For complex formation and structural studies, we recombinantly expressed and purified the soluble construct of LbpB from N. meningitidis (NmLbpB) (residues 20–737) and N. gonorrhoeae (NgLbpB) (residues 20–728) (Figure 1B). Solid-phase-binding assays showed that these purified LbpB constructs were sufficient for binding holo-Lf (Figure 1C and Figure 1—source data 1). For large-scale studies, Lf was purchased and further purified using SEC. Complexes of LbpB with Lf were formed upon incubation of purified LbpB with an excess of Lf and separated using SEC, with the complex eluting as a left-shifted peak (Figure 1D). Sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS–PAGE) analysis confirmed the peak contained both components and suggested a 1:1 stoichiometric complex for both the NmLbpB and NgLbpB complexes with Lf (Figure 1E and Figure 1—source data 1).

It has been previously reported that NmLbpB may have multiple Lf-binding sites and may form higher-order complexes (Ostan et al., 2017). However, in our studies with complex formation by SEC, we observed that NmLbpB appears to form a stable 1:1 complex with Lf. To further confirm this, we studied the stoichiometry of LbpB (Nm and Ng) in complex with Lf in solution using inline SEC-SAXS. The SAXS curve provides information about the shape, molecular weight, and maximum length of a protein or protein complex. It is a great tool for comparing the structural effects of different ligands bound LbpB and for comparing the structures of LbpB from Nm and Ng to determine how closely they match. Further, SAXS allows us to compare how closely our structures match what is observed in solution by seeing how well the simulated SAXS curve from our structure fits the experimental SAXS curve. For our experiments here, our SEC-SAXS studies showed that the individual LbpB and Lf proteins existed as monomers in solution, while the NmLbpB–Lf complex had an experimental molecular weight of 159.6 kDa corresponding to 1:1 stoichiometry complex of NmLbpB and Lf (calculated molecular weight is 156.6 kDa) (Figure 2A and B and Supplementary file 1a). The calculated radius of gyration (R_g) and maximal intramolecular dimension (D_max) are both consistent with a 1:1 stoichiometric complex. Further, we used SEC-SAXS analysis to show that in solution, NgLbpB alone and a complex with Lf match closely to those with NmLbpB (Figure 2A and B, and SI Supplementary file 1).

Static small-angle X-ray scattering (SAXS) analysis showed that NmLbpB alone tends to aggregate in solution, while the NmLbpB–Lf complex forms a monodisperse stable complex (Figure 2—figure supplement 1). Therefore, crystallization screening focussed on both Nm and NgLbpB–Lf complexes using commercial broad-matrix screens. Initial lead conditions were optimized, however, diffraction quality crystals could only be grown of the NmLbpB–Lf complex. Data were collected at the GM/CA ID-D beamline at the Advanced Photon Source and the structure of the NmLbpB–Lf complex solved to 2.85 Å resolution using molecular replacement in space group P4₃2₁2. Model building was performed in COOT and the structure refined using phenix.refine to final R_work/R_free values of 0.20/0.25 (Figure 3A and Supplementary file 1b). One complex was observed in the structure containing one molecule of Lf (consisting of ordered N- and C-lobes) and one molecule of NmLbpB (consisting of an ordered N-lobe, but partially ordered C-lobe).

The overall structure of the NmLbpB–Lf complex has an overall L-shape (Figure 3B). The N-lobe of NmLbpB is composed of residues 1–342 forming two domains, an N-terminal handle domain (residues 45–173) and an eight-stranded β-barrel domain (residues 174–342). The handle domain contains four anti-parallel β-strands and an α-helix between β-strand 1 and 2. The β-barrel domain is packed against the β-sheet of handle domain. The C-lobe of NmLbpB is composed of residues 359–718 and found largely disordered in the crystal structure as evident from higher B-factors for this region compared to the rest of the structure (Figure 3C). The C-lobe also contains a handle domain (residues 359–540) and an eight-stranded β-barrel domain (residues 541–718). While the C-lobe β-barrel domain shares homology with the N-lobe β-barrel domain, the C-lobe handle domain is different as it is composed of a six-stranded β-sheet packed against the β-barrel domain and flanked by two antiparallel β-strands. Importantly, the C-lobe of LbpB contains several long anionic loops, however, none of these were observed in the electron density likely due to their flexibility.

