doi:10.1006/jmbi.2000.4145 available online at http://www.idealibrary.com on J. Mol. Biol. (2000) 303, 423±432

NMR Structure of Oxidized Glutaredoxin 3 fromEscherichia coli

Kerstin Nordstrand1, Anna SandstroÈ m2, Fredrik AÊ slund2,3,Arne Holmgren2,3, Gottfried Otting2 and Kurt D. Berndt1,4*

1Center for StructuralBiochemistry KarolinskaInstitutet, S-141 57 HuddingeSweden2Department of MedicalBiochemistry and BiophysicsKarolinska Institutet, S-171 77,Stockholm, Sweden3Medical Nobel Institute forBiochemistry, KarolinskaInstitutet, S-171 77, StockholmSweden4Faculty of Natural Sciences,SoÈdertoÈrns HoÈgskola, Box 4101S-141 04, Huddinge, Sweden

Present address: A. SandstroÈm, DInnovation Policy Studies, NUTEK,Sweden; F. AÊ slund, Center for GenKarolinska Institutet, S-171 77, Stoc

Abbreviations used: 2D, two-dim3D, three-dimensional; NOE, nucleaNOESY, 2D NOE spectroscopy; COspectroscopy; 2QF COSY, double-quCOSY; 3QF COSY, triple-quantum-HSQC, heteronuclear single quantuHN(CO)HB, correlation spectroscopresonances using magnetization thrcarbon.

E-mail address of the correspondkurt.berndt@csb.ki.se

0022-2836/00/030423±10 $35.00/0

A high precision NMR structure of oxidized glutaredoxin 3 [C65Y] fromEscherichia coli has been determined. The conformation of the active siteincluding the disulphide bridge is highly similar to those in glutaredoxinsfrom pig liver and T4 phage. A comparison with the previously deter-mined structure of glutaredoxin 3 [C14S, C65Y] in a complex with gluta-thione reveals conformational changes between the free and substrate-bound form which includes the sidechain of the conserved, active sitetyrosine residue. In the oxidized form this tyrosine is solvent exposed,while it adopts a less exposed conformation, stabilized by hydrogenbonds, in the mixed disul®de with glutathione. The structures furthersuggest that the formation of a covalent linkage between glutathione andglutaredoxin 3 is necessary in order to induce these structural changesupon binding of the glutathione peptide. This could explain the observedlow af®nity of glutaredoxins for S-blocked glutathione analogues, in spiteof the fact that glutaredoxins are highly speci®c reductants of glutathionemixed disul®des.

# 2000 Academic Press

Keywords: disulphide bond; conformational change; induced ®t;thioredoxin superfamily; thioltransferase

*Corresponding author

Introduction

Glutaredoxins (Grxs) are found in most livingorganisms, viruses, bacteria, plants and animals,where they act as thiol-disul®de oxidoreductases.These enzymes are implicated in vital cellular func-tions such as the synthesis of ribonucleotides byproviding reducing equivalents to ribonucleotidereductase (RR) and regulation of enzymes byspeci®c reduction of protein-glutathione mixeddisul®des (reviewed by Holmgren et al., 1999). In

epartment ofS-117 86 Stockholm,

omics Research,kholm, Sweden.ensional;r Overhauser effect;SY, correlationantum-®ltered

®ltered COSY;m coherence;y for HN and Hb

ough the carboryl

ing author:

Escherichia coli, three different Grxs have beenidenti®ed (AÊ slund et al., 1994, 1996; Vlamis-Gardikas et al., 1997). Grx1 and Grx3 are 9 kDaproteins with large structural similarities, whileGrx2 is an atypical Grx of 24 kDa. In spite of thestructural similarities, Grx1 and Grx3 have a 35mV difference in redox potential (AÊ slund et al.,1997), and display different ef®ciencies in thereduction of RR and glutathione-containing mixeddisul®des.

NMR structures of Grx1 in the reduced (Sodanoet al., 1991) and oxidized form (Xia et al., 1992) aswell as in complexes with glutathione (Grx1-SG,Bushweller et al., 1994) and a 25 residue peptidefrom ribonucleotide reductase (Grx1-B1, Berardi &Bushweller, 1999) have been determined. While theglutathione binding mode in Grx1-SG is similar toGrx3-SG, the peptide binding in Grx1-B1 is funda-mentally different. The B1 peptide does not bind asan intermolecular b-bridge to the loop containingthe conserved cis proline (the ``cis-Pro loop'') inthis structure, as is usually found for peptide bind-ing in the thioredoxin superfamily. Instead peptidebinding results in burial of the Grx1-peptide disul-®de, thus promoting the intra-molecular reactionrequired for complete reduction of the peptide dis-

# 2000 Academic Press

Table 1. Structural statistics

A. Input dataNumber of non-redundant distance restraints 1046

Intraresidual 237Medium range (ji ÿ jj < 5) 511Long range (ji ÿ jj5 5) 298

B. Residual restraint violations for the final 20 conformersAverage residual NOE restraint violations

Sum (AÊ ) 8.26(�0.27)Maximum (AÊ ) 0.10(�0.01)

Average residual angle restraint violationsSum (deg.) 35.70(�4.27)Maximum (deg.) 2.14(�0.20)

C. RMSD of the final 20 conformersAverage RMSD to the mean structure

All backbone atoms (N, Ca, C0 ofresidues 1 to 82) (AÊ )

0.57(�0.11)

All heavy atoms (residues 1 to 82) (AÊ ) 0.97(�0.09)D. Ramachandran plotPercentage in

Most favored regions (%) 83Additionally allowed regions (%) 15Generously allowed regions (%) 2Disallowed regions (%) 0

