The Protein BpsB Is a Poly-β-1,6-N-acetyl-d-glucosamine Deacetylase Required for Biofilm Formation in Bordetella bronchiseptica*

Background: The Bordetella polysaccharide (Bps) is involved in Bordetella biofilm formation. Results: BpsB is a periplasmic metal-dependent poly-β-1,6-N-acetyl-d-glucosamine (PNAG) deacetylase that has unique structural and functional features from known PNAG deacetylases. Conclusion: BpsB-dependent deacetylation of Bps is required for Bordetella bronchiseptica biofilm formation. Significance: Deacetylated Bps is a key component for the structural complexity of Bordetella biofilms. Bordetella pertussis and Bordetella bronchiseptica are the causative agents of whooping cough in humans and a variety of respiratory diseases in animals, respectively. Bordetella species produce an exopolysaccharide, known as the Bordetella polysaccharide (Bps), which is encoded by the bpsABCD operon. Bps is required for Bordetella biofilm formation, colonization of the respiratory tract, and confers protection from complement-mediated killing. In this report, we have investigated the role of BpsB in the biosynthesis of Bps and biofilm formation by B. bronchiseptica. BpsB is a two-domain protein that localizes to the periplasm and outer membrane. BpsB displays metal- and length-dependent deacetylation on poly-β-1,6-N-acetyl-d-glucosamine (PNAG) oligomers, supporting previous immunogenic data that suggests Bps is a PNAG polymer. BpsB can use a variety of divalent metal cations for deacetylase activity and showed highest activity in the presence of Ni2+ and Co2+. The structure of the BpsB deacetylase domain is similar to the PNAG deacetylases PgaB and IcaB and contains the same circularly permuted family four carbohydrate esterase motifs. Unlike PgaB from Escherichia coli, BpsB is not required for polymer export and has unique structural differences that allow the N-terminal deacetylase domain to be active when purified in isolation from the C-terminal domain. Our enzymatic characterizations highlight the importance of conserved active site residues in PNAG deacetylation and demonstrate that the C-terminal domain is required for maximal deacetylation of longer PNAG oligomers. Furthermore, we show that BpsB is critical for the formation and complex architecture of B. bronchiseptica biofilms.

Bacteria belonging to the genus Bordetella are Gram-negative coccobacilli that cause respiratory symptoms and illnesses in humans, animals, and birds (1,2). Currently, nine species belong to this genus, three of which have been studied in detail as follows: Bordetella pertussis, Bordetella parapertussis, and Bordetella bronchiseptica. B. pertussis and B. parapertussis are the causative agents of whooping cough in humans. In contrast, B. bronchiseptica has a broad host range, colonizing and causing disease in a wide variety of animals. B. bronchiseptica can also infect immunocompromised and healthy humans thereby demonstrating zoonotic transmission (2,3).
Despite high levels of vaccination coverage, the incidence of pertussis is increasing steadily in the United States and other developed countries, leading the Centers for Disease Control to classify pertussis as a re-emerging disease (4 -6). Similarly, although multiple vaccines are used with variable success for B. bronchiseptica, animals continue to be carriers and frequently shed bacteria resulting in outbreaks among herds (2). This has generated renewed interest in identifying and understanding the role of new virulence determinants, and in uncovering mechanisms employed by these bacteria for host survival and persistence.
It has been proposed that the colonization and persistence of Bordetella are enhanced by the formation of biofilms in the respiratory tract. Several studies have shown that Bordetella spp. form biofilms on artificial surfaces (3,(7)(8)(9)(10)(11)(12) and that both B. pertussis and B. bronchiseptica form biofilms in the mouse respiratory tract (13)(14)(15)(16)(17). A biofilm is composed of a complex extracellular matrix that encases microcolonies of bacteria (18,19), which limits the diffusion of antimicrobials and occludes phagocytosis (20). The composition of the biofilm matrix can be quite diverse but often includes key components such as proteinaceous adhesins, nucleic acids, and exopolysaccharides (20 -22). We have shown that the major Bordetella adhesin FHA, DNA, and the Bps 6 exopolysaccharide are critical for the structural stability of Bordetella biofilms both in vitro and in the mouse trachea and the nose (13)(14)(15).
The bpsABCD operon, which encodes the machinery responsible for Bps production, is highly conserved in Bordetella species (15,16). In addition to biofilm formation, this operon is essential for innate immune resistance and for the colonization of the mouse respiratory tract (15,16,23). Based on immune reactivity and enzymatic susceptibility (8,15), Bps appears to be similar to poly-␤(1,6)-N-acetyl-D-glucosamine (PNAG or PGA), also known as polysaccharide intercellular adhesin that was originally isolated from the biofilms of Staphylococcus epidermidis (24) and Staphylococcus aureus (25). The genetic locus is homologous to the pgaABCD operon in Escherichia coli (26) and other Gram-negative bacteria (8). Previous bioinformatics analysis (8) and homology to the pgaABCD operon suggest that Bps biosynthesis occurs in a synthase-dependent manner similar to PNAG in E. coli (27). BpsC is predicted to be an inner membrane protein that contains a cytosolic glycosyltransferase domain that likely synthesizes Bps and transports it across the inner membrane. BpsD is predicted to be a small inner membrane protein that likely interacts with BpsC to stimulate Bps biosynthesis and membrane translocation, as shown for PgaC and PgaD (28). BpsA is predicted to be an outer membrane protein that encodes a ␤-barrel domain that could facilitate the export of Bps across the outer membrane and a periplasmic domain that has sequence homology to the tetratricopeptide repeat (TPR) protein interaction motif. BpsB is predicted to be a two-domain periplasmic protein. The N-terminal domain belongs to the family four carbohydrate esterases (CE4s) as defined by the CAZy database (29), and it is likely responsible for the partial deacetylation of Bps similar to the role of PgaB on PNAG (30 -33). The C-terminal domain of BpsB is annotated as a member of the glycoside hydrolase 13-like family (29) and may play a role in modifying or binding Bps as seen for PgaB with PNAG (32). Thus, although the function of the individual genes of the bpsABCD operon can be inferred from bioinformatics analysis, none of these genes have been characterized experimentally.
