The Streptococcus pyogenes Shr protein captures human hemoglobin using two structurally unique binding domains

In order to proliferate and mount an infection, many bacterial pathogens need to acquire iron from their host. The most abundant iron source in the body is the oxygen transporter hemoglobin (Hb). Streptococcus pyogenes, a potentially lethal human pathogen, uses the Shr protein to capture Hb on the cell surface. Shr is an important virulence factor, yet the mechanism by which it captures Hb and acquires its heme is not well-understood. Here, we show using NMR and biochemical methods that Shr binds Hb using two related modules that were previously defined as domains of unknown function (DUF1533). These hemoglobin-interacting domains (HIDs), called HID1 and HID2, are autonomously folded and independently bind Hb. The 1.5 Å resolution crystal structure of HID2 revealed that it is a structurally unique Hb-binding domain. Mutagenesis studies revealed a conserved tyrosine in both HIDs that is essential for Hb binding. Our biochemical studies indicate that HID2 binds Hb with higher affinity than HID1 and that the Hb tetramer is engaged by two Shr receptors. NMR studies reveal the presence of a third autonomously folded domain between HID2 and a heme-binding NEAT1 domain, suggesting that this linker domain may position NEAT1 near Hb for heme capture.

In order to proliferate and mount an infection, many bacterial pathogens need to acquire iron from their host. The most abundant iron source in the body is the oxygen transporter hemoglobin (Hb). Streptococcus pyogenes, a potentially lethal human pathogen, uses the Shr protein to capture Hb on the cell surface. Shr is an important virulence factor, yet the mechanism by which it captures Hb and acquires its heme is not well-understood. Here, we show using NMR and biochemical methods that Shr binds Hb using two related modules that were previously defined as domains of unknown function (DUF1533). These hemoglobin-interacting domains (HIDs), called HID1 and HID2, are autonomously folded and independently bind Hb. The 1.5 Å resolution crystal structure of HID2 revealed that it is a structurally unique Hb-binding domain. Mutagenesis studies revealed a conserved tyrosine in both HIDs that is essential for Hb binding. Our biochemical studies indicate that HID2 binds Hb with higher affinity than HID1 and that the Hb tetramer is engaged by two Shr receptors. NMR studies reveal the presence of a third autonomously folded domain between HID2 and a heme-binding NEAT1 domain, suggesting that this linker domain may position NEAT1 near Hb for heme capture.
Streptococcus pyogenes (Group A Streptococcus) each year causes an estimated 700 million infections worldwide (1). Although most of these infections lead to nonlife-threatening acute pharyngitis, many are lethal and cause ϳ500,000 deaths annually (2). These more severe and invasive infections have a high mortality rate (ϳ30%) and include necrotizing fasciitis and streptococcal toxic shock syndrome (3,4). Infections caused by this deadly microbe are presently among the top 10 causes of death from an infectious disease worldwide (5). S. pyogenes and other pathogenic bacteria require iron to proliferate, as it is a versatile metal that functions as a key biocatalyst and electron carrier in enzymes that mediate metabolism (6). During infections, iron is frequently foraged from human hemoglobin (Hb), 3 which contains ϳ75-80% of the body's total iron content in the form of heme (iron-protoporphyrin IX). Recent studies have shown that S. pyogenes uses an array of surface and membrane-associated proteins to acquire Hb's heme. Understanding how these proteins function at a molecular level is of fundamental importance and could facilitate the discovery of new anti-infective agents that work by limiting microbial access to iron.
Surface and membrane-associated proteins are used by S. pyogenes to capture heme. The streptococcal hemoprotein receptor (Shr) is displayed on the bacterium's surface, where it binds to Hb and acquires the oxidized form of heme (called hemin) (7)(8)(9)(10)(11). The Shr Hb receptor is an important virulence factor, as ⌬shr mutant strains of S. pyogenes exhibit reduced growth in human blood and attenuated virulence in murine and zebrafish models of infection (10,12). In vitro, Shr transfers hemin directly to Shp, a cell wall-associated hemoprotein that uses a bis-methionine ligation mechanism to interact with hemin (13)(14)(15). Shp then passes hemin to HtsA/SiaA, the lipoprotein component of the ABC transporter HtsABC/SiaABC (16). A binary Shp-HtsA complex mediates this transfer step, in which the methionine axial ligands in Shp are displaced by methionine and histidine axial ligands within HtsA (17)(18)(19)(20)(21). This process occurs very rapidly, Ͼ100,000 times faster than the rate at which Shp spontaneously releases hemin into the solvent. Hemin is then presumably transported across the membrane into the bacterial cytoplasm where it is degraded by a yet to be identified oxygenase(s) that releases hemin's iron for use by the microbe (6). Although hemin transfer from Shp to HtsA/SiaA has been extensively studied by Lei and co-workers (20), the first step in the pathway, the extraction of hemin from human Hb by Shr, is less well-understood and the focus of this study.
Many Gram-positive bacterial pathogens display Hb receptors that contain near-iron transport (NEAT) domains (22,23). NEAT domains have been shown to bind Hb, hemin, and in many instances are able to rapidly transfer hemin among one another (24 -35). Hemin acquisition by Staphylococcus aureus has been studied extensively. This microbe uses an array of iron-regulated surface determinant (Isd) proteins that each contain one or more NEAT domains (24, 29, 36 -40). Hb is captured on the cell surface by the closely related IsdB and IsdH proteins (41). Our studies have shown that IsdH captures and extracts hemin from Hb using a conserved tri-domain unit that contains two NEAT domains, called N2 and N3, that are separated by a helical linker domain (42). As compared with the rate of hemin released into the solvent by Hb, binding of the tridomain unit accelerates hemin release 13,400-fold (43). Atomic structures of IsdH-Hb complexes reveal that its N2 and N3 NEAT domains adopt related ␤-sandwich structures that are decorated by short helices, whereas the intervening linker domain adopts a three-helix bundle (44,45). The N2 domain engages Hb's A-helix with high affinity, enabling the helical and N3 domains to distort Hb's F-helix, thereby promoting hemin transfer to IsdH's N3 domain (43,45,46). IsdB shares significant primary sequence homology with IsdH and also contains a tri-domain unit that captures hemin from Hb through a similar mechanism (31,34,42,(47)(48)(49)(50). Other Gram-positive pathogens use NEAT domains to bind hemin and/or Hb, and include among others Listeria monocytogenes, Bacillus anthracis, Bacillus cereus, and S. pyogenes (23). However, the mechanisms used by these domains to capture hemin from Hb may be distinct from that in IsdH and IsdB, because in some instances a single NEAT domain has been proposed to bind to both Hb and heme (51)(52)(53)(54). Notable examples are IsdX1, IsdX2, and Hal in which single NEAT domains from these proteins have been reported to bind Hb and extract hemin from Hb (52)(53)(54). However, complexes of these NEAT domains bound to Hb have not been visualized at atomic resolution.
Unlike other Hb receptors in Gram-positive bacteria, the S. pyogenes Shr protein does not use NEAT domains to engage Hb. Instead, it binds Hb via an N-terminal region (NTR) that is predicted to contain two domains of unknown function, called DUF1533 domains ( Fig. 1) (8,11). Released hemin is then bound by residues located within Shr's C-terminal region (CTR) (residues 365-1275), which contains two heminbinding NEAT domains (NEAT1 and NEAT2) that are separated from one another by a series of leucine-rich repeats (8). Interestingly, similar to the staphylococcal IsdB and IsdH proteins, transfer experiments have shown that the isolated Shr protein increases the rate of hemin release from Hb in vitro, suggesting that the domains within the receptor function synergistically to extract hemin (9). The NEAT domains within Shr can rapidly transfer hemin among one another, but only NEAT1 is able to efficiently transfer hemin to Shp (55). These findings led to the suggestion that NEAT2 stores hemin near the cell surface (55).
Here, we show that Shr uses two related DUF1533 domains to engage Hb, which we rename Hb-interacting domains (HIDs). Biochemical studies and the atomic structure of a HID reveal that it is a structurally novel Hb-binding module that employs a conserved tyrosine residue for Hb recognition. Biochemical studies indicate that the HIDs within Shr bind Hb with differing affinities and suggest that each engages a single globin subunit. Interestingly, NMR studies reveal the presence of a third autonomously folded domain that is located between the HID2 and NEAT1 domains. The implications of this arrangement in hemin acquisition from Hb are discussed.

