Oligopeptide-binding protein from nontypeable Haemophilus influenzae has ligand-specific sites to accommodate peptides and heme in the binding pocket

In nontypeable Haemophilus influenzae (NTHi), the oligopeptide-binding protein (OppA) serves as the substrate-binding protein (SBP) of the oligopeptide transport system responsible for the import of peptides. We solved the crystal structure of nthiOppA in complex with hydrophobic peptides of various sizes. Our novel hexapeptide complex demonstrates the flexibility of the nthiOppA-binding cavity to expand and accommodate the longer peptide while maintaining similar protein–peptide interactions of smaller peptide complexes. In addition to acquiring peptides from the host environment, as a heme auxotroph NTHi utilizes host hemoproteins as a source of essential iron. OppA is a member of the Cluster C SBP family, and unlike other SBP families, some members recognize two distinctly different substrates. DppA (dipeptide), MppA (murein tripeptide), and SapA (antimicrobial peptides) are Cluster C proteins known to also transport heme. We observed nthiOppA shares this heme-binding characteristic and established heme specificity and affinity by surface plasmon resonance (SPR) of the four Cluster C proteins in NTHi. Ligand-docking studies predicted a distinct heme-specific cleft in the binding pocket, and using SPR competition assays, we observed that heme does not directly compete with peptide in the substrate-binding pocket. Additionally, we identified that the individual nthiOppA domains differentially contribute to substrate binding, with one domain playing a dominant role in heme binding and the other in peptide binding. Our results demonstrate the multisubstrate specificity of nthiOppA and the role of NTHi Cluster C proteins in the heme-uptake pathway for this pathogen.

In nontypeable Haemophilus influenzae (NTHi), the oligopeptide-binding protein (OppA) serves as the substrate-binding protein (SBP) of the oligopeptide transport system responsible for the import of peptides. We solved the crystal structure of nthiOppA in complex with hydrophobic peptides of various sizes. Our novel hexapeptide complex demonstrates the flexibility of the nthiOppA-binding cavity to expand and accommodate the longer peptide while maintaining similar protein-peptide interactions of smaller peptide complexes. In addition to acquiring peptides from the host environment, as a heme auxotroph NTHi utilizes host hemoproteins as a source of essential iron. OppA is a member of the Cluster C SBP family, and unlike other SBP families, some members recognize two distinctly different substrates. DppA (dipeptide), MppA (murein tripeptide), and SapA (antimicrobial peptides) are Cluster C proteins known to also transport heme. We observed nthiOppA shares this heme-binding characteristic and established heme specificity and affinity by surface plasmon resonance (SPR) of the four Cluster C proteins in NTHi. Ligand-docking studies predicted a distinct heme-specific cleft in the binding pocket, and using SPR competition assays, we observed that heme does not directly compete with peptide in the substrate-binding pocket. Additionally, we identified that the individual nthiOppA domains differentially contribute to substrate binding, with one domain playing a dominant role in heme binding and the other in peptide binding. Our results demonstrate the multisubstrate specificity of nthiOppA and the role of NTHi Cluster C proteins in the heme-uptake pathway for this pathogen.
Nontypeable Haemophilus influenzae (NTHi) 2 is a widespread Gram-negative commensal that colonizes the human nasopharynx. As an opportunistic pathogen, NTHi is one of the principal bacteria isolated from the middle ear during otitis media infections. NTHi infections can also exacerbate other upper and lower respiratory tract diseases and induce complications for patients with chronic obstructive pulmonary disease and cystic fibrosis (1)(2)(3). The host immune system limits access to essential nutrients and produces antimicrobial peptides to prevent infections. To circumvent host-mediated defenses, NTHi manages a complex network of transport proteins to scavenge nutrients from the host environment.
The oligopeptide (Opp) transport system is an ATP-binding cassette (ABC) transporter responsible for the uptake of peptides, supplying pathogens with essential nutrients as a source of carbon, nitrogen, and amino acids. These bacterial importers are composed of five subunits. In Gram-negative bacteria, the substrate-binding protein (SBP) binds the substrate in the periplasm and shuttles it to the transporter. Transmembrane domains (TMDs) form a dimer to create the translocation channel through the membrane. Energized by ATP binding and hydrolysis, a dimer of nucleotide-binding domains (NBDs) is coupled to the TMDs to drive conformational changes to import the substrate into the cell. Additionally, nutrient acquisition through the Opp system influences many cellular processes, including internalization of quorum-sensing peptides, biofilm production, modifying the cell surface, and antibiotic resistance (4 -7).
OppA, the SBP of the Opp transport system, belongs to a group of structurally related SBPs involved in the uptake of a wide range of nutrients from nickel to peptides known as the Cluster C family. In Gram-negative bacteria, OppA has been shown to select for peptides between 3 and 5 amino acids with high affinity for tri-and tetrapeptides (8). Crystal structures of peptide-bound OppA from Salmonella typhimurium, Yersinia pestis, Escherichia coli, and Burkholderia pseudomallei show peptide specificity is independent of amino acid sequence of the peptides as protein-peptide interactions occur mainly through the peptide backbone (9 -12). Gram-positive Lactococcus lactis OppA binds peptides 4 -35 amino acids long with a preference for nonapeptides (13). Large binding cavities of oligopeptide homologs llOppA, Bacillus subtilis AppA, and Enterococcus