424 NMR Structure of Oxidized E. coli Glutaredoxin 3

ul®de. The Grx1 structures clearly show that onlysmall structural changes accompany the redoxreactions. Slight reorientation of the active-sitehelix was observed upon complex formation, whilethe fate of individual side-chains was dif®cult todeduce due to the conformational variability in theoxidized and reduced form structures. Structuresof reduced (Sun et al., 1998) and glutathione mixeddisul®de (Yang et al., 1998) forms of human Grxhave also been determined by NMR spectroscopy.As for Grx1, the difference between the structurein the reduced form and the complex is very small.The same feature has been found for other redoxactive enzymes that share the glutaredoxin fold.The thioredoxin structure has been determined inthe reduced and oxidized forms (Holmgren et al.,1975; Jeng et al., 1994; Katti et al., 1990; Qin et al.,1994; Saarinen et al., 1995; Weichsel et al., 1996) aswell as in Trx-peptide mixed disulphides (Qin et al.,1995, 1996). The structure of DsbA is also knownin both reduced (Guddat et al., 1998; Schirra et al.,1998) and oxidized (Guddat et al., 1998) forms. Theoverall structural similarity between these enzymesin their different redox states underscores the in¯u-ence of subtle changes in structure on the redoxproperties. This demonstrates the importance ofaccurate and precise structures of both the freeenzymes and the corresponding complexes inorder to understand biological activity in atomicdetail.

We have previously determined a high precisionNMR structure of a Grx3-glutathione mixed disul-®de, trapped by a Cys to Ser mutation of the moreC-terminal active-site Cys, and identi®ed a numberof well-de®ned interactions between Grx3 and thecovalently bound glutathione (Nordstrand et al.,1999a). The current report concerns the NMR struc-ture of the internally oxidized form of this protein,which is determined to equally high precision,allowing a detailed comparision to the Grx3-SGcomplex. Signi®cant conformational changes in theactive site are discussed in terms of their in¯uenceon both redox properties and substrate speci®city.

Results

Structure determination

The wild-type E. coli Grx3 contains threecysteine residues. Two of these form the active sitecysteine pair, Cys11 and Cys14, and the third is asingle cysteine, Cys65. This non-active site cysteinehas been shown to have no signi®cant in¯uence onthe activity of Grx3 (Nordstrand et al., 1999a) butcould lead to unwanted dimerization, and was forthe structure determination replaced with tyrosineby site-directed mutagenesis. The structure ofGrx3-SG, which was previously determined(Nordstrand et al., 1999a), used the same replace-ment for Cys65, which facilitates comparisonsbetween the two forms.

The structure calculation was based on 1046non-redundant distance restraints derived from

integrated cross peaks in NOESY spectra (Table 1).In addition, 318 torsion angle restraints derivedfrom the analysis of intraresidual and sequentialdistance restraints and coupling constants (3JHNHa,3JHaHb,

3JNHb and 3JC0Hb) using the FOUND algor-ithm (GuÈ ntert et al., 1998), implemented in theDYANA program (GuÈ ntert et al., 1997), were alsoincluded in the calculations. By this analysis, 18stereospeci®c assignments of b-methylene protonswith non-degenerate chemical shifts were ident-i®ed. Using the ``glomsa'' routine in the programDYANA additional stereospeci®c assignmentswere obtained and a total of 49 stereospeci®cassignments were used in the ®nal structure calcu-lation.

A ®nal set of 20 conformers from the DYANAcalculations with target functions <0.73 AÊ 2 and lowresidual restraint violations were selected forrestrained energy minimization using the programOPAL (LuginbuÈ hl et al., 1996).

Description of the NMR structure

The RMSD to the mean structure of all backboneatoms (residues 1 to 82; N, Ca, C0) of the 20 energyminimized conformers of oxidized Grx3 is 0.57(�0.11) AÊ . The corresponding number for all heavyatoms is 0.97 (�0.09) AÊ . In a Ramachandran plot,83 % of the non-glycine residues have backbonetorsion angles in the allowed regions, 15 % in theadditionally allowed regions and 2 % in the gener-ously allowed regions. No residue has f and cangles in the disallowed regions of the Ramachan-dran plot.

The structure of Grx3 in the oxidized form issimilar to the previously determined Grx3-SG(Nordstrand et al., 1999a). The secondary structureis essentially the same as in Grx3-SG, with four b

NMR Structure of Oxidized E. coli Glutaredoxin 3 425

strands forming a central mixed b sheet (residues 2to 7, 29 to 33, 54 to 58 and 60 to 63), ¯anked bythree a helices (residues 11 to 25, 37 to 48, 64 to74). The short C-terminal helix was found to be a310 helix, involving residues 77 to 81, in 11 of the20 conformers, and a regular a helix, involvingresidues 77 to 82, in the remaining nine confor-mers. In the NMR structure of Grx-SG, a majorityof the conformers have a regular a helix at the Cterminus, and a smaller fraction (5/20) containsthe 310 helix. As in Grx-SG, a cis peptide bondbetween Val52 and Pro53 was identi®ed from thecharacteristic cross peaks involving these tworesidues in the NOESY spectrum. A classic b bulgeinvolving residues 55 and 62 to 63 was found in allconformers of the Grx3 structure of the oxidizedform.

The active site is located at the N terminus ofhelix a1. The disul®de bond between Cys11 andCys14 is largely buried, exposing only slightly thesulfur atom of Cys11, while the sulfur atom ofCys14 remains completely buried. The c angle ofCys11 is well de®ned with an average value of 101(�6) � whereas the f angle is largely dispersedamong the 20 energy minimized conformers. Theimprecision in f of Cys11 is a consequence of thefact that no amide proton resonance could beidenti®ed for Cys11, preventing the measurementof 3JHNHa, and the low number of NOEs to Thr10which is located in a solvent exposed loop region.This did, however, not affect the precision of thedisul®de bond, since the torsion angle w3 is pre-cisely de®ned with an average value of 74 (�2) �and an average distance between the Ca atoms of

Figure 1. Superposition of the 20 conformers of oxidizedusing all backbone atoms (N, Ca, C0 of residues 1 to 82) fGrx3-SG. Thick lines represent mean structures of oxidized G

the two cysteines of 5.1 (�0.05) AÊ . This motif hasbeen classi®ed as a right-handed hook (Hutchinson& Thornton, 1996).

