Structural Variations within the Transferrin Binding Site on Transferrin-binding Protein B, TbpB*

Pathogenic bacteria acquire the essential element iron through specialized uptake pathways that are necessary in the iron-limiting environments of the host. Members of the Gram-negative Neisseriaceae and Pasteurellaceae families have adapted to acquire iron from the host iron binding glycoprotein, transferrin (Tf), through a receptor complex comprised of transferring-binding protein (Tbp) A and B. Because of the critical role they play in the host, these surface-exposed proteins are invariably present in clinical isolates and thus are considered prime vaccine targets. The specific interactions between TbpB and Tf are essential and ultimately might be exploited to create a broad-spectrum vaccine. In this study, we report the structure of TbpBs from two porcine pathogens, Actinobacillus pleuropneumoniae and suis. Paradoxically, despite a common Tf target, these swine related TbpBs show substantial sequence variation in their Tf-binding site. The TbpB structures, supported by docking simulations, surface plasmon resonance and hydrogen/deuterium exchange experiments with wild-type and mutant TbpBs, explain why there are structurally conserved elements within TbpB homologs despite major sequence variation that are required for binding Tf.

While colonizing or infecting the host, bacterial pathogens utilize surface proteins to acquire nutrients and interact with host components to survive the antibacterial properties of the host milieu. For bacteria, the acquisition of the essential nutrient iron is of critical importance as vertebrates sequester free iron within the extracellular milieu using specific iron carrying glycoproteins, lactoferrin (Lf), and transferrin (Tf) 3 (1).
To colonize, bacteria have selectively evolved with their host to express specific mechanisms for iron uptake. The pathogenic Gram-negative bacteria from the Neisseriaceae and Pasteurel-laceae families have developed a host Tf-specific iron uptake system mediated by a bacterial receptor complex, the Tf-binding proteins A and B (TbpA, TbpB) (2). TbpA is a TonB-dependent outer-membrane protein, strictly essential for iron uptake during bacteria infection (3). TbpA binds both of the holo or apo forms of Tf (4), and it is proposed to serve as a channel for iron transport across the outer membrane (5). TbpB, the second component of the receptor complex, is a surface exposed lipoprotein that binds holo-Tf in a selective manner (4,6). This enables the bacterial receptor to recruit iron more efficiently from the partially iron-saturated forms of Tf in vivo (7). TbpB is essential for colonization of the host (8), and depending on the strain and species, TbpB-defective bacteria present several different growth-impaired phenotypes. TbpBs display species specificity (9), binding only to their host Tf, while the human Tf-receptor has no species specificity retrieving Tf through endocytosis where the low pH of the endosome drives the iron delivery from Tf (10,11).
Surface lipoproteins such as TbpB have been targeted for vaccine development as they elicit a strong immune response, and antibodies to specific surface lipoproteins are bactericidal (12)(13)(14). Because TbpB functions to specifically associate with the host protein Tf, it is likely that the site of interaction may harbor a conserved binding interface (15) that could be exploited to create a broad spectrum vaccine.
Recently the structure of the TbpB from the porcine pathogen Actinobacillus pleuropneumoniae together with functional binding studies has illustrated that the residues from the site of interaction with porcine Tf (pTf) do not appear to be conserved among porcine pathogens (16). The structure illustrates that TbpB is a bilobed protein requiring the N-lobe domain for Tf binding function. Both the N-and C-lobe adopt a similar overall fold; the C-lobe is the most conserved domain in sequence along all TbpB proteins, while N-lobes show paradoxically a less well-conserved sequence even between strains targeting the same host and thus the same Tf (16).
