Structural Characterization of Closely Related O-antigen Lipopolysaccharide (LPS) Chain Length Regulators*

Background: The length of the O-antigen polysaccharide is determined by polysaccharide co-polymerases(PCPs). Results: We have determined structures of several chimeric PCPs that lead to different polysaccharide lengths and determined their oligomerization state. Conclusion: The polysaccharide length distribution depends primarily on the residues on the outer surface of the PCP oligomer. Significance: Our data provide support for the octameric assembly of WzzB PCPs. The surface O-antigen polymers of Gram-negative bacteria exhibit a modal length distribution that depends on dedicated chain length regulator periplasmic proteins (polysaccharide co-polymerases, PCPs) anchored in the inner membrane by two transmembrane helices. In an attempt to determine whether structural changes underlie the O-antigen modal length specification, we have determined the crystal structures of several closely related PCPs, namely two chimeric PCP-1 family members solved at 1.6 and 2.8 Å and a wild-type PCP-1 from Shigella flexneri solved at 2.8 Å. The chimeric proteins form circular octamers, whereas the wild-type WzzB from S. flexneri was found to be an open trimer. We also present the structure of a WzzFepE mutant, which exhibits severe attenuation in its ability to produce very long O-antigen polymers. Our findings suggest that the differences in the modal length distribution depend primarily on the surface-exposed amino acids in specific regions rather than on the differences in the oligomeric state of the PCP protomers.

Bacterial surface is lined with complex carbohydrate polymers that form a protective barrier against the external environment. For pathogenic bacteria, these polymers are an essential component of the survival toolkit within the host and frequently are critical determinants of pathogen-host interactions. In particular, the LPS O-antigen (OAg) 2 is emerging as an important pathogenesis factor contributing to virulence of a number of human pathogens (1)(2)(3)(4)(5)(6)(7). The O-antigen is the most chemically diverse component of the outer membrane lipopolysaccharide and varies among even the same species of bacte-ria. The mature O-antigen chain is a polymer made of the repeating units, with each O-unit containing 4 -6 sugar residues (8). The proteins involved in the biosynthesis of the nucleotide sugars, glycosyl transfers, and OAg processing machinery are encoded on one or more gene clusters (9,10). The OAg processing machinery is responsible for the assembly of the polymer in the periplasmic space and for the transport to the cell surface (11). In a system referred to as Wzy-dependent synthesis, the individual O-units are synthesized at the cytoplasmic side of the inner membrane, then flipped to the periplasm by a flippase Wzx, and assemble into a high molecular mass polysaccharide by Wzy polymerase (9). The assembly is completed by the ligase WaaL, which transfers the mature O-antigen to the sugar moiety of the outer core of the lipid A molecule (9).
The number of the O-units in a given O-antigen polymer exposed on bacterial surface falls within a narrow range distribution, resulting in polymers of a well defined modal length. Much like the chemical composition of the individual sugar moieties making up the polymer, the length of the polysaccharide is species-and serotype-specific (10,12). Early genetic studies identified a single gene wzzB (previously cld or rol) to be responsible for defining the specific number of the repeat units in a given O-antigen polymer (13). Subsequent biochemical investigations have found that the product of wzzB gene is a homo-oligomeric protein embedded in the inner membrane through two transmembrane helices and containing a large periplasmic domain (14). Shortly after, the presence of similar structural features was discovered in seemingly unrelated proteins involved in trafficking a range of complex polysaccharides in both Gram-negative and Gram-positive bacteria (14 -16). These functionally related proteins were named polysaccharide co-polymerases (PCPs) (17) and are further subdivided into various subfamilies depending on the type of the polysaccharide biosynthetic pathway they are part of and the presence of a cytosolic tyrosine kinase domain (11,17).
