Characterization of the Bacterioferritin/Bacterioferritin Associated Ferredoxin Protein–Protein Interaction in Solution and Determination of Binding Energy Hot Spots

Abstract

Mobilization of iron stored in the interior cavity of BfrB requires electron transfer from the [2Fe–2S] cluster in Bfd to the core iron in BfrB. A crystal structure of the Pseudomonas aeruginosa BfrB:Bfd complex revealed that BfrB can bind up to 12 Bfd molecules at 12 structurally identical binding sites, placing the [2Fe–2S] cluster of each Bfd immediately above a heme group in BfrB [Yao, H., et al. (2012) J. Am. Chem. Soc., 134, 13470–13481]. We report here a study aimed at characterizing the strength of the P. aeruginosa BfrB:Bfd association using surface plasmon resonance and isothermal titration calorimetry as well as determining the binding energy hot spots at the protein–protein interaction interface. The results show that the 12 Bfd-binding sites on BfrB are equivalent and independent and that the protein–protein association at each of these sites is driven entropically and is characterized by a dissociation constant (K_d) of approximately 3 μM. Determination of the binding energy hot spots was carried out by replacing certain residues that comprise the protein–protein interface with alanine and by evaluating the effect of the mutation on K_d and on the efficiency of core iron mobilization from BfrB. The results identified hot spot residues in both proteins [L_B⁶⁸, E_A, and E_A⁸⁵ in BfrB (superscript for residue number and subscript for chain) and Y² and L⁵ in Bfd] that network at the interface to produce a highly complementary hot region for the interaction. The hot spot residues are conserved in the amino acid sequences of Bfr and Bfd proteins from a number of Gram-negative pathogens, indicating that the BfrB:Bfd interaction is of widespread significance in bacterial iron metabolism.

Iron is essential for bacteria due to its involvement in multiple metabolic processes including respiration (heme-containing proteins, [Fe–S]-containing ferredoxins) and key enzymatic reactions ([Fe–S]-containing proteins, such as fumarase and aconitase of the TCA cycle).¹ Pathogenic bacteria must obtain iron from the host to support growth, but humans maintain extremely low concentrations of free iron, which are further reduced during infections.^2,3 In bacterial cells, iron homeostasis plays a pivotal role in guarding against iron-induced toxicity caused by the propensity of free iron to stimulate the formation of reactive oxygen species (ROS) such as superoxide, hydrogen peroxide, and the highly destructive hydroxyl radical.⁴ Consequently, free levels of iron in bacteria are tightly regulated to ensure sufficiency for metabolic needs while preventing iron-induced toxicity.⁵ In humans, iron is stored in ferritin (Ftn),⁶ whereas in bacteria, iron reserves are stored in a homologous bacterial Ftn and in bacterioferritin (Bfr), of which the latter is present only in bacteria.^7,8 Ftns, bacterial Ftns, and Bfrs assemble from 24 subunits into a hollow spherical structure with an outer diameter of 120 Å and an inner diameter of 80 Å, in which ≈3500 iron atoms can be stored in the form of an Fe³⁺ mineral. Despite their similar spherical architecture, eukaryotic and prokaryotic Ftns and Bfrs share <18% amino acid sequence homology. The unique primary sequence of each type of ferritin-like molecule strongly influences the packing of the corresponding 24-mer protein shells as well as their function.^9,10

Bfrs are unique in possessing intrinsic heme groups, which are bound at 2-fold intersubunit sites by axial coordination of the heme-iron by methionine residues from adjacent subunits (Figure 1A). Hence, a 24-mer Bfr can be thought of as being assembled from 12 subunit dimers, each harboring a heme, which is buried deep below the protein surface, such that the heme propionates reach into the interior cavity (Figure 1B).⁷ Bfrs function by (i) utilizing O₂ or H₂O₂ to oxidize Fe²⁺ and store Fe³⁺ in their internal cavity (core) and (ii) accepting electrons to reduce Fe³⁺ in their core and mobilize Fe²⁺ to the cytosol.^11,12 In bacteria, the concentration of iron not incorporated in proteins (free iron), Fe²⁺, is maintained by a dynamic equilibrium between Fe³⁺ stored in ferritin-like molecules and Fe²⁺. Hence, one function of ferritin-like molecules is to maintain the cytosolic Fe²⁺ concentrations that enable Fur, the major iron-responsive regulator which is conserved in many bacteria,¹³ to perform its broad range of regulatory functions that link iron homeostasis to broader bacterial metabolism.¹⁴⁻¹⁶ For example, Bfr mutants of Mycobacterium tuberculosis suffer from iron-mediated oxidative stress and are unable to persist in mice and guinea pig models of infection.^17,18 A Bfr-deficient mutant of Nisseria gonorrheae is growth-impaired in iron-limited medium and more sensitive to killing by hydrogen peroxide,¹⁹ and in the plant pathogen Erwinia chrysanthemi, mutation of the bfr gene results in impaired iron utilization and growth defects.²⁰

Figure 1.

Structural organization of BfrB (PDB: 3IS8) and schematic representation of the process of iron mobilization from BfrB. (A) View of a subunit dimer harboring a heme molecule (red), which is coordinated by a conserved methionine from each of the subunits. (B) View of the BfrB interior cavity where iron mineral is stored; heme (red) is buried below the protein shell surface, with the heme propionates extending into the interior cavity. (C) Model of iron mobilization from the BfrB interior cavity, where electrons from NADPH are transferred via FPR and Bfd into the Fe³⁺ mineral stored in the core of BfrB.

Early studies with Escherichia coli suggested that mobilization of iron from Bfr requires interactions with a ferredoxin, dubbed Bfd (bacterioferritin-associated ferredoxin) because the bfd gene is adjacent to the bfr gene.²¹⁻²³ While mining the global genetic response of Pseudomonas aeruginosa to high and low iron conditions,²⁴ we noticed that of the ≈120 genes reported to be upregulated by low-iron conditions, the expression of bfd is upregulated 200-fold and that of a gene coding for a ferredoxin reductase (fpr) is increased 3-fold.²⁴ In contrast, the bfrB gene is downregulated under iron-limiting conditions.²⁵ The strong upregulation of the bfd gene in response to low iron led us to hypothesize that the Bfd protein functions in the mobilization of iron stored in BfrB by accepting electrons from the ferredoxin reductase (FPR) and transferring these to the ferric iron in BfrB for subsequent mobilization of Fe²⁺ (Figure 1C).²⁶ We cloned the genes, characterized the P. aeruginosa BfrB, Bfd, and FPR proteins biochemically and structurally,^27,28 and showed that Bfd, FPR, and NADPH are sufficient to mobilize iron from BfrB in vitro.²⁶

More recently, we reported the X-ray crystal structure of the BfrB:Bfd complex.¹² The asymmetric unit contains three BfrB subunit dimers, each associated with a Bfd molecule, giving rise to a biological assembly consisting of a nearly spherical 24-mer BfrB with 12 Bfd molecules (Figure 2A). Each Bfd binds at an identical site on BfrB, at the interface of a subunit dimer, above each of the heme molecules, placing the [2Fe–2S] cluster of Bfd approximately 22 Å from the heme-iron in BfrB (Figure 2B). Studies in solution demonstrated that formation of the BfrB:Bfd complex enables the heme in BfrB to conduct electrons from the [2Fe–2S] cluster in Bfd to the mineral core, thus accelerating reduction of the Fe³⁺ mineral core and the mobilization of Fe²⁺.^12,26 A zoomed-in view of one of the 12 identical Bfd-binding sites on BfrB (Figure 2C) illustrates the proximity of the [2Fe–2S] cluster to the BfrB surface and shows the interactions that are likely important to stabilize the complex: (i) The side chain of Y² in Bfd anchors in a cleft on the BfrB surface, which is formed by the side chains of E⁸¹ in subunit A (E_A⁸¹) and L⁶⁸ in subunit B (L_B), (ii) the side chain of L⁵ in Bfd fits in a pocket formed mostly by L_B⁶⁸, N_B, L_B⁷¹, and L_A, (iii) the backbone NH group of M¹ forms H-bonds with carboxyl side chains in E_A⁸⁵ and E_A on BfrB and its side chain interacts hydrophobically with L_A⁷⁷, L_A, I_A⁷⁹, G_A, and C_A⁸⁹ on the BfrB surface, and (iv) the side chain of K⁴⁰ in Bfd interacts hydrophobically with the side chain of K_A in BfrB and forms hydrogen bonds with carbonyl oxygen of BfrB Q_B⁷² and L_A.