A. pleuropneumoniae along with Actinobacillus suis and Haemophilus parasuis are important pathogens in pigs that share a common mechanism for acquiring iron directly from the host Tf (17)(18)(19). This parallels the situation with human pathogens in which Neisseria meningitidis, Moraxella catarrhalis, and H. influenzae acquire iron from human Tf through a conserved, highly specific interaction common to these species (15 pathogens provides the greatest opportunity to probe the details of this interaction and determine how a conserved interaction can be maintained in the face of considerable sequence diversity. In this study, we describe crystal structures of two divergent TbpB receptors from the porcine pathogens A. suis (H57) and A. pleuropneumoniae (H87) and provide models of the pTf-TbpB complexes that are supported by hydrogen-deuterium exchange coupled to mass spectrometry (H/DX-MS) and binding experiments with site-directed mutant TbpBs.

EXPERIMENTAL PROCEDURES
Sequence Analysis of tbpB Genes-The tbpB genes from all of our clinical isolates of A. pleuropneumoniae and A. suis were amplified from whole colonies using a set of primers designed to bind ϳ100 bp outside of the gene. These sequences were combined with all of the unique publicly available tbpB sequences, and a ClustalW2 multiple sequence alignment was generated. The alignment was manually inspected and edited, and a bootstrapped neighbor-joining unrooted phylogenetic tree was generated from this alignment using the program Geneious.
Expression and Purification of Recombinant TbpB-Protein was expressed and purified essentially as described previously (16). The genes encoding TbpB from A. suis strain H57 (AsH57) and A. pleuropneumoniae strain H87 (ApH87) were cloned in to an expression vector encoding an N-terminal maltose-binding protein fusion partner containing a polyhistidine tag and a TEV cleavage site preceding the region encoding the mature TbpB protein. The recombinant fusion proteins were isolated on a pTf-Sepharose column as described previously (16) or by Ni-NTA chromatography. After cleavage by TEV protease the TbpB proteins were purified by Q-Sepharose chromatography as described previously (16).
Protein Crystallization-Purified AsH57 and ApH87 TbpBs were screened against the JSCGϩ suite commercial screen using hanging drop vapor diffusion at 10 mg/ml. Initial AsH57 TbpB crystals were observed in 0.2 M tri-potassium citrate pH 8.3 and 20% PEG 3350 then optimized with a 1:1 ratio sitting drop at 20°C in a precipitant solution composed of 0.1 M Bis-Tris pH 5.5, 0.4 M lithium sulfate, 17% PEG 3350 and 20% glycerol yielding crystals in space group P2 1 . ApH87 TbpBs were first observed in 0.16 M calcium acetate, 0.08 M sodium cacodylate pH 6.5, 14.4% PEG 8000, and 20% glycerol and optimized at 20°C using 0.16 M sodium acetate, 0.08 M calcium cacodylate pH 6.5, 23% PEG 3350 and 15% glycerol to yield crystals in space group C222 1 . Unit cell dimensions and statistics are summarized in Table 1.
Structure Determination-Data were gathered on crystals frozen at 105 K on beamline 08ID-1 at the Canadian Light Source (CLS). Diffraction data sets of AsH57 TbpB and ApH87 TbpB were collected at a wavelength of 0.9795 Å using 1°oscillations with 360 and 300 images, respectively. Data were processed with XDS (20) to resolutions of 2.1 Å for ApH87 TbpB and AsH57 TbpB. The first structural model was obtained by molecular replacement based on ApH49 TbpB (PDB code 3HOL) using PHASER (21), yielding an initial R work /R free 0.29/ 0.32 and 0.34/038 for AsH57 and ApH87, respectively. Final models were generated following several rounds of model building and refinement using Coot (22) and Refmac5 (23) programs. A Ramachandran plot calculated 3.9% and 3.0% of residues in the allowed region and 0.4% and 0.2% as outliers, for AsH57 and ApH87 TbpB, respectively. The numbering of residues initiates with the first mature amino acid (Cys-1) remaining after signal peptide cleavage.
