Structural insight into the high reduction potentials observed for Fusobacterium nucleatum flavodoxin

Abstract

Flavodoxins are small flavin mononucleotide (FMN)‐containing proteins that mediate a variety of electron transfer processes. The primary sequence of flavodoxin from Fusobacterium nucleatum, a pathogenic oral bacterium, is marked with a number of distinct features including a glycine to lysine (K13) substitution in the highly conserved phosphate‐binding loop (T/S‐X‐T‐G‐X‐T), variation in the aromatic residues that sandwich the FMN cofactor, and a more even distribution of acidic and basic residues. The E _ox/sq (oxidized/semiquinone; −43 mV) and E _sq/hq (semiquinone/hydroquinone; −256 mV) are the highest recorded reduction potentials of known long‐chain flavodoxins. These more electropositive values are a consequence of the apoprotein binding to the FMN hydroquinone anion with ~70‐fold greater affinity compared to the oxidized form of the cofactor. Inspection of the FnFld crystal structure revealed the absence of a hydrogen bond between the protein and the oxidized FMN N5 atom, which likely accounts for the more electropositive E _ox/sq. The more electropositive E _sq/hq is likely attributed to only one negatively charged group positioned within 12 Å of the FMN N1. We show that natural substitutions of highly conserved residues partially account for these more electropositive reduction potentials.

Short abstract

PDB Code(s): 6OHK and 6OHL;

Abbreviations

FLD

flavodoxin

FMN

flavin mononucleotide

FnFLD

flavodoxin from Fusobacterium nucleatum

hq

hydroquinone

ox

oxidized

sq

semiquinone

Introduction

Fusobacterium nucleatum is an anaerobic Gram‐negative bacterium that is naturally found in the oral cavity of humans. The bacterium is implicated in periodontal disease and gingivitis, which stems from its ability to act as a bridge organism between primary colonizing bacteria such as Streptococcus and late colonizing pathogenic anaerobes (i.e., Treponema denticola and Porphyromonas gingivalis) within the dental plaque biofilm.1 Fusobacterium nucleatum is also an emerging pathogen based on a growing body of research that links the bacteria to colorectal cancer and intra‐amniotic infections.2, 3, 4, 5, 6 Genes critical for F. nucleatum survival have not been experimentally identified, but a recent subtractive genomics study listed flavodoxin (FN0724) as essential for growth.7 In Helicobacter pylori, flavodoxin is required for bacterial survival as it acts as an electron acceptor for pyruvate flavodoxin oxidoreductase, which catalyzes the oxidative decarboxylation of pyruvate.8, 9 The genome of F. nucleatum also carries the gene for pyruvate flavodoxin oxidoreductase (FN1170), along with genes for other enzymes known to participate in electron transfer with flavodoxin, including biotin synthase (FN1000), ribonucleotide reductase (FN0102/3), and pyruvate‐formate lyase activating enzyme (FN0621). Flavodoxin has also been shown to be essential for number of pathogens, including Escherichia coli,10, 11 Streptococcus pneumonia,12 Salmonella enterica,13 Haemophilus influenzae,14 and Vibrio cholera,15 which opens the possibility of using flavodoxin as a drug target. Progress in this area has been made with the discovery of specific inhibitors of H. pylori flavodoxin that elicit bacteriocidal properties.16, 17

Flavodoxins are small acidic proteins (14–23 kDa) that use a noncovalently bound flavin mononucleotide (FMN) cofactor as an electron transfer mediator. A hallmark feature of the flavodoxin family is the ability of the protein to dramatically modulate the redox potentials of the FMN cofactor. The cofactor can exist in three oxidation states, the oxidized quinone, the one‐electron reduced neutral semiquinone, and the two‐electron reduced hydroquinone. The oxidized/semiquinone (ox/sq) and semiquinone/hydroquinone (sq/hq) reduction potentials are −325 and −150 mV versus the standard hydrogen electron, respectively, for FMN in solution at pH 7.0. However, binding of FMN to apoflavodoxin inverts the two reduction potentials with the E _ox/sq rising to as high as −50 mV and the E _sq/hq lowering to as low as −500 mV. The dramatic shift in both couples enables flavodoxin to effectively function as a low‐potential one‐electron reductant, often substituting for ferredoxin under iron‐starvation conditions.

Flavodoxin forms a typical three‐layered α/β protein, with a central five‐stranded β‐sheet surrounded by two helical layers. Flavodoxins are divided into long‐chain and short‐chain subfamilies, with the former group possessing a ~20‐amino acid insert that bisects the 5th β‐strand. FMN is tightly bound to the protein via three loops, the phosphate‐binding loop (residues 10–15 in F. nucleatum), a so‐called 50's loop (residues 56–62) that connects β3 and α3, and a loop connecting β4 and α4 (residues 90–99) referred to as the 90's loop. The latter two loops typically contain two aromatic side chains that sandwich the flavin isoalloxazine ring. The 5′‐ribityl phosphate group is bound to the highly conserved phosphate‐binding loop, which serves as a fingerprint motif for flavodoxins with the following consensus sequence T/S‐X‐T/S‐G‐X‐T/S.18 Hydrogen bonds are formed between the phosphate and the side chains of serine/threonine residues and backbone NH groups of residues within the loop.

