Characterization of Escherichia coli NrdH

Ribonucleotides are converted to deoxyribonucleotides by ribonucleotide reductases. Either thioredoxin or glutaredoxin is a required electron donor for class I and II enzymes. Glutaredoxins are reduced by glutathione, thioredoxins by thioredoxin reductase. Recently, a glutaredoxin-like protein, NrdH, was isolated as the functional electron donor for a NrdEF ribonucleotide reductase, a class Ib enzyme, from Lactococcus lactis. The absence of glutathione in this bacterium raised the question of the identity of the intracellular reductant for NrdH. Homologues of NrdH are present in the genomes of Escherichia coli and Salmonella typhimurium, upstream of the genes for the poorly transcribed nrdEF, separated from it by an open reading frame (nrdI) coding for a protein of unknown function. Overexpression of E. coli NrdH protein shows that it is a functional hydrogen donor with higher specificity for the class Ib (NrdEF) than for the class Ia (NrdAB) ribonucleotide reductase. Furthermore, this glutaredoxin-like enzyme is reduced by thioredoxin reductase and not by glutathione. We suggest that several uncharacterized glutaredoxin-like proteins present in the genomes of organisms lacking GSH, including archae, will also react with thioredoxin reductase and be related to the ancestors from which the GSH-dependent glutaredoxins have evolved by the acquisition of a GSH-binding site. We also show that NrdI, encoded by allnrdEF operons, has a stimulatory effect on ribonucleotide reduction.

Ribonucleotides are reduced to deoxyribonucleotides by ribonucleotide reductases, which are radical containing enzymes that may be divided into three main classes (1,2).
The electrons for this reaction are supplied by small redoxactive proteins such as thioredoxin (Trx) 1 or glutaredoxin (Grx) in the case of class I and II ribonucleotide reductase (3,4), whereas formate fulfills this function for the anaerobic class III enzyme (5). Thioredoxin and glutaredoxin both contain two redox-active cysteine thiols in their reduced form, which by dithiol-disulfide interchange reduce an acceptor disulfide in the active center of ribonucleotide reductase. The active site sequences of thioredoxins and glutaredoxins are conserved among species, being Cys-Gly-Pro-Cys for thioredoxin and Cys-Pro-Tyr-Cys for glutaredoxin (3,4). The disulfide in oxidized thioredoxin is regenerated to a dithiol by thioredoxin reductase (TR) and NADPH, whereas oxidized glutaredoxin is reduced by 2 mol of GSH with the formation of GSSG, which is reduced by glutathione reductase (GR) and NADPH.
Escherichia coli contains three different glutaredoxins (called Grx1, -2, and -3 (6)), which, like all glutaredoxins from other species, show high activity as general GSH-disulfide oxidoreductases in a coupled system with GSH, NADPH, and glutathione reductase (7). Three-dimensional structures for thioredoxins (8) and glutaredoxins (9,10) show that they have essentially completely unrelated amino acid sequences but a similar overall fold (often referred to as the thioredoxin fold), consisting of a central four-stranded ␤-sheet flanked by three helices in the order ␤␣␤␣␤␤␣ (the alignment in Fig. 1 includes the location of secondary structure elements in glutaredoxins). Within the thioredoxin superfamily of proteins (for review, see Ref. 11) two distinct subtypes are the thioredoxin proteins and the glutaredoxin proteins, the latter having a conserved GSHbinding site (boxed in Fig. 1), which has been experimentally determined by NMR (12) for E. coli Grx1.
E. coli and Salmonella typhimurium contain the genetic information for two different ribonucleotide reductases belonging to class I, NrdAB (class Ia) and NrdEF (class Ib), with a limited sequence similarity (13) and differences in their allosteric regulation (14). The expression of NrdEF is repressed and insufficient to allow growth of NrdAB-defective cells under aerobic conditions, unless the expression of NrdEF is increased by the presence of additional copies of the nrdEF genes either on the chromosome or on a plasmid (13,15). Characterization of the proteins encoded by the S. typhimurium nrdEF genes showed a fully functional ribonucleotide reductase that could use E. coli Grx1 (but not thioredoxin) as a hydrogen donor with an apparent K m of 5 M (16) (cf. 0.15 M for NrdAB; Ref. 17).