Surface Plasmon Resonance-Surface Plasmon resonance experiments were performed using BiacoreX (GE Healthcare) at 25°C. Porcine transferrin was coupled to the sensor chip (CM5 research grade) via standard N-hydroxysuccinimde and N-ethyl-N-(dimethylaminopropyl) carbodiimide activation. Wild-type TbpB and mutant variants were diluted at various concentrations with the mobile phase buffer (10 mM Hepes pH 7.5, 150 mM NaCl, and 0.005% surfactant P20). Samples were injected as analytes at a flow rate of 20 l/min, and bound analytes were subsequently removed by washing with the mobile phase at 90 or 240 s after the injection. Regeneration buffer (10 mM Hepes, pH 7.5, 2 M NaCl) was injected prior to each analyte injection. Kinetic constants were calculated from the sensorgrams using the simulated BiacoreX evaluation software, version 4.0.1 (Biacore). The AsH57, ApH87, and ApH49 TbpBs data were fitted to a "heterogeneous ligand" binding model. The ApH87 TbpB R179E mutant was analyzed using the saturation method and fitted with a quadratic equation with BiacoreX evaluation software.
Docking Simulation-As the structure of TbpB bound to pTf has not yet been elucidated, the interaction was modeled using RosettaDock (24) incorporating the information generated by SPR analysis. RosettaDock was run in full atom mode, allowing spin around the axis connecting the two proteins, and using standard Monte Carlo movements. A custom program was written to allow experimentally determined important residues to be taken into account during a docking run. The general algorithm of the custom program was to first run RosettaDock, generating 3,000 high-resolution decoys. Each decoy was then parsed, checking whether each TbpB residue designated as important was within 10 Å of pTf, and deleting those decoys that failed on any residue. RosettaDock was then run again to generate as many new decoys as necessary to bring the total back up to 3,000, and the cycle repeated until all decoys met each of the designated distance constraints.
This program was used to generate models for both of the new structures ApH87 and AsH57 TbpB, as well as the previously solved structure from ApH49 (16). For ApH87, the residues deemed important were Tyr-95, Tyr-121, Tyr-174, and Arg-179, for AsH57 the important residues were Phe-63 and Phe-152, and for ApH49 they were Glu-112 and Phe-171. Initial perturbations were set to randomize both partners, move 3°and 8 Å, with a spin along the protein-protein axis. Monte Carlo minimization was used, and the decoys were generated in fullatom mode using multiple threads in parallel. For both the N-lobe and the full-length ApH87 TbpB with pTF 1500 decoys were generated, running on 6 threads at once. The 10 decoys were chosen by their total score, which is determined by the scoring function.
Hydrogen/Deuterium Exchange (H/DX-MS) Experiments-Each H/DX-MS experiment consisted of two samples: free TbpB (AsH57, ApH49, or ApH87) and the TbpB-pTf complex (AsH57, ApH49, or ApH87). Each experiment was completed in quadruplicate, alternating between the receptors and complexes, therefore making future receptor cross-comparison possible. Ten microliters of TbpB was mixed with 5 l HEPES buffer (10 mM HEPES, pH 7.0). Five microliters of heavy water (deuterated water, D 2 0) was then added to this solution, mixed, and incubated immediately in a water bath at 20°C, for 2 min. The labeled sample was then removed from the water bath, and 10 l of cold quench solution (100 mM glycine hydrochloride, pH 2.3) was added. This sample was immediately injected onto a 10-l loop, connected to an online immobilized pepsin reactor (2.5 cm by 200 m internal diameter, assembled in-house) (Applied Biosystems, Concord, ON, Canada). An isocratic gradient of 3% ACN, 0.03% TFA, and 0.02% FA, at 4 l/min was used to flow the intact labeled protein through the digestor column, and onto a C18 reversed-phase chromatography column (assembled in-house) (Applied Biosystems, Concord, ON, Canada). The peptides were separated with a step gradient of 5 to 90% ACN in 0.03% TFA and 0.02% FA over 14 min, and directly infused into a QSTAR Pulsar i quadrupole time-offlight (QqTOF) mass spectrometer. All chromatography was carried out under ice, at 0 -4°C. Data were collected using Analyst QS TM version 1.1 (Applied Biosystems).