Comparison with the structure of theglutaredoxin 3-glutathione mixed disulfide

The RMSD between protein backbone atoms(residues 1 to 82) of the mean structure of oxidizedGrx and the mean structure of Grx3-SG is 1.1 AÊ .This can be compared to the average pairwiseRSMD of the same region in the 20 conformers ofeach structure separately, which is 0.83 AÊ in Grx3-SG and 0.83 AÊ in oxidized Grx3. The major contri-bution to the difference in these numbers is a sig-ni®cant shift in position of helix a1 (Figures 1 and2), which occurs mainly along the long axis of thehelix, and is most pronounced at the N terminus.The shifted position of the helix seems to affect theC-terminal end which includes Gly25 in oxidizedGrx3, whereas in Grx3-SG the corresponding helixends with Lys24. Also helix a2 and the followingloop region (ending at Thr51) is signi®cantlyshifted between oxidized Grx3 and Grx3-SG(Figures 1 and 2). The movements of the twohelices can be explained by the adaptation of Grx3to its substrate in Grx3-SG.

A small difference in secondary structurebetween oxidized Grx3 and Grx3-SG is found instrand b4, which in a majority of the conformers ofoxidized Grx3 includes residues 60 to 63 and inGrx3-SG is one residue shorter, including residues60 to 62. However, both variants occur in each setof conformers although with different proportions;

Grx3 (red) with the 20 conformers of Grx3-SG (blue)or alignment. The glutathione moiety is excluded fromrx3 (yellow) and Grx3-SG (turquoise), respectively.

Figure 2. Comparison of the mean structures of oxi-dized Grx3 and Grx3-SG: (a) average global displace-ment of the backbone atoms (squares) and all heavyatoms (triangles) between the mean structures of oxi-dized Grx3 and Grx3-SG following superposition of allbackbone atoms (N, Ca, C0 of residues 1 to 82). (b) Aver-age local displacement of the backbone atoms (squares)and all heavy atoms (triangles) in the mean structures ofoxidized Grx3 and Grx3-SG. The local displacement iscalculated by superimposing the backbone atoms (N,Ca, C0) in a three-residue segment, and calculating theRMSD of the middle residue.

Figure 3. Superposition of a selected conformer of oxi-dized Grx3 (red) with a selected conformer of Grx3-SG(blue), showing the conformational changes in Tyr13and Asp66. The backbone atoms of residues 10 to 16were used for alignment. The bonds between Asp66 Nand HN and Tyr13 OZ and HZ are included (gray) forclarity. The hydrogen bond between HZ of Tyr13 andOd of Asp 66 is present in 17/20 conformers of Grx3-SG,and the hydrogen bond between OZ of Tyr13 and HN ofAsp 66 in 7/20 conformers. Both are indicated as brokenlines.

426 NMR Structure of Oxidized E. coli Glutaredoxin 3

notably, the structure of the oxidized Grx3 is betterde®ned in this region.

In addition to the small changes in secondarystructure orientation in the oxidized form com-pared to Grx3-SG, a few side-chains have a largeaverage displacement, in particular Tyr13 andHis15 (Figure 2). In Grx3-SG the Tyr13 side-chainis hydrogen-bonded via the phenolic oxygen to theamide of Asp66 in a few of the 20 conformers, andvia the phenolic hydrogen to the side-chain carbox-ylate in a majority of the conformers. In oxidizedGrx3, Tyr13 is no longer at a suitable distance forsuch interactions (Figure 3). Due to the altered pos-ition of the active-site helix, the distance betweenTyr13 and Asp66 is shorter in oxidized Grx3. Thew1 rotamer position of Tyr13 in Grx3-SG thatallows the side-chain of Tyr13 to pack close to theprotein is sterically prevented in oxidized Grx3,and consequently the aromatic ring is directed out

towards the solvent (Figure 4). As a result, theaverage solvent accessible area of the Tyr13 side-chain is increased from 18 to 68 %. In addition, theside-chain of Asp66, no longer involved in ahydrogen bond in oxidized Grx3, moves from a tconformation in Grx3-SG to a g conformation inoxidized Grx3 (using nomenclature as rec-ommended by Markley et al., 1998). Similarly, theside-chain of His15, which is well de®ned in bothoxidized Grx3 and Grx3-SG, turns away from theactive site in the oxidized protein by going from ag conformation in Grx3-SG to a t conformation inoxidized Grx3 (Figure 4).

Comparison with other oxidized glutaredoxins

Three oxidized glutaredoxin structures havebeen determined previously: Grx from phage T4(PDB code 1AAZ) determined to 2.0 AÊ resolution(Eklund et al., 1992); pig liver Grx (PDB code1KTE) determined to 2.2 AÊ resolution (Katti et al.,1995); and E. coli Grx1 (PDB code 1EGO) deter-mined by NMR spectroscopy (Xia et al., 1992).Despite a number of differences between these glu-taredoxins, the conformation of the active site isvery similar. In both crystal structures, the disul-®de bonds are right-handed hooks (Hutchinson &Thornton, 1996). In T4 Grx, the w3 angles of thetwo molecules in the assymetric unit are 57 � and

Figure 4. Comparison of the active site region of a superposition of all 20 conformers in: (a) oxidized Grx3 and (b)Grx3-SG. All heavy atoms of residues 7 to 19, 52 to 54 and 63 to 67 are displayed. Side-chains that change confor-mation between the structures are displayed with different colors: Cys11 (yellow); Tyr13 (red); Cys14 (orange); His15(violet); Asp66 (green). In (b), glutathione is colored gray. All backbone atoms (N, Ca, C0) of residues 7 to 19, 50 to 54and 63 to 67 were used for alignment.