The Lf structure also consists of N- and C-lobes with each lobe further divided into two subdomains termed N1 and N2 (N-lobe) and C1 and C2 (C-lobe). In the presence of iron, the subdomains of each lobe are bridged together forming the closed state (Baker and Baker, 2005; Sill et al., 2016). However, in absence of iron, these lobes are known to undergo conformational changes to adopt an open conformation (Baker and Baker, 2005; Sill et al., 2016). Structural alignment of Lf alone (PDB ID 2BJJ) with Lf in our complex (RMSD 1.3 Å) shows that both lobes are in closed conformations, as expected since we observe an iron in each lobe, and that Lf undergoes minimal changes upon binding NmLbpB (Figure 3D). Further, the calculated SAXS scattering curve from our crystal structure agrees well with our experimental SAXS scattering curve in solution (Figure 2H1).

The interaction of NmLbpB with Lf is mediated exclusively through the N-lobe of NmLbpB with the C-lobe of Lf (Figure 3B). This interface consists of regions along both the C1 and C2 domains of Lf including residues D630, S637, T639, K640 (C1 domain) and E512, R525, E538, N539, D561 (C2 domain) (Figure 3B and E). Interacting simultaneously with the C1 and C2 domains of Lf, the N-lobe of NmLbpB has an extended interaction interface that includes residues N154, R193, S201, D204, N213, R223, K230, Y253, Q255, and S258 (Figure 3B and E). Given our crystal structure indicated that binding to Lf was exclusively along the N-lobe of NmLbpB, we used solid-phase-binding assays to show that indeed Lf could only interact with the full length and N-lobe only constructs, but not with the C-lobe only (Figure 3F and Figure 3—source data 1f).

Crystallization with NgLbpB was unsuccessful, therefore, we used cryoEM to determine the structure of NgLbpB in complex with Lf to 3.65 Å resolution (Supplementary file 1c). Very good density was observed for side chains of both lobes of Lf and for the N-lobe of NgLbpB allowing unambiguous model building and refinement, however, almost no density was observed for the C-lobe of NgLbpB, unless map contours were reduced drastically (Figure 4A and B and Figure 4—figure supplement 1). An alignment of the Nm and NgLbpB complexes with Lf had an overall RMSD of 3.7 Å, however, an improved RMSD of 1.4 Å when aligning just along the interacting domains (C-lobe of Lf and N-lobe of LbpB), indicating a conserved binding interface between Nm and Ng (Figure 4C and D). Additionally, the N-lobe of Lf in the NgLbpB–Lf cryoEM structure is positioned ~12° relatively to the NmLbpB–Lf crystal structure. Whether this structural difference in the N-lobe of Lf is just an artifact of crystal packing or if it is truly important for function in LbpB remains to be determined.

In the Nm/NgLbpB–Lf complex structures, LbpB interacts with both the C1 and C2 subdomains within the C-lobe of Lf in an extended interface (Figures 3B, E ,, 4C and D). In the NmLbpB–Lf complex, the interaction interface is stabilized by extensive hydrogen bonding and salt bridge interactions encompassing a buried surface area of ~1760.8 Ų (Supplementary file 1d). Based on analysis of our crystal structure, we performed mutagenesis of interface residues in NmLbpB important for mediating Lf binding (Figure 5). Those included K71, Q73, R135, K140, K143, R193, D204, T209, Y210, R223, and D227, most of which are conserved in NgLbpB (Figure 4D and Supplementary file 1e). These mutants all expressed similar to wild type and could be purified using the same method, indicating they were stably folded. The effect of these mutations on Lf binding was first tested with solid-phasing binding assays (Figure 4—figure supplement 2 and Figure 4—figure supplement 2; Source data 1) and with enzyme-linked immunosorbent assays (ELISAs) and compared to wild type (Figure 5B). The results showed that all of the mutations had a significant effect on Lf binding with most resulting in more than ~50% reduction in binding except for mutations Q73E, K143E, and R223E. Those having at least a ~75% reduction in binding included R135E, D140K, R193E, D204K, D227K, R135E/R223E, and T209N/Y210K. To further characterize Lf binding to the mutants, we performed isothermal titration calorimetry (ITC) analysis using a MicroCal iTC200 to determine thermodynamics parameters of binding, including K_d values, and compared the results of the mutants to wild type (Figure 5C and Supplementary file 1f). Representative ITC plots show a K_d value of 0.14 μM for Lf binding to wild-type NmLbpB with a ΔH of −15.8 kcal/mol and an n value of ~1. Agreeing well with the ELISA analysis, ΔH values for the all the mutants except for K143E were significantly reduced ranging from −5.4 to −11.8 kcal/mol (R135E, D140K, R193E, and D227K) and accompanied by increases in K_d values ranging from 9- to 115-fold over wild type.

