Phosphorylation of Wzc, a tyrosine autokinase, is essential for assembly of group 1 capsular polysaccharides in Escherichia coli.

Wzc proteins are tyrosine autokinases. They are found in some important bacterial pathogens of humans and livestock as well as plant-associated bacteria, and are often encoded within gene clusters determining synthesis and assembly of capsular and extracellular polysaccharides. Autophosphorylation of Wzc(cps) is essential for assembly of the serotype K30 group 1 capsule in Escherichia coli O9a:K30, although a genetically unlinked Wzc(cps)-homologue (Etk) can also participate with low efficiency. While autophosphorylation of Wzc(cps) is required for assembly of high molecular weight K30 capsular polysaccharide, it is not essential for either the synthesis of the K30 repeat units or for activity of the K30 polymerase enzyme. Paradoxically, the cognate phosphotyrosine protein phosphatase for Wzc(cps), Wzb(cps), is also required for capsule expression. The tyrosine-rich domain at the C terminus of Wzc(cps) was identified as the site of phosphorylation and autophosphorylation of Wzc requires a functional Walker A motif. Intermolecular transphosphorylation of Wzc(cps) was detected in strains expressing a combination of mutant Wzc(cps) derivatives. The N- and C-terminal domains of Wzc(cps) were expressed independently to mimic the situation found naturally in Gram-positive bacteria. In this format, both domains were required for phosphorylation of the Wzc(cps) C terminus, and for capsule assembly. Regulation by a post-translational phosphorylation event represents a new dimension in the assembly of bacterial cell-surface polysaccharides.

Capsular and exopolysaccharides play crucial roles in the biology of many bacteria, acting either as virulence determinants that withstand host-cell defenses, or in establishing symbiotic relationships between bacteria and plants. More than 80 different capsular structures (known as K antigens) are produced by Escherichia coli isolates and these are subdivided into four different groups based on genetic and biochemical criteria (1). Surface polysaccharides with similar features are formed by other bacterial genera. The his-linked cps loci encode enzymes for the assembly of group 1 capsular K-antigens in E. coli and Klebsiella pneumoniae. The cps loci all contain a conserved region comprising the first 4 genes (orfX, wza, wzb, and wzc cps ) (2), indicating a shared role in CPS 1 expression. Following the conserved genes is a serotype-specific region encoding enzymes that participate in synthesis of polysaccharide repeat units and their polymerization via a Wzy-dependent mechanism (3). The Wzy-mediated polymerization reaction is thought to result in formation of an undecaprenyl pyrophosphate-linked glycan at the periplasmic face of the plasma membrane. The nascent polymer is then translocated to the cell surface via a process that requires outer membrane complexes formed by multimers of Wza cps (4). These complexes resemble the "secretins" for secretion of proteins via type II and type III systems.
A minor form of the group 1 K antigen is expressed on the cell surface in the form of low molecular weight K-antigenic oligosaccharides (one or a few K-repeat units) linked to LPS lipid A-core and termed K LPS (5). Surface expression of K LPS follows a distinct LPS translocation pathway that does not require Wza cps complexes in the outer membrane (4). The capsule structure evident in electron micrographs is formed from high molecular weight capsular K antigen (i.e. CPS) that, unlike K LPS , is not linked to lipid A-core (5). Despite these organizational differences, undecaprenyl pyrophosphoryl-linked intermediates provide the substrate for the Wzy polymerase required for synthesis of both CPS and K LPS (3).
The Wzc cps proteins in E. coli and K. pneumoniae strains that produce group 1 capsular polysaccharides are essentially identical. Homologues are also found in plant-associated Gramnegative bacteria that synthesize related extracellular polysaccharides. Examples include Erwinia amylovora, Xanthomonas campestris, Ralstonia solanacearum, and Sinorhizobium meliloti (reviewed in Ref. 6). In E. coli, Wzc cps participates in the surface assembly of CPS but is not required for K LPS synthesis (3). Deletion of the wzc homologue (exoP) in S. meliloti also eliminates high molecular polysaccharide formation without affecting the ability to synthesize repeat units (7). The Wzc proteins therefore likely play a role in high-level polymerization and it has been suggested that they belong to a larger family of "polysaccharide copolymerase" proteins that dictate chain length in surface polymers including capsules, exopolysaccharides, and LPS O antigens (8).
Investigation of the purified Wzc cps homologue (Ptk) in Acin-etobacter johnsonii by Cozzone and co-workers (9) first showed that Wzc proteins possessed tyrosine autokinase activity. Subsequent studies by the same group confirmed similar properties for purified Wzc ca from E. coli K-12 (10). The wzc ca gene is located in the locus responsible for synthesis of the slime exopolysaccharide known as colanic acid, whose production is dependent on growth conditions (e.g. temperatures below 30°C) (11). Phosphorylation of Ptk and Wzc ca appears to proceed by a novel phosphorylation mechanism because, unlike eukaryotic kinases, Ptk and Wzc ca contain a Walker A box ATP-binding motif (12) that is required for phosphorylation at multiple, but presently unidentified, tyrosine residues (13). Ptk and Wzc ca are dephosphorylated by the cognate phosphotyrosine protein phosphatases Ptp and Wzb ca , respectively, and these activities have been confirmed in vitro using purified proteins (10,14). Based on conserved gene products and preliminary biosynthesis data, colanic acid appears to be synthesized by a Wzy-dependent pathway similar to group 1 capsules (reviewed in Ref. 1). However, despite the fact that the colanic acid biosynthesis locus is widespread in E. coli, it is absent in E. coli strains with group 1 capsules (1). The Wzc homologues from the group 1 CPS and colanic acid synthesis loci are well conserved (51.9% identity; 83.1% total similarity), as are the corresponding Wzb proteins (51.0% identity; 76.2% similarity) (3). Thus they likely play similar roles in both biosynthetic systems.
Interestingly, homologues of Wzc are also found in important Gram-positive pathogens including Streptococcus pneumoniae (15,16), Streptococcus agalactiae (17,18), and Staphylococcus aureus (19,20). However, in these bacteria the N-and Cterminal domains of Wzc are represented in two separate polypeptides (15,20). The CpsC protein of S. pneumoniae is equivalent to the membrane-associated N-terminal domain of Wzc while the CpsD protein is homologous to the C-terminal Walker box-containing domain of Wzc. Recent investigations have shown that CpsD is phosphorylated at a tyrosine-rich C-terminal domain (21). The product of the cpsB gene may be involved in dephosphorylation of CpsC (21), although sequence analysis shows it lacks the sequence features (and catalyticsite motifs) found in other small phosphotyrosine protein phosphatases. Cognate phosphatases have not (yet) been identified in all Wzc-containing systems. Intriguingly, phosphorylation of CpsD is proposed to negatively regulate capsule biosynthesis in S. pneumoniae (21).