NMR Structure of Oxidized E. coli Glutaredoxin 3 427

67 � respectively, with Ca-Ca distances of 5.3 and5.2 AÊ , respectively. In pig Grx the w3 angle is 73 �and the Ca-Ca distance is 5.3 AÊ . In the NMR struc-ture of E. coli Grx1, the disul®de bond is not pre-cisely de®ned, but in ten out of 20 conformers thedisul®de bond can be classi®ed as a right-handedhook. The similarity of the active sites in the well-de®ned Grx structures includes the backbone con-formation and the direction of the side-chains (i.e.similar w1 angle) of all residues in the active siteloop. In Grx3 these are Pro12, Tyr13 and His15,while in T4 Grx the corresponding residues are

Val, Tyr and Asp, in pig Grx Pro, Phe and Arg andin E. coli Grx1 Pro, Tyr and Val.

Discussion

Conformational changes upon binding ofglutathione can explain low affinity forglutathione analogues

Redox reactions of Grxs and related enzymes arecharacterized by small conformational changes,which are in general localized to the active site.

428 NMR Structure of Oxidized E. coli Glutaredoxin 3

This is also found in Grx3 when the structure ofoxidized Grx3 is compared to Grx3-SG(Nordstrand et al., 1999a). In Grx3-SG, a number ofresidues important for the interaction with gluta-thione were identi®ed. A detailed comparison ofGrx3-SG with the oxidized form of Grx3 presentedhere show that slight movements of secondarystructure elements occur upon binding of gluta-thione (Figure 1). Although small, these move-ments are suf®cient to allow a less exposed andthus presumably energetically more favorable con-formation of the Tyr13 side-chain in the complex.Speci®c substrate recognition thus appears todepend on both interaction with the cis-Pro loopand the presence of the intermolecular disul®debond which is the only link between the gluta-thione ligand and helix a1. Assuming a similarbehavior in other Grxs, this could be the reasonthat the glutathione analogues S-methylglutathioneand S-nitrosoglutathione do not seem to inhibithuman Grx even at high (mM) concentrations(Srinivasan et al., 1997). Likewise, only weak inhi-bition (Ki � 0.4-1.2 mM) by S-blocked glutathioneanalogues was found in Grx1 (HoÈoÈg et al., 1982).In vivo, low af®nity in Grxs for non-covalentlybound glutathione should be bene®cial since thehigh levels of GSH in the cytosol (Gilbert, 1984)would otherwise have an inhibitory effect. The dis-crimination between different substrates can thusbe proposed to occur in a transition state, in whichthe conformation of Grx resembles that of theenzyme-substrate mixed disul®de, rather thandepending on tight binding of glutathione to thefree enzyme.

Regulation of redox potential by the activesite residues

The two residues between the active-site cysteineresidues have a large impact on the redox poten-tial, not only in Grx but also in structurally relatedproteins (Grauschopf et al., 1995; Huber-Wunderlich & Glockshuber, 1998; Joelson et al.,1990; Krause et al., 1991; MoÈssner et al., 1998). Thiseffect has been attributed to the in¯uence of theseresidues on the pKa of the thiolate in the reducedform of the enzymes, which determines the leavinggroup ability of the thiolate anion (Grauschopfet al., 1995). Attempts to exploit such a relationshipin the design of enzymes with known redox poten-tial by substituting the active site sequence of thewild-type protein to that of another member of thethioredoxin superfamily have been made (Huber-Wunderlich & Glockshuber, 1998; MoÈssner et al.,1998). These experiments show that while qualitat-ively in agreement with the design, such a simpli-®ed model does not predict redox potential of themutants with good precision. In particular, thecorrelation is poor when Grxs are considered,indicating that other factors are involved.

In Grx3, the two intervening residues in theactive site are Pro and Tyr, which is the consensussequence among Grxs. Pro is known to be an ef®-

cient N-terminal helix capping residue (Richardson& Richardson, 1988). Using the nomenclature ofthese authors, the ®rst residue of a helix is positionN, the N-cap, which in Grx3 is Cys11. Pro hasbeen shown to have a marked preference for pos-ition N1, which can be rationalized by the fact thatit does not have an exposed amide proton for con-tinuation of the helix in the N-terminal direction(Richardson & Richardson, 1988). Residues Tyr orPhe on the other hand has no preference for theN2 position, and must be conserved for otherreasons. Analysis of the Tyr13 side-chain in Grx3shows that it changes from an extended, highlysolvent exposed conformation in the oxidized formto a more stable arrangement in Grx3-SG. Thisstabilization is a consequence of ®tting glutathionein the binding site of Grx3, and includes movinghelix a1 and the loop containing Asp66 furtherapart, thereby providing enough space for the Tyrside-chain to pack against the protein surface(Figure 3). This much reduces the solvent accessi-ble surface of the hydrophobic tyrosine side-chain(see Results). In addition, Asp66 provides stabiliz-ing interactions for the Tyr hydroxyl by hydrogenbonding to the side-chain carboxylate and prob-ably also to the backbone amide.

A previous X-ray crystallography study of a glu-taredoxin-like DsbA mutant, H32Y (Guddat et al.,1997) gives an interesting perspective on the move-ments of this side-chain. While His32 in wild-typeDsbA was found to be in an extended confor-mation in both oxidized and reduced DsbA, thisside-chain was packed against the bulk of the pro-tein in a different crystal form of DsbA, which theauthors presumed resembles the enzyme withbound substrate. In contrast, the Tyr residue in theH32Y mutant was found to pack closely againstthe protein in all crystal forms which was attribu-ted to a favorable interaction between the tyrosineand a phenylalanine ring, Phe36 (Guddat et al.,1997). Therefore, a common feature of wild-typeGrx and DsbA seem too emerge: in the unboundforms, a bulky residue in the active site extendsfrom the protein into the solution. When the sub-strate binds, this side-chain packs against the pro-tein, diminishing the solvent accessible area.