Crystal structure of co-purified peptide nthiOppA complex
We expressed nthiOppA without the periplasmic signal sequence in E. coli and natively purified the protein using a C-terminal His tag. Crystallized nthiOppA was co-purified with bound endogenous peptide, indicated by additional electron density in the substrate-binding pocket. SBPs are frequently co-purified with their substrates, and OppA homologs have demonstrated a broad specificity for peptides (23,24). The structure was solved by molecular replacement to a resolution of 1.85 Å with space group P2 1 2 1 2 1 and one monomer in the asymmetric unit. In general, SBPs have two structurally conserved globular domains consisting of a ␤-sheet that is flanked by ␣-helices (␣/␤-domains), and the five-strand sheet at the core of each domain is connected by two strands to the opposite ␣/␤-domain. The substrate-binding pocket is formed at the interface of these domains. Specific to the Cluster C SBP family, domain I of nthiOppA is divided into two subdomains: domain I A is stitched together from residues 26 to 41, 210 to 290, and 510 to 541, and domain I B includes residues 42 to 209.
Domain II spans residues 291 to 509 (Fig. 1A). The addition of domain I B makes Cluster C proteins larger than other SBPs in other clusters, and this allows for an expanded binding cavity to accommodate the larger substrates associated with this family of SBPs (25).
When compared with open unbound ecOppA (RMSD ϭ 4.42 Å), structural alignment of nthiOppA indicates a conformational change to the closed peptide-bound state (Fig. 1B). In the ligand-bound nthiOppA complex, domains I and II rotate toward the center of the protein to bury the peptide in the substrate-binding pocket. Interestingly, even though domains I A and I B are structurally independent, they rotate as a rigid domain in respect to domain II. The angle of rotation between the two domains is ϳ35°from the open to the closed conformations. The co-purified peptide binds near the hinge region at the interface between the domains, and domain II provides a binding cleft with the majority of protein-peptide interactions for the bound peptide in the substrate-binding pocket.

Characterization of nthiOppA peptide interactions
Using a range of peptides with varying physical properties, we tested the specificity of nthiOppA. Peptides identified as Gram-negative or Gram-positive OppA substrates as well as peptides with an increase in hydrophobicity and length were used in a thermal shift assay. Peptides that increased the stability of nthiOppA (melting temperature, T m ) were identified as

Multifunctional substrate binding of OppA
binding candidates (Fig. S2). The following six binding candidates raised the T m between 2.5 and 6°C: P1 (KKK); P2 (MGG); P3 (LGG); P4 (GIINTL); P5 Long (YLGANGRGGGS); and P6 Brady (bradykinin, RPPGFSPFR). P1 KKK was one of the initial peptides discovered to bind OppA, and peptides P1 KKK and P6 Brady were first crystallized bound to seOppA and llOppA, respectively (9,14,26). Additionally, ecOppA has also been shown to have a preference for positively charged peptides with approximately half of co-purified peptides having at least one arginine, histidine, or lysine (11). The identified binding candidates indicate peptide specificity of nthiOppA is not limited to 3-5 amino acids peptides. nthiOppA can bind positive and hydrophobic peptides as well as longer peptides similar to Gram-positive OppA homologs.
The novel binding candidate peptides were co-crystallized with nthiOppA. Chemically denatured and refolded nthiOppA co-crystallized with P2 MGG , P3 LGG , and P4 GIINTL in the same condition as natively purified nthiOppA with similar crystal morphology and diffraction quality. However, P5 Long co-crystallization produced poorly formed and cracked crystals that reduced diffraction to 9 Å. For the endogenous peptide bound to the natively purified nthiOppA, a 4-residue peptide backbone was built and refined in the electron density. The observed electron density of the second residue of the co-purified peptide is large enough to fit a bulky side chain. However, the density for the side chains cannot be easily interpreted to confidently identify residue assignments, which is likely caused by a heterogeneous mixture of co-purified peptides bound to nthiOppA. Hydrophobic peptides P2 MGG , P3 LGG , and P4 GIINTL have wellresolved electron density representing the bound peptides (Fig.  S2). The overall structure of nthiOppA does not change dramatically between the different bound peptides with RMSD values from 0.07 to 0.33 Å compared with the co-purified peptide complex. While maintaining the protein-peptide interactions, the flexible binding cavity can accommodate longer peptides.
Broad substrate specificity of OppA is mainly regulated through hydrogen bonds to the peptide backbone, with few peptide side chain interactions (11,27). In all of the nthiOppA complexes, the N-terminal end of the bound peptide is secured by a salt bridge (Fig. 2). In addition to the salt bridge, four hydrogen bonds secure the first residue of peptides in the substrate-binding pocket (Tyr-130, two from His-441, and Asp-443). A second salt bridge also contributes to the binding of the C-terminal ends of the tripeptides. For P2 MGG , P3 LGG , and the co-purified peptide, hydrogen bonds to both amino and carboxyl groups of the remaining residues secure the substrate in the binding pocket. P4 GIINTL has the most peptide-backbone hydrogen bonds involving an additional eight residues: Glu-52, Val-54, Asn-388, and Asn-392; two from Arg-437, and Gly-439 and Tyr-509 (Fig. 2D). The P4 GIINTL complex shows nthiOppA-peptide interactions occur with at least the first six residues of a peptide, providing a platform for nthiOppA to bind longer peptides such as P Long and P Brady . Like the tripeptide complexes, the amino and carbonyl groups of the first three P4 GIINTL residues form hydrogen bonds in the binding cavity, and only the carbonyl groups of the remaining peptide residues hydrogen bond with nthiOppA. With extensive peptide back-bone interactions, a salt bridge is not required to stabilize the charged C-terminal end of the hexapeptide.

