Enhanced Binding of TonB to a Ligand-loaded Outer Membrane Receptor

The ferric hydroxymate uptake (FhuA) receptor from Escherichia coli facilitates transport of siderophores ferricrocin and ferrichrome and siderophore-antibiotic conjugates such as albomycin and rifamycin CGP 4832. FhuA is also the receptor for phages T5, T1, Φ80, UC-1, for colicin M and for the antimicrobial peptide microcin MccJ21. Energy for transport is provided by the cytoplasmic membrane complex TonB·ExbB·ExbD, which uses the proton motive force of the cytoplasmic membrane to transduce energy to the outer membrane. To accomplish energy transfer, TonB contacts outer membrane receptors. However, the stoichiometry of TonB· receptor complexes and their sites of interaction remain uncertain. In this study, analyses of FhuA interactions with two recombinant TonB proteins by analytical ultracentrifugation revealed that TonB forms a 2:1 complex with FhuA. The presence of the FhuA-specific ligand ferricrocin enhanced the amounts of complex but is not essential for its formation. Surface plasmon resonance experiments demonstrated that FhuA·TonB interactions are multiple and have apparent affinities in the nanomolar range. TonB also possesses two distinct binding regions: one in the C terminus of the protein, for which binding to FhuA is ferricrocin-independent, and a higher affinity region outside the C terminus, for which ferricrocin enhances interactions with FhuA. Together these experiments establish that FhuA·TonB interactions are more intricate than originally predicted, that the TonB·FhuA stoichiometry is 2:1, and that ferricrocin modulates binding of FhuA to TonB at regions outside the C-terminal domain of TonB.

These proteins bind and transport iron-chelating siderophores from the external milieu into the periplasm in an energy-dependent manner. One such high affinity uptake system relies upon FhuA, receptor for the hydroxamate siderophore ferricrocin (Fc). 1 FhuA also facilitates the transport of siderophoreantibiotic conjugates such as albomycin and rifamycin CGP 4832, is the receptor for phages T5, T1, ⌽80, UC-1, for colicin M and for the antimicrobial peptide microcin MccJ21 (1)(2)(3). Energy for siderophore transport is provided by the cytoplasmic membrane (CM) complex TonB⅐ExbB⅐ExbD, which exploits the electrochemical potential from the proton motive force of the CM and transduces energy to the outer membrane (OM) (4).
Although many proteins involved in TonB-dependent siderophore transport have been identified and studied, the molecular mechanism of siderophore uptake from the external environment to the periplasm of the bacteria remains obscure. Evidence for conformational changes occurring in FhuA following siderophore binding was provided by differential recognition by monoclonal antibodies of ligand-free versus ligandloaded receptor (5). At the atomic level, x-ray crystallographic structures of FhuA (6,7), and the receptor FecA (8) displayed conspicuous structural changes upon ligand binding. A switch helix (residues 24 -29) on the periplasmic face of FhuA unwound to a random coil, and there was a 17-Å translocation of the extreme N terminus, proximal to the Ton box of FhuA. This conformational change was proposed to be a signal reporting the ligand-loaded status of FhuA to TonB (6,7). However, two other OM receptors, FepA and BtuB (9,10), whose structures have been published at atomic resolution, do not contain a switch helix. The unwinding of a switch helix is therefore not a common mechanistic feature for all OM receptors.
In vivo cross-linking studies of TonB and FepA provided the first biochemical evidence of interactions between TonB and OM receptors (11). These results corroborated genetic analyses where point mutations in the Ton box of FhuA were suppressed by mutations in the tonB gene, suggesting a functional interaction near Gln160 of TonB (12,13). Additional cross-linking studies demonstrated that specific ligands of the OM receptors enhance their association with TonB (14 -16).
Detailed analyses of TonB⅐BtuB interactions by site-directed spin labeling revealed that TonB requires a specific orientation for functional contact with the Ton box. Changes in conformation in the Ton box region caused by proline substitutions abrogated transport of the ligand (17). The crystal structure of C-terminal TonB (residues 164 -239) provided first evidence that TonB forms a dimer (18). To date, two models for TonB-OM receptor interaction have been proposed. In what has been termed the propeller model (19), two TonB monomers are intertwined, interacting with OM receptors. This model suggested that dimerized TonB undergoes rotary motion, similar to the mechanism described for the bacterial flagellar motor that is powered by MotA and MotB. ExbB and ExbD are homologous to MotA and MotB (20). An alternate model describes the shuttling of TonB between the CM and OM (19,21). The shuttle model is supported by in vivo labeling experiments that demonstrate periplasmic accessibility of the extreme N terminus of TonB to the cysteine-specific marker Oregon Green 488 maleimide. According to this model, TonB in complex with ExbB and ExbD in the CM are in an unenergized conformation. ExbB⅐ExbD use the proton motive force to energize TonB, allowing its C-terminal portion to interact with the OM. This may cause the release of the N terminus of TonB from the CM and transfer of stored potential energy from TonB to OM receptors, thereby facilitating ligand import.
