Structure of the Hemoglobin-IsdH Complex Reveals the Molecular Basis of Iron Capture by Staphylococcus aureus*♦

Background: IsdB and IsdH proteins from Staphylococcus aureus strip heme iron from human hemoglobin. Results: The IsdH·hemoglobin complex shows how globin-binding and heme-binding NEAT domains of IsdH cooperate to remove heme from both chains of hemoglobin. Conclusion: The supradomain architecture of IsdH confers activity by precisely positioning the heme acceptor domain. Significance: Multiple IsdH·hemoglobin interfaces may be targets for new antibiotics. Staphylococcus aureus causes life-threatening disease in humans. The S. aureus surface protein iron-regulated surface determinant H (IsdH) binds to mammalian hemoglobin (Hb) and extracts heme as a source of iron, which is an essential nutrient for the bacteria. However, the process of heme transfer from Hb is poorly understood. We have determined the structure of IsdH bound to human Hb by x-ray crystallography at 4.2 Å resolution, revealing the structural basis for heme transfer. One IsdH molecule is bound to each α and β Hb subunit, suggesting that the receptor acquires iron from both chains by a similar mechanism. Remarkably, two near iron transporter (NEAT) domains in IsdH perform very different functions. An N-terminal NEAT domain binds α/β globin through a site distant from the globin heme pocket and, via an intervening structural domain, positions the C-terminal heme-binding NEAT domain perfectly for heme transfer. These data, together with a 2.3 Å resolution crystal structure of the isolated N-terminal domain bound to Hb and small-angle x-ray scattering of free IsdH, reveal how multiple domains of IsdH cooperate to strip heme from Hb. Many bacterial pathogens obtain iron from human hemoglobin using proteins that contain multiple NEAT domains and other domains whose functions are poorly understood. Our results suggest that, rather than acting as isolated units, NEAT domains may be integrated into higher order architectures that employ multiple interaction interfaces to efficiently extract heme from host proteins.

Staphylococcus aureus is a Gram-positive bacterial pathogen that causes infections of the skin and invasive disease in many tissues and organs. S. aureus is the leading cause of surgical site infections, skin and soft tissue infections, and infective endocarditis (1). Methicillin-resistant S. aureus strains that are resistant to all ␤-lactam antibiotics cause hospital-and community-acquired infections with high mortality rates (2,3). New antibacterial treatments are urgently needed, but their development requires a better understanding of the mechanisms that underlie S. aureus pathogenesis.
Proteins displayed on the surface of pathogenic bacteria are at the frontline of the host-pathogen interface and interact with tissues of the host to carry out functions in adhesion, cell invasion, and acquisition of nutrients. Prominent among these functions is the capture of iron from the host (4). As well as being required at the active sites of many essential proteins, iron performs an important signaling role in pathogenesis and regulates the activation of nearly 400 S. aureus genes, including many required for colonization of the host (5). Under normal physiological conditions, all iron in the human body exists in complex with proteins. Successful pathogens have therefore evolved effective mechanisms to capture iron directly from these proteins. For example, Gram-negative pathogenic Neisseria species express the outer membrane transferrin-binding protein A (TbpA) and TbpB, which extract iron from serum transferrin (6). Hb is the most abundant iron source in humans, and many bacteria express proteins to capture this iron, such as iron-regulated surface determinant H (IsdH) 5 expressed on the surface of S. aureus.
In mammals, ϳ70% of the total body iron is iron protoporphyrin IX (heme) that is bound to Hb. S. aureus can obtain all of its iron requirements from Hb (7)(8)(9), which it gains access to by secreting hemolytic toxins that lyse erythrocytes. Hb is cap-tured onto the surface of S. aureus through the Hb receptors IsdB and IsdH (9 -13), and heme is extracted via an unknown mechanism. The captured heme is then transferred to heme carrier proteins, IsdA and IsdC, which relay heme to the transmembrane heme permease complex, IsdE/F (14 -18). IsdA, IsdB, IsdC, and IsdH are covalently attached to the bacterial peptidoglycan cell wall via a C-terminal amide linkage (19,20). Members of the Isd pathway are the most highly up-regulated genes in S. aureus in response to iron starvation (5, 19 -21), a condition that the bacteria experience during the early stages of an infection. Deletion of IsdH reduces S. aureus bacteremia in a mouse model (22), and deletion of IsdB protects against infections in heart (13,23) and kidney (9,24), indicating that Hb receptors have an important role in infection and that the Isd pathway might represent a therapeutic target for infections such as endocarditis, where S. aureus currently causes high mortality rates (25,26).