Figure 2.

Structure of the BfrB:Bfd complex (PDB: 4E6K). (A) Twelve Bfd molecules bind a 24-mer BfrB at 12 identical binding sites. (B) Each Bfd molecule (cyan) binds at the interface of a BfrB subunit dimer, above a heme molecule. (C) Zoomed-in view of the BfrB:Bfd interface depicting residues in Bfd (cyan sticks) interacting with the BfrB surface; subunit A in BfrB is rendered green, and subunit B is gray. O, N, S, and Fe are shown in red, blue, yellow, and orange, respectively.

In the present study, we examined the BfrB:Bfd interaction in solution with the aid of surface plasmon resonance (SPR) and isothermal titration calorimetry (ITC) and determined that the 12 Bfd-binding sites on BfrB, which are structurally identical, are also independent. Bfd binds at each of the sites on BfrB with a K_d of approximately 3 μM, and the interaction, although endothermic, is favored by a relatively large change in entropy. Having obtained a benchmark for the interaction between the two proteins, we mutated several residues that comprise the interface to alanine. These experiments revealed how particular side chains and the Bfd binding surface of BfrB contribute to the stability of the complex and to the Bfd-promoted mobilization of core iron from BfrB. While conducting this work, we noticed that recombinant BfrB used in our prior studies contained an extra methionine at position zero (M⁰), immediately upstream from the initiator methionine. We used site-directed mutagenesis to eliminate M⁰ and demonstrated that M⁰ BfrB is structurally identical to wt BfrB and that M⁰ BfrB binds Bfd with a K_d indistinguishable from the K_d obtained for the association between Bfd and wt BfrB.

Experimental Procedures

Site-Directed Mutagenesis

Mutations were introduced to the recombinant bfd and bfrB genes²⁶ with the aid of the Quik Change mutagenesis kit (Stratagene, La Jolla, CA) using the manufacturer’s instructions. Primer pair sequences are provided in Supporting Information. Mutations were confirmed by sequencing (ACGT Inc., Wheeling, IL). The C⁴³S mutant of Bfd is more stable to purification and manipulation but has the same functional and spectroscopic properties as those of wt Bfd; just as important, C43 is not a ligand to the [2Fe–2S] cluster, and it is far from the area used by Bfd to bind BfrB.^12,26 Consequently, the C⁴³S mutant will be referred to as wt Bfd in this article. The following mutations were introduced to the wt bfd gene: Y²A, Y²F, L⁵A, and K⁴⁰A. The extra methionine (M⁰) in recombinant BfrB was removed by eliminating the corresponding codon in the bfrB gene. BfrB with the correct sequence is termed wt BfrB, and BfrB harboring the extra methionine will be referred to as M⁰ BfrB.

Protein Expression and Purification

Recombinant FPR and Bfd (wt and mutants) were expressed and purified as described previously.²⁸ Recombinant BfrB (wt and all mutants) were expressed and purified using previously reported protocols^12,26 with some modifications: E. coli Arctic express RIL cells harboring the recombinant pET11a/bfrB construct were cultured overnight at 37 °C in 50 mL of LB medium containing 100 μg/mL ampicillin and 20 μg/mL gentamicin. The cells were then cultured in 1 L of fresh LB containing no antibiotics for 3 h at 30 °C, transferred to a shaker incubator pre-equilibrated at 10 °C, and incubated for 45 min before protein expression was induced by addition of 1 mM 1-thio-d-galactopyranoside (IPTG). Cells were cultured for an additional 48 h at 10 °C before they were harvested by centrifugation and stored at −20 °C. Cell paste was suspended in a solution containing 50 mM Tris-Base (pH 7.6), 5 mM dithiothreitol (DTT), 0.5 mM phenylmethanesulfonylfluoride (PMSF), and 0.5% Triton X-100 (Sigma) and sonicated on ice. Cell debris were pelleted by centrifugation at 4 °C and 19 500 rpm, and the supernatant was dialyzed against 50 mM Tris-Base (pH 7.6), loaded onto a Q-Sepharose Fast Flow column (12 cm × 2.5 cm i.d.) equilibrated with the same buffer, and eluted with a gradient of 0 to 600 mM NaCl. Fractions containing BfrB were dialyzed against 50 mM potassium phosphate buffer (pH 7.6) containing 1 mM tris(2-carboxyethyl)phosphine (TCEP), loaded onto a hydroxyapatite Bio-Gel column (10 cm × 2.5 cm i.d.) equilibrated with the same buffer, and eluted with a 50 to 800 mM potassium phosphate gradient (pH 7.6), 1 mM TCEP. Isolating fractions containing BfrB and repeating this chromatographic step yield pure BfrB. The separation of 24-mers from incompletely assembled protein was carried out by FPLC (AKTA GE Healthcare) in a Hiload 16/600 superdex 200 pg column equilibrated and eluted with 100 mM potassium phosphate (pH 7.0). BfrB purified in this manner is nearly devoid of heme. To reconstitute with heme, a previously described method²⁹ was used with some modifications: Hemin chloride was dissolved in 0.1 M sodium hydroxide solution and diluted with potassium phosphate buffer (100 mM, pH 7.0) to a final concentration of 1.5 mM. Insoluble material was removed by centrifugation. BfrB (3 μM) in 100 mM potassium phosphate (pH 7.0) and 1 M NaCl was equilibrated at 80 °C for 1 min in a dry bath (LabDoctor Dry Bath Plus), followed by the addition of a first aliquot of heme, which delivered 6 heme molecules per 24-mer BfrB, and the mixture was incubated for 10 min at 80 °C. The temperature was then increased to 90 °C prior to addition of a second aliquot of heme delivering 6 heme molecules per 24-mer BfrB, followed by incubation at 90 °C for 10 min. The temperature was then increased to 95 °C prior to addition of a third aliquot of heme delivering 6 heme molecules per 24-mer, followed by incubation of the mixture at 95 °C for 10 min. The resultant mixture was cooled on ice and loaded onto a Hiload 16/600 Superdex 200 pg column equilibrated and eluted with 100 mM potassium phosphate (pH 7.6) at 4 °C. The desired fractions were collected and supplemented with 1 mM TCEP. The heme content of BfrB was analyzed according to a previously described method.^26,30 Reconstitution of BfrB with an iron core was carried out as reported previously.²⁶ Upon addition of 500 Fe ions/BfrB, the protein solutions were incubated overnight at 4 °C and then passed through a Superdex 200 10/300 GL column. The content of iron in bacterioferritin, before and after reconstitution with an iron core, was measured using a previously reported protocol.^26,31