While autokinase and phosphotyrosine protein phosphatase activity has been demonstrated for some Wzc and Wzb homologues, the relationship between Wzc phosphorylation and capsular polysaccharide assembly has not been investigated in Gram-negative bacteria. Here, we show that two genetically unlinked wzc homologues on the chromosome of E. coli O9a: K30 contribute to different extents in the assembly of the K30 capsule. The phosphorylated tyrosine residues of Wzc cps are located to the C-terminal 17 residues of Wzc cps and assembly of the E. coli group 1 capsule is demonstrated to require both phosphorylation-competent Wzc cps and active Wzb cps phosphatase. The effect of phosphorylated Wzc cps on capsule assembly differs from that of phosphorylated CpsD in S. pneumoniae.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Plasmids, and Growth Conditions-The bacterial strains and plasmids used in this study are listed in Table I. All strains were grown and maintained in Luria-Bertani (LB) medium at 37°C. Where appropriate, media were supplemented with antibiotics to the following concentrations: ampicillin (Ap, 100 g ml Ϫ1 ), chloramphenicol (Cm, 34 g ml Ϫ1 ), gentamicin (Gm, 30 g ml ϪϪ1 ), kanamycin (Km, 50 g ml Ϫ1 ), and spectinomycin (Sp, 100 g ml Ϫ1 ). Expression of genes cloned in pET and pBAD derivatives was induced using 0.1-0.5 mM isopropyl-1-thio-␤-D-galactopyranoside and 0.02% L-arabinose, respectively.

Construction of Chromosomal Insertion Mutations in etk and wzb-
The etk gene is located in a gene cluster at 22 min on the E. coli K12 chromosomal map and is downstream of wza and wzb homologues. A chromosomal wza 22 min ::aadA insertion mutation (strain CWG258) was reported previously (4). The spectinomycin-resistance (aadA) gene cassette is flanked by transcriptional terminators (22) and its polarity therefore eliminates expression of downstream genes (i.e. wzb 22 min and etk). The pWQ129 suicide delivery plasmid carrying wza 22 min ::aadA was used to mutate E. coli CWG315 (wzc cps ::aacC1) to generate CWG285 (wzc cps ::aacC1 wza 22 min ::aadA), by allelic exchange. To generate the chromosomal wzb cps mutation, a 1.25-kilobase fragment spanning wzb cps was amplified by PCR and cloned. The nonpolar aph3A kanamycin-resistance cassette was inserted into the XbaI site within wzb cps . Plasmid pWQ146, a pMAK705-based suicide-delivery construct, was then used to transfer the mutation onto the chromosome of CWG343 (wza 22 min ::aadA wzb cps ::aph3A) by allelic exchange. Details of the approach for allelic exchange are described elsewhere (4). In all cases, correct insertions in chromosomal genes were confirmed by analysis of products of PCR reactions and by sequencing across the insertion using primers flanking the targeted region.
PCR Amplification and Cloning of etk, wzc cps , and Its Deletion Derivatives-Restriction and DNA-modifying enzymes were used according to the manufacturer's instructions. Transformation was done by electroporation using a Gene Pulser from Bio-Rad (23). PCR amplifications were carried out using a Perkin-Elmer GeneAmp PCR System 2400 thermocycler. Oligonucleotide synthesis and automated DNA sequencing services were provided by the Guelph Molecular Supercentre (University of Guelph, Ontario, Canada). The sequences of oligonucleotides are listed in Table II. All amplification reactions were performed in 50-l reactions using Pwo polymerase (Roche Molecular Biochemicals) and conditions optimized for each primer pair. PCR products and plasmid DNA fragments were purified using a QIAquick PCR purification kit (Qiagen) and/or Geneclean II (Bio101). To clone wzc cps , the 2.5-kilobase open reading frame was amplified by PCR using primers TW6 and JD102. The PCR product was digested with EcoRI and PstI and cloned into pBAD18-Km to give plasmid pWQ130. The wzc cps gene was also amplified as an N-terminal His 6 -tagged fusion protein using primer pair TW20-TW19. These primers introduced NdeI and EcoRI restriction sites facilitating cloning in pET28a(ϩ) (Novagen), resulting in plasmid pWQ141. The etk gene was amplified using the primer pair JD141-JD142. These primers introduced flanking MfeI and PstI restriction sites, allowing the product to be ligated to the EcoRI and PstI sites of pBAD24, resulting in plasmid pWQ131. Derivatives of wzc cps with C-terminal deletions were generated by PCR, using primer TW6 and reverse primers that introduced novel termination codons. Primers TW7 (for Wzc  ) and TW8 (for Wzc 1-480 ) also provided appropriate restriction sites to facilitate cloning of the amplified fragments in pBAD18-Cm. The resulting plasmids were pWQ133 (expressing Wzc 1-704 ) and pWQ139 (Wzc 1-480 ). The DNA fragment encoding the C-terminal domain of Wzc cps (Wzc 486 -721 ) was amplified using the primer pairs TW9-JD102 and the product was cloned in pBAD24, giving plasmid pWQ140. In all cases, the sequences of PCR-amplified genes were determined to ensure no errors were introduced during amplification.
Localization of Wzc cps in Subcellular Fractions-Expression of Wzc cps and its mutant derivatives was achieved using the pBAD arabinoseinducible expression vectors. Bacteria were grown to mid-exponential phase and expression of the Wzc cps derivative was induced by adding 0.02% L-arabinose. After an induction period of 15 min to 2 h, the cells were harvested and resuspended in 100 mM Tris-HCl, pH 7.5, containing 100 mM MgCl 2 and 1 mM phenylmethylsulfonyl fluoride (buffer A). Subcellular fractionation was performed at 4°C. The cell suspension was lysed by ultrasonication, after which unbroken bacteria and large debris were removed by centrifugation at 4,000 ϫ g for 8 min. The cell-free lysate was centrifuged at 40,000 ϫ g for 30 min, resulting in a cytosol-periplasm fraction (supernatant) and a pellet containing cell envelopes. The cell envelopes were resuspended in buffer A and further separated into inner and outer membranes by solubilization with 2% (w/v) Sarkosyl in buffer A for 30 min (24). The outer membrane is insoluble under these conditions, and was collected as a pellet at 40,000 ϫ g for 30 min. Wzc cps was identified in the cellular fractions by SDS-PAGE and Western immunoblotting.
