Staphylococcus aureus Uses a Novel Multidomain Receptor to Break Apart Human Hemoglobin and Steal Its Heme*

Background: During infections, Staphylococcus aureus acquires heme-iron from human hemoglobin using the receptor proteins IsdH and IsdB. Results: A conserved multidomain unit in IsdH and IsdB synergistically captures heme and destabilizes the hemoglobin tetramer. Conclusion: Receptor domain synergy and hemoglobin dissociation allow efficient heme uptake by S. aureus. Significance: IsdH and IsdB may represent novel targets for antibiotics that limit microbial access to iron. Staphylococcus aureus is a leading cause of life-threatening infections in the United States. It requires iron to grow, which must be actively procured from its host to successfully mount an infection. Heme-iron within hemoglobin (Hb) is the most abundant source of iron in the human body and is captured by S. aureus using two closely related receptors, IsdH and IsdB. Here we demonstrate that each receptor captures heme using two conserved near iron transporter (NEAT) domains that function synergistically. NMR studies of the 39-kDa conserved unit from IsdH (IsdHN2N3, Ala326–Asp660) reveals that it adopts an elongated dumbbell-shaped structure in which its NEAT domains are properly positioned by a helical linker domain, whose three-dimensional structure is determined here in detail. Electrospray ionization mass spectrometry and heme transfer measurements indicate that IsdHN2N3 extracts heme from Hb via an ordered process in which the receptor promotes heme release by inducing steric strain that dissociates the Hb tetramer. Other clinically significant Gram-positive pathogens capture Hb using receptors that contain multiple NEAT domains, suggesting that they use a conserved mechanism.

into the cytoplasm. In the cytoplasm, the heme oxygenase IsdG or its paralog, IsdI, degrades the tetrapyrrole ring to release free iron for use by the bacterium (15). A molecular level understanding of the Isd system could facilitate the development of new anti-infective agents that work by disrupting heme uptake, because several studies have shown that its components are required for S. aureus virulence (12, 16 -19), and related systems are present in a number of other important pathogens, including Listeria monocytogenes (20,21), Bacillus anthracis (22), and Streptococcus pyogenes (23)(24)(25).
In the Isd system, both Hb and heme are captured by near iron transporter (NEAT) domains that are located within the IsdA, IsdB, IsdC, and IsdH proteins. These conserved binding modules are ϳ125 residues in length and are named for the location of their genes, which are proximal to putative Fe 3ϩ siderophore transporter genes (26). Biochemical studies of isolated NEAT domains indicate that they have distinct binding specificities that enable interactions with one or more distinct ligands, including heme, Hb, haptoglobin, and other host proteins. The atomic structures of several isolated NEAT domains have now been determined, revealing the mechanism of heme and Hb binding (27)(28)(29). In addition, recent studies have shown that heme transfer from IsdA to IsdC occurs when their NEAT domains transiently associate via an ultra-low affinity hand clasp complex (30,31).
The first step in heme acquisition is the capture of Hb and the extraction of its heme molecules by IsdH and IsdB. Both receptors are potential targets for the development of novel antibiotics because isdH and isdB mutant strains of S. aureus are reduced in their ability to infect mice (16,19,32,33), and purified antibodies against IsdH and IsdB confer protection from staphylococcal infections in animal models (33). IsdH and IsdB share a significant degree of primary sequence homology, and, unlike other components of the Isd system, they contain multiple NEAT domains. IsdB has been shown to bind Hb and capture its heme at least 150 times faster than the rate at which Hb spontaneously releases heme into the solvent, suggesting that the receptors capture heme via an activated receptor-Hb complex (14,34). Here we demonstrate that heme capture by IsdB and IsdH is mediated by a conserved structured unit that contains two NEAT domains that are connected by an ␣-helical linker domain. We show, based on absorbance spectroscopy and electrospray ionization mass spectrometry (ESI-MS) measurements, that the linker domain in IsdH forms a three-helix bundle structure that is essential for efficient heme capture. NMR studies of a 39-kDa polypeptide containing the conserved unit from IsdH indicate that it adopts an extended but ordered structure. A model of the heme extraction process is presented in which IsdH dissociates the Hb tetramer to promote heme release.

