Overall, folding of Rc-NifF is conserved with respect to other flavodoxins, consisting of a central parallel five-stranded beta sheet (β1–β5B) flanked by five alpha helices (α1–α5) (Figure 1A). Rc-NifF displays the highest structural homology with the long-chain flavodoxin from A. vinelandii[13] with an Rmsd value of 0.55 Å (see Figure 1) Both present an insertion of eight amino acids at the C-terminal of the 50’s loop (NifF insertion) following the β3 strand (residues 65–72), previously observed in other nif flavodoxins (Figure 1).

FMN occupies a cavity located at the C-terminal region of the central β-sheet with the isoalloxazine ring stacked between two hydrophobic residues, the Leu58 (β3–α3 loop) and the Tyr103 (β4–α4 loop) (Figure 1A). Residues sandwiching the isoalloxazine ring mask the prosthetic group from the solvent, creating a negatively charged environment around the FMN that destabilizes the HQ form. Tyr103 is conserved in Flds, and it stabilizes the si face of the isoalloxazine ring by π–π stacking interactions. However, while long-chain flavodoxins present a Trp at the re face of the isoalloxazine, nif Flds carries a Leu residue (Leu58 in R. capsulatus) (Figures 1B and 2A). This position has been reported to be involved in the regulation of the redox potential of the cofactor by creating a hydrophobic environment that destabilizes the HQ and favors the SQ^· form [14]. The presence of this Leu at the re face makes the isoalloxazine more accessible to the solvent compared to non-nif Flds containing a Trp in this position, which should provide a major stabilization of the HQ. Another factor proposed to play an important role in tuning FMN redox state is the conformation of 50’ loop [3]. During FMN reduction from Q state to both SQ^· and HQ states, N(5) atom of the isoalloxazine becomes protonated, and a peptide bond in the 50’ loop (Gly59-Asp60 in Rc-NifF) flips from an O-down conformation to an O-up conformation. Rc-NifF was crystallized in the oxydized state, but, unexpectedly, the 50’ loop was found adopting a trans O-up conformation, which should favor FMN reduction in Rc-NifF.

Analysis of the electrostatic potential surface of Rc-NifF reveals a total of 27 acidic residues and 14 basic residues, giving a net charge of −13. Ten over the twenty-seven acidic residues situate within a radius of 20 Å from the isoalloxazine, creating an electronegative environment around the redox cofactor. These residues are: Asp11 (β1–α1 loop), Glu73 and Glu77 (N-terminal of the α3 helix), Glu37 and Asp60 (β3–α3 loop), Asp99 and Glu106 (β4–α4 loop), Glu137 and Asp138 (β5A–β5B loop) and Asp155 (β5B–α5 loop). Five Asp/Glu amino acids (Asp11, Asp60, Asp99, Glu106 and Asp155) are in the vicinity of the isoalloxazine ring at distances shorter than 7 Å to the redox center, so their carboxyl groups would be directly involved in the tuning of the redox potential of Rc-NifF (Figure 2).

Rc-NifF presents a basic patch adjacent to the FMN-binding site formed by Arg16, Lys17 and Lys20 (α1 helix), Lys33 (α1–β2 loop) and Arg39 (β2–α2 loop). These residues introduce large differences in the charge distribution on the surface of the protein, which change the orientation of the dipole moment with respect to other Fld structures reported so far (Figure 2, panels B–D). This basic patch is also present in nif flavodoxin of A. vinelandii (PDB entry code 1YOB), where five of these six basic residues are conserved [13]. Dipole-moment orientation is also very similar in both R. capsulatus and A. vinelandii Flds. In addition, this particular dipole-moment orientation and conservation of this basic patch have been predicted for A. chroococum and K. pneumoniae Flds [15,16].

The eight-residue insertion present in Rc-NifF comprises five backbone carbonyl groups (Leu66, Leu67, Ans69, Ala70 and Ala71) and one side-chain carbonyl group (Asn69) protruding to the solvent. Interestingly, orientation of all these backbone carbonyls is structurally conserved in A. vinelandii flavodoxin [13]. Also comprised in the 50’ loop insertion of nif flavodoxins, residues 68 and 72 have been shown to be directly involved in the interaction with nitrogenase in A. chroococcum by NMR experiments [16]. All together, these observations support that nif Flds share a peculiar electrostatic potential surface that, as proposed, could play a role in the interaction of these Flds with other nif proteins [15,16].

A phylogenetic analysis carried out with 64 flavodoxin sequences from 52 bacterial species allowed the construction of an unrooted tree using the Neighbor-Joining method (Figure 3). The displayed clustering corresponds to four well-defined groups of microorganisms (firmicutes, α-proteobacteria, γ-proteobacteria and cyanobacteria), the most containing Flds with assigned biological function. Sequences split into two large groups with good statistical support: short chain Flds, corresponding to gram-positive bacteria (firmicutes) and long chain Flds comprising gram-negative organisms (proteobacteria and cyanobacteria) where Rc-NifF is included.

The short-chain sub-family (Firmicutes phylum) includes Bacillus subtilis, Bacillus cereus and Streptococcus pneumoniae, among others. A biological role was reported only in the case of B. subtilis Flds YkuN and YkuP, both capable of supporting biotin synthesis as electron donors to Cyt P450 BioI and biotin synthase [17]. A similar Cyt P450 reductase activity was reported for Clostridium Fld [18], although it came out as a single isolated branch in the phylogenetic tree, suggesting a previous deviation from the ancestor of the Bacillus Flds (Figure 3). A similar divergence is observed by the Megasphaera elsdenii sequence, a flavoprotein commonly used as structural model for studies of its redox activity and modulation [19]. No Fld from Actinobacteria phylum came out during this sequence collection after applying the BLAST protocol with the Rhodobacter NifF as query.

