Solution Structure of Escherichia coli FeoA and Its Potential Role in Bacterial Ferrous Iron Transport

Abstract

Iron is an indispensable nutrient for most organisms. Ferric iron (Fe³⁺) predominates under aerobic conditions, while during oxygen limitation ferrous (Fe²⁺) iron is usually present. The Feo system is a bacterial ferrous iron transport system first discovered in Escherichia coli K-12. It consists of three genes, feoA, feoB, and feoC (yhgG). FeoB is thought to be the main transmembrane transporter while FeoC is considered to be a transcriptional regulator. Using multidimensional nuclear magnetic resonance (NMR) spectroscopy, we have determined the solution structure of E. coli FeoA. The structure of FeoA reveals a Src-homology 3 (SH3)-like fold. The structure is composed of a β-barrel with two α-helices where one helix is positioned over the barrel. In comparison to the standard eukaryotic SH3 fold, FeoA has two additional α-helices. FeoA was further characterized by heteronuclear NMR dynamics measurements, which suggest that it is a monomeric, stable globular protein. Model-free analysis of the NMR relaxation results indicates that a slow conformational dynamic process is occurring in β-strand 4 that may be important for function. ³¹P NMR-based GTPase activity measurements with the N-terminal domain of FeoB (NFeoB) indicate a higher GTP hydrolysis rate in the presence of potassium than with sodium. Further enzymatic assays with NFeoB suggest that FeoA may not act as a GTPase-activating protein as previously proposed. These findings, together with bioinformatics and structural analyses, suggest that FeoA may have a different role, possibly interacting with the cytoplasmic domain of the highly conserved core portion of the FeoB transmembrane region.

INTRODUCTION

Iron is an essential nutrient for most organisms (1). Its importance arises from its role as a protein cofactor in processes such as DNA precursor synthesis, respiration, and oxygen transport or as an electron carrier (2–4). Two forms of iron exist naturally, ferric (Fe³⁺) and ferrous (Fe²⁺), where ferric iron is the form that is dominant under oxidizing conditions. In Gram-negative bacteria, Fe³⁺ is transported into the periplasm through dedicated outer membrane receptors that are part of the TonB-ExbB-ExbD-dependent transport systems (3, 5–7). Anaerobiosis and acidification lead to the transition of ferric to ferrous iron (3, 8). Different bacterial ferrous iron transport pathways exist: the metal-ABC transporters YfeABCD and FutABC, MntH, ZupT, and EfeUOB as well as the ferrous iron transport system, Feo (9–18). YfeABCD is part of the SitABCD family of ATP-binding cassette (ABC) transporters that regulate manganese and iron levels within the cell (15, 17). FutABC is also an ABC transporter that selectively transports ferrous iron (13, 14). MntH and ZupT are part of the NRAMP (natural resistance-associated macrophage proteins) and the Zip transporter family, respectively, and both have broad-spectrum specificity for various divalent metal ions (10, 12, 16). The EfeUOB transport system is specific for ferrous iron transport; however, it is present only in specific pathogenic species (9, 11, 12). The Feo system is the only transport pathway that is widely distributed and that is dedicated to the transport of ferrous iron (19, 20).

The Feo system was first discovered in Escherichia coli K-12 (E. coli), and it consists of the feoA, feoB, and feoC (yhgG) genes (19–21) (Fig. 1A). Residing upstream of feoA, feoB, and feoC are the fnr and fur regulatory elements (Fig. 1A) (5, 20). Fnr is an anaerobically induced transcriptional activator (22), which acts as an oxygen sensor for the feo operon and, in the absence of oxygen, activates transcription of the feo genes. Fur is an iron-dependent regulator of iron transport genes. In the presence of iron, the Fe²⁺-Fur complex binds to the fur box to inhibit transcription of the feo genes (5, 23). Iron deprivation results in the dissociation of Fur from the fur box to allow for transcription of iron transport genes. Fur and Fnr play a crucial role in maintaining cellular iron homeostasis and preventing the formation of damaging free radicals. The presence of another regulatory element, the rstA box, has also been found in the Salmonella enterica feo operon, which is absent in its E. coli counterpart, to precede the fur binding sequence (24). RstA was shown to regulate the expression of the feoB gene in S. enterica (24). The Feo system is important for the virulence of various pathogenic bacterial strains such as Helicobacter pylori, Legionella pneumophila, and Campylobacter jejuni (25–27). As such, it is important to gain a better insight into the mode of action of the Feo proteins.

Fig 1.

(A) The Feo operon in E. coli K-12. (B) Different feo operon organizations in all sequenced bacterial genomes to date (October 2011). Frequency indicates prevalence of each organization in the NCBI Gene database. NfeoB represents the gene encoding the soluble N-terminal cytoplasmic G protein and GDI domains of FeoB. CfeoB represents the gene encoding the C-terminal transmembrane region of FeoB.