Recently, it was shown by enzyme fractionation that the functional ribonucleotide reductase in the Gram-positive bacterium Lactococcus lactis grown under microaerophylic conditions is an NrdEF type of enzyme and that its hydrogen donor protein (NrdH, 72 amino acids) with the active site sequence Cys-Met-Gln-Cys shows appreciable homology to glutaredoxins (18). Coding sequences for proteins homologous to NrdH are present upstream of the nrdEF genes in E. coli and S. typhimurium (81 amino acids), separated by an open reading frame (Orf2; 136 amino acids). Since we have found that this latter protein seems to be involved in reduction of ribonucleotides, the protein will henceforth be referred to as NrdI. An alignment of NrdH and NrdI proteins from different species are presented in Figs. 1 and 2. The four genes, nrdH-nrdI-nrdE-nrdF, constitute a unique transcriptional unit (15). On the assumption that the gene organization was the same in L. lactis, it was possible to clone the analogous operon for the L. lactis NrdEF enzyme including the gene (nrdH) for the hydrogen donor protein and nrdI (18). Since L. lactis has no GSH and the NrdH proteins also appeared to lack the typical GSH-binding site identified in Grx1 (12), we wondered how NrdH is reduced. To solve this question we have cloned and expressed the NrdH protein from E. coli, an organism with a high content of GSH. Characterization of the NrdH protein showed that it lacked activity with GSH but was a substrate for thioredoxin reductase. Additionally, and with the aim of completing the understanding of the function of all the genes that constitute the conserved nrdEF operon, we have overexpressed the NrdI protein from S. typhimurium and investigated its influence on the activity of the NrdEF reductase.

EXPERIMENTAL PROCEDURES
Materials-Plasmids pET-24a and pET-15b were from Novagen, and pGEM-T was from Promega Corp. E. coli DH5␣ (CLONTECH) was used for general cloning procedures. BL21(DE3) was from Novagen. Taq DNA polymerase and T4 DNA ligase were from Pharmacia Biotech Inc. Oligonucleotide primers were from the Department of Cell and Molecular Biology, Karolinska Institute. Sephadex G-50 resin was from Pharmacia, and DEAE-cellulose (DE-52) anion exchanger was from Whatman. Ni 2ϩ -NTA agarose was from Qiagen. E. coli Grx1, Trx, and TR were obtained as described previously (3,19,20), and S. typhimurium NrdEF proteins (R1E and R2F) were purified as described (16). E. coli NrdAB RR subunits (R1 and R2) were available from this laboratory. The University of Wisconsin Genetics Computer Group (GCG) package (version 8.0, Open VMS) was used for sequence analysis.
Overexpression of E. coli NrdH-The entire gene for E. coli NrdH was amplified by PCR using plasmid pUA523 (15) as a template and two primers containing restriction sites for NdeI and BamHI, respectively (underlined): 5Ј-ATACGACATATGCGCATTACTATTTA-3Ј and 5Ј-ACAGGGATCCTCATGCACTGGCCGCG-3Ј (antisense). Thirty cycles of amplification were used with annealing at 55°C. The amplified DNA fragment was gel-purified and cloned into vector pGEM-T according to the manufacturer's protocol, giving rise to plasmid pUA624. To ascertain that no Taq polymerase-induced mutations were introduced, the cloned fragment was sequenced with fluorescent pUC/M13 universal primers, and the sequence was determined using an automated laser fluorescent DNA sequencer (Pharmacia). pUA624 was digested with NdeI and BamHI, and the nrdH coding sequence was cloned into the expression vector pET-24a digested with the same restriction enzymes, downstream from the inducible T7 promoter and a strong ribosome binding site. Plasmid (pUA625) from one clone unambiguously confirmed to code for NrdH was transformed into BL21(DE3) cells, which carry an IPTG-inducible T7 RNA polymerase gene.