To gain more insight into the binding of Lfcn with LbpB, we performed SEC-SAXS experiments with and without Lfcn, finding that the presence of Lfcn produced a small, but reproducible, change in the scattering profile at lower q values in the range of 0.04–0.08 (Figure 2C and D). A similar change along lower q values in the range of 0.05–0.07 was also observed for the NmLbpB–Lf complex when Lfcn was added (Figure 2E). Together, these data support the hypothesis that Lfcn binding is distinct from Lf binding, further supporting that Lfcn binding is along the C-lobe of LbpB. While we were able to co-crystallize Lfcn with NmLbpB–Lf and solve the structure, the C-lobe of LbpB remained highly disordered and therefore, we were unable to determine the binding interactions. Therefore, to probe the putative binding site of Lfcn, we analyzed the NmLbpB–Lf structure to determine the boundaries of the C-lobe loops (Figure 6A). An electrostatic surface representation of this region shows it to be highly charged with at least one large strongly electronegative region along loops consisting of residues 372–383, 416–418, and 445–526; a more accurate analysis of the electrostatics here is not possible given the lack of order in the loops (Figure 6B).

Based on these loop boundaries, we made deletion mutants with each lacking one of the loops of the C-lobe of NmLbpB. Albeit to varying levels, these mutants all expressed and could be purified using the same method as for wild type, indicating they were stably folded. We then analyzed the effects of these deletions on Lfcn binding in comparison to wild type using ITC analysis on a Nano ITC instrument (Figure 6C and SI Table S7). Our results show that Lfcn binds to wild-type NmLbpB with a K_d of 3.7 μM, a ΔH of −11.4 kcal/mol, and an n value of ~1. The loop deletions Δ372–383, Δ416–418, Δ561–565, Δ594–599, and Δ665–698, however, showed only minimal differences in ΔH and no significant change in K_d, except for Δ445–526 (ΔH of −9.6 kcal/mol and a ~2.5-fold increase in K_d). All n values here are larger than 1 and roughly twofold higher than that observed with Lf (n value slightly below 1), an observation we attribute to inaccuracies in using absorbance measurements at λ₂₈₀ for determining protein and peptide concentrations. Loops containing residues 445–526 and 665–698 contain charged patches that have been shown to affect Lfcn binding (Morgenthau et al., 2014a; Ostan et al., 2017), however, individually removing them did not appear to abrogate binding in our study. However, a double loop-deletion mutant Δ445–526_Δ665–698 showed a significantly reduced ΔH with only barely detectable binding. This confirms that these two loops together contribute to Lfcn binding in LbpB, which is likely mediated by the charged residues they contain. Exactly how Lfcn binding is mediated though by these loops will require additional structural studies.

