NMR structure note: the ferrous iron transport protein C (FeoC) from Klebsiella pneumoniae

Biological context

Iron is an essential nutrient involved in many important biological processes ranging from electron transfer, oxygen transport, gene regulation, photosynthesis, N₂ fixation to DNA biosynthesis (Andrews et al. 2003). Thus, the ability to acquire sufficient quantities of iron from the environment is vital for the survival of living organisms. This is particularly true for pathogens that must compete with the host iron-withdrawal response for their iron supplies. However, iron is also toxic and excess iron can be detrimental for the living system (Touati 2000). It is, therefore, essential for almost all life forms to maintain proper iron homeostasis (Andrews et al. 2003).

Feo is a unique ferrous iron transporter commonly utilized by bacterial systems for acquiring ferrous iron from the environment (Cartron et al. 2006). It was first discovered in Escherichia coli K-12 (Hantke 1987). Subsequently the feo locus was cloned and sequenced (Kammler et al. 1993). The importance of the Feo system has been confirmed in several systems. Both E. coli and Salmonella feoB mutants are attenuated in their abilities to colonize the mouse intestine presumably due to their inability to transport ferrous iron within the anaerobic environment of the mouse intestine (Stojiljkovic et al. 1993; Tsolis et al. 1996). In H. pylori, FeoB appears to provide the main route of iron uptake. It is required for H. pylori colonization of mouse gastric mucosa, as well as for normal growth and iron-uptake under iron restricted conditions (Velayudhan et al. 2000). FeoB is also required for intracellular growth of Legionella pneumophila (Robey and Cianciotto 2002). Thus, various studies have clearly established an in vivo role for Feo in colonization of the gut and in virulence.

The feo operon of enterobacteria encodes three proteins: FeoA, FeoB and FeoC. FeoA is a small SH3-like protein predicted to act as GTPase activating protein (GAP) and/or Fe-dependent repressor (Cartron et al. 2006). FeoB is a large protein (773 a.a.) composing of a 270-residue cytosolic N-terminal domain (NFeoB, residues 1–270) that contains a G-protein domain (a.a. 1–170) and a helical bundle S-domain (a.a. 171–270) presumed to be a GDP-dissociation inhibitor (GDI). The C-terminal region of FeoB is a helical transmembrane domain which is likely to act as the ferrous iron permease and essential for ferrous iron transport activity. FeoC (also called yhgG) is a small 78-residue, hydrophilic winged-helix domain (WHD) found only in γ-proteobacteria. The feoC gene was found preceding the feoB gene in the feo operon. Multiple sequence alignment of FeoC proteins shows that they possess four conserved cysteine residues (CxxGxCKxCPx_4-7C) that are likely to provide a binding site for an [Fe–S]-cluster. Thus, FeoC is presumed to be a [Fe–S]-dependent transcriptional regulator directly controlling the expression of the feo operon (Cartron et al. 2006).

Given its importance and novelty, there has had increasing interest in dissecting the molecular basis of the Feo system. In particular, the presence of a G-protein motif in the intracellular domain of FeoB (NFeoB) has spurred considerable interest in understanding its molecular basis and the structures of NFeoB from several species have been published within the last year (Guilfoyle et al. 2009; Hung et al. 2010). We have also determined the structures of NFeB from K. pneumoniae (kpNFeoB) and Pyrococcus furiosus (pfNFeoB) with and without bound ligands (Hung et al. 2010). The NFeoB from E. coli (ecNFeoB) was proposed to form a trimer with the center forming the cation binding site and the GDI domain working as a switch to control the cation transport (Guilfoyle et al. 2009). However, no Fe²⁺ binding was detected, probably due to difficulty of growing crystals of the FeoB/Fe²⁺ complex since Fe²⁺ is readily oxidized to Fe³⁺ which is insoluble. Furthermore, no interaction among the three Feo components has been reported and no information on the role of FeoA and FeoC is available. Here we report the solution NMR structure Klebsiella pneumoniae FeoC (kpFeoC) and discuss its possible functional roles.

