The IsdC Protein from Staphylococcus aureus Uses a Flexible Binding Pocket to Capture Heme

Abstract

Staphylococcus aureus scavenges heme-iron from host hemoproteins using iron-regulated surface determinant (Isd) proteins. IsdC is the central conduit through which heme is passed across the cell wall and binds this molecule using a NEAr Transporter (NEAT) domain. NMR spectroscopy was used to determine the structure of IsdC in complex with a heme analog, zinc-substituted protoporphyrin IX (ZnPPIX). The backbone coordinates of the ensemble of conformers representing the structure exhibit a root mean square deviation to the mean structure of 0.53 ± 0.11Å. IsdC partially buries protoporphyrin within a large hydrophobic pocket that is located at the end of itsβ-barrel structure. The central metal ion of the analog adopts a pentacoordinate geometry in which a highly conserved tyrosine residue serves as a proximal ligand. Consistent with the structure and its role in heme transfer across the cell wall, we show that IsdC weakly binds heme (K_D = 0.34 ± 0.12 μm) and that ZnPPIX rapidly dissociates from the protein at a rate of 126 ± 30 s^-1. NMR studies of the apo-form of IsdC reveal that a 3₁₀ helix within the binding pocket undergoes a flexible to rigid transition as heme is captured. This structural plasticity may increase the efficiency of heme transfer across the cell wall by facilitating protein-protein interactions between apoIsdC and upstream hemoproteins.

Staphylococcus aureus is an opportunistic Gram-positive pathogen that causes lethal infections such as toxic shock syndrome, meningitis, and endocarditis (1, 2). The bacterium needs the essential nutrient iron to grow and although the human body contains large quantities of this metal, little is directly available to S. aureus as it is sequestered intracellularly (3) or bound to transferrin and lactoferrin (4, 5). During infections, S. aureus procures iron from heme (protoporphyrin IX + iron), which contains ∼80% of the total iron in the body (6). Heme-loaded hemoglobin (Hb)5 is released into the blood plasma by the action of microbial hemolysins that rupture erythrocytes (7). A group of newly discovered proteins called iron-regulated surface determinant (Isd) proteins then scavenge heme and transfer it into the cytoplasm where it is degraded to liberate iron (8, 9). The heme-binding IsdC protein plays an essential role in the transfer of heme across the cell wall peptidylglycan (6, 10). Proteins homologous to IsdC are also present in a number of other important human pathogens (Bacillus anthracis and Listeria monocytogenes) (6, 11). Therefore, compounds that inhibit their ability to capture heme may be useful antibiotics.

In Gram-negative bacteria, the process of heme-iron acquisition is reasonably well understood. Heme is captured from hemoproteins or hemophore-hemoprotein complexes by specific outer membrane receptors (5). It is then transferred into the periplasm in a Ton-B-dependent manner, where it is moved across the inner membrane by specific ABC-dependent permeases (5). Heme acquisition mechanisms used by Gram-positive bacteria are only beginning to be understood. In several species, heme is directly imported into the cytoplasm by ABC transporters positioned in the membrane (Corynebacterium diphtheriae (HemTUV), Streptococcus pyogenes (HtsABC), and S. aureus (HtsABC)) (12–14). To more effectively capture heme, some pathogens supplement these transporters with cell wall-associated proteins that bind host hemoproteins (5). Two such systems have been characterized in detail. In S. pyogenes, the cell wall-associated hemoprotein Shp passes heme to the lipoprotein component of an ABC transporter through direct protein-protein interactions (13, 15). S. aureus uses a more elaborate system to acquire heme in which an array of different Isd proteins bind heme, hemoglobin, and the haptoglobin-hemoglobin complex (8). They then pass heme across the cell wall and membrane into the cytoplasm where it is degraded to release iron (8, 9, 16).

The S. aureus Isd determinant system consists of nine proteins that mediate the delivery of heme-bound iron into the cytoplasm (17). Three Isd proteins are attached to the cell wall by the SrtA sortase: IsdA, IsdB, and IsdH (8). These receptors bind to other proteins and heme. IsdH and IsdB presumably capture heme-laden Hb released from erythrocytes after lysis by hemolysins (8). The IsdH protein binds Hb, the serum glycoprotein haptoglobin (Hp), and the Hb·Hp complex (18, 19). Similar to IsdH, the IsdB protein binds Hb, but is incapable of binding Hp (20). IsdA may function in bacterial adhesion as it interacts with a range of extracellular matrix proteins (21–23). In contrast to the protein/heme receptors, IsdC is embedded within the cell wall by the SrtB sortase (10, 24). It is thought to function as the central conduit through which heme is passed across the cell wall to the membrane IsdDEF complex, which pumps heme across the membrane (8). Once inside the cell, iron is liberated from heme by the IsdG monoxygenase, or its paralog, IsdI (16, 25). Recent studies have also shown that captured heme can be directly incorporated into microbial proteins (14).

The cell wall receptors of the Isd system (IsdA, IsdB, IsdC, and IsdH) contain one or more NEAT domains (NEAr Transporter domains), a ∼125 residue motif found in proteins encoded by genes that are frequently proximal to Fe³⁺ siderophore transporter genes (26). Amazingly, NEAT domains exhibit a wide-range of ligand binding specificities. For example, some domains only bind proteins (e.g. the first and second NEAT domains within IsdH) (19), some may only bind heme (e.g. third NEAT domain from IsdH or IsdC),6 whereas others may bind both of these ligands (e.g. IsdA) (14, 18, 21, 28). Recent NMR and x-ray crystallography studies have revealed that NEAT domains adopt an immunoglobin-like fold, although they share no significant primary sequence homology with members of this family (18, 28, 29). To gain insight into how IsdC transports heme across the cell wall, we determined the three-dimensional structure and dynamics of its NEAT domain (IsdC^N) in complex with zinc-substituted protoporphyrin IX (ZnPPIX). We show that IsdC^N recognizes ZnPPIX through a large cleft located at the end of its β-barrel structure in which the zinc ion is bound in a pentacoordinate manner by a highly conserved tyrosine residue. In contrast to previously characterized NEAT domains, NMR measurements indicate that binding occurs through an induced-fit mechanism. This structural plasticity may increase the efficiency of heme transfer by facilitating interactions with upstream hemoproteins.