Hb-interacting N-terminal region within Shr contains three autonomously folded domains
Previous studies have shown the S. pyogenes Shr protein interacts with Hb via residues located at its N terminus (8). This N-terminal region is formed by residues Lys-25-Val-364 and is hereafter referred to as the NTR (Fig. 1). An inspection of its primary sequence suggests that the NTR contains three folded regions. These include two domains of unknown function (DUF1533 domains) that are formed by residues Gly-61-Phe-123 and Ile-203-Val-269 (8), as well as residues Val-299 -Ser-358 that are predicted to form several ␣-helices. Constructs containing residues Gly-61-Phe-123 and Ile-203-Val-269 were insoluble. Secondary structural prediction algorithms such as PSIPRED (56) and JPred (57) were used to guide the design of additional constructs until highly soluble ones were obtained and characterized by NMR: Shr HID1 , residues Ser-26 -Lys-148 that correspond to the first DUF1533 domain; Shr HID2 , residues Lys-155-Gln-285 that correspond to the second DUF1533 domain; and Shr L , residues Ser-294 -Val-364 that contain the predicted ␣-helical region in Shr that links Shr HID2 to the first NEAT domain (Fig. 1). In each case, constructs included at least two residues beyond the predicted secondary structure elements at both the N and C termini. If ambiguous, extra residues were included to ensure that no potentially important structural elements were truncated. Each polypeptide was uniformly 15 N-labeled and purified, and its 1 H-15 N HSQC spectrum was recorded ( Fig. 2, a, c, and e). All three polypeptides adopt a folded structure, as the cross-peaks in each spectrum are well-dispersed and have appropriate line Previous studies have shown that Shr contains an Hb-binding NTR and a hemin-binding CTR. Polypeptides used in this study contain regions within the NTR and include: Shr HID1 and Shr HID2 that contain the first and second DUF1533 domains (renamed HIDs), respectively; Shr L , which is predicted to contain several ␣-helices; and Shr HID1-2 , which contains both of the DUF1533 domains. The DUF1533 domains are HIDs based on findings reported in this paper.