Crystallization and X-ray Diffraction Data Collection

Crystallization experiments were conducted in Compact 300 sitting drop vapor diffusion plates (Rigaku Reagents) at 20 °C. Equal volumes of BfrB (10 mg/mL in 100 mM potassium phosphate, pH 7.6 and 1 mM TCEP) and crystallization solution were mixed and equilibrated against 75 μL of crystallization solution. All crystals were obtained from the Wizard 3-4 screen (Rigaku Reagents) except for the wt BfrB crystals, which employed the Cryo HT screen (Rigaku Reagents). All crystals were observed within 1 to 2 days. The observed morphology, crystallization conditions, and cryoprotectants are as follows. wt BfrB: Crystals displaying plate morphology were obtained in 1 to 2 days from condition H6 (30% (v/v) PEG 200, 100 mM Na acetate, pH 4.5, 200 mM sodium chloride). Crystals were transferred to a fresh drop of crystallant, which served as the cryoprotectant. L⁶⁸A BfrB: Prismatic crystals from well E10 (50% (v/v) PEG 200, 100 mM Na cacodylate, pH 6.5, 200 mM magnesium chloride) were transferred to a drop of crystallant, which served as the cryoprotectant. L⁶⁸A/E⁸¹A BfrB: Prismatic crystals from well A2 (30% (v/v) 2-methyl-2,4-pentanediol, 100 mM Na acetate, pH 4.6, 20 mM calcium chloride) were transferred to a fresh drop of crystallant, which served as the cryoprotectant. E⁸¹A BfrB: Plate shaped crystals from well F12 (2.4 M sodium malonate) were transferred to a drop of 3.4 M sodium malonate, which served as the cryoprotectant. E⁸⁵A BfrB: Plate shaped crystals from well F4 (15% (v/v) PEG 550MME, 100 mM MES, pH 6.5) were transferred to a drop containing 25% PEG 550MME and 75% crystallant. Fe-soaked crystals of wt and L⁶⁸A/E⁸¹A BfrB were prepared by soaking native crystals for 10 min in 50 mM FeCl₂ freshly dissolved in crystallization solution and frozen in liquid nitrogen. X-ray diffraction data were collected at the Advanced Photon Source Beamline 17-ID (Argonne National Laboratory) using a Dectris Pilatus 6 M pixel array detector. All diffraction data sets were collected at λ = 1.0000 Å except for the iron-bound crystals, which were collected at λ = 1.720 Å in order to obtain the maximum iron anomalous signal. The anomalous difference maps were used to assign iron atoms to specific sites, and peak heights and B-factors of the iron atoms were compared for each particular structure.

Structure Solution and Refinement

Intensities were integrated using XDS,³² and the Laue class analysis and data scaling were performed with Aimless.³³ Structure solution was conducted by molecular replacement with Phaser³⁴ using a single subunit of a previously determined BfrB structure (PDB: 3IS7) as the search model. Refinement and manual model building were conducted with Phenix³⁵ and Coot,³⁶ respectively. Structure validation was conducted with Molprobity.³⁷ Disordered side chain atoms were truncated to the point where electron density could be observed. Further details are provided below, and relevant crystallographic data are provided in Table S1 of the Supporting Information.

L⁶⁸A BfrB

The highest probability Laue class was mmm, and the suggested space group was P2₁2₁2₁, with a = 117.86 Å, b = 125.66 Å, and c = 169.99 Å. On the basis of the Matthew’s coefficient³⁸ (V_m = 2.9 ų/Da, 58% solvent), the asymmetric unit likely contained 12 molecules of BfrB. The top solution was obtained in the space group P2₁22₁, with 12 molecules in the asymmetric unit. Since P2₁22₁ is a nonstandard setting of P2₁2₁2, the data were reindexed using the operator (h,–l,k) to transform the reflections into the standard space group setting P2₁2₁2 with a = 117.86 Å, b = 169.99 Å, and c = 125.66 Å

Following initial refinement, large peaks of electron density were observed in the 4-fold pores and B-pores. The 4-fold pores of BfrB are typically occupied by K⁺ ions.²⁷ However, large peaks of positive electron density (F_o – F_c) greater than 3σ were observed at these sites when modeled as K⁺ ions. A phased anomalous difference map revealed peaks at the K⁺ binding sites of the 4-fold pore that were between 10σ and 11σ. Given that the crystals were obtained from cacodylate buffer and the maximum anomalous signal for arsenic is at approximately λ = 1.047 Å (f″ ∼ 3.9e^–), it appeared that arsenic ions of some form occupied the 4-fold pore sites. The data for this structure were collected at λ = 1.0000 Å, which would still yield an appreciable anomalous signal (f″∼ 3.6e^–). It was clear from the electron density, however, that a cacodylate ion was not bound to this region and thus the electron density was ultimately modeled as AsH₃ (Figure S1). It is unclear how this form of arsenic would be bound at this site, but perhaps it is due to impurities in the cacodylate buffer or due to reactivity with the TCEP that is present in the protein storage buffer. The B-pores also displayed large peaks of electron density, which were modeled as Mg²⁺ ions obtained from the crystallant. Most of the Mg²⁺ are hexacoordinated by water molecules, which also form contacts with residues in the B-pores (Figure S2).

L⁶⁸A/E⁸¹A BfrB

The highest probability Laue class was −3, and the suggested space groups were P3, P3₁, or P3₂. On the basis of the Matthew’s coefficient³⁸ (V_m = 2.9 ų/Da, 57% solvent), the asymmetric unit likely contained 24 molecules of BfrB. All space groups with 3 point symmetry were tested, and the top solution was obtained in space group P3₂ with 24 molecules in the asymmetric unit.

E⁸¹A BfrB

The highest probability Laue class was mmm, with possible space groups C222 or C222₁. On the basis of the Matthew’s coefficient³⁸ (V_m = 2.9 ų/Da, 55% solvent), the asymmetric unit likely contained 12 molecules of BfrB. The top solution was obtained in space group C222₁.

E⁸⁵A BfrB

The highest probability Laue class was 4/mmm, with the highest probability space group being P4₂2₁2. On the basis of the Matthew’s coefficient³⁸ (V_m = 2.8 ų/Da, 56% solvent), the asymmetric unit likely contained 12 molecules of BfrB. All space groups with 422 point symmetry were tested, and the top solution (12 molecules/ASU) was obtained in space group P4₂2₁2.

Measurement of the BfrB:Bfd Binding Affinity by Surface Plasmon Resonance (SPR)

SPR experiments were performed at 22 °C using a Biacore 3000 instrument (GE Healthcare). BfrB was immobilized on a CM5 sensor chip using amine coupling chemistry.³⁹ The sensor chip surface was preconditioned with 50 mM NaOH, 10 mM HCl, 0.1% SDS, and 0.085% H₃PO₄ and activated by injecting 0.1 M N-hydroxysuccinimide (NHS) and 0.4 M 1-ethyl-3-(3-(dimethylamino)propyl) carbodiimide (EDC) in water. The running buffer was PBS with 1.5 mM TCEP, pH 7.4. Immobilization of BfrB was carried out by flowing (10 μL/min) a 100 nM solution of BfrB in 10 mM sodium acetate buffer (pH 5.0) for 32.5 min, followed by quenching surface-activated sites not coupled to BfrB by flowing (10 μL/min) 1.0 M aqueous ethanolamine-HCl (pH 8.5) for 7 min. A cell activated by NHS/EDC and quenched by ethanolamine but not containing BfrB was used as the reference surface, and a cell immobilized with bovine serum albumin (BSA) was used as control. To determine the BfrB:Bfd binding affinity, a solution of Bfd in 50 mM potassium phosphate (pH 7.0), 150 mM NaCl, 5 mM DTT was simultaneously flowed (25 μL/min) over the cell containing immobilized BfrB, the reference cell, and the cell containing immobilized BSA at 25 μL/min. Bfd solutions with the following concentrations were used to construct a binding curve: 0.5, 1.0, 2.0, 5.0, 10.0, 20.0, and 40.0 μM. The corresponding responses were double-referenced by subtracting the bulk refractive index change caused by difference in buffer compositions as well as by subtracting the response from the reference cell. Experiments were conducted in triplicate.

