabbit secondary antibody.
Bacteriophage Techniques-Plaque assays were used to determine the sensitivity of bacteria to phage K30 (specific for the serotype K30 capsular polymer; Ref. 31) and phage O9a (specific for the serotype O9a LPS; Ref. 32). 0.1-ml aliquots of overnight cultures of E. coli CWG281 containing different pBAD24 derivatives were added to 5 ml of TB soft agar, poured onto Luria-Bertani agar plates supplemented with gentamicin, ampicillin, and different concentrations of L-arabinose (0 -0.2%). 0.01-ml aliquots of 10-fold dilutions (10 0 to 10 Ϫ6 ) of phage lysates were spotted onto the inoculated plates, and the results were read after incubation for 6 h at 37°C. 6 Is Functional in Capsule Assembly-The OM lipoprotein Wza is encoded by the K30 CPS biosynthesis (cps) locus and exists as a multimer forming a ringlike shape with a central cavity (17). To aid purification, a derivative with a C-terminal hexahistidine tag was constructed. The Wza His 6 protein was expressed at a level comparable with native Wza and complemented a wza mutation in E. coli CWG281 to restore expression of the K30 antigen ( Fig. 2A). In Western immunoblots, the amount of K30 antigen produced by the strain complemented by Wza His 6 was slightly less than that seen with the native Wza protein. This was consistent with thin section electron micrographs that revealed a corresponding slight reduction in the amount of capsule on the cell surface (Fig. 2B). However, Wza His 6 is clearly functional in CPS assembly.

Wza His
The reduction in capsule assembly directed by Wza His 6 in E. coli CWG281 could reflect either steric problems arising from the C-terminal hexahistidine tag, or an essential requirement for specific C-terminal residues. To provide further insight, a C-terminal deletion derivative (Wza ⌬348 -359 ) was constructed. This derivative was expressed and generated near wild-type levels of K30 CPS in E. coli CWG281 (Fig. 2A). The capsule on the cell surface was indistinguishable from that resulting from activity of the native Wza (Fig. 2B), indicating that the terminal 12 residues of Wza are not essential for its function.
Purification and Properties of Wza and Wza His 6 -The Wza and Wza His 6 proteins were purified from solubilized OM extracts by chromatography (Fig. 3). Both proteins were obtained with Ͼ98% purity (based on Coomassie Blue-stained SDS-PAGE gels). The purified proteins were analyzed by mass spectrometry (data not shown), and the observed molecular masses corresponding to the main peaks were 40,240.6 Da for Wza K30 and 41,077.1 Da for Wza His 6 . These values are consistent with the calculated theoretical molecular mass of the respective mature lipoproteins, modified at the N-terminal cysteine with diacylglycerol and palmitate, supporting previous labeling studies (17). Wza and Wza His 6 both showed the typical SDSresistant multimers reported previously (17) (Fig. 3), suggesting that the C-terminal hexahistidine tag did not influence folding and assembly into multimers in any detectable way and consistent with the functionality of Wza His 6 in CPS assembly.
Wza is a member of the outer membrane auxiliary protein family (33). These proteins are proposed to be ␤-barrel channel proteins that transport polysaccharides across the OM (1, 33). To provide insight into the secondary structure of Wza, purified Wza and Wza His 6 were subjected to CD analysis. The spectra of the proteins were highly similar with a minimum at 208 nm (data not shown), indicating that both proteins are folded and that Wza does not comprise solely ␤-structure. Analysis of the CD spectra for secondary structure using the SELCON procedure (34) gave the following approximate values for Wza: 26% ␣-helices, 14% antiparallel ␤-sheets, 4% parallel ␤-sheets, 22% turns, and 33% other structures. The corresponding recorded values for Wza His 6 were: 19%, 23%, 3%, 21%, and 34%, respectively. The small differences in the deduced structures might be because of the presence of hexahistidine tag in Wza His 6 or differences in the concentration or buffer composition.