From a structural perspective, flavodoxin from F. nucleatum appears unique in that sequence alignment with structurally determined flavodoxins reveals substitutions of a number of highly conserved residues (Fig. 1). First, the phosphate‐binding loop in F. nucleatum (₁₀T‐L‐T‐K‐T‐T₁₅) contains a lysine at position 13, in place of the normally occurring glycine. A glycine at this position of the loop enables the peptide backbone to adopt the torsion angles (ψ = −11° and Φ = 103°) normally disallowed for side chain bearing amino acids.18 Second, F93 replaces the highly conserved tyrosine that flanks the si‐face of the FMN, while Y56 is in place of the normally occurring tryptophan residue on the re‐face. Third, there is a more equal distribution of acidic and basic residues in FnFld, resulting in a higher calculated pI (Table S1). Specifically, FnFld contains N89 in place of a highly conserved aspartic acid that is known to influence the E _sq/hq redox potential in other flavodoxins.19, 20 Herein, we reverted these natural sequence variations to the conserved residues observed for the flavodoxin family to assess their impact on the binding of FMN and the redox properties of the protein. We also present the structure of wild‐type FMN and the K13G variant, which provide insight into how these structural deviations impact the biophysical and thermodynamic properties of the protein.

Figure 1.

Amino acid sequence alignment of a selection of flavodoxins from different species aligned with Clustal MUSCLE: FUSON (Fusobacterium nucleatum strain ATCC 25586), PSEUD (Pseudomonas aeruginosa), CLOSB (Clostridium beijerinckii), DESUV (Desulfovibrio vulgaris), CHOND (Chondrus crispus), HELPI (Helicobacter pylori), ECOLI (Escherichia coli), NOSTC (Nostoc sp PCC 7120), SYNEL (Synechococcus elongates), AZBVI (Azotobacter vinelandii), and RHODO (Rhodobacter capsulatus).

Results

Binding of FMN to apoflavodoxin

Stopped‐flow spectrofluorimetry was employed to determine the rate constants associated with the binding of FMN to apoflavodoxin. In these experiments, the apoprotein was rapidly mixed with a 7.5–50 molar excess of FMN to maintain pseudo‐first‐order conditions. As shown in Figure 2(a), the time course of the fluorescence emission was monophasic under our experimental conditions (10 mM HEPES, pH 7.0, 25°C, 50‐fold excess of FMN over apoprotein), and a fit of the trace to a single exponential equation gave an observed rate constant of 1.04 ± 0.01 s⁻¹ for the wild‐type protein. As shown in Figure 2(b), these observed rate constants are linearly dependent on the concentration of FMN. The slopes of a plot were used to calculate microscopic k _on values of 20.1 ± 0.8 mM ⁻¹ s⁻¹ (wild type), 46.4 ± 1.3 mM ⁻¹ s⁻¹ (K13G), 22.7 ± 2.1 mM ⁻¹ s⁻¹ (Y56W), 23.0 ± 0.9 mM ⁻¹ s⁻¹ (N89D), and 37.9 ± 0.9 mM ⁻¹ s⁻¹ (F93Y; Table 1). Complex formation occurs in a single step as evidenced by the lines passing through the origin of the plot. The binding of riboflavin (the FMN precursor lacking the phosphate group) was not detected for either wild‐type apoflavodoxin or the variants, revealing that the phosphate moiety is a major binding determinant.

Figure 2.

(a) Binding of 50 μM FMN to 1 μM wild‐type apoflavodoxin. The reaction was performed in 50 mM HEPES, pH 7.0 at 25°C. The horizontal line is the fluorescence emission after mixing buffer with FMN. The pseudo‐first‐order rate constant (k _obs) was obtained by fitting an average of five traces to a standard single exponential equation. (b) Plot of the pseudo‐first‐order rate constant as a function of FMN concentration for wild type (squares), K13G (open circles), Y56F (closed triangle), F93Y (hexagon), and N89D (down triangle).

Table 1.

Association (k _on) and Dissociation Rate Constants (k _off) and Equilibrium Dissociation Constant (K _d) for FMN Binding to Apoflavodoxin Performed in 10 mM HEPES pH 7 at 25°C

Protein	k on (mM −1 s−1)	k off (×10−4 s−1)a	K d (nM)
Wild type	20.1 ± 0.8	6.8 ± 1.5	34 ± 6
K13G	46.4 ± 1.3	4.0 ± 0.2	9.0 ± 0.8
Y56W	22.7 ± 2.1	157 ± 14	692 ± 61
N89D	23.0 ± 0.9	4.4 ± 0.2	19 ± 1
F93Y	37.8 ± 2.0	3.6 ± 0.9	9.4 ± 2.3

^aDetermined from the product of k _on and K _d. The errors were determined by propagating the errors in k _on and K _d.

The dissociation constants (K _d) of the FMN–apoflavodoxin complex were also determined for wild type and the variants using a fluorescence quenching experiment (Fig. 3). As shown in Table 1, wild‐type apoflavoxin elicits binding affinity of 34 ± 6 nM for FMN. There was a modest increase in the stability of the FMN–apoflavodoxin complex for K13G (fourfold), F93Y (fourfold), and N89D (twofold). The Y56W variant elicited a 20‐fold FMN weaker binding affinity, destabilizing the complex by 7.5 kJ/mol at 25°C. The dissociation constants (k _off) calculated from the K _d and k _on values further reveal that the release of FMN is 23‐fold slower in Y56F (k _off = 157 ± 14 s⁻¹) compared to wild‐type FnFLD (k _off = 6.8 ± 1.3 s⁻¹). The remaining variants exhibited k _off values that were comparable to wild‐type FnFLD.

Figure 3.

Change in fluorescence emission of FMN upon addition of apo‐forms of wild type (squares), K13G (open circles), Y56F (closed triangle), F93Y (hexagon), and N89D (down triangle). Measurements were made in 50 mM HEPES, pH 7.5 at 25°C. The data were fitted to a quadratic binding isotherm [Eq. (1)] to determine the dissociation constant.

