Structural Basis for Hemoglobin Capture by Staphylococcus aureus Cell-surface Protein, IsdH

Background: Bacteria need iron from the host to establish infection. Results: We report the first structure of hemoglobin bound to a bacterial protein and show that targeted disruption of this interaction can reduce Staphylococcus aureus growth when hemoglobin is the sole iron source. Conclusion: Physical capture of hemoglobin is important for iron uptake by S. aureus. Significance: Hemoglobin receptors may be targets for new antibacterial agents. Pathogens must steal iron from their hosts to establish infection. In mammals, hemoglobin (Hb) represents the largest reservoir of iron, and pathogens express Hb-binding proteins to access this source. Here, we show how one of the commonest and most significant human pathogens, Staphylococcus aureus, captures Hb as the first step of an iron-scavenging pathway. The x-ray crystal structure of Hb bound to a domain from the Isd (iron-regulated surface determinant) protein, IsdH, is the first structure of a Hb capture complex to be determined. Surface mutations in Hb that reduce binding to the Hb-receptor limit the capacity of S. aureus to utilize Hb as an iron source, suggesting that Hb sequence is a factor in host susceptibility to infection. The demonstration that pathogens make highly specific recognition complexes with Hb raises the possibility of developing inhibitors of Hb binding as antibacterial agents.

Pathogens must steal iron from their hosts to establish infection. In mammals, hemoglobin (Hb) represents the largest reservoir of iron, and pathogens express Hb-binding proteins to access this source. Here, we show how one of the commonest and most significant human pathogens, Staphylococcus aureus, captures Hb as the first step of an iron-scavenging pathway. The x-ray crystal structure of Hb bound to a domain from the Isd (iron-regulated surface determinant) protein, IsdH, is the first structure of a Hb capture complex to be determined. Surface mutations in Hb that reduce binding to the Hb-receptor limit the capacity of S. aureus to utilize Hb as an iron source, suggesting that Hb sequence is a factor in host susceptibility to infection. The demonstration that pathogens make highly specific recognition complexes with Hb raises the possibility of developing inhibitors of Hb binding as antibacterial agents.
Bacteria that cause human disease must acquire iron from their host to establish infection (1). Host strategies to restrict iron availability therefore constitute a form of innate immunity. Most iron is sequestered within cells, and serum iron is kept at an extremely low concentration (ϳ10 Ϫ24 M) by the iron transport protein transferrin (2). A related protein, lactoferrin, is present in exocrine solutions such as tears and milk and is also released from neutrophils at sites of infection (3). Iron uptake and homeostasis are even tuned toward anemia in chronic infection by the hormone hepcidin (4). In cases where serum iron levels are either artificially or genetically elevated, there are dramatic increases in levels of bacteremia (5).
Bacteria in all iron-restricted environments, including the human body, employ a variety of generic mechanisms to scavenge this essential element. These mechanisms include the production of siderophores, small (typically Ͻ1000 Da) soluble factors that bind with very high affinity (ϳ10 Ϫ22 -10 Ϫ50 M) to Fe(III). Siderophores are subsequently reabsorbed by the bacteria in the iron-charged form (6). Alternatively, iron-reducing compounds can be released into the environment to generate soluble Fe(II) that is absorbed through the FeoB permease, escaping the requirement for metabolically expensive siderophore production (7).
Naturally, the rich iron reservoirs found in mammals and other animals has driven evolution of iron-scavenging mechanisms that are specific to the pathogen-host relationship. Perhaps the most distinguishing feature of the mammalian iron profile is that ϳ75% of the total body-iron is locked up in heme, a cyclic organic compound (porphyrin) with central squareplanar Fe(II/III) ion coordinated by four nitrogen atoms. This iron pool is considered largely inaccessible to siderophores (6), and pathogens have therefore evolved specialized pathways dedicated to heme uptake. As with all forms of iron, heme is retained tightly sequestered into host proteins at all times. Hb is the most abundant heme protein in humans, accounting for ϳ70% of total body iron, making it a particularly attractive iron source for invading microbes.