EXPERIMENTAL PROCEDURES
Cloning, Protein Expression, and Purification-Plasmids were generated encoding IsdH and IsdB receptor constructs as small hexahistidine-ubiquitin-like modifier (SUMO)-tagged proteins under control of an inducible promoter: pRM208 coding for amino acids 326 -660 in IsdH (IsdH N2N3 ), pRM213 coding for amino acids 326 -466 in IsdH (IsdH N2 ), pRM214 coding for amino acids 466 -660 in IsdH (IsdH linker-N3 ), pRM234 coding for amino acids 326 -543 in IsdH (IsdH N2-linker ), pRM219 coding for amino acids 467-543 in IsdH (IsdH linker ), and pRM221 coding for amino acids 544 -600 in IsdH (IsdH N3 ). Briefly, the DNA was amplified from the S. aureus RN4220 genome by polymerase chain reaction (PCR) and cloned into the vector pHis-SUMO using BamHI and XhoI restriction enzymes (35). pRM233 coding for amino acids 326 -660 (IsdH N2-GS-N3 ) was generated from pRM208 using two-step PCR, such that the linker was replaced with a nine-amino acid artificial linker (GSGSGSGSG). The sequence of all plasmids was verified by DNA sequencing. Generation of the plasmid pRM216 coding for amino acids 326 -660 in IsdH with an alanine substitution of Tyr 642 (Isd N2N3(Y642A) ) has been described earlier (36). Protein expression in Escherichia coli BL21(DE3) cells (New England BioLabs) transformed with the overexpression plasmids in LB/kanamycin (50 g/ml) was induced with 1 mM isopropyl-␤-D-thiogalactoside for 4 h at 37°C. For production of isotopically labeled [ 13 C, 15 N]protein, the cells were grown in M9 minimal medium containing 15 NH 4 Cl and [ 13 C]glucose (Cambridge Isotope Laboratories). The bacterial cells were harvested by centrifugation, resuspended in 50 mM NaH 2 PO 4 , 300 mM NaCl, pH 7.0, and ruptured by sonication. The cell debris was removed by centrifugation, and the supernatant containing the SUMO-tagged proteins was purified using a Co 2ϩ -chelating column (Thermo Scientific). After cleavage of the fusion proteins with ULP1 protease for 2 h at 4°C, they were reapplied to the Co 2ϩ chelating column to remove the protease and cleaved SUMO-affinity tag. The receptor proteins were further purified by gel filtration on a Superdex 75 column (GE Healthcare) equilibrated with 20 mM NaH 2 PO 4 , 50 mM NaCl, pH 6.0. Heme contents of holo-and apoproteins were determined with the pyridine hemochrome assay, and homogeneous apo forms of the heme-binding proteins were generated by extraction with methyl ethyl ketone (37). Expression and purification of [U-2 H, 13 C, 15 N]Isd N2N3(Y642A) was performed according to a previously published protocol (36).
Preparation of Human Hemoglobin-Human blood (30 -40 ml) was collected with heparin anticoagulant by a health practitioner following appropriate institutional protocols. Red blood cells were collected by centrifugation at 700 ϫ g for 10 min at 4°C. The cells were washed three times with 0.9% NaCl and bubbled with carbon monoxide (CO) for 5 min. The cells were then collected by centrifugation and lysed by resuspension in five volumes of water, followed by incubation on ice for 30 min. NaCl was added to a final concentration of 0.9%, resulting in aggregation of the membrane fractions into a gelatinous phase, which was removed by centrifugation for 15 min at 9500 ϫ g. The supernatant containing Hb was supplemented with 1 mM EDTA and bubbled with CO for 5 min. After adjusting the pH to 6.9, the hemolysate was applied to an SP Sepharose Fast Flow column (GE Healthcare) equilibrated with 10 mM NaH 2 PO 4 , 1 mM DTT, pH 6.9, and Hb was eluted with 10 mM Tris-HCl, pH 8.5. The fractions containing Hb were pooled, supplemented with 1 mM EDTA, and bubbled with CO for 5 min. Subsequently, the sample was applied to a Q Sepharose Fast Flow column (GE Healthcare) equilibrated with 10 mM Tris-HCl, pH 8.5. Pure Hb was eluted with 30 mM NaH 2 PO 4 , pH 6.9, with a yield of 100 mg of protein/ml of blood. Hb concentrations were determined using Drabkin's reagent (Sigma).
Electrospray Ionization Mass Spectrometry and Circular Dichroism Spectroscopy-Purified human Hb and IsdH proteins were prepared in 10 mM ammonium acetate buffer at pH 6.9, subsequently mixed to concentrations of 10 and 20 M, respectively, and incubated at 25°C for 1 h. MS measurements of protein samples were performed on a Waters Synapt G1 QTOF mass spectrometer (Waters Corp., Milford, MA) (35). The protein solutions were electrosprayed using Proxeon glass capillary nanoelectrospray emitters at flow rates between 30 and 50 nl/min. Quantification of Hb and Hb-receptor complexes was performed based on the Waters Synapt data by comparing summed peak heights. Higher resolution mass spectrometry experiments were performed using a 15-tesla Fourier transform ion cyclotron resonance instrument (SolariX hybrid Qq-FTMS, Bruker Daltonics, Billerica, MA). Circular dichroism spectra of 0.2 mg/ml IsdB linker and IsdH linker in 10 mM NaH 2 PO 4 , 50 mM NaF, pH 6.8, were recorded on a JASCO J-715 spectropolarimeter (JASCO Corp.) at 25°C with a scan rate of 20 nm/min.
Heme Transfer Kinetics and Affinity Measurements-Heme transfer reactions from Hb to IsdB N1N2 and various IsdH protein constructs were monitored by following absorbance changes using a conventional spectrophotometer (Shimadzu UV-1700 PharmaSpec), as described previously (14). Human hemoglobin was purchased from Sigma and dissolved in 20 mM NaPO 4 , pH 7.5, 150 mM NaCl. Briefly, 1 M holo-Hb (expressed in tetrameric units) was mixed with 10 M apo-receptor protein in 20 mM NaPO 4 , pH 7.5, 150 mM NaCl. Entire absorbance spectra were recorded for hemin transfer from holo-Hb to apo-IsdH N2N3 over time. To compare the heme transfer rates from Hb to the various acceptor proteins, changes in absorbance at 371 and 406 nm were recorded over time at 25°C for up to 2 h. Apparent rate constants for the heme transfer reactions were obtained by fitting the time courses of the absorbance changes ⌬A 406 -371 to single or double exponential curves with Sigma-Plot (Systat Software Inc.). Affinities of IsdH N3 and IsdH N2N3 for heme were determined by fluorescence spectroscopy as described (38).