Redox potentials

The redox potentials of flavodoxin were determined by titration of the oxidized protein with dithionite under anaerobic conditions. The UV–visible absorbance spectra of the wild‐type flavoprotein during the redox titration are shown in Figure 4 and those of the variants are shown in Figure S1. The addition of the first equivalent of dithionite led to the appearance of a broad absorbance band with a maximum at 580 nm and an isosbestic point at 502 nm, signifying formation of the blue neutral semiquinone. Addition of a second reducing equivalent led to a decrease at 580 nm as the blue neutral semiquinone was reduced to the hydroquinone. The midpoint potential was determined by fitting the absorbance change at 580 nm versus potential (normalized to the standard hydrogen electrode) to the modified form of the Nernst equation [Eq. (2)]. For wild‐type flavodoxin the E _ox/sq was −43 mV and the E _sq/hq was −256 mV (Table 2).

Figure 4.

Absorption spectra of wild‐type FnFld (a) recorded during anaerobic redox titration. The spectra and reduction potentials were recorded after each addition of dithionite. Spectra associated with transition of the FMN_ox to FMN_sq are represented with solid lines, while the transition from FMN_sq to FMN_hq is denoted with dotted lines. Plots of absorbance at 580 nm for wild‐type FnFld (b) versus the potential (normalized to the standard hydrogen electrode) were fitted to Eq. (2) to extract reduction potentials for the E _ox/sq and E _sq/hq couples, which are presented in Table 2.

Table 2.

Redox Potentials of Wild Type and the Single‐Site Variants and the Calculated Free Energies Associated with the FMN_ox‐apoFnFld (ΔG _ox), FMN_sq‐apoFnFld (ΔG _sq), and FMN_hq‐apoFnFld (ΔG _hq)

Protein	E ox/sq (mV)	E sq/hq (mV)	ΔG ox a (kJ/mol)	ΔG sq b (kJ/mol)	ΔG hq c (kJ/mol)
Wild type	−43 ± 2	−256 ± 3	−43	−61	−53
K13G	−11 ± 3	−262 ± 3	−46	−67	−59
Y56W	−51 ± 3	−280 ± 3	−38	−56	−45
F93Y	−97 ± 2	−259 ± 3	−46	−59	−51
N89D	−39 ± 2	−301 ± 2	−44	−63	−51
FMN free	−238	−172

^aCalculated from the experimentally determined K _d shown in Table 1 using the equation ΔG _ox = −RTln1/K _d.

^bCalculated from Eq. (3).

^cCalculated from Eq. (4).

The E _ox/sq redox couple of the K13G variant (−11 mV) is slightly more electropositive compared to the wild‐type protein, while the E _sq/hq couple was similar at −262 mV. The Y56W substitution resulted in 5–10 mV more electronegative E _ox/sq and E _sq/hq values. The E _sq/hq reduction potential of the F93Y variant was unchanged compared to the wild‐type protein, but the E _ox/sq couple shifted by ~50 mV to a more electronegative value. The N89D variant resulted in a 45 mV reduction in E _sq/hq, while the E _ox/sq was largely unchanged compared to the wild‐type protein with a reduction potential of −39 mV.

Shifts in E _ox/sq and E _sq/hq observed for flavodoxins relative to that observed for the free FMN in solution are a direct consequence of the affinity apoflavodoxin for the three redox states of the cofactor. This relationship is best represented through a thermodynamic box, which connects E _ox/sq and E _sq/hq of unbound FMN, the corresponding potentials found for FnFld, and the dissociation constant for apoFnFld‐FMN_ox experimentally determined through the fluorescence titration experiment (Fig. 5). Using this thermodynamic box and Eqs. 3, 4, the free energies associated with the apoFnFld‐FMN_sq (ΔG _sq) and apoFnFld‐FMN_hq (ΔG _hq) were determined. As shown in Table 3, ΔG _sq = −61 kJ/mol and ΔG _hq = −51 kJ/mol (Table 2). These free energy values correspond to K _d values of 0.017 and 0.42 nM for the apoFnFld‐FMN_sq and apoFnFld‐FMN_hq complexes, respectively. According to these calculations, apoFnFld preferentially binds to the semiquinone form of FMN over the other two redox states of the cofactor, which is typically observed for the flavodoxin family. Uncharacteristically, apoFnFld has a ~70‐fold tighter binding affinity for the hydroquinone anion compared to the oxidized FMN. Likewise, all four variants follow the same trend with the strength of protein cofactor interaction proceeding as follows: apoFnFld‐FMN_sq > apoFnFld‐FMN_hq > apoFnFld‐FMN_ox.

Figure 5.

A thermodynamic box connecting the reduction potentials of FMN in solution (E ^free _ox/sq, E ^free _sq/hq) with the reduction potentials of FnFld (E _ox/sq, E _sq/hq) and the free energies associated with the binding of the oxidized (ΔG _ox), semiquinone (ΔG _sq), and hydroquinone (ΔG _hq) forms of FMN.

Table 3.

Comparison of the Redox Potentials and the Free Energies Associated With Binding of the Three Redox Forms of FMN to Long‐Chain and Short‐Chain Apoflavodoxin from Different Organisms

Organism	E ox/sq (mV)	E sq/hq (mV)	ΔG ox a(kJ/mol)	ΔG sq b(kJ/mol)	ΔG hq c(kJ/mol)	References
Fusobacterium nucleatum	−43	−256	−42.6	−61.4	−53.3	This work
Synechococcus elongatus	−221	−442	−50.5	−52.1	−26.0	[20]
Nostoc sp PCC 7120	−207	−439	−50.9	−53.8	−28.1	[21]
Escherichia coli (FldA)	−260	−452	−51.3	−49.2	−22.1	[22, 23]
Desulfovibrio vulgaris	−143	−440	−54.9	−64.0	−38.1	[24]
C. beijerinckii	−92	−399	−44.2	−58.2	−36.3	[25]
Desulfovibrio desulfuricans	−58	−387	−35.7	−53.0	−32.2	[26]

