Molecular Basis for the Evolution of Species-Specific Hemoglobin Capture by Staphylococcus aureus

During infection, bacteria must steal metals, including iron, from the host tissue. Therefore, pathogenic bacteria have evolved metal acquisition systems to overcome the elaborate processes mammals use to withhold metal from pathogens. Staphylococcus aureus uses IsdB, a hemoglobin receptor, to thieve iron-containing heme from hemoglobin within human blood. We find evidence that primate hemoglobin has undergone rapid evolution at protein surfaces contacted by IsdB. Additionally, variation in the hemoglobin sequences among primates, or variation in IsdB of related staphylococci, reduces bacterial hemoglobin capture. Together, these data suggest that S. aureus has evolved to recognize human hemoglobin in the face of rapid evolution at the IsdB binding interface, consistent with repeated evolutionary conflicts in the battle for iron during host-pathogen interactions.

ing have been the subject of intense study (4). Many vertebrate-associated bacteria harbor high-affinity uptake systems targeting heme, an abundant iron-containing porphyrin cofactor (5).
The most abundant source of heme iron in the mammalian host is hemoglobin, which mediates oxygen transport within circulating erythrocytes. The predominant adult hemoglobin consists of a tetramer containing two ␣-globin and two ␤-globin protein subunits, each of which binds a single heme molecule for the coordination of oxygen. The Gram-positive bacterium Staphylococcus aureus is well adapted to the human host and is a leading cause of skin and soft tissue infections, endocarditis, osteomyelitis, and bacteremia (6). To acquire iron during infection, S. aureus has evolved a high-affinity hemoglobin binding and heme extraction system, termed the iron regulated surface determinant (Isd) system (7). Following the lysis of proximal erythrocytes via secreted bacterial toxins, the released hemoglobin is captured by receptors at the S. aureus cell surface (8,9). The Isd system of S. aureus in part consists of cell wall-anchored IsdB and IsdH, which bind hemoglobin and haptoglobinhemoglobin, respectively (9,10).
We and others have shown that IsdB is the primary hemoglobin receptor for S. aureus and critical for pathogenesis in murine infection models (9,(11)(12)(13). Additionally, IsdB is highly expressed in human blood (14) and a promising vaccine target (15), underscoring its importance in human disease. IsdB extracts heme from hemoglobin, and heme is subsequently passed across the cell wall and into the cytoplasm for degradation by the heme oxygenases IsdG and IsdI, liberating iron (16)(17)(18)(19)(20). Underscoring the importance of IsdB for pathogenesis, heme is the preferred iron source of S. aureus during murine infection (21). The cell wall-anchored IsdABCH proteins share between one and three NEAT (near transporter) domains for the coordination of hemoglobin or heme. IsdB NEAT1 binds hemoglobin while NEAT2 binds heme, tethered by an intervening linker (22). Consistent with the adaptation of S. aureus to colonize and infect humans, we previously found that S. aureus IsdB binds human hemoglobin more effectively than mouse hemoglobin, the common laboratory animal used to model S. aureus infection (13). These results suggest that hemoglobin variation among mammals dictates effective heme acquisition by S. aureus and other Gram-positive bacteria.
Previous work has demonstrated that pathogens can promote a rapid adaptation of host immunity genes through repeated bouts of positive selection (23)(24)(25). While adaptation during such evolutionary conflicts can take many forms, theoretical and empirical studies indicate that an elevated rate of nonsynonymous to synonymous substitutions in protein-coding genes is often indicative of recurrent positive selection (26,27). To date, most empirical studies of host-pathogen "arms races" have focused on viruses (28)(29)(30)(31). Recently, we showed that the transferrin family of iron-binding proteins has undergone extremely rapid evolution in primates at protein surfaces bound by iron acquisition receptors from Gram-negative bacteria (32,33). These findings are consistent with the existence of a long-standing evolutionary conflict for nutrient iron, whereby mutations in iron-binding proteins that prevent bacterial scavenging protect the host from infection and are favored by natural selection. While these studies have expanded our understanding of how pathogens shape the evolution of host genomes, they also raise the question of whether other components of nutritional immunity might be subject to similar evolutionary dynamics.
