The Klebsiella pneumoniae O2a Antigen Defines a Second Mechanism for O Antigen ATP-binding Cassette Transporters*

An ATP-binding cassette (ABC) transporter-dependent mechanism is responsible for the biosynthesis of polysaccharide O antigens in the lipopolysaccharides of many Gram-negative bacteria. In the Escherichia coli O9a prototype, addition of a non-reducing terminal modification regulates chain length. The terminal residue is an export signal recognized by the nucleotide-binding domain component of the cognate ABC transporter. The Klebsiella pneumoniae O2a antigen lacks a capping residue, and the corresponding nucleotide-binding protein does not contain a separate substrate-binding domain. Unlike the O9a transporter, the O2a transporter can export heterologous O antigens. Export of the O2a antigen is obligatorily coupled to chain elongation, and the O2a transporter is essential for establishing O antigen chain length. The E. coli O9a transporter operates after chain length has been determined. Furthermore biosynthesis and export can be uncoupled. O antigen chain length is a critical element in the ability of lipopolysaccharides to confer resistance to complement-mediated killing. This study establishes that two distinctly different approaches are available for the regulation of O antigen chain length and export via an ABC transporter.

Lipopolysaccharide (LPS) 4 is the major component of the outer leaflet of the outer membrane of Gram-negative bacteria (1). The prototypical LPS molecule consists of three distinct regions: a hydrophobic domain known as lipid A, a core oligosaccharide (OS), and an O-polysaccharide (O-PS) (2). LPS molecules isolated from bacterial cultures form a complex mixture. Some comprise only the lipid A and core OS and are known as "rough" LPS, whereas others are "smooth" (S-LPS) and capped by O-PS. When separated by SDS-PAGE, most S-LPS profiles reveal a strain-or serotype-specific range of modal O-PS chain lengths that vary by repeat unit increments.
Lipid A and the core OS are synthesized by a sequential assembly process, and the product is exported across the inner membrane by MsbA (2,3). Synthesis of the O-PS occurs separately by one of two principal pathways that involve fundamentally different processes: the ABC transporter-dependent pathway or the Wzy-dependent pathway (2,4). In the ABC transporter-dependent pathway, polymerization of the O-PS occurs at the cytoplasmic face of the inner membrane. Initiation of synthesis requires an undecaprenol pyrophosphatelinked (und-PP) N-acetamido sugar. In Escherichia coli, this lipid-linked residue is typically N-acetylglucosamine (GlcNAc). Und-PP-GlcNAc is formed by the action of WecA, the N-acetylglucosamine:undecaprenylphosphate N-acetylglucosamine-1-phosphate transferase first identified in the biosynthesis of enterobacterial common antigen (5). WecA involvement has been established for ABC transporter-dependent O-PS pathways in E. coli and Klebsiella pneumoniae (5)(6)(7)(8)(9), and the essential features are conserved in Yersinia enterocolitica O:3 (10) and the Pseudomonas aeruginosa A-band antigen (11). Serotype-specific glycosyltransferases then extend the und-PP-GlcNAc acceptor by sequentially transferring sugar residues to the non-reducing terminus of the growing chain. After polymerization is complete, the O-PS is exported to the periplasmic face of the inner membrane by the action of an ABC transporter (10 -14). O antigen ABC transporters are comprised of two transmembrane domains and two nucleotide-binding domains (NBDs) encoded by wzm and wzt, respectively. Once at the periplasmic face of the inner membrane, the O-PS is ligated to lipid A-core OS and translocated to the cell surface (2). Although some components of the later translocation steps have been identified (15)(16)(17)(18)(19)(20) the details of the LPS translocation process are largely unknown.