Overexpression of S. typhimurium NrdI-The S. typhimurium nrdI gene was amplified using plasmid pUA335 (13) as a template and two primers containing restriction sites for NdeI and BamHI, respectively (underlined): 5Ј-CGACATATGAGCGCGCTCGTCTAC-3Ј and 5Ј-CGCGGATCCATGGTTTCCTGC-3Ј (antisense). The same general procedure described above was used for cloning the nrdI gene into vectors pGEM-T (pUA626) and pET-15b (pUA627), consecutively. The cloning into NdeI-BamHI-digested pET-15b introduces a sequence coding for six His codons upstream the nrdI gene, facilitating purification of the recombinant protein on Ni 2ϩ -NTA-agarose.
Purification of Overexpressed NrdH Protein-E. coli BL21(DE3) cells containing the pUA625 plasmid were grown in LB medium at 37°C in the presence of 50 g/ml of kanamycin to an optical density at 600 nm of 0.5 and subsequently induced with 0.4 mM IPTG (final concentration). Four hours after induction, the cells were harvested by centrifugation. The bacterial pellet (4 g) was dissolved in 6 volumes of 50 mM Tris-HCl, 1 mM EDTA, pH 7.5, and lysed by a combination of lysozyme (0.2 mg/ml) and sonication. The crude extract obtained after centrifugation was dialyzed extensively against 20 mM Tris-HCl, 1 mM EDTA, pH 9.5, and applied to a 100-ml column of DEAE-cellulose equilibrated with the same buffer. The column was eluted with 1 liter of 50 mM Tris-HCl, 1 mM EDTA, pH 8.0. The eluted protein was concentrated by ultrafiltration using a YM-3 membrane (Amicon) after the addition of 10% glycerol and 0.5 M NaCl (final concentrations). The concentrated protein was applied to a column (100 ϫ 3 cm 2 ) of Sephadex G-50 equilibrated with 50 mM Tris-HCl, pH 8.0, 10% glycerol, and 0.5 M NaCl. The eluted protein was concentrated and stored at Ϫ20°C.
Purification of Overexpressed NrdI Protein-E. coli BL21(DE3) cells harboring plasmid pUA627 were grown in LB medium at 30°C in the presence of 50 g/ml ampicillin to an optical density of 0.5 and subsequently induced with 0.4 mM IPTG for 3 h. After centrifugation and extraction by a combination of lysozyme (0.1 mg/ml) and sonication in 5 volumes of binding buffer (20 mM Tris-HCl, pH 8.0, 0.5 M NaCl, 5 mM imidazole, 0.5 mM CHAPS, 20% glycerol, 0.1% Nonidet P-40, 10 mM ␤-mercaptoethanol), the soluble crude extract was loaded into a Ni 2ϩ -NTA agarose column according to the manufacturer's protocol. The column was washed with the same buffer containing 60 mM imidazole, and NrdI was eluted with 300 mM imidazole. Prior to use, the sample was dialyzed to remove NaCl, imidazole, and ␤-mercaptoethanol.
Determination of Protein Concentration and N-terminal Sequence Analysis-The concentration of the purified NrdH protein was calculated from the absorption at 280 nm using a molar extinction coefficient of 7210 or by Bradford determinations (21) and total amino acid composition. As a complement, the concentration of NrdH was determined from the amount of NADPH consumed upon the addition of TR. The sequence of the first 5 amino acids of the purified NrdH protein was determined by Edman degradation using an ABI 491 Procise instrument.
Assays with Ribonucleotide Reductase-The ability of Grx1, NrdH, and Trx to serve as hydrogen donors for the NrdEF and AB reductases was determined using the standard ribonucleotide reductase assay (16,22). Stoichiometric amounts of R1E/R2F or R1/R2 were incubated for 20 min at 30°C with the chosen reductant system in a final volume of 0.05 ml containing 0.5 mM [ 3 H]CDP (22 cpm/pmol), 10 mM MgCl 2 , 0.1% Nonidet P-40, 50 mM Tris-HCl, pH 8.0, and 0.3 mM dATP (for NrdEF) or 1.5 mM ATP (for NrdAB). The activity of NrdH was determined in the presence of either of the three different reducing systems: (i) 4 mM GSH, 1 mM NADPH, 0.1 M glutathione reductase (ii) 1 mM DTT, or (iii) 1 mM NADPH, 0.1 M thioredoxin reductase (final concentrations). The activity of Grx1 was determined in the presence of 4 mM GSH, 1 mM NADPH, 0.1 M glutathione reductase, and the activity of Trx was determined in the presence of 1 mM NADPH, 0.1 M thioredoxin reductase. One unit is defined as the formation of 1 nmol of dCDP/min. For simplicity of comparison, the data for NrdH and Grx1 have been presented as percentages of the extrapolated V max obtained from Lineweaver-Burk plots (except for the data of Grx1 with NrdAB, where the last data point was set to 100%). Since Trx was inactive with NrdEF, the activity in Fig. 3C is presented in units (as specified above).