In this light it is interesting to note that onedifference between the enzyme-substrate com-plexes of Grx and DsbA is the change in stabilityeffected by burial of the bulky residue. In Grx,reduced solvent accessibility of the Tyr side-chainshould stabilize Grx-SG, while burial of a morepolar and potentially charged His residue in DsbAshould rather destabilize the complex. This view issupported by kinetic data for DsbA, which showsthat the DsbA-glutathione complex is an unstableintermediate (Zapun et al., 1993). In DsbA, the Hisside-chain was shown to destabilize the oxidizedform by unfavorable interaction with the helixdipole, while it interacts favorably with the Cys31thiolate in the reduced form (Guddat et al., 1997).In Grx3, the exposure of the aromatic ring of Tyr13would be expected to destabilize the reduced and

NMR Structure of Oxidized E. coli Glutaredoxin 3 429

oxidized forms by similar amounts, while the com-plex would be signi®cantly stabilized.

His15 is another active site residue in which theside-chain undergoes a large movement betweenoxidized Grx3 and Grx3-SG (Figure 4). His15 hasbeen shown to in¯uence the speci®c activity ofGrx3 in the glutathione-disu®de oxidoreductaseassay (HED assay, Nordstrand et al., 1999b). Thiseffect was suggested to arise from participation ofthis residue in the stabilization of the Cys11 thio-late in reduced Grx3, similar to the situation inreduced DsbA (Guddat et al., 1998). The compari-son of the structures of oxidized Grx3 and Grx3-SGshows that the side-chain of His15 is capable oflarge amplitude motions, which could allow it toact as a base for the thiol proton of Cys14 in theGrx3-SG of the wild-type protein during oxidation.

In general, redox potential and protein stabilityare thermodynamically linked for the thiol-disul-®de oxidoreductases (Creighton, 1986). The effectof stabilization or destabilization of the reducedand oxidized forms as well as the enzyme-sub-strate complexes is to modulate the redox poten-tials of these proteins. The above analysis of themolecular details of Grx3 and DsbA demonstratesthat the importance of the active site residues isnot limited to their in¯uence on thiol pKa. Localinteractions and small conformational changes inthe active site modulate redox potential by affect-ing the stability of each of the oxidized, reducedand mixed-disulphide forms.

Materials and Methods

Sample preparation

Grx3 C65Y was prepared from E. coli cells (HMS-174DE3) transformed with the pET-3 plasmid containing thegene for Grx3 C65Y. The cells were grown in the pre-sence of 50 mg/ml of ampicillin. Preparation of a uni-formly 15N-labeled sample was performed as describedpreviously (AÊ slund et al., 1994, 1996). Material uniformlylabeled with both 15N and 13C was prepared from cul-tures grown on modi®ed M9 medium in which the solesource of nitrogen and carbon were supplied as 15Nlabeled ammonium sulfate and 0.02 % (w/v) 13C glucose,respectively. The M9 medium was modi®ed to contain asupplement of vitamins and trace elements and 0.1 mMof calcium chloride. A two liter culture of this mediumwas induced at mid-logarithmic growth by addition ofIPTG to a ®nal concentration of 0.4 mM, and harvestedafter an additional 12 hours of growth at a ®nalabsorbance of 1.2 at 600 nm. From the resulting 3.5 g ofbacterial pellet, a total of 40 mg of puri®ed Grx3 C65Ywas obtained as described by Nordstrand et al. (1999b).Unlabelled Grx3 C65Y was prepared from culturesgrown on LB medium in a 15 liter fermentor, using thesame puri®cation procedure as for the labeled material.The puri®ed protein was analyzed by N-terminal aminoacid sequence determination, reverse phase HPLC andSDS gel electrophoresis in order to con®rm the hom-ogeneity of the material. The puri®ed material wasentirely recovered in its disul®de form, since no freethiol groups were present as determined using DTNB(5,50-dithio-bis(2-nitrobenzoic acid), Ellman, 1959). TheNMR samples were prepared at pH 5.5 with a 5 mM

protein concentration in 5 % 2H2O/95 % H2O. A samplein 2H2O was prepared by lyophilisation of an oxidizedaqueous solution of Grx3 C65Y, followed by re-dissol-ving the sample in 2H2O.

NMR data collection

All NMR experiments were recorded at a proton fre-quency of 600 MHz on a Bruker DMX600 NMR spec-trometer.

The following spectra were recorded to obtainsequence speci®c resonance assignments: 2QF-COSY(Bodenhausen et al., 1984; Rance et al., 1983); 3QF-COSY(MuÈ ller et al., 1986); clean CITY (Briand & Ernst, 1991)with a mixing time of 80 ms; NOESY (Ernst et al., 1987)with a mixing time of 40 ms, 960 � 4096 data points,t1max � 50 ms and t2max � 300 ms; o1-decoupled NOESY(Otting et al., 1990) with a mixing time of40 ms, 1024 � 2048 data points, t1max � 70 ms andt2max � 150 ms; NOESY in 2H2O solution with amixing time of 40 ms, 1024 � 4096 data points,t1max � 70 ms and t2max � 300 ms. In addition, a 3DNOESY-15N-HSQC experiment (Messerle et al., 1989)with a mixing time of 50 ms, 64 � 360 � 2048 datapoints, t1max � 16 ms, t2max � 29 ms and t3max � 150 mswas recorded using the 15N enriched sample. NOEdistance restraints were collected from all four NOESYspectra.