In addition to its role as the principal bloodstream oxygen transporter, hemoglobin has provided an important biological model for diverse areas of the life sciences. Elegant studies have illustrated how hemoglobin variation underlies multiple instances of adaptation to high altitudes in diverse vertebrate taxa (34)(35)(36)(37). Hemoglobin alleles have also likely been subject to the balance of selection in human populations, where mutations that produce sickle-cell disease also confer resistance to severe malaria (38) and have reached high frequencies in regions where malaria is endemic. Despite its long history of study, the consequences of hemoglobin evolution for vertebrate nutritional immunity remain unclear. In the present study, we set out to investigate the evolution of hemoglobin family proteins in primates and determine whether primate hemoglobin evolution impacts the ability to sequester heme iron from bacterial pathogens. generated using PhyML. All tests detected significant evidence of positive selection acting on both ␣and ␤-globins using both species and gene phylogenies (see Data Set S1 in the supplemental material). Multiple analyses repeatedly identified two sites in ␣and ␤-globins exhibiting strong signatures of positive selection (Fig. 1A). It became apparent that these rapidly evolving sites localized to similar regions of the ␣and ␤-globin proteins, specifically, the N-terminal A helix and the hinge region between the E and F helices (Fig. 1B). In fact, the two sites exhibiting signatures of selection in the ␣and ␤-globin A helices are at homologous positions. These parallel signatures of selection between ␣and ␤-globins might indicate that a similar selective pressure has driven this divergence between primate species. To investigate whether bacterial heme-scavenging receptors are one such selective pressure, rapidly evolving sites were mapped onto a recently solved cocrystal structure between human hemoglobin and the IsdB protein from S. aureus (22). Notably, all four rapidly diverging hemoglobin residues are localized to the IsdB binding interface, in close proximity to the NEAT1 domain (Fig. 1C). Altogether, these findings indicate that primate globins have undergone rapid divergence at specific sites proximal to the binding interface of the S. aureus hemoglobin receptor IsdB.
Primate hemoglobin variation dictates S. aureus binding and heme iron acquisition. To assess how hemoglobin divergence among primates impacts recognition by S. aureus, recombinant hemoglobin from human, white-cheeked gibbon, baboon, talapoin, and marmoset sources were purified, providing a broad representation from our phylogenetic data set. An established biochemical assay was used to measure the binding of hemoglobin by S. aureus, in which S. aureus cells recognize recombinant human hemoglobin as well as hemoglobin purified from blood in an IsdB-dependent manner (see Fig. S1). S. aureus exhibited significantly reduced binding of baboon and marmoset hemoglobin to the cell surface ( Fig. 2A and S2). It was noted that the binding patterns did not strictly match the predictions based on host phylogeny, suggesting discrete large-effect substitutions in hemoglobin may contribute disproportionately to the recognition by S. aureus. We next determined the ability of primate hemoglobins to support growth of S. aureus as the sole iron source. Consistent with the whole-cell binding data, hemoglobins that were bound by S. aureus with low affinity were unable An iron-starved S. aureus wild type was incubated with purified recombinant hemoglobin from representative species across hominoid (orange), Old World monkey (purple), and New World monkeys (brown). Hemoglobin bound to the surface of S. aureus was eluted and analyzed by SDS-PAGE; relative hemoglobin abundance was measured by densitometry analysis (Image J) and compared to human hemoglobin for each replicate. (B) Growth of S. aureus in iron-depleted medium with 2.5 g/ml of purified recombinant hemoglobin as the sole iron source. Shown is the final growth yield of S. aureus after 48 h. Growth of each replicate is compared to growth using human hemoglobin. Panel A shows the means from two independent experiments in biological triplicates, panel B shows the means from three independent experiments with 2 to 3 biological replicates, Ϯ SEM; **, P Ͻ 0.005; ***, P Ͻ 0.0005 by two-way analyses of variance (ANOVA) with Sidak's correction for multiple comparisons comparing transformed (percent value) data.
to support optimal bacterial growth, indicating that the capability to bind hemoglobin is a measure of the ability to utilize hemoglobin as an iron source ( Fig. 2B and S3). Altogether, these results demonstrate that variation among primate globins dictates bacterial hemoglobin capture and heme-dependent growth.