Here, we present the structure of BpsB's N-terminal domain (BpsB(35-307)) from B. bronchiseptica, provide enzymatic evidence that BpsB is a periplasmic PNAG deacetylase, and evaluate its role in Bps export and Bps-dependent biofilm formation. The structure of BpsB  shows a circularly permuted CE4 domain similar to PgaB (30,34) and IcaB (35) with several distinct structural differences. BpsB displays length-dependent PNAG deacetylation activity with the highest rates measured when Ni 2ϩ is present. The N-terminal domain of BpsB shows ϳ46% of the deacetylation activity observed in the full-length protein, and mutational analysis identifies key catalytic residues. Studies in vivo show that a ⌬bpsB mutant strain is still able to export Bps to the cell surface but displays a biofilm-defective phenotype and significantly altered biofilm architecture.

Experimental Procedures
Bacterial strains, plasmids, and oligonucleotide primers used in this study are described in Table 1. The purified BpsB constructs used for crystallization and enzymatic assays carry the C48S mutation to prevent cross-linking and to avoid the use of reducing agents.
Cloning The following protocol was used to express and purify all the BpsB constructs. E. coli BL21-CodonPlus cells transformed with the appropriate plasmid were grown in Luria-Bertani (LB) broth plus 50 g/ml kanamycin at 37°C with shaking at 200 rpm to an A 600 of ϳ0.4 -0.5 and moved to 18°C. After 20 min, the cultures reached an A 600 of ϳ0.6 -0.7, and protein expression was induced by the addition of isopropyl D-1-thiogalactopyranoside to a final concentration of 1 mM. The cells were incubated overnight at 18°C, harvested by centrifugation at 5000 ϫ g for 20 min, and frozen on dry ice. Frozen cell pellets 6 The abbreviations used are: Bps, Bordetella polysaccharide; PNAG, poly-␤-1,6-N-acetyl-D-glucosamine;   MMS for metal-mediated symmetrization) was concentrated to ϳ9 -10 mg/ml and incubated with 2 M eq of either NiCl 2 , CuCl 2 , or CdCl 2 . The protein solutions were then screened for crystallization conditions using the  Table 2). The data were indexed, integrated, and scaled using HKL2000 (36). The structure was determined by molecular replacement with PHENIX AutoMR (37) using a modified version of the N-terminal domain of PgaB(42-655) (Protein Data Bank code 4F9D) as a search model (residues 310 -646 were removed from the model). The resulting electron density map enabled PHENIX AutoBuild (37) to build ϳ 90% of the protein. The remaining residues were built manually in COOT (38) and alternated with refinement using PHENIX.REFINE (37). Translation/Libration/Screw groups were used during refinement and determined automatically using the TLSMD web server (39,40). Structure figures were generated using PyMOL Molecular Graphics System (DeLano Scientific), and quantitative electrostatics were calculated using PDB2PQR (41,42) and APBS (43). Programs used for crystallographic data processing and analysis were accessed through SBGrid (44).
Preparation of PNAG Oligomers-␤-1,6-GlcNAc oligomers were prepared and purified, and their identities were confirmed as outlined previously (45). The ␤-1,6-GlcNAc oligomers were stored as lyophilized powders at room temperature and dissolved in deionized water for use in assays. Accurate oligomer concentrations were determined by 1 H NMR using dimethylformamide as an internal standard. B. bronchiseptica Growth Conditions-Strains were maintained on Bordet-Gengou (BG) agar supplemented with 7.5% (v/v) defibrinated sheep's blood and 50 g/ml streptomycin. B. bronchiseptica strains were grown overnight at 37°C in Stainer-Scholte (SS) broth supplemented with 40 mg/liter L-cysteine, 10 mg/liter FeSO 4 , 4 mg/liter niacin, 150 mg/liter glutathione, and 400 mg/liter ascorbic acid. The overnight culture was used to inoculate fresh medium at a 1:100 dilution and grown at 37°C to an A 600 ϭ 1 (corresponding to 1 ϫ 10 9 cells). For plasmid where I (k) and ͗I͘ represent the diffraction intensity values of the individual measurements and the corresponding mean values. The summation is over all unique measurements. b R work ϭ Α ʈF obs ͉ Ϫ k͉F calc ʈ/͉F obs ͉, where F obs and F calc are the observed and calculated structure factors, respectively. c R free is the sum extended over a subset of reflections (5%) excluded from all stages of the refinement. d Data were calculated using MolProbity (68). e Maximum-likelihood based coordinate error, as determined by PHENIX (37). maintenance, 50 g/ml chloramphenicol was included in the media for strains harboring the GFP-expressing and complementation (⌬bpsB comp ) or vector (⌬bpsB vec ) plasmids.