Currently, atomic-level structural data are available only for periplasmic domains of a few family members. Crystallographic studies of the WzzB from Salmonella typhimurium (Wzz ST ), FepE (Wzz FepE ) from Escherichia coli O157:H7, and WzzE from E. coli revealed that these proteins adopt a very similar three-dimensional fold despite a very limited sequence similarity (18). The oligomeric chain length regulators resemble elongated bell-shaped structures, with the protomers composed of an ␣-␤ base domain and a long ␣-helix extending about 100 Å away from the inner membrane into the periplasm (18). In the crystal structures, soluble domains of WzzB, Wzz FepE , and WzzE were found in different oligomeric organizations composed of five, nine, and eight protomers, respectively. The size of the oligomers was addressed independently also by electron microscopy and small angle scattering. The study by electron microscopy of the full-length WzzB and Wzz FepE reconstituted in proteoliposomes led Larue et al. (19) to suggest a hexameric structure for these oligomers. A small angle x-ray scattering study of the full-length E. coli (O86:H2) WzzB solubilized in n-dodecyl-␤maltoside suggested on the other hand the presence of tetramers in solution (20), although higher order oligomers were observed when the salt concentration was increased. The contradictory results of these studies performed away from the natural environment within the bacterial membrane leaves uncertain the native oligomeric state of the PCPs within the inner membrane.
Despite the current functional and structural data, little is understood about the precise molecular mechanism underlying the chain length control. The specific modal length appears to be associated with each chain length regulator because when expressed in a closely related bacterial species, the PCP imposes its own characteristic modal length distribution despite the drastically different chemical structure of the carbohydrate polymer (14,21,22). Previously, we have used a structureguided chimeric protein approach to establish which regions of the chain length regulator are responsible for the control of modal length. By swapping various structural elements between two closely WzzB molecules yet imposing distinctly different modal length distributions, Shigella flexneri WzzB SF and S. typhimurium WzzB ST , we were able to produce a number of functional chimeric proteins exhibiting a range of intermediate modal lengths (22). These findings posed a series of further questions, one of which was whether the observed differences in phenotypes could be the result of significant local conformational differences between the wild-type protein and the corresponding chimeric chain length regulators.
To address this question, we determined the structures of the periplasmic domain of the S. flexneri WzzB SF and three chimeric chain length regulator proteins generated from parental WzzB SF and WzzB ST and compared them with the parental structures. The structural alignments demonstrated little differences in the main-chain conformations, suggesting that a given modal length is determined by the side chains of residues in selected regions. Most of these residues were found to be surface-exposed on the external face of the oligomer. The chimeras form oligomers with a variable number of protomers, bell-shaped octamers similar to the oligomers of WzzE, pentamers similar to the previously observed E. coli WzzB, and open-face trimers. The size of the octamers corresponds very well to the electron microscopy data, and we surmise that this is the most likely arrangement of WzzB in the bacterial cell. Lastly, we determined the structure of a Wzz FepE loop deletion mutant, which confers strikingly different modal length from the wild-type protein, and discovered that it forms a nonameric structure very similar to its wild-type counterpart. All of these observations suggest that the modality is dictated predominantly by the nature of the external surface amino acids rather then by structural alterations of the chain length regulator itself.

Construction of Expression Vectors Encoding Periplasmic
Domains of Chimeric Proteins-All plasmids were purified from E. coli DH5␣ using the plasmid mini prep kit (Sigma-Aldrich). The sequences corresponding to amino acids (Glu-54 -Lys-293) were PCR-amplified from the corresponding plasmids pSK10 and pSK11 (22) using the following primer set: 5Ј-GAG CAG GGA TCC GAA AAA TGG ACA TCC ACG GCG (8) and 3Ј-GAC CAG GAA TTC TTA TTT CGG ACT ATC GCG ACG AAT AGG. The resultant PCR product was cut with BamHI and EcoRI and ligated into the pET-based histidine expression vector pFO4. The sequence corresponding to the periplasmic domain of WzzB from S. flexneri was amplified from the genomic DNA and was subcloned in pET-based expression vector as described above. All constructs were verified by sequencing and confirmed to contain a periplasmic domain of the chimeric WzzB with a noncleavable N-terminal hexahistidine tag. Wzz FepE (⌬258:266::GSG) mutant was constructed using the overlap PCR method (40). The sequence (Lys-64 -Pro-330) corresponding to the periplasmic domain was then subcloned into the pET-based expression vector containing N-terminal hexahistidine tag.