The O-PSs of E. coli serotypes O8, O9, and O9a are linear polymers of mannose that differ in linkage sequence (21) and represent the best established examples of ABC transporter-dependent O-PS export systems (12, 13,22,23). A novel feature of the polymannose O antigens and some other O-PSs assembled by ABC transporter-dependent pathways is the presence of unique residues at the non-reducing termini of the glycan chains. The polymannose O chains terminate at the non-reducing end in a methyl group (E. coli O8) or both a phosphate and a methyl group (E. coli O9 and O9a) (24,25). In the E. coli O9a prototype, the wbdD gene product is a bifunctional kinasemethyltransferase with the kinase activity being a prerequisite for the addition of the methyl residues (23). The action of WbdD is central to the establishment of the unique modal chain length distribution of the O-PS, but chain termination also couples biosynthesis to export because wbdD mutants can synthesize the O9a polysaccharide but are unable to export it (23). This phenotype was explained by the finding that the NBD (Wzt) protein contains an extended C-terminal domain that binds and is specific for the chain-terminating residue (12, 13). Specificity is imparted by an extended C-terminal region in the Wzt protein that folds to form a substrate-specific O-PS-binding domain (13).
Some O-PSs that follow an ABC transporter-dependent assembly pathway lack a defined non-reducing terminal residue; the polygalactose O-PSs of K. pneumoniae provide one example (24). These O-PSs all contain a disaccharide repeat unit structure (D-galactan I) linked to the reducing terminal GlcNAc (26 -28). D-Galactan I defines the O2a antigen of K. pneumoniae, and in some isolates, it is the only O-PS (26). However, in other serotypes, D-galactan I may be capped by an additional serotype-specific domain that can define the structure of additional O antigens (e.g. O1 and O2c) (26,29). LPS isolated from cultures of these bacteria contains two types of S-LPS; one contains only D-galactan I, and the other contains short chains of D-galactan I capped by the additional antigen molecules (27,28). Six genes form a locus required for D-galactan I biosynthesis and export (7,8,30,31). When expressed in E. coli, the recombinant bacteria produce S-LPS containing only D-galactan I. The genes involved in the formation of the additional repeat unit structures are unknown, but their synthesis (and addition) is not required for D-galactan I O-PS export (30).
The absence of a chain-terminating residue in some O-PSs suggests a different recognition process in coupling biosynthesis to export. Here we examine the assembly-export coupling system for the D-galactan I O-PS from K. pneumoniae and establish that this represents a second mechanism for ABC transporter-dependent O-PS assembly. The process is substantially different from the model established for the polymannose O-PSs.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Plasmids, and Growth Conditions-The bacterial strains and plasmids used in this study are described in Table 1. Bacteria were grown at 37°C in either Luria-Bertani (LB) medium or minimal medium containing M9 salts (32,33).
DNA and Mutagenesis Methods-DNAzol reagent (Invitrogen) was used to purify chromosomal DNA. DNA fragments were PCR-amplified using Pwo polymerase (Roche Applied Science) with custom oligonucleotide primers (Sigma). Where appropriate, the primers included restriction sites to facilitate cloning. Purification of DNA fragments from agarose gels was carried out using the Ultraclean15 DNA purification kit (MOBIO Laboratories). The Qiaquick PCR purification kit (Qiagen) was used to purify DNA fragments directly from PCRs or restriction digestions. Plasmid DNA of pBAD24 (34) or pACYC 184 (35) was purified using either the GeneElute plasmid purification kit (Sigma) or the Qiagen HiSpeed Midi kits (Qiagen). Restriction digestions and DNA ligation reactions were performed according to the manufacturer's instructions. All constructs were confirmed by DNA sequencing at the Guelph Molecular Supercenter (University of Guelph). The red recombinase system (36) was used to create a ⌬wzx mutation in E. coli CS1776 (37).
