Structural Basis for the Thioredoxin-like Activity Profile of the Glutaredoxin-like NrdH-redoxin from Escherichia coli*

NrdH-redoxin is a representative of a class of small redox proteins that contain a conserved CXXC motif and are characterized by a glutaredoxin-like amino acid sequence and thioredoxin-like activity profile. The crystal structure of recombinantEscherichia coli NrdH-redoxin in the oxidized state has been determined at 1.7 Å resolution by multiwavelength anomalous diffraction. NrdH-redoxin belongs to the thioredoxin superfamily and is structurally most similar to E. coliglutaredoxin 3 and phage T4 glutaredoxin. The angle between the C-terminal helix α3 and strand β4, which differs between thioredoxin and glutaredoxin, has an intermediate value in NrdH-redoxin. The orientation of this helix is to a large extent determined by an extended hydrogen-bond network involving the highly conserved sequence motif61WSGFRP(D/E)67, which is unique to this subclass of the thioredoxin superfamily. Residues that bind glutathione in glutaredoxins are in general not conserved in NrdH-redoxin, and no glutathione-binding cleft is present. Instead, NrdH-redoxin contains a wide hydrophobic pocket at the surface, similar to thioredoxin. Modeling studies suggest that NrdH-redoxin can interact with E. coli thioredoxin reductase at this pocket and also via a loop that is complementary to a crevice in the reductase in a similar manner as observed in the E. colithioredoxin-thioredoxin reductase complex.

Thioredoxins (Trx) 1 and glutaredoxins (Grx) are two classes of redox active proteins, which are members of the thioredoxin superfamily (1)(2)(3). Characteristics of this family are the common fold and a conserved sequence motif, CXXC, which contains the redox active cysteine residues. A distinctive feature between thioredoxins and glutaredoxins is the pattern of residues linking these two cysteine residues, i.e. CGPC for thioredoxin and CPYC for most glutaredoxins (3)(4)(5). These two ubiquitous small redox proteins (molecular mass, 9 -12 kDa) have been identified in numerous organisms and are, in their dithiol form, the major disulfide reductases in cells (2,6). They perform a large number of functions in cell growth, such as redox control of transcription factors (7), electron transport to ribonucleotide reductase (8), or defense against oxidative stress and apoptosis (6). Glutaredoxins are specifically reduced by glutathione (GSH), whereas thioredoxins depend on thioredoxin reductase (TrxR) for reduction.
Recently, members of a new class of small redox proteins, glutaredoxin-like proteins, NrdH-redoxin (NrdH), have been found in Escherichia coli and several other organisms, in particular organisms lacking GSH (9,10). NrdH proteins typically show sequence identities in the range of 34 -85% and are homologous to glutaredoxins. For instance, the Lactococcus lactis NrdH redoxin has 27% sequence identity to E. coli glutaredoxin-3 and 17% identity to glutaredoxin-1 (9). NrdHredoxin also contains the active site motif CXXC as in Trx and Grx, but the intervening residues (valine and glutamine) are different from both these classes. Although NrdH is related in amino sequence to glutaredoxins, it behaves functionally as a thioredoxin. It is not reduced by GSH but by thioredoxin reductase, it has a low redox potential, and it can reduce insulin disulfides; these biochemical features are characteristic of thioredoxin (10). The in vivo function of NrdH is not completely clear, but it can act as the functional hydrogen donor for class Ib ribonucleotide reductase and is part of a nrdHIEF operon (9 -11). Recently, E. coli nrdhHIEF mRNA levels were shown to be strongly enhanced in cells treated with oxidant (12).
To elucidate the structural basis of the thioredoxin-like activity profile of the glutaredoxin homologue NrdH, we have solved the crystal structure of E. coli NrdH to 1.7 Å by multiwavelength anomalous diffraction. Comparison of the threedimensional structure of NrdH with that of other redoxins suggests a hydrophobic pocket at the surface and a loop between helix ␣2 and strand ␤3 as common determinants in the recognition and binding of Trx and NrdH to thioredoxin reductase. The architecture of this hydrophobic pocket is completely different from the glutathione binding cleft in glutaredoxins, thus preventing binding of glutathione.