Expression and Purification of the Chimeric Proteins-E. coli BL21 (DE3) (Invitrogen) transformed with the above expression plasmids were grown to mid-log phase, cooled to 20°C, and induced with 0.1 mM isopropyl-D-thiogalactoside. Growth was continued for 20 h before the cells were collected by centrifugation at 4,000 ϫ g for 20 min at 4°C. Cell pellets were resuspended and washed twice with ice-cold wash buffer A (200 mM NaCl, 10 mM imidazole, 40 mM Tris, pH 8.0, supplemented with 1 mM p-aminobenzamidine and 100 M leupeptin protein inhibitors). Cells were broken by sonication for a total of 6 ϫ 10-s bursts. Cell debris was removed by centrifugation for 45 min at 15,000 ϫ g in the cold, and the supernatant was incubated with 0.5 ml of 50% nickel-nitrilotriacetic acid-agarose beads for 2 h at 4°C pre-equilibrated with wash buffer A. The beads were washed with 10 column volumes of buffer A containing 20 mM imidazole and 10 column volumes of buffer B (200 mM NaCl, 30 mM Tris at pH 8.0) containing 40 mM imidazole, and chimeric WzzB were eluted in buffer A containing 250 mM imidazole. The purity of the eluted protein was analyzed on 12% sodium dodecyl sulfate (SDS)-polyacrylamide gels and stained with Coomassie Blue, and the buffer was exchanged into 200 mM NaCl, 40 mM Tris, pH 8.0, using a 10DG desalting column (Bio-Rad, Mississauga, Ontario, Canada). The protein was concentrated 10-fold with a Centricon concentrator (30,000 molecular weight cut-off) and was further purified by gel filtration on a Superdex S200 column (GE Healthcare) in buffer B. The purified protein was concentrated to 10 mg/ml in the same buffer, flash-frozen in liquid nitrogen, and stored at Ϫ80°C. Se-Met-containing proteins were obtained using a similar protocol by growing BL21(DE3) on minimal medium containing L-selenomethionine. Protein concentrations were determined with the Bio-Rad protein assay, using bovine serum albumin as a standard, and dynamic light scattering experiments were carried out using a DynaPro plate reader (Wyatt Technologies, Santa Barbara, CA) to assess sample polydispersity prior to crystallization.
Crystallization-The initial crystallization conditions for WzzB SF were identified using Qiagen JCSG Core suite screens (Qiagen, Mississauga, Ontario, Canada) and optimized using the hanging drop vapor diffusion method. The best WzzB SF crystals were grown by equilibration of 1 l of protein (at 6 mg/ml) in buffer 30 mM Hepes, pH 7.5, 200 mM NaCl, with 1 l of reservoir solution containing 20% PEG 4000, 0.1 M Hepes, pH 7.5, and 0.1 M MgCl 2 and equilibrating over the reservoir solution. The crystals belonged to P2 1 with a ϭ 81.5, b ϭ 62.6, c ϭ 90.7 Å with three molecules in the asymmetric unit and V m of 2.75 Å 3 Da Ϫ1 . For each chimeric protein, a number of commercial crystallization kits from Qiagen, Hampton Research, and Molecular Dimensions were screened using the sitting drop vapor diffusion method at 298 K, mixing 400 nl of concentrated protein solution (10 mg/ml) with 400 nl of the corresponding reservoir solution. For SF4, a number of initial hits in various high salt conditions of the MbClass screen (Qiagen) could be seen 5 days later. Most of these were further optimized in Linbro 24-well plates using the hanging drop vapor diffusion method, and the conditions that led to the best diffracting crystal were 3.0 M sodium formate, pH 6.0 (HNaO 2 ). The crystals belonged to the space group I422 with a ϭ 115.7, c ϭ 210.9 Å with two molecules in the asymmetric unit and V m of 3.14 Å 3 Da Ϫ1 . A similar crystal form was obtained