Electron Crystallography of Wza His 6 Reveals an Octameric Structure-Initial electron microscopy of purified Wza His 6 revealed the same multimeric ringlike structures, mainly arranged in arrays, as observed previously with native Wza (17). Purified Wza His 6 produced ringlike structures indistinguishable from the native Wza and also formed higher order aggregates that gave no significant structural information by electron microscopy (data not shown). The lack of any discernible structural alterations in Wza His 6 allowed the exploitation of the hexahistidine tag in electron crystallographic approaches to gain more detailed two-dimensional information concerning the quaternary structure of the multimer. In this approach, Wza His 6 was first bound to a lipid monolayer containing a mixture of nickel-chelating lipid and total E. coli lipids, and then reconstituted into lipid bilayers upon removal of the detergent by BioBeads (25,26). The two-dimensional crystals were transferred from the air-water interface to a plasticcoated grid, negatively stained, and examined by transmission electron microscope. Although planar arrays of Wza His 6 complexes with varying degrees of regularity were routinely obtained, they frequently did not withstand transfer to the grid without some disruption. A representative array is shown in Fig. 4A (recorded at 20,000ϫ magnification). This array diffracted to 2 orders with the outermost spot corresponding to a spatial resolution of 4.2 nm (Fig. 4B). Fourier filtering of a representative crystalline array yielded an image showing the square lattice more clearly (Fig. 4C). A segment of the filtered image is shown enlarged in Fig. 4D. The lattice spacing was ϳ10 nm. The outer diameters of the individual ring structures were 8.84 Ϯ 0.43 nm, and the central cavity had a diameter of 2.28 Ϯ 0.28 nm. These crystals were reproducible, but better crystalline order could not be obtained under any of the various conditions evaluated.
To obtain structural detail to higher resolution, a series of planar arrays of Wza His 6 were imaged at higher magnification (40,000ϫ), and 1591 individual unit cells were extracted from the micrographs for single particle analysis. The latter process comprised two cycles of unbiased multireference alignment, multivariate statistical analysis, and classification (35). This approach allows the decomposition of the image data set into groups (or classes) that are most similar to one another. Averaging the aligned images within each class yields a result with an enhanced signal-to-noise ratio. Here, the best six class averages (in terms of internal homogeneity) are shown in Fig. 4E with isodensity contour lines superimposed. The variability between classes is the result of conformational variability of individual unit cells, which explains the limited order proffered by the two-dimensional crystals. Although the class-averaged images in Fig. 4E are of better resolution than those in Fig. 4D, exceeding 3 nm, some major peaks were blurred. Nevertheless, the images, in conjunction with the rectangular symmetry of the planar crystals, are entirely consistent with a model in which each ring consists of 8 individual Wza His 6 subunits.

N-terminal Acylation Is Critical for the Formation of Functional Wza Multimers-
The role of acylation of the N-terminal cysteine of Wza in multimer formation and CPS assembly was investigated. A modified derivative, Wza*, was constructed by replacing the signal sequence of Wza with that of the outer membrane porin protein, OmpF, from E. coli. To provide a signal sequence recognized by the signal peptidase I, the 5 N-terminal residues of Wza were also replaced by those from OmpF in mature Wza*. To assess expression of Wza*, whole cell lysates were probed in Western immunoblots using anti-Wza antibody (Fig. 5A) and the protein was expressed. The presence of two bands in the blot is indicative of both precursor and processed forms of Wza* present in whole cell lysates, as often is the case for native Wza (17). In qualitative evaluations of Western immunoblots, the amount of Wza* expressed in CWG281 was typically less than the amount of native Wza or Wza His 6 in the same background. However, Western immunoblot signal obtained with expression of all three derivatives exceeded that seen from chromosomal copy (data not shown).
Expression of Wza* in E. coli CWG281 led to restored synthesis of K30 antigen, evident in Western immunoblots of the appropriate whole cell lysates (Fig. 5A). However, the amount of immunoreactive K30 polymer was reduced and the polymer showed a reduction in apparent molecular mass in PAGE analysis, compared with the control. Surface expression of the K30 antigen is routinely assessed by examining the sensitivity of the organism to bacteriophages specific for the capsular K30 antigen and the LPS O9a antigen (8); the O9a receptor is masked when a capsule structure is elaborated and exposed if the capsule is either reduced in amount or absent. Bacteria expressing Wza were only infected by bacteriophage K30, as anticipated from the data in Fig. 1. In contrast, E. coli CWG281 expressing Wza His 6 that produces a reduced amount of CPS is sensitive to both bacteriophages, indicative of the reduced barrier provided by the capsule in this strain. Surprisingly, E. coli CWG281 expressing Wza* was sensitive to phage O9a but resistant to phage K30. This phenotype is identical to defined acapsular mutants and suggested that no CPS was assembled on the cell surface of Wza* expressing cells. This conclusion was confirmed by electron micrographs of cationized ferritinstained cells that showed no visible CPS on the surface of E. coli CWG281 cells expressing Wza* (Fig. 5B).