Expression and Partial Purification of GST-Wzc cps Fusion Protein-To generate an N-terminal glutathione S-transferase (GST)-Wzccps fusion derivative, the wzc cps open reading frame was amplified by PCR using primers AP05 and AP06 (Table II). After digestion of the fragment with EcoRI and SalI at sites introduced in the primers, the fragment was cloned into pGEX-4T3 (Amersham Pharmacia Biotech) giving plasmid pWQ144. E. coli BL21 (DE3) transformed with pWQ144 was used to express and partially purify GST-Wzc cps by affinity chromatography on glutathione-Sepharose 4B (Amersham Pharmacia Biotech), using a modification of the approach described by Vincent et al. (10). Briefly, cell-free lysates were prepared by sonication in phosphate-buffered saline. After sonication, Triton X-100 was added to give a final concentration of 1%. Cellular debris were then removed by centrifugation at 30,000 ϫ g for 30 min and the cell-free lysate was mixed with chromatography matrix on ice for 30 min, before packing in a column for washing and elution. The column was washed with 3 ϫ 2 ml of phosphate-buffered saline containing 1% Triton X-100. Elution was performed using buffer B (50 mM Tris-HCl, pH 8.0) containing 10 mM glutathione and 0.1% Triton X-100. Fractions containing the GST-Wzc cps protein were pooled and dialyzed against buffer C (25 mM Tris-HCl, pH 7.0, 1 mM dithiothreitol, 5 mM MgCl 2 ).
Expression and Purification of His 6 -Wzb cps -The wzb cps gene was amplified as a 506-bp fragment using primers JD152 and JD153 and cloned into pBAD24 to give pWQ147. For purification, a His 6 -Wzb cps derivative was used. The wzb cps open reading frame was amplified by PCR using primers AP03 and AP04 (Table II). After digestion of the product at the NdeI and BamHI sites introduced in the primers, the fragment was cloned in pET28a(ϩ) giving plasmid pWQ145 and generating an N-terminal His 6 -tag on Wzb cps . The same approach was used to generate pWQ149 expressing the His 6 -tagged version of the mutated derivative Wzb C13S . The protein was purified from cell-free lysates of E. coli BL21 (DE3) containing pWQ145, prepared in buffer D (50 mM Na phosphate buffer, pH 8.0, 300 mM NaCl) containing 10 mM imidazole. After removing cellular debris by centrifugation at 30,000 ϫ g for 30 min, the His 6 -Wzb was purified by Ni 2ϩ affinity chromatography. The Ni 2ϩ -NTA matrix was washed 2 times in 2 ml of buffer D containing 20 mM imidazole, and His 6 -Wzb cps was eluted with 4 ϫ 250 l of buffer D containing 250 mM imidazole.
Site-directed Mutagenesis-Specific mutations were constructed in vitro using a modified procedure of the QuikChange TM site-directed mutagenesis kit from Stratagene. Briefly, complementary oligonucleotides were designed to contain the desired codon change (Table II). To generate Wzc cps K540R , the template consisted of plasmid pWQ130 (wzccps cloned in pBAD18-Km). The Wzb cps catalytic site mutant, Wzb C13S , was generated from plasmid pWQ147 by site-directed mutagenesis Tyrosine Phosphorylation and Capsule Assembly in E. coli using primers AP01 and AP02, to give pWQ148. Following PCR amplification, the reaction products were purified using the QIAquick PCR purification kit. The DNA was digested twice with DpnI to eliminate template, and the remaining DNA was ligated and introduced into E. coli DH5␣ by electrotransformation. The mutated derivative was resequenced (both strands) to confirm the mutation and verify that no other changes were introduced. Production of Antibodies Specific for Wzc cps -Wzc cps was overexpressed as an N-terminal His 6 -tagged derivative from plasmid pWQ141 in E. coli BL21 (DE3). Membranes were prepared as described above and the His 6 -Wzc cps protein was solubilized by incubation in buffer containing 1% Triton X-100 and 8 M urea for 20 min on ice. The solubilized protein extract was applied to a Ni 2ϩ -NTA column (Qiagen) and His 6 -Wzc cps was eluted using 250 mM imidazole. After dialysis, purified His 6 -Wzc cps was used to immunize a New Zealand White rabbit. Immune serum was absorbed against E. coli BL21 (DE3) whole cells, essentially removing all nonspecific antibodies.
Western Immunoblot Analysis of Wzc cps -The protein content of separated cell fractions or whole cell protein lysates was analyzed by SDS-PAGE (25) using 10% polyacrylamide resolving gels stained with Coomassie Brilliant Blue. For Western blotting, samples were electrophoretically transferred to Westran TM polyvinylidene difluoride membranes (Schleicher and Schuell). The transfer buffer was 25 mM Tris, 192 mM glycine, 0.1% SDS, and 20% methanol. Wzc cps was detected in Western blots using the rabbit polyclonal antiserum and a goat antirabbit secondary antibody (Caltag, Burligame, CA). To detect proteins containing phosphotyrosine residues, Western blots were probed using PY20 monoclonal anti-phosphotyrosine antibody (Transduction Laboratories, New York) and an anti-mouse secondary antibody (Jackson ImmunoResearch Laboratories, Inc.). Both secondary antibodies were conjugated to alkaline phosphatase and nitro blue tetrazolium chloride/ 5-bromo-4-chloro-3-indolyl phosphate was used for detection.