The following coupling constants were measured:3JHNHa was measured by inverse Fourier transformationfrom the 2D NOESY spectrum in H2O solution(Szyperski et al., 1992); 3JHaHb was measured from asmall ¯ip angle COSY spectrum in D2O solution (Aueet al., 1976; Mueller, 1987); 3JNHb was estimated qualitat-ively from a 3D HNHB experiment (Archer et al., 1991)recorded with the 15N labeled sample; 3JC0Hb was esti-mated qualitatively from a 3D HN(CO)HB experiment(Grzesiek et al., 1992) recorded with the 13C/15N labeledsample.

All spectra were processed using the program PROSA(GuÈ ntert et al., 1992), and analyzed using the programXEASY (Bartels et al., 1995).

Structure calculations

Structure calculations were performed using the pro-gram package DYANA (version 1.5, GuÈ ntert et al., 1997).Cross-peak volumes were converted to upper distancelimits (2.4-5.5 AÊ ). The FOUND algorithm (GuÈ ntert et al.,1998) was used to perform a gridsearch of allowed con-formations using short range NOE distance restraints aswell as 3JHNHa and 3JHaHb coupling constants to obtainangle restraints for the dihedral angles f, c and w1 andstereospeci®c assignments for b-methylene proton pairs.A total of 200 random starting structures were subjectedto 8000 steps of simulated annealing in torsion anglespace, followed by 500 steps of conjugate gradient mini-mization. Of these, the 20 conformers with the lowesttarget function were immersed in a 6 AÊ water shell andsubjected to 5000 steps of conjugate gradient minimiz-ation using the program OPAL (LuginbuÈ hl et al., 1996).The resulting 20 energy minimized conformers wereselected to represent the NMR structure of glutaredoxin3 in its oxidized form.

430 NMR Structure of Oxidized E. coli Glutaredoxin 3

Structure analysis

The analysis of secondary structure and structuralmotifs in Grx3 were performed using the program PRO-MOTIF (Hutchinson & Thornton, 1996). For compari-sons, the disul®des of T4 Grx (1AAZ), pig Grx (1KTE)and E. coli Grx1 (1EGO) were analyzed using the sameprogram. The program PROCHECK (Laskowski et al.,1996) was used to examine the quality of the structure,including the Ramachandran plot appearance. Calcu-lation of RMSD values and solvent accessible surfaceareas was performed using the program MOLMOL(Koradi et al., 1996). The same program was used forpreparation of all pictures.

Protein Data Bank accession number

The coordinates of the 20 energy minimized confor-mers of Grx3 in the oxidized form has been deposited inthe RCSB Protein Data Bank, with the accession code1FOV.

Acknowledgments

Financial support by grants from the Swedish NaturalScience Research Council (10161 and 11146), the SwedishCancer Society (961) and EU grant BIO4-CT960436 isgratefully acknowledged.

References

Archer, S. J., Ikura, M., Torchia, D. A. & Bax, A. (1991).An alternative 3D NMR technique for correlatingbackbone 15N with sidechain Hb resonances in lar-ger proteins. J. Magn. Reson. 95, 636-641.

AÊ slund, F., Ehn, B., Miranda-Vizuete, A., Pueyo, C. &Holmgren, A. (1994). Two additional glutaredoxinsexist in Escherichia coli: glutaredoxin 3 is a hydrogendonor for ribonucleotide reductase in a thiore-doxin/glutaredoxin 1 double mutant. Proc. NatlAcad. Sci. USA, 91, 9813-9817.

AÊ slund, F., Nordstrand, K., Berndt, K. D., Nikkola, M.,Bergman, T. & Ponstingl, H., et al. (1996). Glutare-doxin-3 from Escherichia coli. Amino acid sequence,1H and 15N NMR assignments, and structural anal-ysis. J. Biol. Chem. 271, 6736-6745.

AÊ slund, F., Berndt, K. D. & Holmgren, A. (1997). Redoxpotentials of glutaredoxins and other thiol-disul®deoxidoreductases of the thioredoxin superfamilydetermined by direct protein-protein redox equili-bria. J. Biol. Chem. 272, 30780-30786.

Aue, W. P., Bartholdi, E. & Ernst, R. R. (1976).Two-dimensional spectroscopy. Application tonuclear magnetic resonance. J. Chem. Phys. 64, 2229-2246.

Bartels, C., Xia, T.-H., Billerter, M., GuÈ ntert, P. &WuÈ thrich, K. (1995). The program XEASY for com-puter-supported NMR spectral analysis of biologicalmacromolecules. J. Biomol. NMR, 6, 1-10.

Berardi, M. J. & Bushweller, J. H. (1999). Binding speci-®city and mechanistic insight into glutaredoxin-cat-alyzed protein disul®de reduction. J. Mol. Biol. 292,151-161.

Bodenhausen, G., Kogler, H. & Ernst, R. R. (1984). Selec-tion of coherence transfer pathways in NMR pulseexperiments. J. Magn. Reson. 58, 370-388.

Briand, J. & Ernst, R. R. (1991). Computer-optimizedhomonuclear TOCSY experiments with suppres-sion of cross relaxation. Chem. Phys. Letters, 185,276-285.

Bushweller, J. H., Billeter, M., Holmgren, A. &WuÈ thrich, K. (1994). The nuclear magnetic reson-ance solution structure of the mixed disul®debetween Escherichia coli glutaredoxin(C14S) and glu-tathione. J. Mol. Biol. 235, 1585-1597.

Creighton, T. E. (1986). Disul®de bonds as probes ofprotein folding pathways. Methods Enzymol. 131, 83-106.

Eklund, H., Ingelman, M., SoÈderberg, B. O., Uhlin, T.,Nordlund, P., Nikkola, M., Sonnerstam, U., Joelson,T. & Petratos, K. (1992). Structure of oxidizedbacteriophage T4 glutaredoxin (thioredoxin).Re®nement of native and mutant proteins. J. Mol.Biol. 228, 596-618.

Ellman, G. L. (1959). Tissue sulfhydryl groups. Arch. Bio-chem. Biophys. 82, 70-77.