Chromosomal Deletion of bpsB-An in-frame nonpolar bpsB deletion in B. bronchiseptica RB50 was constructed using allelic exchange as described previously for the bps locus (8). This plasmid was constructed by ligating the regions flanking 5Ј of bpsB and 3Ј of bpsC for allelic exchange. The 584-bp XbaI-HindIII upstream fragment, including the first 16 bpsB codons, was amplified from B. bronchiseptica RB50 genomic DNA using primers SMP31 and SMP32. The 528-bp HindIII-KpnI downstream fragment, including the last 19 codons of the bpsB reading frame, was similarly amplified using primers SMP33 and SMP34. Using the XbaI-KpnI-digested allelic exchange vector pRE112 together with the XbaI-HindIII and HindIII-KpnI-digested fragments, a three-way ligation was performed resulting in plasmid pSM11. This plasmid was transformed into E. coli strain SM10pir and subsequently conjugated into B. bronchiseptica RB50. Exconjugants were selected on BG agar containing 50 g/ml chloramphenicol and streptomycin to select for single and double cross-over recombinants. Double crossover recombinants were selected on LB agar containing 7.5% (w/v) sucrose. Sucrose-resistant colonies that were streptomycin-resistant but chloramphenicol-sensitive were putative bpsB deletions. The genotype of the ⌬bpsB strain was confirmed by PCR and sequencing.
Genetic Complementation of bpsB-The bpsB gene plus 23 bp upstream of the translational start site and 33 bp downstream of the termination codon were amplified from B. bronchiseptica RB50 genomic DNA using primers SMP3 and SMP4. The resulting PCR fragment was confirmed by sequencing and contained flanking KpnI and HindIII restriction sites that were used to subclone the bpsB gene into the vector pBBR1MCS. This created the complementation plasmid pSM8 that was transformed into the B. bronchiseptica RB50 ⌬bpsB strain.
Construction of the B. bronchiseptica ⌬bpsB GFP Strain-The GFP plasmid, pTac-GFP described previously (8), was transformed into the B. bronchiseptica RB50 ⌬bpsB strain, and transformants were selected on BG agar containing 50 g/ml chloramphenicol and streptomycin. Randomly picked colonies containing pTac-GFP were grown in SS broth with 50 g/ml chloramphenicol and were analyzed for GFP expression utilizing a Nikon Eclipse TE300 inverted microscope, and one of the GFP-expressing clones was chosen for experimental analysis. Comparison of the B. bronchiseptica RB50 ⌬bpsB GFP-expressing strain with respective parental strains not containing the GFP plasmid revealed no differences in growth in batch cultures or colony morphology on BG agar containing blood.
Subcellular Localization of BpsB-The bpsB gene starting 23 bp upstream of the translational start site was amplified from B. bronchiseptica RB50 genomic DNA using primers SMP67 and SMP68. The resulting PCR fragment was confirmed by sequencing and contained flanking XbaI and SacI restriction sites that were used to subclone into the vector pBBR1MCS. This produced the plasmid pSM27 that encodes bpsB with a C-terminal FLAG tag. Plasmid pSM27 was then transformed into the B. bronchiseptica ⌬bpsB strain. Bacteria were grown in SS medium supplemented with 50 g/ml chloramphenicol at 37°C until the A 600 ϳ1.5. Approximately 30 ml of cells at an A 600 ϳ1.5 were harvested by centrifugation (30 min, 3000 ϫ g, 4°C), and pellets were resuspended in 1.25 ml of 10 mM Tris-HCl, pH 8.0, 20% (w/v) sucrose, 1 mM EDTA, and 0.1 mg/ml lysozyme. Following incubation on ice for 10 min, another 1.25 ml of 10 mM Tris-HCl, pH 8.0, was added. The resuspension was then subjected to a freeze/thaw cycle and was further lysed by sonication. The cell lysate was clarified by centrifugation (10 min, 20,000 ϫ g, 4°C), and the supernatant was further centrifuged (60 min, 150,000 ϫ g, 4°C) to separate the bacterial membranes from the soluble cytoplasmic and periplasmic proteins. The soluble protein fraction was carefully withdrawn with a syringe and stored at Ϫ80°C for further analysis. The membrane pellet was washed once with 1 ml of 10 mM Tris-HCl, pH 8.0, and then centrifuged (45 min, 100,000 ϫ g, 4°C), and the supernatant discarded. The membrane pellet was resuspended in 1 ml of 50 mM Tris pH 8.0, 2% (v/v) Triton X-100, 20 mM MgCl 2 , and 1 mM PMSF. The resuspended membranes were incubated for 30 min on ice and then centrifuged (60 min, 100,000 ϫ g, 4°C) to separate the inner and outer membranes. The supernatant containing the solubilized inner membranes was isolated and stored at Ϫ80°C. The outer membrane pellet was resuspended in 500 l of 50 mM Tris-HCl, pH 8.0, and stored at Ϫ80°C. The isolated protein fractions were analyzed by SDS-PAGE and Western blot analysis using an ␣-FLAG antibody.
Purification and Immunoblot Detection of Bps Polysaccharide-B. bronchiseptica RB50, ⌬bpsABCD, and ⌬bpsB strains were grown to stationary phase (A 600 ϳ3) and harvested by centrifugation (20 min, 3000 ϫ g, 4°C). The cell pellet was resuspended in 30 l of 0.5 M EDTA, boiled for 10 min, and centrifuged (5 min, 12,000 ϫ g, 25°C) to remove cell debris. The supernatant was transferred to a new tube and treated with 1 ml of phenol/chloroform to remove proteins. The aqueous phase was extracted with an equal volume of chloroform, precipitated using 100% ethanol overnight at Ϫ20°C, and then centrifuged (5 min, 12,000 ϫ g, 25°C). The pellet was washed with 70% ethanol and resuspended in one-tenth volume of water (ϳ100 -250 l). The solution was then treated with DNase I (0.1 mg/ml) and RNase (0.1 mg/ml) (2 h at 37°C) to remove nucleic acids, followed by proteinase K treatment (2 h, 65°C) to degrade any remaining proteins. A final phenol/chloroform extraction and ethanol precipitation were performed before the pellet was resuspended in water and stored at Ϫ20°C. Purified Bps polysaccharide from the B. bronchiseptica RB50, ⌬bpsABCD, and ⌬bpsB strains were spotted as 10-l samples on a nitrocellulose membrane and allowed to dry overnight. The samples were then probed using a goat antibody raised against S. aureus deacetylated PNAG conjugated to diphtheria toxoid (8,48).