In Vitro Phosphorylation and Dephosphorylation of GST-Wzc cps -Autophosphorylation of GST-Wzc cps was performed using a modification of the approaches described by Vincent et al. (10). Approximately 80 g of GST-Wzc cps was incubated for up to 15 min at 37°C with 10 Ci of [␥-33 P]ATP (PerkinElmer Life Sciences, 3000 Ci/mmol) in a reaction volume of 0.2 ml. Samples were removed at intervals, transferred to 2 ϫ SDS-PAGE sample buffer and heated at 100°C for 5 min. The phosphorylated GST-Wzc cps was detected by SDS-PAGE followed by autoradiography. Dephosphorylation of Wzc cps -GST was assessed using in vitro phosphorylated substrate. After 15 min incubation in the phosphorylation assay, ϳ200 g of His 6 -Wzb cps (or catalytically inactive Wzb C13S ) was added at 37°C and 10 ϫ concentrated buffer was added to give a final concentration of 100 mM sodium citrate, pH 6.5, and 1 mM EDTA. At intervals, samples were removed, transferred to 2 ϫ SDS-PAGE sample buffer and heated at 100°C for 5 min prior to SDS-PAGE and autoradiography.

SDS-PAGE Analysis of Cell-surface
Polysaccharides-Cell surface polysaccharides from proteinase K-digested whole cell lysates were isolated as described by Hitchcock and Brown (26), and analyzed by electrophoresis on 10 -20% Tricine SDS-PAGE gels from Novex (San Diego, CA). Following electrophoresis, the LPS-containing molecules were visualized by silver staining (27). For Western blotting, samples were electrophoretically transferred to 0.45-m BioTrace NT membranes (Gelman Science) and probed with polyclonal rabbit anti-K30 serum (28), which recognizes the identical serotype K30 repeat unit found in both K30 LPS and K30 CPS. Rabbit polyclonal antibodies specific for the K40 antigen were described previously (29).
Bacteriophage Sensitivity Assays-To test for assembly of the K30 capsular layer, strains were assessed for their sensitivity to lysis by specific bacteriophages. The presence of surface-expressed K30 capsular antigen was determined using bacteriophage K30, which requires the K30 antigen as its receptor (30). Bacteriophage O9-1 is specific for the LPS O9a antigen (31). In the O9a:K30 wild-type background, the capsular layer masks the LPS O9a antigen and this strain is O9-1 resistant, but a reduction or absence of K30 capsular antigen unmasks the bacteriophage O9-1 receptor.

Two Functional wzc Homologues on the Chromosome of E.
coli O9a:K30 -The wzc cps gene is located in the K30 antigen biosynthesis (cps) locus of a prototype group 1 capsule-producing strain, E. coli E69 (O9a:K30) (3). When examined in Western blots using anti-K30 antigen antibodies, the cell-surface polysaccharides of the wild-type strain (E. coli E69) showed reactivity in both the K30 LPS fraction (containing primarily a single K30 repeat unit) and in high molecular weight CPS (Fig.  1). Previously, we showed that cell lysates of E. coli CWG315 (wzc cps ::aacC1) still contained immunoreactive K30 capsular antigen, but the amounts were significantly lower than those observed in wild-type lysates (Fig. 1). In addition, E. coli CWG315 was unable to assemble a capsular structure that masked the underlying LPS O9a antigen. The wzc cps mutation in CWG315 has no discernible effect on K LPS (3).
Subsequent studies by others have shown that many E. coli strains contain a homologue of wzc cps , designated etk (32). The etk gene is located in the appA (22 min) region of the sequenced E. coli K12 genome. The predicted Etk protein is an autokinase that is only expressed in some E. coli strains. It shares 74% similarity (57% identity) with Wzc cps (data not shown). Although the biological role of Etk is unknown, the 22-min locus  contains several open reading frames whose predicted products are homologous to proteins involved in polysaccharide expression (4). The products of the last three open reading frames are part of the same transcriptional unit and encode homologues of Wza, Wzb, and Wzc (i.e. Etk). This entire region was amplified by PCR and shown to have essentially the same sequence in E. coli K12 and E. coli O9a:K30 (data not shown). Furthermore, it was demonstrated that the wza 22 min gene product functions with low efficiency in the translocation of E. coli K30 CPS (4). To assess whether the small amount of K30 CPS remaining in E. coli CWG315 (wzc cps ::aacC1) resulted from etk expression, we constructed CWG285 (wzc cps ::aacC1 wza 22 min ::aadA) in which wzb 22 min and etk expression is eliminated by the polar aadA cassette inserted in wza 22 min . The capsule phenotypes of these strains were assessed by Western blotting and bacteriophage-sensitivity assays.
Western blot analysis of E. coli CWG258 (wza 22 min ::aadA) using anti-K30 antigen antibodies showed no significant reduction in synthesis of the CPS compared with the wild-type (Fig.  1) and the mutant strain showed no reaction with bacteriophage O9-1, indicating the maintenance of a protective K30 capsule. In the double mutant, CWG285 (wzc cps ::aacC1 wza 22 min::aadA), formation of CPS was completely abolished. No reaction was detected with the capsule-specific bacteriophage K30 and the strain was sensitive to bacteriophage O9-1. Although no CPS was synthesized, CWG285 showed a significant increase in the extent of polymerization (8 -10 repeat units) of K30 LPS reflected in the ladder of immunoreactive material.
Plasmids carrying the wzc cps and etk genes were used to complement the K30 capsule synthesis defect in CWG285 (wzc cps ::aacC1 wza 22 min ::aadA). Expression of plasmid pWQ130-encoded Wzc cps restored synthesis of K30 CPS and reduced the degree of polymerization of K LPS to the typical wild-type 1-2 repeat units (Fig. 1). While K30 CPS synthesis was restored in the complemented strain, the amount never achieved that seen in the wild-type strain. The reason(s) for this is unclear. While phosphorylated Wzc is certainly overexpressed in the complemented strain (see below), the precise amount of Wzc cps that is competent in CPS synthesis is unknown. This result could also reflect issues relating to stoichiometry of the components in a putative multienzyme complex. Expression of plasmid pWQ131-encoded Etk resulted in smaller amounts of CPS being formed (relative to Wzc cps ) and the polymerization of K30 LPS remained at the higher level. The lower activity of Etk is consistent with the absence of a detectable CPS phenotype of CWG258 (wza 22 min ::aadA). From these data and the phenotypes of CWG258 and CWG315 (above), we conclude that etk encodes a functional homologue of Wzc cps that participates with low efficiency, relative to Wzc cps , in assembly of the K30 capsule. The influence of Etk is not apparent in the presence of functional Wzc cps . In all subsequent experiments, strains carrying a wza 22 min ::aadA mutation were used to simplify the analysis and isolate the effects of Wzc cps and Wzb cps , without the complications of etk and wzb 22 min . Subcellular Location of Phosphorylated Wzc cps -To examine Wzc cps localization in a known genetic background, it was overexpressed in E. coli DH5␣. Although E. coli K-12 derivatives like DH5␣ contain functional copies of wzc ca and etk, these are not evident in control samples. Etk may not be expressed in E. coli DH5␣ (32) and there is no significant transcription of the colanic acid genes (including wzc ca ) in E. coli K-12 at 37°C (reviewed in Ref. 33). To determine the subcellular localization of Wzc cps , cell envelopes from arabinose-induced E. coli DH5␣ (pWQ130) cells were separated into the inner membrane (Sarkosyl soluble) and outer membrane (Sarkosyl insoluble) fractions. A Western blot of the membrane fractions was probed with a commercial monoclonal anti-phosphotyrosine antibody (PY20) which is known to react with homologues of Wzc cps (32). The phosphorylated Wzc cps protein (predicted molecular weight ϭ 79,558) is localized in the inner membrane (Fig. 2), as expected from sequence features and the relationships shared among the Wzc family of tyrosine autokinases. This location was further confirmed by examining membranes fractionated by isopycnic sucrose gradient centrifugation (data not shown).
