Crystal structures of the outer membrane transporter FoxA provide novel insights into TonB-mediated siderophore uptake and signalling

Many microbes and fungi acquire the essential ion Fe3+ through the synthesis and secretion of high-affinity chelators termed siderophores. In Gram-negative bacteria, these ferric-siderophore complexes are actively taken up using highly specific TonB-dependent transporters (TBDTs) located in the outer bacterial membrane (OM). However, the detailed mechanism of how the inner-membrane protein TonB connects to the transporters in the OM as well as the interplay between siderophore- and TonB-binding to the transporter is still poorly understood. Here, we present three crystal structures of the TBDT FoxA from Pseudomonas aeruginosa (containing a signalling domain) in complex with the siderophore ferrioxamine B and TonB and combine them with a detailed analysis of binding constants. The structures show that both siderophore and TonB-binding is required to form a translocation-competent state of the FoxA transporter in a two-step TonB-binding mechanism. The complex structure also indicates how TonB-binding influences the orientation of the signalling domain.


Introduction
Iron is one of the most abundant elements on earth and is essential for life. However, the bioavailability of iron in the environment is extremely low, and under aerobic conditions iron is found mostly as insoluble hydroxides. The demand for the ionic form of iron by all microorganisms growing in iron-limited conditions has led to the evolution of several efficient iron scavenging strategies. One of the predominant mechanisms by which microbes and fungi acquire iron is through the synthesis and secretion of small, specific high-affinity chelators termed siderophores, which keep iron in a chelated, soluble state (1). In Gramnegative bacteria, these ferric-siderophore complexes are actively taken up into cells using highly specific TonB-dependent transporters (TBDTs) situated in the outer bacterial membrane (OM) as well as specific transporters present in the bacterial inner membrane (IM) (2). The energy for this uptake process is derived from the proton motive force and relies on the energizing complex consisting of TonB/ExbB/ExbD situated in the bacterial IM (3,4).
TonB acts as a physical link between the transporters in the OM and the energizing complex in the IM (5). Siderophore binding facilitates TBDT contact with the C-terminus of TonB through an allosteric mechanism, which exposes the TonB-binding site within TBDT, known as the TonB box, to the periplasm (6). Association of TBDTs with TonB establishes the main point of contact with ExbB/ExbD and the proton motive force provides the energy needed for the translocation of siderophores through the lumen of the OM barrel. The precise mechanism of this translocation process is not yet fully understood but is thought to involve a mechanical extraction or unfolding of the plug region from within the barrel lumen (7). Furthermore, a subclass of TBDTs possesses an N-terminal signalling domain, which regulates gene transcription of target operons, often participating in siderophore uptake and processing (8).
The activation of these signalling cascades is both ligand-and TonB-dependent, however the molecular details of this signalling process and its activation remain highly elusive (9,10). It is speculated that the N-terminal pocket of the signalling domain is involved in the interactions with the σ-factor regulator proteins. To date, the crystal structures of the intact TBDT FpvA with the fully-resolved signalling domain suggest that there is a high degree of flexibility at the N-terminal region of the transporter (11). However the putative site involved in contacting the regulator protein is tucked away beneath the barrel lumen.
Pseudomonas aeruginosa is a Gram-negative bacterium and an opportunistic human pathogen, which is a major cause of hospital-acquired infections in immunocompromised patients. In patients with cystic fibrosis P. aeruginosa lung infections are usually associated with increased mortality rates. P. aeruginosa is able to utilise a range of xenosiderophores, i.e siderophores produced by other bacteria or fungi in order to scavenge free iron. Such instances of so-called 'siderophore piracy' highlight bacterial adaptability and potential for colonising in an extremely broad range of environmental niches. One example of siderophore piracy is the uptake of ferrioxamine B, a hydroxamate siderophore produced by many Streptomyces species, by the specific OM transporter FoxA (12). FoxA belongs to the family of TBDTs (transducers) and is involved in ferrioxamine B transport as well as modulation of transcriptional cascades in the bacterial cell. Ferrioxamine B uptake comes at a relatively low energetic cost, when compared with the production of native siderophores such as pyoverdin and pyochelin (13). Indeed, when grown in the proximity of Streptomyces ambofacients, several Pseudomonas species do not produce their own siderophores and instead parasitize on the siderophores of their neighbour by expressing the ferrioxamine B transporter, FoxA (14).
Here, we determined several crystal structures of FoxA from P. aeruginosa, in the apo state as well as in complex with the siderophore ferrioxamine B and TonB. Using a hybrid approach combining X-ray crystallography with biophysical interaction studies, we provide novel insights into TonB-mediated siderophore uptake across the bacterial OM as well as TonBdependent signalling. Our results indicate that the transporter can exist in several different conformations, and that both substrate-and TonB-binding is required to form a translocationcompetent state of the FoxA transporter in a two-step TonB-binding mechanism necessary for transport function.