Biofilm Formation on Glass Coverslides-Log phase cultures of RB50 and mutant strains (A 600 ϳ0.05) were inoculated into a 50-ml conical tube (GeneMate), containing a sterile vertically submerged glass coverslip (rectangular, 22 ϫ 60, Fisher) and 15 ml of SS medium. Conical tubes were incubated at 37°C under static growth for 96 h, with medium being changed every 24 h. At specific time points, coverslips were taken out, and unattached bacteria were removed by three washes with sterile phosphate-buffered saline (PBS). The coverslips were then transferred into a new 50-ml conical tube containing 10 ml of PBS and vortexed for 2 min at high speed, resulting in detachment of bacteria from the glass surface. Detached bacteria were diluted and plated on BG blood agar plates containing streptomycin.
Scanning Electron Microscopy (SEM) of B. bronchiseptica Biofilms-B. bronchiseptica strains were inoculated at an A 600 of 0.05 into 50-ml conical tubes containing 5 ml of SS broth and a vertically submerged sterile glass coverslip (rectangular 22 ϫ 22 mm, Fisher), resulting in the formation of biofilms at the air-liquid interface. After incubation under static growth conditions for 96 h at 37°C, the coverslips were removed, washed gently with sterile PBS, fixed with 2.5% (v/v) glutaraldehyde, and processed for SEM as described previously (8).
Continuous Flow Confocal Microscopy-Three-chambered flow cells were obtained as sterile units from Stovall. Exponentially grown (A 600 ϭ 0.1) B. bronchiseptica RB50 and ⌬bpsB strains containing a GFP plasmid were inoculated into the chambers. Bacteria were allowed to attach to the chamber without flow for 2 h at 37°C. After attachment, the chamber was inverted, and medium flow was initiated using SS broth containing 50 g/ml chloramphenicol at a rate of 0.5 ml/min over a time course of 96 h. Biofilms were observed every 24 h using a Zeiss LSM 510 confocal scanning laser microscope as described previously (8). Biofilm images were displayed and analyzed using the LSM Image Browser software. The COMSTAT computer software package (49) was used to calculate the average and maximum thickness of the biofilms.
Detection of Bps by ELISA-Bacterial strains (10 7 cfu) resuspended in 100 l of PBS or 100 l of PBS were added into the wells of a 96-well enzyme-linked immunosorbent assay (ELISA) plate (Nunc). Samples were incubated at 4°C overnight. The samples were then aspirated, and the plate was blocked with 5% (w/v) skim milk for 1 h and then washed three times with PBS/Tween 20 (PBST). Indicated dilutions of the human serum (15) that had been repeatedly absorbed against the ⌬bps strain (50) were used as the primary antibody diluted in PBST with 5% skim milk for 1 h. The plate was then washed five times with PBST before addition of the goat anti-human horseradish peroxidase detection antibody at a 1:1000 dilution in PBST for 1 h at room temperature, followed by five washes with PBST. Finally, 100 l of tetramethylbenzidine (Sigma) was added to the wells and incubated for 15 min in the dark before the addition of 50 l of 1.0 M H 2 SO 4 . The plate was then read at 450 nm using a plate reader.
Transmission Electron Microscopy (TEM) of B. bronchiseptica Cells-TEM analysis for the presence of Bps polysaccharide on the surface of B. bronchiseptica was carried out by adsorbing 1 ϫ 10 6 bacteria onto carbon-coated gold grids (Electron Microscopy Sciences) in a humidified chamber for 60 min. The bacteria on the grids were exposed to 10% (v/v) heat-inactivated Bps-enriched human serum for 60 min after blocking with 2% (w/v) BSA in PBS. The bound antibodies were detected with 6 nm gold-labeled anti-human polyclonal antibodies at a dilution of 1:10 in the blocking buffer. Finally, the bacteria were subjected to negative staining with 2% (w/v) phosphotungstic acid, pH 6.6, and analyzed with a Tecnai transmission electron microscope.

Domain Analysis and Subcellular Localization of BpsB-
Full-length BpsB from B. bronchiseptica is homologous to E. coli PgaB, and its N-terminal domain is also homologous to S. epidermidis IcaB with sequence identities determined using ClustalW (51) for their mature sequences of 37 and 17%, respectively. Blast searches and structural prediction using Phyre 2 (52) suggest BpsB contains two domains. The N-terminal domain encodes a CE4, and the C-terminal domain is predicted to be similar to GH13 family members that are primarily ␣-amylases (Fig. 1A) (53). Despite significant sequence homology, BpsB has three distinct differences from PgaB. First, residues 1-26 of BpsB are predicted to encode a signal sequence by SignalP (54). The protein does not contain an outer membrane lipidation site found in PgaB (26,55). Second, the linker region between the N-and C-terminal domains is 2-3 times longer in BpsB than PgaB (ϳ12-24 versus 7 residues). Third, there are an additional 30 residues at the C terminus of BpsB that are predicted to adopt a random coil structure. Compared with IcaB, BpsB does not contain the hydrophobic residue loop L1 that is proposed to retain the enzyme to the outer leaflet of the membrane (35), and the sequence homology is predominantly based on the five canonical CE4 motifs.