Identification of the hemoglobin binding domains in Shr
widths. Based on the Pfam prediction for their domain boundaries, the two DUF1533 domains within the NTR are separated by ϳ80 amino acids. To determine whether the domains interact with one another, a longer polypeptide fragment containing both domains was characterized by NMR (Shr HID1-2 , consisting of residues Ser-26 -Gln-285). Superposition of the NMR spectra of [ 15 N]Shr HID1 and [ 15 N]Shr HID2 onto the spectrum of [ 15 N]Shr HID1-2 reveals that the cross-peaks have similar chemical shifts (Fig. S1). Thus, in the context of the longer Shr HID1-2 polypeptide, the Shr HID1 and Shr HID2 domains do not significantly interact with one another, and they exhibit the same structural fold as they do in isolation. The Shr HID2 and Shr L domains also do not interact with one another, as the NMR spectra of a polypeptide containing both of these domains ([ 15 N]Shr HID2-L , residues Gly-61-Asn-368) can be recapitulated by summing the spectra of the isolated [ 15 N]Shr HID2 and [ 15 N]Shr L domains (Fig. S1). We conclude from the NMR data that the Hb-interacting region within Shr contains three independently folded and noninteracting structural domains, two DUF1533 domains, and a linker domain that is predicted to contain several ␣-helices.

DUF1533 domains bind Hb
We used NMR to determine whether the domains within the NTR interacted with Hb. The 1 H-15 N HSQC spectrum of each 15 N-labeled domain was recorded in the presence or absence of Hb (1:1 stoichiometry based on tetramer units) (Fig. 2 (Fig. 2, compare a and c with b and d), indicating that both DUF1533 domains interact with Hb. In contrast, adding Hb to the [ 15 N]Shr L domain causes only minimal changes in the NMR spectrum indicating that the proteins do not interact (Fig. 2f).
Having established that DUF1533 domains bind to Hb, we next measured their affinities using ITC (Fig. 3, a-d, and Table  1). These data indicate that Shr HID2 binds Hb with a K D ϭ 16.0 Ϯ 3.4 M in a process that is enthalpically favorable (Fig. 3b and Table 1). Shr HID1 also binds Hb, but the interaction is weaker (K D ϭ ϳ140 M). The Shr HID1-2 polypeptide containing both DUF1533 domains binds Hb with a K D ϭ 5.1 Ϯ 0.4 M ( Fig. 3a and Table 1). Considering that the binding parameters for Shr HID1-2 are distinct from either HID domain in isolation, it seems likely that Shr HID1-2 exhibits binding avidity. If HID1 and HID2 were binding completely independently, one would expect the Shr HID1-2 -Hb binding parameters to more closely resemble the higher affinity Shr HID2 -Hb interaction. From these data, we conclude that both of Shr's DUF1533 domains bind Hb, and the second domain (Shr HID2 ) is the most important for binding. Based on these results, we propose that the DUF1533 domains within Shr be renamed hemoglobin interacting domains (HIDs).
A NMR experiment was performed to ascertain whether the HIDs (DUF1533 domains) bind to the same site on Hb. In this experiment, the ability of unlabeled Shr HID2 to displace Hbbound [ 15 N]Shr HID1 was determined. Initially, unlabeled Hb was titrated into a solution of [ 15 N]Shr HID1 such that the signals in its 1 H-15 N HSQC spectrum were broadened beyond detection as a result of complex formation (Fig. 4b). Unlabeled Shr HID2 was then titrated into the solution, and the 1 H- 15 (Fig. 4c). When 4-fold Shr HID2 is added, the spectrum of [ 15 N]Shr HID1 is almost completely recovered, yet the signal intensities are approximately one-third as intense as in the original HSQC (Fig. 4d). Importantly, the reappearance of the NMR signals is not caused by simply diluting the NMR sample, as the signals from [ 15 N]Shr HID1 remain broadened in a control experiment in which the [ 15 N]Shr HID1 -Hb complex was diluted only with buffer. As unlabeled Shr HID2 displaces [ 15 N]Shr HID1 from Hb, we conclude that the isolated domains can compete for the same site(s) on Hb. However, the fact that this resulted in only a partial recovery of signal could indicate that Shr HID1 binds a second site on Hb and binds to Hb more promiscuously than Shr HID2 .

Identification of the hemoglobin binding domains in Shr Two Shr HID1-2 proteins bind to tetrameric Hb
The binding stoichiometry of the Shr HID1-2 -Hb complex was determined using size-exclusion chromatography with inline multiangle light scattering (SEC-MALS). When Shr HID1-2 is mixed with the Hb tetramer at a ratio of 2:1 (Shr HID1-2 to Hb tetramer units), a complex at 101 kDa is observed (Fig. 3e). This is consistent with ϳ1.8 Shr HID1-2 proteins binding to one Hb tetramer (calculated as follows: (101 kDa (complex) Ϫ 56 kDa (Hb alone) )/25 kDa (Shr HID1-2 alone) ϭ 1.8). SEC-MALS experiments using higher Shr HID1-2 /Hb ratios (4:1 and 6:1) were also performed and yielded similar results, indicating that Hb is saturated, and a maximum of two Shr HID1-2 proteins bind to one Hb tetramer. Notably, the stoichiometry predicted by SEC-MALS is similar to that obtained from curvefitting of the ITC data (Table 1); two Shr HID1-2 proteins bind to one Hb tetramer.
Analytical ultracentrifugation (AUC) was used to determine the stoichiometry of the Shr HID2 -Hb complex. This method was used instead of SEC-MALS because the molecular weights of Hb and its complex with Shr HID2 are too similar, making it difficult to resolve these species by column chromatography. Similar to Shr HID1-2 , Shr HID2 is monomeric in solution (Fig. 3f); it has a weighted average molecular weight of 13.3 kDa (Table  S1), which corresponds closely to its theoretical molecular weight of 14.4 kDa (58). To simplify the analysis of the AUC data, binding experiments employed Hb0.1, a stabilized tetra- In each titration experiment, Shr polypeptides were injected into a cell containing Hb. In each panel, at the top is shown the time course of the titration (black) and baseline (red). The bottom part of the panel shows the integrated isotherms (squares) and the curve fit (line). As is standard, the first data point in each experiment was eliminated prior to analysis. The fifth data point in the Shr HID1 titration was also eliminated due to a spurious double peak. Origin software was used to analyze the data. e, SEC-MALS data defining receptor-Hb interactions. Elution profiles of Hb alone, Shr HID1-2 alone, and Hb in combination with Shr HID1-2 . Refractometer voltage is indicated on the left y axis, and the corresponding trace is shown in black. The elution volume is shown on the x axis. ASTRA software was used to analyze the data. The molecular weight is indicated on the right y axis. Interestingly, the measured Hb molecular weight is slightly smaller than its predicted molecular mass of 64.5 kDa based on its primary sequence, a discrepancy that has also been observed by others (65,66). f and g, sedimentation equilibrium data defining receptor-Hb interactions. The panel shows profiles of Shr HID2 alone (gray) at 26,000 rpm, as well as Hb alone (red) and in complex with Shr HID2 (blue) at a 1:45 ratio of Hb0.1 (heme basis)/Shr HID2 at 13,000 rpm. The top panels show the residuals of the fit of the experimental data, and the bottom panels show the absorbance readings on the y axis versus the radial position on the x axis. Circles and curved black lines correspond to experimental data and calculated fits to binding models described in the text, respectively.