Species-specific diversity in ␣-globin restricts heme scavenging by S. aureus. The identification of rapidly evolving sites at the IsdB binding interface in both ␣-globin and ␤-globin suggests that both globin subunits contribute to S. aureus species-specific hemoglobin capture. We therefore exploited the enhanced binding of human hemoglobin relative to that from baboon to examine the role of each globin subunit in this biochemical interaction. The ability of S. aureus to bind chimeric hemoglobins was measured, which revealed that both globins contribute to species specificity ( Fig. 3A and S4A), as chimeras containing either human ␣or ␤-globin were bound more effectively than baboon hemoglobin. However, ␣-globin appears to have a greater effect on human-specific capture, as the ␣-human ␤-baboon chimera bound significantly better than ␣-baboon ␤-human. Focusing on phylogenetic variation at the protein binding interface, ␣-globin T8 and N78 were found to both be proximal to the NEAT1 domain of IsdB ( Fig. 3B and C). Mutagenesis of the N-terminal alpha helix of human ␣-globin revealed that substituting the Thr residue of human ␣-globin with the Lys residue of baboon ␣-globin at position 8 reduced binding by S. aureus ( Fig. 3D and S4B and C). Additionally, replacing A5D, T8K, and N9H in human ␣-globin, which converts this seven-amino-acid region (Fig. 3B), with those of baboon ␣-globin leaves S. aureus binding nearly indistinguishable from that of baboon hemoglobin. These results demonstrate that the N-terminal helix of ␣-globin makes a major contribution to human-specific hemoglobin recognition by S. aureus. Next, the relative importance of the rapidly evolving N78 residue in ␣-globin was assessed, which lies N terminal to the sixth alpha helix (Fig. 3B). The substitution of N78 for a glutamine (present in baboon, talapoin, and other Old World primates) or a histidine reduced the binding of human hemoglobin ( Fig. 3E and S4D). Thus, substitutions at multiple residues in ␣-globin that exhibit signatures of repeated positive selection are sufficient to disrupt the ability of S. aureus to recognize human hemoglobin.
␤-Globin divergence contributes to S. aureus hemoglobin binding. S. aureus was capable of binding the baboon ␣-globin human ␤-globin chimeric hemoglobin with higher affinity than baboon hemoglobin (Fig. 3A), signifying that ␤-globin also contributes to S. aureus species-specific hemoglobin capture. Therefore, the contribution of rapidly evolving residues in ␤-globin to this binding interaction was investigated (Fig. 4A). Both S9 and A76 interact closely with the NEAT1 domain of IsdB (Fig. 4B). The effect of replacing the human ␤-globin S9 and A76 with residues found in other primate species analyzed in this work was systematically tested, which revealed that A76 is particularly important for binding by S. aureus ( Fig. 4C and S5). Notably, baboon and human ␤-globins differ at both positions 9 and 76, suggesting that these residues contribute to the inability of IsdB to bind baboon hemoglobin. These differences might also explain the different binding affinities between human hemoglobin and the human ␣-globin baboon ␤-globin chimera, observed in Fig. 3A. As for ␣-globin, no single residue substitution improved binding by S. aureus IsdB, consistent with the hypothesis that IsdB has specifically adapted to bind human hemoglobin. Taken together with earlier data, residues at the IsdB interface of both ␣-globin and ␤-globin contribute to the recognition of hemoglobin by S. aureus. This is consistent with the NEAT1 domain of multiple IsdB monomers engaging in hemoglobin capture by binding both ␣and ␤-globins, as observed in the reported cocrystal structure (22).
