Dimer-Oligomer Interconversion of Wild-type and Mutant Rat 2-Cys Peroxiredoxin

Rat heme-binding protein 23 (HBP23)/peroxiredoxin (Prx I) belongs to the 2-Cys peroxiredoxin type I family and exhibits peroxidase activity coupled with reduced thioredoxin (Trx) as an electron donor. We analyzed the dimer-oligomer interconversion of wild-type and mutant HBP23/Prx I by gel filtration and found that the C52S and C173S mutants existed mostly as decamers, whereas the wild type was a mixture of various forms, favoring the decamer at higher protein concentration and lower ionic salt concentration and in the presence of dithiothreitol. The C83S mutant was predominantly dimeric, in agreement with a previous crystallographic analysis (Hirotsu, S., Abe, Y., Okada, K., Nagahara, N., Hori, H., Nishino, T., and Hakoshima, T. (1999) Proc. Natl. Acad. Sci. U. S. A. 96, 12333–12338). X-ray diffraction analysis of the decameric C52S mutant revealed a toroidal structure (diameter, ∼130Å; inside diameter, ∼55Å; thickness, ∼45Å). In contrast to human Prx I, which was recently reported to exist predominantly as the decamer with Cys83-Cys83 disulfide bonds at all dimer-dimer interfaces, rat HBP23/Prx I has a Cys83-Cys83 disulfide bond at only one dimer-dimer interface (S–S separation of ∼2.1Å), whereas the interactions at the other interfaces (mean S–S separation of 3.6Å) appear to involve hydrophobic and van der Waals forces. This finding is consistent with gel filtration analyses showing that the protein readily interconverts between dimer and oligomeric forms. The C83S mutant exhibited similar peroxidase activity to the wild type, which is exclusively dimeric, in the Trx/Trx reductase system. At higher concentrations, where the protein was mostly decameric, less efficient attack of reduced Trx was observed in a [14C]iodoacetamide incorporation experiment. We suggest that the dimerdecamer interconversion may have a regulatory role.

distance from Cys 52 to Cys 173 is ϳ13 Å. In the oxidized dimeric structure, however, Cys 52 is separated from Arg 128 by about 7.4 Å (24). On the basis of biological and structural findings with Trx, it has been suggested that a mixed disulfide bond between Trx Cys 33 and Prx Cys 173 , formed during the catalysis, is attacked by Trx Cys 36 , generating reduced Prx and oxidized Trx (numbering in Escherichia coli, Refs. 29 -33). Trx, which is a small protein (molecular mass ϳ12 kDa), plays an important role in cellular processes via redox regulation by undergoing reversible oxidation/reduction reaction of two cysteine residues, Cys 33 and Cys 36 . The 2-Cys Prxs are abundantly expressed in mammalian cells, and their content is estimated to be Ͼ20-fold higher than that of Trx (1,23,34). These findings suggest that there is some modulating system for the interaction between the two proteins.
It seems reasonable to suppose that the interaction between Prx and Trx is in part regulated through their quaternary structures. Oligomers larger than the dimer have been observed in solution for 2-Cys Prxs from bacteria, plants, and humans (27,28,(35)(36)(37)(38)(39). Redox state, protein concentration, pH, and ionic strength are reported to influence the oligomeric state but differ in their effects upon different 2-Cys Prxs. Furthermore, the state of amino acid residues at the dimer-dimer interface also influences oligomerization and peroxidase activity (20,28,39). Further, human Prx I and Prx II have distinct quaternary structures; Prx I was recently reported to exist predominantly as the decamer with a Cys 83 -Cys 83 disulfide bridge at the dimer-dimer interface, whereas Prx II exists as a dimer, because the corresponding Cys residue is replaced with Thr in Prx II (40). Based on this feature, it was proposed that the function of Prx I is distinct from that of Prx II.