To further investigate the discrepancy between the Western immunoblotting data showing K30 polymer synthesis and the apparent absence of a capsule structure on the cell surface, the K30 antigen was localized by immunogold electron microscopy with anti-K30 monoclonal and anti-mouse-gold antibodies (Fig.  5C). Bacteria expressing native Wza exhibited the expected gold particles on the surface of the cell. No labeling was evident with E. coli CWG281 as a negative control (Fig. 5C). In contrast to the E. coli CWG281 cells expressing Wza, many of those expressing Wza* showed enlarged electron transparent domains within the periplasm. Immunogold labeling of these domains identified them as the site of accumulated K30 antigen, consistent with the interpretation that CPS is polymerized in strains expressing Wza* but not transported across the OM.
To determine whether the aberrant localization of K30 polymer in E. coli CWG281 expressing Wza* was the result of differences in location or structure of the protein, the IM and OM were separated by sucrose density gradient centrifugation. Wza* was clearly evident in the OM (Fig. 6A). To ensure that the protein was processed as expected, a C-terminal hexahistidine tag was added to Wza*, allowing it to be purified for N-terminal sequencing. The obtained N-terminal sequence, AEIYNVDGLN, was that predicted from sequence data for processed Wza*. The initial 7 residues (underlined) in Wza* are derived from ompF and restriction site sequence and replace the first 7 residues of native Wza (whose N-terminal sequence is CTIIPGQGLN). Despite its appropriate processing and localization, Wza* was unable to form SDS-stable multimers. When incubated at 22°C in SDS-PAGE sample buffer, only monomers of Wza* were evident. As shown in Fig. 6A and reported previously (17), the majority of Wza exist as multimers under similar conditions. Wza* His 6 was purified in the absence of SDS according to the protocol established for Wza His 6 (see above) and attempts were made at two-dimensional crystallization. Under these conditions, multimers were obtained and some ringlike structures were evident (Fig. 6B). However, the Wza* His 6 multimers often showed irregular shapes and tended to form poorly resolved aggregates rather than the organized arrays typical of Wza His 6 . From the limited well resolved ring structures, the outer diameters were measured to be 7.85 Ϯ 0.70 nm and the central cavity diameter was 3.07 Ϯ 0.45 nm. These results reflect a structure with a smaller overall size but containing a larger central cavity, in comparison to Wza His 6 multimers. In summary, the Wza* variant can be processed by the signal peptidase I and localized to the OM, suggesting that acylation is not a prerequisite for transport to the OM. However, the SDS instability of Wza* multimers at 22°C and the altered size of the structures suggests aberrant folding and/or multimerization that leads to loss of function in CPS translocation. Evidence That Wza Is Part of a Transmembrane Complex-We were interested in determining whether Wza K30 is part of a transmembrane complex acting in CPS translocation. One candidate for an IM interaction partner for Wza is Wzc. The Wzc protein is essential for capsule assembly and the close homolog Wzc from E. coli K-12 contains a large periplasmic loop (16); the corresponding region in Wzc from the K30 cps locus comprises 374 amino acids. The periplasmic loop is predicted to be mainly ␣-helical and contains a predicted coiledcoil motif (36), a feature that could be involved in proteinprotein interactions. To stabilize putative protein complexes, in vivo cross-linking experiments were performed with intact cells using DSP. Initial experiments were performed using the wza mutant E. coli CWG281 expressing Wza His 6 and Wzc from arabinose-inducible expression vectors, to enrich the amount of both Wza His 6 and Wzc in the cells. Cross-linked membrane complexes were purified by nickel chelation chromatography. After cross-linking, predominantly high molecular weight complexes were identified that barely migrated into the PAGE gel, although some non-cross-linked Wza His 6 was still evident (Fig.  7A). The amount of the free Wza His 6 varied from experiment to experiment and appears to be a common occurrence when this protein is overexpressed. On treatment with DTT, the complexes released Wza His 6 in increased amounts, as well as Wzc a few random gold particles from the nonspecific labeling typical in such experiments. A capsule is formed in E. coli CWG281 expressing Wza (see Fig. 1) and the amount of gold marker is increased and is confined mainly to the periphery of the cell. In contrast, cells expressing Wza* produce K30 antigen confined to the periplasm and often located in characteristic extended periplasmic domains. Bars on the micrographs represent 0.5 m.  6. Wza* forms OM multimers with altered stability. Panel A shows a Western immunoblot in which OM samples from E. coli CWG281 expressing either Wza or Wza* were probed with anti-Wza antibodies. Note that Wza* is unable to form multimers that are stable in SDS-containing sample buffer at 22°C. Using the purification protocol described under "Experimental Procedures," multimers of Wza* His 6 could be isolated. These were used in two-dimensional crystallography preparations and formed aggregates of ringlike structures rather than organized arrays when visualized by electron microscopy (panel B). The bar represents 0.5 m.