Flexible binding cavity accommodates bulky ligands
In the ligand-bound closed conformation, several Gram-positive OppA homologs have flexible substrate-binding pockets that vary in size. The binding cavities efPrgZ, bsAppA, and llOppA range in volume even when bound to long peptides, 1600, 2500, and 4900 Å 3 , respectively (16). Alignment of heptapeptide-bound efPrgZ with the P4 GIINTL complex (RMSD ϭ 1.30 Å) shows a conserved salt bridge and hydrogen bonds to the N-terminal residue of the peptide (Fig. S3). Similar to other Gram-positive OppA homologs, the P4 GIINTL complex demonstrates the flexibility of the Gram-negative OppA-binding cavity by expanding to accommodate a larger substrate and maintaining the peptide interactions in the substrate-binding pocket. Comparing the P2 MGG and the P4 GIINTL complexes, rearrangement of backbone loops and side chains in the substrate-binding pocket widen the binding cavity to prevent steric clashes with the side chains of residues 4 and 5 of P4 GIINTL . The side chains of Tyr-267 and His-441 adopt different rotamer conformations by rotating 89 and 88°, respectively, creating a larger binding pocket for the bound peptide (Fig. 3).

Heme specificity of Cluster C family in NTHi
The flexible binding cavity of nthiOppA and the established heme binding of other Cluster C SBPs led us to explore the ability of nthiOppA to bind heme. ABC transporters play a vital role in heme uptake in bacteria. To uncover how heme is shuttled to the inner membrane of this "heme-loving" pathogen, we looked at the conservation of heme ABC transport systems in NTHi. Homologs of other heme SBPs, including Yersinia enterocolitica HemT, Y. pestis HmuT, Shigella dysenteriae ShuT, Bordetella pertussis BhuT, Pseudomonas aeruginosa PhuT, E. coli ChuT, and Vibrio cholerae HutB have not been identified in the H. influenzae genome. NTHi Cluster C proteins provide the vital link in the uptake pathway between the OM receptors scavenging heme from the host and the ABC importers delivering the nutrient to the cytoplasm. We surveyed all of the identified NTHi Cluster C proteins. Despite the low sequence identity between the Cluster C SBP proteins in the NTHi genome (nthiHbpA (24.9% identity), nthiSapA (22.2% identity), and NTHI0310 (26.8% identity) compared with nthiOppA), these proteins have high sequence similarity (Fig. S1) and are predicted to have a similar overall structural topology.
Additional co-crystallization of nthiOppA with heme was attempted in the optimized crystallization condition and was screened for in a new crystallization condition, but no heme co-crystals were obtained. Therefore, we did in silico hemebinding studies of nthiOppA and two other solved Cluster C structures, Glaesserella (Haemophilus) parasuis HbpA and ecNikA. ROSIE ligand-docking program was used to predict the most likely interactions between the co-purified peptidebound nthiOppA complex and heme (28). Representative topranked solutions for the SBPs show a similar location and ligand position of heme in each protein (Fig. 4). Based on the docking solutions, movement of backbone loops and side chains in the Multifunctional substrate binding of OppA binding cavities of all three SBPs allows for the docking of heme near the canonical substrate-binding sites. The top-ranked solutions for nthiOppA indicate the substrate-binding pocket could be large enough to accommodate both heme and peptide (Fig. 4A). The docked heme demonstrates the substrate-binding pocket extends into domain I and is adjacent to the peptide-binding site (Fig. 6B). This suggests heme binding of nthiOppA does not conflict with peptide-protein interactions and the possibility of a heme-specific cleft in the substrate-binding pocket.
Previous studies have reported a wide range of heme equilibrium dissociation constants in the Cluster C SBPs. The heme affinity of ecNikA has previously been measured by tryptophan fluorescence quenching, K D value of 530 nM (29). Two other E. coli SBPs, DppA (dipeptide) and MppA (murein tripeptide), bind heme with estimated binding constants of 10 and 50 M, respectively; a mutant strain lacking both of these SBPs was not able to use heme as an iron source during growth (30). The first identified Cluster C heme-binding protein, hiHbpA, was reported to have weak heme affinity with a K D value of 655 M (19, 31). The equilibrium dissociation constants of ecDppA, ecMppA, and hiHbpA were calculated using native PAGE gelshift assays. One major disadvantage of these assays is that the samples were not at equilibrium during electrophoresis (32). Heme-binding affinities of homologs for the other NTHi Cluster C SBPs have not previously been published. The limitations of the calculated heme affinity for some of the Cluster C SBPs led us to investigate the heme specificity and affinity of each of the NTHi SBPs.
To determine the shared heme-binding functionality of nthiOppA and other NTHi Cluster C proteins, we examined the heme specificity and binding affinity of the SBPs. The heme-protein binding kinetics of nthiHbpA, nthiOppA, nthiSapA, NTHI0310, and ecNikA were measured by SPR. The single-cycle kinetics method was used to measure heme bind-

