Báo cáo khoa học: Structure analysis of the flavoredoxin from Desulfovibrio vulgaris Miyazaki F reveals key residues that discriminate the functions and properties of the flavin reductase family pdf

The aim was to elucidate whether flavoredoxin has structural similarity to ferric reductase and ferric reductase activity, based on the sequence similarity to ferric reductase from Archae

Desulfovibrio vulgaris Miyazaki F reveals key residues

that discriminate the functions and properties of the

flavin reductase family

Naoki Shibata1, Yasufumi Ueda1, Daisuke Takeuchi2, Yoshihiro Haruyama2, Shuichi Kojima3, Junichi Sato4, Youichi Niimura4, Masaya Kitamura2and Yoshiki Higuchi1

1 Department of Life Science, University of Hyogo, Japan

2 Department of Applied Chemistry and Bioengineering, Osaka City University, Japan

3 Institute for Biomolecular Science, Gakushuin University, Tokyo, Japan

4 Department of Bioscience, Tokyo University of Agriculture, Japan

Keywords

crystal structure; electron transfer; flavin

mononucleotide; flavin reductase family;

sulfate-reducing bacterium

Correspondence

M Kitamura, Department of Applied

Chemistry and Bioengineering, Graduate

School of Engineering, Osaka City

University, 3-3-138 Sugimoto, Sumiyoshi-ku,

Osaka 558-8585, Japan

Fax: +81 666 05 2769

Tel: +81 666 05 3091

E-mail: kitamura@bioa.eng.osaka-cu.ac.jp

Y Higuchi, Department of Life Science,

Graduate School of Life Science, University

of Hyogo, 3-2-1 Koto, Kamigori-cho,

Ako-gun, Hyogo 678-1297, Japan

Fax: +81 791 58 0177

Tel: +81 791 58 0179

E-mail: hig@sci.u-hyogo.ac.jp

Database

The coordinates and structure factor data

have been deposited in the PDB, under the

accession number 2D5M The nucleotide

and amino acid sequence data may be

found in the DDBJ, EMBL and GenBank

sequence databases under the accession

numbers AB214904 and BAD99043,

respectively

(Received 30 March 2009, revised 28 May

2009, accepted 29 June 2009)

doi:10.1111/j.1742-4658.2009.07184.x

The crystal structure of flavoredoxin from Desulfovibrio vulgaris Miyazaki

F was determined at 1.05 A˚ resolution and its ferric reductase activity was examined The aim was to elucidate whether flavoredoxin has structural similarity to ferric reductase and ferric reductase activity, based on the sequence similarity to ferric reductase from Archaeoglobus fulgidus As expected, flavoredoxin shared a common overall structure with A fulgidus ferric reductase and displayed weak ferric reductase and flavin reductase activities; however, flavoredoxin contains two FMN molecules per dimer, unlike A fulgidus ferric reductase, which has only one FMN molecule per dimer Compared with A fulgidus ferric reductase, flavoredoxin forms three additional hydrogen bonds and has a significantly smaller solvent-accessible surface area These observations explain the higher affinity of flavoredoxin for FMN Unexpectedly, an electron-density map indicated the presence of

a Mes molecule on the re-side of the isoalloxazine ring of FMN, and that two zinc ions are bound to the two cysteine residues, Cys39 and Cys40, adjacent to FMN These two cysteine residues are close to one of the puta-tive ferric ion binding sites of ferric reductase Based on their structural similarities, we conclude that the corresponding site of ferric reductase is the most plausible site for ferric ion binding Comparing the structures with related flavin proteins revealed key structural features regarding the discrimination of function (ferric ion or flavin reduction) and a unique electron transport system

Abbreviations

DvMF, Desulfovibrio vulgaris Miyazaki F; FeR, ferric reductase; Fre, flavin reductase; PDB, Protein Data Bank.

Flavins play major roles as cofactors for a wide variety

of redox proteins and enzymes; these reactions depend

on the redox ability of the flavin species Although the

basic redox reactions are identical or similar, it is of

interest to understand the molecular bases for the

dif-ferent reactivities displayed by flavins in difdif-ferent

pro-tein contexts Flavoredoxin is an electron-transfer

protein that has one FMN molecule per subunit or

monomer [1] The only flavoredoxins characterized to

date are from the sulfate-reducing bacterium

Desulfo-vibrio gigas [1–3] and an Archaeon Methanosarcina

acetivorans [4] Deletion and mutation analyses of this

bacterium have indicated that flavoredoxin is involved

in the thiosulfate reduction process [3] It has been

proposed that flavoredoxin receives an electron, which

was originally generated by a hydrogenase, from

flavodoxin or ferredoxin, and transfers it to a sulfite

reductase, desulfoviridin [3]

The D gigas flavoredoxin has an apparent sequence

identity of 22% with Archaeoglobus fulgidus ferric

reductase (FeR) [1] FeR catalyzes the reduction of

Fe(III) chelates, such as Fe(III)–EDTA, in a

NAD(P)H-dependent manner [5,6] The crystal structure of FeR

has been determined with and without NADP+ [6]