Highly pathogenic species such as Staphylococcus aureus secrete hemolysins (8) to release Hb into the serum where it is accessible to the bacteria. S. aureus displays a preference for heme as an iron source (9) and can grow on Hb as the sole source of iron (10,11). Utilization of Hb in S. aureus is mediated by Isd proteins (12,13), which are also found in a large number of other Gram-positive pathogens (14,15). The Isd pathway in S. aureus comprises nine proteins IsdA-I (12). Of these, the four □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental "Experimental Procedures," proteins IsdA, IsdB, IsdC, and IsdH are expressed on the bacterial cell-surface, anchored through a C-terminal covalent linkage to the peptidoglycan cell wall. Notably, the surface exposed Isd proteins are the most highly up-regulated genes in response to iron starvation in S. aureus (16) and Bacillus anthracis (17). IsdB and IsdH are Hb-binding proteins (10,11). IsdA and IsdC are heme-binding proteins that cooperate with IsdB/H through an unknown mechanism to transfer heme in a unidirectional manner to the membrane-associated lipoprotein IsdE (18 -20). A dedicated ABC transporter complex (IsdD and IsdF) transfers heme from IsdE into the S. aureus cytoplasm, where the porphyrin macrocycle is cleaved by a heme oxygenase complex (IsdG and IsdI), releasing its iron (21). Genetic inactivation of IsdA (22), IsdB (11,22,23), IsdH (24,25), IsdG, or IsdI (26) reduces the ability of S. aureus to cause infections in mice, confirming the Isd system as an important virulence mechanism. It has been demonstrated that immunization with IsdA, IsdB or IsdH antigens (24,25,(27)(28)(29) or administration of purified antibodies to IsdA, IsdB, IsdC, and IsdH (22) confers some protection from infection in various animal models. These studies suggest that blocking the Isd pathway of heme/iron uptake could have therapeutic benefit in human infections.
To transfer the heme group from Hb to the bacterial membrane, the cell wall-anchored proteins IsdA, IsdB, IsdC, and IsdH each possess one ϳ130-residue NEAT (near iron transporter) domain that binds to heme (30 -34). Extraction of heme from Hb relies upon a physical interaction of Hb with the Hb receptors IsdB and IsdH (20). IsdB and IsdH contain, respectively, one or two variant NEAT domains that possess Hb-binding activity but do not bind to heme (see Fig. 1A) (10,(35)(36)(37). The structural basis for heme binding by NEAT domains is comparatively well understood (30 -34), but the molecular mechanism of Hb recognition has remained elusive and is the subject of the current investigation.
Many pathogens express Hb-binding proteins, but the molecular details underlying Hb recognition are unknown in every case. Here, we show that the first NEAT domain from IsdH (IsdH N1 ) binds to a site on the ␣-chain of Hb (␣Hb) and determine the x-ray crystal structure of the IsdH N1 ⅐Hb complex. Using mutant forms of Hb that are defective in IsdH N1 binding, we demonstrate that physical capture of Hb by S. aureus is important for the utilization of this iron source.

EXPERIMENTAL PROCEDURES
Protein Preparation-The DNA sequence encoding the IsdH N1 (IsdH residues 86 -229) and IsdH N2 (residues 321-467) domains from S. aureus strain TCH1516 were cloned into pET15b (Novagen) for expression with an N-terminal His 6 tag. The proteins were expressed and purified as described previously (35) to yield a final product with the additional N-terminal sequence MGSSHHHHHHSSGLVPRGSHM. Native human Hb A was prepared and separated into its constituent ␣Hb and ␤Hb chains as described previously (38). Hb, ␣Hb, and ␤Hb were maintained in the carbonmonoxy-liganded state during purification and subsequent analysis, unless otherwise specified.