Multifunctional substrate binding of OppA
ing of the immobilized SBPs, which included five serial injections with increasing concentrations of heme followed by an extended dissociation step. All four of the NTHi SBPs bind heme with variable affinity (Fig. 5). The heme affinity calculated by spectroscopic analysis of ecNikA provided a reference for the heme affinity measured by SPR. The SPR-calculated heme affinity of ecNikA, K D value of 526 nM, closely matches the previously reported K D value of 530 nM (29). Heme binds to nthiOppA with the highest affinity, K D value of 244 nM. The heme affinities of nthiHbpA and NTHI0310 are similar with K D values of 382 and 420 nM, respectively. nthiSapA showed a slightly lower heme binding with a K D value of 1.1 M. The SPR sensograms for each SBP were fit to a two-state reaction model and indicate a 1:1 binding ratio of heme to SBP. The kinetic rate constants and equilibrium dissociation constant for each of the SBPs are summarized in Table 1. The differences in the heme affinities between the SBPs are largely explained by variability in the initial association rates. Despite their variation in their individual canonical substrates, the NTHi SBPs share specificity for heme.

Accommodating heme and peptide in nthiOppA substratebinding pocket
To further elucidate the multisubstrate specificity of nthi-OppA, we compared the heme and peptide interactions with the substrate-binding pocket. Using an SPR assay, we measured the competitive binding between heme and peptide by injecting either ligand alone or together over immobilized nthiOppA. The sum of the individual responses of heme and peptide was used to determine the theoretical response of both heme and peptide binding to the substrate pocket simultaneously. In the case of heme binding independent of peptide binding, we expect the observed SPR response of the combined ligands to correspond to the theoretical sum of the independent heme and the peptide responses. For competitive binding of heme, we expect a reduction of the observed SPR response of an injection with both ligands compared with the theoretical sum of the individual responses of heme and peptide. The SPR competition assay probes the availability of a heme-specific cleft in the presence of bound peptide in the substrate-binding pocket.
Using this experimental design, P1 KKK , P2 MGG , and P5 Long were injected at a high concentration to load the peptide-specific site of the nthiOppA-binding pocket. The SPR sensograms of the individual peptide response and the combined heme and peptide response were collected for each peptide. The theoretical sum of the individual peptide and heme responses was calculated and compared with the observed combined heme and peptide response for each peptide (Fig. 6). The observed SPR response of the combined heme and P1 KKK injection matched the theoretical sum of the responses for both ligands. For P1 KKK , heme does not compete for binding at the peptide-specific binding site, and heme binding is independent of bound peptide (Fig. 6F). This is evidence of a heme-specific cleft in the substrate-binding pocket and corresponds to the heme-docking studies that predict bound peptide does not exclude heme binding. The combined heme and peptide injections of P2 MGG and P5 Long are lower than the theoretical sum of the individual responses. In these cases, the presence of peptide does limit but does not abolish heme binding (Fig. 6, G and H). In the ligandbound closed conformation of P2 MGG and P5 Long complexes, steric hindrance of the bulky and rigid side chains or occlusion of the channel to the heme-specific binding cleft reduces heme binding. These competition assays show heme does not directly compete with peptide binding, and disruptions in heme and

Multifunctional substrate binding of OppA
peptide binding are likely influenced by a heme-specific binding cleft in close proximity to the peptide-binding site.