NMR Spectroscopy and Solution Structure Determination-NMR spectra of IsdH linker were acquired at 25°C on cryoprobeequipped Bruker Avance 500-, 600-, and 800-MHz spectrometers. Backbone and side-chain chemical shift assignments were obtained by analyzing the following experiments: 1 H, 15  TALOSϩ was used to obtain a majority of the and dihedral angle restraints (40). Additional dihedral angle restraints were obtained by analyzing HNHA spectra (41). Stereo-specific assignments of methylene protons were obtained by analyzing HNHB and 15 N-edited TOCSY spectra. Distance constraints were identified in three-dimensional 15 N-and 13 C-edited NOESY spectra with mixing times of 125 and 130 ms, respectively. NMR spectra were processed using NMRPipe and analyzed using the CARA and PIPP software packages (42,43). The program UNIO was used for automated NOE assignments and structure determination of IsdH linker (44). The NOESY data were manually inspected to verify all NOE assignments and identify additional NOE restraints. The structures were improved by iterative rounds of structure calculations in which hydrogen bond restraints, backbone torsion angle restraints obtained from the program TALOSϩ, and side-chain torsion angle restraints were added (45). A final set of 100 conformers was generated with a standard simulated annealing protocol, as implemented in the program NIH-XPLOR, of which 51 had no NOE, dihedral angle, or scalar coupling violations greater than 0.5 Å, 5°, or 2 Hz, respectively (46). Of these, the 20 conformers with lowest overall energy were chosen to represent the structure of IsdH linker . The quality of the structural ensemble was evaluated with PROCHECK and visualized with PyMOL (47,48). Statistics for the linker structure are presented in Table 1. Details on the NMR experiments used to obtain the backbone chemical shifts of IsdH N2N3(Y642A) have been described (36). 1 D NH residual dipolar couplings were measured using protein samples partially aligned in PEG C12E5/hexanol using two-dimensional 15 N-coupled IPAP 1 H-15 N HSQC experiments. Steady-state 15 N heteronuclear NOE values for the IsdH linker and IsdH N2N3(Y642A) were acquired on cryoprobe-equipped Bruker Avance 600-and 800-MHz spectrometers, respectively. The heteronuclear NOE experiments were carried out in an interleaved manner, with and without proton saturation, and analyzed using the program SPARKY (49).

A Conserved Unit in IsdB and IsdH Containing Two NEAT Domains Rapidly
Captures Heme from Hb-The S. aureus Hb receptors IsdB and IsdH contain two and three NEAT domains, respectively (Fig. 1A). Isolated domains from these receptors have been characterized in vitro and bind to either Hb or heme; the IsdH N1 , IsdH N2 , and IsdB N1 NEAT domains bind to Hb, whereas the C-terminal NEAT domains in both proteins interact with heme (IsdH N3 and IsdB N2 ) (32, 38, 50 -54). Interestingly, a sequence alignment reveals that IsdB and IsdH share 64% primary sequence identity with one another over a region that encodes two NEAT domains (Figs. 1A and 2). This conserved unit contains two NEAT domains that are joined by a ϳ70-amino acid segment, hereafter referred to as the "linker." In IsdH, the unit corresponds to the N2 and N3 domains, which are homologous to the N1 and N2 domains in IsdB, respectively (Fig. 1A, enclosed in a dashed box). In vitro, full-length IsdB rapidly captures heme from Hb (14). To determine if the conserved unit within IsdB and IsdH is responsible for efficient heme capture, UV-visible absorption spectroscopy was used to measure the rate of heme transfer from heme-loaded methemoglobin (MetHb) to either IsdB N1N2 (residues Thr 121 -Asn 458 , containing the N1 and N2 domains in IsdB) or IsdH N2N3 (Ala 326 -Asp 660 , containing the N2 and N3 domains in IsdH). All studies were performed under oxidizing conditions, in which heme is in its ferric form. Upon mixing of MetHb with apo-IsdH N2N3 , a rapid shift of the UV absorbance spectrum of Hb to the heme bound spectrum of IsdH N2N3 is observed (Fig.  1B). This spectral change is indicative of heme transfer to IsdH and is most pronounced at 371 and 406 nm where the absorb-ance increases and decreases, respectively. A kinetic analysis of the heme transfer data indicates that IsdB N1N2 and IsdH N2N3 capture heme from Hb at similar rates, 0.055 Ϯ 0.001 and 0.048 Ϯ 0.001 s Ϫ1 , respectively (Fig. 1C). These rates are similar to those measured for intact IsdB and are up to 580 times faster than the rate at which tetrameric Hb spontaneously releases heme into the solvent, suggesting that heme transfer to IsdB and IsdH occurs via a MetHb-receptor complex (14,34).