The domain architectures of S. aureus IsdH and IsdB ( Fig. 1) are similar, with an extended region of homology that includes all three recognized domains of IsdB, including two near iron transporter (NEAT) domains and a recently identified threehelix linker domain (27). A homologue of IsdB/H occurs in the related species Staphylococcus lugdunensis (28), which also causes infective endocarditis (29). Functionally related proteins that contain NEAT domains and capture heme from Hb are found in many Gram-positive bacteria including Bacillus anthracis (30 -33), Bacillus cereus (34), and Streptococcus pyogenes (35). The C-terminal NEAT domain of IsdB/H (Fig. 1, black boxes) and NEAT domains in IsdA and IsdC (not shown) bind heme with nanomolar to subnanomolar affinity (12, 18, 27, 36 -38). A transfer of heme between these domains occurs through formation of transient protein complexes (14, 16, 18, 36, 39 -41). IsdB/H contain a second class of NEAT domains that bind to the Hb protein, but do not bind heme (Fig. 1, white boxes). The two N-terminal NEAT domains of IsdH (IsdH N1 and IsdH N2 ) bind to the ␣ subunit of Hb with an affinity of 17-100 nM (10 -12), and IsdH N2 also binds ␤ Hb, albeit with a lower affinity (11). Together, these binding activities allow IsdB and IsdH to capture heme from Hb at up to 500ϫ the rate of simple heme dissociation from Hb chains (18,27,42). However, although the structures and ligand binding activities of the IsdH N1 , IsdH N3 , and IsdB N2 NEAT domains are known (11,12,27,39,43), the isolated domains do not release or capture heme from Hb, and the heme capture process is still poorly understood.
Because individual domains do not recapitulate Hb-receptor function, we have analyzed a functional three-domain fragment of IsdH (IsdH N2-N3 ; Fig. 1). Here we report the structures of IsdH N2 and IsdH N2-N3 in complex with Hb using x-ray crystallography, providing the first insights into how the domains of the Hb receptors of S. aureus cooperate to extract heme from Hb.

EXPERIMENTAL PROCEDURES
Protein Production-IsdH N2 (residues 321-467), human Hb, and individual Hb chains were obtained as described previously (11). For x-ray crystallographic and SAXS studies, IsdH (residues 326 -660) carrying a Tyr-642 to Ala mutation (IsdH N2-N3(Y642A) ) was expressed and purified as described previously (27). For heme transfer experiments, IsdH N2-N3 (residues 321-655) and IsdH N3 (residues 542-655) from S. aureus strain TCH1516 were cloned into pET15b (Novagen) for expression with an N-terminal hexahistidine tag. The proteins were expressed and purified to yield final products with the additional N-terminal sequence MGSSHHHHHHSSGLVPRGSHMLE. IsdH N2-N3 was purified over immobilized metal affinity chromatography resin (HIS-Select Nickel Affinity Gel, Sigma). The load condition was 50 mM sodium phosphate, pH 7.4, 500 mM NaCl, 20 mM imidazole. The bound protein was washed in equilibration buffer containing 25 mM imidazole and eluted in 100 mM imidazole. Additional purification was performed by anion-exchange (Q-Sepharose, GE Healthcare); proteins were loaded in 10 mM sodium phosphate, pH 7.0, and eluted over a gradient of 100 -250 mM NaCl. A final gel filtration step over a Superose 12 column (GE Healthcare Life Sciences) equilibrated in 150 mM sodium phosphate, pH 7.0, was performed.