To determine the TbpB peptides arising from the online pepsin digests, unlabeled peptides were prepared under the same conditions as mentioned above. Separated peptides were subjected to tandem mass spectrometry (MS/MS). The MS/MS spectral data were searched against the TbpB sequence (either ApH49, AsH57, or ApH87), using an in-house licensed version of MASCOT 2.1, with conventional identification criteria (Ϯ0.5amu mass tolerance in MS and 0.8amu mass tolerance in MS/MS). The MASCOT search identified 58 usable peptides for AsH57 and peptides were then manually verified. Mass spectral data from each replicate H/DX-MS run was analyzed using Hydra software (25), using the method described by Ling et al. (26).

RESULTS
Phylogenetic Analysis-Through sequencing tbpB genes from our strain collection and combining them with sequences available from public databases, we compared TbpBs from A. pleuropneumoniae and A. suis to gauge the degree of sequence diversity (supplemental Fig. S1). Three TbpBs divergent from the one previously structurally characterized (16) were produced for structural studies and the structures of two of these TbpBs were obtained. Together with the published structure, these provide a reasonable representation of the sequence diversity (red arrows, supplemental Fig. S1). A sequence alignment between the three TbpBs for which structural information is available is shown in Fig. 1. The results illustrate that there is significant sequence variation in the handle (blue color) and barrel (green color) domains of the N-lobe of TbpB ( Fig. 1, upper panel) that primarily resides in loops on cap area of the N-lobe (lower panel).
The N-terminal Anchoring Peptides Have Different Conformations-The conformation of the N-terminal anchor peptide (the residues preceding the leucine 50 residue of the first ␤-strand) of the ApH57 TbpB differs from the two other TbpBs. The ApH87 TbpB N terminus rotates 118 degrees at Lys-43 and extends into a long ␣-helix stacked on the N-lobe (Fig. 2). In the crystal, the N-terminal peptide of ApH49 and AsH87 TbpB (respectively residues 1-41, and 1-43) is oriented toward the C-lobe, interacting the C terminus by intra-chain contact. Crystallized in a different space group C222 1 , the ApH57 TbpB N terminus (residues 1-43) extends along the N-lobe and also interacts the C-lobe terminal residues but from the adjacent monomer in the unit cell. These two alternative conformations support the argument that TbpB uses its N-terminal region as a flexible anchoring peptide to extend above the capsular layer to bind Tf from the extracellular millieu and recruit it toward the outer membrane surface.
N-lobe of TbpB Varies in Sequence and Structure between Porcine Pathogens-The "cap" area is composed of the loops protruding from the N-lobe ␤-barrel and handle domain (loop 1, 3, 5, 8, 10, 12 in addition to the long insertion within the loop 4 that points toward the Cap area in ApH57 TbpB) ( Fig. 1 and  supplemental Fig. S2). This cap region represents the most variable region between the three known TbpB structures, and between all TbpBs in general. Structural alignments of N-lobe domain of AsH57 and ApH87 TbpBs yield an RMSD of 1.7 Å between these two domains, and 3.4 Å and 0.78 Å when respectively compared with ApH49 TbpB.
Focusing on the putative Tf-binding site previously proposed on the cap area (16), a comparison of the cap surfaces reveals many differences between AsH57 TbpB and the two other TbpB receptors. Long insertions in the outward facing loop region within the N-lobe ␤-barrel and handle domain of AsH57 TbpB results in a larger surface exposed area that is 69 Å in length compared with that of ApH49 or ApH87 TbpB, whose cap areas extend 38 and 44 Å, respectively (Fig. 3A).
ApH49 TbpB-pTf Sites of Interaction-Based on our previous model of the Tf-ApH49 TbpB recognition, we mapped the key binding elements involved in the pTf-binding site onto the two new TbpB structures. The first common characteristic shared between the three putative binding pockets is the large positive surface spread on the cap region, which was previously pro-posed to interact with the Tf C-lobe. This positively charged interface involved loops from both the ␤-barrel and the handle domain but, in the case of AsH57 TbpB, the site is exclusively localized on its large loop and ␤-strand (␤4 and ␤5) insertions extended from the handle domain (Fig. 3B).