IsdB diversity among related staphylococcal strains impacts primate-specific hemoglobin capture. Given the observed differences in S. aureus binding between diverse primate hemoglobins, we considered how genetic variation in IsdB might impact this interaction. The IsdB NEAT1 subdomain Q162R-S170T is critical for hemoglobin recognition and is completely conserved among more than three thousand S. aureus clinical isolates (11). Therefore, IsdB variation among congeneric Staphylococcus argenteus and Staphylococcus schweitzeri was assessed. These recently diverged taxa (D) An iron-starved S. aureus wild type was incubated with purified recombinant hemoglobin, including mutagenized human hemoglobin, and bound hemoglobin was quantified. (E) An iron-starved S. aureus wild type was incubated with purified recombinant hemoglobin, including mutagenized human hemoglobin, and bound hemoglobin was quantified. Panel A shows the means from 3 independent experiments with 2 to 3 biological replicates, panel D shows the means from 6 independent experiments with 2 to 3 biological replicates, and panel E shows the means from 2 independent experiments with 3 biological replicates Ϯ SEM; ns, no significance; *, P Ͻ 0.05; **, P Ͻ 0.005; ***, P Ͻ 0.0005 by two-way ANOVA with Sidak's correction for multiple comparisons comparing transformed (percent value) data. (Fig. 5A) are both primate associated and, unlike most other staphylococci, harbor IsdB. We measured the ability of IsdB from S. argenteus and S. schweitzeri to bind hemoglobin by expressing them ectopically in S. aureus lacking the native isdB gene. Consistent with their overall high sequence identity, S. schweitzeri and S. argenteus IsdB bound primate hemoglobin with a similar pattern of species preference as S. aureus ( Fig. 5B and S6). However, both the IsdB of S. schweitzeri and S. argenteus displayed reduced binding of talapoin hemoglobin, and S. argenteus IsdB did not bind marmoset hemoglobin significantly less than human hemoglobin. These data indicate that the variation among IsdB sequences impacts species-specific hemoglobin capture. Evolution of Species-Specific Hemoglobin Capture ® IsdB NEAT1 domain diversity among staphylococci modulates human hemoglobin recognition. A closer examination of the Q162R-S170T region of IsdB NEAT1 revealed variation between related staphylococci but no variation in the critical hemebinding region of NEAT2 (Fig. 6A). This region of NEAT1 closely interacts with the N-terminal helices of either ␣-globin or ␤-globin, in close proximity to both discrete sites bearing signatures of adaptive evolution in ␣-globin and ␤-globin (Fig. 6B). To determine the functional consequences of variation in this NEAT1 domain, Q162 and S170T were mutagenized in S. aureus IsdB to mimic the sequence of S. schweitzeri and S. argenteus. These residues are not expected to disrupt IsdB tertiary structure, as they already exist in related IsdB proteins. Variations at both of these positions reduced the affinity for human hemoglobin, showing that in the context of S. aureus IsdB, Q162 and caseolyticus was included as an outgroup. The similarity of IsdB in S. argenteus and S. schweitzeri relative to S. aureus is shown on the right. Bootstrap values above 80 are indicated. (B) S. aureus lacking native isdB but harboring constitutively expressed plasmid-borne isdB variants were incubated with purified recombinant hemoglobin and from hominoid (orange), Old World monkey (purple), and New World monkeys (brown) and bound hemoglobin was quantified. The means from three independent experiments with 3 biological replicates Ϯ SEM are shown; *, P Ͻ 0.05; ****, P Ͻ 0.0001 by two-way ANOVA with Sidak's correction for multiple comparisons, comparing transformed (percent value) data. S170 are required for high-affinity hemoglobin binding ( Fig. 6C and S7). Additionally, mutagenized IsdB failed to fully support the growth of S. aureus using human hemoglobin as the sole iron source (Fig. 6D), which supports the conclusion that this NEAT1 subdomain of S. aureus has evolved for optimal binding and utilization of human hemoglobin.

DISCUSSION
In this work, we report that recurrent positive selection acting on primate ␣and ␤-globin proteins restricts hemoglobin binding and nutrient acquisition by pathogenic S. aureus. Estimations of divergence in the Staphylococcus genus have been lacking; however, the Kloos hypothesis (41) contends that staphylococci have coevolved with their mammalian hosts over long evolutionary timescales. In support of this concept, primate specificity among staphylococci has been reported, including S. aureus, S. epidermidis, and S. warneri, as well as avian (S. gallinarum), equine (S. equorum), and other taxa. Indeed, it has been proposed that the canine-associated S. pseudintermedius diverged from S. aureus simultaneously with the divergence of Primate and Carnivora orders (42). Most staphylococci are commensal organisms, while S. aureus is uniquely adapted to infect deep tissue and cause disease. As such, the IsdB system is only harbored by S. aureus and closely related primate-associated staphylococci. By narrowing our analysis of hemoglobin evolution to primates, we were thus able to assess specific biological features of primate-associated staphylococci.