To analyze interactions between FhuA and TonB, we selected two complementary biophysical methods. Analytical ultracentrifugation (AUC) was used to determine the buoyant molecular weights and stoichiometries of two genetically engineered, soluble derivatives of TonB: a hexahistidine-tagged full-length TonB (H6.ЈTonB) consisting of residues 32-239 of the mature protein; and a hexahistidine-tagged C-terminal TonB (H6.ЈTonB (CT)), residues 155-239, both purified to homogeneity. Sedimentation velocity experiments were conducted for the two TonBs and for FhuA, either individually or as protein⅐protein complexes. Surface plasmon resonance (SPR) optical biosensors (Biacore) were used to derive thermodynamic parameters of the interacting proteins. Our results demonstrate that the TonB⅐FhuA stoichiometry is 2:1, that TonB interacts with FhuA in an Fc-independent manner, that, in addition to its C-terminal portion, the N-terminal region of TonB participates in binding to FhuA, and that Fc modulates interactions between FhuA and the N-terminal region of TonB.

EXPERIMENTAL PROCEDURES
Strains-Escherichia coli AW740 harbors plasmid pHX405 and expresses recombinant FhuA with a hexahistidine tag at position 405 (22). E. coli ER2566, transformed with pET28 plasmid containing H6.ЈTonB, was similar to the construct described by Moeck and Letellier (15) and was corrected to reflect the wild-type sequences of residues 32-239 of TonB. In addition, E. coli ER2566 was the host strain for pET28 into which was cloned the gene for the C terminus of TonB (residues 155-239). All plasmids were confirmed for their sequence fidelity by DNA sequencing at Sheldon Biotechnology Center, McGill University.
Protein Purification-FhuA was purified (6,22) in N-lauryldimethylamine oxide (LDAO; Fluka) and detergent-exchanged using Q-Sepharose anion exchange media (Amersham Biosciences) into 100 mM HEPES (pH 7.4), 150 mM NaCl, 0.1% Tween 20 (Calbiochem), henceforth designated Biacore running buffer. H6.ЈTonB and H6.ЈTonB (CT) were purified using Ni 2ϩ -nitrilotriacetic acid Superflow resin (Qiagen) followed by cation exchange on SP-Sepharose (Amersham Biosciences). Prior to SPR experiments, each TonB protein was dialyzed into Biacore running buffer. To remove the hexahistidine tag, H6.ЈTonB was incubated with thrombin protease (Amersham Biosciences): 1 unit of protease per mg of H6.ЈTonB. After 3 h at ambient temperature, the reaction mixture was applied to a Ni 2ϩ -nitrilotriacetic acid column, capturing uncleaved H6.ЈTonB; the flow-through was applied to an SP-Sepharose column. Eluted protein was assayed by Western blotting with an anti-His6 monoclonal antibody (Cedarlane Laboratories Limited, Mississauga, Ontario, Canada) for the absence of the hexahistidine tag. TonB-reactive protein was confirmed by a cross-reactive monoclonal antibody that was raised against Trypanosoma brucei procyclin (CLP001A, Cedarlane) and that recognizes the proline-glutamic acid repeat portion of TonB protein. Protein concentrations were determined using the protein dye binding assay (Bio-Rad) and BCA assay (Pierce).
Analytical Ultracentrifugation-Samples were prepared for AUC by extensive dialysis against an AUC buffer: 100 mM HEPES (pH 8.0), 150 mM NaCl. In some experiments, AUC buffer was supplemented with either LDAO at 0.3% or n-octyltetraoxyethylene (C8E4; Bachem) at 0.4%. Protein concentrations were adjusted between 0.3 and 0.9 mg/ml. Ligand-loaded receptor was prepared by mixing Fc and FhuA in a 10:1 molar ratio, 30 min at room temperature. Excess Fc was removed by dialysis against AUC buffer using a Spectrapor dialysis membrane (POR-6, 25-kDa cutoff).