In Vitro Autophosphorylation of Wzc cps and Its Dephosphorylation by Wzb cps -To confirm the autophosphorylation of Wzc cps without ambiguity, we employed the strategy used to establish autophosphorylation of Wzc ca (10). The GST-Wzc cps fusion protein was purified by chromatography on glutathione-Sepharose 4B. Based on Coomassie Blue-stained SDS-PAGE gels, this protein preparation was Ͼ90% pure (Fig. 3 panel A).
Phosphorylation of Wzc cps Is Essential for Capsule Synthesis-Sequence alignments revealed that Wzc cps (like its homologues) contains a Walker A or kinase-1a motif, commonly found in kinases (Fig. 4). The sequence Ala 534 -Ser-Pro-Ser-Ala-Gly-Lys-Thr 541 fits the general consensus sequence motif (AG(X 4 GK)ST). Two putative Walker B or kinase-2 motifs (hh- hhD; where h represents a hydrophobic amino acid) were detected. The optimal spatial distances between Walker A and B motifs are approximately 61 or 145 amino acid residues (34). While neither of the two Walker B motif candidates (VLFID 562 and LIIID 642 ) provides such optimal spacing, the aspartic acid residue within the sequence VIID 651 of the A. johnsonii Ptk protein (which corresponds to the LIIID 642 motif in Wzc cps ) was found to be essential for ATP binding (13).
Tyrosine phosphorylation of Ptk requires a functional Walker A box for ATP binding and hydrolysis (13). To confirm that Wzc cps autophosphorylation follows a similar mechanism, PCR-based site-directed mutagenesis was used to generate Wzc K540R (Fig. 4). Residues equivalent to lysine 540 in the Walker A kinase-1a-motif are believed to be required for ATP hydrolysis in the phosphotransfer reaction (34,35). The Wzc K540R derivative was detected with anti-Wzc cps antibodies (Fig. 5A) and was expressed at levels equivalent to the wildtype Wzc cps protein. When probed with monoclonal PY20 antiphosphotyrosine antibody, only the wild-type protein was phosphorylated (Fig. 5B). The function of Wzc K540R in capsule assembly was also completely abolished. No K30 capsular antigen could be detected in Western blots using polyclonal anti-K30 serum (Fig. 5C) or in phage sensitivity assays (data not shown) and the increased level of K LPS polymerization was not altered by this derivative. These data show that a functional Walker box, capable of ATP binding, is necessary for Wzc cps function in capsular assembly.

C-terminal Tyrosine Residues Provide the Sites of Phosphorylation and Are Required for Wzc cps to Function in Capsule
Formation-Previous studies suggested that Ptk, and presumably its homologues, are phosphorylated at multiple tyrosine residues (9) but the site of phosphorylation was not determined. Sequence alignments including Wzc cps homologues from Gram-negative and Gram-positive bacteria revealed several highly conserved C-terminal tyrosine residues (Fig. 4) and while this work was in progress phosphorylation of CpsD was shown to occur in the C-terminal tyrosine-rich domain (21). Seven of the last 17 amino acid residues of Wzc cps are tyrosine residues. To address the relevance of this tyrosine-enriched domain, the 17 C-terminal amino acids were removed in a truncated version of Wzc cps designated Wzc 1-704 (Fig. 3). Wzc 1-704 was reactive with anti-Wzc cps antibody (Fig. 5A) and localized to the inner membrane (data not shown) but the derivative could no longer undergo tyrosine phosphorylation (Fig. 5B). Furthermore, the Wzc 1-704 protein was no longer able to complement the CPS-deficient phenotype in strain CWG285 (wzc cps ::aacC1 wza 22 min ::aadA) (Fig. 5C), consistent with the proposal that phosphorylation of these tyrosine residues is required for normal Wzc cps function.
To determine whether the truncated Wzc 1-704 derivative still retains the ability to bind ATP, it was coexpressed in CWG285 with the Wzc K540R protein. This resulted in transphosphorylation of Wzc K540R (Fig. 5B) and generation of Wzc activity that restored the assembly of CPS (Fig. 5C).
Activity of the Phosphotyrosine-protein Phosphatase, Wzb cps , Is Required for Capsule Assembly-Since Wzb homologues are known to dephosphorylate their cognate Wzc proteins (10,14) and the phosphotyrosine-protein phosphatase activity for Wzb cps was confirmed as described above, we constructed CWG343 (wza 22 min ::aadA wzb cps ::aph3A) to examine the effect of Wzc cps dephosphorylation on CPS assembly. The presence or absence of Wzb cps has no significant effect on the amount of Wzc polypeptide produced (data not shown), or on the amount of phosphorylated Wzc cps (Fig. 6). However, CWG343 retains only trace amounts of K30 CPS synthesis, detectable as immunoreactive material in Western blots (Fig. 6) and by sensitivity to the K30 CPS-specific bacteriophage K30 (data not shown). Introduction of pWQ147 (Wzb cps ϩ ) restored synthesis of K30 CPS indicating that the defect in CWG343 was attributable only to the single nonpolar insertion in wzb cps . Coomassie Bluestained SDS-PAGE gels of CWG343(pWQ147) whole cell protein lysates showed the presence of an overexpressed protein with the apparent molecular weight (16,604 predicted) ex- pected of Wzb cps (data not shown).