Conditional Expression of D-Galactan I Biosynthesis and ABC Transporter Genes-Cultures of E. coli CWG869 containing pWQ290 and pWQ289 were grown overnight in LB medium containing 0.4% glucose (to repress expression from the pBAD promoter and prevent O2a transporter expression from pWQ290) with the appropriate antibiotics. The cultures were then diluted 1:100 into 5-ml batches of fresh medium. To one culture, arabinose and galactose were added simultaneously once the cells reached an A 600 of ϳ0.4. For conditions involving induction of the O2a transporter prior to D-galactan I synthesis, cultures were grown to an A 600 of ϳ0.4, and arabinose (0 -0.2%, w/v) was added to initiate O2a transporter expression. Incubation was continued for 10 min, and galactose (0.1%, w/v) was then added to initiate D-galactan I synthesis. The cells were This study allowed to grow to an A 600 of ϳ0.6. For induction of the O2a transporter after D-galactan I synthesis, the order of addition of arabinose and galactose was reversed. In an experiment designed to show whether synthesis of the polymer and export were coupled, a "pulse-chase" approach was used. Galactose (pulse) was added to the growth medium at an A 600 of ϳ0.4, and cells were grown for 10 min to allow for D-galactan I synthesis. Cells were then collected by centrifugation at 5,000 ϫ g for 3 min and washed twice in prewarmed medium to prevent further de novo synthesis of polymer. The cells were then resuspended in medium supplemented with arabinose (chase). In all cases, incubation was continued for 10 min at which time cells were harvested for LPS analysis. Comparable experiments were performed using E. coli O9a except galactose was replaced by mannose, and the cells were grown in supplemented M9 minimal medium containing 0.2% glucose or 0.2% glycerol where appropriate. To determine whether synthesis and export of the polymannose O-PS were coupled, cells were also washed in M9 minimal salts supplemented with glycerol prior to mannose addition to remove any residual glucose that would suppress mannose uptake. The cells were held in the M9 minimal medium supplemented with mannose for 2 min and then collected by centrifugation at 5,000 ϫ g for 5 min. The cells were then washed twice in prewarmed medium supplemented with glycerol, then resuspended in M9 minimal medium with glycerol as the carbon source and supplemented with arabinose or glucose, and allowed to grow for 10 min before the cells were harvested. In the pulse-chase experiment the low copy number pBAD322-Cm derivative (38) pWQ346 was used. Complete repression of O9a transporter expression from the equivalent (high copy) pBAD24 construct was difficult to achieve.
LPS Analysis-Proteinase K-digested whole-cell lysates were prepared to examine LPS by SDS-PAGE by the methods of Hitchcock and Brown (39). LPS was visualized by silver staining (40). Western immunoblots were prepared by transferring the samples to Protran nitrocellulose membranes (PerkinElmer Life Sciences). The primary antibody, D-galactan I-specific polyclonal antiserum, was raised in New Zealand White rabbits immunized with formalin-killed whole cells of K. pneumoniae CWK49. To prepare D-galactan I-specific antibodies, immune serum was adsorbed with whole cells of E. coli CWG869. Alkaline phosphatase-conjugated goat anti-rabbit secondary antibody (Sigma) was used for detection.
Immunofluorescence Microscopy-Cultures were grown overnight in LB broth supplemented with 0.4% glucose. Bacteria were subcultured 1:50 into 5 ml of LB broth supplemented with either galactose or mannose (0.1% final concentration) and varying concentrations of arabinose as required. Cultures were then grown for ϳ2.5 h. Immunofluorescence labeling of intact and permeabilized cells was performed as described previously (23) using either D-galactan I-specific or O9a-specific rabbit polyclonal antibodies. Detection was performed using rhodamine red-conjugated goat anti-rabbit antibody (Jackson ImmnoResearch Laboratories). Visualization was carried out using a Zeiss Axiovert 200 microscope using a 100ϫ objective lens, and the images were processed using Openlab software (Improvision).