Sedimentation velocity experiments were performed on all protein samples using a Beckman XL-I Analytical Ultracentrifuge. The sample and the reference sectors of 1.2-cm path length double-sector ultracentrifuge cells were filled with 400 l of protein and AUC buffer, respectively. All sedimentation velocity runs were performed at 40,000 rpm, with absorbance scans monitored at 280 nm in 10-min intervals over a total spin time of 4 h at 24.6°C. Protein complexes were formed immediately prior to each spin by mixing H6.ЈTonB:FhuA at 1:1 and at 2:1 molar ratios, or by mixing H6.ЈTonB (CT):FhuA at 4:1 and at 8:1 molar ratios. In both cases, H6.ЈTonB and H6.ЈTonB (CT) were varied relative to a fixed concentration of FhuA or FhuA plus Fc.
Analysis of AUC Data-Sedimentation velocity data were analyzed by the computer program SEDFIT (23). Initial sedimentation profiles were obtained by fitting the data to the Continuous c(S) Model. Global frictional ratios determined by c(S) analysis were allowed to float to convergence. Buoyant molecular weight (M b ) and final s values were determined using the discrete non-interacting species model of SED-FIT. In sedimentation experiments involving LDAO, the detergent was modeled as a known sedimenting species with an s value of Ϫ0.1 and a micellar buoyant molecular mass of 2700 Da. Sedimentation coefficients of uncomplexed FhuA, H6.ЈTonB, and H6.ЈTonB (CT) were initially estimated from c(S) analyses and then refined by direct fitting to the Lamm equation using the non-interacting discrete species model of SEDFIT. Initial M b values were obtained from the known molecular weights of each protein. Both s and M b were allowed to float to convergence using the non-linear regression algorithms of SEDFIT.
Sedimentation velocity data of mixtures of FhuA plus H6.ЈTonB and FhuA plus H6.ЈTonB (CT) were fit to the discrete non-interacting species model using constrained analysis. From spins of the individual species, initial sedimentation parameters (s and M b ) of uncomplexed components (s1 and s2) were entered. The initial sedimentation parameters of predicted complexes (s3) were input based on the predicted M b as integral additions of the M b values of s1 and s2; initial sedimentation coefficients of s3 were estimated from c(S) analyses. The sedimentation parameters of s3 were then floated to convergence, whereas s1 and s2 parameters remained constrained. Fits to the model were evaluated on the basis of the distribution of residuals and r.m.s.d. errors. Optimization of the fits was performed by varying stoichiometric combinations of s1, s2, and s3 and repeating the constrained analyses until random distributions of residuals and minimal r.m.s.d. errors were obtained. As an unbiased test, data from mixtures of FhuA plus H6.ЈTonB and FhuA plus H6.ЈTonB (CT) were fit to the non-interacting discrete species model, assuming the presence of species with sedimentation parameters equivalent to uncomplexed s1 and s2. The r.m.s.d. errors and residuals were compared with those obtained from optimized fits in which the presence of s3 was modeled.
Preliminary and Steady-state Assays by SPR-SPR measurements were performed using a Biacore 2000 and a Biacore 3000 (Biacore AB) and were carried out in triplicate at 25°C. The data collection rate was set to 10 Hz. Biacore running buffer was used to dilute FhuA. Steadystate experiments were conducted with a flow rate of 5 l/min. To reach steady-state equilibrium, the injection time was 1200 s followed by injections of buffer for 240 s. Regeneration was achieved by three pulses (1 min) of 5 mM NaOH, 0.1% Tween 20, and a subsequent EXTRA-CLEAN procedure.
Biacore Data Preparation and Analysis-Data were prepared using the double referencing method (25). For global analysis, the sensorgrams were transformed to concentration units using the molecular weights of injected proteins. All curves were reduced to 700 evenly spaced sampling points. For each set of individual curves corresponding to injections of various concentrations of FhuA over the same surface, global fitting was carried out using a simple Langmuirian model in the SPRevolution software package of De Crescenzo and colleagues (26,27). For steady-state analysis, the apparent thermodynamic dissociation constants were determined by plotting the control corrected plateau value (R eq ) versus the injected concentration of FhuA. In the case of H6.ЈTonB, the K d values were derived by fitting the experimental R eq values to a model of two independent population interactions, where C corresponds to the injected FhuA concentration, R max1 and R max2 correspond to the maximal amount of FhuA that can be bound to each active H6.ЈTonB population, and K d1 and K d2 to their respective thermodynamic dissociation constants. Alternatively for H6.ЈTonB (CT), the experimental data were adequately fit with a simple interaction model in Equation 2 as follows.
The fitting procedure was performed in Microsoft Excel by non-linear regression with R max values and K d values as floating parameters.