Three‐dimensional structure

The crystal structures of FnFld and the K13G variant were solved to 1.85 and 1.20 Å resolution, respectively, using molecular replacement. The data collection and refinement statistics are presented in Table S2. Both the wild type and the K13G variant contained one monomer in the asymmetric unit and the three‐dimensional structures are essentially superimposable with an R.M.S. deviation of 0.14 Å for the Cα atoms. The overall structure of FnFld conforms to the prototypical α/β‐sandwich fold of the flavodoxin family with five parallel β‐strands (β1–β5) flanked by five α‐helices [α1–α5; Fig. 6(a)]. FnFld is a long‐chain flavodoxin with a 22‐amino acid stretch that splits the fifth β‐stand. A structural similarity search of the Protein Data Bank identified the long‐chain flavodoxin from Nostoc sp PCC 7120 (Anabaena, PDB entry:1FLV) as having the highest similarity with an R.M.S deviation of the Cα of 1.3 Å.27

Figure 6.

(a) A cartoon structure of wild FnFld. The FMN is shown in stick form with the carbon atoms in gray. Loops that form interactions with the FMN are blue, while the remaining protein is pink. (b) A superposition of residues in wild‐type FnFld and K13G that form the phosphate‐binding loop and their hydrogen bonding interactions with the ribityl phosphate group. Hydrogen bonds are denoted with black dotted lines. (c) Close‐up of the aromatic residues that sandwich the FMN isoalloxazine ring, residues (Q57 and V58) that likely rotate upon formation of the semireduced form of the cofactor, and E145, the only residue with a negatively charged group that is within 12 Å of the FMN N1 (d) charge distribution map of FnFld with the surface colored by electrostatic potential from −5 kT/e (red) to +5 kT/e (blue).

The binding site for the FMN is located at the C‐terminal end of the β‐sheet. Table S3 shows the FMN group interactions to neighboring residues in wild‐type FnFld and the K13G variant in comparison to that of Nostoc sp PCC 7120.27 The phosphate group engages in O—HO hydrogen bonds with the side chains of Y56, T10, T12, T14, and T15, in addition to NH—O hydrogen bonds to the backbone atoms within the phosphate‐binding loop [L11, T12, T14, and T15; Fig. 6(b)]. Overall, the number of hydrogen bonds between the ribityl phosphate and the protein is unchanged with the K13G substitution. In flavodoxin from Nostoc sp PCC 7120 and E. coli there is a hydrogen bond between the hydroxyl of T10 and the NH of G13 (Nostoc numbering) that appears to stabilize the loop.22, 27 In other flavodoxins, including long‐chain flavodoxins from Synechococcus elongatus (formally Aspergillus nidulans), H. pylori and the short‐chain flavodoxins from Desulfovibrio vulgaris, Desulfovibrio desulfuricans, and Clostridium beijerniki, the amide of the corresponding glycine residue appears to directly engage in hydrogen bond contact with one of the oxygen atoms of the phosphate group.8, 18, 25, 28, 29 FnFld mirrors the former group as the amide group of K13 (wild type) and G13 (in the K13G variant) form a hydrogen bond with the side chain of T10 (OH—N distance of 2.8 Å in both cases). The Ψ and ϕ angles of the backbone of G13 in the K13G variant are −11° and 101°, in the favored region of the Ramachandran plot for glycine. These two torsion angles rotate slightly in the wild‐type protein to −4° and 79° to avoid unfavorable steric interactions between the Cβ and the neighboring carbonyl group.

The ribityl hydroxyl groups engage in a hydrogen bond with the solvent plus the backbone or side chains of residues T55 and E145. The atoms of the isoalloxazine ring form interactions with the backbone atoms of N89, S96, C98, and G59. A notable difference between FnFld and the flavodoxin from Nostoc sp PCC 7120 is the absence of a hydrogen bond between the protein and the N5 of the FMN. In flavodoxins from Nostoc sp PCC 7120, E. coli and S. elongates, the amide of residue ~59 in the 50's loop forms a weak hydrogen bond to the FMN N5, which tunes E _ox/sq.20, 22, 27, 30 In FnFLd, the corresponding NH group (from V58) is >4 Å from the FMN N5 and the N—H—N geometry is unfavorable for hydrogen bond formation [Fig. 6(c)]. The N5 of FnFld instead forms a hydrogen bond with a water molecule (WAT106), which also interacts with the O4 of the FMN ring. The FMN O4 also forms a hydrogen bond with backbone amide of G59 and the side chain of S96. The loop comprising residues Q57–G59 is well defined in the electron density map (Fig. S2). No difference density is observed that would suggest alternate conformations within this loop.

The isoalloxazine ring is further secured to the protein by stacking between two aromatic side chains. F93 is nearly parallel with the si‐face of the FMN isoalloxazine ring forming extensive π‐π stacking interactions. The coplanar orientation of F93 is similar to that of the normally occurring tyrosine residue that appears in other structurally determined flavodoxins [Fig. 6(c)]. Likewise, the rotated conformation of the Y57 phenolic group with respect to the re‐face of the FMN ring is similar to that observed for flavodoxins that possess the more frequently observed tryptophan residue in this position. The ring in the wild type structure is planar consistent with being in the oxidized state. In the K13A mutant, the flavin ring is bent ~5° suggesting partial reduction of the cofactor likely as a result of photoreduction by the synchrotron X‐ray beam.