These authors were unsuccessful in their attempts to

solve the structure in the Fe(III) ion-bound state The

catalytic mechanism of ferric ion reduction by this

enzyme has been proposed based on biochemical [5] and

structural studies of FeR [6], although the key residues

for ferric ion binding need to be identified to elucidate

the complete reaction mechanism It was surprising that,

as pointed out by Chiu et al [6], FeR revealed a

com-mon overall fold with the FMN-binding protein from

Desulfovibrio vulgarisMiyazaki F (DvMF), whose

crys-tal structure was determined by our group [7] However,

FeR and the FNM-binding protein from have relatively

low sequence identity (12%) FeR also has structural

similarity to the flavin reductase (Fre, NADH : flavin

oxidoreductase) family, which includes the Fre

compo-nent of the two-compocompo-nent flavin-diffusible

monooxy-genase [5,6] The Fre component reduces a flavin with

NADH or NADPH to provide a reduced flavin, which

is used to activate molecular oxygen for the oxygenase

reaction [8] In this family, neither component of the

enzyme binds flavin tightly as a cofactor, but rather

utilizes it as another substrate [8]

Considering the amino acid sequence similarities

between flavoredoxin and related flavin proteins, the

question arises as to whether flavoredoxin possesses

ferric ion or flavin reductase activity Which structures

determine the unique functions of these flavin proteins?

In this study, we present the crystal structure of DvMF flavoredoxin and discuss the key residues for ligand binding and metal ion binding, based on the crystal structures

Results

Cloning and sequencing of the flavoredoxin gene

We determined the nucleotide sequence of the entire flavoredoxin gene (accession number AB214904 in the DDBJ, EMBL and GenBank nucleotide databases) The ORF that encodes flavoredoxin comprises 190 amino acid residues A potential ribosome-binding site (GAGG, nucleotides 737–740 in the PstI–KpnI frag-ment) is present upstream of the initiation codon (ATG), and there are potential promoter regions at nucleotides 643–648 (TTGCCG) and 666–671 (CAA-ACT) in the PstI–KpnI fragment Nucleotides 1339–

1371 comprise the putative transcriptional terminator, forming a stem-and-loop structure The results of a BLAST homology search indicate that the product of this ORF is highly homologous to flavoredoxins from other bacteria, especially that of D vulgaris (Hilden-borough), with an identity of 71%; therefore, we confirmed this ORF to be the flavoredoxin gene

Recombinant flavoredoxin purification

We used an Escherichia coli expression system to express the flavoredoxin gene Recombinant flavo-redoxin was detected in transformed E coli crude cell extracts by SDS⁄ PAGE Through chromatographic steps using DE52 and Superdex 75, a large amount of flavoredoxin was purified to homogeneity by SDS⁄ PAGE (Fig S1) The molecular mass of the expressed flavoredoxin was estimated to be

 23 000 Da by SDS ⁄ PAGE, which is different from the value calculated based on the amino acid sequence (20 800 Da) We also estimated the molecular mass in the native state to be  37 000 Da using a Superdex

75 gel-filtration column This value is about twice that calculated from the amino acid sequence, indicating that the native form of flavoredoxin is a dimer

Amino acid sequence analysis of flavoredoxin The N-terminal amino acid sequence of flavoredoxin from DvMF was determined to be Met–Lys–Lys–Ser– Leu–Gly–Ala, and the Met was formylated When the flavoredoxin amino acid sequences of DvMF and other

organisms were compared, they were found to be

highly conserved The three characteristic

co-ordina-tion motifs (36TSKP–62FGVSVL–124GTHTL) of the

FeR from A fulgidus, which is linked to FMN or

NAD binding [5], were also found in DvMF

flavore-doxin (40CSQP–66FTISIP–128GLHTQ) These

co-ordi-nation motifs are not homologous to those of

flavodoxin or FMN-binding protein

Identification of the prosthetic group

To identify the prosthetic group bound to the

recombi-nant flavoredoxin, UV-visible spectra of the purified

holoprotein were recorded (Fig S2) In the visible

region, absorption maxima were observed at 381 and

452 nm, which are characteristic of proteins that bind

to flavin derivatives The recombinant flavoredoxin

was subjected to reverse-phase HPLC on a C8 column,

and the retention time of the obtained prosthetic group

was compared with those of flavin derivatives The

retention time of the prosthetic group bound to

recom-binant flavoredoxin was identical to that of FMN

(Fig S3) The A448: A268ratio of the holoprotein was

0.267, suggesting that the flavoredoxin expressed in

E colias a holoprotein binds to FMN as a prosthetic

group at a molar ratio of 1

Overall structure of DvMF flavoredoxin

DvMF flavoredoxin was crystallized in the P3121 space

group with one molecule in the asymmetric unit The

structure was refined to a crystallographic R factor of

0.135 and Rfreeof 0.162 at 1.05 A˚ resolution (Table 1)