Recombinant human Hb (rhHb) 5 was produced in Escherichia coli strain BL21(DE3) from the pHb0.0 plasmid, a gift of Dr. John Olson (39). pHUG21 (a gift of Dr. Doug Henderson) harboring the Plesiomonas shigelloides heme transport system was co-transformed to enhance rHb expression (40). The heme transport system was induced by iron restriction with 50 g/ml of the iron chelator ethylenediamine-di-(o-hydroxyphenyl acetic acid) (EDDHA, LGC Standards GmbH). rHb expression was performed at 16°C overnight, and bacteria were lysed by passage through a French press twice at 1200 psi. Hb was purified in a single step over nickel-nitrilotriacetic acid beads (Qiagen) by virtue of interactions with naturally occurring His residues on the surface of Hb and dialyzed twice against PBS. Substitution mutations within the Hb genes were generated using PCR-based mutagenesis and confirmed by sequencing.
For protein crystallography, native Hb purified in the carbonmonoxy-liganded state was converted to the oxygenated form by passing a pure stream of oxygen over a protein solution held on ice and illuminated with a focused beam from a 50-watt halogen lamp. The oxy-Hb was converted to met-Hb by addition of excess potassium ferricyanide in 20 mM sodium phosphate, pH 7.0. The reaction was monitored to completion by UV-visible spectroscopy, at which point the Hb protein was isolated over Sephadex G-25. Met-Hb was combined with purified IsdH N1 in a 1:1 molar ratio (with respect to ␣Hb⅐␤Hb dimers) for crystallization.
Light Scattering-Samples were separated on a Superose 12 column (GE Healthcare) equilibrated in 20 mM sodium phosphate, 150 mM sodium chloride, pH 7.0, with in-line MALS (mini-DAWN; Wyatt Technology Corp., Santa Barbara, CA) and refractive index (Optilab differential refractometer, Wyatt Technology Corp.) measurements. The refractive index increment with respect to mass concentration (dn/dc) was taken to be 0.19 ml g Ϫ1 for all proteins/complexes.
Isothermal Titration Calorimetry (ITC)-ITC was carried out on a MicroCal iTC 200 instrument. All components were dialyzed into 20 mM sodium phosphate, pH 7.5, 150 mM sodium chloride. Multiple injections of 1.1 l were performed at 150-s intervals, with continuous stirring at 20°C. The volume of the reaction cell was 350 l. Heats of dilution were determined from injections of protein solution into buffer and subtracted from the experimental data.
UV-visible Spectroscopy-Changes in heme coordination of ␣Hb (20 M) were measured by UV-visible absorption spectroscopy (Shimadzu UV1800) at 30°C in the presence/absence of NEAT domains in 20 mM sodium phosphate, 10 M diethylenetriaminepenta-acetic acid, pH 7.0.
Structure Determination-Crystals were prepared by hanging-drop vapor diffusion at 293 K in which 2 l protein solution (10.9 mg/ml) was mixed with 2 l of precipitant (0.2 M potassium thiocyanate, 0.1 M Bis-tris propane, pH 7.5, 20% PEG). Crystals of 150 -200 m appeared within 2 days. Seeding was done to grow diffraction quality crystals. The crystals were cryoprotected with 30% glycerol and flash-cooled in a cold nitrogen stream (100 K). X-ray diffraction data were collected in-house using copper K␣ x-ray produced by a Rigaku 007HF rotating-anode generator with Osmic Varimax optics and recorded on a MAR345 image plate (Marresearch GmBH).
Data were collected over a phi-range of 180°and were indexed and scaled using MOSFLM (41) and Scalepack (42). Structure was solved by molecular replacement using PHASER (43), which gave a unique solution when using ␣Hb or ␤Hb (PDB code 1IRD) as independent search models. The structure was refined using REFMAC5 (44), with manual map inspection and model building performed in COOT (45). The quality of the model was regularly checked for steric clashes, incorrect stereochemistry, and rotamer outliers using MolProbity (46).
Bacterial Strains and Growth Conditions-All experiments were carried out with S. aureus strain Newman (47) or with mutants generated in its background. All cultures were inoculated from a single colony and grown overnight (ϳ20 h) in 5 ml of RPMI medium (Thermo) supplemented with 1% casamino acids in 15-ml conical tubes at 37°C with shaking at 180 rotations per minute (rpm). The isogenic variant in which the Hb receptors IsdB and IsdH have been deleted (⌬isdBH) has been described previously (11).