for SF5 with initial crystals grown in 1.0 M sodium citrate (unbuffered) and was further optimized by a hanging drop vapor diffusion by equilibration of 1 l of protein concentrated to 8 mg/ml in buffer B with 1 l of reservoir solution. The best diffracting crystals were grown in a condition containing 2.0 M sodium citrate ϩ 5% (v/v) glycerol and, as SF4, belonged to space group I422 with a ϭ 116.9, c ϭ 219.1 Å, and V m of 3.34 Å 3 Da 3 . Native periplasmic Wzz FepE (⌬258:266::GSG) was concentrated to 18 mg/ml in the buffer containing 0.1 M Hepes, pH 7.5, 200 mM NaCl, and 2% w/v glycerol. The crystals were grown by equilibration of 1 l of protein mixed with 1 l of reservoir solution (0.5 M KCl, 0.05 M MOPS, pH 7.0, 12% (w/v) PEG 4000, and 20% w/v glycerol over 500 l of reservoir solution) and belonged to space group C2 with a ϭ 208.3, b ϭ 141.23, and c ϭ 136.9 Å, with nine molecules in the asymmetric unit and V m of 3.82 Å Da Ϫ1 .
X-ray Data Collection, Structure Solution, and Refinement-The crystal, a Se-Met-labeled wild-type S. flexneri WzzB SF , was flash-cooled in a nitrogen stream to 100 K, and diffraction data were collected to 2.8 Å resolution at the CMCF-1 beamline, Canadian Light Source. Ten of the expected 12 selenium atoms were found using ShelxD (23). Initial phasing was performed by Phaser (24), and the dual-space fragment extension procedure was used to improve the initial phases (25). The initial model was built using the autobuild option in Phenix and subsequently was completed manually using the program Coot. Iterative cycles of refinement and manual building were carried out using Phenix (26) and COOT (27), respectively. The final R work and R free were 24.5 and 30.3%, respectively.
Diffraction data for a Se-Met-labeled SF4 crystal flash-cooled in a nitrogen stream at 100 K were collected to 1.6 Å at 31-ID beamline, Advanced Photon Source, Argonne National Laboratory. Data integration and scaling were performed with the program suite HKL3000 (28). Initial phases were determined by single-wavelength anomalous dispersion at the selenium peak energy ( ϭ 0.9794 Å). Eight of the 10 expected selenium atoms in the asymmetric unit were located with the program SHELXD (23), and the initial phases were calculated and improved using the program MLPHARE (29). The model was built by ARP/ wARP (30) using the data set to 1.6 Å resolution and completed manually using the program COOT and refined with Phenix. All measured reflections were used in the refinement. The water molecules were added, and the translation-librationscrew model for anisotropic temperature factors was applied (31). The refinement converged to an R work value of 21.8% and an R free value of 23.8%. The N-terminal histidine tag was not visible in the electron density map and was not included in the final model. Data sets for native SF5 crystals were collected at the CMCF-1 beamline of the Canadian Light Source to 2.6 Å resolution. SF5 structure was solved by molecular replacement with a monomer of SF4 as a search model using the program Phaser. Residues Lys-126 -Gln-129 were also poorly resolved and were not built in either subunit. The refinement was carried out using Phenix with translation-libration-screw model for anisotropic temperature factors. The final R work values were 23.5 and 29.6%, respectively. In each structure, the tips of several side chains, mostly lysines, arginines, and glutamates, were also disordered and were not included in the models. Both SF4 and SF5 have good stereochemistry as analyzed by the program PROCHECK (32).