Residues 128–130 and 187–190 (four C-terminal

resi-dues) were excluded from the structural model because

of poor electron densities in these regions DvMF

fla-voredoxin contains four a helices (a1–4), two 310

heli-ces (3101–2) and 12 b strands (b1–12) as secondary

structural elements; it also has a Greek key motif with

seven anti-parallel b strands (Figs 1 and 2A), which is

also found in DvMF FMN-binding protein [7] and

A fulgidus FeR [6] Flavoredoxin contains two FMN

molecules per dimer, unlike FeR, which has only one

FMN molecule per dimer (Fig 2B) The FMN

mole-cule is located in the hollow, encompassed mainly by

a1, a2 and b3

A structural homology search was carried out using

the DALI server [9] Among the proteins of known

function, the M acetivorans flavoredoxin exhibited

the lowest rmsd (1.5 A˚) and the highest Z score

(25.5), as expected from the highest sequence identity

(30%) of the known structures A fulgidus FeR

showed the second lowest rmsd (1.9 A˚) and the

second highest Z score (18.8), although the sequence identity between DvMF flavoredoxin and A fulgidus FeR is low (17%) Among the flavin-containing electron-transfer proteins of Desulfovibrio species, the structures of the FMN-binding protein and flavodoxin were determined by X-ray crystallography; the structure of DvMF flavoredoxin resembles the former (rmsd = 2.5 A˚) rather than the latter (rmsd = 3.4 A˚)

As deduced by gel-filtration chromatography, DvMF flavoredoxin forms a dimer, as evidenced by the crystallographic two-fold axis in the crystal When the dimeric structure of DvMF flavoredoxin was com-pared with those of FeR and FMN-binding protein, the flavoredoxin dimer was superimposed on the former (Fig 2B) but not on the latter (Fig 2C) In flavoredoxin, the twofold axis associated with the dimer passes through the vicinity of the side chains of Pro13, Ile126, Gln133 and Ile163 The corresponding

Table 1 Summary of x-ray data collection, phasing and refinement statistics.

Native

Methylmercuric chloride derivative Data collection

Unit-cell parameters (A ˚ )

a = b = 53.35,

c = 116.22

a = b = 53.5,

c = 116.2

Resolution range (A ˚ ) 50–1.05 (1.09–1.05) 50–1.71 (1.77–1.71) Measured reflections 950,955 205,263

Completeness (%) 99.6 (97.0) 99.6 (97.6)

Rmerge 0.088 (0.573) 0.087 (0.158)

SAD phasing for methylmercuric chloride derivative Figure of merit,

centric ⁄ acentric

Refinement Resolution range (A ˚ ) 10–1.05 (1.09–1.05)

-R.m.s deviations from ideal values Bond lengths (A ˚ ) ⁄

angle distances ()

0.016 ⁄ 0.031 -Ramachandran plot

B

C

Fig 2 Structures of flavoredoxin and other

related proteins (A) Overall structure of the

flavoredoxin dimer Each subunit is shown

in green–cyan and violet–red models FMN

molecules are depicted as ball-and-stick

models (B) Superimposed Ca-traces of

flavoredoxin (green and violet) and FeR (light

gray) (C) Superimposed Ca-traces of

flavoredoxin (green and violet) and

FMN-binding protein (light gray).

Fig 1 Amino acid sequence alignment of

DvMF flavoredoxin, Methanosarcina

acetivo-rans flavoredoxin, ferric reductase and HpaC

component of Escherichia coli

4-hydroxyph-enylacetate 3-monooxygenase Secondary

structure elements of flavoredoxin are

shown on the lines of residue numbers.

Residues involved in binding of FMN are

shown in red Residues shown in bold are

aligned based on crystal structures

Resi-dues shown in regular characters indicate

that structural information is unavailable or

that structurally equivalent residues are not

present Alignment for HpaC was performed

with CLUSTAL W [48].

residues in the FMN-binding protein are exposed to

the solvent

In terms of dimer interactions, both N- and

C-termi-nal loops (Met1–Pro13 and Val174–Lys186,

respec-tively) appear to play important roles The six

N-terminal residues (Met1–Gly6) are extended in the

opposite direction along the b10 of the other

mono-mer, forming an anti-parallel intermolecular b sheet

(Fig 2A) The subsequent residues of the loop (Ala7–

Pro13) turn into the interior of the dimer interface,

and Leu10 and Tyr12 form a hydrophobic core with

Ile70, Met116, Val141, Pro154, Ile156 and Pro161 of

the other monomer Similarly, Gly175–Ala181 forms

an anti-parallel intermolecular b sheet with the b12 of

the other monomer, and towards the C-terminus, the

subsequent residues, Phe182–Lys186, pass through the

a2 vicinity of the other monomer For the FeR dimer,

the N-terminal loop is replaced by an a helix, and side

chain-to-side chain interactions play a major role in

dimer interactions through this region By contrast, a

similar intermolecular b sheet through the C-terminal

loop is conserved

Structure of the FMN-binding region

The hydrogen bonds and salt bridge that encompass

the ribitol moiety and the phosphate group of

FeR and M acetivorans flavoredoxin are moderately

and completely conserved in DvMF flavoredoxin

(Fig 3A,B and Fig S4A), respectively In the case of

FeR, Ser84 replaces Asn29 of DvMF flavoredoxin,

which forms a hydrogen bond to the O3P atom of

FMN (Fig 3A,B) DvMF flavoredoxin forms three

additional hydrogen bonds between the NH2 moiety

of Arg51 and three atoms of FMN (N1, O2 and O3’)