Flavodoxins from Gamma proteobacteria include the two archetypical proteins from enterobacteria. The FldA and FldB from Escherichia coli are both members of the SoxRS regulon, an adaptive system responsive to oxidative damage modulated by redox-cycling agents [20]. Several biochemical and molecular results pointed towards the involvement of FldA in the antioxidant response of E. coli[21]. Previous in vitro studies revealed that FldA is also necessary for providing “low potential” electrons to the reductive activation of enzymes, such as pyruvate-formate lyase and the anaerobic ribonucleotide reductase [22,23]. Although FldB is also a SoxRS responsive protein, its putative role during oxidative stress is still controversial, as a high copy number plasmid carrying fldB gene did not complement the fldA mutation [24]. In this frame, various structural differences cause FldA and FldB to cluster in different groups (Figure 3).

The sequence of Rhodobacter NifF fits into a cluster containing all other nif responsive Flds, within the group of alpha-proteobacteria. Nif flavodoxins have been involved in the reduction of nitrogenase in diazotrophs like A. chroococcum[25], A. vinelandii[26] and K. pneumoniae[27]. The presence of an eight amino acid insertion in the 50’ loop (NifF insertion), probably involved in the interaction with nitrogenase [16], is a typical structural feature of this group, as well as the conservation of a Leu at the re face of the isoalloxazine that substitutes the Trp highly conserved in other long chain Flds (Figures 1 and 2). This Leu residue increases the exposition of the isoalloxazine to the solvent in both R. capsulatus and A. vinelandii Flds, which should therefore provide a major stabilization of the HQ in nif Flds.

The distribution of the surface charged residues found in nif Flds remarkably differs from that observed in cyanobacterial Flds (Figure 2). Sequence comparison analysis shows the presence of numerous basic residues all along the region from α1 to α2 helices (residues 14–41) in nif Flds. Most of the analyzed sequences display few basic residues located down-stream the α1 helix, and only nif Flds present such a high concentration of conserved positive charges in the α1 helix, as marked in Figure 2B (R16, K17, K20, K33 and R39). Structural analysis of R. capsulatus and A. vinelandii Flds shows that these basic residues located in α1 helix generate large deviations of dipole-moment orientation with respect to other Flds (Figure 2B,C). This deviation would be distinctively conserved in nif Flds, supporting its biological relevance. In this regard, evidence reported so far support that this biological meaning is related with the interaction of nif Flds with other nif proteins. In the case of Rc-NifF, those proteins would be the nitrogenase and its proposed natural electron donor, the flavodoxin:NADP(H) reductase [12].

Finally, all eleven cyanobacterial ones are clustered in a definite out-group (see Figure 3). Under iron deficit and consequent ferredoxin scarcity, Fld is synthesized and replaces the Fe–S protein acting as an electron carrier between photosystem I and the ferredoxin-NADP(H) reductase in vegetative cells [28,29]. A possible role of Fld as electron donor to nitrogenase in cyanobacteria is still a matter of debate. While in vitro experiments showed a discrete ability of Anabaena Fld to reduce nitrogenase [30], no stimulation of diazotrophic growth under iron limiting conditions was observed in a heterocyst ferredoxin mutant [31]. The sequence and crystal structure comparison detailed above, together with the differently oriented dipolar moment vectors (Figure 2) and the phylogenetic relationships (Figure 3), depict molecular features that exclude Anabaena Fld from the group of nif flavodoxins.

nifF-coding sequence from R. capsulatus[6] was cloned into pET-32a vector (Novagen, Madison, WI, USA) and expressed as a His₆-Trx tag recombinant protein in E. coli[12]. The tagged protein was purified after Ni-NTA affinity chromatography (Qiagen, Hilden, Germany), and subsequent enterokinase treatment of the fusion protein to render free Rc-NifF, as previously described [12].

Rc-NifF was crystallized using the hanging drop vapor diffusion method at 293 K, as described before [32]. An X-ray data set was collected up to 2.17 Å resolution using the CuKα radiation (λ of 1.5418).

Rc-NifF structure was solved at 2.17 Å resolution by the Molecular Replacement Method using the program MOLREP [33] and coordinates of Anabaena PCC7120 flavodoxin (PDB entry 1FLV). A model consisting of a single molecule in the asymmetric unit was subjected to alternated cycles of refinement with programs CNS [34] and REFMAC [35] and manual model building with the software package O [36]. The geometry of the final model was checked using the program PROCHECK [37], finding all residues to be included in permitted regions of the Ramachandran Plot. An electron density map allowed the determination of the complete polypeptide chain of the flavodoxin (182 residues), one FMN molecule bound to the protein and 76 water molecules (statistics of model refinement data are summarized in Table 1). Coordinates were deposited in t he Protein Data Bank with entry code 2WC1.

The amino acid sequences of 64 Flds from 52 bacteria species analyzed in this work were obtained from the National Center for Biotechnology Information by the BLAST network service. The Flds from Rhodobacter capsulatus, Anabaena, Escherichia coli and Bacillus subtilis were used as the query. To construct the phylogenetic trees, the sequences were aligned in the CLUSTAL X 2.0.9, the windows interface for the Clustal W multiple sequence alignment program [38]. Analyses were performed by the Neighbor-Joining distance method [39] and TreeView X Version 0.5.0 was used to display phylogenies. Confidence limits to the inferences obtained were placed by the bootstrap procedure.