ORIGINAL ARTICLE

Maddalena Cross Æ Zhiguang Xiao Æ Estelle M. MaesRoman S. Czernuszewicz Æ Simon C. DrewJohn R. Pilbrow Æ Graham N. George

Anthony G. Wedd

Removal of a cysteine ligand from rubredoxin: assemblyof Fe2S2 and Fe(S-Cys)3(OH) centres

Received: 11 December 2001 /Accepted: 8 February 2002 / Published online: 15 March 2002� SBIC 2002

Abstract The electron transfer protein rubredoxin fromClostridium pasteurianum contains an Fe(S-Cys)4 activesite. Mutant proteins C9G, C9A, C42G and C42A, inwhich cysteine ligands are replaced by non-ligating Glyor Ala residues, have been expressed in Escherichia coli.The C42A protein expresses with a FeIII2 S2 cluster inplace. In contrast, the other proteins are isolated incolourless forms, although a FeIII2 S2 cluster may beassembled in the C42G protein via incubation withFeIII and sulfide. The four mutant proteins were iso-lated as stable mononuclear HgII forms which wereconverted to unstable mononuclear FeIII preparationsthat contain both holo and apo protein. The FeIII

systems were characterized by metal analysis and massspectrometry and by electronic, electron paramagneticresonance, X-ray absorption and resonance Ramanspectroscopies. The dominant FeIII form in the C9Apreparation is a Fe(S-Cys)3(OH) centre, similar to thatobserved previously in the C6S mutant protein. Related

centres are present in the proteins NifU and IscU re-sponsible for assembly and repair of iron-sulfur clustersin both prokaryotic and eukaryotic cells. In additionto Fe(S-Cys)3(OH) centres, the C9G, C42G and C42Apreparations contain a second four-coordinate FeIII

form in which a ligand appears to be supplied by theprotein chain. Electronic supplementary material tothis paper can be obtained by using the Springer Linkserver located at http://dx.doi.org/10.1007/s00775-002-0355-1.

Keywords Rubredoxin Æ Mutant proteins ÆClostridium pasteurianum

Abbreviations BCA: bicinchoninic acid Æ Cp: Clostridi-um pasteurianum Æ CtoG,A: cysteine to glycine or alaninemutations Æ CtoS: cysteine to serine mutations ÆRd: rubredoxin Æ TCA: trichloroacetic acid

Introduction

The electron transfer protein rubredoxin (Rd) fromClostridium pasteurianum (Cp) features a single Fe(S-Cys)4 active site (Fig. 1) [1, 2, 3]. It expresses inEscherichia coli as a mixture of Fe and Zn centres[4, 5, 6, 7]. For mutant proteins, the relative contentof isolated FeIII, ZnII and apo forms varies from caseto case [8]. In an interesting variation, the assembly ofa FeIII2 S2 cluster into the C42A molecular variant ofRdCp was reported by Meyer et al. in 1997 [9]. Thus,substitution of a cysteine ligand by a non-ligatingalanine allowed the incorporation of a binuclear Fe2S2cluster into a protein that normally accommodates amononuclear Fe(S-Cys)4 site. Although this wouldseem to require a major change in stereochemistry, theb-loops of the protein which bind the [Fe2S2(S-Cys)2(N-His)2] centre in the Rieske protein of cyto-chrome bc1 complexes are related closely to those(residues 5–11, 38–44; Fig. 1) which bind the Fe(S-Cys)4 centre in Rd [10]. In related work, generation of

J Biol Inorg Chem (2002) 7: 781–790DOI 10.1007/s00775-002-0355-1

Electronic supplementary material to this paper, comprisingFigs. S1–S7 and Tables S1–S3, can be obtained by using theSpringer Link server located at http://dx.doi.org/10.1007/s00775-002-0355-1

M. Cross Æ Z. Xiao Æ A.G. Wedd (&)School of Chemistry, University of Melbourne,Parkville, Victoria 3010, AustraliaE-mail: agw@unimelb.edu.auTel.: +61-3-83446813Fax: +61-3-93475180

S.C. Drew Æ J.R. PilbrowSchool of Physics and Materials Engineering,Monash University, Victoria 3800, Australia

E.M. Maes Æ R.S. CzernuszewiczDepartment of Chemistry, University of Houston,Houston, TX 77204-5641, USA

G.N. GeorgeStanford Synchrotron Radiation Laboratory,SLAC, Stanford University, PO Box 4349,MS 69, Stanford, CA 94309, USA

a Rd-like centre in a site designed for a FeIII2 S2 clusterhas been accomplished [11].

The possibility of assembling mononuclear or binu-clear ferric centres in the one site is relevant to thechemistry of the NifU and IscU proteins involved iniron-sulfur cluster assembly [12, 13, 14, 15, 16, 17, 18,19]. In particular, available evidence indicates that theNifU protein of Azotobacter vinelandii features two do-mains, the first of which provides a labile Rd-likeFeIII(S-Cys)3 site and the second an FeIII2 S2 clusterbinding site [15]. Expression of the separate domainsconfirmed their ability to independently assemble theirindividual centres.

Mutation of each of the four RdCp cysteine ligandsC6, C9, C39 and C42 in turn to serine (CtoS mutants)allows isolation of FeS3O centres [20, 21]. Their indi-vidual properties depend upon whether a surface (C9,C42) or an interior (C6, C39) ligand is substituted(Fig. 1). The three proteins C9S, C39S and C42S containFeIII(S-Cys)3(O-Ser) centres, while the C6S case featuresa FeIII(S-Cys)3(OH) site in which the S6 sidechain is nota ligand [22].

The present paper reports the generation and isola-tion of the four mutant proteins C9G, C9A, C42G andC42A in which exterior ligands C9 and C42 (sidechains-CH2S-Fe) are replaced by non-ligating glycine (-H) andalanine (-CH3) residues. The primary aim was toincorporate an exchangeable coordination position thatmight allow construction of reactive centres. The abilityof these proteins to assemble FeIII2 S2 and (S-Cys)3FeIII(OH) centres is assessed.