EXPERIMENTAL PROCEDURES
Sample Preparation-Plasmid pUA625, containing the gene for E. coli NrdH (10) was transformed into BL21-CodonPlus (Stratagene), and the cells containing pUA625 were grown as described (10). To produce the NrdH selenomethionine derivative, E. coli BL21-CodonPlus cells with the expression vector pUA625 were grown at 37°C in 1-liter flasks in M9 medium (13) containing 50 g/ml kanamycin. After ϳ4 -5 h, (A 578 ϭ 0.6), methionine biosynthesis was suppressed using an amino acid mixture as described (14) including L-selenomethionine (50 mg liter Ϫ1 ). After 15 min the culture was induced by adding isopropyl-␤-Dthiogalactopyranoside to a final concentration of 1.0 mM. The cells were harvested after 4 h by centrifugation, and the pellet (2.4 g) was stored at Ϫ20°C.
Purification of E. coli NrdH followed a protocol described elsewhere * This work was supported by the Swedish Medical Research Council. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Protein Analysis-The homogeneity of the purified proteins was analyzed with SDS gel electrophoresis. The protein concentration was determined with Coomassie Plus protein assay reagent (Pierce). The incorporation of four selenomethionine residues was confirmed by mass spectrometry at the Protein Analysis Center, Department of Medical Biochemistry and Biophysics, Karolinska Institute Stockholm.
Crystallization and Data Collection-Crystals of native and Se-methionine substituted E. coli NrdH were obtained by vapor diffusion. The native protein solution was dialyzed against 0.05 M CHES buffer, pH 9.1 and concentrated to 16 mg/ml. 3 l of protein solution was mixed with 3 l of mother liquor (1.20 M sodium citrate, 200 mM Tris/HCl, pH 7.9, 5% 2-methyl-2,4-pentane-diol). Hanging drops were equilibrated against the mother liquor at 20°C. Needles, up to 1 mm in length and 0.05 mm in diameter, were obtained after 1-2 weeks. Se-methioninesubstituted NrdH was dialyzed against 0.05 M CHES buffer, pH 9.1, 5 mM dithiothreitol and concentrated to 16 mg/ml. 2 l of the protein solution were mixed with 2 l of mother liquor (95 mM CHES, pH 8.7, 950 mM sodium citrate, 4.7% isopropyl alcohol) and equilibrated against 1 ml of mother liquor at 20°C. Thin plates of 0.02 ϫ 0.02 ϫ 0.05 mm were obtained after 1 week.
A native diffraction data set was collected at 100 K at beam line 711 of MAX-LabII (Lund, Sweden) with a MAR Research imaging plate. The MAD-diffraction data set was collected at 100 K at beam line BM14 of the European Synchrotron Radiation Facility (ESRF, Grenoble, France) with a MAR charge-coupled device detector. Before transfer into the nitrogen gas stream the crystals were soaked for a few seconds in a solution of 69 mM Tris/HCl, pH 7.9, 1.0 M sodium citrate, 17.3% glycerol, 0.045 M spermidine, 3.5% 2-methyl-2,4-pentanediol. All data sets were processed with DENZO and scaled with SCALEPACK (15), or the CCP4 suite of programs (16). The native crystals belong to space group P2 1 2 1 2 1 with cell dimensions of a ϭ 31.2 Å, b ϭ 41.3 Å, c ϭ 58.7 Å and contain one molecule per asymmetric unit. The selenomethionine crystals belong to space group P2 1 with cell dimensions a ϭ 41.2 Å, b ϭ 29.9 Å, c ϭ 55.4 Å, ␤ ϭ 89.6°and two molecules per asymmetric unit.
Phasing, Model-Building, and Refinement-Data for MAD phasing were collected at three different wavelengths to 2.9 Å resolution (Table  I). Six selenium sites were identified using the program SOLVE (17). The positions and occupancies of the selenium sites, as well as values for fЈ and fЈЈ for two wavelengths were refined further in SHARP (18). Density modification and NCS-averaging to 3 Å in the program DM (19) resulted in a well defined electron density map used for initial tracing of the polypeptide chain. The protein model was built using the graphics program O (20) and refined to 2.9 Å resolution with CNS (21), using the experimental phases and tight non-crystallographic symmetry restraints. This model was used for molecular replacement and subsequent refinement to 1.7 Å resolution using the native data set (Table I).