EXPERIMENTAL PROCEDURES

Protein Purification and NMR Sample Preparation—The gene sequence encoding the IsdC NEAT (IsdC^N) domain was cloned into pET15b (Novagen) using standard methods. This generated an expression plasmid that produces the NEAT domain within IsdC (residues Ser²⁵ to Gly¹⁵⁰ containing the additional sequence MGSSHHHHHHSSGLVPRGSHM at its N terminus). Isotopically labeled protein was produced from Escherichia coli BL21(DE3) cells grown at 37 °C. The cell culture consisted of M9 minimal media that was supplemented with either ¹⁵NH₄Cl, or both ¹⁵NH₄Cl and [¹³C₆]glucose as previously described (30). When the cells reached a density of ∼0.6 A₆₀₀ units they were induced by the addition of 1 mm isopropyl β-d-thiogalactoside. Cells were then harvested after 4 h by centrifugation at 8,500 × g for 20 min at 4 °C in a JA-10 rotor. The pellet was then resuspended in lysis buffer (50 mm NaPO₄, 300 mm NaCl (pH 7), 1 mm phenylmethanesulfonyl fluoride, and 1 mm protease inhibitor mixture II (Calbiochem)) and lysed by sonication. The lysed sample was then centrifuged at 11,000 × g for 30 min at 4 °C in a JA-20 rotor. The soluble fraction was then incubated with Talon beads (Clontech) for 45 min. After washing the column, the protein was eluted in the following buffer: 30 mm sodium phosphate, 300 mm NaCl, 150 mm imidazole (pH 7). Fractions containing protein were then pooled and dialyzed against 50 mm sodium acetate, 5 mm EDTA, and 1 m urea (pH 5.1). The sample was then applied to a SP-Sepharose Fast Flow XK-50 column (Amersham Biosciences) equilibrated with SP-buffer (50 mm sodium acetate, 5 mm EDTA, 1 m urea, pH 5.1). IsdC^N was then eluted using a 0–1 m gradient of NaCl dissolved in SP-buffer. Peaks containing IsdC^N were pooled and dialyzed against NMR buffer (50 mm sodium phosphate, 100 mm NaCl, 0.01% sodium azide (pH 6)). The results of a pyridine hemochrome assay indicate that only ∼3% of purified IsdC^N contained heme (31). This form of the protein is hereafter referred to as apoIsdC^N.

Complexes containing IsdC^N bound to ZnPPIX were studied by NMR. The complexes contain a 1:1 ratio of IsdC^N and ZnPPIX (IsdC^N·ZnPPIX complex). Complexes containing either ¹⁵Nor ¹⁵N- and ¹³C-labeled IsdC^N dissolved in NMR buffer were prepared by adding small aliquots of a stock solution of ZnPPIX (50 mm ZnPPIX dissolved in 0.1 m NaOH) to a solution of apoIsdC^N. The pH of the solution after each addition of ZnPPIX was adjusted and the progress of complex formation was monitored by recording a series of ¹H-¹⁵N HSQC spectra. All samples containing ZnPPIX were shielded from light to prevent degradation. Concentrated samples for NMR studies were produced using a centriprep filter device (Amicon; YM-10). The NMR samples contained 1.1 mm of the IsdC^N·ZnPPIX complex. One sample was dissolved in H₂O and NMR buffer, and contained 7% D₂O. A 1:1 IsdC^N:ZnPPIX dissolved in D₂O was also produced by freeze-drying the complex dissolved in NMR buffer, followed by resuspension with an equal volume of 99.999% D₂O.

NMR Spectroscopy Experiments and Structure Determination—NMR spectra were acquired at 302 K on CryoProbe-equipped Bruker Avance 500, 600, and 800 MHz spectrometers. Unless referenced individually, NMR experiments are described in Refs. 32 or 33. Protein resonances (¹H, ¹⁵N, ¹³C) were assigned by analyzing two-dimensional ¹H-¹⁵N HSQC and three-dimensional-CBCA(CO)NH, HNCACB, HNCA, HN(CA)CO, HNCO, HBHACONH (32), CC(CO)NH (34, 35), HNHB (36), HNHA, HCCH-total correlation spectroscopy (TOCSY), HCCH-COSY, ¹⁵N-edited NOESY, and ¹⁵N-edited TOCSY spectra. Two-dimensional (Hβ)Cβ(CγCδCε)Hε and two-dimensional (Hβ)Cβ(CγCδ)Hδ experiments were used to facilitate aromatic side chain assignments (37). The program TALOS was used to obtain φ and Ψ dihedral angle restraints (38). ³J_HNa coupling constants were measured using a three-dimensional HNHA experiment. Stereospecific assignment of methylene protons were obtained by analyzing three-dimensional ¹⁵N-edited ROESY, HNHB, and ¹⁵N-edited TOCSY (35 ms mixing time) spectra. Distance restraints were identified in three-dimensional ¹⁵N- and ¹³C-edited NOESY spectra of the complex. A two-dimensional (F1 and F2 ¹³C filtered) NOESY (39) spectra was used to assign the proton chemical shifts of the ZnPPIX molecule in the complex. Intermolecular NOEs between IsdC and ZnPPIX were identified by analyzing three-dimensional (F1) ¹³C,¹⁵N-filtered, (F2) ¹³C-edited NOESY, three-dimensional ¹³C-edited NOESY, and two-dimensional (F2 ¹³C-filtered) NOESY (39) spectra of the complex. NMR data were processed using NMRPipe (40), and analyzed using the CARA (version 1.4.1) (41) and PIPP (42) software packages.