In both FeR and M acetivorans flavoredoxin, the

cor-responding residue is asparagine (Asn47 and Asn52,

respectively) to which only the O2 atom of FMN

forms a hydrogen bond (Figs 3A,B and Fig S4A)

The isoalloxazine ring of FMN is surrounded by

hydrophobic residues, Leu16, Trp35, Ile84, Phe164,

Tyr171 and Phe182, the first five residues of which

correspond to Leu13, Thr31, Phe81, Tyr147 and

Tyr150 in FeR (Fig 3A,B), and Val18, Trp36, Leu85,

Leu162, Tyr169 and Leu180 in M acetivorans

flavo-redoxin (Fig S4A)

It should be noted that a positively charged residue,

Lys92, is involved in the binding of the phosphate

group(s) of FMN or FAD Lys92 forms a salt bridge

with the O3P of FMN A salt bridge that involves

fla-vin species has not been reported in the structures of

the other electron-transfer flavoproteins; however, a

salt bridge between the lysine⁄ arginine residue and

FMN⁄ FAD is found frequently in flavin-dependent enzymes To date, from the 130 FMN protein PDB entries 22 have at least one FMN–lysine and 62 have

at least one FMN–arginine interaction For FAD proteins, from the 210 entries 9 have at least one FAD–lysine and 55 have at least one FAD–arginine interaction One of these proteins, FeR, has a Lys89 residue that interacts with the phosphate group of FMN As pointed out by Chiu et al [6], the structure

of FeR resembles the flavin-binding domain of ferre-doxin : NADP+ reductase [10] The lysine residue is not conserved; instead, an arginine residue interacts with the FAD molecule of the enzyme In the case of FMN-binding protein, Lys53 is adjacent to FMN, forming a hydrogen bond with the phosphate group of FMN through its main-chain N atom, whereas the side chain amino group is 7 A˚ from the FMN molecule The surface charge models of these proteins indicate that the level of positive charge at the phosphate group binding site is considerably higher in both flavoredoxin and FeR than in the FMN-binding protein (Fig S5A–C)

The accessible surface areas of FMN in flavoredoxin and FeR have been calculated to be 57 and 95 A˚2, respectively Three residues, Trp35, Arg51 and Phe182, are responsible for this difference (Fig 3A,B) Both Trp35 and Arg51 of flavoredoxin have larger volumes than the corresponding residues (Thr31 and Asn47) of FeR No residue in FeR corresponds to Phe182 (Fig 3A,B) The surface model of flavoredoxin indi-cates that the re-side of the isoalloxazine ring is par-tially covered by these residues (Fig S5A) By contrast, the re-side of the isoalloxazine ring of FeR is completely exposed to the solvent (Fig S5B)

Resolution of the structure of NADP+-bound FeR revealed that the nicotinamide moiety of NADP+ faces the re-side of the isoalloxazine ring, and that the 2¢-P-AMP moiety is held in the groove between the 310 helix and the third a helix [6] Unexpectedly, in fla-voredoxin, this site is occupied by Mes, which was added to crystallization buffer solution The Mes mole-cule is held in place through a salt bridge with Arg169, hydrogen bonds with Thr9 and Val167, and hydropho-bic interactions with Trp35 and FMN (Fig 3C)

Structure of the metal ion binding site The electron-density map displayed three isolated spheres with significantly greater density than normal water oxygen atoms Two of these are close to the FMN binding site, and the other is on the opposite surface of the protein An anomalous-difference map calculated from the native dataset showed significant

peaks at each of these sites These densities were

assigned as zinc ions derived from the crystallization

solution as an additive The zinc ion closest to FMN,

Zn201, is coordinated by the Sc atom of one of the two conformers of Cys40 and two water molecules (Fig 3D) Zn203, which is 5.1 A˚ from Zn201, is also

A

B

C

D

Fig 3 FMN binding site (A,B)

Superim-posed models of FMN binding sites of

fla-voredoxin and FeR, showing residues that

interact with the FMN molecule Hydrogen

bonds are shown as green dotted lines

Fla-voredoxin is displayed by atom color, with

the exception that the carbon atoms of

FMN are in yellow FeR is shown as a

trans-parent model The Mes molecule is omitted

for clarity The view in (B) is rotated 180

about the vertical axis Residue labels are

shown as flavoredoxin ⁄ FeR (C) Residues of

flavoredoxin involved in Mes binding.

NADP+derived from the superimposed

model of FeR is also shown as a

transpar-ent model (D) Zinc ion-binding site Zinc

ions are depicted as green spheres The

remaining color codes are the same as in

(A) and (B).

coordinated by Cys39 and the other conformer of

Cys40 This zinc ion is bound to His131, which

corre-sponds to the histidine residue that is completely

con-served among the FeR homologs [5,6] Interestingly,

the residues corresponding to metal ion binding are

totally different in both FeR and M acetivorans

fla-voredoxin These cysteine residues are replaced by

threonine and leucine in FeR and asparagine and

valine in M acetivorans flavoredoxin (Figs 1 and 3D;

Fig S4C) In the case of FeR, however, Cys45 is

adja-cent to this site instead (Fig 3D)