An outstanding question in the study of S. aureus evolution has been determining the selective pressures responsible for human-specific virulence factors. S. aureus subdomain critical for hemoglobin binding shows variation among staphylococcal IsdB, while no variation was observed for the NEAT2 subdomain required for heme binding. (B) The Q162 to S170 subdomain of NEAT1 (cyan) is proximal to helices containing T8 and N78 of ␣-globin (red). (C) S. aureus lacking native isdB but harboring constitutively expressed plasmid-borne S. aureus isdB variants was incubated with purified recombinant human hemoglobin, and bound hemoglobin was quantified. (D) The growth of S. aureus lacking native isdB but harboring constitutively expressed plasmid-borne S. aureus isdB variants using hemoglobin as the sole iron source was monitored over time. Panel C shows, the means from three independent experiments with 3 biological replicates Ϯ SEM; **, P Ͻ 0.005; ***, P Ͻ 0.0005 by two-way ANOVA with Sidak's correction for multiple comparisons, comparing transformed (percent value) data. Panel D shows the results of two independent experiments with six biological replicates each Ϯ standard deviations.
Evolution of Species-Specific Hemoglobin Capture ® asymptomatically colonizes the anterior nares of approximately one-third of the human population yet is capable of causing a wide range of invasive diseases. While some bacterial colonization factors have been implicated in the pathogenesis, many virulence factors have evolved highly specific targets that are not obviously involved in nasal colonization (43). As such, we cannot definitively conclude that IsdB evolution has been driven by selection during invasive disease. It is also likely that variation across IsdB sequences of S. aureus, S. argenteus, and S. schweitzeri may be the results of antigenic variation to evade the immune system. By focusing on the hemoglobin binding pocket of IsdB, we have been able to pinpoint critical variations for hemoglobin specificity.
Our phylogenetic analyses revealed strikingly parallel signatures of positive selection between the ␣and ␤-globin genes across primates. In particular, rapidly evolving sites in the ␣and ␤-globin A-helices are predicted to be homologous on the basis of the predicted protein alignments. Our results suggest that these correlations reflect selection in response to NEAT domain-containing bacterial receptors like IsdB with conserved globin-binding sites. A well-established body of literature has shown that other selective pressures play important roles in the patterns of hemoglobin polymorphism and divergence across vertebrates, including the adaptation to high altitude and malaria resistance (34,38). It is therefore possible that the signatures of selection detected in our study were driven by pressures other than nutritional immunity. Nonetheless, our empirical results demonstrate that variation in hemoglobins at discrete sites has important functional consequences for bacterial iron acquisition.
Previous studies have illustrated how mutations in hemoglobin coding or regulatory regions can have highly deleterious effects on heme binding, oxygen affinity, and protein stability (44,45). In addition to aforementioned sickle-cell alleles, dozens of hemoglobin mutations in humans have been reported that contribute to genetic disease, including anemia and thalassemia (46). Thus, despite identifying particular sites that are highly divergent among primates, much of the globin gene content is constrained due to purifying selection. In future work, it would be useful to determine how variation among primate globins impacts other biochemical functions, including heme binding and oxygen affinity. Such insights might improve our fundamental understanding of hemoglobin biology and the mechanisms underlying human hemoglobinopathies.
In conclusion, this work illustrates how rapid site-specific hemoglobin variation restricts heme acquisition by the prominent human pathogen S. aureus. These findings provide a fundamental new perspective on vertebrate globin evolution, highlighting nutritional immunity as a selective pressure that might strongly impact divergence and natural selection. Future studies will assist in illuminating how these combinations of adaptive mutations contribute to hemoglobin function and host physiology. An understanding of the genetic and molecular determinants of bacterial pathogenicity is critical for developing new antimicrobial treatment strategies, particularly as major pathogens such as S. aureus continue to develop resistance to existing antibiotics. Combining comparative genetics with molecular experimentation in turn provides not only a historical perspective of host-microbe evolutionary conflict but also mechanistic insights on modern human infectious disease.

MATERIALS AND METHODS
Bacterial strains. The bacterial strains and plasmids used in this study are listed in Table 1. For Escherichia coli strains, LB agar and broth (Fisher, Hampton, NH) were routinely used, and growth was at 37°C. For the selection of pHUG21, 12.5 g/ml of carbenicillin (Fisher) was used, for the selection of pHb0.0, 5 g/ml of tetracycline hydrochloride (Alfa Aesar, Haverhill, MA) was used, and for the selection of pOS1 P lgt , 50 g/ml of carbenicillin was used. Staphylococcus strains were grown at 37°C using tryptic soy agar and broth (Fisher), except where noted throughout. For the selection of pOS1 P lgt , 10 g/ml chloramphenicol (Fisher) was used except where noted. The strains were streaked onto agar from stocks stored at Ϫ80°C 2 days prior to each experiment.