X-ray Crystallography-Hb and IsdH N2 were mixed together in 1:2 molar ratio (5 mg/ml Hb) and crystallized by hanging drop vapor diffusion. Crystallization conditions included 0.2 M sodium formate, 0.1 M Bis-Tris propane, pH 7.5, 20% (w/v) PEG3350, and produced crystals of 50 -100 m. The crystals were cryoprotected with 30% glycerol and flash-cooled in a cold nitrogen stream (100 K). Diffraction data to 2.3 Å resolution were collected at 100 K with an x-ray beam wavelength of 0.95370 Å at the Micro Crystallography MX2 beamline at the Australian Synchrotron (Clayton, Australia). Data were indexed and scaled using HKL2000 and SCALEPACK (44), respectively. The structure was solved by molecular replacement using PHASER (45), which gave a unique solution when using ␣⅐␤ Hb dimer (Protein Data Bank (PDB) code 2DN1) and an alanine model of IsdH N1 (PDB code 3SZK) generated by CHAINSAW (46) as independent search models. The structure was refined using REFMAC5 (47), with manual map inspection and model building being performed in COOT (48). The quality of the model was regularly checked for steric clashes, incorrect stereochemistry, and rotamer outliers using MolProbity (49). The final structure had 98.57% of residues in the Ramachandran preferred region, with no outliers according to MolProbity. Coordinates and structure factors can be found at Research Collaboratory for Structural Bioinformatics (RCSB) PDB entry 4FC3.
The IsdH N2-N3(Y642A) ⅐Hb complex was buffer-exchanged into 20 mM HEPES, pH 7.5, and crystallized by hanging drop vapor diffusion. Crystals of 200 -500 m grew in 0.2 M diammonium citrate, 13% (w/v) PEG3350, 0.7% 1-butanol, pH 4.6, at 294 K. Crystals were cryoprotected in stabilizing solution containing 25% ethylene glycol before flash freezing in liquid nitrogen. Diffraction data to 4.2 Å resolution were collected at 100 K with an x-ray beam wavelength of 0.95369 Å on the Australian Synchrotron MX2 beamline. Integration and scaling were performed with HKL-2000 and SCALEPACK (44). Molecular replacement was performed with PHASER (45) using the crystal structures of Hb (PDB 3P5Q), IsdH N2 (PDB 4FC3), and IsdH N3 (PDB 2E7D and 2Z6F) as search models. Clear regions of helical density corresponding to the helical linker domain, which was left out of molecular replacement, provided confidence in the experimental model. The NMR structure of the linker (PDB 2LHR) (27) was placed manually into the model. Anomalous signal was observed at the predicted positions of the iron atoms in the heme groups coordinated by the four globin chains. Although the IsdH N2 and IsdH N3 domains are structurally similar, PHASER unambiguously placed them in unique locations within the complex. The solution was refined with Buster version 2.10.0 (50). Refinement was weighted heavily toward the geometry of the high-resolution structures used for molecular replacement, and included noncrystallographic symmetry restraints (51) and translation/liberation/screw refinement (groups were defined as single domains). Difference electron density joining the domains of IsdH indicated the positions of linking residues, but these could not be built with confidence and so were not included in the model. MolProbity (49) was used to verify the geometry, and the Ramachandran statistics are as follows; 96.2% of residues were found in favored regions; 99.9% were found in allowed regions; 0.06% were found in disallowed regions. Coordinates and structure factors can be found at RCSB PDB entry 4IJ2.
SAXS-Samples of IsdH N2-N3(Y642A) were buffer-exchanged by gel filtration to obtain matched buffer controls. SAXS data were collected on an Anton Paar SAXSess instrument with a sealed tube source. I(0) values and P(r) curves were calculated using GIFT (Anton Paar, Graz, Austria) and PRIMUS (72), and experimental molecular weights were calculated from the calibrated I(0) values (52). No significant aggregation or interparticle interference was found by Guinier or P(r) analysis. Ab initio shape reconstruction from the experimental scattering data was carried out using DAMMIF (53), and 10 independent models were averaged using DAMAVER (54). Although there was some variation between the models (normalized spatial discrepancy ϭ 0.983), no outliers were identified, and filtering of the average envelope to retain the most highly occupied area gave a dumbbell-shaped envelope, which was superposed with the IsdH N2-N3 crystal structure using SUPCOMB (55).
Heme Transfer-Heme transfer was monitored by UV-visible spectroscopy. Apo-IsdH N2-N3 was produced using the acid acetone heme extraction method of Ascoli et al. (56). Apo-IsdH N2N3(Y642A) was obtained heme-free upon purification from Escherichia coli. Hb samples were converted to ferric Hb (Hb containing ferric heme), a form of Hb that is produced following lysis of red blood cells, by the addition of a 5-fold molar excess of potassium ferricyanide and monitored using UV-visible spectrophotometry. Excess oxidant was removed by buffer-exchange over G-25 Sepharose (GE Healthcare). Hb was mixed with apo-IsdH N2-N3 at the ratios indicated under "Results." Reactions were performed in 150 mM sodium phosphate, pH 7.0, at 4°C. Absorbance spectra (350 -700 nm) were recorded at 40-s intervals on a JASCO UV-630 spectrophotometer equipped with a temperature-controlled sample chamber.