Within the Feo system, the FeoB protein is thought to be the main ferrous iron transporter in the cytoplasmic membrane. E. coli FeoB is a 773-residue transmembrane protein comprised of two main regions: an N-terminal cytoplasmic domain and a C-terminal polytopic transmembrane domain. Within the N-terminal domain (NFeoB) reside the G-protein and the guanine nucleotide dissociation inhibitor (GDI) domains (28). The G domain of FeoB is essential for ferrous iron uptake (29). It is thought that the G domain provides energy for the transport process or that it senses the energy state of the cell and relays this information to the transmembrane domain to regulate transport. The GDI domain is an internal regulator that specifically stabilizes the binding of GDP to the G domain (28). The polytopic transmembrane region of FeoB is proposed to act as a permease for the diffusion of Fe²⁺ into the cell (21, 29).

E. coli FeoC is a small 78-residue protein with a winged-helix fold (5, 30). This fold belongs to the helix-turn-helix family that is usually involved in DNA binding (5, 31). It has been suggested that FeoC may act as a transcriptional regulator (5). In comparison to FeoA and FeoB, FeoC is not well conserved between species, and it is thought to be present only in gammaproteobacteria (5) (Fig. 1B).

The role of FeoA is not well understood within the Feo system. E. coli FeoA is a 75-residue cytoplasmic protein with an unknown function. It was proposed that FeoA may interact with FeoB to take part in ferrous iron transport (32). Surveying the available sequenced bacterial genomes to date (NCBI database, accessed October 2011), we find that ∼45% possess one or more Feo systems (Fig. 1B). In some Feo systems, multiple FeoA proteins exist that are covalently attached to each other (Fig. 1B). Sequence analysis of the double and triple FeoA proteins reveals that they have low amino acid sequence homology; typically, there is only ∼30% identity or less between the individual domains. Further analysis of the operons reveals that FeoA occurs together with FeoB 89% of the time (Fig. 1B). In prokaryotes, interacting proteins are often located adjacent to one another within one operon (33). Consequently, the fact that the FeoA and FeoB genes are being expressed together suggests intertwined functional roles. Here, we present the solution structure of E. coli FeoA along with a dynamics analysis of the behavior of the protein in aqueous solution. Moreover, we also performed ³¹P nuclear magnetic resonance (NMR) studies to examine the role of FeoA within ferrous iron transport. Our results show that FeoA may not act as a GTPase-activating protein (GAP) as was earlier suggested (5); instead, we propose that it may function by interacting with the highly conserved core region in the transmembrane domain of FeoB.

MATERIALS AND METHODS

Cloning of E. coli FeoA and NFeoB.

FeoA and NFeoB (residues 1 to 276) were amplified from the E. coli K-12 genome through PCR with oligonucleotides engineered to contain 5′ NdeI (CATATG) and 3′ XhoI (CTCGAG) (New England BioLabs) restriction sites. Amplified FeoA and NFeoB were cloned into pET30a and pSUMO expression vectors, respectively. The pET30a vector (Novagen) attaches a hexahistidine tag to the C terminus of the expressed protein while the pSUMO (small ubiquitin-like modifier) system (LifeSensors) attaches a hexahistidine SUMO tag on the N terminus (34). The original pSUMO vector was modified to contain an NdeI restriction site in the multiple-cloning site. Constructs were verified through nucleotide sequencing at the University of Calgary Genetic Analysis laboratory. Vectors with constructs are maintained in E. coli DH5α cells. Plasmids were transformed to E. coli BL21(DE3) cells for protein expression.

Site-directed mutagenesis of FeoA.

The C-terminal cysteine residue of wild-type FeoA was mutated to a serine (C75S) to prevent any undesired spontaneous disulfide formation and covalent dimerization of the protein during the lengthy NMR experiments. The cytoplasm of E. coli is a reducing environment; thus, Cys75 would be in its reduced form in vivo. The C75S mutation ensures that FeoA is always present in its intracellular form. Moreover, Cys75 is the terminal residue, and it does not appear to be conserved among FeoA proteins from different species; thus, C75S can be considered to be equivalent to FeoA. Preliminary structural data on wild-type FeoA were identical to those for C75S; therefore, we will refer to the mutant protein as FeoA henceforth. The pET30a vector with wild-type FeoA was used as a template for site-directed mutagenesis. Mutagenesis was performed using a protocol outlined in the QuikChange site-directed mutagenesis system (Stratagene).

Overexpression and purification of FeoA and NFeoB.

For overexpression of FeoA, E. coli cells were grown in 1 liter of Luria Bertani (LB) liquid medium in the presence of 30 μg/ml of kanamycin. Cells were expressed to an optical density at 600 nm (OD₆₀₀) of ∼0.4 at 37°C and induced with 0.8 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) at room temperature for 5 h. Cells were harvested and resuspended in 20 mM Tris, 0.5 M NaCl, and 40 mM imidazole (pH 8) and lysed by a French pressure cell. The cell lysate was centrifuged at 18,500 × g for 45 min at 8°C. All histidine-tagged proteins were purified with a chelating Sepharose column (GE Healthcare) through nickel affinity chromatography. Bound proteins were washed with 20 mM Tris, 0.5 M NaCl, and 100 mM imidazole and eluted off the column with 20 mM Tris, 0.15 M NaCl, and 300 mM imidazole (pH 8). Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) was used to visualize protein homogeneity. The FeoA protein concentration was determined by measuring the A₂₈₀ using an extinction coefficient calculated through the ExPaSy ProtParam program. The C-terminal hexahistidine tag of the FeoA construct was not removed for these studies as it enhanced the solubility for the in vitro studies.