Measurement of the BfrB:Bfd Binding Affinity by Isothermal Titration Calorimetry (ITC)

Prior to each experiment, BfrB and Bfd were exchanged into 100 mM potassium phosphate buffer (pH 7.0), 150 mM NaCl, and 1 mM TCEP by passage through a Superdex 200 10/300 GL or superdex 75 10/300 GL column, respectively. ITC was performed at 22 °C using a MicroCal auto ITC instrument (GE Healthcare). The cell was loaded with 1.45 mL of 18 μM Bfd and titrated with a total volume of 0.25 mL of 20 μM BfrB in 25 injections, each lasting 20 s and delivering 10.02 μL, with 300 s between injections. A similar titration of BfrB into buffer served as a control to measure the heat of dilution. The experiment was repeated three times.

Mobilization of Iron from BfrB

These experiments were carried out in an anaerobic chamber according to a previously described method²⁶ with small modifications: A 1 cm path-length cuvette was filled with a solution consisting of BfrB (0.18 μM) reconstituted with 500 ± 20 Fe³⁺ ions/BfrB, Bfd (0.9 μM), FPR (7.2 μM), and α,α′-bipyridine (bipy), 3 mM. The reactions leading to mobilization of Fe²⁺ from BfrB (Figure 1C) were initiated by addition of 1 mM NADPH, and the process was monitored by tracking the time-dependent formation of the [Fe(bipy)₃]²⁺ complex, which absorbs at 523 nm. The percentage of iron released was calculated by normalizing the intensity of the 523 nm band to the intensity expected upon mobilization of the 500 iron ions stored in BfrB.

Results and Discussion

Characterization of the BfrB:Bfd Interaction by SPR Shows That the 12 Bfd-Binding Sites on BfrB Are Identical and Independent

The X-ray crystal structure of the BfrB:Bfd complex showed that each Bfd binds at the interface of a BfrB subunit, near the 2-fold axis of symmetry, placing the [2Fe–2S] cluster of Bfd approximately 22 Å from the heme-iron in BfrB.¹² Studies in solution showed that Bfd binding to BfrB makes the BfrB-heme conductive, accelerates reduction of the Fe³⁺ mineral core, and facilitates mobilization of Fe²⁺.^7,26 As part of this study, we investigated the binding affinity of the BfrB:Bfd complex using SPR. Figure 3A shows the reference- and baseline-subtracted responses obtained from flowing wt Bfd over wt BfrB immobilized on a sensor chip. At each Bfd concentration, the corresponding response reaches a plateau, indicating steady-state equilibrium. Plotting each response at steady state as a function of Bfd concentration results in the hyperbolic binding curve defined by the black circles in Figure 3B. The interpretive model used to fit these data is described below.

Figure 3.

Measuring the affinity of the BfrB:Bfd interaction using SPR. (A) Overlay of reference- and baseline-subtracted sensograms resulting from flowing wt Bfd solution of the indicated concentration over immobilized wt BfrB. (B) BfrB:Bfd binding affinity determined by steady-state affinity analysis. Responses at steady state are plotted as a function of Bfd concentration (black circles in top panel) and fitted to the model described by eq 4 (solid line). The bottom panel shows the relative differences between the fitted and experimental data. A Scatchard plot is shown in the inset of the top panel.

In the SPR experiments, BfrB was immobilized on the surface and free Bfd was in the flow. Hence, [BfrB_f] is the concentration of immobilized BfrB not bound to Bfd, [Bfd_f] is the concentration of Bfd in the flowing solution, which is maintained constant by the flow system, and [BfrB:Bfd] is the concentration of the protein complex at the surface. For a case where Bfd binding is homogeneous and the stoichiometry is 1:1, the expression for the dissociation constant (K_d) can be written as in eq 1,⁴⁰ where the total concentration of immobiled BfrB is given by [BfrB_t] = [BfrB_f] + [BfrB:Bfd].

1

Since the X-ray crystal structure showed that Bfd binds to multiple sites on BfrB, eq 1 was modified to eq 2, which indicates that n sites in BfrB can be occupied by Bfd, with corresponding dissociation constants.

2

The X-ray crystal structure of the BfrB:Bfd complex also showed that the Bfd molecules bind at 12 structurally identical binding sites on BfrB. Thus, assuming that the 12 Bfd binding sites on BfrB are identical and noninteracting, eq 2 can be reduced to eq 3,⁴⁰ which is similar to eq 1, except for the inclusion of the complex stoichiometry, n.

3

At each of the Bfd concentrations ([Bfd_f]) used in the SPR experiments shown in Figure 3A, when the system reaches steady-state equilibrium (at the plateau), the concentration of the BfrB:Bfd complex at the surface is proportional to the magnitude of the SPR response at the pleateau, which is termed R_eq. Hence, eq 3 can be rewritten as eq 4, where R₀, corresponds to the response when every immobilized 24-mer BfrB molecule is bound by one Bfd molecule, and the remainder terms are defined as above. The value of R₀ can be estimated from eq 5,⁴¹ where R_immo is the SPR response obtained upon immobilizing BfrB, MW_BfrB is the molecular mass of BfrB (452 672 Da), and MW_Bfd is the molecular mass of Bfd (7984 Da).

4

5

Setting initial values of n = 12 and R₀ = 520, respectively, and fitting the values of R_eq obtained from each [Bfd_f] (circles in the plot of Figure 3B) to eq 4, values for K_d = 3.3 ± 0.5 μM (Table 1), n = 12.7, and R₀ = 624 ± 18 were obtained. As can be seen from the fitted line in Figure 3B and from the small relative residuals, the binding curve is well-fitted by eq 4. In addition, the linear behavior observed from analysis of the binding curve by the Scatchard plot, which is a powerful tool for identifying deviations from simple binding models,⁴⁰ supports the idea implicit in eq 4, namely, that the 12 binding sites in BfrB are equivalent and independent.

Table 1. Thermodynamic and Kinetic Parameters for the BfrB:Bfd Interaction^a.

	Kd (μM)	ka (μM–1 s–1)	kd (s–1)	ΔH (kcal/mol)	ΔS (cal/(mol K))b	ΔG (kcal/mol)c
SPR equilibrium analysis	3.3 ± 0.5d
SPR kinetic analysis	2.6 ± 0.4d	0.026 ± 0.002	0.069 ± 0.01
ITC analysis	4.2 ± 0.4e			1.6 ± 0.3	30.1 ± 0.7	–7.3 ± 0.1

^aThe reported thermodynamic and kinetic parameters obtained from triplicate experiments at 22 °C.

^bCalculated from ΔG = ΔH – TΔS

^cCalculated from ΔG = −RT ln(1/K_d)

^dK_d obtained at pH 7.4

^eK_d obtained at pH 7.0

Having shown that the 12 binding sites on a BfrB molecule are identical and independent, kinetic analysis of the BfrB:Bfd interaction was carried out assuming the 1:1 binding model described by eqs 6 and 7, where BfrB_s indicates one of the 12 binding sites in each immobilized BfrB molecule, [BfrB_ts] is the total concentration of binding sites (12/BfrB molecule) on the surface, [BfrB_bs] is the concenration of BfrB sites bound to Bfd, k_a is the association rate constant, and k_d is the dissociation rate constant. Equation 7 can be rewritten in terms of the SPR response as eq 8, where R is the response caused by Bfd binding and R_max is the response when all the binding sites are occupied by Bfd.⁴² Hence, values of k_a and k_d (Table 1) were obtained from global fitting of the SPR data to eq 8 (Figure 4) with the aid of BIAevaluation software. Note that the ratio of the constants (k_a/k_d = 2.6 μM) is very similar to the value of K_d obtained from steady-state analysis described above, an observation that strengthens the conclusion that the Bfd binding sites on BfrB are independent and equivalent.