To determine if the NEAT domains within IsdH need to be part of the same polypeptide to rapidly extract heme, we measured the rate of heme transfer from MetHb to the isolated N3 domain in the presence and absence of the Hb binding N2 domain (Fig. 3, A and B). Compared with IsdH N2N3 , the isolated heme binding domain IsdH N3 (Leu 544 -Asp 660 ) acquires heme from MetHb very slowly, which is consistent with an indirect transfer mechanism in which heme is first released from MetHb into the solvent before it is acquired by IsdH N3 (Fig. 3B). The addition of the MetHb binding IsdH N2 protein (Ala 326 -Pro 466 ) in trans to this transfer reaction fails to increase the rate of heme transfer from MetHb to IsdH N3 (Fig. 3B, N2 ϩ N3). This indicates that IsdH N2 binding to MetHb itself does not perturb its structure so as to promote heme release and subsequent capture by the isolated N3 domain. The presence of the N2 domain also does not significantly alter the heme binding affinity of the N3 domain within IsdH N2N3 , because IsdH N2N3 and IsdH N3 bind heme with similar affinities, K D values of 3.2 Ϯ 0.7 and 3.4 Ϯ 0.6 M, respectively (Fig. 4). Combined, these data strongly suggest that the N2 and N3 domains of IsdH need to reside within the same polypeptide to efficiently capture heme from Hb. Because IsdB contains this conserved unit, its NEAT domains also presumably synergistically extract heme (Fig. 2).
The NEAT Domains within IsdH and IsdB Are Connected by a Functionally Important Helical Linker-The ϳ70-amino acid linker segments that connect the NEAT domains in IsdB and IsdH share 70% sequence identity (Fig. 2). To investigate their structure, we purified polypeptides containing this segment from IsdB (IsdB linker , Ser 263 -Ser 361 ) and IsdH (IsdH linker , Pro 466 -Val 564 ). Their circular dichroism (CD) spectra indicate that IsdB linker and IsdH linker adopt a helical conformation, which is evident by negative bands in their CD spectra at 222 and 208 nm and a positive band at 193 nm (Fig. 5A). This is consistent with secondary structure predictions, which propose that amino acids in this region form several ␣-helices. To explore the functional role of the linker domain in IsdH, UVvisible absorbance spectroscopy was used to follow heme capture from MetHb. IsdH linker was unable to acquire heme from MetHb ( Fig. 3B). Moreover, the presence of IsdH linker and IsdH N2 did not accelerate the rate at which IsdH N3 captures heme from MetHb (Fig. 3C, N2 ϩ linker ϩ N3). This indicates that the isolated components of the conserved unit in IsdH are unable to associate with one another via non-covalent interactions to form a fully functioning receptor. To further investigate the function of the linker, polypeptides in which the linker was fused to either the N2 (IsdH N2-linker , Ala 326 -Gln 543 ) or N3 (IsdH linker-N3 , Leu 544 -Asp 660 ) domains were studied. Slow transfer from MetHb to the isolated N3 domain was observed when IsdH N2-linker was added in trans, indicating that MetHb binding by IsdH N2-linker did not significantly promote heme release and subsequent capture by IsdH N3 . Similarly, IsdH linker-N3 captures heme slowly from MetHb in either the presence or absence of IsdH N2 , indicating that the presence of the helical linker does not influence the N3 domain's ability to scavenge heme (Fig. 3C). To determine if the structure of the linker is important for function, we studied IsdH N2-GS-N3 , which replaces the linker with a nine-residue glycine-and serine-rich polypeptide (GSGSGSGSG). Spectroscopic measurements reveal that IsdH N2-GS-N3 captures heme slowly from MetHb at a rate that is similar to that of the isolated N3 domain Arrows indicate the increase and decrease in absorbance over time at 371 and 406 nm, respectively. C, time courses of ⌬A 406 -371 for the heme transfer reaction from Hb to IsdH N2N3 or IsdB N1N2 . The symbols and curves represent the observed data and the single-exponential fitting curves, respectively, yielding heme transfer rates of 0.048 Ϯ 0.001 s Ϫ1 and 0.055 Ϯ 0.001 s Ϫ1 for IsdH N2N3 and IsdB N1N2 , respectively. (Fig. 3, compare B and C). Combined, these data indicate that the NEAT domains in IsdB and IsdH are connected by a helical linker, whose primary function is to properly position the domains so as to specifically facilitate heme transfer from MetHb to the N3 domain.