For the production of ¹³C, ¹⁵N isotope-enriched FeoA, cells were grown under the conditions described above with the exception that M9 minimal medium supplemented with ¹⁵NH₄Cl (1g/liter) and [¹³C]glucose (3g/liter) (Sigma-Aldrich) were used instead of the LB medium. Isotope-labeled protein was purified as per the protocol described above.

NFeoB was expressed in 1 liter of LB medium with 30 μg/ml of kanamycin and induced at an OD₆₀₀ of ∼0.6 with 0.25 mM IPTG for 4 h. Cells were harvested and resuspended in 20 mM Tris, 0.5 M NaCl, and 20 mM imidazole, pH 8, and lysed by a French pressure cell. The lysed cells were centrifuged at 18,500 × g for 45 min at 8°C and applied to a nickel-chelating Sepharose column. Bound proteins were washed with resuspension buffer, and SUMO-NFeoB fusion proteins were eluted with 20 mM Tris, 0.15 M NaCl, and 300 mM imidazole at pH 8. The SUMO-NFeoB fusion protein was cleaved with SUMO protease (100 μg of fusion protein/unit of protease) (LifeSensors). Undigested fusion proteins along with the cut SUMO proteins were removed from NFeoB through reapplication to the nickel-chelating Sepharose column, and NFeoB was eluted using resuspension buffer. SDS-PAGE was used to visualize protein homogeneity; proteins were >95% pure. The protein concentration of NFeoB was also determined through A₂₈₀ measurements.

NMR measurements of FeoA.

NMR samples contained approximately 0.5 mM uniformly ¹⁵N, ¹³C-labeled FeoA, 1 mM 2,2-dimethyl-2-silapentane-5-sulfonate (DSS), and 10% D₂O buffered in 20 mM HEPES, 100 mM NaCl at pH 7. All NMR experiments were performed at 298 K on a Bruker Avance 500 MHz NMR spectrometer equipped with a triple-resonance inverse Cryoprobe with a single axis z-gradient. The sequential backbone resonance assignment of FeoA was obtained through HNCACB, CBCA(CO)NH, HNCO, and HN(CA)CO experiments. Aliphatic side chain assignments were obtained through three-dimensional experiments including C(CCO)NH-total correlation spectroscopy (TOCSY), H(CCO)NH-TOCSY, HBHA(CBCACO)NH, CCH-TOCSY, and HCCH-TOCSY. Aromatic side chain assignments were determined through two-dimensional (HB)CB(CGCD)HD and (HB)CB(CGCDCE)HE (35) and double-quantum-filtered correlation spectroscopy (DQF-COSY) experiments. All nuclear Overhauser effect spectroscopy (NOESY) experiments including three-dimensional ¹³C-edited NOESY-heteronuclear single-quantum coherence (HSQC), three-dimensional ¹⁵N-edited NOESY, and two-dimensional NOESY were acquired with a mixing time of 120 ms.

Backbone dynamics studies of ¹⁵N-labeled FeoA were acquired at 50.68 MHz for the ¹⁵N frequency. Longitudinal relaxation (T₁) experiments were performed using sensitivity-enhanced inversion-recovery pulse sequences (36). Each spectrum was acquired with 1,024 points in the ¹H dimension and 64 points in the ¹⁵N dimension. Sweep widths of 8,012 Hz and 912 Hz were employed for the ¹H and ¹⁵N dimensions, respectively, with 24 scans per t₁ point. Delay times of 14 (twice), 98, 350 (twice), 490, 700, 896, and 1,190 ms were used. ¹⁵N spin-spin relaxation time (T₂) data were acquired using sensitivity-enhanced Carr-Purcell-Meiboom-Gill (CPMG) pulse sequences with pulsed-field gradients (36). All T₂ spectra were acquired with the same parameters as above except with 32 scans per t₁ point. Delay times of 6.55 (twice), 13.09, 26.17 (twice), 39.25, 52.33, 65.41, 78.49, 91.57, 117.737 ms were used. {¹H}-¹⁵N heteronuclear NOE data were acquired with a 5-s presaturation period with the reference spectrum acquired with a 5-s delay time. NMR spectra were processed with NMRPipe/NMRDraw (37) and analyzed with NMRView software (38).

Structure calculation.

The initial structure of FeoA was calculated with CYANA (version 2.1) (39) using distance restraints generated from the automated NOE assignment protocol implemented in CYANA and dihedral angle restraints predicted by TALOS+ (40). Further structural refinement with the addition of hydrogen bond restraints based on secondary structure from the chemical shift index was performed with XPLOR-NIH (version 2.26) (41). The 30 lowest-energy structures from a total of 200 calculated were used for analysis. Structures were validated with the Protein Structure Validation Software (PSVS) suite (42). All molecular graphics used in the manuscript were generated with MOLMOL (43) or PyMOL (version 1.3r1; Schrodinger, LLC) software.

Relaxation data analysis.