To determine the percentage of heme transferred from Hb to the Isd protein at each time point, the acquired UV-visible spectrum was fit to a linear combination of ferric Hb and fully hemeloaded holo-IsdH N2-N3 spectra.

IsdH N2 Binds to a Site Comprising Portions of the A and E
Helices of ␣ Hb-To begin to understand the function of IsdH N2-N3 , we crystallized an IsdH N2 ⅐Hb complex from a mixture containing a 2:1 molar ratio of IsdH N2 to Hb tetramer and determined the x-ray crystal structure to 2.3 Å resolution (Table 1 and Fig. 2). Phases were obtained by molecular replacement using the structures of Hb (PDB 2DN1) (57) and IsdH N1 (PDB 3SZK) (11). The asymmetric unit of the IsdH N2 ⅐Hb crystals comprises one Hb dimer with one IsdH N2 molecule bound to the ␣ Hb subunit ( Fig. 2A). IsdH N2 (Fig. 2B) binds in the same position on ␣ Hb as IsdH N1 (Fig. 2C), and a comparable number of residues at both interfaces participate in hydrogen bonds and salt bridges, including interactions with Asp-74, Lys-11, and Thr-8 of ␣ Hb (Fig. 2, B and C). A short ␣ helix in IsdH N2 (primary sequence FYHYAS), containing a series of aromatic side chains, forms part of the IsdH N2 ⅐␣ Hb-binding interface (Fig. 2D). A helix with similar hydrophobic character, but different primary sequence (YYHFFS), occurs in IsdH N1 (11,12). A Phe side chain from a different position in the IsdH N2 or  (11,27). In addition, the IsdH N2 -binding site is distant from the entrance to the globin heme pocket, making it unlikely to be directly involved in heme extraction. We conclude that IsdH N2 performs an Hb recognition/targeting role within the native IsdH receptor.
The Three Domains of IsdH N2-N3 Are Assembled into a Higher Order Structure That Binds to ␣ and ␤ Hb Chains and Positions the Globin Heme Pocket to Achieve Heme Transfer-To trap a stable IsdH N2-N3 ⅐Hb complex for x-ray crystallographic studies, we expressed IsdH N2-N3 with a Tyr-642 to Ala mutation in the heme pocket of the IsdH N3 domain (IsdH N2-N3(Y642A) ); this inhibits heme binding (27). IsdH N2-N3(Y642A) was mixed with ferric human Hb at a 2:1 molar ratio and subject to extensive crystallization trials. X-ray diffraction data were collected to 4.2 Å resolution using synchrotron radiation (Table 1), and the structure (Fig. 3A) was solved using molecular replacement by searching in a stepwise fashion for domains using the crystal structures of Hb (PDB 3P5Q) (58), IsdH N2 (PDB 4FC3), and IsdH N3 (PDB 2E7D and 2Z6F) (43) as search models.
In the IsdH N2-N3(Y642A) ⅐Hb complex, each ␣ or ␤ Hb globin chain is bound independently by one IsdH N2-N3(Y642A) receptor molecule (Fig. 3A). The three domains of the receptor are arranged in a dumbbell structure with the globular IsdH N2 and IsdH N3 domains separated in space by the intervening helical linker domain (Fig. 3B). This arrangement is consistent with interdomain interactions detected in an NMR spectroscopic analysis of the free receptor (27). Despite differences in the ␣ and ␤ Hb chain sequences, IsdH is clearly bound to equivalent sites on both globin subunits (Fig. 3C). On the ␣ subunit, the IsdH N2 domain docks in the same position in the IsdH N2-N3(Y642A) ⅐Hb and IsdH N2 ⅐Hb structures, suggesting that the Hb-targeting function of IsdH N2 is preserved in the context of the full IsdH receptor and providing independent validation of both structures. At the IsdH N2-N3(Y642A) ⅐␤ interface, the IsdH N2 domain binds at a site on the A and E helices of ␤ Hb, analogous to the binding site on ␣ Hb. Of the 15 residues that comprise the IsdH N2 -interacting face of ␣ Hb, there are three nonconservative substitutions in ␤ Hb: ␣Ala-5 is substituted by ␤Glu-6; ␣Asn-9 is substituted by ␤Ala-10; and ␣Lys-11 is substituted by ␤Thr-12. The latter two changes are expected to disrupt hydrogen-bonding interactions with IsdH N2 (Fig. 2B). Nevertheless, interactions between isolated IsdH N2 and ␤ Hb can be detected in gel filtration (11), suggesting that the IsdH N2-N3(Y642A) ⅐Hb crystal captures a weak mode of interaction that is mechanistically important (as described below).