Assays with Thioredoxin Reductase-The reduction of NrdH, T4-Grx, and Trx by thioredoxin reductase (0.1 M final concentration) was determined in the presence of 0.5 mM of DTNB, 240 M NADPH, and 0.1 mg/ml bovine serum albumin in 100 mM Tris-HCl, pH 8.0, 2 mM EDTA in a final volume of 100 l. The activity was monitored as the increase of absorbance at 405 nm using a Thermomax microplate reader (Molecular Devices).
Determination of the Redox Potential for NrdH-This determination was performed essentially as described (23,24) using an AVIV spectrophotometer model 14DS (AVIV, Lakewood, NJ) at 25°C. Absorbance at 340 nm was determined as the average of 150 data points recorded during 30 s after each step in the experiment. The assays were performed in a degassed and N 2 -equilibrated buffer composed of 100 mM potassium phosphate, 1 mM EDTA, pH 7.0. In a typical experiment, additions of 6 -10 M NADPH (final concentration) were made to the sample cuvette, and the absorbance at 340 nm was used to determine the concentration in the cuvette, followed by the addition of the same amount of NADPH to the reference cuvette. An equivalent volume of NrdH protein (final concentration, 6 M) or buffer (10% glycerol, 0.5 M NaCl in 50 mM Tris-HCl, pH 8.0) was then added to the sample and reference cuvettes, respectively. Equilibration between NrdH and NADPH was allowed to start by the addition of 10 nM (final concentration) of thioredoxin reductase to both cuvettes. The equilibrium was then reversed by two successive additions of NADP ϩ to final concentrations of 125 and 250 M. As a control, the redox potential of Trx (Ϫ270 mV) was determined with the same procedure.
Reduction of Insulin Disulfides-Reduction of insulin disulfides (50 M final concentration) was performed in parallel for Grx1, NrdH, T4-Grx, and Trx using a final concentration of 5 M of each enzyme in the presence of 1 mM DTT in 100 mM potassium phosphate, 1 mM EDTA, pH 6.0. The reaction was followed as the increase of absorbance at 600 nm due to the precipitation of insulin when reduced to A and B chains (25). The activity of NrdH, T4-Grx, and Trx (final concentration of each enzyme, 5 M) as reductants of insulin disulfides (100 M final concentration) was also determined in the presence of 200 M NADPH, 0.1 M thioredoxin reductase in 100 mM potassium phosphate, 1 mM EDTA, pH 7.0. The activity was in this case monitored as the decrease in absorbance of NADPH at 340 nm.
GSH-Disulfide Oxidoreductase Assays-GSH-disulfide oxidoreductase assays were performed as described (17), measuring the reduction of ␤-hydroxyethyl disulfide by GSH at the expense of NADPH as monitored at 340 nm. A standard of purified E. coli Grx1 was used in each experiment as a positive control.

RESULTS
Overexpression and Purification of the E. coli NrdH Protein-The gene for E. coli NrdH was amplified by PCR using primers with designed restriction sites. The PCR product was cloned into the T7 RNA polymerase-dependent expression vector pET-24a. Using this system, the NrdH protein was expressed to levels of around 30% of total soluble protein, as judged by densitometry scans of SDS-polyacrylamide gels. The M r of the overexpressed polypeptide (9 kDa) was in accordance with the expected size for NrdH. The protein was purified to homogeneity but showed poor solubility. Thus, to avoid precipitation of the protein when concentrated, the final steps of the purification were done in a buffer containing 10% glycerol and 0.5 M NaCl. Purified NrdH protein was subjected to N-terminal amino acid sequence determination, confirming the homogeneity of the protein and also showing an unprocessed initiator Met residue as expected.