Redox potential of recombinant flavoredoxin

Figure 4 shows the results of linear regression analysis

of the logarithms for the redox ratio of the mediator

versus that of recombinant flavoredoxin The redox

potential of oxidized flavoredoxin⁄ reduced

flavoredox-in (Eflr) was calculated as )343 mV at pH 7.0,

deter-mined using Neutral Red (Em,7=)325 mV, n = 2)

[11] or benzyl viologen (Em,7=)359 mV, n = 1) [12]

An n value of 2 was used in these experiments, which

fit the experimental data closely Although

recombi-nant flavoredoxin was fully reduced by sodium

dithio-nite, no semiquinone intermediate was found

Flavoredoxin reduction by NAD(P)H

Reduction experiments were performed under

anaero-bic conditions substituted by oxygen-free argon, and

NADH or NADPH was used as the reductant The

observed increase in absorption around 340 nm is caused by the addition of NADH or NADPH Based

on the amino acid comparison (Fig 1) and the crystal structure of flavoredoxin, we propose that flavoredoxin has a NAD(P)H binding site similar to that of A fulgi-dus FeR, and therefore we expect that NAD(P)H is bound to this site and reduces the FMN of flavo-redoxin; however, no decrease in absorption in the visible region because of flavoredoxin reduction was observed, even when NADH or NADPH was added (data not shown); therefore, we conclude that flavore-doxin has no oxidase activity

Ferric reductase and flavin reductase activities of recombinant flavoredoxin and related proteins Flavoredoxin uses both NADH and NADPH as elec-tron donors to reduce Fe3+–EDTA and FMN; how-ever, we found that both FeR and Fre activities using NADH were lower than those using NADPH (data not shown) FeR and Fre activities of flavoredoxin were 2.58· 10)3 and 2.70· 10)3 unitsÆmgÆprotein)1, respectively, whereas those of DrgA from Synechocystis

sp PCC6803, which was used as a positive control, were 8.66· 10)2 and 3.22 unitsÆmgÆprotein)1 under aerobic conditions, respectively [13] Even though both activities of flavoredoxin could be detected, FeR and Fre activities of flavoredoxin were 33.6-fold and 1200-fold lower than those of DrgA, respectively However, FMN-binding protein [14] showed neither FeR nor Fre activity

Discussion

Based on the high structural similarity between DvMF flavoredoxin and A fulgidus FeR, we expected that the flavoredoxin would have FeR and Fre activities, both

of which confer reduction of a flavin by receiving elec-trons from NADH or NADPH; however, these activi-ties of flavoredoxin were much lower than those of

A fulgidusFeR The FeR activity of A fulgidus, which

is also a sulfate reducer, was reported to be 3503 unitsÆmgÆprotein)1 at 85C using NADPH and FMN [5] This value is  1.36 · 106-fold higher than that of DvMF flavoredoxin DvMF FMN-binding protein, which also shares structural similarity with A fulgidus FeR, did not exhibit any detectable FeR activity It should be noted that the FeR activity of the A fulgidus enzyme is at least 1000-fold higher than all other bacterial enzymes [5] In addition, the FeR activity

of A fulgidus enzyme was  1070-fold higher than DrgA, which we used as a positive control, even under anaerobic conditions (3.28 unitsÆmgÆprotein)1) [13]

Fig 4 Linear regression analysis of the logarithms of concentration

ratios for oxidized and reduced forms of a mediator versus that of

recombinant flavoredoxin () Neutral Red used as the mediator; (D)

Benzyl viologen is used as the mediator.

Thus, it appears that only A fulgidus FeR possesses

extremely high FeR activity, compared with other

FeRs The FeR activity of flavoredoxin is 102–104times

lower than that of other bacterial FeR-active enzymes

Taking this into account, we postulate that DvMF

fla-voredoxin reacts similarly to FeR, although its activity

is not comparable with that of A fulgidus FeR

The same may be true of flavin reductase activity

Two explanations can be proposed for the large

differ-ences noted in ferric ion and flavin reductase activities

between DvMF flavoredoxin and A fulgidas FeR The

first possibility is that NAD(P)H binds to flavoredoxin,

but hardly transfers an electron to FMN The fact that

DvMF flavoredoxin has a lower redox potential than

NADH ()343 versus )320 mV) supports this

explana-tion The second possibility is that the binding of

NAD(P)H is sterically hindered by the surrounding

residues Regarding this second possibility, when the

flavoredoxin–Mes complex is superimposed on the

FeR–NADP+ complex, steric hindrances occur

between flavoredoxin and NADP+ at three different

sites (Fig 3C) First, the adenine ring of NADP+has

severe steric contacts with Pro166, Val167 and Ser168,

which comprise the loop between b11 and b12

Sec-ond, the phosphate and ribose groups overlap with the

phenyl group of Phe182 Third, the nicotinamide

moi-ety is immediately adjacent to Trp35 At the first site,

the adenine ring has to move away from these

overlap-ping residues to bind to the corresponding site of

fla-voredoxin At the second site, such steric repulsions

could be relieved by displacement of the residues

and⁄ or a modified configuration of NADP+, because

most Phe182 is exposed to solvent and the side chain

rotates freely about its Ca–Cb and Cb–Cc bonds The

steric contacts of Trp35 at the third site are not severe

Chiu et al [6] have suggested that the dimethylbenzene

moiety is a candidate for ferric ion binding If this is

the case, flavoredoxin cannot bind a ferric ion at this

site unless Phe182 moves away from the site Then,

because NAD(P)H was not the most suitable electron

donor for DvMF flavoredoxin, the FeR and flavin

reductase activities of DvMF flavoredoxin might have

been underestimated, and we cannot ignore that the

physiologic function of flavoredoxin is to transfer an

electron from reduced ferredoxin or flavodoxin to a

sulfite reductase, i.e desulfoviridin in the thiosulfate

reduction process [3]