Molecular replacement was done with the program AMORE (22), which gave an R free of 28.3%. The subsequent refinement was done using CNS (21). The values for R cryst and R free after refinement were 20.0 and 21.5%, respectively, with the test set used for R free calculation containing 10% of the reflections. Model geometry was analyzed with PROCHECK (23). Further details of the refinement statistics are given in Table II.
Modeling of the complex of NrdH with TrxR was carried out with O, using the coordinates for the Trx-TrxR complex, accession code 1F6M (24). Structure comparison and structural alignments were carried out with the programs TOP and MAPS using default parameters (25). The figures were produced using the programs Grasp (26), O (20), and Bobscript (27). The coordinates and structure factors have been submitted to the Protein Data Bank, accession code 1H75 and r1H75sf, respectively.

RESULTS
Overall Structure-The refined model of NrdH at 1.7 Å resolution contains residues 1-76 of the protein and 82 water molecules. The polypeptide chain is well defined in the electron density map (Fig. 1), except for the last five residues, which were not visible in the electron density map, probably because of disorder. The side chain of Arg 35 has been modeled in two different conformations, with relative occupancies of 0.7 and 0.3, respectively.
NrdH is a compact molecule with overall dimensions of 18 ϫ 24 ϫ 37 Å and a surface of 5141 Å 2 . The structure consists of a four-stranded ␤-sheet with topology Ϫ1x, ϩ2x, ϩ1 and three flanking helices, a fold that is characteristic of the thioredoxin superfamily (3) (Fig. 2). Helix ␣2 runs roughly perpendicular to helices ␣1 and ␣3. A cis-proline (Pro 52 ), found at the beginning of strand ␤3, is highly conserved among proteins of the thioredoxin family.
A NrdH-specific Sequence Motif-In NrdH-like proteins the chain comprising the end of ␤4, the following loop, and the beginning of helix ␣3 contains a highly conserved sequence motif, 61 WSGFRP(D/E) 67 (Fig. 3A). A pattern scan in Swiss-Prot and TrEMBL showed that this pattern is unique to NrdH proteins. The position of helix ␣3 relative to the remaining part of the molecule is to a large extend determined by an intricate network of hydrogen bonds formed by the WSGFRP(D/E) motif, two water molecules, and surrounding protein residues (Fig. 3). A central role is played by a water molecule, which forms Additional allowed regions (%) 9.1 Disallowed regions (%) 0.0  (Table III). These proteins all belong to the thioredoxin superfamily and share the thioredoxin fold. Although NrdH does not bind GSH, the overall structure is most similar to Grx and is less similar to Trx (Table III).
A comparison of NrdH with the structures of Grx and Trx from E. coli shows that despite the high overall similarity, all three proteins have distinctive features, which are typical for their subclass (Fig. 4). The N-terminal part is structurally most similar among all three proteins. Strands ␤1, ␤2, and ␤3, as well as helix ␣1 and the N termini of the helices ␣2 align well among the three proteins (Fig. 4). The C-terminal part is structurally less conserved. The most striking difference between these structures is the orientation of helix ␣3 in relation to strand ␤4 (Fig. 4B). Trx has a rather long connection between strand ␤4 and helix ␣3, and ␣3 is almost antiparallel to the strand. In Grx the chain instead immediately adopts helix conformation after strand ␤4 and this helix is then approximately perpendicular to the helix in Trx. In NrdH the connecting loop is longer than in Grx but shorter than in Trx resulting in a helix orientation in between that of Trx and Grx3 (Fig. 4B).
Trx contains one additional ␤-strand and one helix at the N terminus, which are not present in NrdH and Grx (29,33). Trx also has a longer helix ␣1 and a shorter helix ␣2 than NrdH and Grx. The loop from helix ␣2 to strand ␤3 in Trx (residues 70 -75) occupies a complementary groove on the surface of thioredoxin reductase (24) upon binding. The longer helix ␣2 of NrdH shifts the position of this loop compared with its location in Trx (Fig. 4). Strand ␤4 and the following loop to helix ␣3 are more similar between NrdH and Trx than between NrdH and Grx3 (Fig. 4B). At the tip of strand ␤4 is a glycine residue (position 63 in NrdH, and 92 in E. coli Trx), which is conserved in the superfamily (34). This residue has been shown to be important in binding of Trx to other molecules, such as T7 DNA polymerase, TrxR (35), or in the assembly of filamentous phages (36).