In three of the IsdH receptor molecules, the helical linker domain positions the IsdH N3 domain directly over the heme pocket of the bound globin subunit. The electron density observed for the fourth IsdH N3 domain, which was expected to lie over the heme pocket of the other ␤ Hb subunit, was not sufficiently strong to place this IsdH N3 domain with confidence (Fig. 3A, dashed circle), and it was therefore not included in the model. The three complete IsdH N2-N3(Y642A) molecules in the complex have essentially identical structures (backbone r.m.s.d. Ͻ1.2 Å in pairwise comparisons), indicating that, in addition to binding the same site on ␣ and ␤ Hb, IsdH N2-N3 binds both globin chains in the same conformation (Fig. 3C). In this conformation, the IsdH N3 domain is positioned directly over the heme pocket on the globin, such that the heme would need to move only ϳ12 Å, through the entrance to the globin heme pocket, to effect transfer to IsdH (Fig. 3D, dashed arrow). Interestingly, electron density at the heme-binding site of the IsdH N3 domains (Fig. 3D, green), together with weak anoma-lous signal, suggests that the receptors have partial heme occupancy despite the inactivating Tyr-642 to Ala mutation. At the resolution of our x-ray data, however, it is not possible to say whether the globin heme pocket structure is altered by interaction with IsdH N3 . In addition, the heme-binding ␤ strands (residues Val-637-Gln-645) of IsdH N3 are not well defined in the electron density. As a result, we are not able to discern whether localized structural changes occur in Hb to promote heme transfer. Nevertheless, the close proximity of IsdH and Hb heme-binding sites in the IsdH N2-N3 ⅐Hb complex is consistent with a direct protein-to-protein heme relay.
The Three Domains of IsdH N2-N3 Are Pre-organized to Position the Heme Acceptor Site over the ␣/␤ Hb Heme Pocket-To investigate whether changes in IsdH N2-N3 conformation occur upon binding to Hb, we performed SAXS on the free IsdH N2-N3(Y642A) receptor. Model-free analysis of the SAXS FIGURE 3. Crystal structure of IsdH N2-N3 bound to Hb. A, four IsdH N2-N3 receptors bind to one Hb tetramer in the asymmetric unit of the crystal (PDB 4IJ2). Three of the IsdH N2-N3 receptors are complete, comprising IsdH N2 , linker, and IsdH N3 domains. The position of the IsdH N3 domain in the fourth receptor molecule is not well defined (the expected location is marked by a dashed ellipse). B, stereo diagram showing the 2F o Ϫ F c electron density map for one full IsdH N2-N3 receptor contoured at 1 . C, detail of the IsdH N2-N3 ⅐␣ (blue/orange) and IsdH N2-N3 ⅐␤ (teal/yellow) interaction complexes. Residues from the loop regions, which are not modeled in the structure, are shown with dashed lines. D, the IsdH N3 ⅐␣ interface with F o Ϫ F c difference map shown at 3 (green), indicating electron density that is not accounted for by atoms of the IsdH N2-N3 ⅐Hb model. The structure of heme-bound IsdH N3 (PDB 2Z6F, red) is superimposed to indicate the expected binding position of the heme group (lines), which coincides with a peak of F o Ϫ F c difference electron density in the IsdH N2-N3 ⅐Hb complex (green). Heme transfer from globin to the IsdH receptor requires a relatively small translation of only 12 Å between the E and F helices of the globin (dashed arrow). data yielded the expected molecular weights for all samples, which were within 10% of the theoretical molecular weights calculated from the sequences ( Table 2). We generated theoretical scattering curves for the three full receptors present in the complex with Hb using the program CRYSOL (59). These theoretical curves fit extremely well to the solution scattering of IsdH N2-N3(Y642A) , with values of 0.924, 0.904, and 0.866 (Fig.  4A), indicating that no major structural change occurs upon receptor binding and providing independent validation of the 4.2 Å resolution crystal structure. Ab initio shape reconstruction of the free IsdH receptor reveals close agreement with the crystal structure of the Hb-bound receptor (Fig. 4B). These results argue strongly that the receptor does not undergo a major conformational change upon binding to Hb. Instead, the domains of the receptor are pre-organized to position the heme acceptor site over the globin heme pocket.