Characterization of ApH49 TbpB demonstrated Phe-171 as a crucial residue for Tf-binding function. Based on structure overlapping, we detected an equivalent aromatic residue, Tyr-174 in ApH87 at the similar position to ApH49 TbpB Phe171 (Fig. 3B), which is not apparent from sequence alignments (Fig.  1, upper panel). The cap region of AsH57 TbpB is more divergent than the two other crystallized TbpBs and there were no obvious residues in the binding interface that corresponded to Phe-171 or other critical residues from ApH49 TbpB. These variations are essentially due to major structural differences in loop organization of the cap region (Fig. 3B).
Although the cap interface of AsH57 TbpB does not appear to be conserved with ApH49 and ApH87, it still has a conserved feature of two surface exposed hydrophobic residues (Phe-63 and Phe-152) located within the positively charged cap area (Fig. 3B). However, these phenylalanines are localized on the handle domain loop instead of ␤-barrel as in the ApH87 and ApH49 TbpB structures.
Surface Plasmon Resonance Reveals the Binding Affinities of TbpBs with pTf-Surface plasmon resonance was used to quantify binding affinity between different TbpB preparations and covalently immobilized pTf on a BiacoreX (Table 2 and the supplemental Fig. S3). ApH49 and ApH87 TbpB share closely related cap areas, and an equally similar pTf affinity of 44 and 60 nM, respectively. The AsH57 TbpB has a lower binding affinity of 120 nM. Our SPR data are in agreement with a previous characterization of ApH49 TbpB by isothermal titration calorimetry that revealed a dissociation constant of 55 nM (16).
SPR data from the TbpB receptors did not match to a 1:1 Langmuir binding model. Difficulty in fitting our kinetic data to a 1:1 model could be a consequence of ligand/analyte heterogeneity, or two-step binding kinetics. To discriminate between these models we monitored the effect of several saturation times on the dissociation events of TbpBs (supplemental Fig. S3 and Table S1): in case of heterogeneous analyte or two-step binding kinetics, the dissociation will slow as the duration of the injection is increased. Comparing the dissociation phase that occurs after variation in the saturation time, we show that the dissociation curves are not affected by the contact time (supplemental Fig. S3) and that the "heterogeneous ligand" binding model best fits the data.
Mutational Analysis of AsH57 and ApH87 TbpB-The pTf-binding site determined for the previously characterized ApH49 TbpB has common features with the two homologous TbpB structures presented herein. They all share the feature of a positively charged region in the N-lobe cap sur-   face and all contain surface exposed aromatic residues within this region. In ApH49 TbpB this surface exposed aromatic was shown to be essential for binding pTf. Thus all exposed aromatic amino acids from the putative binding sites of AsH57 and ApH87 TbpBs were mutated to alanine and the resulting mutant proteins tested for binding by SPR. TbpB. Tyr-174 from ApH87 has the strongest impact on the pTf interaction and is localized near to the essential Phe-171 from ApH49 TbpB (Fig. 4B) (16). In the case of AsH57 TbpB, direct binding experiments identify two phenylalanines involved in pTf recognition; Phe-63 and Phe-152 are located in the large protruding loop L1 and the extended ␤5 strand from the N-lobe handle domain with their side-chains solvent exposed (Fig. 4C). The loops protruding from the N-lobe ␤-barrel were also proposed to mediate electrostatic interactions between pTf and TbpB (16). To probe the role of the electropositive surface of the TbpB N-lobe Cap region on pTf affinity, we mutated the conserved, solvent-exposed Arg-179 to Glu within AsH87 TbpB, (equivalent to Arg-175 in ApH49 TbpB). The resultant mutant protein was purified and examined using direct binding assays by SPR. The pTf-binding experiments indicate a drastic lost in binding efficiency of the R179E mutant compared with wild type (60 nM to 6.1 M binding affinities) ( Table 2). The substitution of a single positively charged residue in the cap area to glutamic acid is sufficient to abrogate the pTf binding activity of the ApH87 TbpB.