Archaebacteria contain small redox proteins of as yet unknown function, which are similar in sequence to the glutaredoxins (19 -30% sequence identity), but do not react with glutathione (37,38). These redoxins are also related in amino acid sequence to NrdH proteins, with about 20% sequence identities, and it has therefore been suggested that they may form a common subgroup in the thioredoxin superfamily (38). Comparison of the three-dimensional structure of NrdH with the solution structure of the reduced form of an archaebacterial redoxin, Mj0307 from Methanococcus jannaschii (38), reveals that the structures are not as similar as one might expect for members of the same subgroup. In fact, superposition gives a r.m.s. of 2.0 Å with 45 equivalent C␣ atoms, significantly worse than superposition of NrdH with Grx or Trx (Table III). Furthermore, these archaebacterial sequences do not contain the 61 WSGFRP(D/E) 67 sequence motif typical of the mesophilic NrdH redoxins. These differences suggest that the two proteins are not members of one subgroup but represent different sub-groups of the thioredoxin family.
The Active Site-The active site of E. coli NrdH is located at the beginning of helix ␣1. The electron density map clearly shows formation of a disulfide bond between residues Cys 11 and Cys 14 , and the structure of NrdH described here is thus that of the oxidized protein.
The N-terminal cysteine of the CXXC motif provides the reactive thiolate in the reduced protein (2,39,40) and is more solvent accessible whereas the C-terminal cysteine side chain is buried. Although the active sites of NrdH, Trx, and Grx are quite similar, NrdH is more similar to Trx than to Grx. In particular, the , , and angles for the first cysteine of the active site are more similar to Trx than to Grx (Table IV).
Compared with Grx the angle shows a difference of about 40°b ut only 18°difference compared with Trx. The angle of NrdH is 108°and is similar to E. coli Trx (108°), whereas the glutaredoxins have angles about 86 -96°. Also in the angle, NrdH and Grx differs about 18 -24°, whereas NrdH and E. coli Trx have an identical 164° angle. In all compared redoxins, the bond angles of the second active site cysteine are more similar to each other than the bond angles of the first cysteine (Table  IV). We find no obvious correlation between the conformational angles of the first cysteine and the reduction potential. T4 Grx with E°Јϭ Ϫ240 mV has the most similar reduction potential to NrdH (E°Јϭ Ϫ248.5 mV), but differs most in the active site cysteine angles. However, T4 Grx, like NrdH, has a valine adjacent to the first cysteine residue in the CXXC motif. This is consistent with observations that the residues between the two active site cysteines are important determinants of the redox potential (41).
Surface Topology and Glutathione Binding Site-Whereas the first cysteine of the active site, Cys 11 , is surface accessible, Cys 14 lies in a groove formed by the residues Arg 8 , Cys 11 , and Gln 13 . NrdH has a deep hydrophobic pocket next to the active site; Val 54 and Ile 55 form the bottom of this cavity, surrounded by the hydrophobic residues Val 33 , Leu 43 , Phe 48 , and Tyr 6 (   (Fig. 5). The hydrophobic pocket of Trx is not surrounded by as many positively charged residues as observed in NrdH, only one positively charged surface residue, Arg 73 , is structurally conserved and aligns to Arg 49 in NrdH. Instead of a hydrophobic pocket, Grx has a narrow groove where GSH is binding. Tyr 13 , Thr 50 , Thr 51 , Val 52 , and Arg 40 form the walls of a bent cleft, which contains the glutathione binding site (42,43). As in the hydrophobic pocket of NrdH, arginines, Arg 16 , Arg 40 , and Arg 49 (Fig. 5) surround the channel. In NrdH, side chains block part of this cleft, in particular by Gln 50 . Residues interacting with glutathione in Grx are in general not conserved in NrdH (Fig. 4a).