IsdH N2-N3 Acquires Heme from ␣ and ␤ Hb-The structure of the IsdH N2-N3 ⅐Hb complex shows how IsdH could access heme in both the ␣ and the ␤ subunits of Hb. Physical interaction between IsdH N2 and isolated ␣ and ␤ Hb subunits has been demonstrated in solution, with the binding to ␤ Hb being considerably weaker (11). To investigate whether IsdH N2-N3 can remove heme from ␣ and ␤ chains of Hb, and hence probe the functional significance of ␤ chain interactions, we studied the transfer of the heme from ferric Hb to apo-IsdH N2-N3 using UV-visible spectroscopy. In the absence of heme, apo-IsdH N2-N3 shows negligible absorption from 350 -750 nm (Fig. 5A, dashed line). Holo-IsdH N2-N3 , when bound to ferric heme (Fig. 5A, dotted and  dashed line), has a characteristic spectrum that distinguishes it from ferric Hb (Fig. 5A, solid line). After mixing ferric Hb with apo-IsdH N2-N3 , UV-visible spectra were acquired as a function of time, and the fraction of ferric Hb and holo-IsdH N2-N3 at each time point was determined by least squares fitting to a linear combination of the spectra shown in Fig. 5A. At a mixing ratio of 5 apo-IsdH N2-N3 to 1 Hb tetramer, there was rapid and quantitative transfer of heme from Hb to IsdH N2-N3 , with only ϳ5% ferric Hb remaining after 10 min at 4°C (Fig. 5B, bottom,  dotted and dashed curve). As an alternative approach to assess the number of Hb heme groups accessed by IsdH N2-N3 , we measured the final UV-visible spectrum after a 10-min incubation of ferric Hb with different amounts of apo-IsdH N2-N3 . We observed a linear relationship between apo-IsdH N2-N3 :Hb molar mixing ratio and the total spectral change at 406 nm, up to a molar ratio of ϳ3.5:1 (Fig. 5C), which was consistent with a 4:1 IsdH N2-N3 :Hb interaction model. Together, these experiments show that heme is effectively removed from both ␣ and ␤ chains under these conditions. Although IsdH can quantitatively deplete heme from Hb, it is possible that heme is released into solution from destabilized Hb and scavenged by IsdH N2-N3 receptors that are not physically bound to Hb chains. To investigate whether a physical interaction between the IsdH receptor and Hb is required to extract heme from all sites (␣ and ␤ chains), we utilized the isolated IsdH N3 heme-binding domain. The isolated IsdH N3 domain is fully functional to bind heme from solution but is unable to capture heme from ferric Hb over the course of 10 min (Fig. 5B, gray circles), as shown previously by Spirig et al. (27). When IsdH N2-N3 was mixed with Hb in a 2:1 ratio, in the presence or absence of additional IsdH N3 , the heme transfer curves were identical, with ϳ50% of heme groups removed from Hb in each case (Fig. 5B, compare dashed and solid  curves), indicating that IsdH N3 only captures heme from Hb in the context of the intact receptor. The absence of any detectible competition for heme binding between IsdH N3 and IsdH N2-N3 confirms that interaction between free IsdH N3 and the globin heme pocket is extremely weak in the absence of the IsdH N2 domain. Together, the above data indicate that heme extraction from Hb (i) takes place directly within an IsdH⅐Hb complex, (ii) requires contacts mediated through the IsdH N2 domain, and  (iii) occurs at both ␣ and ␤ Hb subunits, as predicted by the IsdH N2-N3 ⅐Hb crystal structure.

DISCUSSION
We demonstrate here how multiple domains in IsdH cooperate to effect heme extraction from Hb. Excellent agreement between the IsdH N2-N3 ⅐Hb crystal structure and the solution x-ray scattering of the free receptor indicates that the three domains of IsdH N2-N3 are pre-organized to position IsdH N3 at the entrance to the globin heme pocket when the IsdH N2 domain docks with its cognate site between the A and E helices of ␣ or ␤ Hb. The isolated heme-binding IsdH N3 domain interacts with Hb too weakly to function in heme uptake, indicating that physical positioning of this domain in the context of the intact receptor is an essential feature of the heme transfer mechanism. A high level of sequence homology between IsdH N2-N3 , IsdB, and an IsdB homologue from the related pathogen S. lugdunensis (28) indicates that our results are relevant to all three proteins.