Typically, flavodoxins possess charge asymmetry on the surface of the protein, with a large negative patch surrounding the FMN cofactor. This clustering of acidic residues serves to lower the E _sq/hq reduction potential by destabilizing the binding of the hydroquinone anion.20, 31 It also mediates recognition with redox partner proteins.32, 33 Analysis of the electrostatic potential surface of FnFld reveals that the protein does not possess this large negative patch around the FMN cofactor [Fig. 6(d), Fig. S3]. Two smaller negative patches are located on the surface of the protein above and below the FMN. FnFld also contains small basic cleft comprising residues K13, K37, N34, and H62 that is filled with solvent. These residues introduce a notable difference in the charge distribution on the surface of the protein, which changes the orientation of dipole moment from that normally observed for this protein family. FnFld only has one negatively charged group (E145) within 12 Å of the FMN N1‐O4 (where the negative charge on the FMN hydroquinone localizes). By comparison, flavodoxin from Nostoc sp PCC 7120 contains five carboxyl groups positioned <12 Å from the same region of the FMN.27

Discussion

FMN binding to apoFnFld

The FMN cofactor is tightly secured to the apoprotein through hydrogen bond interactions and van der Waals contacts between hydrophobic side chains and the isoalloxazine ring. The phosphate‐binding site in flavodoxin is the most highly conserved structural region of the flavodoxin protein family. It also represents an atypical phosphate‐binding motif as it does not contain any positively charged residues to stabilize and neutralize the negatively charged phosphate dianion. Instead, it is stabilized by a network of NH—O and OH—O hydrogen bonds supplied by the backbone NH atoms and the side chains of neutral serine and threonine residues. These interactions contribute significantly to the binding energy of FMN as riboflavin does not form a complex with wild‐type apoFnFld or any of the variants studied. Although, the phosphate‐binding loop of FnFld does not adhere exactly to the consensus sequence with a lysine at position 13, this substitution does not appear to greatly affect interactions with the phosphate group. Consequently, its inclusion into the phosphate‐binding loop only has a modest effect on FMN binding affinity. In the wild‐type protein, K13 does form part of a positively charged cleft on the same protein face as the exposed FMN cofactor, so it may be part of a recognition interface for redox partner proteins.

Y56W was the only variant in this study to substantially weaken FMN binding affinity. The 20‐fold increase in K _d observed for Y56F is entirely attributed to a faster dissociation rate constant of the cofactor from the protein scaffold. This implies that the mutation does not alter factors (electrostatic and/or hydrophobic) that govern the initial encounter between the apoprotein and the FMN cofactor. In the wild type structure, the phenolic group of Y56 engages in hydrogen bond contact with the phosphate. In structures of long‐chain flavodoxins from Nostoc sp PCC 7120, S. elongates, E. coli, and Chondrus crispus (which harbor the more frequently observed tryptophan), the NH of the indole ring engages in a hydrogen bond with the ribityl phosphate.18, 27, 34, 35 Thus, it is conceivable that the Y56W mutation in FnFld maintains the hydrogen bond to the phosphate, but this is not confirmed without the structure of this variant. Likely, its absence would lead to a faster k _off. Alternatively, weaker binding may be a consequence of the substitution altering the conformation of the 50's loop, which has shown to be a highly flexible region of protein and important for binding the oxidized form of the cofactor.36, 37, 38

Perturbation of the E_ox/sq redox potential

The shift in the FMN E _ox/sq to a more electropositive value from that observed for the cofactor in solution is attributed to the preferred stabilization of the FMN semiquinone state by the protein scaffold over that of the oxidized form.20, 24, 25, 29 Single electron reduction of the oxidized FMN to the semiquinone is accompanied by protonation of the FMN N5 to N(H)5, converting this atom to a hydrogen bond donor. In the oxidized structures of flavodoxin from S. elongatus, E. coli, and Nostoc sp PCC 7120, the FMN N5 engages in a weak hydrogen bond with the NH group of residue ~59 (Fig. 7).18, 20, 25, 28 The interaction is considered weak as the distance between the two heteroatoms is 3.5–3.7 Å.39 Conversion of S. elongatus flavodoxin to the semiquinone induces rotation of the N58‐V59 peptide backbone to allow for formation of a stronger hydrogen bond between the carbonyl group of N58 and the FMN N(H)5 (Fig. 7). Formation of this new and stronger hydrogen bond is believed to result in preferred binding of the semiquinone over that of the quinone.20, 24, 25, 28, 29, 40, 41 Short‐chain flavodoxins from Clostridium beijerinckii and D. desulfuricans also undergo an analogous peptide flip upon formation of a semireduced state; however, these flavodoxins do not form a hydrogen bond to the oxidized FMN N5.25, 28 As a consequence, the difference in the free energy associated with binding of the FMN quinone (ΔG _ox) compared to that of the FMN semiquinone (ΔG _sq) is >9 kJ/mol in short‐chain flavodoxins and 1.5–3 kJ/mol in long‐chain flavodoxins (Table 3). The larger energy difference associated with the former group is partially due to the absence of an energetic penalty associated with breaking a hydrogen bond in the oxidized state.42 Consequently, it is easier to reduce short‐chain flavodoxins, and as a result their E _ox/sq is ~150 mV more electropositive.

Figure 7.

Comparison of the dipeptide–FMN interaction in the 50's loop among four long‐chain flavodoxins. (a) Superposition of the structures of oxidized (pink, PDB entry: 1OFV) and semiquinone (cyan, PDB entry: 1CZL) forms for flavodoxin from S. elongatus. (b) Comparison of the distance between the NH of residue 59 (or 58 in the case of FnFld) and the FMN N1 for three long‐chain flavodoxin (PDB codes are given in parenthesis).