In Vitro Galactosyltransferase Assays-Membranes were prepared from 500-ml cultures of CWK49 transformed with pWQ391. Cultures were grown to an A 600 of 0.6 in M9 minimal medium containing 0.2% glycerol; where appropriate, O2a transporter expression was induced by adding 0.2% L-arabinose. Cells were collected by centrifugation at 5,000 ϫ g for 10 min and washed twice with 50 mM Hepes buffer, pH 7.5. The cells were then resuspended in 20 ml of 50 mM Hepes, pH 7.5. Ultrasonication was used to lyse cells. Unbroken cells and inclusion bodies were removed by centrifugation at 12,000 ϫ g for 15 min. Membranes were then collected by ultracentrifugation at 100,000 ϫ g for 1 h. The resulting membrane pellet was resuspended in 0.5 ml of 50 mM Hepes, pH 7.5, and stored at Ϫ80°C. Protein concentration was determined using the Bio-Rad DC protein assay kit. Galactosyltransferase assays were performed at 30°C in 0.6-ml reaction volumes containing 50 mM Hepes, pH 7.5, 20 mM MgCl 2 , 2 mM dithiothreitol, 606 nM UDP-[ 14 C]galactose (258 mCi/mmol; PerkinElmer Life Sciences), and the membrane equivalent of 0.6 mg of total protein.
In one set of experiments, bacitracin (500 g/ml final concentration) was added to the reactions. Samples of 0.1 ml were withdrawn at intervals over a 30-min period and added to 0.3 ml of cold 12% acetic acid to stop the reaction. Samples were filtered onto MicroSep 0.45-m cellulose filters (Osmonics) and washed twice with 12% acetic acid to remove residual UDP-[ 14 C]galactose. Filters were dried and placed into 5 ml of Ecolite scintillation fluid (ICN Biomedicals), and radioactivity incorporated into polymer was measured by scintillation counting.

Comparison of ABC Transporter Sequences-The
NBDs from E. coli serotypes O8 and O9a share a highly conserved N-terminal domain containing all of the essential motifs that define ABC proteins (12). They diverge in the serotype-specific C-terminal O-PS-binding domain (13). Wzt EcO9a is 413 amino acids in length, whereas wzt from K. pneumoniae encodes a significantly shorter protein of 246 amino acids. Wzt KpO2a and the N-terminal domain of Wzt EcO9a are highly conserved (see Fig. 2). The absence of a comparable O-PS-binding domain in Wzt KpO2a raises questions concerning substrate recognition and specificity in the shorter Wzt proteins. The size and sequence of Wzt KpO2a are very similar to those of KpsT, the NBD protein involved in group 2 capsular polysaccharide (CPS) assembly systems in E. coli (14,41). Interestingly the CPS transporters are conserved between different CPS serotypes and have no specificity for the repeat unit structure of their exported substrates.
The transmembrane domains of ABC transporters are generally poorly conserved in primary sequence, but they have similar predicted secondary structures (12). Hydropathy plots illustrate the characteristic six transmembrane helices typical of the protein family. The distribution of these helices is similar in Wzm KpO2a and Wzm EcO9a (data not shown).
The D-Galactan I Transporter Can Functionally Replace the ABC Transporter of E. coli O9a-The polymannose ABC transporters for the E. coli O8 and O9a O antigens are specific for their substrate (12), a property imparted by a recognition domain of Wzt interacting with the terminating residues (13). The observation that D-galactan I in K. pneumoniae is often capped with additional repeat unit structures ( Fig. 1) together with the absence of a C-terminal polymer recognition domain on the NBD (Fig. 2) suggested that the corresponding D-galactan I (O2a) ABC transporter may parallel the group 2 CPS ABC transporters in having relaxed specificity. To test this hypothesis, the ability of the K. pneumoniae O2a transporter to export the O9a polymannose O-PS was investigated. An E. coli CWG638 (O9a; manA,wzm-wzt::aphA-3) was transformed with pWQ290 containing the O2a transporter cloned behind an arabinose-inducible pBAD promoter. In this strain, polymannose biosynthesis is dependent on the addition of mannose to the growth medium. This is essential because in the absence of export synthesis of the O9a polymer is toxic and results in aberrant cell morphology (Fig. 3B) as reported previously (12, 13). In the absence of transporter expression, only rough LPS is formed (Fig. 3A); however, expression of the O2a transporter resulted in the formation of S-LPS. Trace amounts of S-LPS (at the limits of detection in silver staining) were detectable after induction with 0.02% arabinose, and a clear ladder of S-LPS was evident with 0.2%. Notably the S-LPS exhibited the same O-PS chain length modality as the authentic product, indicating that normal synthesis and chain termination were unaffected by the heterologous transporter. Immunofluorescence completed on CWG638 harboring pWQ290 induced with 0.2% arabinose confirms that the O-PS is evident on the surface of the cells (Fig. 3B).