that is not present in samples without DTT. As a control for nonspecific binding to the Ni-NTA resin, parallel experiments were performed with E. coli CWG281 overexpressing native Wza and Wzc; neither protein was detected in samples processed with, or without, DTT (data not shown). These experiments provided direct evidence that Wzc is involved in a Wza His 6 -containing complex. The results were confirmed by identification of interactions in an E. coli LE392 background, devoid of the remaining members of the K30 CPS assembly machinery. In these experiments, Wzc His 6 was used to trap the complexes on Ni-NTA resin. After treatment with DTT, two predominant proteins were visible in a silver-stained gel, corresponding to Wzc His 6 and Wza, and both were clearly identified in Western immunoblots (Fig. 7B). In a control experiment where Wza was expressed alone in LE392, no Wza was captured on Ni-NTA (data not shown).
In previous cross-linking experiments, Wzc was shown to oligomerize independent of Wza (13,16). To determine whether the interaction between Wza and Wzc is essential for the formation of stable Wza multimers, Wza was expressed in the wzc-null strain, E. coli CWG285. Western immunoblot analysis of Wza in this background showed that SDS-stable multimers were still formed in the absence of Wzc (Fig. 8).

DISCUSSION
In Gram-negative bacteria, macromolecules such as CPS destined for the cell surface, or the extracellular environment, must cross both the IM and the OM. The Wza protein represents the best candidate for the OM channel protein for translocation of group 1 CPS in E. coli and K. pneumoniae. Furthermore, Wza from E. coli K30 also shares sequence similarity with a number of OM proteins associated with capsule and (secreted) extracellular polysaccharide production in other bac-terial species (8,17), suggesting a conserved function. The results presented here represent the first description of the secondary and quaternary structure of a member of this class of proposed OM polysaccharide export proteins (33).
The presence of significant ␣-helical domains in Wza indicates that it possesses a secondary structure distinct from OM porin proteins that have a predominantly ␤-barrel secondary structure (37). The multimeric structure formed by Wza is certainly more complex than porins. Single particle analysis of samples from two-dimensional crystals strongly suggests that Wza His 6 forms a multimer of eight identical monomers. The images of the Wza multimers resemble those obtained with members of the OM "secretin" family (38), involved in type II and III protein export, type IV pilus assembly, and filamentous phage assembly. Secretins are multimeric OM proteins in which a conserved C-terminal ␤-barrel rich domain is implicated in forming the multimeric OM channel (39 -42). Typically, 6 -14 identical subunits form the ringlike structure of the secretin (39,(42)(43)(44)(45)(46)(47)(48)(49)(50)(51)(52). Despite similarities in overall architecture of the secretin and Wza multimers, the respective monomers share no primary sequence similarity. With an average diameter of 8.84 Ϯ 0.43 nm, Wza His 6 complexes are smaller than those observed for many secretins. For example, PilQ from Neisseria meningitidis has a diameter of 16.5 nm (43). The central cavity in the Wza multimer is also smaller than that of PilQ (2.28 Ϯ 0.28 nm for Wza His 6 , compared with 6.5 nm for PilQ). Such differences may reflect the types of substrates for these putative channels. The group 1 CPS polymeric product is large (Ͼ100,000 Da; Ref. 53), but polymers of this type tend to adopt a flexible random coil structure in solution (54) and it is certainly conceivable that the polymer could be "threaded" through the channel as a linear strand.
The channels formed by protein export secretins are gated (48,55), and low resolution three-dimensional structures of three secretins suggest an open state for these channel proteins in the OM and a closed state in the periplasm entrance (42,44,49). Gated channels are also formed by TolC, an outer membrane protein involved in type I protein secretion and multidrug efflux (56). The solved crystal structure of the TolC trimer reveals a channel composed of a 40-Å ␤-barrel OM anchor domain and a 100-Å ␣-helical tunnel domain, long enough for a contiguous gated channel across the periplasm and interaction with the IM protein components (57). The TolC channel is proposed to be closed at the ␣-helical periplasmic domain by FIG. 7. In vivo interaction between Wza and Wzc demonstrated in chemical cross-linking experiments. In panel A, Wza His 6 and Wzc were overexpressed from the arabinose-inducible plasmids pWQ300 and pWQ130 respectively in CWG281. In panel B, Wzc His 6 and Wza were expressed together in E. coli K-12 strain LE392 from plasmids pWQ305 and pWQ306, respectively. DSP cross-linking was performed with intact cells, and cross-linked complexes were purified using Ni-NTA-agarose after isolation and solubilization of membrane proteins as described under "Experimental Procedures." The samples were divided, and one aliquot was treated with DTT to cleave the reversible cross-links. The samples were analyzed by SDS-PAGE on an 8% polyacrylamide resolving gel and stained with silver (upper frames). The lower frames show the relevant portions of Western immunoblots of the same samples probed with antibodies against Wza or Wzc. intra-and intermolecular hydrogen bonds and salt bridges, and is thought to be opened by an "iris-like" action (58,59). It remains to be determined whether Wza multimers form a gated channel and, if so, whether the ␣-helices detected in Wza play a similar role to those in TolC.