Methods and results

Protein expression and purification

Isotopically labeled kpFeoC proteins were overexpressed in E. coli as GST fusion proteins and purified through Glutathione Sepharose 4B column (GE). The GST-tag was removed by PreScission Protease and kpFeoC was eluted from GST-affinity column. The protein was further purified through a HiLoad 16/60 Superdex 75 column. The purified protein contains 5 extra residues (GPLGS) at the N-terminus with a total length of 83 residues. The purity of the protein products were checked by SDS-PAGE and confirmed by mass spectrometry.

NMR spectroscopy

NMR spectra were acquired at 15 °C on Bruker Avance 500 and 600 MHz spectrometers equipped with triple resonance cryogenic probes. Samples for NMR experiments contain 0.2–1.0 mM protein in 20 mM phosphate buffer (pH 6.5), 50 mM NaCl, 0 (oxidized) or 10 mM (reduced) DTT and 10 % D2O. Sequence specific assignments of the polypeptide backbone were made from analysis of HNCA, HN(CO)CA, HNCO, HN(CA)CO, CBCA(CO)NH and HNCACB spectra. Side chain resonances were assigned from the combined information content of the ¹⁵N-edited NOESY-HSQC, ¹³C-edited NOESY-HSQC, HCCH-TOCSY and HBHA(CO)NH spectra. ¹H chemical shifts were externally referenced to 0 ppm methyl resonance of the 2,2-dimethyl-2-silapentane-5-sulfonate, whereas ¹³C and ¹⁵N chemical shifts were indirectly referenced according to the IUPAC recommendations (Markley et al. 1998). NMR data were processed using software Topspin and analyzed by software SPARKY (T. D. Goddard and D. G. Kneller, SPARKY 3, University of California, San Francisco, CA, USA).

NOE distance constraints were obtained from ¹⁵N-edited NOESY-HSQC and ¹³C-edited NOESY-HSQC spectra. The dihedral angle restraints on phi torsion angles (ϕ) and psi torsion angles (ψ) of the protein backbone were empirically predicted using the TALOS software program (Cornilescu et al. 1999). Hydrogen bonds derived from CSI prediction (Wishart and Sykes 1994) were introduced as a pair of distance restraints. Families of structures were calculated by a simulated annealing (SA) approach followed by a refinement procedure using software ARIA 1.2 (Linge et al. 2001). Energy minimization was carried out using torsion angle dynamics (TAD). PARALLHDG force field was employed as non-bonded interactions for the structure calculation. In the first ARIA round, manually-assigned NOEs were included for the generation of initial-folded structures. The unambiguous and ambiguous NOE restraints derived from ARIA outputs were further analyzed and employed as inputs for the next round of calculations. Slightly modified slow-cooling standard SA protocols were used for the calculations. The quality of the constraints was checked by analyzing the violations of the calculated conformers using MolMol (Koradi et al. 1996) and Procheck (Laskowski et al. 1996) software programs.

NMR resonance assignments and structure determination of kpFeoC_R

The purified recombinant kpFeoC protein exists as a monomer in solution over a wide range of pH conditions, as monitored by size-exclusion chromatography (pH 6.5–8.5). The well-dispersed ¹H-¹⁵N HSQC spectra of kpFeoC contain many peaks beyond 8.5 ppm in the ¹H dimension, indicating that kpFeoC is well-structured (blue peaks in Fig. 1a). Addition of 10 mM dithiothreitol (DTT) to the sample to reduce the four cysteine residues (Cys⁵⁶, Cys⁶¹, Cys⁶⁴ and Cys⁷¹) resulted in the disappearance of a set of resonances and the appearance of a new set of resonances in the 8.0–8.8 ppm region. The ¹H, ¹⁵N and ¹³C resonances of the DTT-reduced kpFeoC (kpFeoC_R) at pH 6.5, 288 K have been completely assigned using standard multidimensional heteronuclear NMR techniques (Deposited in the BioMagResBank under accession number 15634). The backbone ¹H, ¹⁵N and ¹³C resonances of the non-reduced kpFeoC at pH 6.5, 288 K have also been assigned. The results showed that the loop region exists in a mixture of two conformers, as indicated by the presence of two sets of resonances, presumably due to the presence of mixed oxidation/disulfide states. The secondary structure of kpFeoC_R, deduced from the consensus chemical shift index (Wishart and Sykes 1994), comprises of three helices (H1, a.a. 4–13; H2, a.a. 19–25; and H3, a.a. 33–42) and three β-strands (B1, a.a. 16–18; B2, a.a. 46–52; and B3, a.a. 73–77) in a H1-B1-H2-H3-B2-B3 topology.