IsdH Destabilizes Hb to Promote Heme Release-We used ESI-mass spectrometry to investigate the mechanism through which IsdH accelerates heme release from Hb. ESI-MS allows the quantification of different Hb oligomers in the presence and absence of IsdH (55). Hb consists of ␣and ␤-globin chains each bound to a heme. The globins assemble into a noncovalently bound (␣␤) 2 tetramer that dissociates into (␣␤) dimers with a dissociation constant (K D ) of 2 M (56). This is evident from the ESI-MS spectrum of a 10 M solution of Hb; from the ratio of the signal from the dimer and tetramer ions, the dimer/ tetramer ratio is 1:1.4. This is consistent with previously reported studies (57,58) and validates the use of ESI-MS to estimate the relative abundances of Hb species. ESI-MS spectra of Hb were acquired in the presence or absence of either wildtype IsdH N2N3 , IsdH N2-GS-N3 , or IsdH N2N3(Y642A) , which con-tains a Y642A mutation in the N3 domain that disrupts heme binding (Fig. 6, A-D). In all of the experiments, the receptors were present at a 2-fold molar excess relative to Hb (expressed in tetrameric units). The ESI-MS data are summarized in Fig.  6E, which shows a histogram plot of the relative abundances of the various forms of Hb (the sum of the monomeric ␣or ␤-globins (M); (␣␤) dimer (D); and (␣␤) 2 tetramer (T)), as well as receptor-bound forms of the (␣␤) 2 tetramer (T:R), and (␣␤) dimer (D:R). Incubation of IsdH N2N3 with Hb substantially reduces the amount of dimeric and tetrameric Hb, which is converted into monomeric globins. This is consistent with previous studies that have shown that Hb dissociates into its component globins upon heme removal (58,59). Because substoichiometric amounts of the receptor were used, Hb dissociation is not complete, leaving mostly a mixture of dimeric Hb and the (␣␤) dimer-receptor complex. Importantly, after the IsdH N2N3 addition, most of the Hb tetramer disappears, and very little (␣␤) 2 tetramer-receptor complex is formed. This indicates that receptor binding and/or heme removal significantly destabilizes the tetramer. A primary sequence alignment of IsdH (Q99TD3) and IsdB (Q7A656) was generated using ClustalW (69). Conserved residues are indicated by gray boxes. The predicted NEAT domains IsdH N1 , IsdH N2 /IsdB N1 , and IsdH N3 /IsdB N2 are highlighted by yellow, red, and green boxes, respectively. The IsdH and IsdB linker domains are indicated by blue boxes.
To gain insight into the role of the linker and heme binding in the acquisition process, ESI-MS spectra of Hb in the presence of IsdH N2-GS-N3 or IsdH N2N3(Y642A) were acquired. Unlike the wild-type receptor, when the IsdH N2-GS-N3 linker mutant is incubated with Hb, the majority of the receptor binds to the (␣␤) 2 tetramer to form a (␣␤) 2 -IsdH N2-GS-N3 complex, and a significant fraction of the tetramer remains intact (Fig. 6E). Moreover, smaller amounts of Hb are converted to its monomeric globins, whereas roughly similar amounts of (␣␤) dimer and (␣␤) dimer-receptor complex are present. The absence of monomeric globins is compatible with the kinetic data that showed that IsdH N2-GS-N3 extracted heme from Hb inefficiently. The fact that the linker mutant does not disrupt the tetramer suggests that it adopts a unique structure as compared with the wild-type protein, such that the mutant receptor can no longer impart sufficient structural strain to rupture the tetramer. To determine if the receptor needs to bind heme in order to dissociate the Hb tetramer, we studied IsdH N2N3(Y642A) , which contains a Y642A mutation in the N3 domain that disrupts heme binding. When Hb is incubated with IsdH N2N3(Y642A) , the amount of tetrameric Hb is significantly reduced as a result of its conversion into the (␣␤)-IsdH N2N3(Y642A) complex. However, only small amounts of monomeric globin are produced. This suggests that heme removal from the tetramer by the receptor is not required to dissociate it into its dimeric state. However, heme removal appears to be required to convert Hb into its monomeric units. A working model of the extraction process is presented under "Discussion." Structure of the Linker Domain-To gain a better understanding of the molecular basis of heme capture, we determined the NMR solution structure of IsdH linker (Protein Data Bank accession code 2LHR). The NMR spectra of IsdH linker are well resolved, enabling nearly complete 1 H, 13 C, and 15 N resonance assignments (Fig. 5B). A total of 1793 experimentally derived restraints were used to determine the structure, including 1469 interproton distance restraints, 118 dihedral angle restraints, 54 3 J HN␣ restraints, and 152 13 C secondary shift restraints. An ensemble of 20 conformers representing the structure of IsdH linker is displayed in Fig. 7A. The structure is well defined by the NMR data; the backbone and heavy atom coordinates of the structured residues Val 470 -Val 531 can be superimposed with a root mean square deviation of 0.42 Ϯ 0.10 and 0.87 Ϯ 0.07 Å, respectively (experimental and structural parameters are presented in Table 1).
The linker forms a three-helix bundle that is composed of helices ␣1 (Asp 471 -Lys 486 ), ␣2 (Leu 490 -Lys 503 ), and ␣3  (Glu 506 -Ala 530 ) (Fig. 7B). In the bundle, the long axes of the helices are co-linear and are connected by short reverse turns. The structure is stabilized by a hydrophobic core that is formed by nine leucine and tyrosine residues (Leu 477 , Leu 480 , Leu 481 , Tyr 484 , Leu 497 , Leu 500 , Leu 504 , Tyr 508 , and Tyr 512 ; Fig. 7C). Although each helix contributes residues to the hydrophobic core, helix ␣3 is longer than the other helices, such that its C terminus projects from the bundle. This region and residues immediately following it presumably facilitate interactions with the N3 domain in the intact receptor (see below). { 1 H} 15 N heteronuclear NOE measurements are compatible with the structure because residues Val 470 -Val 531 , whose coordinates are precisely defined in the ensemble, exhibit large magnitude NOE values, indicating that they are immobile on the picosecond time scale (Fig. 5C).