Functional replacement of the O9a transporter required the complete O2a transporter; transformation of E. coli O9a wzm and wzt single mutants with the corresponding K. pneumoniae O2a cloned gene did not restore polymannose O-PS export (data not shown). This suggests that specific interactions between the transmembrane domains and NBDs of the transporter are required for a functional transporter, an observation made previously with the E. coli O9a and O8 ABC transporters (12, 13).
Overexpression of the O2a Transporter Leads to a Change in O-PS Modality-Differences between the processes coupling O-PS biosynthesis and export for the polymannose and D-galactan I O-PS systems are clearly illustrated in a simple experiment involving overexpression of the cognate transporters  (29) or O2c (26) specificity. Note that the capping domain is not required for export as these bacteria produce a mixture of S-LPS species, some containing only D-galactan I. B, organization of the gene cluster responsible for D-galactan I biosynthesis (7,8,14). The additional genes responsible for capping serotype-specific domains have not been identified. Also shown is KpsT, the NBD required for E. coli K1 capsule export. Identical residues are highlighted in black. Note the extended serotype-specific C-terminal domain in the Wzt homologs involved in polymannose export (12, 13). (Fig. 4A). The O9a and O2a transporters were overexpressed in their native backgrounds. The LPS profile (and O-PS modality) isolated from E. coli O9a overexpressing the native transporter was unchanged from the wild type. In contrast, overexpression of the O2a transporter in its native strain led to a substantial reduction in the average O antigen chain length. This phenotype requires the complete functional transporter; overexpression of either Wzm or Wzt alone in the wild type K. pneumoniae strain had no effect on the average O-PS chain length (data not shown).
One interpretation of the data is that elevated expression of the O2a transporter reduces, in some manner, the O-PS biosynthesis activity. To examine this possibility, membranes were prepared from K. pneumoniae (CWK49) with varying O2a transporter expression and used as a source of galactosyltransferase activity (Fig.  4B). Paradoxically strains with higher levels of O2a transporter induction actually showed significantly elevated levels of galactose incorporation. It is conceivable that this result reflects increased O2a transporter activity enhancing transfer of the und-PP-linked glycan across the membrane to facilitate recycling of und-PP and its reentry into the pool for further rounds of polymerization. To test this hypothesis, bacitracin was added to the reaction. Bacitracin inhibits the dephosphorylation of und-PP, therefore preventing its return to the active pool of lipid carrier (42,43). As shown previously in studies with the E. coli O9a polymannose antigen (44), bacitracin significantly reduces mannose incorporation consistent with inhibition of und-PP recycling in the membrane preparation. However, the failure of bacitracin to completely eliminate incorporation was also observed in the polymannose system (44). Bacitracin does reduce the amount of galactose incorporation, indicating that recycling does occur in vitro, but it does not lower the activity to the levels seen in the wild type strain in the absence of the overexpressed O2a transporter (Fig. 4B). This could imply that recycling does not fully account for the increased galactose incorporation seen in cells overexpressing the O2a transporter. However, the result must be interpreted with caution because it is possible that the bacitracin does not access all of its potential targets in these experiments.