6

7

8

Figure 4.

Kinetic analysis of the BfrB:Bfd interaction. Top: Reference- and baseline-subtracted sensograms obtained from flowing wt Bfd solution of the indicated concentration over immobilized wt BfrB shown in different colors. Corresponding fitted traces are shown in black. Relative residuals are shown in the bottom panel.

Characterization of the BfrB:Bfd Interaction by ITC

The interaction between wt BfrB and wt Bfd was also studied by ITC at 22 °C. The top panel in Figure 5, which shows the binding isotherms obtained from titrating Bfd with BfrB, indicates that formation of the BfrB:Bfd complex is an endothermic process. The ITC data were integrated, normalized for nonspecific heat effects, and fitted using a nonlinear least-squares algorithm (minimization of χ²) with the aid of the Origin software package provided by MicroCal. Since the 12 binding sites on BfrB are equivalent and independent, the data were fitted to the 1:1 binding model described by eq 6, where the concentration of titrant (BfrB) is the concentration of Bfd-binding sites (subunit dimers). Fitting the integrated heats (bottom panel in Figure 5) allowed us to obtain the ΔH, ΔS, and K_d values listed in Table 1, which indicate that the BfrB:Bfd association is entropically driven. Note that the value of K_d obtained from these measurements is in good agreement with the K_d value obtained from the SPR experiments, thus providing additional support for the idea that the 12 Bfd-binding sites in BfrB are identical and independent. Identical experiments (SPR and ITC) were carried out with M⁰ BfrB (Figures S3 and S4). The results show that the K_d values obtained by SPR (K_d = 3.8 ± 0.2 μM) and ITC (K_d = 2.6 ± 1.0 μM) are within error, indistinguishable from those obtained with wt BfrB. Hence, the M⁰ residue has no influence on the BfrB:Bfd association. Moreover, as will be shown below, M⁰ has no influence on the BfrB structure.

Figure 5.

Analysis of the BfrB:Bfd interaction using ITC. Top: Heat absorbed upon titrating BfrB into Bfd. Bottom: Integrated heats plotted against the BfrB_s/Bfd mole ratio and fitted to the 1:1 binding model described by eq 6.

Dissecting the Interaction Surface in the BfrB:Bfd Complex

The structure of the BfrB:Bfd complex shows that L⁶⁸, E⁸¹, and E⁸⁵ in BfrB interact with M¹, Y², and L⁵ in Bfd (see Figure 2C), forming a contiguous set of interactions that is likely to contribute significantly to the stability of the complex. To study the relative contribution of these interactions to the formation of the complex, we prepared the L⁶⁸A, E⁸¹A, E⁸⁵A, L⁶⁸A/E⁸¹A mutants of BfrB and the Y²A, Y²F, L⁵A, and K⁴⁰A mutants of Bfd. It has been shown that in BfrB breathing motions of the protein shell link relatively remote sections in the structure and that, for example, mutations in some of the pores can affect the reactivity of relatively distant ferroxidase centers.⁹ The main objective of the mutations introduced in wt BfrB for this study is to determine their effect on the stability of the BfrB:Bfd interaction and concomitant inhibition of mobilization of iron stored in the BfrB core. Consequently, we characterized the BfrB mutants to determine that the site-directed mutations, which were designed to affect only the BfrB:Bfd interaction, did not cause significant changes in the structure of the protein or its ability to oxidize Fe²⁺ and store Fe³⁺. The biochemical characterization is described immediately below, and the structural analysis is presented later, after we have discussed the functional implications.

Biochemical and Functional Characterization of BfrB Mutants

The ability of the mutant proteins to assemble into 24-mers was analyzed with the aid of a calibrated size exclusion column. The results show that retention volumes exhibited by all of the mutants are nearly identical to the retention volume displayed by wt BfrB (Figure S5), indicating that all mutants assemble into 24-mers. Additionally, the UV–vis absorption spectrum displayed by each of the BfrB mutants is identical to that of wt BfrB and M⁰ BfrB (Figure S6): The heme in the oxidized form gives rise to a Soret band at 418 nm, whereas in its reduced form, the Soret band is at 425 nm with peaks in the visible at 527 and 567 nm. In wt BfrB, the ratio A₂₈₀/A₄₁₈ = 0.67 is indicative of a 24-mer protein harboring 12 heme molecules.²⁶ In the UV–vis spectra obtained from each of the BfrB mutants, the A₂₈₀/A₄₁₈ ratio is also ≈0.67, indicating that the mutants also harbor 12 hemes in a 24-mer assembly. This conclusion is also supported by analysis of heme content.

The Fe²⁺ oxidation and Fe³⁺ uptake function of each mutant were evaluated by reconstituting each of the proteins with 500 iron ions and monitoring the process by UV–vis spectrophotometry, as described previously.^26,27 The spectral data (Figure S7) obtained from titrating a solution of BfrB with aliquots delivering 50 Fe²⁺ ions/BfrB show a gradual increase in the absorption at 300 nm, which is associated with the growth of a Fe³⁺ mineral in the core of ferritin-like molecules.²⁶ After addition of 500 Fe ions/BfrB, the solution was incubated overnight at 4 °C and then passed through a Sephadex G-25 column. Subsequent determination of iron content in BfrB showed that each of the proteins captured >450 Fe³⁺ ions/BfrB. Having determined that the BfrB mutants assemble into fully functional 24-mers capable of oxidizing Fe²⁺ and storing Fe³⁺ in their core, the effect of the mutations on the ability of Bfd to bind BfrB mutants and promote the mobilization of core iron was examined.

Mobilization of Core Iron from the BfrB Mutants Is Compromised

We have previously demonstrated that mobilization of Fe³⁺ stored in BfrB requires Bfd^12,26 and that the process of iron mobilization (shown schematically in Figure 1C) involves Bfd binding to BfrB, which permits electron transfer from the [2Fe–2S] cluster in Bfd to the core Fe³⁺ mineral in BfrB via its heme. Consequently, to determine the effect of the BfrB mutations on the ability of Bfd to bind and promote the mobilization of core Fe³⁺, we reconstituted wt and each of the BfrB mutants with 500 ± 20 Fe³⁺ ions in their core (Figure S7). Mobilization of the iron core was monitored by UV–vis spectrophotometry, where the reaction mixture contained wt, BfrB, FPR, Bfd, and an excess of bipyridine (bipy), a Fe²⁺ chelator. Addition of excess NADPH initiates the reactions shown in Figure 1C, and the release of Fe²⁺ from BfrB is monitored by ΔA₅₂₃, which tracks the time-dependent formation of the [Fe(bipy)₃]²⁺ complex. The black circles in Figure 6A track ΔA₅₂₃ normalized to the total absorbance change expected upon removal of the 500 iron ions from wt BfrB core. In comparison, the reaction carried out in the absence of Bfd (red circles, Figure 6A) clearly demonstrates that there is no formation of [Fe(bipy)₃]²⁺ and, consequently, no mobilization of BfrB from the iron core. Thus, Bfd is necessary for the efficient and quantitative mobilization of Fe²⁺ from BfrB.

Figure 6.