S. aureus Growth Curves-Single colonies of S. aureus were inoculated into RPMI medium plus casamino acids, supplemented with 0.5 mM EDDHA and grown overnight. One ml of overnight cultures was normalized to A 600 of 3.0, and bacteria were sedimented (9000 ϫ g, 3 min) and resuspended in 1 ml of NRPMI with 0.5 mM EDDHA. NRPMI was prepared in advance by treating RPMI plus casamino acids with Chelex 100 (Sigma) according to the manufacturer's recommendations and supplementing the resulting ion-deficient medium with 25 mM ZnCl 2 , 25 M MnCl 2 , 100 M CaCl 2 , and 1 M MgCl 2 . The resulting suspension of S. aureus was subcultured (1:100) into 1 ml of NRPMI with 0.5 mM EDDHA and rHb at 2.5 g/ml. One-ml cultures were incubated at 37°C in 15-ml conical tubes with shaking at 180 rpm. A 600 measurements were taken at the indicated time points by mixing 20-l aliquots of the culture with 180 l of PBS in 96-well plates. The graphs represent a mean of three independent experiments. Error bars represent S.D.; asterisks denote values upon mutant rHb supplementation significantly different from values upon wild type rHb supplementation at the same time point (Student's two-tailed t test, p Ͻ 0.05).

First and Second NEAT Domains of IsdH Bind to ␣Hb
Chain of Hb-To investigate the molecular mechanism of Hb recognition by S. aureus, we produced the Hb-binding domains, IsdH N1 and IsdH N2 (Fig. 1A), in an E. coli expression system. Recombinant IsdH N1 and IsdH N2 are pure monomers as determined by SEC (Fig. 1B) and Rayleigh light scattering (supplemental Table 1). Both NEAT domains bind to native adult Hb A, shifting the SEC elution peak to an earlier elution time (Fig.  1C, solid line). Formation of a protein complex was confirmed by light scattering measurements (supplemental Fig. 1). To determine the binding sites on Hb, we first separated the constituent ␣Hb and ␤Hb chains. Mixing of IsdH N1 or IsdH N2 with purified ␣Hb, at a 1:1 molar ratio, lead to the formation of a distinct protein complex (Fig. 1D), with a molecular weight close to that expected for a heterodimer (supplemental Fig. 1). The binding affinity for the IsdH N1 :␣Hb was 100 nM, determined by isothermal titration calorimetry (ITC, supplemental Fig. 2). Interaction between IsdH N1 and ␤Hb could not be detected using SEC (Fig. 1E), light scattering or ITC (supplemental Figs. 1 and 2). Some evidence of complex formation was detected upon mixing equimolar quantities of IsdH N2 and ␤Hb, but the mixture retained a substantial proportion of free monomers suggesting a weaker interaction (Fig. 1E). Thus, the IsdH N1/N2 domains show a clear preference for ␣Hb binding, although it is conceivable that interaction of IsdH N2 with ␤Hb might also be significant in vivo. In the absence of heme, the resulting apo-␣Hb chain is no longer competent to bind IsdH N1 (supplemental Fig. 3), suggesting that Hb is released from the Hb receptor following heme removal.
IsdH N1/N2 Do Not Induce Large-scale Changes in Heme Pocket Structure of ␣Hb-To determine whether IsdH N1 or IsdH N2 domains induce structural changes that might promote heme release, we employed UV-visible absorption spectros- copy as a sensitive probe of heme coordination chemistry. These studies revealed that IsdH N1/N2 had no substantial effect on ␣Hb heme pocket structure or heme release over a period of 8 h (Fig. 2A). Small absorbance changes were observed in all samples consistent with slow autooxidation. After 48 h the spectrum of IsdH N1 :␣Hb contained distinct peaks at 630, 500, and 405 nm (Fig. 2B, solid line), characteristic of the auto-oxidation product, met-Hb (48). The sample of free ␣Hb contained precipitate and following filtration, yielded a spectrum indicative of a mixture of heme-ligand states (Fig. 2B, dashed line). Our interpretation is that met-␣Hb forming in the free ␣Hb sample was being lost as precipitate (met-␣Hb is well known to be highly unstable) and that IsdH N1 was stabilizing met-␣Hb. ITC revealed that IsdH N1 binds carbonmonoxy-␣Hb or met-␣Hb with the same affinity (supplemental Fig. 2). These data indicate that IsdH N1 and IsdH N2 capture Hb, independent of ligand/oxidation state, but do not induce conformational changes in Hb that cause heme release. Hence, rapid heme release must rely upon recruitment of the heme acceptor domain of IsdH.