T₁ and T₂ data were fitted with the program CurveFIT (A. G. Palmer, Columbia University). The uncertainties for {¹H}-¹⁵N NOE values were evaluated from spectra with and without proton saturation. Residues that show NOE values of <0.65 were eliminated from analysis due to possible fast internal motions (44, 45). Residues involved in chemical exchange processes affect T₂ relaxation time and, thus, were removed as described by Tjandra et al. (46). Residue-specific correlation times were determined from the ratio of the spin-spin relaxation rate constant to the spin-lattice relaxation rate constant (R₂/R₁) using the program R2R1_τm (A. G. Palmer, Columbia University), and a single isotropic total correlation time (τ_c) of the protein was calculated. Estimates of axial and fully anisotropic diffusion tensors were obtained using the program QUADRIC_DIFFUSION (A. G. Palmer, Columbia University). Relaxation data were fit to five models according to the model-free formalism (47, 48), where each N-H vector is assigned a specific model from which relevant motional parameters (the squared order parameter, S², and the chemical exchange rate, R_ex) were extracted. Model selection was performed through the FAST-ModelFree program (49) that uses ModelFree, version 4.20 (50).

³¹P NMR.

³¹P NMR spectroscopy was used to monitor the GTPase activity of NFeoB. The ³¹P NMR samples contained 60 μM NFeoB in 20 mM HEPES, 100 mM KCl or NaCl, 3 mM GTP, 4 mM MgSO₄, and 10% D₂O at pH 7. One-dimensional NMR experiments were acquired every hour for 24 h to monitor GTP hydrolysis. All experiments were acquired on a Bruker Avance 400 MHz spectrometer equipped with a 10-mm Broadband probe at 310 K in a 10-mm NMR tube. GTP self-hydrolysis was monitored by performing the same experiment as described above except that no NFeoB was added.

To examine the role of FeoA, it was added to an identical sample to observe its effect on the GTPase activity of NFeoB. FeoA and NFeoB were mixed together in 20 mM HEPES and 100 mM KCl at pH 7 and concentrated down to 60 μM each. GTP (3 mM) and MgSO₄ (4 mM) and D₂O (10%) were added to the concentrated solution, and hydrolysis activity was immediately recorded. The control sample contained the same reagents with the exception of FeoA. These experiments were acquired at 298 K.

Protein structure accession numbers.

The atomic coordinates of the 30 lowest-energy structures have been deposited in the RCSB Protein Data Bank (PDB) under accession number 2lx9. The NMR chemical shifts were deposited in the BioMagResBank (BMRB) under accession number 18668.

RESULTS

Protein purification.

The His-tagged FeoA construct could be expressed with a yield of ∼10 mg of purified protein per liter of cell culture in LB medium; the expression levels in minimal medium were around ∼7 mg of protein per liter of culture. FeoA was usually expressed in the cytoplasm of E. coli BL21(DE3) and was not detectable in inclusion bodies. NFeoB, with a SUMO tag, was expressed in LB medium with a yield of ∼50 mg of tag-free purified NFeoB protein per liter of culture. Both proteins could be purified to homogeneity as determined by SDS-PAGE (see Fig. S1 in the supplemental material). Molecular weights were confirmed with mass spectrometry (data not shown).

Solution structure of FeoA.

The backbone amide resonances of FeoA could be assigned unambiguously with the exception of a few residues that experienced severe broadening through chemical exchange (e.g., Gln2 and Ser75). The NMR solution structure of FeoA was calculated based on 1,197 NOE-derived distance restraints (Table 1). Superposition of the 30 lowest-energy structures reveals a well-defined structure that converges to a backbone root mean square deviation (RMSD) of 0.40 ± 0.09 Å for the folded region (Fig. 2A). FeoA possesses two antiparallel β-sheets with β-strands 1 and 5 being orthogonal to β-strand 2, 3, and 4 (Fig. 2B). There are two α-helices in FeoA; the first α-helix sequentially follows the first β-strand and is situated above the β-barrel structure resembling a lid (Fig. 2B; see also Fig. S2A in the supplemental material). The second α-helix is slightly shorter than the first helix, consisting of only two turns, and it is located between β-strands 4 and 5. Ramachandran analysis shows that 88.2% of the residues are in the favored regions, while 11.8% of the residues are in the additionally allowed regions (Table 1). No residues are present in the disallowed regions of the Ramachandran plot.

Table 1.

Experimental restraints and structural statistics for the 30 lowest-energy structures

Parameter	Value for the parameter
No. of NOE cross-peaks assigneda	1,924
Experimental restraints
No. of distance restraints from NOEsb
Total	1,197
Intraresidue	297
Sequential	351
Medium-range (2–4)	192
Long-range (5+)	357
No. of dihedral angle restraintsc	130
No. of hydrogen bond distance restraints	40
Avg RMSD from experimental restraints
Distance restraint violation (Å)	0.039 ± 0.002
Dihedral angle restraint violation (°)	0.574 ± 0.090
Avg RMSD from ideal stereochemistry
Bond length (Å)	0.003 ± 0.000
Bond angle (°)	0.471 ± 0.010
Improper bond (°)	0.318 ± 0.010
PROCHECK Ramachandran analysis (%)d
Residues in favored regions	88.2
Residues in additional allowed regions	11.8
Residues in generously allowed regions	0.0
Residues in disallowed regions	0.0
Coordinate precision of folded regions (Å)e
Backbone	0.40 ± 0.09
All heavy atoms	1.02 ± 0.11

^aDetermined from CYANA, version 2.1 (39).

^bNumbers of restraints per residue were calculated using FeoA C75S residues 5 to 70 from Xplor-NIH, version 2.26 (41).