Wza is an OM lipoprotein but the function of the lipid moiety had not been investigated. The data for Wza* presented here suggest that the acylation is not required for OM targeting of Wza, but it is essential for assembly of a stable Wza multimer that can function in capsule assembly. Some protein export secretins are also lipoproteins (51, 60) but there is no information concerning the precise role of acylation in these examples. Non-acylated secretins typically require an additional OM lipoprotein ("secretin pilot") that associates with the secretin (61) and is required for formation of a stable multimer as well as, in some cases, localization of the multimer. Examples are found in type II (48,50,62,63) and III (45,46,64,65) protein secretion systems from various bacteria.
The assembly and localization of secretin multimers can also be influenced by other components of the protein secretion complex (65,66). For example, export of filamentous phage f1 requires only four proteins and cross-linking experiments suggest that the transmembrane complex is preformed in the absence of substrate by protein-protein interaction between the OM secretin pIV and the IM protein pI (67). This contrasts with the type I export hemolysin paradigm, where assembly of the functional complex and recruitment of TolC requires binding of the substrate to the IM components (22,68,69). The assembly and localization of Wza multimers shows no absolute requirement for Wzc. In fact, the observations that stable Wza multimers form in E. coli K-12 and that cross-linked Wza and Wzc could be isolated in the K-12 background both suggest that there is no absolute requirement for either the presence of substrate, or of other dedicated components of the capsule assembly machinery. However, as in all such cases, it is impossible to exclude the involvement of conserved "housekeeping" E. coli proteins in the formation of the multimeric complex. In type II protein export systems, the C-terminal domain of the secretin polypeptide contains the site of interaction with other proteins in the complex, including the pilot proteins (61,64). The differences in sequence (and perhaps structure-function) between Wza and the protein secretins are emphasized by the observation that the elimination of the C terminus in Wza ⌬348 -359 does not abrogate its function in capsule assembly.
E. coli K30 mutants deficient in Wza are unable to form a capsule but also do not accumulate detectable amounts of polymer within the cell, despite the fact that the glycosyltransferases for K30 synthesis are still active and K LPS is formed (8,17). A similar phenotype is observed with mutants defective in Wzc and Wzb (8,12). One attractive explanation for these similar phenotypes is some type of feedback regulatory mechanism, potentially involving an enzyme complex. Wzc provides a good candidate for interaction with the OM in a complex because of its large periplasmic domain (16) and the putative coiled-coil structures in this domain (36). The coiled-coil region is not involved in the known oligomerization of Wzc (14) but could certainly participate in interactions with other proteins. Although Wza shows no evidence of coiled-coil motifs itself, we have demonstrated here that Wza can be cross-linked in a complex that includes Wzc. It would be premature to take cross-linking data in isolation as definitive evidence of a direct interaction between Wza and Wzc, because other (unknown) proteins could serve as intermediaries in the interaction. However, the data are entirely consistent with the existence of a complex, minimally involving Wza and Wzc. These results differ from those obtained during attempts to cross-link Wzc and Wza in the E. coli K-12 colanic acid (EPS) system using formaldehyde (16). The reason(s) for the difference in results are unknown. Possibilities include the use of experimental systems with different modes of detection (and presumably different sensitivities), and the study of enriched material captured by exploiting hexahistidine tags, rather than analyzing whole cell lysates.
The normal pathway for E. coli group 1 CPS assembly is uncoupled in strains expressing Wza*. Although the unstable multimers formed by Wza* are unable to support CPS translocation to the cell surface, they are apparently sufficiently well recognized by the assembly system that any feedback regulation of early stages of synthesis is circumvented. As a result the strain with Wza* accumulates periplasmic K30 polymer in enlarged periplasmic bays. Periplasmic polymer has been observed in mutants affecting the translocation of group 2 CPS, and electron transparent domains are evident in mutants accumulating intracellular polymer (70 -73). The periplasmic location of K30 polymer in strains expressing Wza* represents the first direct demonstration that Wza (and by implication, Wza homologs in other CPS systems) is indeed required for CPS translocation across the OM.