Structure calculations were performed using the programs ATNOS/CANDID (43, 44) and XPLOR-NIH (45). Initially, only the structure of IsdC^N in the complex was determined using ATNOS/CANDID. Input into this program included the amino acid sequence, chemical shifts obtained from sequence-specific assignments, and three NOESY experiments: an ¹⁵N-edited NOESY (mixing time 125 ms), an aliphatic ¹³C-edited NOESY (mixing time 125 ms, ¹³C carrier at 35 ppm), and ¹³C-edited NOESY spectrum optimized for aromatic residues (mixing time 120 ms, ¹³C carrier at 100 ppm). Dihedral angle restraints determined by TALOS were also included at the start of the calculations. The boundaries for these restraints were set to ±30° or three times the standard deviation reported by TALOS, whichever value was larger. The standard CANDID protocol of seven cycles of peak identification and assignment, followed by structure calculations, was executed. This process assigned the vast majority of cross-peaks in the NOESY data and produced a set of conformers that converged to the structure of the IsdC^N protein in the complex. Subsequently, the NMR data were manually inspected to identify additional NOE distance restraints, including a total of 31 intermolecular distances. Hydrogen bond restraints were derived from analysis of NOE patterns from secondary structure analysis. In calculations carried out at the final stages of refinement, the side chain of Tyr¹³² chain was invariably proximal to the central zinc ion of the protoporphyrin ring, however, the distance between the hydroxyl oxygen and metal was slightly larger than predicted from high resolution crystal structures of other protein-protoporphyrin complexes. We therefore elected to include a single artificial distance restraint that holds the hydroxyl oxygen and the zinc atoms within 2.2 Å of one another. No significant structural changes are produced with the addition of this restraint and structures calculated using this restraint completely satisfied all of the experimental NMR data. The final structure of the complex was calculated using these additional restraints using the program XPLOR-NIH. A total of 200 conformers were produced. The structure of the complex is represented by 30 conformers that have the lowest overall energy. Statistics for the structure of the IsdC^N·ZnPPIX complex are presented in Table 1. Figures were prepared using the programs MOLMOL or PyMOL (46, 47).

TABLE 1.

Structural statistics of the solution structure of IsdC^N:ZnPPIX The notation of the NMR structures is as follows: <SA> are the final 30 simulated annealing structures; is the average energy minimized structure. The number of terms for each restraint is given in parentheses.

	<SA>
Intramolecular
R.m.s. deviations from NOE restraints (Å)a
All (1449)
Sequential [|i – j = 1] (341)	0.011 ± 0.003	0.024
Medium range [|i – j| ≤ 4] (89)	0.031 ± 0.006	0.057
Long range [|i – j| ≥ 5] (519)	0.020 ± 0.002	0.053
Intra-residue (500)	0.005 ± 0.002	0.018
R.m.s. deviations from hydrogen bonding restraints (Å)b (40)	0.022 ± 0.003	0.043
R.m.s. deviations from dihedral angles restraints (°)c (207)	0.415 ± 0.058	0.620
R.m.s. deviations from 3JHNα coupling constants (Hz) (71)	0.234 ± 0.019	0.727
Intermolecular
R.m.s. deviations from NOE restraints (Å)c Intermolecular (31)	0.038 ± 0.006	0.064
Deviation from idealized covalent geometry
Bonds (Å)	0.002 ± 0.00003	0.005
Angles (°)	0.963 ± 0.003	0.543
Impropers (°)	0.356 ± 0.0404	0.452
PROCHECK results (%)d
Most favorable region	68.0 ± 2.6	79.8
Additionally allowed region	29.8 ± 2.7	17.3
Generously allowed region	2.4 ± 0.09	1.9
Disallowed region	0.0 ± 0.0	1.0
Coordinate precision (Å)e
Protein backbone	0.53 ± 0.11
Protein heavy atoms	0.98 ± 0.11

^aNone of the structures exhibited distance violations greater than 0.5 Å, dihedral angle violations greater than 5°, coupling constant violations greater than 2 Hz

^bTwo distance restraints were employed for each hydrogen bond (r_NH···O < 2.3 Å and r_N···O < 3.3 Å)

^cThe experimental dihedral angle restraints were as follows: 98 φ, 87 ψ, and 22 X₁ angular restraints

^dDetermined using the program PROCHECK (74)

^eThe coordinate precision is defined as the average atomic root mean square deviation (r.m.s) of the 30 individual SA structures and their mean coordinates. These values are for residues Asp³⁰ to Gly¹⁴⁴ of IsdC^N. Backbone atoms are N, Cα, and C′

¹⁵N Relaxation Measurements and Analysis—The relaxation data were collected using the ¹⁵N-labeled sample of the IsdC^N·ZnPPIX complex acquired on a Bruker Avance 600-MHz spectrometer equipped with 5-mm single axis pulsed field gradient at room temperature probe. The ¹⁵N spin-spin/transverse rate constant (R₂), longitudinal rate constant (R₁), and steady-state ¹⁵N heteronuclear NOE values were measured as previously described (48–50). Ten R₂ two-dimensional experiments were performed in random order, with relaxation delays of 17 (duplicate), 35, 52 (duplicate), 69, 86, 104, 121, and 138 ms. Ten R₁ two-dimensional experiments were performed in random order, with relaxation delays of 42 (duplicate), 167, 335, 544 (duplicate), 816, 1172, 1591, and 2428 ms. The heteronuclear NOE experiment was carried out in an interleaved manner, with and without proton saturation and repeated three times. All experiments were acquired with 2048 × 256 complex points in the F2 and F1 dimensions with corresponding spectral widths of 10,000 and 1,886 Hz. The proton carrier frequency was set to the water resonance. Peak heights were determined using Sparky (51). Methods used to extract the relaxation parameters and estimate errors have been described previously (52). The ¹H-¹⁵N HSQC spectra of the IsdC^N·ZnPPIX complex are well resolved, enabling the reliable measurement of relaxation parameters for 94 residues, of a total of 123 (prolines excluded). The average R₁, R₂, and ¹⁵N (53) NOE parameters for the complex are, respectively, 1.59 ± 0.17 s^-1, 10.18 ± 1.80 s^-1, and 0.86 ± 0.11. Similar methods were used to analyze the ¹H-¹⁵N relaxation parameters for apoIsdC^N. The average R₂ and ¹⁵N{¹H} NOE parameters for the apoIsdC^N were 10.76 ± 0.27 s^-1 and 0.86 ± 0.03^-1, respectively.