^cPredicted from TALOS+ (40).

^dValidated through the Protein Structure Validation Suite (42), excluding glycine and proline residues.

^eResidues 5 to 70.

Fig 2.

(A) The 30 lowest-energy structures of FeoA C75S superimposed for the well-folded region (residues 5 to 70). (B) Ribbon representation of the lowest-energy structure. β-Sheet and α-helices are represented in navy and yellow, respectively. (C) Comparison of solution NMR structures of E. coli FeoA (navy) (PDB ID 2lx9) and K. pneumonia FeoA (green) (PDB ID 2gcx). (D) Superposition of NMR structure of E. coli FeoA (navy) (PDB ID 2lx9) and crystal structure of S. maltophilia FeoA (yellow) (PDB ID 3mhx).

Figure 2C shows an overlay of the PDB-deposited solution NMR structure of Klebsiella pneumonia FeoA (PDB identification number [ID] 2gcx) and the E. coli protein. The sequence identity between these two proteins is 90%. The residues that differ are all conservative substitutions with the exception of Asn-38 in E. coli, which is a His in K. pneumonia. This residue is solvent exposed and does not alter the structure significantly. All secondary structural elements are the same for both proteins, with the overall structures agreeing to a backbone RMSD of 2.22 Å.

The crystal structure of Stenotrophomonas maltophilia FeoA complexed with zinc was recently reported (PDB ID 3mhx) (32). The amino acid sequence of S. maltophilia FeoA is significantly different from the sequences of E. coli and K. pneumonia FeoA (data not shown). Superposition of this structure on that of E. coli FeoA shows the typical β-barrel structure being conserved with a backbone RMSD of 2.61 Å (Fig. 2D). In contrast to the two helices and five β-strands of E. coli FeoA, S. maltophilia FeoA contains three helices and five β-strands (Fig. 2D; see also Fig. S2B in the supplemental material). Despite these differences, the core β-barrel fold is conserved.

Dynamic properties of FeoA.

Backbone dynamics experiments including {¹H}-¹⁵N heteronuclear NOE, T₁, and T₂ measurements were carried out at 500 MHz with 68 residues selected for final analysis (Fig. 3). Of note, analysis of the dynamics data excluded the hexahistidine tag region and all five proline residues which do not have an amide proton. NOE and R₂ results reveal a relatively ordered conformation for FeoA, with the exception of the loop region between β-strands 3 and 4, which undergoes chemical exchange. The S² order parameter (average S² [S²_ave] of ∼0.87) derived from model-free analysis agrees well with the R₁/R₂/NOE data, suggesting that FeoA is a rigid protein (Fig. 3). Lower S² values near the C terminus of FeoA indicate a more flexible region (Fig. 3). Four residues with conformational exchange occurring in the milli- to microsecond timescale are observed in the region of β-strands 3 and 4 (Fig. 3). These residues also have a higher effective R₂ [R_2eff] relaxation rate. The higher R_2eff of these residues are a result of the higher R_ex contribution. Of specific interest is Val56, which has the highest R_ex value of all the residues. The total correlation time of FeoA was calculated to be 6.2 ns. Quadratic diffusion (D) analysis reveals an isotropic globular protein with a D_parallel/D_{perpendicular} of 0.87 ± 0.03. All data are in accordance with secondary structural elements as illustrated in Fig. 3. Moreover, they indicate that E. coli FeoA behaves as a monomeric, almost globular, protein in aqueous solution.

Fig 3.

¹⁵N relaxation data and FAST-ModelFree analysis for FeoA. R₁, R₂, NOE values, S² order parameters, and chemical exchange rates (R_ex) are plotted as a function of the residue number. S² order parameters reflect motions on the pico- to nanosecond timescale, and R_ex values reflect motions on the micro- to millisecond timescale. Positions of the secondary structural elements are shown.

³¹P NMR of NFeoB.

To monitor the GTPase activity of NFeoB, GTP hydrolysis was followed over time by ³¹P NMR. This assay simultaneously monitors the levels of GTP, GDP, and P_i in the sample. Each phosphorus atom in the GTP/GDP moieties is identified through its characteristic chemical shift. The first experiment after 1 h shows nonhydrolyzed GTP with three phosphate peaks: γ-, α-, and β-phosphate (Fig. 4A). As time progresses, the P_i peak becomes increasingly visible around 2 ppm. After 21 h, the majority of GTP has been hydrolyzed, with small residual amounts observed through the small β-phosphate GTP peak (Fig. 4A). A corresponding control experiment showed no GTP autohydrolysis under the same conditions (data not shown); thus, the hydrolysis observed is due to the addition of NFeoB.

Fig 4.

GTPase activity assay of NFeoB. ³¹P NMR spectra of NFeoB in the presence of NaCl (A) and KCl (B).

Previously, Streptococcus thermophilus NFeoB was shown to accelerate GTP hydrolysis by 16-fold in the presence of potassium (51). Our initial ³¹P NMR experiments were performed in NaCl. To investigate whether E. coli NFeoB acts through a similar mechanism as S. thermophilus NFeoB, the same ³¹P NMR assay was performed in the presence of KCl. Figure 4B illustrates the complete hydrolysis of GTP after 4 h. The only visible peaks after 4 h represent the P_i and the β- and α-GDP peaks. A comparison of the NaCl and KCl spectra shows that the hydrolysis activity is five times faster in KCl than in NaCl (Fig. 4).