Models of TbpB AsH57 and ApH87 Bound to pTf-Rosetta docking was performed using the C-lobe of pTf (iron loaded) and the structures of the N-lobes of ApH87 or AsH57 TbpBs. In these docking experiments the mutational data were used to add constraints to the docking algorithm. For both TbpBs, the

The Crystal Structures of Porcine Pathogen TbpBs
top 1% of the highest scoring docked models were examined manually, and one heavily favored orientation was found for each (supplemental Fig. S4). These predominant models were homologous within 20 of the top 30 decoys for their respective docking runs. The second most favored orientation had only 3 and 4 homologues each. Though ApH87 and AsH57 TbpB have many major differing structural features, especially around the proposed pTf binding site, the predicted docking orientations of both structures were nearly identical (Fig. 5). Because of this similarity, a new docked model was generated for the previously reported ApH49 TbpB structure, incorporating important residue data as constraints (described under "Experimental Procedures"). The majority of the models (17 of the top 30) resulted in a nearly identical orientation to those of AsH57 and ApH87 models. In our three independently generated models, the TbpBbinding interface on pTf is conserved. The C1 and C2 domains of the transferrin C-lobe bind to the handle and ␤-barrel domain of TbpBs, respectively (Fig. 5). Porcine transferrin docks on TbpB with the ␣-helix H 1 (residues 352-366) and the loop elements L 15 (residues 507-513 and 519 -521), L 18 (residues 555-559), and L 23 (residues 624 -627 and 633-636). A conserved binding pattern emerges from the Rosetta docked models, which suggests the helix and loop elements H 1 , L 15 , L 18 , and L 23 of pTf form the TbpB binding interface on pTf (Fig.  5). The negatively charged ␣-helix H 1 and loop23 (consisting of the motif CSSS, residues 624 -627) of pTf binds the positively charged cap area of the TbpBs. Loop15 of pTf makes several contacts with TbpBs, but one key residue contact that is central to each model is the positively charged Arg-509, which is oriented toward a negative pocket in the canyon formed between the N-lobe ␤-barrel and handle domains. 4 Loop18 (non uniformly charged) of pTf makes contact on the ␤-barrel domain but we note that the electrochemical environment on TbpB is not maintained between the three-studied TbpBs.
Porcine Transferrin Binds TbpB on the Cap Area-To localize and confirm the Tf-binding site location, we performed H/DX-MS. Results from experiments with TbpB from three strains; A. pleuropneumoniae H49 and H87, and A. suis H57 alone, and upon complex formation with pTf, are presented in Fig. 5. For the ApH49, AsH57, and ApH87 TbpBs, a total of 57, 58, and 54 pepsin-digested peptides, respectively, could be utilized, covering 46, 45, and 43% of the protein sequences and 69, 60, 61% of the respective cap region (supplemental Fig. S5). The isotopic mass shift due to deuterium incorporation in TbpB fragments were detected using a QStar pulsar mass spectrometer, and analyzed via Hydra software, as described under "Experimental Procedures." Despite incomplete sequence coverage of the receptors with a pepsin digestion, the protected sequences of the three-studied TbpBs are all located on the N-lobe domain and more precisely localized along the cap area of each of the porcine TbpBs studied (Fig. 4).