Differential substrate binding of nthiOppA domains
To further expand our knowledge of nthiOppA substrate binding, we expressed and purified the individual domains of nthiOppA, nthiOppA 1A1B and nthiOppA 2 (Fig. 7, A and B). Using SPR, we determined nthiOppA 1A1B has about 4-fold higher affinity for heme than nthiOppA 2 with K D values of 577 nM and 2.46 M, respectively (Fig. 7, C and D, and Table S3). nthiOppA 1A1B binds heme with a similar affinity to nthiOppA. This corresponds with our heme-docking studies, which identified domain I as playing a major role in the formation of a heme-specific cleft in the binding cavity. An intrinsic tryptophan fluorescence quenching assay determined P4 GIINTL binds nthiOppA 2 with a K D value of 172 nM, and nthiOppA 1A1B has a weaker affinity with a K D value of 11.7 M (Fig. 7, E and F). Peptide binding of nthiOppA is largely mediated through interactions with domain II, and nthiOppA 2 has a similar heme affinity as nthiOppA with a K D value of 754 nM (Fig. S4). Notably, there are two hydrogen bonds and a salt bridge between residues in domain II and the N-terminal residue of the bound peptide. Based on these binding studies, each domain plays a differential role in binding both substrates with domain I directing heme binding and domain II driving P4 GIINTL binding of nthiOppA.

Discussion
NTHi employs Cluster C SBPs to scavenge essential nutrients and adapt to rapidly changing microenvironments in the host. In Moraxella catarrhalis, oppA is necessary for invasion of human respiratory epithelial cells and persistence during infection (33). In addition to nutrient uptake, Gram-negative OppA plays a role in proteostasis of the periplasm. To aid in protein folding, ecOppA acts as a protein chaperone in the periplasm (34,35). OppA can also assist in the recycling of misfolded periplasmic proteins by transporting protease-degraded peptides into the cytoplasm for reuse as a nutrient source.
Broadening our understanding of Gram-negative OppA peptide specificity is important for determining peptide utilization for essential nutrients and signaling pathways. Our data show nthiOppA bound to a longer peptide than previously observed in other Gram-negative OppA structures. The protein-peptide hydrogen bond interactions of the tripeptide-bound nthiOppA structures are maintained in the novel hexapeptide-bound complex and demonstrate peptide recognition of this SBP is independent of the length and amino acid composition of these peptides. The P4 GIINTL nthiOppA complex illustrates the flexibility of the binding cavity that expands to accommodate the hexapeptide. Our findings bridge the peptide specificity of Gram-negative and Gram-positive OppA proteins and highlight the similarities of these SBPs.
During further study of the flexible binding cavity of nthi-OppA, we identified heme as a novel substrate for this SBP. NTHi lacks the necessary enzymes for heme biosynthesis and has a strict growth requirement for heme as a source of iron to sustain aerobic respiration (36). This led us to survey all the Cluster C proteins in NTHi, and we identified that they all have specificity for heme with affinities ranging from 244 nM for nthiOppA to 1.1 M for nthiSapA. Interestingly, the heme affinity of the individual SBP is not a determinant of the role these

Multifunctional substrate binding of OppA
proteins play in the heme-uptake pathway. Considered a lowaffinity heme-binding protein, nthiSapA plays a crucial role in the heme-uptake pathway for NTHi survival after heme starvation (22). Even in bacteria that employ high-affinity heme SBPs, such as HmuT, PhuT, and ShuT, there are functional overlapping heme transport systems. Deletion of the hmu locus in Y. pestis did not eliminate the ability of the mutant strain to utilize heme or colonize mice in the systemic infection model, indicating another heme transport system is sufficient for virulence in the mouse model (37). Functionally similar heme trafficking proteins in the cytoplasm, such as P. aeruginosa PhuS, have comparable heme affinity to Cluster C SBPs (38). These heme-binding proteins demonstrate heme affinity does not dictate functionality or limit the essential role of these proteins in the heme-uptake pathway for bacterial survival and pathogenesis.
NTHi Cluster C SBPs deliver substrate to three ABC transporters in the PepT (peptide/opine/nickel) family: gene clusters dpp, opp, and sap (Fig. 8). In vivo studies have demonstrated the Dpp and Sap importers are important for heme uptake. Deletion of components in the dpp or sap operons reduced the ability of the bacteria to utilize heme and recover after heme-iron starvation, respectively (19,22,44,45). With more Cluster C SBP genes than PepT transporter gene clusters in NTHi, orphan SBPs have versatility in complex assembly within this family and share importers for substrate uptake. NTHi lacks the gene encoding dppA, and nthiHbpA (55% sequence identity to ecDppA) is an orphan SBP without an encoded transporter in its operon. In H. influenzae, HbpA recognizes the Dpp transporter and delivers substrate for import into the cell (19). EcMppA, another orphan SBP, delivers its peptide substrate to the Opp transporter, and both ecMppA and ecDppA are dependent on the Dpp transporter for heme uptake (30,46). Similar to hiHbpA and ecMppA, orphan SBP NTHI0310 likely utilizes one of the PepT importers for substrate delivery. These functionally overlapping Cluster C SBPs are conserved in Haemophilus (Fig. S5).
In the dynamic and perilous host environment, it is advantageous for bacteria to manage multiple heme SBPs to ensure the pathogen maintains access to the essential nutrient. Despite the canonical substrates of Cluster C SBPs ranging in diversity and size from a nickel ion to antimicrobial peptides, the unifying characteristic of these proteins has previously been their overall structure conserved from a common ancestor (47). Here, we have identified multisubstrate specificity for heme is a shared characteristic of NTHi Cluster C SBPs. This functional overlap between the Cluster C SBPs, particularly the previously unknown heme-binding capability of nthi-OppA and NTHI0310, has made it a challenge to fully characterize the heme-uptake pathway. Further studies of the unique and essential role each SBP plays in the transport of heme need to be explored. Better understanding of the interplay between these multifunctional Cluster C SBPs will help us uncover how pathogens adapt and overcome host-mediated defenses.