It has been reported that A fulgidus FeR activity

decreases during purification because of the loss of a

flavin molecule, and is restored by adding FMN or

FAD, which suggests that the affinities for flavin

species are not sufficiently high In D gigas

flavore-doxin, however, FMN binds tightly with a dissociation

constant of 0.12 nm [1], i.e 2500-fold lower than the

Kmvalue (0.3 lm) for FMN of FeR [5] This large dif-ference can be explained in part by the difdif-ference in the exposed surface areas of FMN, as described above Among the NADH : flavin oxidoreductase family, one structure, the HpaC component of Sulfolobus tokodaii 4-hydroxyphenylacetate 3-monooxygenase, which is a member of the two-component flavin-diffusible mono-oxygenases, in the FMN-bound form has been reported [15] The Kmof S tokodaii HpaC for FMN is unknown, but that of E coli HpaC is reported to be 2.1 lm [16], which is comparable with the values obtained for other NADH : flavin oxidoreductases [16–23] The accessible surface area of FMN of

S tokodaii HpaC was calculated to be 152 A˚2 As expected, this area is significantly larger than those of flavoredoxin (57 A˚2) and FeR (95 A˚2) The large acces-sible surface area for FMN of HpaC is quite reason-able considering that HpaC releases the FMN molecule immediately upon reduction, with which the oxygenase component catalyzes the oxygenation reaction

A fulgidusFeR is reported to be able to use FMN, but not riboflavin, as the electron acceptor, although most FeR molecules can use both For example, the

Km and kcat values of E coli Fre for riboflavin are reported as 2.5 lm and 52.4 s)1 with NADPH as the electron donor and 1.3 lm and 30.6 s)1 with NADH [13] For FMN, Kmand kcatvalues of E coli Fre were calculated as 2.67 lm and 0.724 s)1 with NADPH as the electron donor and 0.67 lm and 1.84 s)1 with NADH (our unpublished data) These differences may influence FeR and Fre activities

Chiu et al [6] claimed that His126 of A fulgidus FeR is one of the candidates for the ferric ion-binding residue based on the structure of the mercury-bound derivative, although a ferric ion was not found at the site of either the native or NADP+-bound form; instead this residue is bound to the nicotinamide moi-ety of NADP+ in the NADP+-bound form These authors also proposed that Cys45 of FeR, which corre-sponds to Ser49 of DvMF flavoredoxin, is another candidate [6] The Oc atom of Ser49 does not bind to

a zinc ion but is only 5.68 and 5.50 A˚ from Zn201 and Zn203, respectively (Fig 3D) In FeR, Cys39 and Cys40 are replaced by leucine and threonine, and Cys45 rather than threonine would be the ligand for a ferric ion, because the hydroxyl group of threonine has

a lower affinity for ferric ions than the thiol group of cysteine DrgA, which was used as a positive control

in activity measurements, has a cysteine (Cys147) and three histidine (His15, His20 and His45) residues, which could be candidates for ferric ion-binding

residues However, DrgA seems to have a different

folding from flavoredoxin and FeR, as the secondary

structure prediction by PSIPRED [24] suggests that

DrgA has a helix-rich structure (data not shown),

unlike the b-rich structures of flavoredoxin and FeR

Structural analysis needs to be carried out to elucidate

the ferric ion-binding site of DrgA Suharti et al [4]

have recently shown that M acetivorans flavoredoxin

does not transfer an electron to ferric and chelated

ferric ion, and does not have a cystein residue at the

metal-binding site The corresponding region is formed

by Val40, Asn41, Gly50 and Phe127 (Fig S4C), which

are unlikely to have high affinity to ferric ions It

should be noted that E coli HpaC displayed FeR

activity [16] Sequence alignment indicates that it has a

cysteine residue at the position that corresponds to

Cys45 (Fig 1); therefore, flavin proteins, which have

similar folding and ferric reductase activity, have at

least one cysteine residue at or around the site

corre-sponding to the zinc-binding site of flavoredoxin

These lines of evidence raise questions as to why a zinc

ion binds robustly to DvMF flavoredoxin and why no

metal ions were found in the A fulgidus FeR structure

even in a ferric ion atmosphere FeR may have low

affinity for ferric ions in the oxidized state As Chiu

et al [6] have suggested, the solvent-exposed

dimethyl-benzene moiety is one candidate for binding the ferric

ion In this case, the ferric ion would bind only weakly

to FeR, because none of the residues that could act as

a ligand for the ferric ion are found at this position A

zinc ion is preferentially trapped at the metal ion

bind-ing site of DvMF flavoredoxin Indeed, the ferric

reductases of Azotobacter vinelandii [25] and

Legionella pneumophila [26] are inhibited in the

pres-ence of zinc ions This is probably because a zinc ion

is bound tightly at the ferric ion binding site, although

the detailed inhibition mechanism remains unknown

In the case of A vinelandii ferric reductase, the enzyme

was purified as a mixture of two proteins or a

hetero-dimer with molecular masses of 44 600 and 69 000 Da

[25] Although none of the amino acid sequences of

A vinelandii is recorded as ferric reductase, an NCBI

entry, ‘Oxidoreductase FAD⁄ NAD(P)-binding:

Oxido-reductase FAD-binding region’ (ZP_00418100), might

be the amino acid sequence of the smaller component

of A vinelandii ferric reductase This entry contains

443 residues with a calculated molecular mass of

49 180 Da and has apparent identity with the FAD

and NADH domains of BenC, which is the reductase

component of benzoate dioxygenase reductase The

crystal structure of BenC [27] shows that Cys307,

which corresponds to Cys412 of the entry, is only 6 A˚

from the isoalloxazine ring of FAD, which is

compara-ble with flavoredoxin and FeR If the entry corre-sponds to the smaller component of the A vinelandii ferric reductase, Cys412 could be the ferric ion binding site

Experimental procedures

Cloning and sequencing of the flavoredoxin gene

E coli JM109 (recA1, supE44, endA1, hsdR17, gryA96, relA1, thi, D(lac-proAB), F’[traD36, proAB+, lacIq, lacZDM15]) was used for cloning and expression of the fla-voredoxin gene DvMF was grown [28] and used for geno-mic DNA preparation Restriction and modification enzymes were purchased from New England BioLabs (Pic-kering, Ontario, Canada), Nippon Gene (Tokyo, Japan) and Toyobo (Osaka, Japan) The [32P]ATP[cP] (185 TBqÆm-mol)1) was obtained from MP Biomedicals (Irvine, CA, USA) All other chemicals were of analytic grade for bio-chemical use Genomic DNA isolated from DvMF was pre-pared by the method of Saito & Miura [29] To amplify the flavoredoxin gene, we searched for the published amino acid sequences of flavoredoxin from other bacteria, because the DvMF flavoredoxin amino acid sequence was unknown, but we could not find the PCR conditions; how-ever, we noted that the ABC transporter gene, the amino acid sequence of which is highly conserved across species, was in the complementary strand upstream of the flavore-doxin gene in sulfate-reducing bacteria Assuming that gene mapping is similar among sulfate-reducing bacteria, we designed two primers according to the conserved regions of the amino acid sequence of the ABC transporter gene from D vulgaris (Hildenborough), and amplified part

of the ABC transporter gene using PCR with DvMF geno-mic DNA as a template The PCR primer sequences were

as follows: ABC01, 5¢-TGGATAGCCGCCAAGATGGG GTT-3¢ (23-mer corresponding to the amino acid sequence

58W–I–A–A–K–M–G–F); and ABC02, 5¢-GCGAGCGG GGCCGAATCGTAGAA-3¢ (23-mer complementary sequence to the corresponding amino acid sequence

163F–Y–D–S–A–P–L–A) The PCR products were separated

by agarose gel electrophoresis and a fragment of 340 bp was extracted using the MinElute Gel Extraction Kit (Qia-gen, Venlo, Netherlands) The nucleotide sequence of this fragment revealed a putative amino acid sequence that was similar to the amino acid sequence of another ABC trans-porter from sulfate-reducing bacteria We then synthesized five primers, which were used to determine the sequence upstream of the ABC transporter gene using genomic DNA

as the template The primer sequences were as follows: ABC04, 5¢-CCAGCTTCACCTTGCCCTTC-3¢ (20-mer); ABC05, 5¢-CTTGTCCACGTAGGCGAAGG-3¢ (20-mer); Flr05, 5¢-TCTCGTGGGCACATACGACC-3¢ (20-mer); Flr06, 5¢-TCAACAAGGTGGATCCGGTG-3¢ (20-mer); and Flr07,

5¢-ACGTGAAGGTGGACGAATCC-3¢ (20-mer) We

identi-fied the flavoredoxin gene in the complementary strand

upstream of the ABC transporter, and then designed a

30-mer probe DNA (5¢-TCGGAGGTACCGCGCTGCAC

GCCCAGCTTC-3¢), which is a complementary sequence

corresponding the amino acid sequence133V–K–L–G–V–Q–

R–G–T–S–E We carried out Southern hybridization with

this labeled oligonucleotide at 65C and detected a band

that hybridized to an 4.3-kb PstI–KpnI fragment using a

Bioimaging Analyzer (BAS1000; Fuji, Tokyo, Japan) Then,

we digested the genomic DNA with PstI and KpnI, and

fractionated the products on an agarose gel The separated

fragments were ligated into the corresponding sites in

pUC18, and the resultant ligation mixture was transformed

into E coli JM109 One such transformant, isolated by the

colony hybridization method, was found to harbor a

plasmid that carried the  4.3-kbp PstI–KpnI fragment of

DvMF DNA, and was designated pABCFLR The

nucleo-tide sequence of the inserted fragment was determined by

sequencing the restriction fragment that was cloned into

pUC18 using the dideoxy chain termination method [30]

with a DNA sequencer

Expression and purification of recombinant

flavoredoxin

For high-level expression in E coli, we used pMK2 [31],

which is an expression vector that carries the tac promoter

The coding region of the flavoredoxin gene was amplified

by PCR with KOD-Plus- DNA polymerase (Toyobo)