Although FnFld is a long‐chain flavodoxin, the protein scaffold does not form a direct hydrogen bond to the oxidized FMN N5. The distance between the N5 and the closest hydrogen bond donor (the NH of V58) is 4.9 Å and the N‐H‐N angle is unfavorable for formation of a hydrogen bond. The absence of this interaction likely increases the energy difference between ΔG _ox and ΔG _sq to 19 kJ/mol which in turn results in a E _ox/sq value (−49 mV) that is comparable to short‐chain flavodoxins. The flavodoxin domains of neuronal nitric oxide synthase and cytochrome P450 reductase, which are structurally related to short‐chain flavodoxins, also lack a hydrogen bond between FMN N5 and the protein scaffold.43, 44 The absence of this interaction likely accounts for E _ox/sq values of −43 mV and −49 observed for cytochrome P450 reductase and neuronal nitric oxide synthase, respectively.45, 46 FnFld is not unique among structurally determined long‐chain flavodoxins for its lack of a direct hydrogen bond to the FMN N5, as structures of flavodoxins from Rhodobacter capsulatus and H. pylori also lack this specific interaction.8, 47 The redox potentials for these latter two flavodoxins have yet to be determined, so it remains to be confirmed if this structural variation results in more electropositive E _ox/sq values.

Previous mutagenesis studies of the highly conserved FAD‐shielding tyrosine in flavodoxin from Nostoc sp PCC 7120 (Y94) highlights its importance in tuning the FMN redox potentials.48 A Y94F mutant of this protein weakened the binding affinity for the oxidized FMN and caused the E _ox/sq to become 26 mV more electropositive.48 The reverse mutation in FnFld (F93Y) unsurprisingly caused the opposite effect. The K _d for the protein–FMN complex increased fourfold, while E _ox/sq shifted from −43 to −97 mV. In both the Y94F variant from Nostoc sp PCC 7120 and the F93Y variant from F. nucleatum, the E _sq/hq was unaffected. Calculation of the binding energy of the semiquinone (ΔG _sq) and hydroquinone (ΔG _hq) reveals that more electron‐rich tyrosine of the F93Y variant preferentially stabilizes (albeit modestly) the binding of the oxidized form of the cofactor over the other two redox states. The increase in the stability of the FMN_ox‐apoF93Y complex may be attributed to stronger dispersion forces between the more electron‐rich tyrosine and the oxidized isoalloxazine ring.

Perturbation of the E_ox/sq redox potential

Unlike conversion of the FMN quinone to the neutral semiquinone, single electron reduction of the semiquinone to the hydroquinone is not accompanied by protonation of the isoalloxazine ring; the FMN N1 remains unprotonated.49, 50 Consequently, the hydroquinone is anionic. The low reduction potential of the E _sq/hq that typifies the flavodoxin family is linked to the weak binding affinity of the hydroquinone anion to the apoprotein. It is often four‐ to five‐orders of magnitude weaker than the binding of the oxidized or semiquinone forms of the cofactor (Table 3). Charge repulsion between hydroquinone and a clustering of negatively charged residues in the vicinity of the isoalloxazine ring substantially weakens the affinity of the fully reduced form of the cofactor.19, 20 Neighboring hydrophobic residues, including the two aromatics that sandwich the isoalloxazine ring also contribute to the charge repulsion, as they prevent shielding of the negative charged residues from the solvent.51

The effects of these negatively charged residues have been studied through mutagenesis in several flavodoxins. For example, swapping D90 and D100 in the flavodoxin from S. elongatus for an asparagine cause E _sq/hq to become 40 and 25 mV more electropositive, respectively.20 The more pronounced effect of the D90N mutant on E _sq/hq is likely a consequence of the close proximity (~5 Å) of its carboxylate carbon atom to the FMN N1. The carboxylate carbon of D100 is ~6 Å from the FMN N1. A similar effect was observed in D. vulgaris. Neutralizing individual aspartates and glutamates in this short‐chained flavodoxin led to a 10–13 mV increase in E _sq/hq, with the exception of D95, which corresponds to D90 in S. elongates. The D90N variant of D. vulgaris resulted in a 50 mV more positive E _sq/hq, which again is likely due to its closer proximity to the FMN N1.19 D90 (S. elongatus numbering) is highly conserved among long‐chain flavodoxins, whereas D100 is less well conserved. However, in F. nucleatum, the corresponding residues are N89 and G99. Given the above mutagenesis studies of D. vulgaris and S. elongates, it is not surprising that N89D resulted in a ~45 mV more electronegative E _sq/hq value of FnFld. Incidentally, the corresponding residues in human cytochrome P450 reductase and human neuronal nitric oxide synthase are N178 and S886, respectively.43, 44 The absence of negatively charged groups at these positions may partially account for more electropositive E _sq/hq values observed for the eukaryotic enzymes.45, 46 The more electropositive E _sq/hq value observed for FnFld may also attributed to the increase in the solvent accessibility of the FMN cofactor which was calculated to be 127 Ų. By comparison, the solvent accessible surface areas calculated for the flavodoxins of Nostoc sp PCC 7120, (PDB entry: 1FLV), S. elongatus (PDB entry: 1CXK), and E. coli (PDB entry: 1AHN) are 111, 96, and 101 Ų, respectively.

In summary, we have shown that the long‐chain flavodoxin from F. nucleatum elicits unusually high redox potentials as a consequence of the FMN hydroquinone anion having a higher binding affinity to the apoprotein compared to the oxidized form of the cofactor. The E _ox/sq couple is comparable to that observed in some short‐chain flavodoxins and it likely arises from the absence of a hydrogen bond between the oxidized FMN N5 and the protein scaffold. The more electropositive E _sq/hq couple arises from the near absence of negatively charged residues in the vicinity of the N1 atom of FMN. Naturally occurring substitutions of several conserved residues (i.e., Y56, N89, and F93) partially account for the more electropositive reduction potentials. While the presence of K13 in the phosphate‐binding loop does not dramatically alter the FMN binding affinity, it does form part of a positively charged cleft on the surface of the protein, which may facilitate binding with partner proteins.