Mobilization of Fe³⁺ stored in mutant BfrBs designed to block the BfrB:Bfd interaction is impaired. (A) Time-dependent increase in the normalized ΔA₅₂₃ upon addition of excess NADPH (final concentration, 1.0 mM) to 20 mM phosphate buffer (pH 7.6) containing 0.9 μM wt Bfd, 7.2 μM Fpr, and 0.18 μM of wt (black), E⁸¹A (green), L⁶⁸A (cyan), E⁸⁵A (orange), L⁶⁸A/E⁸¹A BfrB (blue) or no BfrB (red). (B) Time-dependent changes in the position of the heme Soret band in BfrB on addition of NADPH to the solutions in panel A; 418 and 425 nm correspond to fully oxidized and fully reduced heme, respectively.

Similar experiments conducted with the BfrB mutants show that the mutations significantly decrease the efficiency of iron mobilization: Replacement of Leu⁶⁸, Glu⁸¹, or Glu⁸⁵ for Ala causes significant decrease in the rate of iron release from BfrB, as shown by the cyan, green, and orange traces in Figure 6A. It is noteworthy that iron release from the Leu⁶⁸Ala/Glu⁸¹Ala double mutant is nearly completely abolished, as shown by the blue trace. To facilitate a more quantitative comparison of the effect of the mutations on the rates of iron mobilization, the initial rates of iron mobilization were obtained from fitting the initial, linear part of the curves (0 to ∼4 min) to a linear function (Table 2).

Table 2. Dissociation Constants for the BfrB:Bfd Interaction Involving Mutant BfrB and Bfd Molecules and Corresponding Initial Rates of Iron Release from the Core of BfrB.

protein	Kd (μM) (pH 7.4)	initial rate (min–1)
wild type	3.3 ± 0.5	0.099 ± 0.007
BfrB E81A	258.5 ± 21.5	0.031 ± 0.0003
BfrB L68A	298.5 ± 20.5	0.019 ± 0.0003
BfrB E85A	590 ± 2	0.018 ± 0.0006
BfrB L68A/E81A	not measurable	not measurable
Bfd Y2F	10.7 ± 0.2	0.097 ± 0.004
Bfd K40A	7.7 ± 0.5	0.090 ± 0.002
Bfd L5A	82 ± 2.5	0.044 ± 0.0005

The accompanying plots in Figure 6B track the time-dependent shift of the Soret band from 418 nm (oxidized heme) to 425 nm (reduced heme). In the case of wt BfrB, as the iron mobilization process progresses and the Fe³⁺ mineral core has been ≈80% mobilized, the heme is completely reduced, indicating that when the iron core is significantly diminished the flux of electron transfer from Bfd to heme in BfrB is faster than the flux of electron transfer from the heme to the Fe³⁺ core. When L⁶⁸A, E⁸¹A, or E⁸⁵A are utilized, reduced heme accumulates more gradually, and the corresponding ferric cores are mobilized more slowly than that in wt BfrB. In the case of the E⁸¹A/L⁶⁸A BfrB mutant, the Fe³⁺ core is not mobilized (Figure 6A), and the heme remains oxidized (Soret band remains at 418 nm). These observations are consistent with the idea that the interaction between L⁶⁸A/E⁸¹A BfrB and Bfd is blocked, which inhibits electron transfer and Fe³⁺ mobilization. Taken together, the observations summarized in Figure 6 indicate that Bfd and BfrB single mutants undergo interactions conducive to electron transfer less frequently than Bfd and wt BfrB and that interactions between L⁶⁸A/E⁸¹A BfrB and Bfd, if they occur at all, are not conducive to electron transfer.

The BfrB Mutants Have Significantly Reduced Affinity for Bfd

The findings described in the section above suggest that, relative to wt BfrB, the BfrB mutants form less stable complexes with Bfd. Experimental support for this idea was obtained by measuring the K_d for the associations between Bfd and each of the BfrB mutants using SPR, in a manner similar to that described above for the association between wt BfrB and Bfd. Figure 7A–D illustrates the reference- and baseline-subtracted responses obtained from flowing Bfd over immobilized L⁶⁸A, E⁸¹A, E⁸⁵A, and L⁶⁸A/E⁸¹A BfrB, respectively. For each of the mutants, the corresponding response reaches a plateau at each of the Bfd concentrations, indicating steady-state equilibrium. Plotting each response at steady state as a function of Bfd concentration produces the binding plots defined by the black (E⁸¹A), red (L⁶⁸A), and green (E⁸⁵A) circles in Figure 7E. Fitting the data to eq 4 produced the corresponding K_d values, which are listed in Table 2. The data shows that the association between Bfd and L⁶⁸A or E⁸¹A BfrB is ≈100 times weaker than the corresponding association with wt BfrB and that the E⁸⁵A mutation results in the lowest affinity for Bfd, which is ≈170 times lower than that measured for wt BfrB. Note that the reference- and baseline-subtracted responses obtained from flowing Bfd over immobilized L⁶⁸A/E⁸¹A BfrB (Figure 7D) indicate that the association between these two proteins is undetectable by SPR. The affinity of Bfd for BfrB decreases in the order wt ≫ E⁸¹A ≈ L⁶⁸A > E⁸⁵A ≫ L⁶⁸A/E⁸¹A (Table 2), which suggests that the efficiency of core iron mobilization might follow a similar order. The pseudo rate constants of iron mobilization, however, follow the order wt ≫ E⁸¹A > L⁶⁸A ≈ E⁸⁵A ≫ L⁶⁸A/E⁸¹A. The explanation for these seeming discrepancies probably resides in the fact that binding affinity is not the only determinant for electron transfer between the two proteins. Rather, interactions conducive to electron transfer must place the [2Fe–2S] cluster of Bfd in relative close (optimum) proximity to the heme in BfrB so that electron transfer is facilitated. Consequently, although the binding affinity of L⁶⁸A > E⁸⁵A, it is possible that E⁸⁵A interacts with Bfd in a manner conducive to electron transfer more frequently than does the L⁶⁸A mutant.

Figure 7.

Binding affinity of Bfd for mutant BfrBs is diminished. Overlay of reference- and baseline-subtracted sensograms resulting from flowing wt Bfd at the indicated concentrations over immobilized (A) L⁶⁸A, (B) E⁸¹A, (C) E⁸⁵A, and (D) L⁶⁸A/E⁸¹A BfrB. (E) Responses at steady state were plotted as a function of Bfd concentration and fitted to the model described by eq 4.

Structural Characterization of BfrB Mutants

The X-ray crystal structure of wt BfrB is nearly identical to the previously reported structure of M⁰ BfrB (PDB 3IS7); superposition⁴³ of subunit A of wt BfrB with subunit A of M⁰ BfrB resulted in 0.20 Å RMSD between C_α atoms for the 153 residues aligned using secondary structure matching.⁴⁴ Superposition of the wt BfrB structure with the structures of each of the BfrB mutants in this study shows that the overall organization of the mutants is nearly identical to that of the wt protein. The RMSD values (C_α atoms) obtained from superposing subunit A of wt BfrB with subunit A of L⁶⁸A, E⁸¹A, E⁸⁵A, and L⁶⁸A/E⁸¹A BfrB are 0.19, 0.25, 0.22, and 0.15 Å, respectively, for 156 residues aligned. Our previous work showed that in all structures of as-isolated M⁰ BfrB the ferroxidase centers are devoid of iron. Soaking crystals of as-isolated protein in crystallization solution containing FeCl₂, however, invariably resulted in iron binding at the ferroxidase centers. We also noted that iron binding is always accompanied by a conformation change of the ferroxidase center ligand H¹³⁰ side chain from a gate-open to gate-closed state.^9,27 Hence, it is not surprising that in this study soaking crystals of wt and mutant BfrBs in FeCl₂ solution resulted in iron binding to the ferroxidase centers and in conformation change of the H¹³⁰ side chain from gate open to gate closed (Figure S8). Interestingly, the ferroxidase iron ions in wt and L⁶⁸A/E⁸¹A BfrB are also bridged by an acetate ion obtained from the crystallization buffer. These observations, which indicate that the ferroxidase centers in the mutant BfrB molecules are functional in crystallo, are in good agreement with ferroxidase center competency in solution, as demonstrated in the above-described Fe²⁺ oxidation and Fe³⁺ uptake efficiency in solution (Figure S7). Consequently, it is clear that the mutations do not affect the iron oxidation function or the global structure of BfrB. As will be shown below, small structural changes observed in each of the BfrB mutants near the mutated residue are consistent with their lower binding affinity for Bfd.