To determine whether FeoA can act as a GAP of NFeoB, the same assay was performed in the presence of FeoA. The addition of an equimolar amount of FeoA had no effect on NFeoB's hydrolytic activity; the amount of time required to hydrolyze GTP was the same as without FeoA (data not shown).

DISCUSSION

Bacterial ferrous iron transport systems are currently much less understood in comparison with ferric iron transport pathways. For example, it is not yet known whether ferrous iron enters the cell through active transport or through a passive diffusion mechanism. In vivo models have shown that FeoA and FeoB are closely linked and that deletion of FeoA reduces ferrous iron uptake while deletion of FeoB completely abolishes ferrous iron transport (20). FeoA has previously been classified as an Src-homology 3 (SH3)-like domain based on its low amino acid sequence similarity to the C-terminal domain of DtxR (5), while FeoB is thought to provide energy and regulate transport (5, 20, 28, 29, 51). Our NMR studies of E. coli FeoA contribute to the understanding of bacterial ferrous iron transport as we have been able to investigate both the solution structure and the dynamic properties of the protein and the role that it may play in ferrous iron transport.

The structure of E. coli FeoA is composed of a β-barrel and two α-helices. The FeoA fold resembles that of a eukaryotic SH3 domain. SH3 domains are a group of interacting modules that are prevalent in eukaryotic organisms (52). They have been noted to interact with small G proteins and may bind GAPs or guanine exchange factors (GEFs) (53). In order to highlight the similarity, the FeoA structure was overlaid with the prototypical SH3 domain of Abl tyrosine kinase (ATK) (PDB ID 1ju5) (Fig. 5A). Superposition of FeoA and ATK reveals an RMSD of 2.64 Å, with the core β-barrel structurally conserved with several distinct differences (54). Specifically, the first α-helix that resides above the barrel and the second α-helix that completes the barrel are absent in the prototypical ATK SH3 domain (Fig. 2B and 5A). Normally, SH3 domains possess three conserved loops that are important for protein-protein interactions: the RT-Src, N-Src, and distal loop (Fig. 5A E) (55). The RT-Src loop of ATK is replaced by an α-helix in FeoA (Fig. 5A and E). While the N-Src loop of ATK is present in FeoA, its position and length are quite different (Fig. 5A and E). The distal loop is present and appears to be in the same orientation in both FeoA and ATK (Fig. 5A and E). Hydrophobic pockets filled with aromatic residues that are typical of SH3 domains (52) are also absent in FeoA. The lack of standard SH3 domain features is consistent with the notion that SH3 domains are less prevalent in prokaryotes than eukaryotes, which suggests that perhaps the SH3-like fold of FeoA could fulfill a distinct role (52, 55, 56).

Fig 5.

Structural comparison of FeoA with prototypical SH3 domain and Dali-determined structural homolog, IdeR. (A) Prototypical SH3 domain Abl tyrosine kinase (PDB ID 1ju5 chain C) superposed on E. coli FeoA (PDB ID 2lx9) in green and black, respectively. The RT-Src, N-Src, and distal loops of SH3 domains are indicated. Val56 of FeoA is highlighted in pink. (B) Structural superposition of E. coli FeoA (PDB ID 2lx9) (black) and IdeR (PDB ID 1u8r chain C) (gray). Ancillary site and primary site of IdeR are shown in blue and pink, respectively. Hydrogen bond contacts are illustrated in yellow. (C) Ancillary site (blue) and primary site (pink) of IdeR (PDB ID 1u8r chain C). E172 and Q175 of the SH3-like domain donate two metal ligands for the ancillary site. Hydrogen bond contacts are highlighted in yellow. (D) Potential metal ligands of E. coli FeoA. The SH3-like domain of IdeR is replaced by FeoA in this structure. Highlighted in green are residues of FeoA close to the binding sites of IdeR. Metal binding sites and contacts of IdeR are colored as in panel C. FeoA is depicted in black for all structures. (E) Three-dimensional structural alignment of E. coli FeoA with Abl tyrosine kinase and the C-terminal SH3-like domains of the three Dali structural homologs. Positions of β-sheets and α-helices are shown in dashed boxes. The location of the RT-Src, N-Src, and distal loops with respect to the secondary structure of Abl tyrosine kinase is indicated. Highlighted in yellow is one of the metal binding ligands in the SH3-like domain of the homologs.

Currently, seven structures for FeoA from different bacterial species have been deposited in the PDB: five NMR solution and two crystal structures. In comparison with E. coli FeoA, the deposited but as yet unpublished NMR structure of K. pneumonia FeoA (90% sequence identity) shows that both structures agree well with each other, with an RMSD of 2.2 Å (Fig. 2C). Of the seven structures, only the S. maltophilia FeoA crystal structure has been published with a suggested function as a bacterial ferrous iron transport-activating factor (32). Compared to E. coli FeoA, S. maltophilia FeoA possesses an additional helix (Fig. 2D). Despite their different sequences, the β-barrel structure is conserved, and thus this feature may be important for the function of this protein.