Materials and methods

Mutagenesis

General materials and DNA manipulations have been describedpreviously [6, 21, 22]. Oligonucleotides were purchased fromGIBCO BRL Custom Primers. The following primers generatedthe four mutants C9G, C9A, C42G and C42A: 3¢-GTACATGT-CATCG/CACCTATATAAATATTA-5¢; 3¢-CATACAGGAAACCG/CACCTCATCCTTTTC-5¢. The emboldened bases are mis-matched and convert a cysteine codon to glycine and alanine. Aftermutagenesis and identification by DNA sequencing, the fourmutants were subcloned into the EcoRI and PstI sites of thepKK223-3 vector for protein expression.

Protein expression and isolation of FeIII forms

Plasmids pKK223-3/RdCp carrying the four mutations weretransformed into E. coli strain JM109. The procedures for proteinexpression were similar to those described previously [6, 21]. TheC42A gene expressed as a brown fraction, a mixture of apo proteinand a form incorporating an [FeIII2 S2] centre [9]. This fraction couldbe isolated by the methods usually employed for Rd after initialanion exchange chromatography (DE52) [9, 21]. The other threesystems produced colourless protein, presumably a mixture of ZnII-substituted and apo forms [21]. For each of the four systems, de-naturation of the lysate with trichloroacetic acid (TCA) in thepresence of b-mercaptoethanol followed by addition of Fe(NO3)3led to orange fractions. These were isolated by anion exchangechromatography (DE52, 10 cm·3 cm; Tris-HCl, 50 mM, pH 7.8;gradient, 0.10–0.30 M NaCl; 100 mL h–1), ammonium sulfateprecipitation (60%) of contaminating proteins and size exclusionchromatography (G-75; 80 cm·1 cm; 20 mL h–1). Pooled fractionswere concentrated (Centricon) to approximately 20 mg mL–1 andstored at –70 �C. Overall yields were 3–5 mg of protein per L ofculture.

SDS-PAGE confirmed the presence of a single protein com-ponent. However, both FeIII and apo forms were present and wereseparated by anion exchange chromatography on a high-resolutionHPLC UNO Q column (3 cm·0.5 cm, Bio-Rad) equilibrated withbuffer (Tris-HCl, 50 mM, pH 7.8). The properties of the columnwere assessed with a mixture of FeIII, ZnII and apo forms of thestable mutant protein V44I [8] (0.5 mg of each form; 2 mL totalsolution; see Supplementary material, Fig. S1). The FeIII and ZnII

forms eluted at 0.23 and 0.26 M NaCl (flow rate 2 mL min–1),respectively, consistent with their expected anionic charges of 10–and 11– at pH 7.8. The apo-protein component elutes over theextended range 0.3–0.6 M NaCl, consistent with the presence of anumber of apo forms differing in the redox levels of the three cy-steine residues. The FeIII preparations (2 mg) of the new proteinswere eluted under the same conditions (e.g. Fig. S2).

Generation of Group 12 derivatives

The protein solutions as prepared above can be converted to Zn,Cd, or Hg forms. The protein (5 mg; 2 mg mL–1) was precipitatedby addition of TCA (ca. 5–10%) and b-mercaptoethanol (0.15 M),producing a white protein precipitate. After centrifugation(7600 rpm, 3 min), the protein pellet was washed with TCA (5%)containing b-mercaptoethanol (50 mM) and re-dissolved in aminimum volume of Tris base (0.5 M) containing b-mercaptoeth-anol (0.15 M). After two-fold dilution with water, the precipitationand dissolution procedure was repeated. The final pH was ap-proximately 8. A two-fold molar excess of the metal salt solution ofZn(NO3)2, Cd(NO3)2 or HgCl2 (200 mM) was added. After incu-bation on ice (1 h), a precipitate was removed by filtration. Thesolution was diluted four-fold with buffer (Tris-HCl, 50 mM,pH 7.8) and applied to a DEAE-52 anion exchange column(1.5 cm·6.0 cm) which had been equilibrated with buffer

Fig. 1. NH� � �S interactions (dashed lines) around the Fe(S-Cys)4center in RdCp (generated from the coordinates of pdb5rxn.ent inthe Brookhaven Protein Databank). A pseudo-two-fold axis isperpendicular to the page, passing though the Fe atom

782

(Tris-HCl, 50 mM, pH 7.6). The protein was batch eluted (9–12 mL, approx. 0.5 mg mL–1) with a solution of 0.3 M NaCl inbuffer, the elution being monitored at 280 nm. The eluted proteinsamples were concentrated (Centricon) and desalted with a Bio-DelP-6 DG gel column (Bio-Rad) for further analysis.

Separation of different forms of the metallated proteins wasachieved on the UNO Q anion exchange column. The ZnII and CdII

preparations of the C9G and C9A proteins each exhibited adominant chromatographic peak at 0.21 M NaCl, assigned to therespective MII holo proteins, which appeared to be accompanied bylow proportions of apo forms (Fig. S3).

On the other hand, stable holo forms only of the HgII proteinswere detected (Fig. S3). The analysis provided conditions for ad-dition of HgCl2 at the lysate stage of the primary isolation forrecovery of a pure HgII-substituted form of each of the proteinsC9G, C9A, C42G and C42A at yields of about 10 mg per L ofculture. Subsequently, each mutant protein was purified initially asa HgII derivative and then converted to the unstable FeIII forms bythe denaturation method described above to achieve the highestpossible Fe content.

Assays for iron and protein content

The Fe-substituted proteins were assayed for both Fe and proteincontent. The protein content was estimated with the bicinchoninicacid (BCA) protein assay [23, 24]. This method is sensitive (<25lg),but is influenced by variation in the content of cysteine, cystine,tryptophan and tyrosine as well as a number of interferences. Forunknown reasons, RdCp did not perform adequately as a standard.As a consequence, the stable mutant proteins C9S and C42S werechosen as standards. The BCA reagent was freshly prepared [25].Sodium deoxycholate (0.15% w/v, 100 lL) was added to the pro-tein solution (1 mL, 0.3–0.4 mg) in an Eppendorf tube and left tostand at room temperature (10 min). TCA (72% w/v, 100 lL) wasadded and the mixture agitated by a vortex mixer before centrifu-gation (6000 rpm, 15 min). The supernatant (1.6 mL) was collectedquantitatively and the protein precipitate was washed withdeionized water (3·200 lL). The washings were added to thesupernatant and the solution assayed for iron (vide infra).