The structure of the BfrB:Bfd complex showed that Bfd binding to BfrB is accompanied by relatively small rearrangements on the BfrB surface.¹² The most significant rearrangements are the reorientation of the L⁶⁸A, E⁸¹A, and E⁸⁵A side chains on BfrB, which enable hydrophobic and hydrophilic interactions with Bfd: rearrangement of the L⁶⁸ and E⁸¹ side chains narrows the cleft formed by the two side chains, which serves as a pocket to receive the Y² side chain from Bfd (Figure 8A). The new conformation of the L⁶⁸ side chain also contributes to the narrowing of the pocket where the L⁵ side chain from Bfd anchors on the BfrB surface. The side chain reorientations of E⁸¹ and E⁸⁵ enable H-bonding interactions between the backbone carbonyl oxygen of Y² in Bfd and the carboxylic group of E⁸¹, as well as the backbone N–H of M¹ in Bfd and the carboxylic group in E⁸¹ and E⁸⁵. These interactions, in turn, allow the side chain of M¹ in Bfd to pack hydrophobically against the BfrB surface. Close inspection of the Bfd-binding site in L⁶⁸A BfrB (Figure 8B) shows that the conformation of the E⁸¹ side chain remains as in the wt protein but that the shorter side chain of A⁶⁸ makes the cleft shallower on one side. This is expected to affect the extent to which the Bfd Y² side chain can bind and also affect the pocket where Bfd L⁵ anchors on the BfrB surface. The structure of the E⁸¹A mutant (Figure 8C) suggests that, in addition to a shallower more open cleft where the anchoring of Y² from Bfd is likely to be less efficient, a hydrogen bond between the NH of M¹ in Bfd and a carboxyl O in BfrB E⁸¹ is lost. The shorter A⁸⁵ side chain in the E⁸⁵A mutant (Figure 8D) allows the E⁸¹ side chain to move away from the cleft and therefore create a wider gap, which is probably unfavorable for efficient anchoring of the Bfd Y² side chain. In addition, the hydrogen bonds between NH of M¹ in Bfd and the carboxyl oxygens in BfrB E⁸⁵ are no longer possible, which, in turn, may have an unfavorable effect on how the side chain of M¹ in Bfd packs against the BfrB surface. In aggregate, the E⁸⁵A mutation eliminates more interactions than the L⁶⁸A or E⁸¹A mutation, which is in agreement with the observation that the K_d measured for the interaction between Bfd and E⁸⁵A BfrB is approximately 2-fold larger than the K_d values measured for the complexes formed between Bfd and L⁶⁸A BfrB or E⁸¹A BfrB (Table 2). Finally, the structure of the L⁶⁸A/E⁸¹A mutant (Figure 8E) shows a very wide cleft, where the Y² side chain in Bfd is unlikely to undergo any meaningful interactions. This is probably the reason that the association between Bfd and the BfrB double mutant is blocked and mobilization of core iron stored in L⁶⁸A/E⁸¹A BfrB is inhibited.

Figure 8.

Structural changes in the Bfd-binding sites of BfrB mutants. (A) View of the wt BfrB:Bfd interface, showing Bfd residues in cyan sticks and BfrB in surface representation, with subunit A in green and subunit B in gray. (B–E) Identical views of the Bfd-binding sites in the BfrB mutants, illustrating the structural changes in each of the BfrB mutants. O and N atoms are in red, and S and Fe atoms in the [2Fe–2S] cluster of Bfd are in yellow and orange, respectively.

Mutations in Bfd

The X-ray crystal structure of the BfrB:Bfd complex revealed, for the first time, the Bfd fold, which consists of a helix-turn-helix motif, where the four Cys ligands to iron in the [2Fe–2S] cluster are located in two hairpin loops (Figure 9). Note that iron ligands C⁴ and C⁶ are located in the same hairpin loop (cyan) that contains M¹, Y², and L⁵, whose side chains anchor Bfd at the BfrB surface. We prepared the Y²A, Y²F, L⁵A, and K⁴⁰A Bfd mutants to evaluate the effect of the mutations on the association with BfrB and their effectiveness at promoting the mobilization of core iron from BfrB. The L⁵A and K⁴⁰A mutants can be expressed and purified in yields comparable to those of wt Bfd. The mutants display UV–vis spectra nearly identical to that of the wt protein (Figure S9), indicating that they bind a [2Fe–2S] cluster. In contrast, Y²A Bfd is expressed in the E. coli host cells but does not incorporate a [2Fe–2S] cluster. Inspection of the wt Bfd structure shows that the Y² side chain contributes to stabilizing the hairpin loop containing C⁴ and C⁶ via packing interactions with G⁸ and T¹⁰ (Figure 9). It is therefore possible that the inability of the Y²A mutant to bind a [2Fe–2S] cluster is a consequence of conformational disorder in the loop containing iron ligands C⁴ and C⁶. Given that Y²A Bfd cannot bind a [2Fe–2S] cluster, we prepared the Y²F mutant in an attempt to study the role played by the phenolic OH in BfrB:Bfd association and in promoting the mobilization of core iron from BfrB.

Figure 9.

Structural organization of the hairpin loops harboring the iron ligands in the [2Fe–2S] cluster of Bfd (PDB 4E6K). The N′-terminus hairpin loop (cyan) contains residues Y² and L⁵ and iron ligands C⁴ and C⁶; packing interactions involving Y², G⁸, and T¹⁰ stabilize the hairpin and the [2Fe–2S] cluster. The hairpin loop containing iron ligands C³⁸ and C⁴¹ is shown in green; O, N, S, and Fe atoms are shown in red, blue, yellow, and orange, respectively.

The dissociation constant for the interaction between BfrB and each of the mutant Bfds was measured using SPR. The corresponding reference- and baseline-subtracted sensograms (Figures 10A and S10) show that each of the systems attains steady-state equilibrium, so K_d values (Table 2) were obtained from fitting the corresponding data to eq 4 (Figure 10B). The K_d obtained with Y²F Bfd is approximately 3 times larger than that measured with wt Bfd, which indicates that the phenol OH in Y² contributes only modestly to stabilizing the BfrB:Bfd complex. This observation is consistent with our analysis of the interface, which shows that anchoring of the Y² side chain engages mainly hydrophobic interactions with L⁶⁸ and E⁸¹ in BfrB (Figure 2C). The K_d measured with L⁵A Bfd, on the other hand, is ≈30-fold higher than that obtained with wt Bfd, a finding that is also consistent with our analysis of the interface, which suggests that packing of the L⁵ side chain on the BfrB surface is important for the stabilization of the BfrB:Bfd complex. The K_d obtained with K⁴⁰A Bfd is 2 times larger than the K_d obtained with wt BfrB. Analysis of the interface shows that the hydrophobic portion of the K⁴⁰ side chain interacts with the hydrophobic portion of BfrB K_A⁷⁶; in addition, K⁴⁰ forms hydrogen bonds with Q_B and L_A⁷⁴ in BfrB. Although it is possible that the A⁴⁰ side chain in K40A Bfd partially substitutes for the hydrophobic stabilization provided by the K⁴⁰ and BfrB K_A packing interaction, the modest increase in K_d observed with the K⁴⁰A mutant suggests that K⁴⁰ is less important to the stabilization of the BfrB:Bfd complex than Y² and L⁵. The relative stability of the complexes formed between BfrB and mutant Bfds (wt ≈ Y²F ≈ K⁴⁰A > L⁵A) is reflected in the ability of each of these mutants to promote the mobilization of core iron from BfrB (Figure 10C and Table 2). It is interesting to note that under the conditions of our iron mobilization assay an increase in K_d of approximately 3 times does not significantly affect the efficiency of iron mobilization. This may be due to the presence of a large excess of NADPH in the cuvette.