In eukaryotes, some SH3 domains are known to bind to G proteins through proline-rich sequences (PXXP), where they can regulate the enzymatic activity or interactions further down the signaling cascade (57–59). G proteins, depending on their nucleotide-bound state, can interact with different effectors to propagate or terminate signals. Intriguingly, the FeoB protein also possesses a cytoplasmic G protein in its N-terminal domain, which has led to the suggestion that the cytoplasmic FeoA protein can interact with the cytoplasmic G domain of FeoB (5). Furthermore, a bioinformatics analysis reveals that FeoA and FeoB occur together in one operon for 89% of all the Feo systems (Fig. 1B). In fact, in a limited number of cases FeoA also appears to be directly fused to FeoB, indicating that these proteins act in a spatially proximal manner (Fig. 1B). As suggested by Lee et al., the presence of two genes within the same operon is a strong indicator of intertwined biological roles (33). We first examined the potential for FeoA-FeoB interactions on the assumption that FeoA may act as a GAP for FeoB. Using ³¹P NMR, we established that the E. coli NFeoB (residues 1 to 276) construct possesses GTPase activity (Fig. 4); the observed GTP hydrolysis activity was extremely slow in the presence of sodium, as noted previously (Fig. 4A) (28, 29, 51). Interestingly, replacing sodium with potassium drastically increased the rate of GTP hydrolysis by NFeoB, similar to what was observed for S. thermophilus FeoB (Fig. 4) (51). It appears that E. coli NFeoB has a preference for potassium like its counterparts in other organisms. Potassium has a larger ionic radius than sodium and thus provides a more ideal geometry in the active site of the G protein for the hydrolysis of guanine nucleotides (51). Unexpectedly, our ³¹P NMR data indicate that no change is observed in the rate of GTP hydrolysis by NFeoB in vitro in the presence of FeoA. This result suggests that FeoA probably does not act as a GAP, as had been suggested (5), or that additional cofactors are required to induce an increased rate of hydrolysis. As previously noted, the affinity of NFeoB and the complete FeoB protein toward their guanine nucleotide substrates may vary depending on the construct being investigated (28). In particular, the transmembrane portion of FeoB could have a pronounced effect in regulating the affinity of the G domain toward its substrates, and, thus, its potential role in regulating protein-protein interactions with FeoA or other interacting species should not be disregarded. To this end, performing future in vitro experiments with full-length FeoB could provide additional insight toward the potential role of FeoA as a GAP. Be that as it may, E. coli NFeoB contains only one PXXP recognition site for typical SH3 domains, but this site is buried inside the protein. Moreover, the regular interactions would be disrupted by the insertion of the extra helix in FeoA compared to eukaryotic SH3 domains. Hence, FeoA and NFeoB binding, if any, could not occur in the same way as described for the eukaryotic SH3 domains.

Our NMR relaxation results indicate that some regions of the E. coli FeoA protein experience flexibility and chemical exchange (Fig. 3). In particular, slow conformational exchange is observed for several residues in β-strand 4 (Fig. 3). Val56 has the highest R_ex value and is situated at the end of the distal loop which is sandwiched between β-strands 3 and 4 (Fig. 5A). The distal loop, as previously mentioned, is a defining feature of eukaryotic SH3 domains that is important for protein-protein interactions (55). With the high flexibility of this region observed in the NMR relaxation data, this suggests that this loop may be important for protein-protein interactions involving FeoA. Apart from V56 experiencing higher conformational exchange, it is also situated in one of the surface-exposed hydrophobic patches on FeoA (Fig. 6A), suggesting that it could partake in protein-protein interactions.

Fig 6.

Electrostatic surface potential plots of FeoA (PDB ID 2lx9) (A) and ScaR (PDB ID 3hru chain B) (B). Val56 of FeoA is highlighted in pink on the cartoon representation in panel A. The SH3-like domain of ScaR is highlighted in black in panel B. Positively charged regions are indicated in blue, while negatively charged regions are shown in red. Hydrophobic areas are represented in white. The orientation of the surface potential plot is represented by its corresponding cartoon depiction.

In an attempt to identify potential alternative functions for E. coli FeoA from its structure, we submitted the FeoA structure to the Dali server (60). The Dali server compares newly determined structures against the structures deposited in the PDB to identify structural homologues, which can lead to the establishment of potential evolutionary relationships (60). These results revealed similarity to streptococcal coaggregation regulator (ScaR) (PDB ID 3hru) (Z-score, 7.1), iron-dependent regulator (IdeR) (PDB ID 1u8r) (Z-score, 6.7), and diphtheria toxin repressor protein (DtxR) (PDB ID 1c0w) (Z-score, 5.6). These proteins are all part of the metal and DNA-binding protein families that function as transcriptional regulators (61–68). In addition, these proteins all have three domains: a DNA-binding, a dimerization, and an SH3-like domain (64, 69, 70), where the structure of FeoA resembles the last domain (see Fig. S3 in the supplemental material). Interestingly, primary sequence alignment of FeoA with these proteins reveals a low overall sequence identity (71, 72) despite the structural similarity. Figure S3 shows the superposed structures of FeoA with the structural homologues. FeoA aligns well with the SH3-like C-terminal domains of all three homologs, with overall RMSDs of 2.73 Å, 2.42 Å, and 2.83 Å for ScaR, IdeR, and DtxR, respectively (see Fig. S3). Structural alignment shows that most secondary structural elements and the global fold are conserved (Fig. 5B and E; see also Fig. S3). Minor differences exist, however; the core β-barrel structure that is reminiscent of SH3 domains is conserved.