Identification of the hemoglobin binding domains in Shr
meric from of Hb in which the ␣-globin chains are part of a single polypeptide. When Shr HID2 is added to Hb0.1 at a 180fold molar excess (Hb tetramer units), its weighted average molecular mass is 97.4 kDa ( Fig. 3g and Table S1), consistent with two Shr HID2 domains binding to one Hb0.1 tetramer. Hb0.1 is saturated with Shr HID2 , because similar molecular weights are measured when Shr HID2 is present at 60-fold molar excess. Interestingly, the binding stoichiometry obtained from fitting the ITC data suggests that four Shr HID2 proteins can bind to a tetramer, which differs from the 2:1 stoichiometry determined by AUC ( Table 1). The origin of this discrepancy is unclear, but it may result from the different conditions employed in each experiment.

HID is a structurally novel Hb-binding domain
To gain insight into the molecular basis of Hb binding, we determined the crystal structure of HID2, the second DUF1533 domain within Shr. To facilitate its crystallization, we first used NMR to delineate regions within the Shr HID2 polypeptide that are structurally ordered. Triple resonance methods were applied to [U-13 C,U-15 N]Shr HID2 , enabling ϳ94% of its backbone resonances to be assigned (Fig. 5a). A heteronuclear { 1 H} 15 N NOE experiment was then performed to identify regions within Shr HID2 that are mobile and unstructured (Fig. 5b). The majority of residues within Shr HID2 are structurally ordered in the absence of Hb, as indicated by heteronuclear NOE values Ͼ0.6 for backbone amide groups that span residues Ile-179 -Gln-283. However, ϳ20 residues located at the N terminus of Shr HID2 are disordered, as they possess low magnitude heteronuclear NOEs. We therefore produced a truncated variant of Shr HID2 for X-ray crystallography studies that removes its unstructured N-terminal residues Lys-155-Ala-174 yielding a new construct (Shr ⌬HID2 , residues Asn-175-Gln-285). Importantly, Shr ⌬HID2 binds to Hb with similar affinity and stoichiometry as Shr HID2 (Fig. 3c and Table 1), indicating that the deleted N-terminal residues are not important for function.
The structure of Shr ⌬HID2 was determined at 1.5 Å resolution using single-wavelength anomalous diffraction methods applied to native and KI-soaked crystals. Two Shr ⌬HID2 domains are present in the asymmetric unit (Fig. 6a), forming a 2-fold symmetry-related dimer that buries 517 Å 2 of surface area (59). AsSEC-MALSandAUCanalysesindicatethatShr ⌬HID2 ismonomeric, the dimer observed in the structure is likely an artifact caused by the high concentration of protein present in the crystal. The two polypeptide chains in the dimer possess nearly identical atomic structures, as their heavy atom coordinates can be superimposed with a r.m.s.d. of 0.80 Å. The polypeptides are well-defined by the electron density, with the exception of residue Asp-226 in chain A and residues Lys-284 and Gln-285 in chain B. Complete structural and data statistics are presented in Table 2.
Shr ⌬HID2 adopts a ␤-sandwich-type fold in which two antiparallel ␤-sheets are flanked by an ␣-helix. One face of the sandwich is formed by a sheet containing strands ␤1, ␤2, ␤6, and ␤5 (␤-sheet 1), whereas the other face is formed by a sheet containing strands ␤9, ␤8, ␤7, ␤3, and ␤4 (␤-sheet 2) (Fig. 6b). Strands ␤1 and ␤2 form an antiparallel ␤-hairpin. The ␤2 strand is then followed by a long loop that is connected to a two-turn ␣-helix that is positioned nearly perpendicular with respect to ␤-sheet 2. Following the helix, strand ␤3 makes antiparallel contacts with strand ␤7 in ␤-sheet 2. Strand ␤4 then forms part of a ␤-hairpin with ␤3 before being followed by an extended loop that winds around and extends over the top of ␤-sheet 1. This is followed by strand ␤5 that forms a ␤-hairpin with strand ␤6 and together with strands ␤1 and ␤2 comprise ␤-sheet 1. The structure is completed by a ␤-meander that is constructed from strands ␤7, ␤8, and ␤9 within ␤-sheet 2. Shr HID1 presumably adopts a similar structure, as it shares 27% sequence identity with Shr HID2 . The solvent-exposed side of ␤-sheet 1 is predominantly acidic, whereas on ␤-sheet 2 there is a greater balance of acidic and basic residues (Fig. 6c). It is worth mentioning that the HID domain is larger than the predicted DUF1533 domain, wherein ␤1, ␤2, and the loop connecting it to ␣1 as well as ␤9 are all located outside of the predicated DUF1533 domain in the primary sequence.
The HID is a structurally unique bacterial Hb-binding domain. Many species of Gram-positive bacteria that are related to S. pyogenes employ NEAT domains to bind Hb. Based on a Dali analysis, the Shr ⌬HID2 and NEAT domains adopt distinct structures; their Z-score does not exceed the threshold for a strong match as defined by Holm et al. (60). Moreover, Shr ⌬HID2 and NEAT domains have distinct secondary structure topologies, and Shr ⌬HID2 lacks the conserved aromatic motif that is present in Hb-binding NEAT domains (Fig. 7, a and b). Shr ⌬HID2 is also structurally distinct from Hb receptors found in other human pathogens, such as HpuA (61), ShuA (62), and HpHbR (63). The closest HID structural homolog in the Protein Data Bank is the CBM46 domain present in the Cel5B cellulase from Bacillus halodurans (PDB code 4uz8); the proteins can be superimposed with an r.m.s.d. of 2.1 Å and have a Dali Z-score of 7.9, which is deemed significant (60). However, the proteins have distinct functions, as key residues in CBM46 that mediate carbohydrate binding are poorly conserved in Shr ⌬HID2 .