The relaxation data were analyzed using programs kindly provided by Prof. Arthur G. Palmer III at Columbia University. Multiple approaches were used to assess the overall motion of the complex (54, 55). The principal moments of the inertia tensor were first calculated using the program Pdbinertia and yielded values of 1.00:0.90:0.20. As the relative moments vary significantly from a perfect sphere, the statistical significance of fitting the relaxation data to either axially symmetric or isotropic models of tumbling were explored using the R2R1 Diffusion program (54). These calculations were performed using a trimmed data set that included only those residues with R₂/R₁ ratios within 1 S.D. from the average R₂/R₁ ratio, and residues that had NOE ratios >0.65 (56). This analysis revealed axial symmetric diffusion and yielded a molecular correlation time (t_m) of 8.1 ns. No significant improvement was observed when more complex models of motion were tested. The tensor parameters were also calculated using the program Quadric Diffusion (57–59). This yielded a correlation time of 7.5 ns, but again showed a preference for the axial symmetric model. The results from this final calculation were used as an initial guess for the model-free analysis described below.

The amplitudes and effective correlation times of internal motions of the backbone amide groups were extracted using the Model-free formalism (60, 61). In this analysis, the internal motion of the NH vector is assumed to occur on two different time scales, fast and slow, which are characterized, respectively, by the square of the order parameters, S²_f and S²_s, and their effective correlation times, τ_f and τ_s (where τ_f << τ_s << τ_m). The square of the generalized order parameters is defined as S² = S²_fS²_s and corresponds to the spatial restriction of the NH bond vector (where 0 ≤ S² ≤ 1). The analysis also accounts for line broadening due to chemical exchange, R_ex. All these motional parameters were fit to the spin relaxation data using the program Model-free 4.01 (62, 63). Of 94 quantifiable residues in the complex, 73 could be satisfactorily fit using Model-free analysis. Model 1 (S²-only) was an appropriate fit for 65 residues, 2 residues fit to model 2 (S² and t_e), 2 residues fit to model 3 (S² and R_ex), 2 fit to model 4 (S², t_e, and R_ex), and 2 residues fit to model 5 (S²_f, S²_s, and t_e). If only residues located in regions of regular secondary structure are considered, the average order parameter is 0.93 ± 0.03.

Measurements of ZnPPIX and Heme Binding—The intensity of the Soret peak at 419 nm in the UV-visible spectrum of the IsdC^N·ZnPPIX complex was used to monitor complex formation. Aliquots of free ZnPPIX (dissolved in 0.1 m NaOH) were added to a 1 μm sample of apoIsdC dissolved in 50 mm Tris (pH 7.5). After each addition, the pH was adjusted, equilibrated for 5 min, and the UV absorbance measured. A total of seven ZnPPIX additions (0.2 to 10.0 μm) were made to generate the binding isotherm. The dissociation constant (K_D) of ZnPPIX binding was determined by non-linear least squares fitting the data to the following equation,

(Eq. 1)

RESULTS

IsdC^N Binds Heme and Zinc-protoporphyrin IX—The IsdC protein is the central conduit through which heme is passed across the cell wall from surface-exposed heme receptors to the underlying transmembrane IsdDEF complex. It captures heme via its NEAT domain (IsdC^N, residues Ser²⁵ to Gly¹⁵⁰). Although heme binding is its only known function, the affinity of IsdC^N for this molecule has not been measured. We therefore used an absorption spectrophotometry assay to determine the dissociation constant (K_D) of binding. IsdC^N was incubated with increasing amounts of hemin and the amount of IsdC^N·heme complex formed was determined by measuring the UV absorbance at 412 nm. Fig. 1a shows a plot of this data as a function of hemin added. Curve fitting, assuming a 1:1 binding stoichiometry, yields a K_D of 0.34 ± 0.12 μm. The rather weak binding affinity is consistent with its function, as IsdC^N must be able to readily capture and release heme as it transits the cell wall. Recently, the B. anthracis IsdC protein was also shown to bind heme weakly (K_D = 3.10 ± 0.42 μm), consistent with the two proteins having similar functions (6).

FIGURE 1.

IsdC^N forms a complex with protoporphyrin IX (ZnPPIX). a, absorption spectrophotometry assay of IsdC^N binding to heme. The plot shows the intensity of the Soret band at 412 nm of the IsdC^N·heme complex after the addition of varying amounts of hemin. Fitting of the binding isotherm yielded a K_D value of 0.34 ± 0.12 μm. b, absorption assay of ZnPPIX binding to IsdC^N. The plot shows the absorbance of the IsdC^N·ZnPPIX complex as a function of ZnPPIX added. In this assay absorbance was monitored at 419 nm, the Soret band of the IsdC^N·ZnPPIX complex. Fitting of the binding isotherm yielded a K_D value of 3.14 ± 0.32 μm. c, strips of ¹H-¹⁵N HSQC spectra are shown depicting the formation of the complex, IsdC^N·ZnPPIX. The left, middle, and right panels show apoIsdC^N to the conversion to the 50% complex, and to holoIsdC^N, respectively. For clarity, only one residue (Ser¹⁰²) is shown. Analysis of this data enabled the rate of protoporphyrin release to be determined. d, representative NMR spectra of the IsdC^N·ZnPPIX complex showing intermolecular NOEs between IsdC^N and ZnPPIX. The left two panels show selected regions of the two-dimensional [F2]¹³C-filtered NOESY spectrum. The right-most panels show selected regions of the three-dimensional [F1-¹³C-filtered]¹³C-edited NOESY spectrum. Intermolecular NOEs are labeled and intramolecular NOEs are unlabeled. A total of 31 intermolecular NOEs were identified in the NMR data.