As BpsB does not have a predicted lipidation site, we used subcellular fractionation to determine the protein's cellular localization. B. bronchiseptica cells were separated into three subcellular compartments, a soluble fraction, an inner and an outer membrane fraction. To ensure the efficiency of separation, the cytoplasmic protein BvgA and the outer membrane protein BcfA (56) were used as controls. The purified BcfA protein runs at a slightly higher molecular weight due to the presence of the T7 tag, whereas the BvgA protein runs at the expected weight as it is untagged and purified from inclusion bodies. Furthermore, because a BpsB-specific antibody was not available, a recombinant FLAG-tagged variant of BpsB was expressed. As shown in Fig. 1B, BpsB was localized both to the soluble and the outer membrane fractions. These results suggest that BpsB is likely a periplasmic protein that can localize to the outer membrane.

Structure of BpsB(35-307) MMS Reveals a CE4 Domain
Similar to E. coli PgaB-To gain structural insight into the function of BpsB, crystallization trials were conducted. BpsB  produced crystals that did not diffract, so a domain isolation approach was utilized. The N-terminal deacetylase domain constructs BpsB(27-311) and BpsB(35-311) produced soluble protein but were recalcitrant to crystallization. A slightly truncated variant, BpsB(35-307), and its surface entropy mutant, K128A K129A, were also recalcitrant to crystallization. As a consequence metal-mediated synthetic symmetrization was utilized (57). BpsB(35-307) MMS was incubated with 2 M eq of NiCl 2 , CuCl 2 , or CdCl 2 directly before crystallization trials. BpsB(35-307) MMS produced crystals suitable for data collection in the presence of NiCl 2 . Diffraction data were collected to 1.95 Å, and the structure was solved using the molecular replacement technique. BpsB(35-307) MMS crystallized in the monoclinic space group P2 1 with two molecules in the asymmetric unit and refinement produced a final model with good geometry and R factors R work and R free of 16.9 and 20.8%, respectively (Table 2). Residues 298 -307 and GSHM left over from the hexahistidine tag cleavage were not included in the final model as there was no interpretable electron density present.
BpsB(35-307) MMS adopts the (␤/␣) 7 barrel fold common to the CE4 family ( Fig. 2A) (58). The BpsB(35-307) MMS (␤/␣) 7 core is composed of seven parallel ␤-strands arranged in a barrel that is surrounded by only six ␣-helices. BpsB(35-307) MMS also contains three ␤-hairpin motifs (␤H1-3, green) located on the top of the (␤/␣) 7 barrel (C termini of the core ␤-strands), and an ␣-helix (␣2, purple) that plugs the bottom on the barrel ( Fig. 2A). Both of these structural elements are commonly found within the CE4 family and are both conserved within the PNAG deacetylases PgaB and IcaB. Electrostatic surface potential analysis of BpsB(35-307) MMS reveals an electronegative face that contains the active site pocket (Fig. 2B). Interestingly, loop L1 (residues 53-70) packs against the electronegative face partially occupying the active site. The density of L1 is poor in both molecules suggesting it is flexible, and its conformation may therefore be an artifact of crystallization. Indeed, examination of interaction between molecules A and B in the asymmetric unit reveals a 2-fold rotational symmetry axis is present between residue Ile-68 of both molecules. Residue Asp-64 from molecule A hydrogen bonds with Arg-44 and Ser-291 from molecule B, and Phe-66 and Ala-67 from molecule A pack into a hydrophobic pocket in molecule B. Equivalent interactions are observed between these residues on molecule B and molecule A.
Structural alignment of BpsB(35-307) MMS with the N-terminal domain of E. coli PgaB, residues 43-309 (Protein Data Bank code 4F9D), shows strong conservation of the CE4 domain with a root mean square deviation of 1.3 Å over 225 eq C␣ atoms (Fig. 3A). BpsB(35-307) MMS contains the same circular permutation of the conserved CE4 motifs present in PgaB and IcaB (30,35). However, two topological differences occur between BpsB(35-307) MMS and PgaB . First, the final ␣-helix of the (␤/␣) 7 barrel fold, ␣8 in PgaB, is missing in BpsB(35-307) MMS (Fig. 3A). Second, loop L1 is longer in BpsB(35-307) MMS and lies across the top of the (␤/␣) 7 barrel partially occluding the active site (Fig. 2B, and 3, A and B). Removing L1 from surface representation reveals a narrow groove prior to the active site pocket (Fig. 3B). This secondary groove is not present in PgaB (Fig. 3B) and could play a role in binding PNAG if loop L1 is, as we predict, flexible in BpsB. Structural alignment of the BpsB and PgaB active sites shows a strong conservation of the CE4 motifs, with the only major differences being the location of His-49 and His-55 (Fig. 3C). Our recent study of IcaB (35) suggested that His-49 is a multifunctional acid/base catalyst in the deacetylation mechanism. Interestingly, the electron density was poor for this residue in both BpsB molecules suggesting it may be inherently flexible.