IsdH N2N3 Adopts an Extended but Ordered Multidomain Structure-We used NMR to investigate the structure and dynamics of IsdH N2N3(Y642A) . It is structurally identical to the wild-type protein based on its HSQC spectrum but is reduced in its ability to bind heme. Previously, we sequence-specifically assigned the chemical shifts of its backbone atoms (36). To learn whether the domains form a rigid unit within IsdH N2N3(Y642A) , we measured { 1 H} 15 N heteronuclear NOE relaxation parameters. As shown in Fig. 8A, residues spanning the N2, linker, and N3 domains exhibit positive and mostly uniform { 1 H} 15 N heteronuclear NOEs, which indicates that they are structurally ordered. Notably, residues that connect the domains also exhibit positive NOEs. Because some of these residues are unstructured in the isolated linker polypeptide (Fig. 5C), this suggests that in the context of IsdH N2N3(Y642A) , they form stabilizing interactions with residues located in the N2 and/or N3 domains and that the domains form a single structured unit.
Although the structure of the full receptor is unknown, the structures of the isolated linker and N3 domains are known, and the structure of the N2 domain can be accurately modeled using the previously determined NMR and crystal structures of IsdH N1 , which shares 54% sequence identity with N2 (50,51,53). To determine whether the domains undergo major structural changes upon incorporation into IsdH N2N3 , we measured 15 N-1 H residual dipolar couplings ( 1 D NH RDCs) in a sample of IsdH N2N3 partially aligned in pentaethylene glycol monododecyl ether (C 12 E 5 PEG)/hexanol. The 1 D NH data provide information about the angle of each backbone N-H bond relative to an alignment tensor. The compatibility of the individual domain structures with the RDC data was evaluated by plotting the back-calculated versus experimental 1 D NH values (Fig. 9). There is good agreement between the experimental data and the individual structures of the N2, linker, and N3 domains, which have calculated Q-factors of 0.28, 0.10, and 0.23, respectively. This indicates that incorporation of the domains into IsdH N2N3 does not significantly alter their structure and is consistent with our previously reported C ␣ and C ␤ backbone secondary chemical shifts of IsdH N2N3 , which suggested that the domains have similar secondary structures in isolation and when incorporated into IsdH N2N3 (36).
The chemical shifts of IsdH N2N3(Y642A) and polypeptides containing its isolated domains were compared with the aim of learning if the domains interact with one another in the context of IsdH N2N3(Y642A) . Fig. 8B shows an overlay of the secondary chemical shifts of IsdH N2N3(Y642A) and IsdH linker . Similar secondary chemical shifts were observed for the structured part of the linker, suggesting that its conformation is preserved in IsdH N2N3(Y642A) . Average chemical shift differences of the backbone amide signals of isolated linker and the corresponding residues in IsdH N2N3 are displayed in Fig. 8C. In general, small chemical shift differences were observed for residues in the core helices of the linker, indicating that they do not form a molecular surface that interacts with the N2 or the N3 domains. However, significant chemical shift differences in the linker occur for residues located at the beginning of helix ␣2 (Leu 490 -Arg 492 ) and at its N terminus (Asp 468 -Glu 472 , Thr 474 -Tyr 475 ) and C terminus (Gln 526 -Ser 529 , Val 531 -Thr 538 , Thr 540 -Gln 543 ). Mapping these changes onto the NMR structure of the linker reveals that they reside at distinct ends of the domain (Fig. 8D). This is consistent with residues at the beginning of helix ␣2 and the N terminus of the linker contacting the N2 domain, while residues at the C-terminal end interact with the N3 domain. Interestingly, comparison of the secondary chemical shifts suggests that helix ␣3 in the linker domain is lengthened at its C terminus when it is incorporated into IsdH N2N3 (Fig. 8B). Moreover, residues immediately following this segment, based on their secondary chemical shifts, do not participate in regular secondary structure when located in IsdH N2N3 but are nevertheless highly ordered, based on the heteronuclear NOE data (Fig. 8A). To further ascertain whether the domains in IsdH N2N3 might be significantly interacting with one another in IsdH N2N3 , we produced 15 N samples of IsdH N2 and IsdH N3 . The 1 H-15 N HSQC spectrum of IsdH N2 is well resolved and, when overlaid with the spectrum of IsdH N2N3 , reveals very similar chemical shifts (data not shown). This suggests that, in the context of IsdH N2N3 , the N2 domain does not contain a large contact surface that interacts with the remainder of the protein. A similar analysis using 15 Nlabeled IsdH N3 was also attempted but did not prove fruitful because the cross-peaks in its spectrum are partially broadened, presumably because of protein aggregation. Combined, the absence of extensive interaction surfaces in the linker and N2 domains suggests that, while ordered, IsdH N2N3 does not adopt a compact structure.