Ternary structure of FoxA bound to ferrioxamine B and TonB Ct
Active uptake of siderophores such as ferrioxamine B across the outer membrane (OM) relies on the establishment of a physical contact between the specific transporter and inner membrane (IM)-tethered TonB. In order understand the mechanism of complex formation between FoxA and TonB we determined the crystal structure of the ternary complex consisting of FoxA bound to ferrioxamine B and the C-terminal TonB domain (residues 251-340, referred to hereafter as TonB Ct ). Two complexes were present in the asymmetric unit, with crystal contacts generated through the exposed soluble regions of both proteins ( Figure   S1A,B). We could resolve almost an entire FoxA molecule including the signalling domain (residues 53-820, with 11 amino acids missing from the N-terminus after the signal peptide) as well as the full TonB Ct . We can identify two main modes of contact between FoxA and TonB Ct . Similar to the FhuA and BtuB-TonB complexes (15,16) the primary interaction site between FoxA and TonB Ct occurs through β-augmentation with parallel strands forming between residues 332-337 of TonB and residues 142-146 of FoxA ( Figure 1A and FoxA D352 , and between TonB R266 and FoxA D355 as well as a side-chain (FoxA E316 ) to the backbone carbonyl (TonB S300 ) via a hydrogen bond ( Figure S2). These contacts provide a secondary site of attachment for the TonB fragment and tether the C-terminal region of TonB to the barrel. This tethering locks the orientation of TonB against the barrel and the membrane plane. An overlay of the two complexes found in the asymmetric unit reveals that TonB Ct and the signalling domain are rotated by 9° with respect to the barrel and the membrane plane ( Figure S2). It is evident that TonB Ct and the signalling domain experience some degree of flexibility in the distal part of the complex, whereas the proximal part of the complex is most likely stabilized by the contacts at the secondary site. In the previous structural models (15,16), TonB sits in close proximity to the β-barrel, almost parallel to the putative lipid bilayer plane. In our crystal structure, the TonB fragment is located almost perpendicular to the βbarrel and the membrane plane highlighting the potentially dynamic nature of TBDR-TonB complexes, which has been suggested by recent EPR experiments (17) ( Figure S3). Overall, the structure of the ternary FoxA-ferrioxamineB-TonB Ct complex presented in this work reveals both the structure of the N-terminal signalling domain as well as a markedly different orientation of the TonB Ct relative to the TBDT compared to previously determined ternary structures of FhuA and BtuB (15,16).
To understand the conformational changes occurring in FoxA in response to ferrioxamine B and TonB Ct we also determined the crystal structure of FoxA in its apo state ( Figure 2A). In this structure a large solvent-exposed lumen faces the extracellular side of the membrane and is filled with solvent molecules. No electron density was present for the last two amino acids of the TonB box region and the N-terminal signalling domain of FoxA (residues 45-143), most likely due to the high flexibility of the linker connecting the plug domain and the signalling domain, as previously observed in the structures of FecA (18,19) and FpvA (11,20,21). Inspection of both apo as well as ligand-and TonB Ct -bound crystal structures of FoxA revealed that in the apo state the TonB box is predominantly occluded in the interior of the barrel domain. A comparison of the plug domain conformations in both of our FoxA structures indicates that the TonB box must be displaced by approximately 22 Å from the folded plug domain in order to allow for β-augmentation to occur with TonB ( Figure 2B).
Compared to the apo FoxA structure the ternary complex of FoxA-FoaB-TonB Ct also reveals substantial loop movements. Displacement of loops 7 and 8 by approx. 7 Å on the extracellular side of the membrane leads to the closure of the barrel lumen on both sides of the membrane, limiting the access to the barrel lumen ( Figure 2C). Mechanistically, this would prevent the dissociation of the siderophore during translocation and opening of the entry channel within the barrel.