Our SEC-SAXS experiments suggested that Lf and Lfcn had distinct, non-competing, binding sites on LbpB. Therefore, we next performed ITC experiments using a Nano ITC instrument to determine the binding properties of Lf to NmLbpB in the absence or presence of Lfcn. Our results show that Lf binds to NmLbpB alone with a K_d of 0.45 μM and a ΔH of −15.1 kcal/mol (SI Table S8). Nearly identical binding parameters are observed for Lf binding to the preformed NmLbpB–Lfcn complex, indicating that the presence of Lfcn has no effect on Lf binding (Figure 6D and SI Table S8). We next tested the effect of the presence of Lf on Lfcn binding. Our earlier results showed that Lfcn binds to NmLbpB alone with a K_d of 3.7 μM and a ΔH of −11.4 kcal/mol, while here we found that Lfcn binds to the NmLbpB–Lf complex with a K_d of 7.3 μM and a ΔH of −10.6 kcal/mol (Figure 6E and SI Table S8). Only ~ twofold difference in K_d was observed here, however, no change in ΔH indicating that binding of Lfcn to NmLbpB is not disrupted in the presence of Lf.

The codon optimized LbpB from N. meningitidis strain MC58 (NmLbpB) was purchased from Bio Basic. For expression, a gene fragment encoding residues 20–737 was subcloned in expression vector pHIS-Parallel2 with N-terminal 6× His-tag and Tobacco Etch Virus (TEV) protease cleavage site. For large-scale expression, the expression vector was transformed into BL21(DE3) cells and protein expression performed at 24°C upon induction with 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) for 9–12 hr. Cells were harvested by centrifugation at 6000 rpm and stored at −80°C until further use.

For purification, frozen pellets were thawed and resuspended in lysis buffer (20 mM Tris–HCl, pH 7.5, 150 mM NaCl, 5 mM β-mercaptoethanol [BME], 25 mM imidazole) supplemented with 10 µg/ml DNaseI and 0.5 mM phenylmethylsulfonyl fluoride. Cells were lysed by three passages through Emulsiflex C3 high pressure homogenizer (Avestin). The cellular debris were removed by centrifugation at 18,000 rpm for 30 min at 4°C. The supernatant was collected and subjected to affinity purification using 5 mL Hi-Trap column containing Ni-NTA resin (Qiagen) on an AKTA Pure 25 L protein purification system (GE Healthcare). The column was washed with 10 column volume (CV) lysis buffer and protein was eluted using elution buffer (20 mM Tris–HCl pH 7.5, 150 mM NaCl, 5 mM BME, 250 mM imidazole). The eluted protein was subjected to TEV protease digestion to remove the 6× His-tag. The digested protein was cleaned up using a Ni-NTA affinity column and the flowthrough collected. The concentrated flowthrough was further purified using SEC on a Superdex 200 10/300 GL Increase (GE Healthcare) column using 20 mM Tris–HCl pH 7.5, 150 mM NaCl. The elution fractions were run on an SDS–PAGE gel to examine purity. The fractions containing NmLbpB were pooled and concentrated for further experiments.

The DNA sequences for the individual lobes (N- and C-lobes) encoding residues 20–362 and 381–737, respectively, were amplified using full-length NmLbpB as template. These fragments were ligated into pHIS-Parallel2 with N-terminal 6× His-tag and TEV protease cleavage site. Also, point mutants were created using wild-type NmLbpB plasmid using standard site-directed mutagenesis methods. These constructs were expressed and purified using the similar approach as full-length NmLbpB as described above.

The NgLbpB gene fragment encoding residues 20–728 was amplified from the genomic DNA of N. gonorrhoeae strain FA19 with primers containing restriction sites for NcoI and XhoI. The gene fragment was cloned into pHIS-Parallel2 plasmid for expression. The construct was confirmed by DNA sequencing. The expression and purification of NgLbpB were performed using the same approach as for full-length NmLbpB as described above.

For complex formation, purified Nm/NgLbpB was incubated with twofold molar excess of human holo-Lf (RayBiotech, Inc) at 4°C for 1 hr. The complex was separated from excess Lf using SEC on a Superdex 200 10/300 GL Increase column (GE Healthcare). Complex formation was confirmed using SDS–PAGE gel. The fractions containing Nm/NgLbpB–Lf were pooled and concentrated for further experiments.