The white pellet of precipitated protein was dissolved for theprotein assay. Sodium dodecyl sulfate (SDS, 50 lL, 5% w/v)containing NaOH (0.1 M) was added to the pellet, which wasdissolved by vortex mixing. The volume was made up to 500 lLwith Milli-Q water. BCA reagent (2.0 mL) was added to the sampleand the solution incubated at 60 �C (30 min). The solution wasallowed to cool to room temperature (30 min) before measurementof absorbance at 562 nm.

Control experiments using stable Rd proteins of known ex-tinction coefficients confirmed the method (Table 1). The estimateof protein mass determined spectroscopically correlated with thatdetermined colourimetrically by the BCA method to better than10% (Supplementary material, Table S1; 0.5 lg in 6.5 lg).

Iron content was assayed with bathophenanthroline reagent[26, 27]. All containers and apparatus were acid washed. Reagents

were treated to remove adventitious iron. A standard curve wasprepared from 10, 20, 40, 60, 80 and 100 lL aliquots (10–100 lgFe) of a stock solution of standard (NH4)2Fe(SO4)2. Each aliquotof supernatant (vide infra) was diluted to 100 lL with deionizedwater before sequential addition of deionized water (2.1 mL),aqueous HCl (1%, 100 lL), saturated sodium acetate solution(750 lL), sodium ascorbate solution (5%, 100 lL) and bath-ophenanthroline reagent (1.5 mM, 400 lL). The solutions were leftat RT for 1 h before measurement of absorbance at 535 nm. Thestandard curves obtained followed Beer’s law up to 100 lg (1.8·10–6moles) of iron. The assay was confirmed by control experiments.Stable Rd proteins were analysed, providing the expectediron:protein ratio of 1 (Table 1).

Generation of Fe2S2 centres

The brown Fe2S2 form of the C42A mutant protein was isolateddirectly from the lysate as described above. Under these sameconditions, the C42G, C9A and C9G proteins expressed as col-ourless products. Brown solutions were generated from either theapo or mononuclear FeIII forms of the C42G and C42A systemsby incubation with a two-fold excess of Na2S and Fe(NO3)3 [28,29]. These were loaded onto a DEAE-52 anion exchange column(10 cm·1 cm) equilibrated with buffer (Tris-HCl; 50 mM,pH 7.8). The brown fraction was eluted with 0.5 M NaCl inbuffer (Tris-HCl; 50 mM, pH 7.8). Solutions were concentrated(Centricon) and stored at –70 �C. The same procedures applied tothe C9G and C9A systems produced the mononuclear FeIII formsonly, suggesting that neither protein can accommodate an Fe2S2centre.

Binding of exogenous ligands

Each of the FeIII forms of the cysteine to glycine or alanine mutant(CtoG,A) proteins (0.16 mM) were titrated with 2–100 moleequivalents of solutions (100 mM) of NaN3, NaCN, NaNCS,NaOAc, MeCN, L-cysteine or b-mercaptoethanol. The titrant so-lution was introduced quantitatively by syringe (Hamilton). Theprotein solution was monitored by UV-visible spectrophotometryin the 260–600 nm region.

Physical measurements

Electrospray ionization mass spectra, electronic spectra, X-rayabsorption spectra and resonance Raman spectra were obtained asdescribed previously [21, 22]. EPR spectra were acquired at X-bandusing a Bruker ESP 380E spectrometer equipped with a dielectricresonator and an Oxford instruments helium flow cryostat. Spectrawere obtained at the same gain, using microwave powers between 2and 20 lW and modulation amplitudes £ 4G.

The resonance Raman samples were lyophilized by freeze-dry-ing under vacuum (5–8 h). Sample conditions were �5 mM; NaCl(0.3 M); Tris-HCl (50 mM, pH 7.8). 54Fe2O3 was converted to54FeCl3.nH2O by dissolving in hot hydrochloric acid (12 M). Thesolution was transferred quantitatively into a round-bottomSchlenk flask and the acid removed by evaporation under vacuum.The dried residue was diluted to the appropriate volume with milli-Q water. The incorporation of 54Fe was achieved by the metalsubstitution technique described above. The incorporation of 18Oor 2H was achieved by dissolving the lyophilized form of the pro-tein in H2

18O (96.7% 18O, Matheson) or 2H2O (99.8% D,Cambridge Isotope). Electronic spectra of dissolved samples oflyophilized protein were indistinguishable from those obtainedbefore freeze-drying.

X-ray absorption spectroscopy employed lyophilized samplesreconstituted by the addition of water. Protein (7–10 mM) waspresent in Tris-HCl buffer (50 mM) and NaCl (300 mM) at pH 8.0.The samples were mixed with glycerol (30% by volume) and loadedinto Lucite sample cells fitted with Mylar tape windows.

Table 1. Ratio of iron to protein in solution

Protein Iron:protein ratio

RdCpa 1.1V8/44Ga,b 1.0C9Ab 0.7C9Gb 0.8C42Ab 0.5C42Gb 0.6

aProtein content determined by absorbance measurement (molarabsorptivity, 6640 cm–1 M–1 at 490 nm)bProtein content determined by BCA assay using the C9S and C42Sproteins as standards

783

Results and discussion

The exterior ligands C9 and C42 in native Rd are relatedby a pseudo-two-fold symmetry [3, 21]. The pair ofproteins (C9G, C42G) might be anticipated to displayproperties in common, as might the pair (C9A, C42A).Each has a Cys ligand sidechain -CH2S-Fe

III replaced bynon-ligating -H or -CH3 sidechains, which are bothmuch less sterically demanding. Their ability to incor-porate an Fe2S2 cluster was explored initially.