Data sets for the native Wzz FepE (⌬258:266::GSG) mutant were collected to 2.8 Å resolution at the CMCF-1 beamline, Canadian Light Source, and its structure was solved by molecular replacement with a nonamer of a WT Wzz FepE (18) as a search model using the program Phaser. Direct methods (OASIS as part of the CCP4 suite) were used to improve the phases obtained by the molecular replacement (33), and the whole model was rebuilt by and subsequently refined using Phenix. The density for residues 257-266 was very ambiguous, and they were not modeled. R work and R free converged to 24.2 and 28.9%, respectively. Data collection and refinement statistics are given in Table 1.

RESULTS
In our previous work, we reported construction of several chimeric chain length regulators based on two parental WzzBs, WzzB SF and WzzB ST , which share 70% amino acid identity and yet lead to distinctly different O-antigen modal length distributions (22). The chimeric WzzBs confer the O-antigen modal lengths that are in between those of the two parental WzzBs. We have previously reported the structure of the WzzB ST (18).
In an effort to understand the structural basis for the observed phenotypic outcomes, we first determined the crystal structure of the periplasmic domain of WzzB SF (amino acids 52-293) and subsequently of two chimeras, which we refer to as SF4,  (Fig. 1A). The two chimeric proteins share 96% sequence identity (Fig. 1B) and are over 80% identical to the WzzB SF and WzzB ST parents (Fig. 1B). Remarkably, these minor differences in the amino acid sequence are sufficient to cause pronounced differences in modal lengths of the resulting O-antigen chains; WzzB SF leads to a modal length of ϳ13 repeating units, WzzB ST leads to ϳ20 repeating units, SF4 leads to ϳ17 repeating units, and SF5 leads to ϳ15 repeating units (Fig. 1C).
Structure of WzzB SF -The crystals of WzzB SF were obtained using PEG 4000 as a precipitating agent, and the structure was solved using the experimental phases derived from a single anomalous dispersion experiment to eliminate model bias. Unlike the WzzB ST (18), this protein forms open trimers in the crystal ( Fig. 2A). The same organization was also previously found for the periplasmic domain of WzzB EC from E. coli W3110, determined at a low resolution (18). 3 Interestingly, all three WzzB SF subunits within the trimer are bent to a different degree near their midpoints of the central, long helix ␣3, resulting in movements of the top parts of the molecules (loop L4) by several Å, with the largest tilt in the C subunit (Fig. 2, B and C). The observed different bending of the central helix may stem from crystal packing differences and a nonidentical environment of the inner and outer protomers in the trimer. This indicates a certain amount of flexibility of the central helix that may play a functional role of the chain regulators in vivo. In the pentameric structure of WzzB ST reported previously by our group (18), these conformational changes could not be discerned as the top parts of all the subunits in the pentamer were disordered. The two outside protomers pack tightly to the central protomer along their entire lengths in a manner very similar to that observed previously in the closed oligomers of Wzz FepE and WzzE and the open WzzB Ec trimer (18). The intersubunit interface area, as calculated by the PISA server (34), is ϳ1950 Å 2 . This interprotomer packing observed in the pentamers of WzzB ST is an exception in that the contacts are made predominantly by the base domains with the intersubunit interface area being only 570 Å 2 (18).
Structures of SF4 and SF5 Chimeras-Both of the chimeric proteins were crystallized under high salt conditions and produced similar crystal forms, both belonging to the I422 space group but with dimensions along the c-axis differing by 8 Å (Table 1). Each crystal contained two protomers in the asymmetric unit (Fig. 3A), which assembled into the bell-shaped octamers through 4-fold crystallographic symmetry (Fig. 3B). This oligomerization state is different from the pentameric assembly previously observed for wild-type WzzB ST and the open trimers of WzzB SF (Table 2) but very similar to the oligomeric form of WzzE. The bell-shaped oligomer has the height of 95 Å and an outer diameter of 90 Å at the base and 55 Å at the top. These dimensions are similar to those seen for the WzzE octamer (95, 113, and 67 Å, respectively (18)) and significantly larger then the pentameric WzzB ST (18). In both chimeric proteins, the loop region facing the internal cavity of the oligomer (loop L1, residues 124 -131) is poorly ordered, and consequently, this region was not modeled.