Materials and methods
All peptides were solubilized according to the manufacturers' recommendations. Peptides P1 KKK , P3 LGG , and P6 Brady (bradykinin, RPPGFSPFR) were purchased from Sigma. Cus-

Multifunctional substrate binding of OppA
tom peptides P2 MGG , P4 GIINTL , and P5 Long (YLGANGRGGGS) were synthesized by GenScript. A 10 mM heme stock solution was prepared by dissolving hemin chloride (Strem Chemicals) in 100% DMSO, and the heme concentration was confirmed using the pyridine-hemochromagen method.

Bioinformatics analysis
The genome of NTHi 86-028NP was searched for Cluster C SBPs using BLAST queries. Four of these SBPs were identified in NTHi, HbpA, OppA, SapA, and a putative peptide-binding protein (NTHI0310). Percent identity and percent similarity comparisons between NTHi SBPs were calculated from pairwise sequence alignments (Table S1) generated by LALIGN using a BLOSUM50 matrix (EMBL-EBI). Alignment of NTHi SBPs was performed using ClustalW and superimposed with the secondary structure of nthiOppA in Jalview version 2 (48).

Expression vectors of SBPs
Expression constructs were amplified from the genomic DNA of the clinical strain of NTHi 86-028NP and E. coli strain K-12 substrain MG1655. Each construct was amplified without their predicted periplasmic signal sequence, nthiHbpA (residues 20 -549), nthiOppA (residues 21-541), nthiSapA (residues 24 -564), NTHI0310 (residues 24 -514), and ecNikA (res-idues 23-524). The SBP constructs were cloned into the pET-21b vector using the NdeI and XhoI restriction sites to create constructs fused to a C-terminal His 6 tag. Domain construct nthiOppA 1A1B was created by site-directed mutagenesis; domain II was deleted and replaced with two glycine residues to fuse the C-terminal end of domain I A to domain I B (21-290 GG 510 -541). nthiOppA 2 (residues 291-509) was cloned using the same method as the other SBP constructs. All plasmids were verified by sequencing (ACGT Inc.).

Protein expression and purification
SBP proteins were expressed in E. coli BL21(DE3) grown in 1-liter cultures of Luria Broth media supplemented with 100 g/ml ampicillin. Cells were grown at 37°C to early log phase (OD 600 of 0.4) and then cooled to 16°C. After the incubator cooled, protein expression was induced (OD 600 of 0.8) with 400 M isopropyl 1-thio-␤-D-galactopyranoside. Cells were cultured overnight and harvested by centrifugation at 5000 rpm and stored at Ϫ80°C. Each step of the protein purification was carried out at 4°C. Bacterial cells were resuspended in buffer A (25 mM HEPES, pH 8, 500 mM NaCl, 15 mM imidazole, pH 8) and lysed with an S-4000 sonicator (Misonix Sonicators). The cell debris was removed by centrifugation at 17,000 rpm for 1 h, and the soluble fraction was loaded on to an equilibrated nickelnitrilotriacetic acid affinity chromatography column (Thermo-Fisher Scientific). The 5-ml column was rinsed with 10 column volumes of buffer A, followed by 5 column volumes of buffer B (buffer A, supplemented to 30 mM imidazole, pH 8). The SBPs were eluted with 8 column volumes of buffer C (buffer A, supplemented to 250 mM imidazole, pH 8). Eluted protein (determined by SDS-PAGE to be Ͼ95% pure) was dialyzed overnight in buffer D (25 mM HEPES, pH 7.5, 500 mM NaCl). The protein was applied to a HiLoad 16/600 Superdex 200 size-exclusion chromatography column (GE Healthcare). Buffers for nthiH-bpA and NTHI0310 were supplemented with 5 mM 2-mercaptoethanol. Protein fractions were pooled and concentrated to 20 mg/ml. Proteins were stored in buffer D at Ϫ80°C until needed.
For co-crystallization of nthiOppA with peptides, the protein was chemically denatured and refolded on the affinity column (see supporting Experimental procedures). The bacteria cells were resuspended, sonicated, and centrifuged as noted above. After the soluble fraction was applied to the affinity column and washed with 10 column volumes of buffer A, the protein was then denatured with 20 column volumes of denaturing buffer (6 M GdnHCl, 25 mM HEPES, 15 mM imidazole, pH 8). The column was rinsed with 10 column volumes of refolding buffers with decreasing concentrations of GdnHCl (buffer A supplemented with 3, 1.5, 1, and 0.5 M GdnHCl). To remove the remaining GdnHCl, the column was washed with 1 column volume of buffer A, and the protein was eluted with 8 column volumes of buffer C. The protein was dialyzed, applied to the Superdex column, and stored as mentioned above.