Isolation of Fe2S2 centres

Typically, molecular variants of RdCp express in E. coliwith zinc bound in the metal site or as a mixture of bothiron and zinc forms. In contrast, the C42A protein ex-pressed with an intact [Fe2S2]

2+ cluster [9].This observation was confirmed with the isolation of

a brown fraction whose electronic spectrum reproducedthat seen previously (Table 2) [9]. Solutions rapidly lostcolour upon dilution, precluding isolation of a singleform of the protein. Analysis of the original data [9]indicates that the best preparations contained about50% Fe2S2 groups. Treatment of the apo or mononu-clear FeIII forms (vide infra) of the C42A protein with asource of FeIII and sulfide under reducing conditions [28,29] also generated the characteristic spectrum of anFe2S2 centre.

In contrast to the behaviour of the C42A system,each of the C9A, C9G and C42G variants expressed byE. coli was colourless. For the C42G system, an elec-tronic spectrum characteristic of an Fe2S2 species wasobserved when the as-isolated sample or the apo ormononuclear FeIII forms were treated with FeIII andsulfide under reducing conditions (Table 2). However,the C9 variants did not incorporate Fe2S2 centres underthese experimental conditions.

The present Rd systems feature three Cys residuesonly. The b-loops of protein which bind the [Fe2S2(S-Cys)2(N-His)2] centre in the Rieske protein are relatedclosely to those (residues 5–11, 38–44; Fig. 1) which bindthe Fe(S-Cys)4 centre in Rd [10]. In fact, the two coordi-nating Cys residues of the Rieske centre superpose withinterior ligands C6 and C39 of Rd while the bridging

sulfido ligands superpose with the Sc atoms of exteriorligands C9 and C42, one of which is missing in the Rdvariant systems studied here. Consequently, it is likelythat the Fe2S2 forms of C42G and C42A will have similarstructural features and at least one non-cysteinyl ligand.

Isolation of mononuclear centres

Denaturation of the lysate under reducing conditionsfollowed by addition of Fe(NO3)3 led to orange solu-tions for each of the four CtoG,A systems. The char-acteristic electronic spectrum of the FeIII(S-Cys)4 activesite in Rd arises from S to Fe(III) charge transfer tran-sitions [31, 32]. Those of the FeIII-CtoG,A proteinsdiffered significantly (Table 2, Fig. 2), being broaderand apparently shifted to higher energy, consistent witha lower site symmetry and the presence of fewer thanfour Cys ligands. Similar shifts were seen in the CtoSproteins which feature FeIII(S-Cys)3)(O-Ser) and Fe

III(S-Cys)3(OH) sites [21, 22].

However, the FeIII centres of the CtoG,A systemswere unstable and the colour bleached upon standing at4 �C (half-life, 4–12 h). Mass spectrometry (Fig. S4,Table S2) and anion exchange chromatography(Fig. S2) confirmed the presence of a mixture of the holoand apo forms in solution. Higher proportions of apoprotein are present in the FeIII-C42 preparations than inthe C9 analogues. A search was initiated for a stablemetallated form that could be isolated conveniently.RdCp expresses in E. coli as a mixture of Fe and Znforms [7] and CdII-substituted forms of RdCp and itsCtoS variants are stable (see, e.g. [21]).

The mononuclear FeIII derivatives were convertedto ZnII, CdII and HgII forms. The ZnII and CdII

Table 2. Electronic absorption spectra in Tris-HCl (50 mM, pH 7.9)

Protein kmax (nm)

RdCp 750, 570, 490, 380, 350, 280Fe2S2-C42G 453, 410, 327, 280Fe2S2-C42A 448, 410, 325, 280Fe2S2-C42A

a 444, 412, 322, 280FeIII-C9A 528, 463 (sh), 423, 336, 280FeIII-C9G 515, 462, 341, 280FeIII-C42A 510, 412, 329, 280FeIII-C42G 529, 421, 320, 280

aRef. [9]

Fig. 2. Electronic absorption spectra of FeIII-RdCp proteins(0.15 mM in 50 mM Tris-HCl, pH 8). The spectra have been offsetby 0.2 absorption units for clarity

784

preparations of the C9G and C9A proteins each exhib-ited a dominant HPLC chromatographic peak at 0.21 MNaCl, assigned to the respective MII holo proteins.These were accompanied by low proportions of apoforms. The ESI-MS spectra detected mixtures of apoand holo forms. On the other hand, only stable holoforms of the HgII proteins were detected by both tech-niques (Fig. S3, Table S3).

Substitution with HgII at the lysate stage of the iso-lation led to pure forms of each of the proteins C9G,C9A, C42G and C42A. A significant increase in yieldwas obtained from the isolation procedure, 10 mg ofprotein per L of culture compared to 3–5 mg for theFeIII forms. The stable HgII derivatives could be con-verted to the FeIII forms when required, expediting studyof these less stable forms. Samples were maintained onice and assayed within 1 h of generation. A higher Fecontent was achieved in comparison with FeIII samplesgenerated directly from the lysate. However, all of theCtoG,A mutants generated contained less than the sto-ichiometric ratio of 1.0 (Table 1) and appear to bemixtures of holo and apo proteins.

EPR of FeIII forms

The EPR spectra of rubredoxins and related systemshave been reviewed and EPR spectra of the CtoS pro-teins reported previously [20, 21, 33, 34]. Like RdCp, thelatter each display resonances at g-values in the ranges9.3–9.6 and 4.0–4.8, consistent with a high-spin FeIII

centre possessing large zero-field splitting and a highrhombic distortion. The former g-value is associatedwith one of the principal directions of the lowest (±1/2 or±5/2) Kramers doublet and the latter with the threeprincipal directions of the middle (±3/2) Kramers dou-blet.

In the present study at 3 K, each of the proteins ex-hibit resonances with geff�9, 7.3 and 5.7, together with asuperposition of broad (geff�5.1 to 3.8) and narrow(geff�4.3) resonances (Fig. 3). The spectra were observ-able even at 77 K and temperature dependence (notshown) indicated that the broad and sharp features inthe geff�4 region are independent. Assuming that thezero-field splitting of each of the systems exceeds themicrowave quantum, this established the presence oftwo sites, centres I and II, in which the ±1/2 doublet lieslowest (i.e. D>0).