The staphylococcal DNA gyrase gene tree was generated using PhyML with 1,000 bootstraps as described above. Macrococcus caseolyticus DNA gyrase was included as an outgroup. The similarity of IsdB sequences in S. argenteus and S. schweitzeri relative to that in S. aureus is shown in Fig. 5.
Recombinant purification of hemoglobin. Hemoglobin expression strains [BL21(DE3) pHUG21 pHb0.0] were streaked onto LB agar containing 12.5 g/ml carbenicillin and 5 g/ml tetracycline hydrochloride. pHb0.0 harbors both ␣and ␤-globin genes, and proper folding and tetramerization require sufficient intracellular heme. Therefore, pHUG21, which encodes a heme uptake system, is coexpressed, and hemin is supplemented in the medium (52). Single colonies were inoculated in 5 ml of LB broth supplemented with 12.5 g/ml carbenicillin and 5 g/ml tetracycline hydrochloride and grown for 14 h at 37°C with shaking. This culture was used for inoculating 1:500 into 1.5 liters of LB with 12.5 g/ml carbenicillin, 5 g/ml tetracycline hydrochloride, 100 M hemin (prepared fresh at 10 mM in 0.1 M NaOH; Sigma, St. Louis, MO), and 50 g/ml of the iron chelator ethylenediamine-di(ohydroxyphenylacetic acid (EDDHA, solid added directly to medium; LGC Standards, Teddington, UK; ) in a 2.8-liter Fernbach flask. Cultures were grown at 37°C until the optical density at 600 nm (OD 600 ) reached 0.6 to 0.8. The expression of hemoglobin was induced with 40 g/ml IPTG (isopropyl-␤-isopropyl-␤-Dthiogalactopyranoside-thiogalactopyranoside; RPI, Mount Prospect, IL). At 16 h postinduction at 37°C, cells were collected by centrifugation. The cell pellet was resuspended in 20 ml phosphate-buffered saline (PBS) containing 10 mM imidazole (Fisher), 5 mM MgCl 2 (Sigma), 1 Roche protease inhibitor tablet (Fisher), approximately 1 mg/ml of lysozyme (Thermo), and 100 g/ml DNase from bovine pancreas (Sigma). The cell pellet resuspended with rocking for 20 min at room temperature after incubating on ice for 20 min. The cells were lysed using an Emulsiflex (Avestin, Ottawa, CA), and then the cell lysate was clarified by ultracentrifugation (60 min at 17,000 ϫ g). The cell lysate was applied to a 3 ml of nickelnitrilotriacetic acid (Ni-NTA) resin (Qiagen) in a gravity column, to which hemoglobin binds, and washed with 50 ml PBS containing 10 mM imidazole. Hemoglobin was eluted with 6 ml PBS containing 500 mM imidazole, with the first 1 ml of eluate discarded. The hemoglobin-containing eluate was dialyzed twice sequentially in PBS at 4°C. Purified hemoglobin was filter sterilized with a 0.45-m filter and stored in aliquots in liquid nitrogen. The hemoglobin concentration was measured with Drabkin's reagent (Sigma) using human hemoglobin as a standard, ranging from 2 to 6 mg/ml. The relative purity was assessed using SDS-PAGE before use in experiments, as shown in Fig. S8.