Here, to address the relationship between the oligomeric properties and peroxidase activity of mammalian 2-Cys Prx, we prepared HBP23/Prx I variants by using site-directed mutagenesis, and we also performed an x-ray structure analysis of decameric Prx I. In contrast to the recently reported result for human Prx I (40), our crystal structure of the decameric rat C52S mutant (Protein Data Bank (PDB) entry 2Z9S) showed that only one out of the five dimer-dimer interfaces involved a disulfide bridge from Cys 83 to Cys 83 of the adjacent dimer, indicating that stabilization of the decamer structure is not due to Cys 83 -Cys 83 disulfide bridges at the dimer-dimer interfaces, but rather, is predominantly due to hydrophobic interaction and structural fitting, as is seen in the decameric structures of bacterial (26 -28) and human (25) Prx II.

EXPERIMENTAL PROCEDURES
Preparation of Recombinant Proteins-The coding region of HBP23/Prx I was amplified by PCR using the forward primer, 5Ј-AGC CAT ATG TCT TCA GGA AAT GCA A-3Ј, and reverse primer, 5Ј-TCG GAT CCT CAC TTC TGC TTA GAG AAA TAC TC-3Ј. The PCR product was digested with NdeI/ BamHI and cloned into the corresponding site of pET3c vector. This expression vector was named pET3c-HBP23. Mutants of HBP23/Prx I were prepared at Cys 52 , Cys 83 , Cys 173 , Arg 128 , and Arg 151 , which were replaced with Ser, Lys, Ala, Glu, and Asp, respectively, using a QuikChange TM site-directed mutagenesis kit (Stratagene). All of the site-directed mutations were con-firmed by DNA sequencing. The E. coli strain BL21/DE3 was transformed with pET3c-HBP23 for overexpression of HBP23/ Prx I and cultured as described previously (41). Each HBP23/ Prx I recombinant protein was purified by the methods described previously (41).
The Trx gene of E. coli (E-TrxA) was prepared from pTrx plasmid (Invitrogen). The NedI/SalI DNA fragment from pTrx was cloned into pET30a or pET28a and designated pET30-ETrxA or pET28-HisETrxA, respectively. pET30-ETrxA plasmid is the expression vector for the wild-type Trx, and pET28-HisETrxA is the expression vector for modified Trx containing a His tag sequence (six histidine residues) at the N-terminal region. Site-directed mutation of E-TrxA, changing Cys 33 or Cys 36 to Ser, was carried out using the QuikChange TM sitedirected mutagenesis kit (Stratagene). All of the variant genes for Trx and His-tagged Trx were confirmed by DNA sequencing. The E. coli strain BL21/DE3 was transformed with pET30-ETrxA or pET28-HisETrxA, cultured in LB medium containing 50 g/ml kanamycin at 37°C to an absorbance of 0.7 at 600 nm, and induced with 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside for 4 h before being harvested by centrifugation. The cell pellets were suspended in 50 mM Tris-HCl, pH 8.0, containing 0.1 M NaCl and 1 mM EDTA (buffer A) and disrupted by sonication. The soluble fraction recovered by centrifugation was treated at 65°C for 5 min. After centrifugation, the supernatant was dialyzed against 10 mM Tris-HCl, pH 8.0, and applied to a DE-52 ion-exchange column (2.5 ϫ 5.0 cm). Proteins were eluted with a gradient of 0 -0.3 M NaCl. His-tagged Trx was purified on a Ni 2ϩ -nitrilotriacetate column according to the procedures recommended by the manufacturer.