The spectrum of the FeIII(S-Cys)4 centre I in thedesulforedoxin from Desulfibrio gigas also exhibitsgeff�7.3, 5.7 and a broad feature centred near geff�4 [35].These spectral features are associated with a singleprotein site possessing a value of E/D<0.1 and are ob-served for centre I in the present systems. Inspection ofrhombograms [33, 35, 36] indicates that the geff�9 fea-ture must originate from a different centre with a largervalue of E/D, centre II. The middle Kramers doublet ofthis second centre will contribute a three-line powderpattern in the vicinity of geff�4, so that it is likely that

the broad resonance in this region is actually a super-position of contributions from centres I and II. Thehigh-field features arising from some of the other prin-cipal directions of the lower and middle doublets ofcentres I and II could not be observed, presumably be-cause of strain broadening of the powder lineshapes.

The very weak contribution from the geff�9 line inthe C9A sample (Fig. 3) indicates a minor contributionfrom centre II in this system. Interestingly, the geff�4.3line is still quite intense. Moreover, the E/D value cor-responding to the geff�9 line of each sample deviatesfrom the condition (E/D�1/3) required to generate asharp geff�4.3 resonance (for the CtoA samples,geff�9.3, whereas for CtoG, geff<9). It is probable thateach of the samples contains traces of adventitious FeIII

that occupies non-protein coordination sites in the fro-zen solution. It appears that the C9A sample contained adominant FeIII protein species (centre I), that each of theC9G, C42G and C42A samples also contains a secondFeIII protein species (centre II) and that each is con-taminated with adventitious FeIII sites.

X-ray absorption spectroscopy

The fine structure (EXAFS) associated with the X-rayabsorption spectrum of the iron atom was useful inprobing the structure at the iron centres of the CtoS

Fig. 3. X-band EPR spectra of Rd proteins at 3 K. The magneticfield axis has been normalized in the dimensionless ratio g0/geff,where g0�2.0 represents the true g-factor of the FeIII. Spectra havebeen normalized to their maximum peak-to-peak intensity. A four-line background CrIII signature on the high-field side of the sharpgeff�4.3 has been subtracted carefully from all spectra

785

single mutant proteins with their FeOS3 centres [21, 22].The present experiments aimed to obtain structural dataon the C9A protein, the system which appeared tocontain a single dominant FeIII protein site, centre I. Inaddition, the C9/42S protein [22] was examined to pro-vide comparative data for an FeO2S2 centre.

Results at pH 8.0 are reported in Table 3. The near-edge spectra are typical of four-coordinate Rd proteins(Fig. 4a) [21]. Fourier transformation of the EXAFSleads to well-resolved peaks with outer-shell featuresbeing found at distances >3 A (see Fig. 4c). The peaksare at distances consistent with the presence of Fe-O,Nand Fe-S bonds.

For the C9/42S protein, the coordination number N,while significant to 0.5 ligand atoms only, indicates thepresence of two O,N and two S ligand atoms (Table 3).This is consistent with the presence of an FeO2S2 centre.The derived FeIII-O and FeIII-S bond lengths at pH 8.0are 1.842(1) A and 2.307(1) A, respectively. The Fe-Odistance is indistinguishable from those in the FeIIIS3Ocentres of the C9S and C42S single mutant proteins. TheFe-S distance is significantly longer.

The data for the oxidized C9A sample at pH 8.0 areconsistent with the presence of a four-coordinateFeS3(O,N) metal centre (Table 3). The relative areasunder the transformed peaks are similar to those seen inthe singly mutated CtoS proteins [21] and contrasts withthat seen for the C9/42S protein with its FeS2O2 centre(Fig. 4). An FeIII-O,N vector is clearly detected. Inter-estingly, the FeIII-X bond length for the exterior ligandvariant C9A is significantly longer [0.037(7) A] than thatobserved for the C9S analogue which has a Fe(S-Cys)3(O-Ser) centre (Table 3). It is similar to or longerthan those Fe-O bonds in the C39S (Fe-OSer) and C6S(Fe-OH) proteins [21, 22].

Resonance Raman spectra

The RR spectra of the FeIII forms of the CtoG,A pro-teins were obtained with a number of excitation wave-lengths in the 476.5–568.2 nm range, with the bestoverall enhancement occurring at 530.9 nm (Fig. 5). The

spectra are dominated by a group of bands in the 320–370 cm–1 region which are sensitive to 54Fe substitution(0.2–2.2 cm–1). These properties associate them with Fe-S(Cys) stretching modes m(FeS3) [22, 37, 38, 39, 40, 41].

The C9A preparation appears to contain a predom-inant FeIII protein species, centre I, whose highest pos-sible site symmetry is C3v, as for the CtoS systems [22].The m(FeS3) features of the C9A and C9S proteins aresimilar with stretching frequencies at �338, �345 and�355 cm–1 being shared by both systems (Table 4).Shoulders at �332, �356 and �373 cm–1 seen for theC9S mutant are also detected, albeit weakly, in the C9Aspectrum. When the C9A protein is excited with the476.5 nm laser line, an additional mode at �364 cm–1 isenhanced, which occurs in the RR spectrum as a sepa-rate band centered around 361 cm–1 by overlapping withthe �355 cm–1 shoulder (Fig. 6). The exact origin ofthese spectral features is presently unclear, but their56/54Fe isotope sensitivities suggest some m(FeS3)involvement (Table 4).

Some of the spectra of the other CtoG,A proteins(centres I and II) appear broader and more complexthan that of the C9A protein, possibly due to overlap-ping features from centres I and II. However, the fre-quencies of the main bands in all these spectra aresimilar (Fig. 5, Table 4).