D-Galactan I Biosynthesis Is Coupled to Export in Vivo-To further study the role of ABC transporter expression on O-PS chain length, a conditional expression system was established to temporally uncouple D-galactan I biosynthesis from export. The four genes required for D-galactan I biosynthesis, wbbM-glf-wbbN-wbbO, were constitutively expressed from the tetracycline promoter of pACYC184 (pWQ289). The O2a transporter genes were expressed under the control of an arabinose-inducible promoter in a compatible pBAD plasmid  (pWQ290). Using E. coli K-12 strain CWG869 (galE:Tn10,⌬wzx), UDP-Galp formation and polymer biosynthesis are dependent on addition of galactose to the growth medium (37). E. coli K-12 isolates are unable to form O-PS because of mutations in the chromosomal O-PS biosynthesis locus (45). However, residual Wzx activity results in nonspecific flipping of short lipid-linked oligosaccharide intermediates synthesized by plasmid-encoded genes, and these can be ligated to lipid A-core OS (8,46). This activity was eliminated by the ⌬wzx mutation to ensure that D-galactan I export occurred only through the O2a transporter. Plasmid pWQ288 containing the entire D-galactan I cluster in pACYC184 served as a control for expression of the O-PS (Fig. 5A).
In initial experiments, induction of O2a transporter expression and D-galactan I biosynthesis were initiated simultaneously. In the absence of the O2a transporter activity (no arabinose), E. coli CWG869 (pWQ289) shows no S-LPS in silver-stained SDS-PAGE of the whole-cell lysate (Fig. 5A, top). However, as expected the cells can synthesize polymer that is visible on Western immunoblots probed with anti-D-galactan I antibodies (Fig.  5A, bottom). This material accumulates intracellularly and can only be identified in immunofluorescence experiments when the cells have been permeabilized (Fig. 5D). This is consistent with previous electron micrographs showing the accumulation of D-galactan I at the inner membrane-cytoplasm interface in the absence of the O2a transporter (14). Surprisingly the apparent size of the accumulating polymer based on SDS-PAGE migration (in the corresponding Western immunoblot) is considerably larger than the O-PS from CWG869 (pWQ288) where D-galactan I O-PS is exported and ligated to lipid A-core OS. This suggests that control over O-PS chain extension activity is lost in the absence of a functional ABC transporter. The amount of S-LPS on the cell surface of CWG869 (pWQ289  A, B, and C, lower panels). Assays were performed in E. coli CWG869 (galE,⌬wzx). The galactosyltransferases (wbbM, wbbN, and wbbO) as well as the UDP-galactopyranose mutase (glf) were cloned behind the tetracycline promoter of pACYC184 (pWQ289). The ABC transporter components (wzm-wzt) were expressed behind the arabinose-inducible promoter in pWQ290. D shows representative immunofluorescence images (with anti-D-galactan I antibodies) of intact cells and permeabilized cells to confirm the surface or intracellular location of the D-galactan I-antigenic polymers revealed by SDS-PAGE and Western immunoblots. For each set of images, a differential interference contrast image is shown on the left, and the corresponding fluorescence image is shown on the right. and pWQ290) was dependent on the amount of arabinose (and hence the level of expression of the O2a transporter) (Fig. 5A). In the absence of arabinose, trace amounts of immunoreactive material were detected on the surface of intact cells (Fig. 5D), but this falls below the sensitivity of silver-stained SDS-PAGE. S-LPS was detected on silver-stained SDS-PAGE in lysates of cells grown with Ն0.002% arabinose, and this material was accessible on the surfaces of intact cells as indicated by the immunofluorescence images (Fig. 5D). Under conditions where the amount of S-LPS was highest (0.2% arabinose), the cells did not accumulate large molecules of D-galactan I O-PS. These experiments suggested that the O2a transporter is required both for export and for determination of D-galactan I O-PS chain length.
The uncoupling of the expression of the O2a transporter and the D-galactan I biosynthesis system facilitated experiments to address the effect of timing of ABC transporter expression on polymer export. When the expression of the O2a transporter genes was induced by arabinose induction prior to the initiation of D-galactan I synthesis, the O-PS modality was shifted downward as the arabinose concentration increased (Fig. 5B); this is similar to the effect observed in K. pneumoniae (Fig. 4A). In contrast, the LPS profiles of cells where O2a transporter expression was induced after the induction of polymer synthesis showed that the very large accumulated intermediates remained suggesting that once synthesized this material is not a substrate for export (Fig. 5C).