The sequence homology between NrdH proteins and glutaredoxins has been noted, with E. coli Grx3 being the closest glutaredoxin homologue (18). A prediction of the secondary structure elements of E. coli NrdH using the PHD program (26) produced a similar pattern of secondary structure elements (with the exception of ␤4) as experimentally determined for Grx1 and Grx3 (9,27). Fig. 1 shows a refined alignment where the predicted secondary structure elements of E. coli NrdH have been matched against the known secondary structure elements of E. coli Grx3 (27). The presence of a proline (Pro-52 in E. coli NrdH) in the NrdH proteins in the same relative position as the conserved cis-proline in glutaredoxins reemphasizes the structural similarity between NrdH proteins and glutaredoxins. The conclusion that NrdH proteins lack the conserved GSH-binding site found in glutaredoxins is still valid in this refined alignment. It is noteworthy that E. coli NrdH may be aligned with E. coli Grx3 with essentially no introduction of gaps. This could provide a basis for further studies of E. coli NrdH intended to introduce a GSH-binding site by changing the relevant amino acids of NrdH to the corresponding ones found in the GSH-binding site of Grx3, which has been shown to be essentially identical to the GSH-binding site of Grx1. 2 Overexpression and Purification of the S. typhimurium NrdI Protein-Since we have found that a sequence homologous to nrdI found between nrdH and nrdE in E. coli and S. typhimurium (15) is present in all known nrdEF operons (e.g. L. lactis (18), Mycoplasma genitalium (28), and Bacillus subtilis (29)), one should suspect a function for this unknown protein in the ribonucleotide reduction reaction. Searching for functional similarities of NrdI in the protein sequence data bases was unsuccessful. Fig. 2 shows the predicted amino acid sequence alignment of all known nrdI products. To investigate the effect of NrdI in the in vitro assay of S. typhimurium NrdEF ribonucleotide reductase, we overexpressed and partially purified the recombinant NrdI protein of this bacterium.
After cloning the PCR-amplified nrdI into the vector pET-15b, the His-tagged NrdI protein was expressed to levels of around 30% of total protein. Only a small fraction of the overexpressed protein was found in the soluble fraction of the crude extract. This soluble NrdI material was purified by one-step chromatography on Ni 2ϩ -NTA resin, resulting in material that was about 50% pure. The size of the recombinant protein was higher (17 kDa) than the size of the expected NrdI polypeptide (15 kDa) due to the His tag. Any attempt to increase the 2 K. Nordstrand, unpublished data.
FIG. 1. Alignment of glutaredoxin-like enzymes of the NrdH type with classical glutaredoxins. The sequences were obtained from the following sources: NrdH from E. coli, S. typhimurium, and L. lactis (18); Grx from phage T4 (43); Grx1 from E. coli (34); Grx3 from E. coli (27). The secondary structure of Grx3 determined by NMR (27) is shown at the bottom of the alignment (dark lines), and the predicted secondary structure of E. coli NrdH obtained as described (26) is shown at the top of the alignment (stippled lines). The sequences have been aligned so that the predicted secondary structure elements of E. coli NrdH match the determined secondary structure elements of E. coli Grx3. The residues that constitute the GSH-binding site of E. coli Grx1 are boxed. Residues conserved both in NrdH proteins and in glutaredoxins are on a gray background.
FIG. 2. Predicted amino acid sequence alignments of all known nrdI products. The sequences from E. coli (Ec), S. typhimurium (St), and L. lactis (Ll) were obtained by the authors (13,15,18). The M. genitalium (Mg) sequence comes from its genome sequence determination (28), and that of B. subtilis (Bs) is from Ref. 29. The size (in amino acids) of each product is indicated as well as the percentage of identical/similar amino acids with respect to the E. coli sequence. Consensus is shown by capital letters (complete conservation among all five proteins) and by lowercase letters (conservation in all but one of the sequences). The alignment has been generated with the program PILEUP of the GCG package. solubility of the recombinant NrdI during growth or extraction (e.g. by decreasing the growth temperature or IPTG concentration or by additions of NaCl, glycerol, detergents, or reducing agents such as DTT) was unsuccessful. NrdI could be purified in the presence of guanidinium hydrochloride or urea, but the protein precipitated when the denaturing agent was removed by dialysis.