To test for the presence of solvent ligands, spectrawere collected in H2

18O and 2H2O solvents. The bandobserved for the 54Fe-substituted C9A mutant at a fre-quency of 587 cm–1 (585 cm–1 in natural abundanceC9A) is sensitive to H2

18O and 2H2O substitution(Fig. 6, Table 5). The shifts to 570 and 548 cm–1 uponincorporation of H2

18O are smaller and larger, respec-tively, than that estimated for a diatomic 54Fe-O stretch(17 and 39 versus 26 cm–1). However, the mean shift ofthe 570/548 cm–1 doublet, which derives from Fermiresonance with a weak protein band at 553 cm–1

(Fig. 6a) [41], corresponds well to the expected value.Upon incorporation of 2H2O, the stretching frequencyat 587 cm–1 shifts by 19 cm–1 to higher frequency(Fig. 6c). These shifts demonstrate that the band mustinvolve vibrational motion of an iron atom and a water-exchangeable oxygen atom. The shift to higher

Table 3. EXAFS curve fitting results for the FeIII centres in mutant rubredoxins at pH 8.0a

Protein Fe-S (A) r2 (A2)b N Fe-X (A)c r2 (A2)b N Non-Cysligand(s) X

RdCpd 2.274(1) 0.0019(1)C9A 2.300(1) 0.0049(2) 2.7(1) 1.878(3) 0.0011(5) 0.7(1) OH?C9Sd 2.285(1) 0.0019(1) 1.843(4) 0.000(3) O-SerC39Sd 2.281(2) 0.0027(2) 1.865(6) 0.0016(6) O-SerC42Sd 2.290(1) 0.0013(1) 1.836(5) 0.0006(5) O-SerC9/42S 2.307(1) 0.0024(3) 2.0(1) 1.842(2) 0.0019(3) 1.8(1) (O-Ser)2C6Sd 2.280(1) 0.0027(1) 1.869(5) 0.0024(5) OH

aValues in parentheses are estimated standard deviations from thediagonal elements of the covariance matrix. These values are pre-cisions, and the accuracies (generally larger) are difficult to esti-mate. For bond lengths, the commonly accepted upper limit isbetween ±0.001 and ±0.002 A

bDebye-Waller factorscX=O or N atomdRef. [21]

786

frequency upon 2H substitution is characteristic of dif-ferential coupling [42] of a stretching m(Fe-O) coordinatewith a bending d(Fe-O-H) coordinate. The coupling isrelieved in 2H2O by the large drop in the d(Fe-O-D)frequency, leading to the observed increase in frequency

Fig. 4a–c. X-ray absorption spectra at pH 8.0 for the C9A (lowerplots) and C9/42S proteins (upper plots). a Near-edge spectra; b k3-weighted EXAFS data (experimental and best fit plots); c EXAFSFourier transforms (phase-corrected for Fe-S backscattering)

Fig. 5. Low-temperature (77 K) resonance Raman spectra (75–825 cm–1) of (a) C9A, (b) C9G, (c) C42A and (d) C42G proteins(�5 mM; NaCl, 0.3 M; Tris-HCl, 50 mM, pH 7.8). Conditions:530.9 nm excitation wavelength, 300 mW laser power, 4 cm–1 slitwidths and 0.5 cm–1 increments

Table 4. Resonance Raman frequenciesa and assignmentsb forC9X mutant rubredoxins in the 300–400 cm–1 region

C9A C9G C9S Assignment

303 (0.8) 311 (0.6) 304 (0.2) d(SCC)325 (0.4) 324 (0.5) 324 (0.3) d(SCC)333/338 (0.2/0.4) 333/340

(0.3/0.4)332/338(0.5/0.7)

m(FeS3)

345 (0.8) 348 (0.8) 346 (0.8) m(FeS3)355/364/373(1.4/0.6/0.8)

356/364(1.4/1.3)

356/373(1.4/1.0)

m(FeS3)

aNumbers in parentheses are positive increases in wavenumbersupon substitution of natural abundant iron with 54Fe. Peak posi-tions were obtained from curve fitting analyses (not shown)bBased upon those assigned to the CtoS mutant proteins [22, 41]

787

of the the m(Fe-OH) stretch. The behaviour allows as-signment of the 587 cm–1 band to the presence of anFeIII-OH moiety. The observed isotope shifts are iden-tical, within experimental error, to those observed forthe C6S protein [22]. Consequently, centre I may beidentified as an FeIII(S-Cys)3(OH) complex.

Bands at 618, 625 and 620 cm–1 in the C9G, C42Gand C42A samples exhibit entirely similar behaviour(Table 5, Figs. S5–S7). Consequently, centre I is presentin all FeIII-CtoG,A mutant proteins.

The m(Fe-OH) mode in the C9A protein occurs at asignificantly lower energy (585 cm–1) compared to those

of the C9G, C42G and C42A samples (618–625 cm–1)and the more completely characterized C6S protein(617 cm–1) [22, 41]. It would appear that the secondarycoordination sphere supplied by the protein matrix dif-fers in the C9A case, altering the environment of theFeIII-OH moiety and its vibrational coupling properties.

No other features in the spectra of the CtoG,A mu-tant proteins depend upon substitution of O and Hisotopes. Consequently, the FeIII four-coordinate centreII does not appear to involve a solvent-based ligand.

Binding of exogenous ligands at FeIII centres

Native RdCp has an iron atom bound by four cysteineresidues in a distorted tetrahedral environment. Theengineered CtoG,A mutants of RdCp were designed toincorporate an iron atom bound by three cysteine resi-dues and an exchangeable coordination position. Afour-coordinate FeIII(S-Cys)3(OH) centre is present ineach of the CtoG,A preparations and was observedpreviously in the C6S protein [22]. The buried interiorOH ligand in the latter appears to be inert to ligandexchange.

The influence of potential exogenous ligands on theelectronic spectra of the FeIII forms of the mutant pro-teins was assessed. Thiocyanate, acetate, acetonitrile andL-cysteine had no effect. The addition of two equivalentsof NaCN (based on FeIII content) to each of the CtoG,Aprotein solutions was sufficient to cause progressive lossof the S(Cys)fiFe charge transfer bands in the visibleregion and apparent destruction of the iron centre.