Figure 10.

Effect of mutations in Bfd on the binding affinity of the BfrB:Bfd complex and on iron mobilization from wt BfrB. (A) Overlay of reference- and baseline-subtracted sensograms resulting from flowing L⁵A Bfd solution of the indicated concentrations over immobilized wt BfrB. (B) Responses at steady state plotted as a function of the concentration of wt (blue), L⁵A (red), Y²F (black), and K⁴⁰A Bfd (green). (C) Time-dependent increase in the normalized ΔA₅₂₃ upon addition of excess NADPH (final concentration, 1.0 mM) to 20 mM phosphate buffer (pH 7.6) containing 0.18 μM wt BfrB, 7.2 μM Fpr, and 0.9 μM wt (blue), Y²F (black), L⁵A (red), or K⁴⁰A Bfd (green).

Concluding Remarks

In this study, we have dissected the BfrB:Bfd interface to validate the model obtained from the X-ray crystal structure and to gain a greater understanding of the interactions that contribute to the stability of this protein–protein complex. The results, which demonstrate that replacing key residues at the protein–protein interface for alanine lowers the strength of the association and impairs core iron mobilization, are in excellent agreement with the structural model, which places the [2Fe–2S] cluster in Bfd immediately above the heme in BfrB to facilitate electron transfer. Our findings also provide significant insight into factors that contribute to the stability of the BfrB:Bfd complex. In this context, it is important to note that although the number of structures of protein–protein complexes is relatively limited, analysis of protein–protein complexes has shown that not all residues participating at protein–protein interfaces contribute equally to the stability of the complex. Rather, there are critical residues that contribute to the majority of the binding energy. Thus, these have been termed hot spot residues. Hot spot residues can be identified by alanine-scanning mutagenesis, if their substitution by alanine leads to a binding energy difference of at least 2 kcal/mol (ΔΔG ≥ 2 kcal/mol).⁴⁵⁻⁴⁷ Hot spot residues tend to occur in clusters, which often form a network of interactions constituting hot regions; importantly, such a hot region includes residues from both proteins.⁴⁸ Our results, which show that replacing L_B⁶⁸, E_A, and E_A⁸⁵ in BfrB and L⁵ in Bfd for alanine results in ΔΔG ≥ 2 kcal/mol, clearly identify these as hot spot residues (Figure 11A). A zoomed-in view of the BfrB:Bfd interface (Figure 11B) illustrates how these critical residues form a clustered network of packing interactions that define a hot region. The figure also makes it evident that Y² in Bfd is probably an integral component of the hot region. As pointed out above, the Y²A Bfd mutant cannot bind a [2Fe–2S] cluster and therefore could not be studied. Nevertheless, the structure suggests that hydrophobic packing of its aromatic side chain with the L_B side chain and the hydrophobic portion of the E_A⁸¹ side chain in BfrB dominate the contribution of Y² to the stability of the complex. Identification of the hot spot residues in both proteins, as well as visualization of the hot region, makes it apparent that a large portion of the hot spot interactions is eliminated from the hot region by replacing E_A and E_A⁸⁵ in BfrB for alanine, which should significantly decrease the binding affinity, a prediction that is reflected in the observation that Bfd cannot bind to L⁶⁸A/E⁸¹A BfrB. Finally, it is noteworthy that all of the hot spot residues identified in this investigation are conserved in the Bfr and Bfd sequences (Figure S11) of a large number of Gram-negative pathogens, where the bfr and bfd genes are also contiguous in the corresponding chromosomes (Table 3). Consequently, conservation in the arrangement of these genes in the chromosomes of many pathogens and conservation of hot spot residues in both proteins indicate that the hot region that stabilizes the BfrB:Bfd association in P. aeruginosa is likely to be of widespread significance in bacterial iron metabolism.

Figure 11.

Identification of hot spot residues at the interface of the BfrB:Bfd complex. (A) Change in binding free energy (ΔΔG) brought by mutations in BfrB (green for subunit A; gray for subunit B) or in Bfd (cyan). (B) Zoomed-in view of the BfrB:Bfd interface depicting how hot spot residues network into a hot region; O is in red, and N is in blue.

Table 3. Pathogens Where Adjacent bfd and bfr Genes Code Conserved Hot Spot Residues in the Corresponding Bfr and Bfd Sequences.

	locus ID
microorganism	bfd	bfr
Pseudomonas aeruginosa PAO1	PA3530	PA3531
Escherichia coli O157	Z4696	Z4695
Escherichia coli K12	b3337	b3336
Salmonella typhimurium	STM3444	STM3443
Salmonella enterica serovar Typhi	t4061	t4062
Salmonella enterica choleraesuis	SCH_3377	SCH_3378
Klebsiella pneumonia	KPN_03722	KPN_03723
Yersinia pestis	YPO0205	YPO0206
Yersinia pseudotuberculosis	YPTB3701	YPTB3700
Yersinia enterocolitica	YE3926	YE3925
Aeromonas salmonicida	ASA_0466	ASA_0467
Vibrio cholerae El	VC_0364	VC_0365
Vibrio cholerae	VC0395_A2775	VC0395_A2776
Vibrio vulnificus	VV1_1340	VV1_1341
Shigella flexneri 2a 2457T	S4406	S4407
Shigella flexneri 2a str301	AAN44819.1	AAN44818.1
Shigella dysenteriae	SDY_3499	SDY_3498
Erwinia carotovora	ECA4033	ECA4034
Enterobacter cloacae	ECL_04701	ECL_04700
Serratia sp.	SerAS12_4646	SerAS12_4645
Citrobacter koseri	CKO_04744	CKO_04743
Acinetobacter sp.	ACIAD3328	ACIAD3330

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.biochem.5b00937.

• Arsenic ion bound to the 4-fold pores of L⁶⁸A BfrB (Figure S1); Mg ion in the B-pore of L⁶⁸A BfrB (Figure S2); ITC and SPR analyses of the interaction between Bfd and M⁰ BfrB (Figures S3 and S4); elution volumes of wt and mutant BfrBs from a calibrated size exclusion column (Figure S5); oxidized and reduced spectra of wt and mutant BfrBs (Figure S6); UV–vis spectra obtained from iron uptake of wt and mutant BfrBs (Figure S7); superposition of the ferroxidase centers of native and iron bound wt and L⁶⁸A/E⁸¹A BfrB (Figure S8); UV–vis spectra of wt and mutant Bfds (Figure S9); overlay of double referenced sensorgrams resulting from injecting Bfd Y²F and Bfd K⁴⁰A over immobilized BfrB (Figure S10); sequence alignment of bfd and bfrb genes from different pathogens (Figure S11); sequence of primer pairs used to introduce the sited-directed mutations; crystallographic data for wt and mutant BfrB structures (Table S1) (PDF).

Accession Codes

Coordinates and crystallographic structure factors for the distinct BfrB structures have been deposited in the Protein Data Bank under accession codes listed in Table S1 of the Supporting Information.

This study was supported by a grant from the National Science Foundation (M.R., MCB-1158469). A 2014 University of Kansas Strategic Grant, Level I, also supported this work.

The authors declare no competing financial interest.