To examine more closely the possible precursor-product relationships between the large intermediates and S-LPS on the cell surface, an additional series of pulse-chase experiments was performed using E. coli CWG869 (pWQ290 and pWQ289) (Fig. 6A). Polymer formation was initiated by the addition of galactose (pulse), and O2a transporter expression was initiated by the addition of arabinose (chase). In control experiments, a pulse with no chase resulted in the production of high molecular weight polymer but no S-LPS as expected (lane 2). In the absence of the pulse, no polymer was formed (lane 3). In samples where the cells were washed to eliminate any further D-galactan I formation prior to arabinose addition and therefore O2a transporter expression (lane 4), the cells retained large intracellular intermediates. No S-LPS was detected by SDS-PAGE (Fig. 6A), and the cells showed no surface D-galactan I in immunofluorescence studies (Fig. 6B). To confirm that these results were not because of an unanticipated effect on the polymer biosynthesis machinery from the washing step, one sample received additional galactose during the arabinose chase (lane 5). These conditions allowed de novo synthesis of D-galactan I and formation of S-LPS (Fig. 6B), although the large pool of accumulated polymer built during the initial pulse was unchanged. These experiments confirm that O-PS polymerization and export must be temporally coupled in the D-galactan I system and that this process establishes O-PS chain length.
The O9a Transporter Is Proficient for Postpolymerization Export-Three lines of evidence predicted that the polymannose O-PS assembly system would not show the same strict temporal coordination of biosynthesis and export. (i) The polymannose O-PSs are terminated with a non-repeat unit residue (methyl or methyl and phosphate groups in O8 and O9a, respectively) (24). (ii) Chain length regulation is maintained in the absence of export provided that chain termination occurs (12, 23). (iii) Chain termination is essential for export (12, 13). To test this prediction, a similar conditional expression system was set up in E. coli CWG638 with the O9a transporter controlled by an arabinose-inducible promoter in pWQ346 and O-PS biosynthesis initiated by mannose addition to the growth medium (Fig. 7). The resulting polymannose polymer accumulates inside the cells when the O9a transporter is not expressed (12). The cells were washed prior to induction of the O9a transporter to eliminate further O-PS biosynthesis. Under these conditions, S-LPS with wild type modality is produced. Although the amount of S-LPS produced is low based on SDS-PAGE, immunofluorescence clearly demonstrates that this material is on the cell surface (Fig. 7B) indicating that, unlike

DISCUSSION
The chain lengths of the E. coli O8 and O9a polymannose O-PSs are regulated by the addition of non-reducing terminal modifications (23). These terminal modifications also serve as an export signal recognized in a serotype-specific manner by the C-terminal domain of the NBD protein Wzt (12, 13). The absence of a non-reducing terminal modification on the K. pneumoniae D-galactan I polymer and the involvement of a Wzt protein lacking an extended C-terminal domain suggest a fundamentally different mechanism for coordinating O-PS synthesis and export. From the studies reported here, the stoichiometry of the ABC transporter and the O-PS biosynthesis machinery is the critical determinant of D-galactan I chain length. Overexpression of the transporter components suggests Wzm and Wzt are both required for this process; neither Wzt nor Wzm alone can effect a change in the average O-PS chain length. In the simplest model, an increase in the relative amount of the O-PS transporter out-competes the biosynthesis machinery for the O-PS substrate, leading to early chain termination.
Coordination of chain extension and export is critical in the K. pneumoniae D-galactan I system. In the absence of the O2a transporter, long unregulated O-PS chains are synthesized, and these are not substrates for export once completed. It is unclear whether this is due only to the disconnection of synthesis and export or whether it reflects a limitation in the maximum size of nascent polymers that the ABC transporter is able to export (or both). These results are in clear contrast to the E. coli O9a system where postsynthetic export is possible even though it is unlikely to be the normal situation in vivo.