DISCUSSION
To successfully mount an infection, S. aureus and other pathogens acquire the essential nutrient iron from human Hb. Two surface-displayed S. aureus receptors capture Hb on the cell surface, IsdB and IsdH. The receptors share a high degree of sequence homology over a region that contains two NEAT domains that are separated by a ϳ70-amino acid "linker" segment (Figs. 1A and 2). The NEAT domains in the conserved units have distinct functions; in each protein, the N-terminal domain binds to Hb, and the C-terminal domain interacts with heme (8,38). Interestingly, the NEAT domains in IsdB appear to function synergistically, because Lei and colleagues (14) have shown that IsdB captures heme from Hb ϳ28 -250 times faster than proteins that contain only a single NEAT domain. To gain insight into the molecular basis of this synergy, we studied the conserved bi-NEAT domain unit located within IsdH (IsdH N2N3 ). UV-visible spectroscopy measurements of heme transfer from Hb indicate that IsdH N2N3 rapidly acquires the heme of Hb at a rate that is 110 -580 times faster than the rate at which Hb spontaneously releases heme into the solvent (IsdH N2N3 acquires heme at a rate of 0.048 Ϯ 0.001 s Ϫ1 , whereas the ␣ and ␤ subunits in tetrameric Hb release heme into the solvent at a rate of ϳ0.000083 and ϳ0.00042 s Ϫ1 , respectively) (34). IsdB and IsdH N2N3 capture heme from Hb at a similar rate, compatible with both proteins forming a receptor-Hb complex in which heme is actively removed. These transfer rates may be slower if the heme iron in Hb is in its reduced state, because IsdH N3 has been shown to bind ferric heme more tightly than ferrous heme (60). Systematic dissection of IsdH N2N3 into its components indicates that its NEAT domains need to be part of the same polypeptide chain in order to rapidly acquire heme from Hb. Moreover, a linker with a specific structure and size   that connects the domains is required for efficient heme capture; an IsdH N2-GS-N3 mutant in which the linker is replaced with a glycine-serine nonapeptide acquires heme slowly from Hb. IsdH N2N3 adopts an ordered elongated dumbbell-shaped structure in which its NEAT domains are separated by a helical linker domain. The NMR structure of the linker domain (called IsdH linker ) reveals that it adopts a three-helix bundle. First observed in the IgG-binding domain of S. aureus, three-helix bundles serve as robust scaffolds for molecular recognition and are ubiquitously found in structural proteins, enzymes, and DNA-binding proteins (61,62). Because the N and C termini in IsdH linker are positioned at opposite ends of the bundle, in the context of the IsdH N2N3 receptor, the linker domain presumably acts as a spacer that holds the N2 and N3 domains apart from another by ϳ40 Å. This is compatible with the assigned NMR spectra of the intact 39-kDa IsdH N2N3 receptor, because a comparison with the NMR spectra of IsdH linker reveals that only residues located at the ends of the helical bundle near the connection points to the N2 and N3 domains exhibit large chemical shift differences. Moreover, the NMR chemical shifts of residues in the isolated IsdH N2 domain and IsdH N2N3 are similar, suggesting that N2 is not involved in extensive interdomain interactions in the structure of IsdH N2N3 . Interestingly, although IsdH N2N3 adopts an elongated structure, the domains do not appear to be connected by flexible loops. Inspection of the heteronuclear NOE data of IsdH N2N3 reveals nearly uniform values over the length of the polypeptide, including amino acids that connect the domains. Notably, several residues at the N and C termini of the linker domain that are unstructured in the isolated IsdH linker become ordered when they are located in IsdH N2N3 (in IsdH N2N3 , 2 and 11 residues preceding and following the linker domain, respectively, exhibit elevated NOE values in IsdH N2N3 as compared with IsdH linker ). Thus, the three domains within IsdH N2N3 adopt an extended conformation in which their positioning is fixed with respect to one another. IsdB can be assumed to adopt a similar structure because it shares significant sequence homology with IsdH N2N3 , and we have shown that its linker region also adopts a helical conformation.
From the ESI-MS data, IsdH N2N3 extracts heme from Hb via the ordered process shown in Fig. 10A. On the cell surface, IsdB and IsdH can be expected to encounter Hb in its (␣␤) 2 tetrameric and ␣␤ dimeric forms, whose relative abundance depends on protein concentration. When IsdH N2N3 binds to the (␣␤) 2 tetramer, it promotes its dissociation into ␣␤ dimers, which is presumably caused by receptor-induced steric strain that ruptures the weaker ␣ 1 ␤ 2 interface of the tetrameric Hb (63). Dimer formation is expected to facilitate heme transfer to IsdH N2N3 because dimeric Hb releases heme more readily than the (␣␤) 2 tetramer; compared with the tetramer, the rate of heme loss from the ␣ and ␤ chains in the isolated (␣␤) dimer is 2 and 10 times faster, respectively (34). In the second step, heme is transferred from the (␣␤) dimer to the N3 domain within the IsdH N2N3 receptor. Our data do not reveal which globin chain, if any, serves as the preferred heme donor for IsdH N2N3 . It is possible that heme is first removed from the ␤ subunit because it has intrinsically weaker affinity for heme as compared with the ␣ subunit (34). Alternatively, structural distortions induced in the dimer by the receptor may trigger heme transfer from the ␣ chain, creating semi-␤ Hb from which heme is known to be rapidly released (34). In the final step, after the loss of one of its heme molecules, the (␣␤) dimer dissociates completely. Formation of monomeric species is probably driven by the greater tendency of Hb dimers to dissociate (59). As the monomeric ␣ and ␤ chains quickly lose their heme to the environment, both globins could be expected to readily release their ligand to IsdH (34). A similar transfer reaction is expected to occur when IsdH encounters an (␣␤) Hb dimer, but it would bypass the need for tetramer dissociation. An alternative heme transfer pathway is also possible. In it, the receptor would remove heme directly from the tetramer or concurrently with tetramer dissociation. Heme removal from the tetramer could be advantageous because it would produce semi-Hb tetramers that are prone to dissociate (64). However, as described immediately below, heme capture from the Hb tetramer is not an obligate step in the transfer reaction.  A schematic diagram shows the binding equilibria involved in the extraction process. Wildtype IsdH N2N3 binds to the ␣ chain of Hb promoting its dissociation into (␣␤) dimers. Heme acquisition by the receptor protein results in further dissociation of Hb into its monomeric subunits. See "Discussion" for details. B, a model of IsdH N2N3 in complex with Hb. IsdH N2 (red) was modeled based on the solution structure of IsdH N1 (Protein Data Bank entry 2H3K). The complex model with Hb was generated by superposition over the crystal structure of the IsdH N1 -Hb complex (Protein Data Bank entry 3SZK). A possible orientation of the linker (blue) and IsdH N3 (green; Protein Data Bank entry 2Z6F) allowing productive heme transfer from a Hb (␣␤) dimer (yellow-orange) to IsdH is indicated. The orientation of the subdomains (N1, linker, and N2) within IsdH N2N3 has not been experimentally determined, and only one possible orientation is shown. The protein backbones are shown as schematics. The heme groups in Hb are shown in stick representation.