Biophysical characterization of the interactions between FoxA and TonB Ct
Previous structural and biochemical investigations into the mechanisms of TBDT activation and substrate uptake have shown that siderophore binding usually leads to an unwinding of either an N-terminal helix or a stretch of polypeptide within the plug domain bearing the TonB box motif. This mechanism of polypeptide unwinding, initiated by concerted small motions throughout the plug, allows the C-terminus of TonB to make contact with the loaded transporter molecule. Moreover, insights into TBDT association with TonB paint a very complex, and often conflicting picture of substrate-dependent transporter activation and TonB binding. One model suggests a constitutively bound TonB-TBDT complex (22,23), whilst another proposes a cooperative mode of transporter-TonB interactions that is driven by initial substrate capture by the TBDT (24,25). Therefore, we sought to understand the nature of FoxA-TonB interactions using isothermal titration calorimetry (ITC) to characterize the thermodynamics of FoxA-TonB interactions.
For this purpose, we have reconstituted FoxA into MSP1D1 nanodiscs in order to minimize the detergent mismatch (26,27). Additionally, nanodiscs (ND) provide a lipidic scaffold and alleviate the deleterious effects detergents might have on the conformation of the transporter.
Titration of TonB Ct into apo FoxA-ND complexes resulted in strong, saturable exothermic heats indicative of protein association. ITC data were fitted to a single binding site model and yielded a K d of 107 ± 30 nM. The binding process is enthalpically driven with a large, negative ΔH of -11.01 ± 0.3 kcal mol -1 and entropically unfavourable with a negative TΔS value of -1.47 kcal mol -1 ( Figure 3A). No binding was observed when TonB Ct was titrated into empty nanodiscs. The observation of tight binding between TonB Ct and apo FoxA in nanodiscs is in contrast to similar experiments performed with FhuA and TonB Ct, for which no binding could be observed (27). We speculate that the presence of the N-terminal domain and the interacting region upstream of the TonB box in FoxA is responsible for the differences in the binding modes between these two transporters.
Next, we analyzed the interactions between TonB Ct and ferrioxamine B-bound FoxA in lipid nanodiscs. The data were also fitted to a single binding site model. Our ITC experiments showed that in the presence of ferrioxamine B the binding affinity is increased 17-fold yielding a K d value of 6 ± 4.5 nM. The thermodynamic parameters of the association reaction also differ such that the ΔH value decreases drastically to -19.16 ± 0.4 kcal mol -1 and the entropic contribution becomes even more unfavourable with TΔS of -7.86 kcal mol -1 ( Figure   3B).
We speculate that the decrease in the entropy of TonB Ct binding in the presence of ferrioxamine B most likely arises from large conformational restrictions of the flexible and highly mobile TonB and signalling domains as well as folding/desolvation events involved in association with TonB Ct . The difference in thermodynamics between the two binding reactions also suggests several distinct binding modes between TonB Ct and FoxA, which rely on the siderophore capture. The negative entropy is compensated for by a large decrease in the enthalpy of binding, which is driving the association reaction. Our ITC data suggest that in the presence of ferrioxamine B, FoxA is able to form a much tighter complex with TonB Ct .
Since β-augmentation between the TonB box of FoxA and TonB Ct is driven predominantly by hydrogen-bonded interactions, the stark decrease in the enthalpy of binding in our ITC could be explained by the formation of these additional contacts. The increased affinity and drastically reduced enthalpy of association is indicative of a much larger surface area participating in the complex formation process compared with the thermodynamics of apo FoxA-TonB Ct complexes. Altogether, the interaction studies presented here strongly support a cooperative mechanism of siderophore-dependent TonB capture by FoxA and that two distinct TonB-binding events can occur at the FoxA transporter.
To delineate the interactions between FoxA and TonB we generated FoxA variants with truncations in the signalling domain and analysed complex formation using analytical sizeexclusion chromatography. Full-length FoxA exhibits three distinct elution profiles corresponding to apo protein, FoxA-TonB Ct complex and the ternary complex FoxAferrioxamine B-TonB Ct confirming our ITC findings of two distinct TonB Ct -bound states, with and without the siderophore ( Figure S4A). Deletion of residues 64-130, which correspond to the majority of the signalling domain, yet retaining the sequence upstream of the TonB box observed in our crystal structure, had no effect on the constitutive and cooperative binding of TonB Ct to FoxA ( Figure S4B). However, the deletion of residues 64-143, which also include the upstream binding motif of the TonB box, abrogated the constitutive binding of TonB Ct to FoxA, whilst retaining the ability to cooperatively form the ternary complex between FoxA, ferrioxamine B and TonB Ct ( Figure S4C). In combination with our structure of the ternary complex, we propose a two-step binding mechanism: The constitutive mode of TonB Ct binding is mediated by the stretch of amino acids located upstream of the TonB box, and the interaction with ferrioxamine B would result in the allosteric release of the TonB box from within the barrel interior allowing the formation of a very tight complex between ferrioxamine B-bound FoxA and TonB, which is necessary for downstream translocation events leading to siderophore uptake.