The addition of NaN3 generated three isosbesticpoints in the C9A case (Table 6), indicating clean con-version of the single FeIII site present, the hydroxocomplex (centre I), to an azido complex (Eq. 1):

FeIIIðS�CysÞ3ðOHÞþN�3 �!FeIIIðS�CysÞ3ðN3ÞþOH�

ð1Þ

Isosbestic points were detected for the equivalent reac-tions with the other FeIII-CtoG,A proteins (Table 6,Fig. 7a), suggesting reaction of centre I only. However,in each case, spectral changes were not complete afterthe addition of 100 equivalents of ligand, so Eq. 1 wouldappear to be an equilibrium.

Fig. 6. Low-temperature (77 K) resonance Raman spectra (260–760 cm–1) of 54Fe-reconstituted C9A protein (�5 mM; NaCl,0.3 M; Tris-HCl, 50 mM, pH 7.8): (a) in H2O; (b) in H2

18O; (c)in 2H2O. The m(Fe-OH) stretch is correlated. Conditions: 476.5 nmexcitation wavelength; 300 mW laser power; 4 cm–1 slit width;0.5 cm–1 increments

Table 5. Frequency shifts (cm–1) in m(Fe-OH) stretches of CtoG,Amutant proteins upon 54Fe, 18O and 2H isotope substitution

Protein C6S C9A C9G C42A C42G

m(Fe-OH) 617 585 618 620 62556Fefi54Fe +3 +2 +2 +2 +316Ofi18O –28 –17/–39a –23 –28 –271Hfi2H +10 +19 +13 –b +9

aFermi doublet (see text)bNot determined

Table 6. Isosbestic points (nm) for the CtoG,A mutantsa

Protein Ligand

Azide HSCH2CH2OH

C9A 352, 418, 552 –C9G 345, 441, 549 –C42A 340, 553 334C42G 345 342

aProtein, 1 mg mL–1; ligand solution, 100 mM; buffer, Tris-HCl,50 mM, pH 7.8

788

Titration of b-mercaptoethanol (HSCH2CH2OH)into solutions of the C9G and C9A proteins caused re-duction of the FeIII forms present. However, for theC42A sample, the resulting spectrum (Fig. 7b, uppertrace) was similar to that observed for RdCp (Fig. 7b,lower trace; Table 2) which possesses an FeIII(S-Cys)4centre (Fig. 7b). An isosbestic point developed at334 nm, suggesting substitution at centres I or II (Eq. 2):

FeIIIðS� CysÞ3XþHSCH2CH2OH

! FeIIIðS� CysÞ3ðSCH2CH2OHÞ ð2Þ

However, spectral changes are not complete at ligand toprotein ratios of �50:1. Above this ratio, reduction oc-curred, preventing quantitative interpretation. Interest-ingly, L-cysteine does not induce any spectral change, i.e.does not bind to nor reduce the FeIII centres.

Small linear ligands only (azide, cyanide) are able togain access, suggesting that the approach to the vacantcoordination position is protected by protein sidechainsand/or backbone. The exception is the C42A protein,which also binds b-mercaptoethanol to reassemble anFeIIIS4 centre (Eq. 2). Note that this form has the ste-reochemical flexibility to accommodate an Fe2S2 cluster.

Structural considerations

The interior ligand variant C6S features a stable FeIII(S-Cys)3(OH) centre. Likely contributors to its stabilityinclude hydrogen-bonding interactions between theFeIII-OH and adjacent S6 functions plus the protectionafforded by its internal position. In addition, seven ofthe ten residues present in the sequence 4–13 C6S (YT-STVCGYTY) carry OH functions. Some of these maycontribute to a stabilizing hydrogen-bonded network.On the other hand, the equivalent centres in the exteriorligand variants C9G, C9A, C42G and C42A are lessstable and presumably less protected. The FeIII(S-Cys)3centres in certain NifU and IscU proteins are also labilein vitro (see, e.g. [15]).

The C42S mutant protein displays a Fe(S-Cys)3(O-Ser) site. The gross structural consequence of sub-stitution of sulfur (covalent radius >1 A) by oxygen(covalent radius 0.73 A) is replacement of an atom ofvolume �1 A3 with one of volume �0.4 A3. The crystalstructure revealed that, while major structural changesare limited to the region of the mutation, the polypeptidechain ‘‘kinks’’, with the Ca and Cb atoms of S42 moving0.4(1) and 0.5(1) A, respectively, towards the iron atom[21]. Closer packing of atoms is apparent in this localsurface region.

For the FeIII(S-Cys)3(OH) centre I in the C9A pro-tein, the combination of an OH ligand and an alanineCbH3 sidechain will have a similar volume to a serinylOCbH2 ligand in the stable mutant C9S, whose struc-ture, presumably, is similar to that of C42S discussedabove (both C9 and C42 are exterior ligands; see Fig. 1).Small, linear, anionic ligands N3

– and CN– appear ca-pable of competing with the OH– ligand at pH 7.8.

In addition to a centre I, the C9G preparation alsoexhibits a centre II which does not appear to involve asolvent-based ligand. The EPR parameters of centre IIare similar to those of the native and C9S protein,consistent with the presence of a four-coordinate com-plex, Fe(S-Cys)3X. The question then arises as to thenature of the fourth ligand X. Replacement of the C9ligand sidechain -CbH2S- by Gly-H

a will leave a signif-icant cavity. A structural rearrangement is likely, leadingto a peptide carbonyl oxygen atom or a peptide amidenitrogen atom acting as the fourth ligand X. In RdCp,the amide NH of residue Tyr11 acts as the proton donorin the NH� � �S hydrogen bond to Sc of ligand C9 (Fig. 1)[3]. Amide N11 is 3.45 A from Sc9. In C9G, NH11 mayinteract directly with the Fe atom or, if deprotonated,the amide itself may act as ligand. The pseudo-two-foldsymmetry means that the C42G and C42A proteins havethe equivalent NH44 centre available.

Centre I [FeIII(S-Cys)3(OH)] or centre II [FeIII(S-Cys)3X; X=protein-based ligand] may be considered aspreliminary representations of the labile FeIII(S-Cys)3Xcentres generated by the NifU and IscU proteins in-volved in iron-sulfur cluster assembly. It is also possiblefor transient FeIII2 S2 centres to be assembled in sitesproviding three Cys ligands.