Materials and Methods

Cloning, expression, and purification of flavodoxin from F. nucleatum

The flavodoxin gene was amplified from the genomic DNA of F. nucleatum strain ATCC 25586 with Pfu Turbo DNA polymerase using the primers 5′‐GGA ATT CCA TAT GAA GAC TAT TGG TAT TTT TTA TGC‐3′ and 5′‐CGG GAT CCT TAT TTG AAT TCT TTT TTT ATT TC‐3′ which harbor the Nde I and BamHI restriction sites (boldface type), respectively. The PCR product was cleaved with NdeI and BamHI and ligated into pet41a also cut with the same restriction enzymes. The resulting plasmid (designated p41FnFld) was transformed into Roseatta(DE3)pLysS for protein expression. Roseatta(DE3)pLysS cells harboring the p41FnFld were grown at 30°C in 0.5 L of terrific broth supplemented with 100 μg/mL of ampicillin and 35 μg/mL of chloramphenicol until the absorbance at 600 nm reached 0.6–0.8. At this stage of cell growth, isopropyl thiogalactoside was added to the growth media to a final concentration of 0.2 mM. The cultures continued to grow at 25°C at 200 rpm for 16 h. Cells from 3 L of culture were harvested by centrifugation (6000g, 15 min, 4°C) and the cell pellet was stored at −80°C.

Flavodoxin was purified following a protocol similar to that described by Genzor et al.52 All purification steps were performed in an ice‐water bath or at 4°C. Flavodoxin was purified by resuspending the cell pellet in 50 mM Tris–HCl pH 8.0 containing 1 mM EDTA, 0.5 mM PMSF, and 1 mM β‐mercaptoethanol. The cell suspension was sonicated for 30 min (30 s pulses at 1 min intervals) and then centrifuged at 38,000g for 60 min. Ammonium sulfate was gradually added to the supernatant to a final concentration of 65% (w/v). The cell suspension was centrifuged at 38,000g for 15 min to remove precipitated proteins. The yellow supernatant was applied to a DEAE Sephacel column (5 × 6 cm) equilibrated with 50 mM Tris–HCl pH 8.0 with 65% ammonium sulfate (Buffer A). The column was washed with two column volumes of Buffer A and then the desired protein was eluted by applying a linear gradient over four column volumes to 50 mM Tris–HCl, pH 8.0. Yellow fractions were pooled, concentrated to 40 mL, and dialyzed against 4 L of 50 mM Tris–HCl, pH 8.0 for 16 h. The dialysate was then applied to a Q‐sepharose column (2.6 × 14 cm) equilibrated with 50 mM Tris–HCl, pH 8.0. The column was washed with two column volumes of 50 mM Tris–HCl, pH 8.0 to remove unwanted proteins. Separate fractions containing flavodoxin and apoflavodoxin were obtained by applying a linear gradient from 0 to 1M NaCl in 50 mM Tris–HCl over 10 column volumes. The total amount of apoflaovdoxin was determined by the Bradford assay using BSA as a standard, and the concentration of flavodoxin was determined using an extinction coefficient of 9.1 mM ⁻¹ cm⁻¹ (ε_{456 nm}). Apoflavodoxin was also obtained by treatment of flavodoxin with trichloroacetic acid.53

Site‐directed mutagenesis

All of the variants of FnFld were created by site‐directed mutagenesis using the pFnFld plasmid as a template and PfuTurbo DNA polymerase. Each variant was expressed and purified using the same protocol as the wild‐type enzyme. The oligonucleotides for the mutagenesis are shown in Table S4.

Stopped‐flow kinetic measurements

A SF‐61DX2 stopped‐flow spectrophotometer (TgK Scientific) was used to monitor the binding of FMN to apoflavodoxin. The excitation wavelength was set at 446 nm, and a 530 nm cutoff filter was inserted in front of the photomultiplier positioned orthogonal to the incoming light. The reactions were performed under pseudo‐first‐order conditions with the concentration of FMN (7.5–50 μM) in excess of the protein (1 μM). Experiments were performed in 50 mM HEPES, pH 7.0, 25°C.

Fluorescence titrations

The dissociation constants (K _d) for the FMN complexes (wild type and the four single‐site variants) were determined by titrating a 0.1 μM solution of FMN with aliquots of apoflavodoxin (5–300 μM). The titrations were performed at 25°C in 50 mM HEPES pH 7.0, using a Perkin‐Elmer LS50‐B spectrophotometer. Quenching of the flavin fluorescence was followed at 525 nm following excitation at 446 nm. The K _d was determined by fitting the change in fluorescence to a quadratic binding isotherm [Eq. (1)]

$$\mathrm{\Delta }F=\left(\frac{\mathrm{\Delta }{F}_{\max }}{2P_o}\right)\left\{P_o+L_o+K_d-{\left[{\left(P_o+L_o+K_d\right)}^2-4P_o{L}_o\right]}^{\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$2$}\right.}\right\},$$ (1)

where ΔF is the change in fluorescence emission, ΔF _max is the maximum change in fluorescence emission, P _o is the total protein concentration, L _o is the total FMN concentration, and K _d is the dissociation constant for the FMN–protein complex.