Identification of the hemoglobin binding domains in Shr HIDs within Shr HID1 and Shr HID2 engage Hb using a conserved tyrosine residue
We attempted to identify surface residues in the HID that mediate Hb binding by aligning the primary sequences of the DUF1533 domains and their flanking regions from putative Hb receptors. The primary sequences of DUF1533 domains from four Shr orthologues present in Clostridium novyi, Streptococcus iniae, Streptococcus equi, and Streptococcus dysgalactiae were aligned to Shr HID1 and Shr HID2 (Fig. S2). Six residues are universally conserved, of which only one has a side chain that is completely surface-exposed, Tyr-197 (Shr HID2 labeling). A Y197A mutation was introduced into Shr HID2 (Shr HID2-Y197A ), and its ability to bind Hb was determined using the aforementioned NMR titration experiment. In marked contrast to WT Shr HID2 , adding Hb to [ 15 N]Shr HID2-Y197A does not cause its NMR signals to broaden significantly, even when Hb is in excess (2:1 Hb/Shr HID2-Y197A ) (Fig. 8a). As the Y197A mutation does not disrupt the structure of Shr HID2-Y197A , the side chain of Tyr-197 presumably forms contacts with Hb that are required for binding. The corresponding tyrosine mutation was also introduced into Shr HID1 (Y55A, Shr HID1-Y55A ) and was also shown to disrupt Hb binding (Fig. 8b). We conclude that a conserved tyrosine residue in Shr HID1 and Shr HID2 is critical for Hb binding and that the domains likely engage Hb in a similar manner. Left, entire HSQC spectrum is shown. Right, inset shows an enlarged view of the center, which has the greatest amount of spectral overlap. Of the Shr HID2 construct, ϳ94% of the residues were assigned. Unassigned residues correspond to two prolines, two residues at the N terminus, and five residues scattered throughout the domain. b, { 1 H} 15 N heteronuclear NOE values (on the y axis) shown on a per residue basis (on the x axis). Residues with missing heteronuclear NOE values were eliminated from the analysis due to either missing backbone assignments or severe spectral overlap. Shown are the average and standard deviation of three replicates.

Discussion
Human hemoglobin is a rich source of iron that bacterial pathogens capture using surface-displayed and secreted Hb receptors (64). In Gram-positive bacteria, these receptors typically interact with Hb via NEAT domains (23). Interestingly, the Hb receptor in S. pyogenes, called Shr, appears to bind Hb via a novel mechanism. Its primary sequence has been divided into two regions: NTR that binds Hb, and CTR that binds hemin (8,11). Intriguingly, based on its primary sequence, the NTR does not contain a NEAT domain suggesting that it engages Hb in a unique manner. To learn how Shr interacts with Hb, we first used NMR spectroscopy to delineate residues within its NTR that form autonomously folded domains. Three regions within the NTR are structured, two modules previously defined as domains of unknown function (DUF1533 domains) and a third module located near the C terminus of the NTR that is predicted to contain several ␣-helices (Fig. 1). NMR titration experiments conclusively demonstrate that only the DUF1533 domains within the NTR interact with Hb (Fig. 2). They therefore represent previously uncharacterized HIDs. In Shr they are HID1 (residues 26 -148) and HID2 (residues 155-285). Inter-actions originating from HID2 contribute significantly to Shr's Hb binding affinity, because based on ITC measurements it binds to Hb with a K D ϭ 16.0 Ϯ 3.4 M, whereas the isolated HID1 domain has significantly weaker affinity (K D ϭ ϳ140 M). Studies of a polypeptide containing both domains (Shr HID1-2 ), suggest avid binding to Hb. The 3-fold decrease in K D (5.1 Ϯ 0.4 M) afforded by HID1 in the Shr HID1-2 construct would seem to indicate that binding of one HID domain facilitates binding of the other resulting in the higher apparent affinity observed for the Shr HID1-2 -Hb interaction. Tsumoto and co-workers (11) have measured the binding affinity of residues Gln-22-Val-364 of Shr, which contains both HIDs and the linker domain within the NTR. This polypeptide binds to Hb with similar affinity as Shr HID1-2 (K D ϭ 6.7 M), further demonstrating that only the HIDs within the NTR mediate Hb binding.