RESULTS
Analytical Ultracentrifugation-Sedimentation velocity experiments on FhuA, H6.ЈTonB, and H6.ЈTonB (CT), individually and in complex, were performed in buffers containing the neutrally buoyant detergent C8E4 such that actual molecular weight values of the protein components could be determined. In advance of determining the stoichiometry of FhuA⅐TonB complexes, sedimentation parameters of the individual uncomplexed species were determined and used as prior knowledge in the analysis of sedimenting complexes. To obtain the best M r estimates of the sedimenting species, apparent vbar (⌽) values were determined from the known molecular weights of each protein, along with the buoyant molecular weights (M b ) determined by fitting to the non-interacting discrete species model of SEDFIT. As observed by Boulanger et al. (28), their experimentally determined vbar of FhuA (0.776 ml/g) is significantly different from the value predicted from the amino acid sequence of the protein (29). Similarly, using the known M r of FhuA and the experimental M b , we determine a ⌽ of 0.775 ml/g (Table I), in contrast to the sequence-predicted value of 0.735. The ⌽ values were also determined for H6.ЈTonB and H6.ЈTonB (CT), and these values (0.719 and 0.714 ml/g; Table I) also differ from vbars predicted from amino acid sequences (0.740 and 0.730 ml/g, respectively). The fits of the sedimentation velocity data of uncomplexed species to the non-interacting discrete species model were in all cases excellent with random distributions of residuals and least-squares (r.m.s.d.) errors of fit equal to or less than 0.0061.
Given these ⌽ values, FhuA sedimented as a monomeric protein with a molecular mass of ϳ80,000 Da; the presence of bound Fc resulted in a small yet reproducible increase in the sedimentation coefficient from 3.50 to 3.60 s, possibly due to structural rearrangements in the protein upon binding to ligand as has been observed in many of the crystal structures of bacterial OM receptors (6 -9, 30). H6.ЈTonB sedimented as a monomer in C8E4-containing AUC buffer with a M r of 27,430. H6.ЈTonB (CT) sedimented as a dimer with a M r of 23,090 (Table I). Using the same data sets, analyses modeling the presence of additional oligomeric states of either TonB species resulted in extremely poor fits to the data (data not shown). Of the three proteins analyzed, c(S) analysis indicates that H6.ЈTonB has the highest frictional ratio: f/f o ϭ 2.39. This is consistent with H6.ЈTonB having an extended conformation in solution. H6.ЈTonB (CT) has a lower frictional ratio (f/f o ϭ 1.96), because it is a truncated form of H6.ЈTonB; furthermore, H6.ЈTonB (CT) is dimeric, resulting in the protein having a globular shape as compared with the more extended H6.ЈTonB.
FhuA and the two TonB species were then mixed together for sedimentation velocity ultracentrifugation and were analyzed for the formation of complexes: FhuA plus H6.ЈTonB (CT) and FhuA plus H6.ЈTonB, both in the absence and presence of Fc. Sedimentation data were analyzed by first fitting to a contin-  uous c(S) model to obtain initial sedimentation parameters. This was followed (Table II) by direct fitting to the Lamm equation by a non-interacting discrete species model using the c(S)-derived sedimentation parameters and known molecular weights of individual proteins as prior knowledge (Table I). Refinement of c(S)-derived sedimentation parameters by the non-interacting species model resulted in converged s values similar to those obtained from the c(S) analysis, indicating a good correlation between the two models. However, unlike c(S) analysis, which employs a global frictional coefficient, the noninteracting species model resolves proteins of different diffusion coefficients, allowing for the calculation of buoyant molecular weight (M b ) for each sedimenting species. In the context of AUC, the term "non-interacting species" refers to sedimenting species that do not reversibly interact over the time frame of the experiment and can include stable protein complexes. In our case, a complex of FhuA and H6.ЈTonB (CT) or of FhuA and H6.ЈTonB is considered "non-interacting" if it is observed to sediment as a single species with a molecular weight that is additive of the uncomplexed components and with a sedimen-tation coefficient that does not significantly change as the ratio of the uncomplexed components is varied.