NMR Solution Structure of the IsdC^N·ZnPPIX Complex—To understand the mechanism of heme capture we used multidimensional NMR spectroscopy and simulated annealing calculations to solve the structure of a 1:1 complex between IsdC^N and ZnPPIX. ZnPPIX was used because it is structurally identical to heme, but lacks a paramagnetic iron center that could complicate and/or diminish the quality of the NMR data by causing line broadening and hyperfine shifts (65). As shown in Fig. 1b, IsdC^N binds ZnPPIX with a K_D of 3.14 ± 0.32 μm. The diamagnetic complex is in slow exchange on the NMR chemical shift time scale as the ¹H-¹⁵N HSQC spectrum of a complex containing substoichiometric amounts of ZnPPIX exhibits cross-peaks that correspond to both the free and bound forms of the protein (Fig. 1c).

The structure of the complex was calculated using: 1,449 intra-protein distances, 40 hydrogen bonds, 71 ³J_HNα couplings, and 207 dihedral angles. In addition, 31 intermolecular NOE distance restraints between the protein and ZnPPIX were included that were identified in two-dimensional F₂-filtered and three-dimensional F₁-filtered ¹³C-edited NOESY spectra of the complex (Fig. 1d). Structure calculations produced an ensemble of 30 conformers that possess good covalent geometry and no NOE, dihedral angle, or scalar coupling violations greater than 0.5 Å, 5°, or 2 Hz, respectively (Fig. 2a). The protein is structured from residues Asp³⁰ to Gly¹⁴⁴, which have backbone and heavy atom coordinate root mean square deviations to the mean structure of 0.53 ± 0.11 and 0.98 ± 0.11 Å, respectively. Complete structure and restraint statistics are summarized in Table 1.

FIGURE 2.

NMR solution structure of the IsdC^N·ZnPPIX complex. a, a cross-eyed stereoview showing the ensemble of 30 lowest energy structures of the IsdC^N·ZnPPIX complex. The backbone atoms of residues Ser²⁵ to Gly¹⁵⁰ are shown in blue and the heavy atoms of ZnPPIX are colored red. b, ribbon drawing of the structure of the complex. Two views are shown that are related by a 180° rotation. The strands in the β-sheet are indicated by arrows and labeled. The view on the left is identical to that shown in panel a. c, electrostatic surface of IsdC^N bound to ZnPPIX. The views are similar to panel a. Positively and negatively charged residues are colored blue and red, respectively. Neutral or hydrophobic residues are colored white. ZnPPIX (colored yellow) is shown surrounded by hydrophobic residues.

IsdC^N captures protoporphyrin through a large cleft located at the top of its β-barrel structure (Fig. 2b). The barrel is formed by two multistranded anti-parallel β-sheets that pack against one another. One face of the structure is formed by a five-stranded β-sheet (β1a, β2, β3, β5, and β6), whereas the other face contains four β-strands (β1b, β8, β7, and β4). The fold is initiated by residues in strand β1, which can be divided into two discrete segments, strands β1a (Ser³¹–Leu³⁴) and β1b (Asn³⁵–Tyr³⁹). These segments bridge distinct faces of the protein, with strand β1a pairing with strand β2 (Ala⁵⁷–Lys⁶¹) on one face, whereas strand β1b pairs with strand β8 (Pro¹²⁹–Asn¹⁴³) on the other face. After strand β1b, the chain forms a short 3₁₀ helix (Ala⁴⁹–Asp⁵¹) that contacts ZnPPIX before descending to form strand β2. A short turn then reverses the chain to initiate strand β3 (Leu⁶⁶–Asn⁷³), which pairs with strand β2. After crossing over to the opposing side of the barrel, the chain forms strand β4 (Ile⁷⁸–Ile⁸³), which pairs with residues in strand β7 (Asn¹¹⁰–Lys¹²⁴). It then crosses back over again to initiate strands β5 (Asn⁸⁹–Asn⁹⁴) and β6 (Glu⁹⁹–Glu¹⁰⁵), which pair with one another and pack against strand β3 through backbone hydrogen bonding to residues in strand β6. These interactions complete one face of the barrel. Interestingly, the backbone that forms chains β3to β6 adopts a (3,1)_N Greek key motif that is characteristic of immunoglobulin-like fold even though IsdC does not share significant sequence homology with members of this superfamily. After strand β6, the chain then progresses to the other side of the barrel to form strands β7 (Asn¹¹⁰–Lys¹²⁴) and β8 (Pro¹²⁹–Asn¹⁴³), which complete the structure by pairing with strands β4 and β1b, respectively.

Structural Basis for the Recognition of Protoporphyrin—Zn-PPIX is stationed within a hydrophobic cleft at one end of the barrel structure. One face of the protoporphyrin ring lies flat on the surface formed by residues in strands β7 and β8, whereas the other side is contacted by the 3₁₀ helix (Fig. 3a). Binding buries only ∼60% of the surface area of the protoporphyrin ring, consistent with the weak affinity of IsdC^N for both heme and ZnPPIX. In the pocket, ZnPPIX is positioned so as to bury the non-polar 2- and 4-vinyl and 3- and 5-methyl groups against residues in strand β4 and the preceding loop. The center of this hydrophobic interface is formed by the 4-vinyl and 3-methyl groups, which are encapsulated by the side chains of Ile¹¹⁷ (β7) and Ile⁷⁸ (β4), and the aromatic side chains of Tyr⁵² (β1a-3₁₀ helix linker), Phe⁵³ (β1b–β2 linker), and Trp⁷⁷ (β3–β4 loop). In addition, the side chains of residues Ile¹²¹ and Trp⁷⁷ surround the 2-vinyl group, whereas the adjacent 1-methyl group packs against Phe¹³⁰. The 1-methyl group of ZnPPIX also contacts the side chain of Val¹²⁵ found in the β7–β8 loop, which is the only region external to the hydrophobic core that shows interaction with the tetrapyrrole (Fig. 3a). On the opposite side of the ring, the non-polar interface is continued by contacts to the 5-methyl, which is juxtaposed with Ile⁴⁸ (3₁₀ helix) and Ile¹³⁸ (β8). This binding arrangement is well supported by the NMR data (Fig. 1d) and exposes the anionic propionates of the protoporphyrin ring to the solvent for favorable interactions.