Interaction of NrdH with E. coli TrxR-Superposition of NrdH on E. coli Trx in the Trx-TrxR complex (24) gives a model that shows only few steric clashes between the two molecules. These occur in the area of the loop that connects helix ␣2 and strand ␤3 in NrdH, and are because of the longer helix ␣2 in NrdH compared with Trx. However, a small rotation of NrdH will dock this loop in a groove on TrxR and at the same time allow interaction of TrxR with the hydrophobic pocket in NrdH, analogous to the Trx-TrxR interaction (Fig. 6). Phe 141 and Phe 142 of TrxR fit very well into this hydrophobic pocket. The distance between Cys 11 of NrdH and Cys 138 of TrxR in this docking model is 4.3 Å but as in the Trx-TrxR complex, the loop containing Cys 11 could bend toward TrxR, enabling the cysteine residues to form a disulfide bond. DISCUSSION Structural analysis of NrdH has shown that it lacks the glutathione binding site seen in the glutaredoxins. The architecture of the surface, which in glutaredoxins defines the binding site of GSH, is completely different, and residues interacting with the electron donor are also not conserved between the two subfamilies of redoxins. NrdH is thus unable to accept GSH as a substrate.
The interaction of thioredoxin with thioredoxin reductase is characterized by three major features: (i) the docking of two phenylalanine residues from TrxR into a hydrophobic pocket of Trx close to the redox active cysteine, (ii) binding of the loop between helix ␣2 and strand ␤3 (nomenclature according to Fig. 4) of Trx into a groove on the surface of TrxR and (iii) interaction of the redox active cysteines enabling transfer of reducing equivalents (24). A striking observation is the pres-  (24). TrxR is colored from blue at the N terminus to yellow at the C terminus. NrdH is shown in red. Right, close-up of the major interface area of NrdH with TrxR, showing the hydrophobic pocket at the surface of NrdH, and the loop of TrxR carrying the two phenylalanine residues that bind in this cavity. ence in NrdH of a large hydrophobic pocket, at the same position as the corresponding pocket in Trx. Modeling experiments showed that it is possible to dock NrdH onto TrxR in a very similar way as observed in the TrxR-Trx complex (24), with the essential features of redoxin-reductase interactions preserved. It thus appears that the architecture of the hydrophobic pocket and the protruding loop, formed mainly by residues 43-55 at the surface of NrdH, are the major determinants of the substrate specificity of this protein and at least in part define the thioredoxin-like activity profile. NrdH is part of the NrdHIEF operon, where NrdEF codes for a class Ib ribonucleotide reductase (9). The NrdEF genes appear not to be expressed under standard growth conditions (44). However, recent results show that basal transcript levels of the NrdHIEF mRNA are strongly enhanced depending on growth medium and phase (12). Furthermore, addition of oxidants also results in strong up-regulation of NrdHIEF transcription suggesting functions under stress conditions. NrdH is a better hydrogen donor for the NrdEF ribonucleotide reductase than for the NrdAB enzyme. The class Ib RNR is efficient at low oxygen levels (9) and it has been suggested that this additional aerobic enzyme could function when the oxygen availability is too low to activate class Ia but high enough to inactivate class III (11). Under microaerophilic conditions most of GSH is conjugated with spermidine (45) and not available for the cell. NrdH lacks a glutathione binding site and instead has a hydrophobic pocket to bind TrxR and can thus function independent of the cellular glutathione status. Thus, under these conditions, type Ib RNR in combination with NrdH is a much more efficient enzyme than type Ia RNR (NrdAB). The NrdEF-NrdH system may therefore be the evolutionary oldest RNR reducing system, capable of functioning in a microaerophilic environment, where glutathione was not yet available.
Given this scenario, it has been proposed that NrdH and similar redox proteins may be related to the progenitor of the thioredoxin family (10). Grx and NrdH proteins are among the structurally simplest members of the Trx superfamily, in fact the common fold motif is a NrdH-redoxin or glutaredoxin rather than a thioredoxin fold. Both in a structural and functional sense, NrdH can be described as a hybrid between the thioredoxins and glutaredoxins and it is conceivable that other members of this family emerged by divergent evolution from a NrdH-like ancestor.