All three of these structural homologues possess two metal binding sites with one residing in between the dimerization and SH3-like domains (Fig. 5B). The presence of metal ions signals these proteins to transition between the DNA-binding and nonbinding forms (61–70, 73). The SH3-like domains of IdeR and DtxR also play an important role in this conversion apart from sole metal binding. They interact with the other domains of the protein either on an intermolecular or intramolecular level that results in global conformational changes, allowing the binding or release of DNA (62, 63, 65–68, 74). Examining our solution NMR studies, done at high protein concentrations, has shown that FeoA is a monomeric protein, making it unlikely that it dimerizes noncovalently in a similar fashion to IdeR for function. Of the three homologues, the SH3-like domain of ScaR is only known to bind metal, with no specific function like the other two (64, 68). FeoA shares greater structural similarity between all three C-terminal domains of these proteins than with the prototypical SH3 domains (Fig. 5A and B; see also Fig. S3 in the supplemental material). Therefore, it is possible that it would bind metal ions and interact with other proteins to function as a metal-sensing transcriptional regulator.

To investigate this, we examined the metal binding sites of all three proteins and compared them to FeoA for similarity. The contributing ligands from the SH3-like domain of IdeR for the ancillary site are Glu172 and Gln175 (Fig. 5B and C). Gln175 is not present in the structural alignment between the two, while Glu172 aligns with Arg16 of FeoA, an acidic-to-basic substitution with an extended side chain (Fig. 5D and E). This is similar to the SH3-like domain of ScaR, which donates Asp160 for metal coordination in the secondary site, which corresponds to Arg16 of FeoA (Fig. 5E). Therefore, E. coli FeoA does not appear to have the appropriate ligands for metal ion coordination, in contrast to IdeR/DtxR/ScaR. The evidence for FeoA being a transcriptional regulator has not been established, and thus the reason for its structural similarity to these C-terminal SH3-like domains remain unclear. A surface potential plot of ScaR reveals that the only positively charged region is the DNA-binding domain, with the dimerization and SH3-like domains being acidic (Fig. 6B). In contrast, the surface of FeoA is mainly basic (Fig. 6). FeoA could perhaps interact with other proteins, similar to the C-terminal domain of ScaR/IdeR/DtxR, to function in a noncovalent transcriptional regulator complex.

It is thought that ferrous iron diffuses into the periplasm through porins to be transported into the cell through the transmembrane domain of FeoB (Fig. 7) (5). The exact mechanism of uptake is not understood; however, two Gate motifs reminiscent of the Saccharomyces cerevisiae iron permease can be identified within the transmembrane domain (5, 75). It is generally assumed that ferrous iron is transported into the cell through the Gate motifs while the G domain relays information about the intracellular environment to the membrane. Although FeoA was first thought to function as a GAP, this has not been observed in our in vitro studies. Perhaps its function as a GAP in vivo could be dependent on other cofactor(s) which are absent under the conditions tested. However, it is equally possible that FeoA interacts with other highly conserved parts of FeoB. To explore other possible interacting regions of FeoB, its primary sequence was submitted to Pfam, a curated database of protein domain organizations (76). A Pfam analysis of FeoB indicates that the protein is composed of four domains: the N-terminal domain (NFeoB), the two Gate motifs, and another domain in between the two Gate motifs annotated as the C domain of FeoB (CFeoB; residues 457 to 507) (Fig. 7), which we have termed the core domain. Interestingly, part of the core CFeoB region and the carboxy-terminal region (residues 743 to 773) are predicted to form cytoplasmic domains, and these could both be potential alternative FeoA interaction partners (Fig. 7). The conserved core CFeoB region, which is present in all FeoB proteins, may play a role in bringing iron into the cytoplasm. The CFeoB core possesses His, Met, and Glu residues that could act as potential iron-coordinating ligands, and it is possible that FeoA interacts with this region to regulate the rate at which iron enters the cell (Fig. 7). The carboxy-terminal region of FeoB (residues 743 to 773) also has some characteristic features that could be important for iron binding, such as multiple cysteines that could form an iron sulfur cluster to move the iron away from the Gate regions (Fig. 7). However, this region, which is prominent in E. coli FeoB, is found only in a small number of FeoB proteins and therefore seems less likely to play a role in interacting with FeoA. Consequently, we propose that the conserved cytoplasmic core CFeoB (residues 457 to 507) region is another potential interaction partner for FeoA. The importance of the conserved core region to the regulation of the transport of Fe²⁺ through FeoB has not been emphasized before and merits further analysis.

Fig 7.

Proposed function for FeoA in E. coli ferrous iron transport. The striped transmembrane regions represent the Gate motifs while the dotted transmembrane regions represent the core CFeoB domain as represented in the Pfam database. FeoB is thought to take up iron through the Gate motifs while the G protein and the GDI domains signal to the membrane regions about the state of the cell or potentially provide energy for the membrane region for active transport. FeoA is thought to fulfill a dual role: (i) to take part in iron transport through interacting with FeoB and (ii) to form a regulatory complex with a transcriptional regulator.