Experiments were performed with FhuA (Ϫ/ϩ Fc) mixed with either TonB species such that immediately prior to each centrifugation, proteins were combined at selected molar ratios of FhuA:H6.ЈTonB (CT) (monomer) ϭ 1:4 and 1:8, or FhuA: H6.ЈTonB (monomer) ϭ 1:1 and 1:2. Sedimentation velocity ultracentrifugation of these mixtures resulted in a combination of complexed species and uncomplexed species, which could be resolved by analysis of the sedimentation boundaries over the time course of the experiment. Sedimentation velocity data of both FhuA plus H6.ЈTonB (CT) and FhuA plus H6.ЈTonB mixtures fit well to the discrete non-interacting species model (Table II and Fig. 1, A-D). In addition to observing sedimenting species corresponding to uncomplexed FhuA (s1) and H6.ЈTonB (CT) or H6.ЈTonB (s2), a third sedimenting species (s3) corresponding to FhuA⅐TonB complex was always observed in the absence and presence of Fc. Modeling interactions without s3 led to dramatically poorer fits with highly non-random distributions of residuals (data not shown). The FhuA plus H6.ЈTonB  (CT) complex (s3) sedimented at ϳ4.5 s, a value significantly higher than that of uncomplexed FhuA (s1; 3.50 s). Uncomplexed dimeric H6.ЈTonB (CT) (s2) was also detected. In the absence of Fc, the abundance of s3 was increased significantly as the FhuA:H6.ЈTonB (CT) ratio increased from 1:4 to 1:8. Upon addition of Fc, a limiting amount of complex appeared to be formed at 1:4, suggesting that, although Fc is not necessary for complexation, it slightly enhanced the binding of H6.ЈTonB (CT) and FhuA. However, even at a FhuA:H6.ЈTonB (CT) ratio of 1:8, a significant proportion (37%) of FhuA remained uncomplexed in the presence of Fc (Table II).
Mixtures of FhuA plus H6.ЈTonB exhibited significantly different sedimentation behavior than FhuA plus H6.ЈTonB (CT). A sedimenting species corresponding to FhuA plus H6.ЈTonB complex (s3) was observed between 3.24 s and 3.53 s. This species was distinct from uncomplexed FhuA, having a M b additive of uncomplexed FhuA and two molecules of H6.ЈTonB. Furthermore, the oligomeric state of uncomplexed H6.ЈTonB (s2) changed in the presence of Fc, resulting in a transition from monomer to dimer. Unlike the FhuA plus H6.ЈTonB (CT) mixtures, the addition of Fc resulted in complete disappearance of uncomplexed FhuA (s1), suggesting that almost all of the OM receptor in the ligand-bound state interacts with H6.ЈTonB. In most cases, the absorbances of s1, s2, and s3 accounted for at least 90% of observed c tot . However, for FhuA plus Fc plus H6.ЈTonB mixtures (Table II, lines 7-8), a fourth sedimenting species was observed at ϳ4.0 s with an M b of ϳ14,000 Da. This species was found to account for ϳ30% of c tot . We previously observed a minor (Ͻ10% c tot ) sedimenting species at 3.6 s having a M b equivalent to monomeric TonB in uncomplexed H6.ЈTonB solutions, likely representing an alternate conformation of H6.ЈTonB (data not shown). Given the presence of dimeric H6.ЈTonB at ϳ1.90 s in FhuA plus Fc plus H6.ЈTonB mixtures, it is consistent that the species at ϳ4.0 s would correspond to a dimeric form of the alternate conformation observed in uncomplexed H6.ЈTonB solutions.
In the case of FhuA plus H6.ЈTonB (CT) mixtures, the M b of the complexes (Table II; 1-4). This reflects the binding of a H6.ЈTonB (CT) dimer to a FhuA monomer. This additivity is also observed for FhuA plus H6.ЈTonB in the presence of Fc. In the absence of Fc, however, some FhuA remains monomeric in solution while the remainder forms a 1:2 complex with H6.ЈTonB. A weight-averaged ⌽ value was determined for FhuA plus H6.ЈTonB and FhuA plus H6.ЈTonB (CT) complexes using the ⌽ values for the individual components and the theoretical molecular weight of the complex given a predicted FhuA:TonB stoichiometry of 1:2. In all cases, M r values of complex determined in this manner reflect the predicted molecular weights of 1:2 FhuA⅐TonB complexes.
The effect of detergent interaction with the two TonB proteins was also examined. Sedimentation behavior of H6.ЈTonB and H6.ЈTonB (CT) in the presence of C8E4 or LDAO or Tween 20 revealed significant changes in M b and s parameters (Table  III). The M b of both TonB proteins slightly decreased (1.40 s) in the presence of LDAO due to the floatation effect of the detergent, whereas the M b of H6.ЈTonB increased significantly in the presence of Tween 20, reflecting the higher density of this detergent. The detergent C8E4 does not affect the M b of either H6.ЈTonB or H6.ЈTonB (CT), because this detergent has a neutral buoyancy in aqueous solution. However, the sedimentation coefficients of both TonB proteins decreased in the presence of C8E4, suggesting that the detergent bound to these proteins and resulted in subtle changes to their overall shape, reflected by changes to the frictional coefficients. From these data, it is clear that all three detergents bound to H6.ЈTonB and that C8E4 and LDAO bound to H6.ЈTonB (CT).