Activity of NrdH and NrdI with the NrdAB and NrdEF Ribonucleotide Reductases-The activity of NrdH was compared with that of E. coli Grx1 and Trx as reductants of the NrdAB enzyme from E. coli and the NrdEF enzyme from S. typhimurium. NrdH was found to be a functional hydrogen donor for both enzymes in the presence of either 1 mM DTT or thioredoxin reductase/NADPH but not in the presence of GSH/ glutathione reductase/NADPH. The V max of NrdH (in units/g of the R1 or R1E subunits) was similar to that of Grx1 and Trx with the NrdAB enzyme and similar to the V max of Grx1 with the NrdEF enzyme. As shown in Table I and Fig. 3A, NrdH showed a lower K m value with the S. typhimurium NrdEF enzyme than with the E. coli NrdAB enzyme, whereas the opposite was found to be the case for E. coli Grx1 (Fig. 3B). This tendency for NrdH was even more pronounced when the assays were performed in the presence of 1 mM DTT instead of thioredoxin reductase/NADPH. Thus, the apparent K m value of NrdH as a hydrogen donor for NrdEF was repeatedly found to be lower (0.3-0.6 M) in the presence of 1 mM DTT than in the presence of NADPH/thioredoxin reductase (K m ϭ 1.2 M). This finding is hard to reconcile with the finding that NrdH is efficiently reduced by thioredoxin reductase (see below). Nevertheless, the results clearly demonstrate that NrdH is a more specific hydrogen donor for NrdEF than for NrdAB, whereas the opposite is the case for Grx1. E. coli Trx was a hydrogen donor for NrdAB but not for NrdEF (Fig. 3C) as observed previously (16).
The partially purified NrdI recombinant protein stimulated the NrdEF ribonucleotide reductase activity when the NrdH protein was used as a hydrogen donor, not only when NrdH was reduced by thioredoxin reductase/NADPH (Fig. 4A) but also, although to a minor extent, when it was reduced by 0.5 mM DTT. On the other hand, there was no stimulatory effect when NrdEF was reduced by Grx1 (either using DTT or GSH/glutathione reductase/NADPH; Fig. 4B) or by DTT alone up to 20 mM. A similar stimulatory effect of NrdI could also be seen for NrdAB in the presence of Trx and thioredoxin reductase/ NADPH, while the effect in the presence of NrdH/thioredoxin reductase/NADPH was almost nonexistent. Even in the presence of NrdI, Trx remained inactive with NrdEF.
Activity of NrdH with Thioredoxin Reductase-Having established that NrdH is a substrate for thioredoxin reductase, we next chose to characterize how NrdH performed as a substrate for this enzyme compared with the other known substrates, Trx and T4-Grx. Using the standard DTNB reduction assay in the presence of NADPH/thioredoxin reductase at pH 8.0, we found that the three enzymes had similar K m values in this assay and that the V max for Trx was only somewhat higher than for NrdH or T4-Grx (Table II).
Since NrdI stimulated both the activity of NrdH plus NrdEF and that of Trx plus NrdAB in the presence of NADPH/thioredoxin reductase, we tested whether NrdI would affect the reduction of NrdH or Trx by thioredoxin reductase in the DTNB assay. No effect by NrdI was seen.
Redox Potential of the NrdH Enzyme-The ability of NrdH to be reduced efficiently by thioredoxin reductase allowed the determination of its redox potential by assessing the equilibrium constant with NADP ϩ /NADPH. Calculated from a standard state redox potential for NADP ϩ /NADPH of Ϫ315 mV, a redox potential of Ϫ248.5 Ϯ 1.5 mV was obtained for NrdH.