Crystal Structure of IsdH N1 ⅐Hb Complex-To reveal the molecular mechanism of Hb binding, IsdH N1 was crystallized in complex with met-Hb, and x-ray diffraction data were collected to a resolution of 3.0 Å (Table 1). Initial phases were obtained by molecular replacement using ␣Hb or ␤Hb (PDB code 1IRD) as independent search models. In the crystal structure, IsdH N1 binds to one ␣Hb⅐␤Hb dimer through a site on the ␣Hb chain (Fig. 3A). There are two independent IsdH N1 ⅐Hb complexes in the crystallographic asymmetric unit. The IsdH N1 and ␣Hb subunits are identical in the two complexes (C ␣ coordinates overlay with a root mean square displacement of 0.35 Å, supplemental Fig. 4). Electron density in one of the ␤Hb subunits is weak, possibly due to whole domain movement. The well defined ␣Hb⅐␤Hb dimer is very similar in structure to the native Hb dimer (C ␣ displacement of ϳ1 Å root mean square) with no dramatic conformational change being induced by IsdH N1 (supplemental Fig. 4). The dimeric form of Hb appears to have been selected by the crystallization process, as IsdH N1 does not overlap with the Hb dimer-tetramer interface. Indeed, molecular weight measurements obtained for the complex in solution were consistent with IsdH N1 bound to a mixture of Hb dimers and tetramers (supplemental Table 1). In any case, the met-Hb dimer-tetramer interaction is relatively weak (K d ϳ 30 M) (49); hence, the ␣Hb⅐␤Hb dimer observed in our structure is likely to be a physiologically relevant species in plasma.
IsdH N1 ⅐␣Hb Interface-The IsdH N1 domain is a ninestranded distorted ␤-barrel with an immunoglobulin-like fold. The conformation is very similar to the heme-binding NEAT domains from IsdA, IsdB, and IsdC (30 -34) but with an additional N-terminal helix loop. The binding site for Hb is formed from loops at one end of the ␤-barrel (Fig. 3A), designated loops 2, 4, 6, and 8 based on the conserved NEAT domain fold. Loop 2 of IsdH N1 contains an unusual string of five consecutive aromatic residues (Tyr-125, Tyr-126, His-127, Phe-128, and Phe-129; Fig. 3, B and C), which are crucial for Hb binding (36). A number of these aromatic side chains make van der Waals contacts with non-polar surfaces on ␣Hb and are partially (Tyr-125 and Phe-128) or completely (Tyr-126 and Phe-129) buried from solvent. Loops 4 and 6 of IsdH N1 account for the majority of electrostatic interactions with ␣Hb (Fig. 3C). Overall, the ␣Hbinteracting face of IsdH N1 displays a negative electrostatic potential that complements a surface with positive potential on ␣Hb. Of particular note, the positively charged ⑀-amino group of ␣Hb Lys-11 protrudes into an electronegative pocket on IsdH N1 (Fig. 3D) lined with hydroxyl and carbonyl groups from Ser-130, Tyr-126, Asn-151, and Thr-152 (Fig. 3C).
A comparison with the NMR structure of the free IsdH N1 domain (Fig. 4, A and B) (36) reveals a substantial conformational change in IsdH N1 loops 2 and 8 upon complex formation. Loop 8 requires a considerable movement of ϳ8-10 Å to accommodate ␣Hb. More strikingly, residues 123-129 of loop 2 are disordered in free IsdH N1 (36) and undergo a transition to an ordered helical conformation upon interaction with Hb. In the absence of ␣Hb, it appears that the aromatic side chains in loop 2 cannot bury sufficient surface area to attain a stable fold. In the helical conformation Gln-124 and His-127 in loop 2 make specific intramolecular H-bonding interactions with the underlying ␤-sheet (Fig. 4C). Remarkably, the Hb-binding site on IsdH N1 occupies the same face of the NEAT fold that is used to bind heme in other NEAT domains (Fig. 4D). We infer from this that structural properties of the NEAT fold might predispose this site to ligand interactions.