Detection of FhuA⅐TonB Interactions by SPR-To determine the relative contribution of different portions of the TonB molecule when binding to FhuA, experiments were conducted on SPR-based biosensors, Biacore 2000 and Biacore 3000. In a typical SPR experiment, one of the binding partners (the ligand) is immobilized and the other interactant (the analyte) is injected in solution over the sensor chip surface. The resulting interaction between the ligand and the analyte is recorded in resonance units (RU), which are directly proportional to the mass accumulation on the surface. The first step in designing a Biacore experiment is to determine which interactant is to be coupled to the surface. Injections of H6.ЈTonB, and H6.ЈTonB (CT), both at 100 nM, over a control surface indicated that the two TonB proteins interacted non-specifically with the CM4 carboxymethylated dextran surface. In contrast, FhuA at the same concentration did not display nonspecific interaction. Thus the two TonB proteins were chosen to immobilize as ligand and FhuA as the detergent-solubilized analyte.
Preliminary experiments were conducted on the Biacore 2000 by coupling 35 RU of H6.ЈTonB. FhuA (12.5-400 nM) was preincubated with Fc (20 M), a 50-fold excess of Fc at the highest concentration of FhuA. Injections at a flow rate of 100 l/min resulted in significant binding when compared with a control surface. After regenerating the surface of immobilized H6.ЈTonB, FhuA that had not been preincubated with Fc was injected at matching concentrations; similar amounts of interacting FhuA were observed (Fig. 2A). These results are in contrast to reported observations that Fc enhances FhuA⅐TonB interactions (14,15). To determine whether trace amounts of Fc in the Biacore 2000 system were contaminating the FhuA injections, we decided to further study the FhuA⅐TonB interac- tion in the absence of Fc using a Biacore 3000, which had previously not been exposed to Fc. This resulted in similar responses, thus confirming the interaction between FhuA and H6.ЈTonB in the absence of Fc. To ensure that the interactions were specific to the TonB portion of the recombinant H6.ЈTonB protein and not influenced by the hexahistidine tag incorporated by the pET28 vector, equivalent amounts (30 RU) of thrombin-cleaved H6.ЈTonB were coupled to the sensor chip surface. Interactions with FhuA (Ϫ/ϩ Fc) gave similar responses as above (data not shown). The resulting sets of control-corrected sensorgrams were globally fit using a simple kinetic model. Deviations from the simple model were observed in the absence or presence of Fc as judged by the non-random distribution of the residuals (S.D. ϭ 0.729 without Fc and 0.675 with Fc). Artifacts such as mass transport and crowding effects were reduced by using a low loaded H6.ЈTonB surface and working at high flow rate. Thus, the observed deviations from the simple model are likely due to the complexity of the FhuA⅐TonB interactions as suggested by the AUC results (see above).
Experiments with H6.ЈTonB (CT) revealed that its interaction with FhuA had an apparent lower affinity than H6.ЈTonB. To enhance signal response, the amount of H6.ЈTonB (CT) coupled to the sensor chip surface was increased to 500 RU. FhuA interacted with H6.ЈTonB (CT) both in the absence and presence of Fc (Fig. 2B). As in the case of H6.ЈTonB, these interactions could not be depicted by a simple model (data not shown).
Steady-state Analysis of FhuA⅐H6.ЈTonB-To examine further the interactions between FhuA and H6.ЈTonB and the effects of Fc, experimental conditions were changed to reach a plateau at the end of each injection. Injection times were increased to 1200 s and flow rate was decreased to 5 l/min thereby reducing material consumption. To enhance signal response, the amount of H6.ЈTonB immobilized on the CM4 surface was increased to ϳ100 RU, and FhuA Ϫ/ϩ Fc injections were varied from 20 to 2000 nM. The experiments were conducted in duplicate over multiple H6.ЈTonB surfaces. Injections of FhuA alone demonstrated an increase in total FhuA bound to H6.ЈTonB compared with injections of FhuA plus Fc over the same surface (Fig. 3). Because Scatchard plots highlighted complex interactions both in the absence and presence of Fc (Fig. 3, insets A and B), the data were fit with a model assuming the presence of two distinct TonB populations displaying different affinities for FhuA. In the absence of Fc, the low affinity H6.ЈTonB population was determined to have an R max of 166.5 RU and a K d of 563.5 nM. The other H6.ЈTonB population was less abundant (R max ϭ 26.2 RU) yet displayed a higher affinity with a K d of 36.5 nM. In the presence of Fc, the affinity related to the low affinity population was decreased by 3-fold (K d ϭ 1530.0 nM) and had a similar R max (179.0 RU) as determined in the absence of Fc. In contrast, the affinity of FhuA for the H6.ЈTonB high affinity population was increased by 7-fold (K d ϭ 5.3 nM) in the presence of Fc and had a R max of 20.0 RU similar to that determined in the absence of Fc (Table IV).