Ernst, R. R., Bodenhausen, G. & Wokaun, A. (1987).Principles of Nuclear Magnetic Resonance in One andTwo Dimensions, International Series of Monographson Chemistry, 14, Clarendon, Oxford.

Gilbert, H. F. (1984). Redox control of enzyme activitiesby thiol/disul®de exchange. Methods Enzymol. 107,330-351.

Grauschopf, U., Winther, J. R., Korber, P., Zander, T.,Dallinger, P. & Bardwell, J. C. (1995). Why is DsbAsuch an oxidizing disul®de catalyst? Cell, 83, 947-955.

Grzesiek, S., Ikura, M., Clore, M., Gronenborn, A. M. &Bax, A. (1992). A 3D triple-resonance NMR tech-nique for qualitative measurement of carbonyl-Hb Jcouplings in isotopically enriched proteins. J. Magn.Reson. 96, 215-221.

Guddat, L. W., Bardwell, J. C., Glockshuber, R., Huber-Wunderlich, M., Zander, T. & Martin, J. L. (1997).Structural analysis of three His32 mutants of DsbA:support for an electrostatic role of His32 in DsbAstability. Protein Sci. 6, 1893-1900.

Guddat, L. W., Bardwell, J. C. A. & Martin, J. L. (1998).Crystal structures of reduced and oxidized dsba:investigation of domain motion and thiolate stabil-ization. Structure, 6, 757-767.

GuÈ ntert, P., DoÈ tsch, V., Wider, G. & WuÈ thrich, K.(1992). Processing of multi-dimensional NMR datawith the new software PROSA. J. Biomol. NMR, 2,619-629.

GuÈ ntert, P., Mumenthaler, C. & WuÈ thrich, K. (1997).Torsion angle dynamics for NMR structure calcu-lation with the new program DYANA. J. Mol. Biol.273, 283-298.

GuÈ ntert, P., Billeter, M., Ohlenschlager, O., Brown, L. R.& WuÈ thrich, K. (1998). Conformational analysis ofprotein and nucleic acid fragments with the newgrid search algorithm FOUND. J. Biomol. NMR, 12,543-548.

Holmgren, A., SoÈderberg, B. O., Eklund, H. & BraÈndeÂn,C. I. (1975). Three-dimensional structure of Escheri-chia coli thioredoxin-S2 to 2.8 AÊ resolution. Proc.Natl Acad. Sci. USA, 72, 2305-2309.

Holmgren, A., Arner, E. S. J. & Berndt, K. D. (1999).Glutaredoxin. In Encyclopedia of Molecular Biology(Creigthon, T. E., ed.), pp. 1020-1023, John Wiley &Sons, Inc. , New York.

HoÈoÈg, J. O., Holmgren, A., D'Silva, C., Douglas, K. T. &Seddon, A. P. (1982). Glutathione derivatives asinhibitors of glutaredoxin and ribonucleotide

NMR Structure of Oxidized E. coli Glutaredoxin 3 431

reductase from Escherichia coli. FEBS Letters, 138, 59-61.

Huber-Wunderlich, M. & Glockshuber, R. (1998). Asingle dipeptide sequence modulates the redoxproperties of a whole enzyme family. Fold. Des. 3,161-171.

Hutchinson, E. G. & Thornton, J. M. (1996). PROMOTIF- a program to identify and analyze structuralmotifs in proteins. Protein Sci. 5, 212-220.

Jeng, M. F., Campbell, A. P., Begley, T., Holmgren, A.,Case, D. A., Wright, P. E. & Dyson, H. J. (1994).High-resolution solution structures of oxidized andreduced Escherichia coli thioredoxin. Structure, 2,853-868.

Joelson, T., SjoÈberg, B. M. & Eklund, H. (1990). Modi®-cations of the active center of T4 thioredoxin bysite-directed mutagenesis. J. Biol. Chem. 265, 3183-3188.

Katti, S. K., LeMaster, D. M. & Eklund, H. (1990). Crys-tal structure of thioredoxin from Escherichia coli at1.68 AÊ resolution. J. Mol. Biol. 212, 167-184.

Katti, S. K., Robbins, A. H., Yang, Y. & Wells, W. W.(1995). Crystal structure of thioltransferase at 2.2 AÊ

resolution. Protein Sci. 4, 1998-2005.Koradi, R., Billeter, M. & WuÈ thrich, K. (1996).

MOLMOL: a program for display and analysis ofmacromolecular structures. J. Mol. Graph. Model. 14,51-55.

Krause, G., LundstroÈm, J., Barea, J. L., Pueyo de laCuesta, C. & Holmgren, A. (1991). Mimicking theactive site of protein disul®de-isomerase by substi-tution of proline 34 in Escherichia coli thioredoxin.J. Biol. Chem. 266, 9494-9500.

Laskowski, R. A., Rullmann, J. A., MacArthur, M. W.,Kaptein, R. & Thornton, J. M. (1996). AQUA andPROCHECK-NMR: programs for checking the qual-ity of protein structures solved by NMR. J. Biomol.NMR, 8, 477-486.

LuginbuÈ hl, P., GuÈ ntert, P., Billeter, M. & WuÈ thrich, K.(1996). The new program OPAL for moleculardynamics simulations and energy re®nements ofbiological macromolecules. J. Biomol. NMR, 8, 136-146.

Markley, J. L., Bax, A., Arata, Y., Hilbers, C. W.,Kaptein, R., Sykes, B. D., Wright, P. E. & WuÈ thrich,K. (1998). Recommendations for the presentation ofNMR structures of proteins and nucleic acids. J. Mol.Biol. 280, 933-952.

Messerle, B. A., Wider, G., Otting, G., Weber, C. &WuÈ thrich, K. (1989). Solvent suppression using aspin lock in 2D and 3D NMR spectroscopy withH2O solutions. J. Magn. Reson. 85, 608-613.