Identification of the hemoglobin binding domains in Shr
The crystal structure of HID2 reveals that it adopts a compact ␤-sandwich type fold that is distinct from previously characterized Hb-binding proteins (Fig. 6b). Structures of three different NEAT domains in complex with Hb have been determined, including the first and second NEAT domains from the S. aureus IsdH receptor (IsdH N1 and IsdH N2 ) and the first NEAT domain from the S. aureus IsdB protein (IsdB N1 ) (44,50,65). These NEAT domains share significant primary sequence homology with one another and engage Hb in a similar manner. The globin chain in Hb is contacted by residues in the NEAT domains that are located in an ␣-helix that is positioned between strands ␤1b and ␤2, and by residues located within nearby loops (loops connecting strands ␤3-␤4 and ␤5b-␤6). Each NEAT domain contains a conserved aromatic signature sequence ((F/Y)YH(F/Y)) that forms the Hb contacting ␣-helix that when mutated disrupts binding. The HIDs within Shr must bind to Hb in a fundamentally distinct manner, as their tertiary structures differ substantially from NEAT domains. Moreover, HID primary sequences do not contain the Hb-contacting aromatic motif. Atomic structures of other Hb receptors from eukaryotic and Gram-negative microbes have also been determined and do not share significant structural homology with the HID.
To identify the surface on HIDs that contact Hb, we compared their primary sequences and performed targeted mutagenesis. Six residues are completely conserved, but only Tyr-197 (HID2 numbering) is fully solvent-exposed in the structure of HID2 and conserved. Tyr-197 is located in the extended loop that precedes the ␣-helix and is surrounded by other less well-conserved residues that may mediate Hb interactions. Mutational exchange of the tyrosine to alanine in either HID1 or HID2 abrogates Hb binding, substantiating its functional importance (Fig. 8). The HID was originally named DUF1533. At present, more than 150 proteins containing DUF1533 domains from 71 distinct species of bacteria are known. Notably, these conserved tyrosine residues lie outside the predicted DUF1533 domains in the N-terminal regions of the HIDs. It is therefore unclear whether all of these sequencerelated domains will also bind Hb, as some lack the key tyrosine residue and they are present in proteins that are unlikely to be orthologs of Shr because they contain other domains that are predicted to have enzymatic functions.
In vitro binding studies indicate that a single Hb tetramer is engaged by two Shr receptors. SEC-MALS experiments demonstrate that a Shr HID1-2 polypeptide containing both HID1 and HID2 forms a complex with the Hb tetramer that has 2:1 stoichiometry (Fig. 3e). It is likely that the intact Shr protein binds with similar stoichiometry, as residues outside the NTR do not mediate interactions with Hb (8). Within the 2:1 Shr HID1-2 -Hb complex, we speculate that HID1 and HID2 domains likely engage distinct sites on the Hb tetramer. In this scenario, each of the domains would bind a globin subunit, with interactions originating from HID2 being the most important. This model is supported by the stoichiometry obtained for the Shr HID1-2 -Hb interaction by SEC-MALS. It is also consistent with the results of the NMR experiments that showed that the isolated domains competitively bind Hb if it is assumed that the high-protein concentrations used in the NMR studies caused the domains to promiscuously bind to the globin subunits and/or the domains bind differently to Hb in isolation as compared with Shr HID1-2 . Furthermore, the mutagenesis studies in combination with the NMR competition experiments suggest that both HID's engage Hb via the same interface. Taken together, this leads us to conclude that each globin chain likely contains a single HID-binding site. It is interesting to note that HID binding to Hb is entropically unfavorable for both the isolated domains and Shr HID1-2 (Table 1). It is possible that residues within the domain(s) undergo a disorder to order transition upon binding Hb, as is the case for the Hb-binding NEAT domains within the IsdH and IsdB S. aureus Hb receptors (50,65). However, the origin of this effect remains unclear, as the heteronuclear NOE data of apo-HID2 do not reveal the presence of large disordered loops. Structures of the Shr-:Hb complex are required to understand the molecular origins of these energetic effects.
Based on the domain architecture that we have elucidated, Shr may use a tri-domain unit to extract hemin from Hb. Only the heme extraction mechanisms used by the S. aureus IsdB and IsdH proteins are well-understood. These proteins share similar primary sequences and extract hemin from Hb using a tri-domain unit, in which a helical domain separates N-and C-terminal NEAT domains that bind to Hb and hemin, respectively (42). Detailed studies of IsdH indicate that its N-terminal NEAT domain (N2 domain) engages the A-helix within a globin chain contributing ϳ95% of the total binding standard free energy (43). This positions the remaining helical and NEAT domains near the hemin pocket on the same globin subunit, thereby enabling them to distort Hb's F-helix to trigger hemin release. IsdB also contains a homologous tri-domain unit that binds to Hb in a similar manner (50). Interestingly, Shr contains a hemin-binding NEAT domain (NEAT1) immediately following the HID2 and linker domains we have identified (Fig. 1). Thus, residues 175-502 in Shr form a