One possible interpretation of these data is that the phenotype of CWG343 is due to a loss of protein-protein interactions, rather than a simple loss of phosphotyrosine protein phosphatase activity. To address this question, we constructed pWQ148 and pWQ149 in which Wzb cps has a C13S mutation. The phosphotyrosine-protein phosphatase signature motif (H/ V)C(X 5 )R(S/T) (36) contains a nucleophilic cysteine residue (Cys 13 in Wzb cps ) that forms a phosphocysteine intermediate during catalysis. This residue is essential for catalysis (37,38) and has been shown to be required for the activity of Wzb ca (14). As expected, the Wzb C13S showed no activity in vitro against phosphorylated Wzc cps (Fig. 3, panel D). Although the mutant protein was made in and readily detectable in CWG343 cell lysates (data not shown), it was unable to restore CPS synthesis (Fig. 6). Thus the defect in CWG343 is attributable to the loss of phosphotyrosine-protein phosphatase activity.
A Functional Wzc cps Comprising Independently Expressed Nand C-terminal Domains-Wzc cps homologues in Gram-positive bacteria are encoded as two distinct proteins with the corresponding genes generally found next to one another in the same cluster. For example, in S. pneumoniae CpsC resembles the N-terminal membrane-anchored part of Wzc cps . The adjacent gene (cpsD) encodes a predicted polypeptide equivalent to the C-terminal hydrophilic domain containing the ATP-binding motif and the conserved tyrosine residues (15). Both polypeptides are required for autophosphorylation (21). To determine whether different domains of Wzc cps would also be functional as separate polypeptides, initiation and termination codons were introduced and fragments expressing the N and C termini of Wzc cps were cloned independently. The break point (between amino acid residues 481 and 494) was dictated by sequence alignments with known Wzc cps homologues in Gram-positive bacteria (data not shown). A termination codon was introduced after Lys 480 generating the N-terminal polypeptide Wzc 1-480 . The Gly 486 codon was replaced with an ATG start codon to express the soluble C-terminal polypeptide Wzc 486 -721 . The Wzc 1-480 and Wzc 486 -721 derivatives were cloned behind the arabinose-inducible promoter in pBAD18-Cm and pBAD24, generating plasmids pWQ139 and pWQ140, respectively.
The Wzc cps N-and C-terminal domains (Wzc 1-480 and Wzc 486 -721 ) were expressed and located in the inner membrane and cytosol, respectively (data not shown). While the truncated Wzc 1-480 N-terminal fragment was readily detected in Western blots with anti-Wzc cps antibodies (Fig. 7A), the Wzc 486 -721 Cterminal fragment was expressed at only low levels and could only be detected in heavily overloaded SDS-PAGE gels (data not shown). When expressed alone in strain E. coli CWG285, neither part of Wzc cps could function in capsule formation (Fig.  7C). However, when both proteins were expressed simultaneously in CWG285, CPS synthesis was restored. The function in capsule formation was correlated with phosphorylation of the Wzc 486 -721 C-terminal domain, as this only occurred when both parts of Wzc cps were coexpressed (Fig. 7B). The amount of CPS was lower than wild-type and a masking K30 capsule was not formed based on the sensitivity of the transformant to bacteriophage O9-1. The low efficiency in the two-part construct may be related to low level expression of soluble Wzc 486 -721 and its phosphorylation in a CWG285 background. It is also conceivable that the stoichiometry of the two Wzc cps domains is important and this cannot be controlled in the two-plasmid system.

The N Terminus of Wzc cps Cannot Serve as a Functional Replacement for the LPS O-Antigen Chain-length Regulator,
Wzz-The N-terminal domains of Gram-negative Wzc proteins contain two putative transmembrane helices flanking a periplasmic loop. This topology resembles that of Wzz, the LPS O antigen chain-length determinant and the two classes of proteins share sequence similarity (6,8,39). Wzz proteins lack the additional C-terminal sequences that, in Wzc, contain the ATP-binding motif and tyrosine-rich domain. Wzz determines the distribution of O-antigen chain lengths and generates a modal pattern in LPS size distribution, evident as clusters of bands in SDS-PAGE. To determine whether the N terminus of Wzc cps can alter chain length modality as does Wzz, we used the E. coli K40 antigen as a reporter system. The K40 antigen is a group 4 capsule (1) in which the majority of the polysaccharide is linked to lipid A-core as an LPS O antigen. Wzz controls the modality of the K40 LPS (29) and the wzz::aacC1 mutant (CWG290) shows a loss of a modal cluster of K40 LPS bands in SDS-PAGE gels and accumulation of lower molecular weight bands (Fig. 8). This is thought to reflect a bias in the polymerization system favoring chain-length termination rather than further polymerization (reviewed in Ref. 6). A modal cluster of bands was restored in the SDS-PAGE profile when CWG290 was transformed by plasmid pWQ30 carrying the wzz gene from E. coli O75 (28). In contrast, expression of the N-terminal domain of Wzc (Wzc 1-480 ) in CWG290 did not bring about the dramatic modality resulting from Wzz O75 expression. However, it did alter the polymerization profile evident in SDS-PAGE. Furthermore, the effect was dependent on the extent of Wzc 1-480 expression (reflecting varying levels of induction by increasing arabinose concentrations). Interestingly, full-length Wzc cps did not generate the same altered profile of K40 LPS even at the highest concentration of arabinose (0.2%) (Fig. 8) and, as expected, the C terminus alone had no effect (data not shown).

DISCUSSION
In this study we investigated the structure of Wzc cps from E. coli E69 (O9a:K30) and its role in capsule formation. Two separable domains were identified and phosphorylation of the Wzc cps tyrosine autokinase was shown to be essential for synthesis of CPS. The conserved features in Wzc homologues suggest a shared and widespread role in prokaryotes. Since many of the Wzc-containing bacteria require capsular or extracellular polysaccharides as essential virulence determinants, Wzc homologues play a crucial role in pathogenesis of these organisms. Regulation by a post-translational phosphorylation event represents a new dimension in the assembly of bacterial cellsurface polysaccharides.