FIGURE 3.

Protoporphyrin binding pocket of IsdC^N. a, view of the ZnPPIX binding pocket. The ZnPPIX molecule is shown in red, except for the pyrrole nitrogens, which are gray. It interacts with non-polar residues within the 3₁₀-helix and strands β4, β7, and β8 (colored yellow). b, as in panel a, but the structure has been rotated by 90° to show an edge-on view of the ZnPPIX.

The zinc ion is coordinated by five substituents; it interacts with the four pyrrole nitrogen atoms of the protoporphyrin and appears to interact with the side chain of Tyr¹³², located at the base of the pocket in strand β8 (Fig. 3b). No potential coordinating atoms for Zn²⁺ are present on the opposite side of the ring. Rather, the side chain of Ile⁴⁸, offered by the 3₁₀ helix, lies perpendicular to the protoporphyrin plane. This finding nicely explains the substantial upfield chemical shift of its δ-methyl group and the observation of NOEs between this methyl and the βH and 5-CH₃ protons within ZnPPIX (Fig. 1d). This interaction surface is completed by contacts from the methyl group of Ala⁴⁹, which lies adjacent to Ile⁴⁸ in the helix and is positioned by several intermolecular NOEs to the β-meso, 4α-vinyl, and 5-CH₃ protons of the protoporphyrin.

IsdC^N Adopts a Rigid Structure in the Complex—Protein dynamics play an instrumental role in molecular recognition and cannot be revealed from structural data alone. To gain insight into the backbone dynamics of the protein in the complex we measured the rates of longitudinal (R₁) and transverse (R₂) relaxation, as well as the {¹H}-¹⁵N NOE values of its backbone ¹⁵N atoms. These data were then interpreted using the Model-free formalism to extract the magnitudes and time scales of motion (61, 66). This yields the S² parameter, which gives a concise account of each mobility of NH bond vector on the picosecond time scale. It ranges from 0 to 1, with values of 1 indicating that the amide is completely immobilized. Fig. 4a shows a plot of the S² values as a function of residue number revealing near uniform picosecond time scale mobility over the structured region of the protein. As expected, the S² values are generally higher in regions of regular secondary structure, and smaller at the flexible N and C termini. Interestingly, residues surrounding ZnPPIX are quite rigid, including those within the 3₁₀ helix and strands β7 and β8 that protrude from the body of the protein (Fig. 2b). Amides undergoing slower micro- to millisecond time scales are indirectly revealed from the Model-free approach as a contribution to the transverse relaxation in the form of the R_ex term. Only four amino acids in the complex exhibit modest R_ex values: Glu⁸⁸ (2.9 s^-1), Val¹⁰⁶ (2.5 s^-1), Lys¹⁰⁸ (4.3 s^-1), and Lys¹²⁸ (1.6 s^-1). However, they are distributed throughout the structure and not localized to the heme pocket. Taken together, the relaxation data strongly indicate that the backbone of IsdC^N in the complex is immobilized on the pico- to millisecond time scales.

FIGURE 4.

Model-free and NMR relaxation data of free and bound IsdC^N. a, S² parameters as a function of residue number for IsdC^N within the IsdC^N·ZnPPIX complex. S² gives a concise account of the mobility of each NH bond vector on the picosecond time scale. It ranges from 0 to 1, with values of 1 indicating that the amide is completely immobilized. b, ¹⁵N{¹H} heteronuclear NOEs as function of residue number for apoIsdC^N. c, R₂, the ¹⁵N spin-spin/transverse rate constant, plotted as a function of residue number for apoIsdC^N.

Comparison to the Crystal Structure of the IsdC^N-Heme Complex—During the refinement of the NMR structure, the crystal structure of the IsdC NEAT domain bound to heme was reported (29). The NMR structure of the IsdC^N·ZnPPIX complex and crystal structure of the IsdC^N·heme complex are very similar to one another, substantiating the use of ZnPPIX as a heme analog. When superimposed, the coordinates of the Cα atoms of residues Asp³⁰ to Gly¹⁴⁴ have an r.m.s. deviation of only 1.5 Å (Fig. 5a). However, slight differences do exist, such as the tilt of the protoporphyrin within the binding pocket, which differs by ∼12 and ∼10 degrees about the z and x axes, respectively (Fig. 5a). ZnPPIX is also buried more deeply in the pocket in comparison to heme (1.23 Å). In addition, there are some variations in the orientation of residues that line the pocket. For example, the side chain of Ser⁴⁷ located adjacent to the helix points toward the porphyrin in the crystal structure, whereas in the NMR structure it points into the solvent. Finally, the lone α-helix is also slightly longer in the crystal structure as compared with the NMR structure (x-ray, Ile⁴⁸–Tyr⁵²; NMR, Ala⁴⁹–Asp⁵¹). Overall, these minor variations do not affect how the protoporphyrin is bound and they likely originate from differences in the ionic radii of the central metal ion and the methods used to determine each structure.

FIGURE 5.