BpsB Is a Metal-dependent PNAG Deacetylase-Diffraction data collected at the nickel-absorption edge (1.485 Å) revealed a total of five nickel ions in the anomalous difference map of the asymmetric unit. Three nickel ion locations were unique. The first unique nickel, present in both molecule A and B, was coordinated by His-235 and His-239, and these were the residues engineered to produce a metal coordination site. The second unique nickel was coordinated by Asp-144 and His-240 in molecule A and formed a crystal contact with Asp-144 and His-240 from molecule B in a symmetry-related asymmetric unit. The third unique nickel, found in molecule A and B, was coordinated by Asp-114, His-184, and His-189 (Fig. 4A). The arrangement of the Asp-His-His residues at this third unique nickel ion was characteristic of the metal coordination found in the CE4 family active site. To determine whether BpsB showed metaldependent deacetylase activity on PNAG, fluorescamine assays with PNAG oligomers were conducted. BpsB  showed low levels of activity as the isolated enzyme, but ϳ3.8and 7.5fold increased activity in the presence of Co 2ϩ and Ni 2ϩ , respectively (Fig. 4B). Unlike PgaB, BpsB(27-701) activity could be abolished when incubated with 1 mM EDTA (Fig. 4B).
BpsB Displays Length-dependent Deacetylation-The structure of BpsB(35-307) MMS revealed the absence of the final helix in the (␤/␣) 7 fold, and the length of the linker between the Nand C-terminal domains of BpsB is roughly double that of PgaB. We hypothesized that the association of the two domains may be significantly weaker, and that if the C-terminal domain plays a significant role in binding PNAG (such as providing additional binding subsites) there should be a difference in lengthdependent deacetylase activity. Thus, fluorescamine assays were conducted using short PNAG oligomers. Our data show a clear trend of increasing rates of deacetylation with increasing PNAG oligomer length (Fig. 5A). This suggests the C-terminal domain provides additional binding sites for PNAG, which is consistent with studies on PgaB (30). However, in contrast to PgaB, the N-terminal domain of BpsB (BpsB(27-311)) displayed ϳ46% activity compared with the full-length protein (Fig. 5B). The isolated N-terminal domain of PgaB does not display deacetylase activity (32). The ability of BpsB  to effectively bind PNAG and deacetylate the polymer may be due to the additional groove that precedes the nickel-active site (Fig.  3B). Furthermore, the likelihood that loop L1 is flexible could be exploited for increased substrate binding by forming an additional structural element to the active site, a component also lacking in PgaB.
Conserved Residues in BpsB Are Required for PNAG Deacetylation-Previous studies on PgaB (30,31), IcaB (35,47), Vibrio cholerae chitin deacetylase (VcCDA) (59), and PgdA (60) suggest BpsB likely uses a metal-assisted general acid/base mechanism common to CE4 members. Based on a number of substrate-bound structures of VcCDA, the metal ion is required for substrate binding and coordinating the 3Ј-hydroxyl of the GlcNAc moiety and the carbonyl of the N-acetyl group, coordinating a water molecule involved in nucleophilic attack on the N-acetyl group, and assisting in the stabilization of the tetrahedral transition state intermediate (59). In addition to a metal co-factor, the CE4 family contains five signature motifs, abbreviated MT1 to MT5 as they occur sequentially in the sequence, that are responsible for substrate binding, metal coordination, and enzymatic hydrolysis. Asp-113 and Asp-114 belong to MT1 and are involved in catalysis and metal coordination, respectively. His-184 and His-189 belong to MT2 and are involved in metal coordination. Tyr-252 and Arg-290 belong to MT3, although Arg-290 is found at the end of the CE4 domain sequence where MT5 is located, and are responsible for creating one side of the active site cavity and coordinating the catalytic base in MT1, respectively. Leu-274 belongs to MT4 and forms one side of the N-acetyl hydrophobic binding pocket. His-49 and Leu-292 belong to MT5 and are involved in catalysis and form the second side of the hydrophobic binding pocket with MT4, respectively. To probe the role of these conserved active site residues (Fig. 3C), mutagenesis studies were con-  ducted on BpsB . The D114A, H184A, and H189A variants all abolished activity showing their importance in metal coordination (Fig. 5B). The H49A and D113A variants also showed no activity suggesting their importance in the catalytic mechanism (Fig. 5B). Interestingly, the D113N mutant showed only 1% of wild-type activity (Fig. 5B). This is in contrast to what has been observed for IcaB, where the comparable variant retained 50% of wild-type activity (35). This suggests that for BpsB Asp-113 may play a significant role in catalysis along with His-49, rather than His-49 acting as a single acid/base catalyst as proposed for IcaB (35).
BpsB Is Not Required for Bps Synthesis or Export-Previous studies in E. coli have shown that PgaB is not required for the production of the PNAG polymer but is required for its modification and export (33). To determine the role of BpsB in Bps biosynthesis and export, a nonpolar in-frame deletion of bpsB in RB50 was constructed (⌬bpsB). An immunoblot assay with antisera raised against deacetylated PNAG from S. aureus (8) was used to compare the production of Bps between RB50 (WT), ⌬bpsB, and ⌬bps strains. Compared with the WT strain, the ⌬bpsB strain showed no significant differences in extracellular Bps production. As expected, the ⌬bps strain showed a severe loss in Bps production (Fig. 6A). The residual immunoreactive material that is observed in the ⌬bps strain has been observed previously (8,15) and probably represents some other material that is weakly cross-reactive with the dPNAG antibody.
As an additional means to confirm that Bps is expressed and exported to the surface of the ⌬bpsB strain, we utilized an enzyme-linked immunosorbent-based assay with a serum that specifically detects Bps (50). At serum dilutions of 10 Ϫ3 and 10 Ϫ5 , there were no statistically significant differences between  the absorbance readings when the WT and ⌬bpsB strains were used to coat the plates (Fig. 6B). Although at the serum dilution of 10 Ϫ4 , the absorbance readings of the ⌬bpsB strain were slightly lower than that observed for the WT strain, it was still considerably higher than that observed when either the ⌬bps strain or PBS was used to coat the plates.