Previously, we showed that a nonpolar insertion mutation in wzc cps significantly lowered the amount of CPS synthesis in E. coli CWG315, resulting in a defective capsular layer that was unable to mask underlying LPS O9a antigen (3). Here, we established that the residual expression of CPS on the cell surface of strain CWG315 is due to a second functional Wzc homologue, encoded by the unlinked etk (formerly ep85) gene. Ilan and colleagues (32) have shown that the Etk tyrosine autokinase is only expressed in a subset of pathogenic E. coli strains, but its functional role in E. coli was not determined and its contribution to virulence has not been tested. The etk gene is located in the appA (22 min) region of the E. coli K-12 genome in a cluster encoding homologues of Wza and Wzb, as well as other genes whose predicted products share similarities with enzymes involved in biogenesis of cell-surface polysaccharides. The 22-min locus lacks either a Wzy-polymerase homologue or an ABC transporter, features that define the primary pathways for bacterial cell-surface polysaccharide biogenesis (reviewed in Refs. 1 and 40). Despite the apparent lack of a full spectrum of biosynthetic components necessary for its function as an independent polysaccharide expression locus, the 22-min locus gene products can cooperate with functions encoded by the cps locus (4). Wza cps forms a multimeric outer membrane complex that resembles the secretins for secretion of proteins by type II and type III secretion systems (4). When expressed from a multicopy plasmid, the wza 22 min gene product can also function with low efficiency in assembly of the K30 capsule. This is consistent with the current finding that Etk participates with low efficiency, relative to Wzc cps , in CPS synthesis. In light of this functional data, Etk should be renamed Wzc 22 min, joining Wzc cps and Wzc ca (colanic acid biosynthesis) as the third known Wzc homologue in E. coli. Although the colanic acid biosynthesis locus is also widespread in E. coli and represents an additional source of Wzc activity, this locus is absent in E. coli isolates with group 1 capsules (reviewed in Ref. 1).
Tyrosine autokinase activity has been documented in Ptk from A. johnsonii (41), Wzc ca from E. coli K-12 (10), Etk from E. coli (32), AmsA from E. amylovora (32), and now Wzc cps from E. coli isolates with group 1 capsules. However, the exact function of these proteins and the role of their phosphorylation in the biology of their respective microorganisms have not been established. The involvement of a Walker A motif in autokinase activity in these enzymes suggests that the kinase catalytic mechanism differs from that in eukaryotic protein kinases (13). Phosphorylation of Ptk requires ATP binding and hydrolysis (13) and a similar activity in Wzc cps was confirmed experimentally by site-directed mutation of the invariant Lys 540 of the Walker A box. This residue is believed to be involved in ATP binding and hydrolysis (34,35). The Wzc K540R mutation prevented phosphorylation of Wzc and completely abolished its function in capsule assembly, providing the first indication that ATP binding is essential, and that the phosphorylated form of Wzc cps is the one that is functional in CPS expression. The truncated Wzc cps derivative (Wzc 1-704 ) identified the highly conserved C-terminal tyrosine-rich domain residues as the site of phosphorylation and confirmed the essential requirement for Wzc cps phosphorylation in group 1 CPS synthesis. While the CPS phenotypes arising from an ATP-binding defect and a lack of the phosphorylation site are identical, we cannot rule out the possibility that they have different detrimental effects on the CPS synthesis process. Although the role of phosphorylation of Wzc cps has not been directly tested in other Gram-negative CPS synthesis systems, production of the exopolysaccharide succinoglycan is drastically reduced in S. meliloti strains expressing a derivative of ExoP (Wzc) lacking the C terminus (42). By analogy to Wzc, this region of ExoP contains the phosphorylated tyrosine residues. As with the E. coli group 1 CPS system, the exoP-delete strain is still able to synthesize lipid-linked glycan repeat units (7).
The effect of Wzc phosphorylation on group 1 CPS expression differs from that seen in the representative Gram-positive system from S. pneumoniae (21), where domains resembling the N and C termini of Gram-negative Wzc are found in separate polypeptides encoded by adjacent genes. Proteins sharing sequence similarity and features with the C-terminal domain of Wzc cps include CpsD in S. pneumoniae (15) and S. agalactiae (17,18), CapB from S. aureus (19), EpsD from Streptococcus thermophilus (43), and EpsB from Lactococcus lactis (44). CpsC, CapA, EpsC, and EpsA encode the corresponding membrane-associated domains, respectively. In S. pneumoniae CpsD, a Walker box mutation renders the protein unable to autophosphorylate and generates bacteria that synthesize only trace amounts of CPS (21). The size of the remaining CPS product was not reported. In contrast, the phosphorylation of the tyrosine residues per se was not required for CPS synthesis in S. pneumoniae. In fact, replacement of the C-terminal tyrosines with phenylalanine leads to a mucoid phenotype, presumed to reflect an increase in CPS expression. The differences in phenotype in the E. coli and S. pneumoniae systems are not simply due to unexpected effects arising from the use of a C-terminal deletion derivative for the E. coli analyses. The ability of Wzc K540R and Wzc 1-704 to undergo intermolecular transphosphorylation and restore CPS synthesis is consistent with the notion that the phenotype arising from Wzc 1-704 is due to a loss of phosphorylation. Preliminary experiments using site-directed mutagenesis to change tyrosine residues to phenylalanines demonstrate that phosphorylation occurs at multiple tyrosines in Wzc cps , and work is in progress to determine which tyrosine residues are important. However, a mutant in which all the tyrosines are mutated lacks phosphorylation and confers a CPS phenotype identical to that of Wzc 1-704 . 2 Thus, the S. pneumoniae system differs from the E. coli group 1 CPS in two respects: CpsD mutants defective in ATP binding/hydrolysis and tyrosine phosphorylation have different effects on CPS phenotype, and phosphorylated tyrosine may act as a negative effector in S. pneumoniae. The role of phosphotyrosine-protein phosphatase activity in S. pneumoniae is currently unclear. Although the product of an adjacent gene (cpsB) is implicated in dephosphorylation of CpsD (21), the putative CpsB protein lacks the motif (and active site cysteine) typical of phosphotyrosine-protein phosphatase. While CpsB might reflect a new type of phosphatase, such an activity has not yet been tested.