DISCUSSION
Structural Features of TbpBs-The two new structures reported in this study demonstrate that the overall structural features of TbpBs are conserved despite considerable sequence variation and size differences. Each lobe is comprised of an 8-stranded ␤-barrel with a 5-stranded handle domain in the N-lobe and a 6-stranded handle domain in the C-lobe (supplemental Fig. S2). The size and sequence differences are largely manifested by variation in the loops that constitute the Cap region of the N-lobe. There is a conserved interaction between the N-terminal lobe that contains the Tf binding interface and the relatively conserved C-lobe, in an orthogonal orientation. Although the C-lobe is not directly involved in Tf binding, its presence and conserved interaction with the N-lobe could be important for the receptor function in vivo. The lack of sequence variation in the C-lobe compared with the N-lobe may infer a conserved interaction with other surface components.
In the ApH49 and AsH57 proteins the extended N-terminal anchoring peptide interacts through hydrogen bonds with the C-lobe. In the ApH87 structure the anchoring peptide assumes a different conformation that appears to be stabilized by interchain contacts with the adjacent TbpB in the crystal lattice. The complete N-terminal anchoring peptides consist of more than 43 residues that could equate to ϳ100 Å in length if in an extended conformation in solution. Thus the two observed conformations (Fig. 2) suggest that the N-terminal membraneanchored peptide may allow the TbpB receptor to adopt an extended orientation to more effectively capture host Tf in the presence of the polysaccharide capsular layer.
TbpB Conservation of the pTf-binding Sites-Based on our pTf-TbpB model, there are substantial variations in the TbpB N-lobe Cap regions, however all strains studied permit high affinity specific binding interactions with Tf. The cap region involved in pTf recognition is directly exposed to the host environment, including the immune defense elements. The sequence and structural variation of this cap region likely is a mechanism for immune evasion, providing decoy epitopes that vary between strains so that a primarily strain-specific immune response is induced (27). Because the different receptor proteins need to retain the ability to bind to transferrin in a similar fashion, there should be conserved features that could otherwise induce a cross-reactive immune response. Thus there would be selective pressure for directing the immune response to variable regions or to accomplish a somewhat conserved Tf interaction by variable means.
Based on the three solved structures of TbpB we can classify the pTf-binding sites into at least two groups: ApH49 and ApH87 TbpBs that share similarities in the loop's cap area, and AsH57 TbpB that has a large extended loop containing some new secondary structural elements. The results from this study suggest that AsH57 TbpB adopts a novel interaction pattern with pTf, as the binding interface appears highly modified compared with ApH87 and ApH49 (Fig. 3A). The cap area of AsH57 TbpB is also identified by H/DX-MS experiments as an interfacial region with pTf. The two protruding loop regions of AsH57 TbpB create a larger surface area buried in the model (ϳ4400 Å 2 ) with pTf (Fig. 4C). The strikingly different loop structure while maintaining interactions with the same regions of pTf has important implications on how this diversity is created and on potential barriers to horizontal exchange of antigenic diversity.
Using a structure-based approach, TbpB mutants were selectively designed to probe related effects on function with the goal to more precisely define the binding activity of these prokaryote pTf receptors. Based on previous studies of ApH49 TbpB (16), we first focused on the solvent exposed aromatic residues accessible from the cap area of AsH57 and ApH87 TbpBs. SPR binding experiments with all the alanine substitutions of phe- nylalanine and tyrosine residues within the putative binding sites, revealed that similar amino acids to the ApH49 TbpB Phe-171 are also required for the pTf binding of AsH57 and ApH87 TbpBs. These binding experiments point to the importance of the Tyr-95 and Tyr-121 and the structurally conserved Tyr-174 on ApH87 TbpB (Fig. 4B), and the non-conserved Phe-63 and Phe-152 on AsH57 TbpB (Figs. 4C). Interestingly, the phenylalanines 63 and 152 are localized on the protruding loop L1 and the extended ␤5-strand from the handle domain instead of the ␤-barrel as in the case of the essential aromatic residues from ApH49 and ApH87 TbpBs (Fig. 4).