The tightly packed protomers form numerous side-to-side contacts via ϳ15 hydrogen bonds, five salt bridges, and numerous van der Waals contacts, which extend along the entire length of each protomer. In the base domain, the contacts are made between strand ␤4 and helix ␣3 of one protomer with helix ␣2Ј and the bottom of strand ␤1Ј of the other. The long central helix ␣3 also contributes to the intermolecular contacts forming numerous interactions with the residues in helices ␣4Ј and ␣5Ј of the neighboring subunit. A significant fraction of residues in loop L4 at the top of the oligomer is also engaged in multiple contacts with the residues in the L4 regions of the neighboring protomers. Overall, the surface area of the inter-   MAY 4, 2012 • VOLUME 287 • NUMBER 19

JOURNAL OF BIOLOGICAL CHEMISTRY 15699
subunit interface seen in the chimeric chain length regulators constitutes 1935 and 1906 Å 2 (or SF4 and SF5, respectively).
In the previous structural studies of class-1 polysaccharide co-polymerases, the regions of the protomers most distant from the membrane, comprising the top of the central helix and the following residues, were disordered to various extents in all of the crystal structures (18). Our subsequent functional studies demonstrated that this part of the molecule plays an essential role in several O-antigen chain length regulators (22). In the structures of both SF4 and SF5, most of this segment could be reliably modeled (Fig. 4A), providing the first view of the stable conformation of the region most distant from the membrane and capping the top of the oligomer. The conformation of this loop differs slightly between the two molecules in the asymmetric unit and is also somewhat different in the two chimeras, reflecting a degree of inherent flexibility of this loop.
Further structural comparisons revealed that the chimeric proteins are very similar, with the root mean square deviation for all C␣ atoms 0.4 Å. The amino acids differing between the two chimeric proteins are predominantly distributed throughout the top of the molecule and are found at the tip of the central long helix ␣3, in the loop L4 region, and within helix ␣4. Out of the 13 amino acid differences, most either represent conservative substitutions (I217V, D234E, R205Q, M228L) or are associated with a loss or a gain of a charge (E201Q, D208N, E209Q, K211Q, Q214K, Q221E). One substitution (K237E) leads to a charge reversal, and two are nonconservative substitutions (Q241L, A253N) (Fig. 4A). Six out of these 13 residues (residues 205, 214, 217, 221, 228, and 234) are involved in intersubunit contacts, yet these substitutions do not affect significantly the protomer association. In all of the proteins examined, the internal cavity is rich in positively charged regions, whereas the external surface contains patches of both positive and negative charges. The profile of the solvent-accessible surface revealed that some of the polar residues, which differ between the two chimeras, are highly solvent-exposed (Fig. 4B). Mainly, these are found at the tip of the central long helix ␣3 (205 and  209), in the first part of the loop L4 region (211 and 214), and within the helix ␣4 (241) (Fig. 4B).

Structure of Wzz FepE Loop L4
Mutant-In our previous work, we found the loop L4 region to be essential for OAg chain length determination as various modifications to this part of the protein resulted in drastic reductions of modality regulated by Wzz FepE and completely abolished the function of WzzB (22). We have further investigated this notion and sought to determine whether a more subtle mutation of the loop L4 would be sufficient to alter the modality. To this end, we substituted the first part of L4 in Wzz FepE , composed of the amino acids Ser-258 -Asp-265, by a Gly-Ser-Gly linker (Fig. 1A). This modification, which affected only the tip of the L4 loop, rendered the chain length regulator to produce the OAg modality comparable with what was observed by the other loop L4 mutants (Fig.  5A). To determine whether this replacement led to conformational changes in the Wzz FepE structure that could account for a significantly reduced modal length, we have solved the crystal structure of the periplasmic domain of this Wzz FepE (GSG) mutant at 2.8 Å resolution. The mutant was found to crystallize in the C2 space group with nine molecules in the asymmetric unit forming a circular, closed nonamer, the same oligomeric assembly as was previously seen for the wild-type protein (18). As in the wild-type Wzz FepE , the electron density for the GSG linker replacing loop L4 could not be unequivocally traced in the electron density map, indicating its flexibility, but the main chain of the wild-type and mutant Wzz FepE were nearly identical with a root mean square deviation of 0.37 Å for all C␣ atoms (Fig. 5B).