Protein crystallization
Crystallization of the co-purified peptide nthiOppA complex (25 mM HEPES, pH 7.5, 500 mM NaCl) was achieved by the vapor diffusion in sitting drops at 22°C. The nthiOppA crystals A, the human host has many iron/heme reservoirs, including hemoglobin, haptoglobin, myoglobin, and serum albumin. NTHi has many OM receptors (Hup, HgpABC, HxuC, and HemR) to scavenge heme, and each OM receptor recognizes a unique combination of host hemoproteins, hemophores, or free heme for utilization in the uptake pathway. The TonB/ExbB/ ExbD complex provides energy to the OM receptors to transport the key nutrient across the membrane. SBPs acquire substrate from other components of the transport system, including metallochaperones, and the receptor/TonB complex (39 -42) and deliver it to their cognate ABC transporter, powered by ATP hydrolysis, for import into the cell (adapted from Faraldo-Gomez and Sansom (43)). B, gene clusters of NTHi Cluster C SBPs and PepT importers. The opp and sap operons contain the SBP, two TMDs and two NBDs. The dpp operon includes the TMDs and NBDs without the corresponding dppA SBP. The remaining gene clusters include the orphan SBPs, hbpA and NTHI0310.
were obtained with a 1:1 ratio of 10 mg/ml protein and reservoir solution containing 0.1 M sodium acetate at pH 4.6 and 2.4 M ammonium sulfate. Crystals usually appeared within 24 h. Cocrystallization of nthiOppA with peptides was achieved using a 1:10 refolded OppA to peptide mixture yielding final concentrations of 1 mM peptide and 6 mg/ml protein in binding buffer (25 mM HEPES, pH 7.5, 150 mM NaCl). The mixture was incubated on ice for 30 min. Co-crystals grew in the same reservoir solution and conditions. All crystals were briefly soaked in cryoprotectant consisting of reservoir solution supplemented with 15% v/v glycerol, harvested with a nylon loop, and flash-cooled in liquid nitrogen.

X-ray diffraction data collection and structure determination
Diffraction data were collected at the Advanced Photon Source (Argonne, IL) LS-CAT beamline 21-ID-D with an Eiger X 9M detector (DECTRIS AG). The diffraction data were integrated using XDS (49) and scaled with AIMLESS, from the CCP4 suite (50). The initial model for nthiOppA was determined using the molecular replacement pipeline Balbes (51), and model building was further improved with ARP/warp (52) from the on-line CCP4 platform. The output model was manually rebuilt over several cycles with Coot (53). Each peptide was built in the observed electron density, and the models were refined with REFMAC (54). Validation statistics of the final models were calculated with Molprobity (55). Details of data quality and structure refinement are summarized in Table 2. Also, see Table S2 for additional details of data quality and structure refinement for P1 KKK , P5 Long , and P6 Brady peptides. Coordinate files have been deposited in the Protein Data Bank under the accession codes 6DQQ, 6DQR, 6DQT, 6DQU, 6DTF, 6DTG, and 6DTH. Structural figures, analysis of nthiOppA substrate-bound states, and sequence-independent structural alignments with RMSD calculations were performed in PyMOL version 2.0 (Schrödinger, LLC). Protein-peptide hydrogen bond lengths were calculated in LigPlotϩ version 1.4 (56).

Thermal shift assay
Peptide binding candidates were identified using a thermal shift assay of nthiOppA. Reactions of 2 g of refolded nthi-OppA, 5ϫ SYPRO Orange (ThermoFisher Scientific), and 1 mM peptide in binding buffer (25 mM HEPES, pH 7.5, 150 mM NaCl) were placed in a 384-well PCR plate. The plate was heated from 25 to 95°C with a heating rate of 0.5°C/min. The fluorescence intensity was measured with an excitation wavelength of 470 nm and emission wavelength of 580 nm. The thermal shift assay was performed using a Bio-Rad CFX384 Real-Time Detection System (Bio-Rad).