Fig. 1

a ¹H-¹⁵N HSQC spectra of 0.5 mM of recombinant oxidized- (blue) and reduced- (red) KpFeoC acquired at 288 K, pH6.5. The reduced spectrum was obtained in the presence of 10 mM DTT. b Overlay of the backbone traces of 20 best NMR structures. c Ribbon representation of the tertiary structure of KpFeoC_R. The secondary structure elements and the four cysteine residues shown in sticks in W1 loop are labeled. The side chain of Trp⁷⁵ was shown in sticks. d The spatial localization of Trp⁷⁵ and Trp⁷⁶ (Magenta). The side chain of Trp⁷⁵ and Trp⁷⁶ (behind B2 strand) are shown in sticks. e The hydrophobic core of kpFeoC. Trp⁷⁶ (Magenta) is located at the center of the hydrophobic core

The solution structure of kpFeoC_R was further computed based on distance geometry calculations and energy minimization with 1,198 experimental and empirical NMR restraints including 1,075 NOE restraints, 23 hydrogen bond restraints and 100 dihedral angle restraints. Figure 1b shows an overlay of the backbone traces of an ensemble of 20 structures (selected from a set of 100 structures) with lowest energies and good agreement with NMR restraints. A summary of the structure statistics for these 20 structures is given in Table 1. The root-mean-square-deviation (RMSD) values of these structures were 0.436 ± 0.083 Å and 2.468 ± 0.518 Å for the backbone atoms of the structured regions and the full length protein, respectively. The large RMSD of the full length protein is mainly due to the poorly defined long W1 loop. The Ramachandran plot analysis indicated that 82.6 % of residues are in the most favored region, 14.5 % in the additionally allowed region, 2.9 % in the generously allowed region and 0.0 % of residues are in the disallowed regions. The experimental constraints and the coordinates of 20 best conformers of kpFeoC_R have been deposited in Protein Data Bank (PDB ID: 2K02).

Table 1 Structural statistics of kpFeoC_R

Overview of the structural of kpFeoC_R

The solution structure of kpFeoC_R protein comprises a three-helix bundle packed against a three-strand anti-parallel β-sheet (Fig. 1c). The B2-B3 strands are connected by a 20-residue long loop (W1 loop). The Trp⁷⁶ residue, which is conserved in all prokaryotes characterized to date except in Yesinia enterocolitica where it is replaced by a tyrosine residue, is located in the middle of a hydrophobic core composing of Val⁷, Leu¹¹, Ala¹⁹, Ile³³, Met³⁶, Leu³⁷, Met⁴⁰, Val⁴⁶ and Leu⁷⁸ (Fig. 1d). Whilst the side chain of the adjacent Trp⁷⁵ residue, which is conserved in most prokaryotes characterized to date but is replaced by a valine in Cronobacter turicensis and a isoleucine in Yesinia enterocolitica sticks out to the solvent accessible surface and interacts with side chains of Glu¹⁸, Lys²⁰, Ile⁴⁹ and Glu⁵¹. These two terminal tryptophan residues likely play an important structural role in the folding of kpFeoC_R.