Identification of the hemoglobin binding domains in Shr
D2-linker-NEAT1 tri-domain segment that is similar to those present in IsdB and IsdH. It is possible that Shr uses a similar mechanism as IsdH/B to accelerate hemin release from Hb in which HID2 binds to a globin chain to "deliver" the NEAT1 domain near Hb's hemin pocket. Interestingly, both the Shr and IsdH receptors supplement their tri-domain segments with an additional N-terminal Hb-binding domain (HID1 in Shr and the N1 NEAT domain in IsdH).
Recent studies indicate that the supplemental N1 domain in IsdH may function to slow the rate of Hb removal from the blood by obstructing interactions with the macrophage-specific endocytic receptor CD163, thereby advantageously prolonging microbial access to Hb's heme iron (66). Additional studies will be needed to determine whether the HID1 domain in Shr performs a similar function.
Collectively, the data presented in this paper provide new insight into how Shr captures Hb on the cell surface. S. pyogenes presumably encounters the dimeric form of Hb, as rupture of the red blood cells is expected to dramatically dilute Hb causing the tetramer to dissociate in ␣␤ dimers. As we have shown that two receptors bind a Hb tetramer, on the cell surface Shr likely engages the Hb ␣␤ dimer with 1:1 stoichiometry, such that its HID1 and HID2 domains each engage a globin subunit. Based on previous hemin transfer studies between Shr and Hb (8,9), it seems likely that Hb's hemin molecule is transferred to Shr's NEAT1 domain, which is proximally positioned to HID2. The domain architecture of Shr HID2-L-N1 is similar to IsdH N2N3 , a Hb-and a hemin-binding domain separated by an ␣-helical linker. We have previously shown that IsdH N2N3 accelerates hemin release from Hb by interactions with the linker domain, which distorts the Hb's F-helix (45,46). The similar domain architecture between IsdH N2N3 and Shr suggests that the linker domain in Shr may also facilitate hemin release after it engages Hb. Further insight into the mechanism of Hb capture will be gained by structural studies on the Shr-Hb complex.

Cloning, protein expression, and purification
Standard methods were used to construct expression plasmids that produced the following polypeptides: Shr HID1 (residues 26 -148); Shr HID2 (residues 155-285); Shr ⌬HID2 (residues 175-285); Shr HID1-2 (residues 26 -285); and Shr L (residues 294 -364). Proteins were expressed from pSUMO-based vectors (LifeSensors) that were transformed into Escherichia coli BL21(DE3) cells (New England Biolabs, Beverly, MA). Cultures were grown at 37°C to an A 600 of 0.6 -0.8 before induction with isopropyl ␤-D-thiogalactoside at a final concentration of 1 mM. Induction proceeded at 25°C overnight before harvesting the cells by centrifugation. The cell pellet was then resuspended in lysis buffer containing: 50 mM NaH 2 PO 4 /Na 2 HPO 4 , pH 7.0, 300 mM NaCl. A protease inhibitor mixture (Calbiochem) and phenylmethylsulfonyl fluoride (Sigma) was also added to the lysis buffer. Cells were lysed by sonication, and the lysate was clarified by centrifugation (27,200 ϫ g). The lysate was then loaded onto a Co 2ϩ -chelating column (ThermoFisher Scien-tific, Waltham, MA), and unbound proteins were removed by washing the column with lysis buffer (5 column volume (CV) equivalents). Nonspecifically bound proteins were also removed by applying 5 CVs of wash buffer (lysis buffer ϩ 10 mM imidazole) and 5 CVs of lysis buffer. The N-terminal His 6 -SUMO tag was then cleaved with the ULP1 protease and eluted by washing the column with lysis buffer. If necessary, proteins were purified using an SEC column as described by the vendor (Sephacryl, GE Healthcare). Human Hb was prepared as described previously from the blood of a healthy donor provided by the CFAR Virology Core Lab at the UCLA AIDS Institute (43). Purified Hb0.1 was a generous gift of the Olson laboratory.

Isothermal titration calorimetry (ITC)
ITC measurements were performed using a MicroCal iTC200 calorimeter (GE Healthcare) at 25°C. The Hb (in the Identification of the hemoglobin binding domains in Shr carbonmonoxy state) and Shr constructs were buffermatched in 20 mM NaH 2 PO/Na 2 HPO 4 , pH 7.5, 150 mM NaCl, 450 mM sucrose. The cell was filled with 25-35 M Hb (concentration reported on a heme/globin chain basis), and the syringe was filled with either 700 M Shr HID1 , 500 M Shr HID2 , 550 -600 M Shr ⌬HID2 , or 350 -500 M Shr HID1-2 . Twenty injections were performed using 2.0-l injection volumes at 180-s intervals. For each of the Shr constructs, an experiment was carried out to control for the heats of dilution in which the respective Shr construct was titrated into buffer, and the control data were subtracted from the experimental data. ORIGIN software was used to fit the data to a single-site binding model.