Steady-state Analysis of FhuA⅐H6.ЈTonB (CT)-To determine an apparent K d , steady-state equilibrium analysis was conducted for H6.ЈTonB (CT) as previously performed for H6.ЈTonB. 500 RU of H6.ЈTonB (CT) was immobilized on the surface of a CM4 sensor chip; flow rates were set to 5 l/min and injection times to 1200 s. FhuA Ϫ/ϩ Fc was injected in duplicate over multiple H6.ЈTonB (CT) surfaces at concentra-  tions ranging from 50 to 5600 nM. Fitting of FhuA Ϫ/ϩ Fc data (Fig. 4) demonstrated a single population interaction with coupled H6.ЈTonB (CT), suggesting a single interaction site that was confirmed by Scatchard plot analysis (Fig. 4, insets A and  B). FhuA⅐Fc was characterized by an R max of 241 RU and a K dapp of 842 nM; the addition of Fc to FhuA resulted in an R max of 296 RU and a K dapp of 1490 nM (Table IV). DISCUSSION Our analytical ultracentrifugation establishes that both H6.ЈTonB and H6.ЈTonB (CT) form stable 2:1 complexes with FhuA, both in the absence and presence of ferricrocin. Although we find that H6.ЈTonB binds to FhuA in a 2:1 complex, the sedimentation coefficient for this complex (3.3 s) is significantly lower than that of FhuA plus H6.ЈTonB (CT) (4.5 s). Given the quality of the fit to our AUC data, we propose that such sedimentation behavior is well described by the noninteracting species model. SPR experiments indicate that FhuA plus H6.ЈTonB complex in the absence and presence of Fc is stable, as indicated by its slow dissociation (Fig. 2A). The sedimentation behavior of s3 therefore reflects a stable complex and not the reversible interactions of s1 (FhuA) with s2 (H6.ЈTonB) (Table II). These observations are consistent with data ( Table I)  In contrast to H6.ЈTonB (CT), we observed that in the absence of Fc, H6.ЈTonB is monomeric in mixtures with FhuA. This suggests that the N-terminal portion of TonB prevents homodimerization of the protein. Given that TonB is highly elongated and rigid structure (31), intramolecular interactions between the N-terminal and C-terminal portions of TonB may prevent dimerization of the C-terminal domain. We also observed a minor (Ͻ10% abundance), higher sedimenting species of TonB with a monomeric TonB molecular weight, suggesting that the protein adopts multiple conformations in solution. In the presence of Fc, however, the FhuA plus H6.ЈTonB mixture shows an uncomplexed H6.ЈTonB dimer. We propose that interaction of H6.ЈTonB with ligand-bound FhuA results in conformational changes in H6.ЈTonB that break its intramolecular interactions, allowing formation of a stable dimer. Apparently TonB dimerization is facilitated by interactions with ligandloaded FhuA and the stable H6.ЈTonB retains its oligomeric state upon dissociation from FhuA.
Steady-state analysis by SPR of TonB interactions with FhuA clearly establishes that the ligand-loaded state of the OM receptor has an effect on the binding of TonB. Our analyses (Table IV) revealed two H6.ЈTonB populations, characterized by different affinities when binding to FhuA both in the absence and presence of Fc (Fig. 3). In contrast, binding of H6.ЈTonB (CT) to FhuA in the absence and presence of Fc is adequately fit as a single TonB population interacting with the OM receptor (Fig. 4). Comparing affinities determined by the steady-state analysis for binding of H6.ЈTonB and of H6.ЈTonB (CT) to FhuA, we conclude that the low affinity population detected for interaction between full-length TonB and FhuA corresponds to the binding of C-terminal TonB domain to FhuA. The additional K d value observed for H6.ЈTonB demonstrates a higher affinity population. We therefore propose that these K d values reflect a single low affinity binding region that resides in the C-terminal domain of TonB and a higher affinity binding region, which occurs outside the C-terminal domain of TonB. Affinity for FhuA may also be enhanced by a more complex set of interactions where the N-terminal binding region synergistically enhances C-terminal TonB interactions with FhuA. SPR experiments show that the low affinity binding region and the high affinity binding region respond to the addition of Fc in an inverse manner (Table IV). Whereas the K d for the low affinity binding region observed in experiments with both H6.ЈTonB and H6.ЈTonB (CT) increases ϳ2-fold in the presence of Fc, the K d for the high affinity binding region in H6.ЈTonB decreases about 7-fold in the presence of ligand. Our AUC results support this observation, because in the presence of Fc almost all FhuA is complexed to H6.ЈTonB. In contrast, even in the presence of excess H6.ЈTonB (CT) and Fc, a significant population of uncomplexed FhuA remains in mixtures of FhuA and H6.ЈTonB (CT). This may reflect conformational rearrangements occurring at the high affinity binding region upon ligand loading of the OM receptor, resulting in the large change in K d observed in the SPR experiments. Taken together, these results provide the first evidence that a FhuA-binding region on TonB exists outside the C-terminal domain of TonB. This high affinity binding region is more sensitive to the ligandloaded state of the OM receptor than the low affinity binding region, which corresponds to the C-terminal domain. We propose that TonB associates with FhuA in two discrete states: a complex that reflects transient encounters and a more stable functional complex, which is necessary for TonB-dependent energy transduction. The Fc-independent encounter complex is not equivalent to the functional complex that involves a modulation of affinities within the two binding regions.