DISCUSSION
What is the oligomeric state of WzzBs in the periplasm? In the previous crystallographic studies (18), the periplasmic domains of the three different PCPs exhibited different oligomerization states, with Wzz FepE being a nonamer, WzzE being an octamer, and WzzB ST being a pentamer. Although the FIGURE 3. Crystal structure of SF4 chimeric chain length regulator. A, structural alignment of the two crystallographically independent subunits (subunit A is orange, and subunit B is blue). B, a side view and an end-on view of the graphic representation of the octamer obtained by the application of the crystallographic symmetry. C, structural superposition of the WzzB SF trimer (red) on the octamer of SF4. The trimer has smaller radius of curvature than the octamer.  . Differences between SF4 and SF5. A, residues differing between the SF4 and SF5 are shown explicitly in stick representation. Helices are colored red, strands are magenta, and loops are green. B, surface accessibility analysis of the two WzzB chimeras with the most buried residues colored in blue, whereas the most solvent-accessible residues colored in red. The most surface-accessible residues differing between the two chimeras are colored in green and are depicted in ball-and-stick representation.
Å. This was smaller than the dimensions of the Wzz FepE nonamer, slightly smaller than the WzzE octamer, and larger than the WzzB ST pentamer. The authors derived a model of a hexamer based on the crystal structure of WzzE that measured 100 Å at the base, 95 Å in length, and 55 Å at the top and concluded that this model had the size compatible with the EM reconstruction. Here, we show that the WzzB chimeras SF4 and SF5 form octamers in the crystal with dimensions of 95 Å across the base, 100 Å in length, and 50 Å across the top of the octameric ring. These octamers are therefore in a very good agreement with the EM data. Moreover, removing two protomers from this octamer would result in a smaller oligomer size than constructed by Larue et al. (19), too small to fit the EM data. To further investigate this, we have analyzed the curvature of the open trimers of WzzB SF . Superposition of this trimer on the SF4 octamer shows that it displays higher curvature (smaller radius) (Fig. 3C). We have computationally extended the trimer following its curvature and found that this accommodates a hexamer. However, the dimensions of this artificially constructed hexamer would be ϳ70 Å, which is inconsistent with the EM studies of the full-length proteins (19) and with the number of protomers in a circular oligomers observed in the crystal structures. Clearly, a crystal structure of the full-length WzzB is needed to clarify the controversy surrounding the real oligomerization state of these molecules and to determine whether the size of the oligomer is indeed playing a role in the determination of the length of the carbohydrate polymer. However, the fact that the size of the octamers is in agreement with that derived for the full-length proteins resuspended in proteoliposomes supports the view that octamers represent the physiologically relevant assembly of WzzB. Interestingly, the cytosolic kinase domain from a PCP-IIa family member Wzc was also found to assemble as octamers (36). Wza, an outer membrane lipoprotein, proposed to form a trans-envelope complex with Wzc, also adopts the octameric organization in the crystals (37).