Loop movements establish additional contacts with the bound siderophore
Siderophore capture by a specific TBDT is an integral part of establishing the necessary contacts with the TonB/ExbB/ExbD complex, which provides the energy for substrate One of the water molecules involved in the coordination is stabilised through hydrogen bonds by Gln 441 (Figure 4B).
The inward movement of loop 8, which is seen in the ternary complex with TonB Ct , places Lys 657 into close proximity of the bound ferrioxamine B with its ε-amino group protruding towards the hydroxamate groups coordinating Fe 3+ on the opposite side of His 374 ( Figure 4C).
These interactions presumably enforce the directionality of siderophore passage in the instance where the high-affinity site on the plug domain is modified during the partial unfolding and provides additional stabilising contacts to the siderophore prior to steps leading to its transport through the lumen of the barrel.
We measured the interaction between ferrioxamine B and purified FoxA using tryptophan fluorescence quenching experiments and ITC. Titration of ferrioxamine B to FoxA purified in nonyl glucopyranoside lead to concentration-dependent quenching of Trp fluorescence and allowed us to calculate the dissociation constant (K d ) of 100 nM, with a 1:1 stoichiometry ( Figure S5C). Our ITC experiments titrating ferrioxamine B into FoxA yielded a K d value of 180 ± 140 nM, which agrees well with our fluorescence experiments ( Figure S5D). Such strong association is also observed in other widely studied TBDT-siderophore complexes (28) and reflects the highly specific nature of these transporters at capturing extremely scarce and remain siderophore-bound until it is able to engage with TonB for transport of the siderophore inside the periplasm (Fig 5). Our structure also offers potential insight into signal transduction necessary for the regulation of transcription of the FoxA operon. The signalling domain, visible in the ternary complex, is exposed N-terminally to the periplasmic space, where is can at some stage of the transport engage with the sigma factor regulator protein FoxR situated in the IM. Such an orientation of the signalling domain differs starkly from previous structural studies involving FpvA (11,21).

Materials
The detergents used for purification were from Anatrace (Maumee, OH, USA) or Glycon (Luckenwalde, Germany). Desferrioxamine B was purchased from Sigma-Aldrich. All other chemicals were of analytical grade and obtained from Roth (Karlsruhe, Germany) or Sigma Aldrich / Merck (Darmstadt, Germany).