SEC-MALS
The analytical size-exclusion column WTC-030S5 (Wyatt Technology) was equilibrated in 20 mM NaH 2 PO/Na 2 HPO 4 , pH 7.5, 150 mM NaCl using an AKTA pure (GE Healthcare). For each protein or complex, a 100-l (buffer-matched) sample was loaded: 1.2 mM HbCO (heme basis), 600 M Shr HID1-2 , and the complex of HbCO at 1.2 mM and Shr HID1-2 at 600 M, 1.2 mM, and 1.8 mM. During elution, light scattering was measured with a miniDAWN TREOS (Wyatt Technology), and the refractive index (n) was measured with an Optilab T-rEX system (Wyatt Technology). Data were analyzed using ASTRA software (version 6.1) to obtain average molecular weights. The dn/dc value (where c is the concentration) for the calculation was set to 0.185 ml/g for Hb (66). Based on the primary sequence, the dn/dc values for Shr HID1-2 and Shr ⌬HID2 were calculated to be 0.187 and 0.188 ml/g, respectively. For the complex of Shr HID1-2 with Hb, the dn/dc value was calculated to be 0.186 ml/g based on the equation for the weight average sum. Theoretical molecular weights were obtained using the ProtParam server based on the primary amino acid sequence (58).

Analytical ultracentrifugation (AUC)
Sedimentation equilibrium experiments were performed at 20°C using a Beckman Optima XL-A analytical ultracentrifuge equipped with an An60-Ti rotor. Native HbA has a dimer/tetramer equilibrium, which complicates the determination of the Shr HID2 -Hb binding stoichiometry by AUC. To simplify the analysis, a mutant variety of Hb, which stabilizes the tetrameric form, was employed (Hb0.1) (71,72). Absorption optics at 280 nm were carried out for Shr HID2 alone and at 412 nm for Hb0.1 alone or in complex with Shr HID2 so that only the heme molecule was detected. All experiments were carried out in 20 mM NaH 2 PO/Na 2 HPO 4 , pH 7.5, 150 mM NaCl. 450 mM sucrose. To prevent heme release, Hb0.1 in the carbon monoxide state was used. Three mm pathlength double sector cells were used for all samples and were purged with CO before sealing the cells to prevent oxidation. Sedimentation equilibrium profiles were measured at 8,000 (ϳ5,200 ϫ g), 11,000 (ϳ9,700 ϫ g), and 26,000 rpm (ϳ54,500 ϫ g) for Shr HID2 alone, with the last being used for the analysis. For Hb0.1 alone and in complex with Shr HID2 , sedimentation equilibrium profiles were measured at 11,000 (ϳ9,700 ϫ g) and 13,000 rpm (ϳ13,600 ϫ g), with the latter being used for the analysis. A third measurement was made at 30,000 rpm (ϳ72,500 ϫ g) and was used to determine the baseline. Samples of Hb0.1 and Shr HID2 alone contained 5.5 and 248 M protein, respectively. For data collected at a 15:1 ratio of Shr D2 /Hb0.1, the sample contained 6.2 and 89 M Hb0.1 and Shr HID2 , respectively. For the 45:1 ratio of Shr HID2 /Hb0.1, the sample contained 5.5 and 248 M Hb0.1 and Shr HID2 , respectively. Weight average molecular masses were determined by fitting with a nonlinear least-squares exponential fit for a single ideal species using Beckman Origin-based software (version 3.01). Partial specific volumes were calculated from the amino acid compositions and corrected to 20°C. They were 0.740 for Shr HID2 , 0.749 for Hb0.1, and 0.746 for the complex.

X-ray crystallography
Crystals of Shr ⌬HID2 were produced from a stock of 52 mg/ml (4.2 mM) dissolved in 20 mM NaH 2 PO 4 /Na 2 HPO 4 , pH 6.5. Crystals were grown using the hanging drop, vapor diffusion method against a mother liquor of 0.1 M sodium acetate, pH 4.5, 0.1-0.2 M lithium sulfate, 30 -50% (w/v) PEG 400. Crystals developed after ϳ24 h at room temperature. Data sets were collected on two crystals: Shr ⌬HID2 (native) and Shr ⌬HID2 soaked for 30 s in mother liquor containing 35.9% (w/v) PEG 400 and 460 mM KI (Shr ⌬HID2 ϩ KI) (73). The KI dataset was collected on a Rigaku FREϩ generator with a copper anode equipped with a Rigaku HTC image plate detector ( ϭ 1.5418 Å). The Matthews coefficient indicated that there were two Shr ⌬HID2 molecules in the asymmetric unit using the online server MATTPROB (74). XDS/XSCALE was used to index, integrate, and scale the data (75). Conservative resolution limits were applied to balance the calculated I/, R sym , and CC1/2 in the highest resolution shell. The structure of Shr ⌬HID2 was solved by single-wavelength anomalous dispersion on Shr ⌬HID2 ϩ KI using HKL2MAP (76), the graphical user interface for the SHELXC/D/E programs. Four iodide sites were found by SHELXD (77). SHELXE (78) was used to assign the hand, produce the first set of phases, and perform solvent flattening. The final figure of merit was 0.767. Further density modification and automatic model building was performed by SHARP (79). SHARP built residues 2-111 of chain A and 8 -109 of chain B in the initial model out of 112 residues in the construct. The Shr ⌬HID2 ϩ KI model was refined to 1.8 Å. This model was further refined against the high-resolution 1.5 Å native data set obtained at the Advanced Photon Source on beamline 24-ID-C on a DECTRIS-PILATUS 6M detector ( ϭ 0.9793 Å). Initial rounds of structure refinement were carried out using Coot (80) and PHENIX (81). Later rounds of refinement were carried out using Coot (80) and BUSTER (82) with TLS refinement (83). Structure statistics are presented in Table 2. The coordinates of the final model and structure factors have been deposited in the Protein Data Bank under PDB code 6DKQ.