Despite the loop variation between the pTf-binding sites, the swine pathogen TbpBs share similar characteristics for capturing porcine Tf. In addition to the essential Phe and Tyr hydrophobic residues on the cap area, the pTf-TbpB recognition seems to depend on a positively charged interface, which is predicted at the pTf interface by computational Rosetta docking and by experimental HDX-MS data. Indeed the putative pTf-binding sites from the three TbpB structures all maintain a positively charged interface on the cap area, which matches to the neighborhood of the essential Phe and Tyr (Fig. 3B). To test the pTf binding dependence toward this positively charged area, we prepared an ApH87 TbpB mutant in which the Arg-179 has been substituted by a glutamic acid. This mutant allowed us to measure the deficiency in pTf-binding caused by a partial charge disruption in the putative binding site. The mutation was sufficient to disrupt the pTf binding function by insertion of a single negatively charged amino acid.
All three TbpB structures present similar characteristics involving aromatic and positively charged residues, but they show variability in localization. The binding contributors from both ApH49 TbpB and ApH87 have partial similarity in arrangement (Fig. 3B), but differ more drastically with the structurally non-conserved AsH57 TbpB binding site (Fig. 3B). The independent Rosetta docked models from ApH49, ApH87, and AsH57 predict the N-lobe Cap area as the pTf-binding pocket of the swine-related TbpBs. The H/DX-MS experimental data validates our three Rosetta docked models, confirming the Cap area as the TbpB/pTf interface. Although H/DX-MS overall TbpB sequence coverage of the three TbpB strains was low, coverage of the cap regions was quite high, with most of the missing coverage being attributed to the rigid ␤-barrels in each of the TbpB lobes. In agreement with our models, most of the protruding loops from the cap area are protected in presence of pTf (Fig. 4).
Despite the divergence in the TbpB interface, the conserved physicochemical properties of the binding sites support idea that these TbpB could target a common binding-site on pTf. Based on our Rosetta docking experiments, TbpB receptors bind to the same surface on pTf (Figs. 4 and 5). Our three independently generated TbpB/pTf binding models suggest the loop elements L15, L18, L23, and the ␣-helix H1 as the TbpBbinding site on pTf.
Despite a high sequence homology between Tfs from different species (70% between pTf and hTf C-lobes), the TbpB binding function for Tf is host specific. Interestingly, amino-acids from pTf that are predicted to contact the porcine pathogen TbpBs are not conserved when compared with hTf (supplemental Fig. S6), sug-gesting that pathogens have adapted to discriminate specific host Tf with higher affinity; or a host evolution as defense mechanism to avoid infection from Tf-based pathogens.
Implication in Vaccine Design-TbpBs have been targeted for development of vaccines for a number of human and veterinary pathogens due to the critical role they mediate and their resulting invariant presence in clinical disease isolates (14,(27)(28)(29). Although they are capable of inducing a protective immune response, there is an insufficient cross-protective response induced by individual intact receptor proteins to be considered as suitable vaccine antigens. The fact that these proteins mediate a conserved interaction with the host Tf provided some optimism of ultimately targeting conserved regions of the proteins for developing a more cross-protective antigen. Furthermore, if common Tf-binding regions could be targeted by the immune response, the bacteria would be unable to evade the immune response by ongoing antigenic variation as this could also compromise the critical function mediated by TbpB.
The results from this study demonstrate that there is no obvious conserved Tf-binding region in the N-terminal lobe common to all the TbpBs from porcine pathogens that could simply be used as vaccine antigen capable of inducing a cross-protective immune response. It seems evident that the divergent TbpBs accomplish binding of porcine Tf with common, but not identical, features and that further study will be needed to fully appreciate the molecular details of this interaction. Clearly the most informative approach would be determining structures of pTf-TbpB complexes but these have been evasive to date, and a combination of alternate approaches may need to be considered. A detailed appreciation of the pTf-TbpB interaction may ultimately provide insights on how to develop one or more receptor derivatives or mimics capable of inducing a cross-protective response against all porcine pathogens. Expanding this approach to developing vaccines against human pathogens will require gathering structural information on a large set of TbpB homologues to develop broad-spectrum protection again multiple pathogenic species.