Protein expression and purification
Full-length FoxA gene from Pseudomonas aeruginosa stain PAO1 was cloned into a modified pET28a vector bearing a C-terminal TEV cleavage site prior a His 6 -tag. Protein overexpression was carried out in Escherichia coli Lemo21 cells (29) in 2xTY media supplemented with NPS (50 mM Na 2 HPO 4 , 50 mM KH 2 PO 4 , 25 mM NH 4 SO 4 ) and 5052 mix (0.05% glucose, 0.2% lactose, 0.5% glycerol) and 0.5 mM L-rhamnose. Cells were grown to an OD 600 of 1 at 37 °C, the temperature was reduced to 20 °C and 0.1 mM IPTG was added for further 16 hours. Cells were lysed in 30 mM Tris pH7.5, 200 mM NaCl, 10% glycerol using the high-pressure homogenizer (EmulsiFlex-C3, Avestin) and cell debris was removed by centrifugation at 22,000 g for 30 min. 1% Triton X-100 was added to the clarified cell lysate and incubated for 1 hr at 4 °C. The outer membrane fraction was isolated by a second centrifugation step at 100,000 g and the resuspended pellet was solubilised overnight in 1% octyl glucopyranoside (OG). Insoluble material was removed by another centrifugation step at 100,000 g for 20 min. Solubilised OM fraction was applied to the Ni-NTA resin followed by subsequent washes with buffer containing 25 mM imidazole and 0.4% nonyl glucopyranoside (NG) or 0.4% C8E4. Protein was eluted with 250 mM imidazole in buffer with 0.4% NG or C8E4 and tobacco etch virus protease (1:10 w/w) was added to the eluted fractions overnight.
After reverse Ni-NTA purification, the protein was concentrated and passed over a Superdex S200 10/300 size exclusion column.
TonB Ct (TonB1 from from Pseudomonas aeruginosa strain PAO1, residues 251-340) was overexpressed in Escherichia coli BL21 Gold cells, grown at 37 °C in LB medium. Cells were induced with 0.2 mM IPTG at OD 600 of 0.6-0.8 and the temperature was reduced to 20 °C. After 12 hours cells were spun down and lysed in 30 mM Tris pH 7.5, 500 mM NaCl, 10% glycerol using the EmulsiFlex-C3 (Avestin) homogeniser and cell lysate was centrifuged at 40,000 g to remove the cell debris. Cleared cell lysate was supplemented with 20 mM imidazole and loaded onto the 5-ml HisTrap Ni-NTA column. After several washes the protein was eluted with resuspension buffer supplemented with 300 mM imidazole. TEV was added to the pooled fractions containing TonB Ct and reverse-purification was performed the next day to remove the TEV and cleaved His 6 tag. TonB Ct was concentrated and stored at -80 ᵒC until further use.
MSP1D1 was expressed and purified as previously described (30,31) (32). Briefly, MSP1D1 in pET28a vector was transformed in E. coli strain BL21 (DE3) and grown in terrific broth (TB) media at 37 °C. At an OD 600 of 1.5 the protein expression was induced by adding 1 mM isopropyl ß-D-1-thiogalactopyranoside (IPTG) and cells were grown for 4 h at 37 °C. Cells were harvested by centrifugation at 3000 g, resuspended in lysis buffer (50 mM Tris pH 8.0, 500 mM NaCl) with 1% Triton X-100 and broken using sonication. The cleared lysate was loaded onto a HisTrap column and washed with ten column volumes each of lysis buffer containing 1% Triton X-100 and 50 mM cholate, respectively. MSP1D1 was eluted with buffer containing 500 mM imidazole, and fractions containing pure protein were pooled and incubated with TEV protease overnight. Subsequently, the protease and cleaved His-tag were separated by applying a second IMAC chromatography step and MSP1D1 without His-tag was concentrated up to 400 µM and stored at -80 °C until further use.