Fig. 7a, b. Electronic spectra for titration of FeIII protein samples.a C9G with azide; b C42A with b-mercaptoethanol

789

Acknowledgements Dr. John Boas is thanked for discussions onEPR spectra. A referee is thanked for pertinent structural sugges-tions. This work was supported by grants from the AustralianResearch Council (A10020211 to A.G.W.) and the Robert A.Welch Foundation (E-1184 to R.S.C.).

References

1. Beinert H, Holm RH, Munck E (1997) Science 4:653–6592. Eaton WA, Lovenberg W (1973) In: Lovenberg W (ed) Iron-sulfur proteins, vol. II. Academic Press, New York, pp 131–162

3. Sieker LC, Stenkamp RE, LeGall J (1994) Methods Enzymol243:203–216

4. Mathieu I, Meyer J, Moulis J-M (1992) Biochem J 285:255–2625. Zeng Q, Smith ET, Kurtz DM Jr, Scott RA (1996) Inorg ChimActa 242:245–251

6. Ayhan M, Xiao Z, Lavery MJ, Hamer AM, Nugent KW,Scrofani SDB, Guss M, Wedd AG (1996) Inorg Chem 35:5902–5911

7. Taylor PK, Parks BA, Kurtz DM Jr, Amster IJ (2001) J BiolInorg Chem 6:201–206

8. Xiao Z, Maher MJ, Cross M, Bond CS, Guss JM, Wedd AG(2000) J Biol Inorg Chem 5:75–84

9. Meyer J, Gagnon J, Gailard J, Lutz M, Achim C, Munck E,Petillot Y, Colangelo CM, Scott RA (1997) Biochemistry36:13374–13380

10. Iwata S, Saynovits M, Link TA, Michel H (1996) Structure4:567–579

11. Kazanis S, Popapsky TC, Barnhart TM, Penner-Hahn JE,Mizra UA, Chait BT (1995) J Am Chem Soc 117:6625

12. Dean DR, Bolin JT, Zheng L (1993) J Bacteriol 175:6737–674413. Zheng L, Cash VL, Flint DH, Dean DR (1998) J Biol Chem

273:13264–1327214. Yuvanyiama P, Agar JN, Cash VL, Johnson MK, Dean DR

(2000) Proc Natl Acad Sci USA 97:599–60415. Agar JN, Yuvanyiama P, Jack RF, Cash VL, Smith DR, Dean

DR, Johnson MK (2000) J Biol Inorg Chem 5:167–17716. Agar JN, Zheng L, Cash VL, Dean DR, Johnson MK (2000)

J Am Chem Soc 122:2136–213717. Foster MW, Mansy SS, Hwang J, Penner-Hahn JE, Surerus

KK, Cowan JA (2000) J Am Chem Soc 122:6805–680618. Nishio K, Nakai M (2000) J Biol Chem 275:22615–2261819. Ollagnier-de-Choudens S, Mattioli T, Takahashi Y, Fontecave

M (2001) J Biol Chem 276:22604–22607

20. Meyer J, Gaillard J, Lutz M (1995) Biochem Biophys ResCommun 212:827–833

21. Xiao Z, Lavery MJ, Ayhan M, Scrofani SDB, Wilce MCJ, GussJM, Tregloan PA, George GN, Wedd AG (1998) J Am ChemSoc 120:4135–4150

22. Xiao Z, Gardner AR, Cross M, Maes EM, Czernuszewicz RS,Sola M, Wedd AG (2001) J Biol Inorg Chem 6:638–649

23. Wiedelman KJ, Braun RD, Fitzpatrick JD (1988) Anal Bio-chem 175:231–237

24. Price N (1999) Biotechnol Appl Biochem 29:99–10825. Smith PK, Krohn RI, Hermansen GT, Malia AK, Gartner FH,

Provenzano MD, Fujimoto EK, Goeke NM, Olson BJ, KlenkDC (1985) Anal Biochem 150:76–85

26. Van de Bogart M, Beinert H (1967) Anal Biochem 20:325–33427. Evans PJ, Halliwell B (1994) Methods Enzymol 223:82–9228. Meyer J, Fujinaga J, Gaillard J, Lutz M (1994) Biochemistry

33:13642–1365029. Brereton PS, Maher MJ, Tregloan PA, Wedd AG (1999) Bio-

chim Biophys Acta 1429:307–31630. Johnson MK (1998) Curr Opin Chem Biol 2:173–18131. Lovenberg W, Sobel BE (1965) Proc Natl Acad Sci USA

54:193–19932. Gebhard MS, Deaton JC, Koch SA, Millar M, Solomon EI

(1990) J Am Chem Soc 112:2217–223133. Hagen WR (1992) Adv Inorg Chem 38:165–22234. Moura I, Pereira AS, Tavares P, Moura JJG (1999) Adv Inorg

Chem 47:361–41935. Moura I, Macedo A, Moura JJG (1989) In: Hoff AJ (ed) Ad-

vanced EPR: applications in biology and biochemistry, chap 23.Elsevier, Amsterdam

36. Troup GJ, Hutton DR (1964) Br J Appl Phys 15:1493–149937. Czernuszewicz RS, LeGall J, Moura I, Spiro TG (1986) Inorg

Chem 25:696–70038. Czernuszewicz RS, Kilpatrick LK, Koch SA, Spiro TG (1994)

J Am Chem Soc 116:7134–714139. Spiro TG, Czernuszewicz RS, Han S (1988) In: Spiro TG (ed)

Biological applications of Raman spectroscopy, vol. 3. Wiley-Interscience, New York, pp 523–553

40. Spiro TG, Czernuszewicz RS (2000) In: Que L Jr (ed) Physicalmethods in bioinorganic chemistry. University Science Books,Sausalito, Calif., pp 59–119

41. Maes EM (2001) PhD dissertation, University of Houston42. Czernuszewicz RS, Su YO, Stern MK, Macor KA, Kim D,

Groves JT, Spiro TG (1988) J Am Chem Soc 110:4158–4165

790