For dot blot assays with NmLbpB wild type and mutants, 10 µl of each sample was spotted on polyvinylidene difluoride (PVDF) membrane and air dried. The membrane was then blocked by incubation of the membrane in 5% bovine serum albumin (BSA) for 1 hr followed by three washes with 1× PBST (1× phosphate buffered saline [PBS] + 0.005% Tween-20) 5 min each. Potential ligand binding was performed by incubating the blots in 20 µg/ml holo-Lf in PBST +0.5% BSA for 1 hr. Unbound ligand was removed by three washes with PBST. The blots were probed using horse radish peroxidase (HRP) enzyme-conjugated rabbit anti-human Lf antibody (RayBiotech, Inc) for 1 hr. Excess antibody was removed by washing using three washes of 1× PBST and 1× PBS each. Finally, enhanced chemiluminescence (ECL) substrate was used for the detection of HRP’s enzymatic activity.

For dot blot assays with NgLbpB wild type and mutants, 10 µl of each sample was spotted on activated PVDF membrane and air dried. 1× PBS with BSA (5%) was used for blocking unoccupied sites followed by three washes with 1× PBST. The HRP-conjugated holo-Lf (Lf-HRP) probe was used to monitor Lf binding. Excess probe was removed by washes with 1× PBST and 1× PBS. Finally, ECL substrate was used for the detection of HRP activity.

All dot blot assays were performed at least in triplicate with representative data shown.

Purified samples were subjected to inline SEC-SAXS. The scattering data were collected at beamline 18-ID of Biophysics Collaborative Access Team (BioCAT) of the Advanced Photon Source, Argonne National Laboratory. The data were analyzed and final plots made using BioXTAS RAW (Hopkins et al., 2017) and ATSAS (Manalastas-Cantos et al., 2021). First, the data were reduced and the data range for scattering curves were selected. Upon averaging of the data, the q-range and molecular weight information were obtained by analyzing Guinier plot. The pair-distance distribution curves were calculated using GNOM. Theoretical scattering for the X-ray and cryoEM structures was calculated and compared with experimental scattering curves using Crysol.

For crystallization, NmLbpB–Lf was concentrated to 7.4 mg/ml. Initial crystallization trials were performed using commercial screens using hanging-drop vapour-diffusion method at 20°C. Crystal hits identified from these screens were optimized. Diffraction quality crystals of NmLbpB–Lf were obtained in 2.0 M ammonium sulphate, 20 mM Tris, pH 8.5 at 20°C. Crystals were harvested and diffraction data were collected at the GM/CA 23ID-D beamline at the Advanced Photon Source, Argonne National Laboratory using Pilatus 6 M detector at 100 K. The data were processed using HKL2000 (Otwinowski and Minor, 1997) in space group P4₃2₁2 with unit cell parameters a = b = 120.39, c = 207.38, α = β = γ = 90.0.

Initial phases were calculated by molecular replacement within PHASER (Phenix) (Liebschner et al., 2019; McCoy et al., 2007) using the structure of diferric human Lf (PDB ID: 2BJJ) and structure of the N-lobe of NmLbpB (PDB ID: 4U9C). Furthermore, the model was build using Coot (Emsley et al., 2010) and refined using phenix.refine (Phenix) (Liebschner et al., 2019). The final model was refined to R_work and R_free values 0.20 and 0.25, respectively, at 2.85 Å resolution. The geometric analysis was performed using Molprobity (Phenix) (Chen et al., 2010; Liebschner et al., 2019). Structural analysis and figure preparation were performed using PyMol (Schrödinger). Final figures were assembled using Adobe Photoshop and Illustrator.