Heme is extracted from Hb and moved across the bacterial cell wall to the bacterial membrane in a series of direct protein-to-protein transfer steps (Fig. 6). Transport by protein-to-protein relay is frequently employed in nature to transfer reactive species or signaling molecules in both prokaryote and eukaryote cells. Examples include copper transport (60), electron transport (61), and relay of phosphoryl groups in signal transduction in plants (62), animals (63), and bacteria (64). In these examples, precise positioning of the donor and acceptor proteins is critical to cargo transfer, and both donor and acceptor proteins have co-evolved to perform efficient ligand transfer. In contrast, Hb has evolved to minimize heme dissociation, and partial unfolding of the globin is required in order for heme to enter/exit the globin heme pocket (65). We speculate that the challenge of extracting heme from the deep binding cleft in Hb cannot be met by interaction through a single interface. Hence, the IsdB/H receptors are anchored through a site distant from the heme pocket (Fig. 6, step 1), which holds the IsdH N3 domain in contact with Hb while at the same time allowing conformational changes in the heme pocket to take place (Fig. 6, step 2). Consistent with this idea, mutagenesis studies of the Hb receptor A (HgbA) from Gram-negative Haemophilus ducreyi indicate that there is likewise a physical separation of the Hb-binding and heme uptake interfaces (66,67). This similarity suggests that an architecture involving multiple interaction interfaces might have a functional advantage for heme extraction from Hb in a range of systems.
Protein-protein interactions that mediate transport or signaling are typically extremely transient in nature (60,61,64). In the Isd pathway, weak interactions (K d Ͼ 5 mM) are responsible for rapid heme transfer between S. aureus IsdA and IsdC (41). The structure of IsdH N2-N3 predicts that the IsdB/H receptors would need to dissociate from Hb to relay heme to IsdA/C (Fig.  6, step 3), and hence there is necessarily a balance between the IsdB/H⅐Hb complex persisting long enough for heme extraction, but not so long that transfer to IsdA/C becomes impractically slow. In this light, it is possible that weaker binding of IsdH to ␤ Hb, as compared with ␣ Hb, is an adaptation to the intrinsically more rapid heme dissociation from the ␤ subunits of Hb dimers/tetramers (42). Notably, IsdH contains an additional N-terminal Hb-binding domain, IsdH N1 , which is not present in IsdB. IsdH N1 and IsdH N2 bind to the same site on ␣ Hb and so may compete for binding under some circumstances. IsdH N1 displays no detectible interaction with ␤ Hb (11) and so is not expected to interfere with IsdH N2-N3 binding to ␤ Hb. The precise role that IsdH N1 plays, in the context of full-length IsdH, will now need to be established. Notwithstanding these factors, our data indicate that heme capture by IsdH and IsdB occurs by a similar mechanism for ␣ and ␤ chains. Kinetic data also support a single mechanism of heme uptake from ␣ and ␤ chains; Zhu et al. (18) showed that heme is almost completely transferred from Hb to IsdB in under 2 min at 22°C with a single rate constant.
In IsdH and IsdB, the same NEAT domain fold has evolved to specifically and exclusively bind either the surface of Hb or a heme molecule. The Pfam protein fold database identifies over 2000 NEAT domain sequences in Gram-positive species across the phylum Firmicutes, including many pathogens that cause severe human disease such as S. pyogenes, B. anthracis, C. perfringens, and L. monocytogenes, as well as S. aureus. Most NEAT was very similar to that of 2:1 IsdH N2-N3 :Hb. Error bars are Ϯ S.E. from two experiments. C, the plot shows absorbance at 406 nm, recorded after a 10-min incubation of apo-IsdH N2-N3 ⅐Hb, prepared at different molar mixing ratios. The linear relationship between spectral change and IsdH N2-N3 :Hb ratio is consistent with a model in which 3-4 IsdH N2-N3 molecules interact with four heme sites per Hb tetramer.