Given the overall similarity in the components for Wzy-dependent CPS biosynthesis in Gram-positive and Gram-negative bacteria, the differential effect of the phosphorylationdeficient Wzc cps and CpsD derivatives was surprising. The two domains of Wzc cps could be expressed as independent polypeptides to mimic the Gram-positive situation and these constructs were able to function in both phosphorylation of the C-terminal domain and in capsule assembly. However, there are some significant structural differences between CpsCD and Wzc cps . The extra-cytoplasmic membrane loop of CpsC and its homologues are usually 270 -300 residues smaller than the corresponding domain in Wzc cps (6,8). Another interesting feature distinguishing the Gram-negative and S. pneumoniae systems is the fact that the tyrosines in CpsD are arranged in a (YGX) 4 motif. Whether these latter features play a role in the apparent differences in the requirement for Wzc cps /CpsD phosphorylation in CPS synthesis remains to be established.
Phosphorylated Ptk and Wzc ca are substrates for the phosphotyrosine-protein phosphatases, Ptp and Wzb, and these enzymes were found to be functionally interchangeable (10). Wzccps is also dephosphorylated by Wzb cps . Unexpectedly, a Wzb cps mutant also lacks K30 CPS biosynthesis. To rule out the possibility that the CPS phenotype in CWG343 (wzb::aacC1) reflected a loss of important protein-protein interactions rather than the phosphotyrosine-protein phosphatase activity per se, we used a catalytically inactive mutant (Wzb C13S ). The inability of this mutant to complement the defect in CWG343 strongly supports the notion that phosphatase activity is essential. The most likely explanation for the CPS phenotype of the wzb cps mutant is an inability to dephosphorylate Wzc cps . This is based on activity of Wzb cps against Wzc cps , the fact that Wzc cps is the only known tyrosine-phosphorylated protein in the CPS biosynthesis system, and the fact that these E. coli strains have no other detectable tyrosine-phosphorylated proteins other than Etk. The requirement for both autokinase and phosphatase activity in CPS biosynthesis raises the interesting possibility that cycling between phosphorylated and dephosphorylated Wzc cps is involved in CPS assembly. However, attempts to demonstrate cycling in vitro were not successful (data not shown). Either the activities are confined to one-time events, or an essential component or condition needed for cycling is absent in the in vitro system. The level of autokinase activity is insufficient to address this question in vivo.
Although the results presented here provide the first indication of the requirement for phosphorylated Wzc cps in assembly of the K30 capsule, its exact role has yet to be resolved. The resemblance of Wzc homologues from Gram-negative and Gram-positive bacteria suggests they may act at conserved stages in the assembly process and the most logical candidate is polymerization. Significantly, the Wzc homologues in both Gram-positive and Gram-negative bacteria are involved in the assembly of polysaccharides that appear to be polymerized by Wzy-dependent polymerization systems. The Wzc cps protein is a member of the cytoplasmic membrane-periplasmic auxiliary protein 1 family (39). More recently, these proteins have been suggested to be part of a larger family including Wzz, a protein associated with chain length determination of Wzy-dependent LPS O antigens. An alternative name, polysaccharide copolymerase was proposed (8). The membrane-periplasmic auxiliary/polysaccharide copolymerase proteins are associated with loci for capsule and exopolysaccharide expression in Grampositive and Gram-negative bacteria and they share similarities in the transmembrane topology. Wzc proteins are distinguished from Wzz by an extended C terminus that contains the ATP-binding motif and phosphotyrosines, suggesting a more complex functional role for Wzc.
Wzy polymerase proteins catalyze glycosidic bond formation and are therefore specific for a given polymer repeat-unit structure. There is little primary sequence relatedness shared by Wzy homologues but their hydropathy profiles are similar. A given Wzz protein can interact with polymerization systems for different LPS O antigen repeat units (see, for example, Refs. 28 and 45-47). This suggests few (if any) limits are imposed on Wzz function by either Wzy primary structure or by the repeat unit structure of the polysaccharide product. However, the chain-length modalities are certainly sensitive to differences in Wzz primary sequences (45)(46)(47). In contrast to the sequence variation in Wzz within different E. coli serogroups, both E. coli and K. pneumoniae have cps gene clusters encoding essentially identical Wzc cps proteins (2). Wzz regulation is confined to lipid A core-linked polymers and while it can generate modality in K LPS , it does not influence assembly of the E. coli group 1 CPS (28). The natural lack of modality in K LPS in the wild-type strains results from the absence of wzz on the chromosome of E. coli isolates with group 1 capsules (28). Conversely, we show here that Wzc cps cannot restore modality in the absence of Wzz. While Wzc 1-480 does not influence O-chain length in precisely the same manner as Wzz, it does alter the profile of O-antigensubstituted LPS molecules in SDS-PAGE, suggesting it can interact with and modulate an O-antigen polymerization system providing that the C-terminal domain is absent.
It should be noted that enzymatic activity directly involving polymerization has not been proven for either Wzz or Wzc and their role could be indirect. In E. coli, polymerization reactions for K30 LPS and K30 CPS share the same Wzy-polymerase enzyme (3) and K30 LPS molecules comprising 8 -10 repeat units of K30 antigen are formed in strains devoid of Wzc cps and Etk. Thus, Wzc cps is not an essential partner for Wzy activity per se. However, Wzc cps might be essential for a specific subset of substrates, or enzyme complexes devoted to assembly of the high molecular weight capsular K30 antigen. For example, Wzc cps may regulate the flow of substrates into high molecular weight K30 CPS, accounting for the increased polymerization of K30 LPS evident in the absence of CPS synthesis. Wzc proteins could also potentially influence the synthesis of CPS by effecting formation of a capsule assembly complex. Both Wzz and Wzc proteins have periplasmic domains that are predicted to form coiled-coils (8), a feature important for protein-protein interactions. In the case of Wzz, oligomerization has been demonstrated by cross-linking methods (47). The demonstration that coexpression of Wzc K540R and Wzc 1-704 allows intermolecular transphosphorylation and restoration of capsule expression strongly suggests that Wzc cps proteins can also interact with one another. If Wzc cps plays a role in assembling a functional complex for assembly of the K30 CPS, the presence of polymerized K LPS can only be explained by separate biosynthetic complexes for the polymerization of K30 CPS (Wzc-dependent) and K30 LPS (Wzc-independent). The phenotype of E. coli CWG285 (wzc cps ::aacC1 wza 22 min ::aadA) would then be explained by the absence of functional CPS polymerization complexes. The availability of additional Wzy enzyme and its lipid-linked substrates for K LPS complexes would be reflected in the higher level of K LPS polymerization. This issue provides the direction for future studies.