The EM images of full-length Wzz FepE reconstituted in lipid bilayers indicated particles of ϳ100 Å diameter at the base (19), as compared with 115 Å of the nonamer in the crystal, which the authors suggested would be more consistent with a hexamer. Here, we report the crystal structure of a Wzz FepE mutant crystallized under significantly different crystallization conditions than the wild type and in a different space group (C2 versus P6 3 22 or P3 2 21) and yet forming a nonamer with the same overall dimensions as that of the wild-type Wzz FepE oligomer. Although we cannot exclude the possibility that the presence of the transmembrane regions might affect the number of protomers within the oligomer through their interaction with the surrounding lipid bilayers, we surmise that the same oligomeric arrangement with nine protomers is also likely for the full-length protein. We have remeasured the distances between the particles in the negatively stained EM micrographs (Fig. 4F in Ref. 19) using the specified calibrated pixel size of 3.84 Å and obtained the distance between the centers of the neighboring subunits of 115 Å for a full-length Wzz FepE , which is very similar to the size of the nonamer observed in the crystals.
Another interesting finding that emerged from the structural analysis of Wzz FepE (⌬258:266::GSG) mutant is that it exhibits the same main-chain conformation and oligomeric state, yet leads to a drastically different modal length distribution. We previously considered that the oligomer size might be a significant factor in the OAg modal length determination mechanism, with the larger oligomers producing polymers containing more repeating O-units (12,18). Kintz and Goldberg (41) recently reported that a residue involved in maintaining the oligomer stability of Wzz FepE homologue from Pseudomonas aeruginosa (Wzz2) is partially responsible for the specification of a given O-antigen chain length. The structure of our Wzz FepE mutant, which leads to much shorter O-antigen chains and yet forms the same oligomer as the wild-type Wzz FepE , strongly suggests that the oligomeric size alone is not a sole factor of the modal length determination. Instead, our finding reiterates the importance of the flexible region at the top of the molecule in the mechanism of the OAg chain length control. This is interesting in light of the reported observation that OAg modality can be reproduced in vitro with only Wzy polymerase and WzzB chain regulator present (38). Loop L4 region is positioned too far from the inner membrane to form physical pro- tein-protein interactions with the Wzy polymerase, and therefore, L4 is likely to serve as one of the sugar-binding faces interacting with the polar O-antigen subunits. Indeed, most of the differences between SF4 and SF5 chimeric proteins, which produce the O-antigens with distinctly different modal lengths, also fall in this region (Fig. 4A). As subtle as the differences are, they would be sufficient to confer differential affinities to the O-antigen-PCP interaction. Recently, it was shown that the Wzy polymerase-O-antigen interactions in Pseudomonas are very sensitive to even conservative amino acid changes in specific positions as even Arg-to-Lys mutation led to a complete loss of the polymerase function, which was attributed to the loss of sugar binding (39). Clearly, the top of the molecule is not the only site of O-antigen-PCP interaction. In our previous study, we established the importance of the 269 -274 segment. This region is the same in SF4 and SF5 and contains only one negatively charged residue (Asp-273); however, in WzzB SF , it contains three negatively charged residues (Asp-270, Asp-271, and Asp-274). This patch would thus be expected to form a different network of interactions with the incoming repeat units, perhaps leading to a weaker affinity between the O-antigen and the chain length regulator.
In summary, our work provides another structural insight into the PCP-1 machinery involved in the secretion of high molecular weight sugar polysaccharides at the atomic resolution. It demonstrates previously unseen oligomerization states for the WzzB periplasmic domains, of which the octameric assembly appears to be more physiologically relevant than the previously reported pentameric arrangement, and is in agreement with the EM data reported for the full-length proteins. The structures of the chimeric domains show for the first time the structural details of the top region of the oligomer, which is involved in PCP function. Our structural analysis of the Wzz FepE mutant also brings forth two important points. First, it shows that the mutation of the L4 region does not exert its phenotypic effect through the structural effects on the individual protomers. Second, the finding that the periplasmic domains of both wild-type and mutant FepE were found to be nonamers despite the wide differences of the crystallization conditions provides the evidence that this arrangement may indeed represent the preferred physiological state of the Wzz FepE O-antigen chain length regulator.