Analytical size-exclusion chromatography (SEC)
Truncation mutants of FoxA (Δ64-130 and Δ64-143 residues) were generated using the standard QuikChange PCR mutagenesis protocols. Analytical SEC analyses of complex formation between FoxA, truncated FoxA and TonB Ct were performed using a Superdex S200

Structure determination
All X-ray diffraction data was collected at 100 K. Data were collected at the PETRA III/EMBL P14, ESRF ID30B and BESSY 14.1 beamlines. All datasets were processed with XDS (33), and merged with AIMLESS (34,35). All final data were merged from two individual datasets. Unit cell parameters and space groups are given in Table I. We used the FhuA model from Escherichia coli (PDB:1BY3) as a molecular replacement candidate (40-45% sequence identity to FoxA). After the successful placement of the model using Phaser (36), the FoxA model was completed using a combination of phenix.autobuild (37) and manual building in Coot (38). Apo FoxA was refined using phenix.refine and REFMAC5 (39) (40). For the FoxA-FoaB complex, refinement was performed initially using phenix.refine, followed by TLS and jelly-body refinement in REFMAC5 (40). For modelling the FoxA-FoaB-TonB Ct complex, apo FoxA was used as a search model in Phaser. Once the MR solution was identified, clear density corresponding to the TonB Ct and the N-terminal signalling domain became evident and they were manually built into the electron density using Coot. Refinement was carried out initially using phenix.refine at the early stages of model building. Once all the backbone poly-Ala stretches were built, Buster-TNT (41) was used for all the subsequent refinement procedures. The final models correspond to residues 44-820 of FoxA with 119-124 and 138-155 being disordered. The TonB Ct model comprises residues 251-340. All data collection and refinement statistics are summarized in Table I.

Isothermal titration calorimetry (ITC)
FoxA was incorporated into MSP1D1 nanodiscs. Briefly, FoxA was mixed with purified MSP1D1 and POPC lipids in 1:2:70 ratio and biobeads were added to initiate the nanodisc assembly. After approximately 4-5 hours, the mixture was concentrated and purified by gel size exclusion chromatography using a Superdex S200 10/300 column. All proteins were extensively dialysed against 20 mM HEPES, 150 mM NaCl, pH 7.5 overnight. 15 µM FoxA or FoxA-FoaB incorporated into MSP1D1 nanodiscs were loaded into the ITC cell, 150 µM TonB Ct was placed in the syringe. All binding reactions were measured at 26 °C. TonB Ct was also titrated against empty nanodiscs as a control. Heat of dilution was obtained by titrating TonB Ct into the dialysis buffer and subtracted from all subsequent measurements. Heats of binding for all the reactions were integrated using Microcal Origin software and all the data were fitted to a single-site binding model.

Tryptophan fluorescence quenching experiments
All fluorescence measurements were performed using a Cary Eclipse fluorescence spectrometer. FoxA purified in nonyl glucopyranoside (NG) was diluted to 100 nM in a 3 ml quartz cuvette. Trp fluorescence was excited at 280 nm and emission spectra were recorded from 310-420 nm. Ferrioxamine B, diluted in the same buffer as FoxA, was titrated into the cuvette until saturation in fluorescence quenching was reached. Control experiments with buffer were performed to account for dilution effects on Trp fluorescence. Buffer conditions were 20 mM Tris pH 7.5, 200 mM NaCl, 0.4% NG. Curves were plotted and analysed in GraphPad Prism 7. The binding curve was fitted to a single-site binding model.

Supplementary Movie
Visualizing the conformational changes occurring in FoxA in response to ferrioxamine B and TonB Ct binding. We observe the closure of extracellular loops 7 and 8 as a prerequisite for translocation of ferrioxamine B. At the periplasmic side, TonB box is expelled from the plug domain in order to make contacts with the TonB molecule.  FoxA-TonB Ct with foaB 420 ± 300 nM -9.37 ± 0.75 -0.77