Mutations in ␣Hb Abrogate Binding to IsdH N1/N2 -To gain insight into the ligand binding specificity of NEAT domains, we compared the IsdH N1 -binding surface of ␣Hb (Fig. 5A) to the surface of ␤Hb, which does not bind IsdH N1 (Fig. 5B). The most outstanding difference between these two surfaces is a substitution of Lys-11 in ␣Hb and for Thr-12 in ␤Hb. Interestingly, Lys-11 is conserved across ␣Hb sequences from distantly related mammalian species (Fig. 5C). To test the role of Lys-11 in IsdH recognition, we produced recombinant Hb (rHb) carrying a Lys-11 to Thr mutation in the ␣Hb chain (rHb(␣K11T)). A second mutant (rHb (␣⌬18,19)) was produced to test the effect of removing the binding site for one of the IsdH N1 loop 2 aromatic residues (Tyr-125) by shortening the linker between ␣Hb helices A and B. Thirdly, rHb(␣A5E) was generated in response to the natural variation in this position across ␣Hb sequences from different species (Fig. 5C). Reciprocal mutations were also introduced into the ␤Hb chains (Fig. 5C). Wild type and mutant rHb samples were purified and converted to the stable cyanomet form (50). All proteins had normal UV-visible spectra (data not shown) and eluted in one major peak from SEC, at the expected mobility (supplemental Fig. 5), indicating no substantial defect in folding or ␣Hb-␤Hb subunit interactions. The individual SEC traces for rHb and IsdH N1 are overlaid in Fig. 5D, upper panel. Mixing of these components lead to formation of the expected IsdH N1 ⅐Hb complex (Fig. 5D, lower panel, black). Strikingly, rHb(␣K11T) showed no detectable interaction with IsdH N1 , yielding two separate elution peaks (Fig. 5D, blue). Similarly, rHb(␣K11T) failed to interact significantly with IsdH N2 (Fig. 5E, blue), strongly suggesting that IsdH N1 and IsdH N2 bind a common site on Hb. The reverse mutation in ␤Hb, rHb(␤T12K), had no detectable effect on IsdH N1 binding (Fig. 5D, red). However, rHb(␤T12K) experienced a larger than normal shift in SEC retention time when mixed with IsdH N2 (Fig. 5E, red), suggesting that the intrinsic weak binding to wild type ␤Hb subunits might be enhanced by the ␤T12K mutation. The ␣A5E mutation partially inhibited binding to IsdH N1 but not IsdH N2 and rHb(␣⌬G18,A19) also displayed reduced binding to IsdH N1 but was not tested against IsdH N2 (Fig. 5F and supplemental Fig. 5). In summary, both IsdH N1 and IsdH N2 are likely to bind Hb through the same site on ␣Hb but display some differences in sequence specificity that could potentially provide greater tolerance to species variation in the Hb sequence.
IsdH N1 -␣Hb Interaction Is Important for S. aureus Growth on Hb as Sole Iron Source-To investigate the importance of NEAT-Hb interactions for iron uptake from Hb, we grew S. aureus strain Newman on medium containing Hb as the sole iron source. When supplemented with rHb, a culture of S. aureus grew over an incubation time of 40 h as shown in Fig.  5G (black). In contrast Hb carrying ␣Hb mutations that inhibit IsdH N1 and/or IsdH N2 binding supported a significantly slower rate of culture growth. An S. aureus strain carrying deletions of both Hb receptors, IsdB and IsdH (⌬isdBH), failed to replicate on wild type or mutant Hb (Fig. 5H), confirming that the IsdB/H Hb receptors are required for iron uptake from Hb in this assay. Compared with isdBH, the residual growth of S. aureus Newman on mutant Hb suggested that our individual ␣Hb mutations are not sufficient to completely abrogate interactions with intact IsdH and/or IsdB. This could be due to avidity affects arising from the multidomain structure of IsdB/H, binding of Hb through ␤Hb chains, or because the mutations that inhibit binding to IsdH have more limited effects on the IsdB-Hb interaction. In any case, our results demonstrate that capture of Hb through NEAT domain receptors is an important step that allows S. aureus to utilize iron from Hb.