DvMF genomic DNA was used as the PCR template The

PCR primer sequences were as follows: flr09,

5¢-CGACCCGGGTCATGAAGAAATCCCTGG-3¢ (27-mer);

and flr10, 5¢-TTTGTCGACTGATCAGGAGCGCAGGC

C-3¢ (27-mer) PCR was carried out at 94 C for 2 min,

fol-lowed by 35 cycles of 94C for 15 s, 55 C for 30 s, and

68C for 1 min The PCR products were digested with

SmaI and SalI and ligated into similarly digested pUC18

The cloned fragment was confirmed by sequencing, digested

with SmaI and HindIII, and then ligated into pMK2, which

was initially digested with EcoRI, blunt-ended with the

Klenow fragment and then digested with HindIII to give

the expression vector pMKFLR9-10 E coli was

trans-formed with pMKFLR9-10, and transformants were grown

in 1.7 mL LB medium containing 100 lgÆmL)1 ampicillin

for 9 h at 37C Twelve flasks containing 167 mL of the

same medium were inoculated with 1.7 mL culture and

incubated overnight with shaking at 37C Cells were

harvested by centrifugation at 4000 g for 10 min at 4C

The cell pellet was resuspended in 10 mm Tris⁄ HCl buffer

(pH 8.0), sonicated using a Model 201M sonicator

(Kubot-a, Tokyo, Japan) at 9000 Hz and 200 W for 10 min, then

ultracentrifuged at 100 000 g for 2 h at 4C The

superna-tant was then dialyzed against distilled water overnight at

4C

For flavoredoxin purification, the dialysate was loaded onto a DEAE-cellulose column (DE52, 2.2· 15 cm) equili-brated with 10 mm Tris⁄ HCl (pH 8.0) The column was washed with 150 mL of 100 mm NaCl and 10 mm Tris⁄ HCl (pH 8.0) Flavoredoxin was eluted with 200 mL of 300 mm NaCl and 10 mm Tris⁄ HCl (pH 8.0) The colored eluent was dialyzed against distilled water overnight at 4C, and then reloaded onto a DE52 column equilibrated with

10 mm Tris⁄ HCl (pH 8.0) Flavoredoxin was eluted with a linear gradient of 100–300 mm NaCl in 10 mm Tris⁄ HCl (pH 8.0) in a total volume of 300 mL The flavoredoxin-containing fractions were identified based on absorbance at

448 nm The colored fractions were collected and dialyzed against distilled water, and then lyophilized or concentrated using Vivaspin (MW 5000 cut-off; Sartorius AG, Go¨ttin-gen, Germany) Gel filtration on a Superdex 75 HR10⁄ 30 column was carried out using the Pharmacia FPLC system (Uppsala, Sweden) in 200 mm NaCl, 10 mm Tris⁄ HCl (pH 8.0) at a flow rate of 0.5 mLÆmin)1, and the purified recom-binant flavoredoxin was eluted after 23 min SDS⁄ PAGE (15%) was carried out according to the method of Laemmli [32] For amino acid sequence analysis, further purification was performed The purified recombinant flavoredoxin was loaded onto a TSK-GEL TMS-250 (Tosoh, Tokyo, Japan) The column was washed with 0.1% trifluoroacetic acid, and then developed with a gradient of 0–80% acetonitrile in 0.1% trifluoroacetic acid The flow rate was 0.8 mLÆmin)1 Purified apo-flavoredoxin was separated by SDS⁄ PAGE and electroblotted For deformylation, a sample was treated with 0.6 m HCl overnight at room temperature The total protein concentration was determined using the BCA pro-tein assay kit (Thermo Scientific Pierce, Waltham, MA, USA) or calculated from the absorbance value at 448 nm

Identification of the prosthetic group

To identify the prosthetic group, purified recombinant fla-voredoxin and a mixture of riboflavin, FAD and FMN were loaded onto an HPLC C8column (NUCLEOSIL 10

C8; Shinwa Chemical Industries Ltd., Kyoto, Japan) The column was washed with 0.1% trifluoroacetic acid, and then developed with a gradient of 0–20% acetonitrile in 0.1% trifluoroacetic acid at a flow rate of 0.8 mLÆmin)1

Crystallization, data collection and processing The details of crystallization and data collection have been reported previously [33] The purified protein solution was concentrated to 25 mgÆmL)1by centrifugation using a Viva-spin (Mr5000 cut-off; Sartorius) Flavoredoxin was crystal-lized using the sitting-drop vapor diffusion method Protein droplets were prepared by mixing 2 lL protein solution with 2 lL reservoir solution and equilibrated against

100 lL reservoir solution containing 10% (w⁄ v) poly(ethyl-ene gycol)8000, 0.2 m zinc acetate and 100 mm Mes (pH