In silico heme-docking studies
Ligand docking of heme was conducted with ROSIE (Rosetta Online Server that Includes Everyone) ligand-docking protocol  (28). Along with the structure of nthiOppA that we solved, we looked at other previously solved structures for gpHbpA (PDB code 3M8U) and ecNikA (PDB code 3DP8). The gpHbpA structure has 74% sequence identity to nthiHbpA. The solvent-free structure of each SBP was used as the template, and heme with a 2 ϩ formal charge was used as the ligand. The substrate-binding pocket was probed with a 7 Å search radius starting at x ϭ 38.0, y ϭ 43.5, and z ϭ 10.5, and the search radius was centered at the interior of the protein. For each protein, 200 structures were generated and ranked based on the calculated lowest interface energy. Of the 20 top-ranked ligand poses for nthi-OppA, heme was docked in the same position in the substratebinding pocket in 18 of the solutions. For hiHbpA and ecNikA there were 11 and 19 solutions with the same location, respectively. For the remaining solutions in the top 20, the heme molecule is overlapped with the most common ligand pose, but the porphyrin ring is slightly rotated. Top-scoring docking models were displayed using PyMOL.

Heme affinity determined by surface plasmon resonance
SBPs were coupled via a standard amine-coupling method in flow channels 2, 3, and 4 on a CM5 sensor chip (GE Healthcare); 1000 -2000 RU of each SBP was immobilized. Flow channel 1 was designated the control channel. The surface of the chip was equilibrated in running buffer (25 mM HEPES, pH 7.5, 150 mM NaCl, 0.1% Tween 20, 2% DMSO) at a flow rate of 100 l/min for at least 3 h. The data were obtained using single-cycle kinetic experiments in triplicate. For each replicate, five analyte (heme) injections were prepared by 2-fold serial dilution. The analyte samples were consecutively injected by increasing heme concentrations over all four channels to determine the binding constants of the SBPs. The association time of each analyte injection was 1 min, followed by a final 5-min dissociation step. SBP kinetic experiments were run at a flow rate of 40 l/min at 25°C. The sensor surface was regenerated after each experiment with two 30-s injections of running buffer, supplemented with 0.1% SDS, at a flow rate of 50 l/min. All experiments were performed with the Biacore T200 instrument (GE Healthcare) according to the manufacturer's instructions. Kinetic rate constants and equilibrium dissociation constants were determined by fitting the data globally to the 1:1 two-state reaction model using the Biacore T200 evaluation software version 3.0 (GE Healthcare).

Surface plasmon resonance competition assay
After nthiOppA immobilization (3300 RU), the chip was equilibrated in running buffer as described above. For the competition assay, analytes were injected for 1 min at a flow rate of 40 l/min at 25°C. A 500 nM heme injection was used to calculate the heme response. Individual injections of P1 KKK , P2 MGG , and P5 Long were measured at 250, 1.5, and 250 nM, respectively. Maintaining the concentration of each analyte, combined mixtures of heme and peptide were injected. The sensor surface was regenerated after each experiment with a 30-s injection of running buffer, supplemented with 0.1% SDS, at a flow rate of 50 l/min. The response for each analyte was observed in triplicate. All experiments were performed with the Biacore T200 instrument (GE Healthcare) according to the manufacturer's instructions.

Circular dichroism (CD) spectroscopy
The nthiOppA domain samples were diluted to 16 M in 5 mM Tris, pH 7.5, 15 mM NaCl. Far-UV CD spectra were collected using a 0.1-cm quartz cuvette from 260 to 190 nm in the step scan mode, with a 2-nm bandwidth, a 4-s response time, and a 1.0-nm step speed. Each spectrum is the accumulation of three scans. CD analysis was performed using a J-815 CD spectrometer (Jasco). The spectra show strong helical and ␤-sheet secondary structure characteristics indicating nthiOppA 1A1B and nthiOppA 2 are well-folded domains (Fig. S4).

Intrinsic tryptophan fluorescence quenching
Solutions of 100 nM nthiOppA, nthiOppA 1A1B , and nthi-OppA 2 were prepared in reaction buffer (25 mM HEPES, pH 7.5, 150 mM NaCl) for steady-state fluorescence experiments. P4 GIINTL was titrated into the protein solution in steps of 0.2 or 10 M. After each titration step, samples were stirred for at least 5 min and then left to rest for at least 5 min at room temperature. Samples were excited at 295 nm, and the fluorescence maximum was observed at 329 nm. UV-grade polyacrylic cuvettes with 1-cm path length and excitation and emission slits of 1 mm were used for data collection. All fluorescence intensity experiments were performed using a PC1 photoncounting steady-state fluorometer (ISS). The percent quenching of tryptophan fluorescence at 329 nm versus the P4 GIINTL concentration was fit to a single-site binding model using GraphPad Prism 6.0.