General Disulfide Reduction Capacity of NrdH-Reduction of insulin disulfides is a classical assay for thioredoxin (25). As shown in Fig. 5, NrdH was almost as potent a reductant as Trx in this system and was much more potent than Grx1 and T4-Grx. The same relative order of activity among Trx, NrdH, and T4-Grx was also obtained when thioredoxin reductase/ NADPH was used instead of DTT.
In contrast to glutaredoxins, the NrdH protein lacked detectable activity in the GSH-disulfide oxidoreductase assay, where glutaredoxins show a high activity (data not shown).

DISCUSSION
The NrdH protein was originally discovered as the hydrogen donor for L. lactis NrdEF, a class Ib enzyme that is the active ribonucleotide reductase under aerobic conditions in this organism (18). The juxtaposition of nrdH and nrdI upstream of nrdEF in a conserved operon suggests a specific involvement of NrdH and NrdI in the ribonucleotide reduction process. Since the NrdEF enzyme from E. coli and S. typhimurium serves as an excellent model system for class Ib enzymes, we have now extended the characterization of the proteins encoded by this operon to include NrdH and NrdI. We have found NrdH to be an efficient hydrogen donor for ribonucleotide reductase, with higher specificity for the NrdEF enzyme than for the NrdAB enzyme. This reaction was stimulated modestly by the addition of NrdI by an as yet unknown mechanism. Furthermore, we show that NrdH is a good substrate for thioredoxin reductase and that it has no detectable activity with NrdAB or NrdEF in the presence of GSH. A summary of the biochemical properties of NrdH in comparison with other redox active proteins from E. coli and phage T4 is presented in Table III.
Functionally, NrdH thus behaves like the classical thioredoxin (3, 30) of E. coli, sharing some of its biochemical properties including a low redox potential and the ability to reduce insulin disulfides. However, the sequences of the two proteins are not related. Instead, NrdH shows sequence homology with E. coli glutaredoxins (Fig. 1). Glutaredoxins function as dithiol shuttles during ribonucleotide reduction. Their distinction is that they contain a glutathione binding site and that glutathione reduces their active cysteines. With one known exception, they are not reduced by thioredoxin reductase, the exception being the glutaredoxin induced by phage T4, which for this reason originally was classified as a thioredoxin (31). This protein was later renamed T4-Grx (32), since thioredoxin re-ductase can be fully substituted by GSH (33) and also because of the sequence homology with E. coli Grx1, which became apparent upon the determination of the primary structure of the latter (34). The three-dimensional structure of T4-Grx is also more related (10) to the known structure of E. coli Grx1 (9) than to thioredoxin. In addition to NrdH, the known substrates for E. coli thioredoxin reductase are Trx and phage T4-Grx. The sequence homology between these three proteins is, however, restricted (21%) between NrdH and phage T4-Grx and essentially insignificant between NrdH (or T4-Grx) and Trx. This prevents us from making any predictions of the residues that are involved in the interaction with thioredoxin reductase. However, given the predicted high structural similarity (Fig. 1) between in particular E. coli Grx3 and NrdH it might be possible to mutate Grx3 with the aim of making it a substrate for thioredoxin reductase. The identity of the residues to be changed, however,   5. Comparison of E. coli Trx (छ), NrdH (q), Grx1 (ࡗ), and T4-Grx (å) as reductants of insulin disulfides. The assay was performed at pH 6.0 in the presence of 1 mM DTT. Activity was monitored as the increase of absorption at 600 nm due to precipitation of insulin monomers. E, the background of the reaction.
is not obvious and will require careful analysis of three-dimensional structures.
Conceivably, the inability of NrdH to use GSH could be an effect of the redox potential of the protein being too low to allow an efficient reduction by GSH. We have found that E. coli NrdH has a redox potential of Ϫ248 mV at pH 7.0 compared with Ϫ270 mV for thioredoxin (23). Since this number is quite similar to the Ϫ240 mV determined for T4-Grx (31,35), which is relatively efficiently reduced by GSH, the lack of activity of NrdH with GSH can best be explained by the protein lacking the residues needed to interact with GSH. Furthermore, NrdH did not show any activity in the general assay for glutaredoxins using the artificial disulfide ␤-hydroxyethyl disulfide as a substrate, supporting our conclusion that the protein does not catalyze GSH-dependent disulfide reductions.