Interaction models for H6.ЈTonB (CT) and H6.ЈTonB with FhuA are illustrated in Fig. 5. Our analyses indicate that the C-terminal domain of TonB exists in solution as a dimer and it forms an encounter complex with FhuA even in the absence of Fc. The affinity of interaction decreases when FhuA is loaded with Fc. Therefore, interactions between FhuA and the Cterminal domain of TonB are stronger when the OM receptor is in a ligand-free state. Upon binding of ligand, the affinity of H6.ЈTonB (CT) for FhuA decreases; however, they remain complexed (Fig. 5A). The solution state of H6.ЈTonB has been shown to be monomeric, indicating that complexation of H6.ЈTonB and FhuA is initiated by a single H6.ЈTonB. Dimerization of TonB may then be facilitated by interactions involving two bound TonB monomers with the OM receptor. Of the two affinities determined for H6.ЈTonB by SPR analyses, the lower affinity corresponds well to the C-terminal portion of the protein; the higher affinity region resides N-terminal to residue 154. In the absence of Fc, both binding regions interact with FhuA with their affinities specified in Table IV. Upon binding of ligand to FhuA, binding at the low affinity site decreases with a concomitant increase in affinity at the high affinity site in the N-terminal portion of TonB, thereby facilitating transition to the functional complex (Fig. 5B).
In the context of TonB-OM receptor interactions, the models of binding proposed in Fig. 5 can be expanded to incorporate known structural changes caused by Fc. The unwinding of the switch helix and the translocation of the N terminus of FhuA upon Fc binding, as resolved by x-ray crystallography (6), may result in the decreased affinity observed at the C-terminal binding region. This large conformational change in FhuA may also allow for the Fc-induced enhancement of binding at the higher affinity region. To date, it has been recognized that TonB-OM receptor interactions are enhanced upon ligand binding, but not at a site other than the C terminus, and it is now evident that other sites within TonB are also critical.
Our proposed models are consistent with certain aspects of the "shuttle model" of TonB interaction with OM receptors proposed by Postle and Kadner (19). In this model, TonB is proposed to make initial contact with the OM receptor through its C-terminal domain. This contact results in dissociation of TonB from the ExbB⅐ExbD complex enabling additional contacts between the OM receptor and sites within the N-terminal domain of TonB. Indeed, our SPR data support this model in that we show potential contacts between the N terminus of TonB and FhuA. Our data also support aspects of the "propeller model" proposed by Chang et al. (18) in which dimerization of TonB is necessary for complexation with the OM receptor such that energy-dependent ligand transfer can occur. This is consistent with our concept of a functional TonB⅐FhuA complex.
Recently, in vivo dimerization studies of TonB demonstrated (32) that TonB (residues 164 -239) fused to the cytoplasmic ToxR (residues 1-182) has a strong propensity to form dimers in the periplasm. This observation matches our AUC results for H6.ЈTonB (CT) dimerization. Their study also demonstrated that TonB (residues 33-239) fused to cytoplasmic ToxR did not dimerize, a result substantiated by AUC analysis of H6.ЈTonB.
Sauter et al. proposed that interactions of TonB with the ligandloaded OM receptor FecA may affect TonB dimerization, consistent with our observations. The in vivo dimerization study, taken together with our results, reconcile aspects of the shuttle and propeller models and suggest that more intricate mechanisms of TonB interactions with OM receptors must be considered.