The coordinates of the solution structure of the reduced FeoC from E. coli (ecFeoC) has been deposited in PDB data bank (PDB ID: 1XN7). However, the oxidation state of ecFeoC is not specified. The kpFeoC_R is highly homologous to ecFeoC (72 % identity), thus their structures are also very similar (Fig. 2a). The RMSD of the N-terminal 42 residues encompassing the helical bundle region of these two proteins is 1.627 Å. However, kpFeoC_R has much more compact β-sheet structure with well-defined β strands whilst the B2–B3 region of ecFeoC is much more open and the β strands are shorter. Comparison of the structure of kpFeoC_R to that of the winged-helix domain of DtxR (PDB ID: 1XCV) (D’Aquino et al. 2005) or the DNA binding domain of BlaI repressor (PBD ID: 1XSD) (Safo et al. 2005) showed that the B2 and B3 strands, as well as the W1 loop are much longer in kpFeoC_R. The winged-helix domain often uses helix-3 to bind to DNA and the DNA binding site often is highly positively charged, as is the case in BlaI repressor (Fig. 2e). In comparison, the corresponding helix-3 region in kpFeoC_R is actually highly negatively charged, thus is unfavorable for interaction with the negatively charged DNA. This is consistent with our inability to detect DNA binding for kpFeoC (unpublished observation). In comparison, the putative DNA binding surface of ecFeoC is less negatively charged because of the following two substitutions in helix-3: Asp³⁴ → Asn³⁴ and Glu³⁸ → Gln³⁸.

Fig. 2

a Ribbon representation of the structure of reduced FeoC from E. coli (ecFeoC_R). b Ribbon representation of the structure of the winged-helix domain of DtxR (PDB ID: 1XCV). c Ribbon representation of the structure of the DNA binding domain of BlaI repressor from Staphylococus aureus (PDB ID: 1XSD). The structure is shown in an orientation such that the DNA binding site is facing outward. d Surface charge distribution of the DNA binding domain of BlaI repressor. The orientation of the structure is the same as shown on (c). e Surface charge distribution of kpFeoC_R. The orientation of the H3 helix of kpFeoC_R is the same as that in BlaI repressor shown in (c). f ¹H-¹⁵N HSQC spectra of kpFeoC_R without (blue) or in the presence of 1:1 molar ratio of Zn²⁺ (red). The surface charge distribution was generated with the program Pymol (The PyMOL Molecular Graphics System, Version 1.2r3pre, Schrödinger, LLC)

Is FeoC a novel transcriptional regulator for feoABC ?

The kpFeoC has been proposed to be a [Fe–S]-dependent transcriptional regulator directly controlling the expression of the feo operon (Cartron et al. 2006) we next tested the metal cation binding affinity of kpFeoC_R. We found that it binds to Zn²⁺ (Fig. 2f). Binding of Zn²⁺ caused the disappearance of the resonances from the W1 loop region and minor changes of few resonances from residues near the W1 loop. Thus, Zn²⁺ is likely to bind to the cysteine residues in the W1 loop. In contrast, we could not detect any change in the ¹⁵N-HSQC spectrum of kpFeoC_R at 1:1 molar ratio of Fe²⁺, (unpublished observation). Addition of Zn²⁺ or Fe²⁺ did not enhance the DNA binding affinity of kpFeoC (unpublished observation). Our results do not rule out the possibility that kpFeoC forms [Fe–S]-cluster in vivo since [Fe–S]-cluster is known to be highly unstable and that formation of [Fe–S]-cluster requires a biosynthetic machinery (Crack et al. 2012; Johnson et al. 2005). One best characterized sensory protein that requires the [Fe–S]-cluster is that of the E. coli O₂-sensing fumarate and nitrate reduction (FNR) regulator. The E. coli FNR becomes activated by insertion of an O₂-labile [4Fe–4S]-cluster into the N-terminal sensory domain by the iron-sulfur cluster (Isc) biosynthetic machinery. This causes conformational changes resulting in dimerization of the FNR monomer. This enables the C-terminal DNA-binding domain to recognize specific binding site within FNR-controlled promoters to regulate an array of >300 genes. The kpFeoC_R protein in the present study is in the apo-from and it is still not clear whether the four cysteines in kpFeoC can form a [Fe–S]-cluster, thus its role as a [Fe–S]-dependent transcriptional regulator needs further examination.