Differences between the solution structure of IsdC^N·ZnPPIX and previously determined structures of NEAT·heme complexes. a, overlay of the NMR structure of the IsdC^N·ZnPPIX complex (red) and the crystal structure of the IsdC:heme (blue). ZnPPIX is colored orange, and heme is colored cyan. Both views differ by a 90° rotation. The x and z axis of rotation are shown to facilitate comparison of heme and ZnPPIX bound to the crystal structure and solution structure of IsdC, respectively. b, overlay of the protoporphyrin binding pockets in the IsdC^N·ZnPPIX (blue) and IsdA·heme (gray) complexes. ZnPPIX is colored red. Side chains of IsdC^N and IsdA are colored blue and gray, respectively. The heme in the IsdA·heme complex is not shown, but is similarly positioned as the ZnPPIX molecule in the IsdC^N·ZnPPIX complex.

ApoIsdC^N Contains a Flexible Binding Pocket—The structure of IsdC^N in the absence of heme has not been determined. Because ligand binding can have profound effects on the structure and dynamics of a protein, we studied apoIsdC^N using NMR (67). Many regions of the ¹H-¹⁵N HSQC spectra of the apo- and ZnPPIX-bound forms of the protein were sufficiently different to preclude assignment of apoIsdC^N using the previously determined chemical shifts of the IsdC^N·ZnPPIX complex. Therefore, the NMR spectra of a ¹⁵N- and ¹³C-labeled sample of apoIsdC^N was assigned using triple resonance experiments (three-dimensional HNCO, HNCACB, CBCA(CO)NH, HNCA, and HN(CO)CA). We were able to assign nearly all of the cross-peaks in the ¹H-¹⁵N HSQC spectrum to specific residues within the protein (data not shown). However, after this point, it became apparent that signals for 29 residues were missing in the NMR spectra (triple resonance and ¹H-¹⁵N HSQC spectra). ¹H-¹⁵N HSQCs acquired at various temperatures (278 to 303 K) also failed to produce extra resonances (data not shown). An inspection reveals that the missing data occurs for residues in the binding pocket, suggesting that it is flexible in the absence of heme (Fig. 6a).

FIGURE 6.

The effects of heme binding on the NMR spectra and secondary structure of IsdC^N. a, residues that could not be assigned in apoIsdC^N are mapped to the backbone structure of IsdC^N solved in complex with ZnPPIX. Unassigned residues that comprise the 3₁₀ helix and surrounding loops are shown in yellow. Unassigned residues from β7 and β8 and a single residue from β5 are shown in red. The views differ by a 90° rotation. b, representative graph showing the TALOS prediction of secondary structure formation in holoIsdC^N and (c) in apoIsdC^N. Residues that could not be assigned are left blank. Residues with a value of 1, 0, and -1, indicate β-strand, loop, and helix formation, respectively.

To gain greater insight into the flexibility of apoIsdC^N, we measured R₂ and ¹⁵N NOE values for its backbone amide atoms (Fig. 4, b and c). This analysis reveals that regions outside of the heme binding pocket remain rigid in the absence of heme. This is evidenced by {¹H}-¹⁵N NOE values near 0.9 and uniform R₂ rates throughout these residues. Of particular interest is the relaxation data from residues Asn¹²⁶ and Gly¹²⁷. These residues are located at the tip of the β7–β8 strand that forms one side of the heme binding pocket. Their relaxation data indicates that they are immobilized, suggesting that this side of the binding pocket is structured.

Secondary Structure Analysis of ApoIsdC^N—The structure of IsdC^N in the absence of heme is not known. The aforementioned line broadening precludes a high resolution structure determination of apoIsdC^N by NMR. Moreover, repeated attempts to measure residual dipolar couplings of a variety of alignment media (DMPC:DHPC and alkyl-poly(ethylene glycol):hexanol liquid crystalline media) (68) proved unsuccessful. We therefore compared the secondary chemical shifts of IsdC^N in the free and bound state. This analysis is an effective way to identify specific secondary structures within a protein and the impact of protoporphyrin binding. Secondary structure of each protein was analyzed using Cα, Cβ, C′, ¹Hα, and ¹⁵N chemical shifts using the program TALOS (38). As expected, an analysis of the complex reveals the presence of eight β strands and a single helix in the protein that coincide well with the structural elements present in the three-dimensional solution structure (Fig. 6b). A similar analysis of the assignable chemical shifts of apoIsdC^N reveals that most of the secondary structural elements are preserved (Fig. 6c). Notably, the sheet forming one side of the heme binding pocket is present in the apoform as residues Asn¹¹⁰–Lys¹¹⁶ in strand β7, and residues Pro¹²⁹-Lys¹³¹ and Tyr¹⁴⁰-Asn¹⁴³ in strand β8, exhibit secondary shifts indicative of their participation in a β-sheet. Combined with the relaxatin data of the apo-form (Fig. 4, b and c), this suggests that the global structure of the protein is preserved in the absence of heme. It also suggests that the helix in the binding pocket is mobile, whereas residues in the β7–β8 platform that binds heme are immobilized and in a β-sheet conformation.

Measurement of ZnPPIX Release from IsdC^N—The kinetics of heme binding by IsdC has not been directly measured. To estimate the dissociation rate of ZnPPIX from IsdC^N, a complex containing ¹⁵N-labeled IsdC^N and 0.5 eq of ZnPPIX was studied (50 mm sodium phosphate, 100 mm NaCl, 0.01% sodium azide (pH 6)). The proton line widths of select cross-peaks within the ¹H-¹⁵N HSQC spectra of apoIsdC^N and the 50% complexed sample were then compared (Fig. 1c). Because the free and bound forms of the protein are of similar size, they presumably have similar transverse relaxation times (T₂). Therefore, for a second-order exchange process describing ZnPPIX binding to the protein, 1/T^*_2EL = 1/T^*_2E + k_-1, where 1/T^*_2EL and T^*_2E are the effective relaxation times of the protons and k_-1 is the rate of ZnPPIX release from the protein (69). This comparative analysis reveals that ZnPPIX dissociates from the protein at a rate of 126 ± 30 s^-1.