DISCUSSION
Direct binding of host iron proteins to bacterial cell-surface receptors is a widespread strategy to capture iron, but despite their importance, these interactions are still poorly understood at the molecular level. For example, the structures of transferrin receptors present on the surface of pathogenic species from the genera Neisseria and Pasteurella have only recently been determined (51,52) and only then in the absence of the transferrin ligand. The importance of Hb-binding proteins is emphasized by their widespread distribution across bacterial and protozoan pathogens. In Gram-negative bacteria, TonB-dependent heme transporters in the bacterial outer membrane act as Hb receptors in Haemophilus (53), Neisseria (54), Pasteurella (55), Porphyromonas (56), and Helicobacter (57) species. In Gram-positive bacteria, the IsdB/H proteins from S. aureus are the most well documented Hb receptors, but cell-surface proteins from Streptococcus pyogenes (58) Streptococcus equi (59) Bacillus cereus (60), and Corynebacterium diphtheriae (61) are reported to bind Hb. In addition to the surface-bound receptors, secreted hemophores produced by Gram-negative (62) and Gram-positive (63,64) bacteria may transiently interact with Hb to facilitate heme transfer. The IsdH N1 ⅐met-Hb complex is the first Hb receptor complex for which the structure has been determined. It reveals a highly specific protein-protein interaction, raising the possibility that inhibitors to this interaction might be developed. If this is so, the diversity of Hb-binding functions described above may provide restriction points through which to inhibit colonization of the host by a range of pathogens.
Loops 2 and 4 account for the majority of IsdH N1 -Hb interactions and are largely conserved in IsdH N2 and IsdB N1 (supplemental Fig. 6), suggesting that all three NEAT domains bind HbA in a similar way. In each case, loop 2 contains an aromatic- rich sequence and is predicted to adopt a similar ␣-helical structure, with conserved Gln and His residues interacting with the underlying ␤-sheet (supplemental Fig. 6). In IsdH N1 , Tyr-126 and Ser-130 form part of the binding pocket for ␣Hb Lys-11, and these residues are conserved in IsdH N2 and IsdB N1 . Similarly, side chains from IsdH N1 loop 2 involved in H-bonding to Lys-11 are conserved in IsdH N2 and IsdB N1 . Thus, it is likely that the preference for binding to ␣Hb observed for IsdH N1 and IsdH N2 will also hold true for IsdB. This, along with the fact that Lys-11 is highly conserved in ␣Hb sequences from many mammalian species, argues that targeting the ␣Hb subunit is of mechanistic importance.
The transfer of heme from met-Hb to the isolated heme acceptor domains of IsdH/B is governed by the simple dissociation of heme from met-Hb (20,35) and is 2-3 orders of magnitude slower than heme transfer to the full-length Hb receptor (20). Thus, physical interaction between Hb and the Hb receptors is crucial for activating heme transfer. Structural and UVvisible spectroscopy data obtained for IsdH N1/N2 argue that this enhancement in heme transfer rate does not arise from a con- negative, red) and ␣Hb-contacting loops from IsdH N1 (same orientation as Fig. 3C). B, a model showing the surface of ␤Hb (PDB 1GZX) overlaid with the ␣Hb-binding residues of IsdH N1 . Dashed ovals mark substantial differences between the molecular surfaces of ␣Hb and ␤Hb. C, helix A sequences of ␣Hb and ␤Hb from divergent mammalian species, showing ␣Hb residues that make H-bonds (stars) or other contacts (dots) with IsdH N1 . D, upper panel, SEC elution traces for IsdH N1 (solid curve) and recombinant wild type Hb (rHb, dashed curve). Lower panel, SEC elution traces for mixtures of IsdH N1 with rHb (black), rHb(␤T12K) (red), or rHb(␣K11T) (blue). E, SEC elution traces showing interactions of IsdH N2 with rHb and rHb mutants. F, summary of binding data for rHb mutants, indicating the fraction of unbound IsdH N1 in chromatographic separations (supplemental Fig. 5). G, growth of S. aureus strain Newman with rHb or rHb mutants as the sole iron source. H, growth of S. aureus ⌬isdBH under conditions as in G.