Whole-cell hemoglobin binding assay. S. aureus strains were streaked on tryptic soy agar (containing 10 g/ml chloramphenicol for strains carrying plasmids) and grown at 37°C for 24 h. Single colonies were used to inoculate 3 ml of RPMI containing 1% Casamino Acids and 0.5 mM 2,2=-dipyridyl (Acros/Fisher) to induce the expression of chromosomal isdB or 10 g/ml chloramphenicol (for strains carrying plasmids with constitutive isdB expression). After 14 to 16 h of growth at 37°C with shaking, 2 OD 600 units (except for experiments shown in part in Fig. 3D and S4C, where 1 OD 600 unit was used) were collected by centrifugation in a 1.5-ml Eppendorf tube. The cell pellet was resuspended with 1 ml PBS or PBS containing recombinant hemoglobin. Then, 10 g/ml (chromosomal isdB) or 2.5 g/ml (plasmid-borne isdB) of hemoglobin was used. The cells were incubated with hemoglobin or PBS for 30 min at 37°C with shaking, and then the cells were collected by centrifugation at 4°C at 8,000 ϫ g. Cells were washed thrice by pipetting with 1 ml ice-cold PBS and centrifuging at 4°C at 8,000 ϫ g. After the final wash, the cells were resuspended in 30 l 0.5 M Tris (pH 8.0; Fisher) containing 4% SDS (Fisher) and heated at 90°C for 5 min to remove surfacebound proteins. The cells were collected by centrifugation at 8,000 ϫ g, and the eluate was added to 6ϫ loading buffer and heated at 90°C for 5 min. The samples were subjected to 12% or 17.5% SDS-PAGE and silver stained (GE, Boston, MA). The gels were imaged using an Alpha Innotech Alpha Imager or Bio-Rad ChemiDoc MP imaging system. Quantification was performed by densitometric analysis with Image J (NIH) according to the software instructions and by quantifying the area under the peak that corresponds to the hemoglobin band, excluding background density. Because of the variation in stain intensity and quantity of nonspecific bands across gels, all comparisons were made within the same gel, and the relative density was calculated for each biological replicate within the same gel; the comparison was either to human hemoglobin or wild-type IsdB, depending on the assay. Additionally, PBS-only samples and S. aureus ΔisdB::erm were used to verify that hemoglobin binding in this assay is IsdB dependent (Fig. S4C, S5, and S8) as previously observed, and that recombinant human hemoglobin is bound equally as well as hemoglobin purified from human blood (Fig. S1) (11,13).
Growth with hemoglobin as sole iron source. For hemoglobin variants ( Fig. 2D and S3), S. aureus Newman WT was streaked onto tryptic soy agar and allowed to grow for 24 h at 37°C. A few colonies were used to inoculate 5 ml of RPMI (Corning, Corning, NY) supplemented with 1% Casamino Acids (Fisher) and 0.5 mM EDDHA (prepared fresh in ethanol). After growth to stationary phase at 37°C with shaking, approximately 16 h, 4 l of culture was inoculated in 196 l of medium in a 96-well plate, and the OD 600 at 37°C with shaking was monitored over time using a BioTek plate reader. The medium was RPMI containing 1% Casamino Acids that had been depleted of cations with Chelex 100 (Sigma) according to the manufacturer's instructions, filter sterilized, and supplemented with 25 M ZnCl 2 , 25 M MnCl 2 , 100 M CaCl 2 , and 1 mM MgCl 2 (all from Fisher) to restore noniron cations, 1.5 mM EDDHA to chelate any remaining free iron, and 2.5 g/ml of recombinant purified hemoglobin as the sole iron source.
For IsdB variants (Fig. 6D), S. aureus strains were streaked onto tryptic soy agar containing 10 g/ml chloramphenicol and allowed to grow for 24 h at 37°C. A single colony was resuspended in 120 l RPMI containing 1 M EDDHA (prepared fresh in 0.1 M NaOH) and 5 g/ml chloramphenicol. Then, 100 l was added to 2 ml of RPMI containing 1 M EDDHA and 5 g/ml chloramphenicol and grown at 37°with shaking in aeration tubes for 8 h. The OD 600 was measured for each culture and normalized to 1, and 5 l was used to inoculate 2 ml of RPMI containing 1 M EDDHA, 5 g/ml chloramphenicol, and 50 nM hemoglobin or no hemoglobin (for ΔisdB pOS1 P lgt isdB). Growth was monitored every 12 h by removing 50 l of culture and adding to 150 l PBS to measure the OD 600 with path length correction in a BioTek plate reader, with the background subtracted from the corrected OD 600 values. Growth using hemoglobin as a sole iron source in both assays is IsdB dependent (11,13,22).
Quantification and statistical analysis. Specific statistical details for each experiment can be found in the corresponding figure legends. Data analysis and statistical tests were performed in Prism 6 (GraphPad).

ACKNOWLEDGMENTS
We thank members of the Skaar and Barber laboratories for critical evaluation of the manuscript. We acknowledge the gift of pHUG21 from Douglas Henderson (University of Texas of the Permian Basin) and pHb0.0 from John Olson (Rice University).