The absence of a discrete O-PSbinding domain in Wzt KpO2a dictates a different mechanism for substrate recognition. Clearly the O2a transporter exhibits no specificity for a particular composition/repeat unit structure but must instead recognize some conserved feature in the lipid-linked O-PS substrates. Similar "relaxed specificity" is also seen with PglK and Wzx (47). PglK is an ABC "half"-transporter from Campylobacter jejuni involved in the export of und-PP-linked glycans destined for protein N-glycosylation (48,49). Wzx exports und-PP-linked O-PS repeat units in the Wzy-dependent pathway (for a review, see Ref. 2). PglK and Wzx are functionally interchangeable (47). Both glycans are initiated with an acetamido sugar (predominantly GlcNAc), and this may be the element recognized by the exporters (50). Relaxed specificity is also seen in the export of group 2 CPS in E. coli. The CPS ABC transporter (KpsMT) is functionally interchangeable between serotypes (51,52) and between species (53). It is assumed that there must be a common recognition event in the export process, and it has been proposed that a conserved reducing terminal diacylglycerol phosphate derivative on the capsular polymer provides the export signal (51). By analogy to these systems, the O2a exporter could recognize the carrier lipid, und-PP, and/or the GlcNAc residue that initiates these O-PSs. There is currently no way to discriminate between these possibilities. Plasmid-encoded kpsMT cannot complement a wzm-wzt mutation and vice versa.
With a recognition element at the reducing terminus of the polymer, there is the intriguing possibility that O-PS export can be initiated before chain extension is complete. The question FIGURE 7. Synthesis and export are not coordinated in the E. coli O9a system. A conditional system was established in E. coli CWG638 (⌬wzm-wzt::aphA-3) transformed with pWQ346 containing the O9a transporter genes cloned behind an arabinose-inducible promoter. E. coli CWG634 provides the O9a control. Polymer synthesis was initiated by the addition of mannose, and cells were then washed. One set of samples then received arabinose (Ara) to induce expression of the ABC transporter, and the other was cultured in glucose (Glc). A shows the silver-stained SDS-PAGE gel of LPS in whole-cell lysates (upper panel). B provides the results of immunofluorescence studies (anti-O9a antibody) on whole cells. For each set of images, a differential interference contrast image is shown in the left column, and the corresponding fluorescence image is shown on the right. remains as to whether the transporter actively sequesters the polymer away from the glycosyltransferases to terminate synthesis (and establish modality) or whether the transferases are physically removed from the growing chain once the appropriate modal length is achieved. These issues have critical impact on pathogenesis where serum resistance is dependent on O-PS chain length in many bacterial pathogens (54). The E. coli group 2 CPS system (also ABC transporter-dependent) may provide some insight into processes involved in the coupling of synthesis and export. It has been demonstrated recently that the prototype E. coli group 2 CPS, polysialic acid, is synthesized in a protected environment within the cell (55). The nascent polymer is not accessible to a highly specific glycanase enzyme that can be expressed in the cytoplasm and is presumably protected by biosynthesis and accessory proteins. This would facilitate temporal coupling of polymer biosynthesis and export.
Database searches (data not shown) identify other examples of O-PS NBDs lacking a discrete binding domain, although no biochemical information is available. The O-PS of Y. enterocolitica O:3 NBD provides one example (10). The Vibrio cholerae O1 antigen is, in a sense, a hybrid of the two O-PS ABC transporter models. The polymer can be capped with a non-reducing terminal methyl group, and the presence or absence of this moiety dictates the serological difference between the Ogawa and Inaba serotypes (56,57). However, in contrast to the E. coli O9a system, export of the V. cholerae polymer is not dependent on methylation (58). This can now be explained by the observation that the cognate NBD lacks the C-terminal substrate recognition domain. In this system, chain termination by terminal modification and export seem to be separable events.
In summary, we have now described two mechanisms for O-PS export via ABC transporters. These differ in the nature and flexibility of the substrate recognition process as well as in the processes coupling polymerization (and chain length control) and export. Clues to which process might be involved for a particular polymer are provided by the sequence of the NBD protein, specifically the presence or absence of an accessory domain. There are interesting parallels with the assembly systems for other surface glycans, and it will be interesting to see whether the export principles seen in the two O-PS models are conserved.