MoÈssner, E., Huber-Wunderlich, M. & Glockshuber, R.(1998). Characterization of Escherichia coli thiore-doxin variants mimicking the active-sites of otherthiol/disul®de oxidoreductases. Protein Sci. 7, 1233-1244.

Mueller, L. (1987). P.E.COSY, a simple alternative to E.COSY. J. Magn. Reson. 72, 191-196.

MuÈ ller, N., Ernst, R. R. & WuÈ thrich, K. (1986). Multiple-quantum-®ltered two-dimensional correlated NMRspectroscopy of proteins. J. Am. Chem. Soc. 108,6482-6492.

Nordstrand, K., AÊ slund, F., Holmgren, A., Otting, G. &Berndt, K. D. (1999a). NMR structure of Escherichiacoli glutaredoxin 3-glutathione mixed disul®de com-plex: implications for the enzymatic mechanism.J. Mol. Biol. 286, 541-552.

Nordstrand, K., AÊ slund, F., Meunier, S., Holmgren, A.,Otting, G. & Berndt, K. D. (1999b). Direct NMRobservation of the Cys-14 thiol proton of reducedEscherichia coli glutaredoxin-3 supports the presenceof an active site thiol-thiolate hydrogen bond. FEBSLetters, 449, 196-200.

Otting, G., Orbons, L. P. M. & WuÈ thrich, K. (1990). Sup-pression of zero-quantum coherence in NOESY andsoft-NOESY. J. Magn. Reson. 89, 423-430.

Qin, J., Clore, G. M. & Gronenborn, A. M. (1994). Thehigh-resolution three-dimensional solution struc-tures of the oxidized and reduced states of humanthioredoxin. Structure, 2, 503-522.

Qin, J., Clore, G. M., Kennedy, W. M., Huth, J. R. &Gronenborn, A. M. (1995). Solution structure ofhuman thioredoxin in a mixed disul®de intermedi-ate complex with its target peptide from the tran-scription factor NFkB. Structure, 3, 289-297.

Qin, J., Clore, G. M., Kennedy, W. P., Kuszewski, J. &Gronenborn, A. M. (1996). The solution structure ofhuman thioredoxin complexed with its target fromRef-1 reveals peptide chain reversal. Structure, 4,613-620.

Rance, M., Sorensen, O. W., Bodenhausen, G., Wagner,G., Ernst, R. R. & WuÈ thrich, K. (1983). Improvedspectral resolution in cosy 1H NMR spectra of pro-teins via double quantum ®ltering. Biochem. Biophys.Res. Commun. 117, 479-485.

Richardson, J. S. & Richardson, D. C. (1988). Amino acidpreferences for speci®c locations at the ends ofalpha helices. Science, 240, 1648-1652.

Saarinen, M., Gleason, F. K. & Eklund, H. (1995). Crystalstructure of thioredoxin-2 from Anabaena. Structure,3, 1097-1108.

Schirra, H. J., Renner, C., Czisch, M., Huber-Wunderlich,M., Holak, T. A. & Glockshuber, R. (1998). Structureof reduced DsbA from Escherichia coli in solution.Biochemistry, 37, 6263-6276.

Sodano, P., Xia, T. H., Bushweller, J. H., BjoÈrnberg, O.,Holmgren, A., Billeter, M. & WuÈ thrich, K. (1991).Sequence-speci®c 1H NMR assignments and deter-mination of the three-dimensional structure ofreduced Escherichia coli glutaredoxin. J. Mol. Biol.221, 1311-1324.

Srinivasan, U., Mieyal, P. A. & Mieyal, J. J. (1997). pHpro®les indicative of rate-limiting nucleophilic dis-placement in thioltransferase catalysis. Biochemistry,36, 3199-3206.

Sun, C. H., Berardi, M. J. & Bushweller, J. H. (1998). TheNMR solution structure of human glutaredoxin inthe fully reduced form. J. Mol. Biol. 280, 687-701.

Szyperski, T., GuÈ ntert, P., Otting, G. & WuÈ thrich, K.(1992). Determination of scalar coupling constantsby inverse Fourier transformation of in-phase multi-plets. J. Magn. Reson. 99, 552-560.

Vlamis-Gardikas, A., AÊ slund, F., Spyrou, G., Bergman,T. & Holmgren, A. (1997). Cloning, overexpression,and characterization of glutaredoxin 2, an atypicalglutaredoxin from Escherichia coli. J. Biol. Chem. 272,11236-11243.

Weichsel, A., Gasdaska, J. R., Powis, G. & Montfort,W. R. (1996). Crystal structures of reduced,oxidized, and mutated human thioredoxins:evidence for a regulatory homodimer. Structure, 4,735-751.

Xia, T. H., Bushweller, J. H., Sodano, P., Billeter, M.,BjoÈrnberg, O., Holmgren, A. & WuÈ thrich, K. (1992).NMR structure of oxidized Escherichia coli glutare-doxin: comparison with reduced E. coli glutaredoxin

432 NMR Structure of Oxidized E. coli Glutaredoxin 3

and functionally related proteins. Protein Sci. 1, 310-321.

Yang, Y., Jao, S., Nanduri, S., Starke, D. W., Mieyal, J. J.& Qin, J. (1998). Reactivity of the human thioltrans-ferase (glutaredoxin) C7S, C25S, C78S, C82S mutantand NMR solution structure of its glutathionyl

mixed disul®de intermediate re¯ect catalytic speci-®city. Biochemistry, 37, 17145-17156.

Zapun, A., Bardwell, J. C. & Creighton, T. E. (1993). Thereactive and destabilizing disul®de bond of DsbA, aprotein required for protein disul®de bond for-mation in vivo. Biochemistry, 32, 5083-5092.

Edited by P. E. Wright

(Received 27 June 2000; received in revised form 30 August 2000; accepted 30 August 2000)