DISCUSSION

S. aureus uses the Isd system to scavenge heme-iron from hemoglobin. The cell wall-embedded IsdC protein plays a central role in this process as it receives heme from surface-exposed receptors such as IsdA, IsdB, and IsdH, before transferring it to the membrane-associated IsdDEF complex for import into the cytoplasm. To understand how IsdC transfers heme, we solved the NMR structure of the IsdC^N:ZnPPIX. Our results reveal that the NEAT domain within IsdC binds protoporphyrin in a large hydrophobic cleft located at the end of its β barrel structure. One face of the porphyrin ring lies flat on the surface formed by strands β7 and β8, whereas the other side is contacted by a 3₁₀ helix (Fig. 2b). This binding mode has also been seen in two recently published crystal structures of the IsdA·heme and IsdC·heme complexes (28, 29). In each structure, the metal ion is bound in a pentacoordinate manner using a tyrosine residue as the axial ligand (Tyr¹³² in IsdC). Primary sequence alignments indicate that this residue is conserved in the 2nd and 3rd NEAT domains within the IsdB and IsdH proteins, respectively. These domains have yet to be characterized, but presumably bind heme. Notably, the Shp hemoprotein from Streptococcus pyogenes adopts a NEAT domain fold and binds heme in a generally similar manner even though it is unrelated at the primary sequence level (17–19% sequence identity with the Isd NEAT domains from S. aureus) (70). As Shp is also a cell wall-associated protein involved in heme capture from hemoglobin, the NEAT domain fold may be especially well suited for this function.

Previous studies have not directly measured the affinity of IsdC for heme or the rate at which heme is released from the protein. Our results indicate that IsdC^N binds heme weakly with a K_D of 0.34 ± 0.12 μm. In addition, the bound ZnPPIX molecule rapidly dissociates at a rate of 126 ± 30 s^-1. These properties should enable IsdC to rapidly capture and release heme as it is transferred across the cell wall. Furthermore, the data are compatible with the structure of the complex, as the protoporphyrin appears to be loosely held by the protein because it is only partially buried within the binding pocket (∼40% is solvent accessible) and its central metal is coordinated by only one axial ligand. Interestingly, the rate of protoporphyrin release measured by NMR is much faster than the value obtained from heme transfer studies (71). The reason for this discrepancy is unclear, although it may be due to the different metalloporphyrins used in the studies.

Directional flow of heme into the cell wall from the surface-exposed IsdA, IsdB, and IsdH heme receptors may in part be driven by heme affinity differences between these proteins and IsdC, which binds heme ∼10-fold more tightly (72). To gain insight into the molecular basis of this affinity difference we compared the solution structure of the IsdC^N·ZnPPIX complex with the crystal structure of IsdA bound to heme. Both proteins bind protoporphyrin in a generally similar manner even though they share only 18% sequence identity. However, two significant structural differences exist that may explain why IsdC^N has higher affinity for heme. First, in IsdC^N the β7–β8 strands extend farther from the body of the protein. This enables IsdC to make additional specific contacts to the 1-CH3 portion of the protoporphyrin ring by the side chain of Val¹²⁵ within the β7–β8 loop. Second, in IsdC^N, the indole ring of Trp⁷⁷ presented from the β3–β4 loop forms a large surface that contacts the 2-β-vinyl group of the protoporphyrin ring. Although in the IsdA·heme complex, less extensive contacts to this portion of the ring are made by a phenylalanine ring (Phe¹¹²) (Fig. 5b). Interestingly, the results of a sequence alignment of the Isd NEAT domains are compatible with this single amino acid playing a major role in modulating affinity as it is not conserved in other potential upstream heme-receptors (2nd and 3rd NEAT domains of IsdB and IsdH, respectively) (18).

NMR studies indicate that IsdC captures heme using a flexible binding pocket. In the apo-form, the resonances from residues that line the heme binding pocket show extensive line broadening, suggesting that their atoms experience micro- to millisecond time scale fluctuations in their magnetic environments. Flexibility is localized to this region as the relaxation properties and the chemical shifts of the remainder of apoIsdC suggest that the global fold is preserved. Moreover, it is likely that only the 3₁₀ helix is in motion in the absence of heme because several residues in the opposing binding surface formed by the β7–β8 sheet exhibit relaxation and chemical shift properties similar to the holo-protein. Two types of helical motions are plausible: 1) the helix may alternate between ordered and disordered states in the absence of heme or, 2) the helix could remain structured and undergo segmental motions (29). Both types of motion would occur in the unfavorable intermediate exchange regime and could cause broadening by affecting the chemical environment of the residues lining the base of the β7–β8 sheet. A mobile 3₁₀ helix in apoIsdC^N is compatible with the structure of the hemoglobin and haptoglobin-binding IsdH protein, because the analogous region in its NEAT domain is structurally disordered in the apo-state (18, 73).

In contrast to IsdC, crystallographic studies indicate that the IsdA protein uses a preformed pocket to bind heme (28). In the structure of apoIsdA the 3₁₀-helix is stabilized by contacts from the side chain of Met⁸⁴, which projects from the helix and interacts with non-polar residues located at the base of β7–β8 strand (28). Interestingly, in IsdC^N, Ala⁴⁹ replaces this methionine and is presumably too short to stabilize helical packing, which may explain why it is mobile in the apo-state (Fig. 5b).

Recent studies suggest that the formation of distinct protein-protein complexes mediate heme transfer from the heme receptors, IsdA and IsdB, to IsdC (71). The flexible binding pocket observed in apoIsdC may be useful in promoting interactions with these proteins. It may also serve to fine tune both the affinity and kinetics of heme binding to maximize the efficiency and directionality of heme transfer across the cell wall. As protoporphyrin-based molecules such as ZnPPIX kill pathogenic bacteria, including S. aureus (27), the structure of the complex presented here may be useful in guiding the design of related metalloporphyrins that selectively disrupt heme import.