Purified NgLbpB–Lf complex was applied to Quantifoil R 3.5/1 Cu 200 grids, that were first glow discharged using a Pelco EasiGlow instrument, and plunge-frozen using Vitrobot Mark IV (Thermo Fisher Scientific). Grids were screened and the grid with optimal particle distribution and ice-thickness was used for data collection using a Titan Krios G1 microscope equipped with a Gatan K3 direct electron detector and a Gatan Quantum energy filter. Movies were collected in super-resolution counting mode (pixel size 0.54 Å) with a total dose of 53.68 e⁻/Ų. A total 4966 movies were collected and subjected to motion correction using MotionCor2 (Zheng et al., 2017). The subsequent data processing was performed using cryoSPARC (Punjani et al., 2017). CTF parameters were obtained using ‘Patch CTF estimation (multi)’. An initial set of templates were made by first using ‘Blob picker’ followed by ‘2D Classification’. The best 2D classes were then fed into ‘Template picker’ and subjected to multiple rounds of 2D classification to remove junk particles. The best classes representing different particle orientation were then used for ‘Ab-initio Reconstruction’. After two rounds of ‘Ab-initio Reconstruction’ followed by ‘Heterogeneous refinement’, the best-looking class was subjected to ‘Non-uniform Refinement’, which yielded a final reconstruction to 3.65 Å resolution. A model of the NgLbpB–Lf structure was prepared based on the NmLbpB–Lf crystal structure and placed and fit within the map using ChimeraX (Pettersen et al., 2021). The cryoEM map contained well-resolved density at the protein–protein interface, however, the C-lobe of NgLbpB could only be placed as a rigid body given only minimal density even when map contours were reduced. The structure was refined using ‘Real-space refinement’ within Phenix (Liebschner et al., 2019) and figures and analysis performed using ChimeraX (Pettersen et al., 2021) and PyMOL (Schrödinger). Final figures were assembled using Adobe Photoshop and Illustrator.

For binding analysis with Lf, we performed sandwich ELISAs. The NmLbpB variants were coated on hydrophobic plates by first incubating the plates with 0.1 mg/ml of each protein at 4°C overnight. Acinetobacter baumannii BamB (a kind gift from Robert Stephenson) was included as a negative control. After incubation, wells were washed three times with 1× PBST and blocked with 5% BSA for 30 min. Subsequently, BSA was removed and the wells were washed three times with 1× PBST. For ligand binding, the wells were incubated with the Lf for 30 min. Following that, plates were washed three times with 1× PBST and probed with HRP-conjugated rabbit anti-human Lf antibody (RayBiotech, Inc) for 30 min. The antibody was washed away with three washes of 1× PBST and 1× PBS. The assay was then developed using 1-Step Ultra TMB-ELISA substrate solution (Thermo Scientific). Finally, the assay was stopped upon addition of 0.18 M sulfuric acid. For quantitative data, the absorbance at 450 nm was measured using a Spectramax M2e microplate reader (Molecular Devices). All the experiments were performed in triplicate and repeated at least twice. Similarly, to study Lf binding of the NgLbpB wild type and mutants, ELISAs were performed using HRP-conjugated holo-Lf (Lf-HRP) as a probe for Lf binding. Pisum sativum Toc75 POTRA (a kind gift from Karthik Srinivasan) was included as a negative control.

All ELISA experiments were performed in triplicate with standard errors shown.

To monitor the properties of Lf binding to the NmLbpB variants, ITC experiments were performed using a MicroCal iTC200 ITC calorimeter (Malvern Panalytical). The NmLbpB variants and Lf proteins were buffer exchanged into 1× PBS (PBS: 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, pH 7.5) by SEC. The NmLbpB variants and Lf were concentrated to 30 and 300 µM concentration, respectively. Lf was titrated into ITC sample cell containing the NmLbpB variants at 25°C. Data analysis was performed using Origin 7.0 (OriginLab).

Lfcn binding to wild type and loop-deletion mutants of NmLbpB was measured using a Nano ITC instrument (TA instruments). The sample cell contained 30 µM of protein with 600 µM of Lfcn injected over 20 injections with stirring at 300 RPM at a temperature of 10°C. The ITC data were analyzed using minimized independent binding model with the instrument NanoAnalyze software.

ITC experiments to study the effect of presence of Lfcn on Lf binding to NmLbpB and vice versa, were performed using a Nano ITC instrument (TA instruments). The protein in the sample cell was loaded at 30 µM concentration with stirring at 300 rpm at 25°C for Lf and at 10°C for Lfcn titrations at concentrations of 300 and 600 µM, respectively. The ITC data were analyzed using a minimized independent binding model using the instrument NanoAnalyze software.

All ITC experiments were performed with at least two replicates with representative data shown.