Several lines of evidence indicate that binding of the IsdH N2N3 receptor to tetrameric Hb induces steric strain in Hb that causes it to dissociate into dimers and that this process does not require heme transfer to IsdH N2N3 (Fig. 10A). The most compelling evidence comes from the ESI-MS data of IsdH N2N3 and IsdH N2N3(Y642A) , which indicate that both proteins readily disrupt the tetramer. Because IsdH N2N3(Y642A) binds heme with lower affinity, this indicates that structural perturbations in Hb induced by receptor binding are sufficient to cause it to dissociate. This process requires two NEAT domains that are connected by a structured linker because the Hb tetramer does not dissociate when it is bound to an IsdH N2-GS-N3 mutant in which the linker domain is replaced with a flexible glycine-serine peptide. The idea that an intact bi-NEAT domain receptor is required to dissociate the tetramer is also consistent with a recent crystal structure of the IsdH N1 -Hb complex, which revealed that binding of the isolated N1 NEAT domain to Hb induced only modest structural changes in Hb (50). As we have shown, IsdH N2N3 adopts a rigid structure in its apo state; this suggests that binding of IsdH N2N3 to Hb results in atomic overlap between the proteins that causes the tetramer to dissociate. A model of the structure of the IsdH N2N3 -Hb complex illustrates a possible orientation of the receptor protein on Hb (Fig. 10B). The orientation of the subdomains (N1, linker, and N2) within IsdH N2N3 has not been experimentally determined, and only one possible orientation is shown. The model was constructed using the NMR structure of IsdH linker , the crystal structure of the isolated N3 domain, and a homology model of the N2 domain based on the structure of IsdH N1 . Based on the recently reported crystal structure of the isolated N1 domain bound to Hb, the N2 domain in IsdH N2N3 can be expected to engage the ␣ subunit of Hb via its A-helix (50). Contacts from N2 presumably originate from residues located within surface loops positioned at one end of its ␤-barrel structure because these residues are conserved in N1 and N2. The relative positioning of the remainder of the IsdH N2N3 protein and its contacts to Hb cannot be predicted from our NMR data. However, assuming that IsdH N2N3 adopts an extended structure, the N3 domain could, in principle, be positioned adjacent to the heme pockets of either the ␣ or ␤ subunits. Unlike IsdB, the IsdH protein contains an N-terminal NEAT domain (N1) that binds to the ␣ subunit of Hb (Fig. 1A) (50). It is possible that the N1 and N2 domains in IsdH simultaneously engage the Hb tetramer via its two ␣ subunits. Alternatively, N1 and N2 may not simultaneously engage the same tetramer. In this scenario, Hb binding by N1 may function to increase the efficiency of heme capture by increasing the local concentration of Hb that is proximal to IsdH N2N3 . A more detailed understanding of the mechanism of extraction and the origin of molecular strain induced by the receptor on Hb will require studies of the full-length IsdH protein and the structure determination of IsdH N2N3 in both its free and Hb-bound states.
We have demonstrated that the NEAT domains within IsdH function synergistically to capture heme from Hb. Interestingly, several other pathogenic species of Gram-positive bacteria display surface proteins implicated in heme capture that contain more than one NEAT domain (26). At present, only a few of these proteins have been characterized biochemically. S. pyogenes encodes the membrane-anchored Shr protein, which has two NEAT domains, and, similar to IsdB and IsdH, it has been proposed to acquire heme via a receptor-Hb complex (65,66). B. anthracis produces a Hb hemophore called IsdX2 that contains five NEAT domains (67,68). All of its domains bind Hb, and some are multifunctional because they can also bind heme. It will be interesting to see if subsets of these domains are also connected by structured linker segments that enable their NEAT domains to function synergistically. Despite the prevalence and importance of multi-NEAT domain proteins in Gram-positive bacteria, this present study is the first to address in detail the possible interactions between NEAT domains, the role of the linker segments, and functional synergy between these regions. Further research will be required to reveal if the mechanism of extraction described here can be generalized to other NEAT-containing Hb receptors. This work could lead to small molecule antibiotics that work by limiting microbial access to heme-iron.