Where does this leave the NrdH protein? It lacks a glutathione binding site, and its active site cysteine residues are not reduced by glutathione. It was actually discovered in an organism that lacks glutathione. Therefore, it may not functionally be classified as a glutaredoxin, despite the sequence homology ( Fig. 1). In the phylogenetic tree (Fig. 6), it is apparent that NrdH proteins form a separate group, on the same branch as glutaredoxins but definitely separated from thioredoxins. Since the two groups are easily distinguished by the presence of several typical conserved residues (8,10,11), we chose to refrain from classifying NrdH as a thioredoxin despite its thioredoxin-like activity profile. E. coli contains only one known thioredoxin so far, but yeast has two isoforms (36) and plants have many isoforms (37). All of these thioredoxins are at least 105 residues long and do in comparison with glutaredoxins contain one additional ␤-sheet and one ␣-helix preceding the ␤␣␤␣␤␤␣ domain, which thioredoxins share with glutaredoxins. They also contain conserved residues (E. coli Trx numbering) such as Asp-26, Trp-31, and Pro-40, which have no equivalents in the NrdH proteins.
We would like to point out that the two glutaredoxin-like proteins, clustered with the NrdH proteins in the phylogenetic tree (Fig. 6), also might be substrates for thioredoxin reductase. As a speculation, we would like to extend this hypothesis to other glutaredoxin-like proteins present in the genomes of many organisms lacking GSH, including archae (38,39). Glutaredoxins are the simplest members of the thioredoxin superfamily, and the glutaredoxin fold is present in all of the members (here including GSH-peroxidases and glutathione S-transferases (11). A reasonable evolutionary scenario could actually be that NrdH and similar glutaredoxin-like proteins are related to the progenitors of the thioredoxin superfamily, from which the other members evolved by divergent evolution (the glutaredoxins simply by the acquisition of a GSH-binding site).
A genetic study has shown that disruption of the gene for thioredoxin reductase (trxB), but not that of thioredoxin (trxA), is accompanied by increased disulfide bond formation in the cytoplasm of E. coli (40). This finding suggested that there are additional substrates for thioredoxin reductase in E. coli that may channel electrons from this enzyme for disulfide reduction (40,41). An obvious candidate for this role is NrdH. A function of NrdH in general disulfide reduction in the cytoplasm is, of course, dependent on the levels of expression of this protein, which are currently unknown. Since extensive studies of E. coli mutants lacking thioredoxin failed to identify any additional thioredoxin reductase-coupled protein (42), the expression of NrdH is likely to be low, as is the case for NrdEF. The poor solubility of the protein is, however, a factor to consider for the lack of detection of NrdH protein in extracts of E. coli.
This paper demonstrates that the NrdH gene located upstream of the nrdEF genes in E. coli codes for a protein that has a higher specificity as a hydrogen donor for NrdEF than for NrdAB. By analogy with the situation in L. lactis, it would thus seem that most NrdEF enzymes use NrdH as the functional in vivo hydrogen donor. However, the recently completed sequence determination of the M. genitalium genome (28) shows the presence of an NrdEF type of ribonucleotide reductase but no NrdH-like protein. This could provide an example of an NrdEF enzyme that interacts with, for example, thioredoxin for which M. genitalium contains a coding sequence. Thus, care should be taken not always to associate NrdEF with NrdH, as the situation in L. lactis, E. coli, and S. typhimurium would suggest.
The mechanism of the stimulatory effect of NrdI on the activity of NrdH with NrdEF and on the activity of Trx with NrdAB remains elusive. Since the presence of nrdI in the known nrdEF loci is more conserved than that of nrdH, we believe that the NrdI protein has an important in vivo function for the activity of NrdEF. It is surprising that all of the gene products encoded by the nrdEF operon are fully functional proteins also in E. coli and S. typhimurium, since this operon seems to be poorly transcribed and knock-out mutants of the nrdEF genes have no phenotype (15). The conservation of the operon does, however, suggest an important in vivo function that at present is not understood.