formational change in Hb that promotes heme release. Instead, we favor an alternative model whereby binding to Hb through the IsdH N1/N2 or IsdB N1 domains enhances the rate of productive encounters between Hb chains and the IsdH N3 or IsdB N2 heme acceptor domains. One potential role for the specific binding of IsdB/H to ␣Hb subunits might be to control the sequential release of heme from the ␤Hb and ␣Hb chains by orienting the heme acceptor domains in a particular way. ␣Hb⅐␤Hb dimers can loose heme from one subunit (semi-Hb) and retain much of their native structure (65), potentially allowing them to remain associated with the Hb receptor until all the heme is removed. IsdH N1 does not bind to free apo-␣Hb, suggesting that the IsdH⅐Hb complex would dissociate following complete heme removal from Hb.
NEAT domains with the characteristic aromatic loop 2 sequence of IsdH N1 are restricted to the genomes of S. aureus and the related Staphylococcus lugdunensis, suggesting that the Hb-binding mechanism described here is restricted to these organisms. However, Hb-binding function has been reported for other NEAT domain containing proteins, including IlsA from B. cereus (60), IsdX1 from Bacillus anthrasis (63) and Shr from S. pyogenes (66), suggesting that NEAT domains have adapted to make different interactions with Hb or that sequences outside the NEAT domains are responsible for the Hb-binding activity of these proteins.
Concerted folding and binding is proposed to allow recognition of diverse molecular targets by a single protein (67), and it is possible that, being an opportunistic pathogen, folding and binding of IsdH might help S. aureus to accommodate some of the natural sequence variation present in Hbs from different host species. IsdH also binds directly to haptoglobin, a mammalian protein that is unrelated in sequence or structure to Hb (10,25,35). Aromatic residues in IsdH N1 loop 2 are required for binding to Hb or haptoglobin (35), and it will now be interesting to establish whether folding and binding has a role in accommodating these different ligands. The function of haptoglobin is to sequester Hb that is released by continuous physiological levels of hemolysis, hence the Hb⅐haptoglobin complex represents an alternative route through which S. aureus might capture Hb.
Among Isd proteins, folding and binding is not limited to IsdH. An NMR study of the heme-binding NEAT domain from IsdC revealed that loop 2 (Fig. 4D) undergoes significant conformational dynamics in the absence of ligand (33) consistent with a folding-and-binding event. Here, loop flexibility may facilitate insertion of the hydrophobic heme molecule "under" loop 2. In contrast, loop 2 from the heme-binding NEAT domain of IsdA adopts a helical conformation in crystal structures of the heme-free and heme-bound forms. The structure of IsdA has not been determined in solution, and it is conceivable that a folded loop 2 conformation might be stabilized during crystallization. Although a more extensive study into the general role of conformational dynamics in NEAT-ligand interactions is required, it is tempting to speculate that structural flexibility in an ancestral NEAT domain may have played a role in the remarkable adaptation of this domain to binding such vastly different molecular targets as heme and Hb protein.
The IsdH N1 ⅐Hb structure reveals a highly specific recognition interface that allows S. aureus to capture heme directly from the richest iron source in the mammalian host. The Hbreceptor mechanism establishes a direct conduit for the transfer of heme groups from host Hb into the bacterial cytoplasm. In this regard, it contrasts with strategies such as siderophore or protease secretion, which mobilize iron into the extracellular milieu where it can be intercepted by competing bacterial populations. The highly specific nature of the IsdH-Hb interaction suggests that sequence variation in the IsdB/H-binding face of Hb could influence susceptibility to S. aureus infection, as described recently for mice and humans (68). Determining the structure of an Hb-capture complex now raises the exciting possibility of mimicking the NEAT-interacting face of ␣Hb as a strategy for developing new anti-virulence treatments for S. aureus infection.