domains are identified as heme-binding modules, based on conservation of key residues involved in heme interactions. The most conserved NEAT-containing protein is IsdC. IsdC contains a single NEAT domain that delivers heme to the heme permease complex in the cytoplasmic membrane (IsdE/F), which is also highly conserved in Firmicutes, suggesting that IsdC/E/F are components of an ancestral heme-scavenging pathway. In contrast, upstream of IsdC, there is variation in the number of NEAT proteins and the domain architectures of these proteins in different bacterial species, reflecting diversification in the mechanisms for heme capture and heme relay to IsdC. NEAT domain proteins that target Hb as an iron source include IsdX1, IsdX2, and Hal from B. anthracis (30 -33), IlsA from B. cereus (34), and Shr from S. pyogenes (35,68). Each of these proteins contains one or more heme-binding NEAT domains and additional domains with poorly characterized function (Fig. 7). Interestingly, the diverse domain architecture suggests that heme capture from Hb may have evolved multiple times. Although the IsdH N2 and linker domain sequences appear to be unique to S. aureus and closely related species, different domains could play conceptually similar roles to target and support heme-binding NEAT domains. For example, in Shr, sequences adjacent to the N-terminal NEAT domain have been implicated in Hb binding (68). The structures and properties of these domains have not been determined. In addition, Shr, IlsA, and Hal contain leucine-rich repeat domains; these domains are well established as versatile protein interaction motifs (69). The secreted hemophore IsdX2 takes heme from Hb via four heme-binding NEAT domains (32,33). A fifth NEAT domain does not bind heme, but does interact with Hb (32), and so may play structural or Hb-targeting function in the context of the full-length IsdX2. IsdH N2-N3 is the first example from Gram-positive bacteria where a structure comprising multiple domains of an Hb receptor has been determined, revealing how these domains cooperate to achieve function.
The IsdH N2-N3 ⅐Hb structure predicts that IsdH can access iron from all forms of Hb that are likely to be encountered in serum following erythrocyte lysis. Dissociation of the Hb tetramer occurs in vivo following erythrocyte lysis, due to dilution effects and autooxidation processes. In addition, removal of heme from one or more globin chains is expected to destabilize Hb, leading to increased formation of monomers (27,42). Importantly, our results suggest that a similar mechanism of heme extraction will operate for tetramer, dimer, or monomer FIGURE 6. Model of heme extraction from human Hb by the S. aureus Hb receptors IsdB and IsdH. Hb is bound, through ␣ or ␤ subunits, by the Hb capture domain of IsdH/B (step 1). An ␣⅐␤ Hb dimer is shown here, but we envisage that the receptor could equally bind to Hb tetramer or individual ␣/␤ Hb monomers. Heme is transferred from the ␣/␤ chain to the heme acceptor domain of IsdH/B, which is positioned over the globin heme pocket (step 2). IsdH/B that dissociates from Hb in the heme-loaded form interacts transiently with, and transfers heme to, IsdA (step 3) (16,18) or IsdC (not shown). Hb may interact sequentially or simultaneously with other IsdH/B receptors until all heme groups are removed and transferred to IsdA/C (step 4). IsdC can receive heme from IsdA (36), or directly from IsdB/H (16,18), and delivers heme to the ligand-binding subunit (IsdE, step 5) (18) of the heme permease (IsdF), which shuttles heme to the cytoplasm (step 6). globin species. Hb released by hemolysis is also bound by the serum scavenger protein haptoglobin, which inhibits globin denaturation and targets Hb to the scavenger receptor CD163 on macrophages. IsdB/H binds Hb⅐haptoglobin complexes and the IsdH N1 domain is reported to bind haptoglobin and hemoglobin independently (7,10,12,22), but heme uptake from Hb⅐haptoglobin complexes has not been measured. The structure of an Hb⅐haptoglobin complex was solved recently (70), and comparison of that structure with the IsdH N2-N3 ⅐Hb complex shows that haptoglobin does not block IsdH N2 or IsdH N3 domain interactions with the globin, suggesting that Hb⅐haptoglobin is available as an iron source for S. aureus.
In summary, our data reveal the mechanism by which S. aureus, an important human pathogen, obtains the iron that is essential for infection. The pre-organized, multidomain architecture of the IsdH/B heme-capturing unit is a solution that couples highly specific recognition of Hb to a functional heme extraction module and might represent a general strategy exploited by a wide range of pathogenic bacteria.