Finally, to ensure the presence of extracellular Bps in the ⌬bpsB strain was not from periplasmic leakage during isolation, TEM on bacterial cells was conducted. Electron microscopic analysis showed abundant colloidal gold labeling of both the WT and ⌬bpsB strains. As expected, no labeling of the ⌬bps strain was observed (Fig. 7). Taken together, the results from immunoblotting, ELISA, and TEM suggest that BpsB is not required for Bps production or export.
BpsB Is Required for Biofilm Formation and Contributes to the Biofilm Structural Complexity-Next, we aimed to determine whether BpsB had a role in biofilm formation in B. bronchiseptica. To address this, a static biofilm assay using glass coverslips was developed. First, to validate the assay, biofilms formed by the WT and the Bvg Ϫ phase-locked strain RB54 were compared. It is well established that the BvgAS locus is required for biofilm formation in B. bronchiseptica (9,11), and RB54 displays a severe defect in biofilm formation compared with the WT strain. As expected, the Bvg Ϫ phase-locked strain was recovered in much lower numbers than the WT strain at all the time points tested. There were no significant differences in biofilm formation between the WT and ⌬bpsB strains at 24 h. However, at 72 and 96 h, the ⌬bpsB strain was recovered at significantly lower numbers compared with WT (Fig. 8A). Complementation of bpsB on a plasmid (⌬bpsB comp ) restored the biofilm defect in the ⌬bpsB strain, as the biofilms either exceeded (72 h) or were similar (96 h) to that formed by the WT strain. In contrast, the ⌬bpsB strain complemented using the empty vector plasmid only (⌬bpsB vec ) formed biofilms similar to that of the ⌬bpsB strain. As expected, the ⌬bps strain where the entire bpsABCD operon has been deleted was recovered from the glass coverslips in lower numbers compared with the WT strain. Interestingly, there were no significant differences in CFU numbers between the ⌬bpsB and ⌬bps strains recovered from the coverslips at any of the time points tested. This suggests that BpsB is required for producing the biofilm-relevant form of Bps, and its absence from B. bronchiseptica severely impairs biofilm formation to similar levels as the ⌬bps strain, which lacks the capacity to produce the Bps polysaccharide.
To further examine the mechanism by which BpsB contributes to biofilm formation, we utilized SEM and flow cells assays followed by confocal scanning light microscopy. SEM analysis of the biofilm coated-coverslips showed that although the ⌬bpsB strain was still able to adhere to the surface, it lost the characteristic three-dimensional biofilm structure displayed by the WT strain (Fig. 8B). Quantitative assessment of the biofilms of the WT and ⌬bpsB strains formed in the flow cell chamber under continuous flow was achieved by analyzing the confocal scanning light microscopy-generated images for biomass, average thickness, and maximum thickness. The COMSTAT computer program was utilized for this purpose. No significant differences were observed after 24 and 48 h between the WT and the ⌬bpsB strain with respect to any of the parameters tested (Table 3). At 72 h however, the biofilms formed by the ⌬bpsB strain exhibited a significantly lower maximum thickness (Table 3). At 96 h the ⌬bpsB mutant formed biofilms that displayed significantly lower values for all the three parameters, biomass, average thickness, and maximum thickness compared with the WT strain (Table 3). Taken together, these results indicate that BpsB-dependent modification of Bps promotes biofilm formation in B. bronchiseptica by contributing to the extracellular matrix structure.

Discussion
Previously, we have shown that Bps is exported to the surface of B. bronchiseptica cells and is critical for biofilm formation (8). In this report, we show that the BpsB protein from B. bronchiseptica is a PNAG deacetylase, and we characterize its role in the deacetylation and export of the Bps polysaccharide and biofilm formation. A, exopolysaccharides were extracted from overnight-grown cultures of B. bronchiseptica strains using EDTA and proteinase K treatment. Extracted samples were spotted onto nitrocellulose membranes and probed with goat antibody raised against S. aureus deacetylated PNAG conjugated to diphtheria toxoid (8). B, B. bronchiseptica strains or PBS were added to an ELISA plate to measure the amount of Bps present on the surface of bacterial cells. The absorbed human serum at indicated dilutions was used to probe for Bps, followed by detection using goat anti-human IgG conjugated to horseradish peroxidase. Error bars are representative of the standard deviation. *** indicates a p value of Ͻ 0.0001 between ⌬bps and other strains, and # denotes a p value of Ͻ0.05 between WT and ⌬bpsB strains. Samples were run in duplicate and are representative of one of two independent experiments. FIGURE 7. Localization of the Bps exopolysaccharide. Electron microscopy analysis for the presence of Bps on the surface of B. bronchiseptica was carried out by adsorbing 1 ϫ 10 6 bacteria onto carbon-coated gold grids in a humidified chamber for 1 h, and then exposed to 10% heat-inactivated Bps-enriched human serum for 1 h after blocking with 2% BSA in PBS. The bound antibodies were detected with 6 nm gold-labeled anti-human polyclonal antibodies at a dilution of 1:10 in the blocking buffer using negative staining with 2% phosphotungstic acid, pH 6.6, and analyzed with a Tecnai transmission electron microscope. Arrows indicate the presence of gold-labeled Bps, which is present on the surface of the WT and ⌬bpsB strains.