Redox Potentiometry

Redox titrations of wild type and the variants of FnFLD were performed in a glove box (Belle Technology, Weymouth, UK) maintained under nitrogen atmosphere with O₂ < 5 ppm. The titration buffer, 50 mM sodium phosphate pH 7.0, was made anaerobic by bubbling it with N₂ for 2 h followed by overnight equilibration in the glove box. Concentrated FnFLD was introduced to the glove box and then gel filtered over a 10 mL size‐exclusion column (Bio‐Rad Econo‐Pac 10 DG column) equilibrated with the anaerobic titration buffer. The protein was diluted using the anaerobic titration buffer to around 50 μM in a total volume of 7 mL. Redox mediators, benzyl viologen (1 μM), phenazine methosulfate (1 μM), and 2‐hydroxy‐1‐4‐napthoquinone (1 μM) were added to the protein solution. The oxidized protein was gradually reduced with the addition of 1‐μL aliquots of sodium dithionite. After each addition of dithionite, the potential was recorded (once the protein solution had equilibrated) and the flavin absorbance spectrum was recorded. The potential was measured using a pH/ORP meter (Orion 3‐Star Benchtop Meter, Thermo Scientific™) equipped with an ORP electrode (Orion 9179 BNMD, Thermo Scientific™) and the flavin absorbance spectra were recorded with a Lambda 265 UV/visible spectrophotometer (Perkin Elmer). Both instruments were housed in the glove box. Titrations were performed to obtain an approximate 20–40 mV change between each data point and were stopped after recording three spectra with no further changes in absorbance at 580 nm. The observed potentials obtained with the Ag/AgCl electrode were normalized to the standard hydrogen electrode by the addition of 204 mV to the potential values. The midpoint potentials for FMN_ox/sq and FMN_sq/hq couples were determined by plotting the absorbance at 580 nm against the normalized potential. The data were fitted to equation for a 2‐electron process [Eq. (2)], which is derived from extension of the Beer–Lambert Law and the Nernst equation.

$$A=\frac{a{10}^{\left(E-E_1^{\text{'}}\right)/59}+b+c{10}^{\left(E_2^{\text{'}}-E\right)/59}}{1+{10}^{\left(E-E_1^{\text{'}}\right)/59}+b+c{10}^{\left(E_2^{\text{'}}-E\right)/59}}$$ (2)

where A is the total absorbance at 580 nm, E is the observed normalized potential. E ₁′ and E ₂′ are the oxidized/semiquinone and semiquinone/hydroquinone midpoint potentials of the flavin. The component absorbance values contributed by the flavin in the oxidized, semiquinone, and hydroquinone states are represented by a, b, and c, respectively.

Free energy calculations

The free energies associated with the binding of the semiquinone and hydroquinone forms of FMN to apoflavodoxin were determined by connecting the redox potentials of the free flavin cofactor with the redox potentials of flavodoxin using a thermodynamic loop. The free energy associated with FMN_sq:ApoFld and FMN_hq:ApoFld was determined with the following equations

$$\mathrm{\Delta }{G}_{\mathrm{sq}}=\mathrm{\Delta }{G}_{\mathrm{ox}}-F\left(E_{\mathrm{ox}/\mathrm{sq}}-E_{\mathrm{ox}/\mathrm{sq}}^{\text{free}}\right)$$ (3)

$$\mathrm{\Delta }{G}_{\mathrm{hq}}=\mathrm{\Delta }{G}_{\mathrm{ox}}-F\left(E_{\mathrm{ox}/\mathrm{sq}}+E_{\mathrm{sq}/\mathrm{hq}}-E_{\mathrm{ox}/\mathrm{sq}}^{\text{free}}-E_{\mathrm{sq}/\mathrm{hq}}^{\text{free}}\right)$$ (4)

where F is the Faraday constant and E _ox/sq ^free and E _sq/hq ^free are the midpoint potentials of the free FMN cofactor.

Crystal growth, data collection, and structure refinement

Crystals of wild‐type FnFld and the K13G variant were obtained by the sitting‐drop vapor diffusion method in 96‐well trays at 4°C with 2 μL drops containing a 1:1 well mixture of reservoir condition and protein. The FnFld crystals used for structure determination were grown from a ~21 mg/mL FnFld solution mixed with a reservoir of 42%–46% PEG 600, 0.16–0.18M calcium acetate, and 0.1M sodium cacodylate. The K13G Fld crystals were grown from a ~8.3 mg/mL protein solution mixed with 21% PEG 3350, 0.23M magnesium chloride, and 0.1M HEPES pH 7.5. For cryoprotection, crystals were soaked in a reservoir solution supplemented with 15% (K13G Fld) or 20% (WT FnFld) glycerol for ~30 s and immersed in liquid nitrogen until data collection. Diffraction data for K13G Fld were collected at the Canadian Light Source beamline 08B1‐1 and processed using XDS to 1.2 Å resolution.54 Diffraction data for wild‐type FnFld were collected on a Rigaku MicroMax 007‐HF generator, VariMax HR optics, and Saturn CCD 944+ detector. Data were processed to 1.85 Å using Mosflm.55 Phase determination by molecular replacement with the K13G Fld dataset was performed using the structure of flavodoxin from Helicobacter pylori (PDB entry code 1FUE) using the AutoSol program in the Phenix suite of programs.56 Model building and refinement were carried out in the Phenix suite and visualized using COOT.56, 57 This structure was used for molecular replacement of the FnFld dataset with model building and refinement performed as for the K13G variant. The full peptide chain (residues M1 to K167) of wild‐type FnFld was included in the model. The K13G variant has an additional three residues at the amino terminus derived from the expression plasmid.

Preparation of figures and electrostatic surfaces

All crystallographic figures were created with PyMOL software (www.pymol.org) and potentials were calculated using the Adaptive Poisson‐Boltzmann Solver plugin implemented in PyMOL using default parameters.58

Conflict of Interest

The authors declare that they have no conflict of interest to declare.

Protein Data Bank Deposition

Atomic coordinates and structure factors for wild‐type FnFld and the K13G variant have been deposited in to RCSB Protein Data Bank with accession codes 6OHL (wild type) and 6OHK (K13G).

Supporting information

Appendix S1: Supporting Information