The structure and functional characterization of BpsB reveal that the deacetylase domain belongs to the CE4 family with strong structural conservation to PgaB (Figs. 2 and 3). BpsB is an inefficient metal-dependent deacetylase of PNAG oligomers (Fig. 4) similar to PgaB and IcaB. BpsB can use a variety of divalent cations for deacetylation with highest activity in the presence of Ni 2ϩ and Co 2ϩ ions. As the PNAG deacetylases are likely expressed during nutrient-limiting conditions (i.e. when a biofilm is needed), it is not surprising that they can utilize a variety of metal cofactors for activity. What remains unknown for the PNAG deacetylase and other periplasmic CE4s is how the enzymes are loaded with their respective metal cofactor. It is possible that the enzymes can scavenge the ions directly from the outer membrane metal ion porins (such as BtuB in E. coli). However, metal ions in the periplasm are usually specifically transported by solute-binding proteins (such as NikA for Ni 2ϩ in E. coli) and delivered to the inner membrane ATP-binding cassette transporters for cytosolic delivery. Thus, it is more likely that the enzymes are directly loaded by the solute-binding proteins or via periplasmic metallochaperones.
Our structural and functional analysis of BpsB shows high conservation to PgaB; however, in contrast to PgaB, the isolated BpsB deacetylase domain, BpsB , is able deacetylate PNAG oligomers (Fig. 5B). The secondary groove preceding the active site in the BpsB(35-307) MMS structure in combination with a significantly altered L1 loop appear to be the key elements that allow for PNAG deacetylation in the absence of the C-terminal domain. Although not required for deacetylation, the C-terminal domain of BpsB is required for increased rates of deacetylation with oligomer length and likely increases the binding affinity for PNAG. These structural and functional differences suggest that the mechanism of periplasmic processing and outer membrane export may be different from that proposed for PgaB (32). Dot-blot assays, ELISA, and TEM analyses showing Bps secretion into the extracellular matrix in a ⌬bpsB mutant (Figs. 6 and 7) further support this hypothesis as a ⌬pgaB E. coli mutant retains PNAG in the periplasm (33). These observations may also be the result of structural/functional differences between the outer membrane proteins BpsA and PgaA. PgaA and BpsA are predicted to contain a periplasmic domain that contains multiple TPR motifs involved in protein-protein interactions and a ␤-barrel porin. Interestingly, PgaA is 146 residues longer than BpsA, and structural prediction servers such as Phyre 2 (52) suggest that the additional 146 amino acids are predominantly located to the C terminus of the periplasmic TPR-motif containing domain. One possible explanation is that the smaller TPR domain in BpsA could affect the ability to form a protein-protein interaction with BpsB. However, our localization data do not support this as BpsB co-fractionates with the outer membrane suggesting it likely forms a  complex with BpsA. Furthermore, a protein-protein interaction has been observed using cross-linking with the orthologous proteins, HmsF and HmsH, from Yersinia pestis (61). An alternative explanation may be that the additional residues in PgaA to the C terminus of the periplasmic TPR domain may occlude the ␤-barrel porin preventing dPNAG export without some conformational change. Thus, in BpsA the porin may be constitutively open and allow for PNAG transport in the absence of a protein-protein interaction driven by Bps deacetylation. Because of difficulties in isolating pure Bps from both WT and ⌬bpsB knock-out strains for compositional analysis, it is still a possibility that Bps could be deacetylated nonspecifically by another unknown protein in the absence of BpsB. This in turn could also be sufficient to modulate the conformation of BpsA to allow for Bps export. BpsB  deacetylation is inefficient, yet the conserved residues in the BpsB(35-307) MMS active site are all required for activity (Fig. 5B). The PNAG deacetylases PgaB, BpsB, and IcaB all have low rates of deacetylase activity in vitro. Maintaining low levels of deacetylation of PNAG in vivo is likely a mechanistic feature to modify the chemical properties of the polysaccharide needed for biofilm formation. For example, partial deacetylation of PNAG from S. epidermidis has been shown to be crucial for adhesin properties, as deletion of the PNAG deacetylase IcaB results in shedding of the polymer from the cell surface and abolishment of biofilm formation (62). Thus, the ability to fine-tune the level of deacetylation/acetylation of exopolysaccharides appears to be an important aspect for biofilm formation.
This report is the first in-depth analysis of a Gram-negative PNAG deacetylase other than PgaB. Our results suggest that there are structural and mechanistic variations between the orthologous PNAG biosynthetic systems of E. coli and B. bronchiseptica. Although BpsB is not required for the biosynthesis and export of Bps (Figs. 6 and 7), it plays an essential role in biofilm formation ( Fig. 8 and Table 3). Our data suggest that Bps or deacetylated Bps is not primarily used as a surface adhesin, as B. bronchiseptica ⌬bps or ⌬bpsB strains can still effectively adhere to surfaces (Fig. 8). However, deacetylated Bps is required for the intricate three-dimensional architecture seen in later stages of biofilm development ( Fig. 8B and Table 3). It may be that evolutionary differences among orthologous PNAG biosynthetic systems have occurred to cause important differences to the dPNAG polymer in vivo, such as polymer length and level of deacetylation. These variations in the structure and chemical properties of PNAG may have evolved so that bacteria can more efficiently colonize distinct microenvironments within the host. We propose that by contributing to biofilm formation in the respiratory tract, BpsB promotes long term survival within the mouse respiratory tract. The contribution of BpsB in promoting the pathogenic properties of B. bronchiseptica is currently under investigation.
The structural and functional characterization of BpsB presented herein provides further evidence that Bps is the same compositional homopolymer as PNAG produced by numerous Gram-negative and -positive bacteria. Understanding how the modification of Bps affects surface adhesion and biofilm formation will help guide the development of anti-biofilm agents against Bordetella. Identifying key links in biofilm formation and Bordetella infections will be pertinent for the identification of vaccine targets or drugs to combat respiratory related illness.