The Trx reductase (TrxR) gene was prepared from genomic DNA of E. coli strain JM109 by PCR (DNA Data Bank in Japan/ European Molecular Biology Laboratory (DDBJ/EMBL) accession number J03762). The primers using for PCR were as follows: forward primer, 5Ј-GGG ATC CCA TGG CAG CAC CAA ACA C-3Ј, and reverse primer, 5Ј-ACC CAT AGT CGC ATG GTG TC-3Ј. The PCR fragment was cloned into the SmaI site of pUC119 vector, and the cloned gene was sequenced. The direction of the TrxR gene was the same as that of the lacZ gene of pUC119, and the plasmid was designated pUC-TrxR. The E. coli strain JM109 was transformed with pUC-TrxR, and the transformant constitutively overproduced the Trx reductase protein. To prepare Trx reductase, the transformant was cultured in LB medium containing 50 g/ml ampicillin at 37°C for 12-18 h. The cell pellets were suspended in buffer A and disrupted by sonication. After centrifugation, the supernatant was fractionated with ammonium sulfate (40 -80% saturation). After dialysis against 20 mM Tris-HCl, pH 8.0, the fraction was applied to a DE-52 ion-exchange column (2.5 ϫ 5.0 cm), and proteins were eluted with a gradient of 0 -0.4 M NaCl. Fractions containing the protein were dialyzed against 20 mM Tris-HCl, pH 8.0, and applied to a 2Ј,5Ј-ADP-Sepharose affinity column (0.6 ϫ 7.5 cm). Trx reductase was eluted with 20 mM Tris-HCl, pH, 8.0, containing 0.1 M NaCl and 10 mM NADP. Examination of Coomassie Blue-stained bands after gel electrophoresis indicated that each isolated recombinant protein was Ͼ90% pure.
The concentration of purified proteins, HBP23/PrxI, Trx, and Trx reductase, was determined with a Coomassie Blue protein assay reagent kit (Pierce). Bovine serum albumin was used as a standard.
Determination of Peroxidase Activity-The peroxidase activity was measured with two systems, the Trx-and DTT-based systems (42,43). The peroxidase activity in the presence of Trx was assayed in 0.7 ml of reaction mixture containing 50 mM HEPES-NaOH (pH 7.0), 0.5 mM EDTA, 60 g of Trx, 2 g of Trx reductase, 0.1 mM H 2 O 2 , 150 M NADPH, and the appropriate amount of HBP23/PrxI at 25°C. Reaction was initiated by the addition of H 2 O 2 , and consumption of NADPH was monitored spectrophotometrically at 340 nm. The peroxidase activity in the presence of DTT was assayed as described previously, with a slight modification. The reaction mixture (0.7 ml) contained 150 mM potassium phosphate buffer, pH 7.4, 50 mM DTT, 30 mM t-butyl hydroperoxide and the appropriate amount of HBP23/PrxI at 25°C. The activity was measured spectrophotometrically at 310 nm.
Size-exclusion Chromatography-Size-exclusion chromatography was performed at room temperature using the fast protein liquid chromatography system (Amersham Biosciences) with a Sephacryl S-200 column. Ferritin (440 kDa), aldolase (158 kDa), ovalbumin (43 kDa), and myoglobin (17 kDa) were used as molecular mass standard markers for calibration of the column. Elution of liver extract or molecular mass standards was monitored at 280 nm. Absorbance at 280 nm was measured with a Hitachi U-3200 spectrophotometer to monitor elution of protein from the column. Protein-containing fractions of liver extracts were subjected to SDS-PAGE (15% gel) according to the method of Laemmli (44). Antibody against HBP23/Prx I was prepared as described previously (45). X-ray Crystallography of HBP23/Prx I-The C52S mutant was crystallized by the hanging drop method at 20°C by mixing 2 l of the protein solution (5 mg/ml in 5 mM sodium acetate, pH 5.0, 2 mM DTT, 1 mM CHAPS) with 2 l of reservoir solution (0.17 M ammonium acetate, 20% glycerol, 25% polyethylene glycol 4000, 84 mM sodium acetate, pH 5.0). Plate crystals appeared within 2 or 3 days and grew to their full size in 2 weeks. The crystals were soaked with a precipitant solution containing 30% glycerol and then plunged into liquid nitrogen. The data set was collected at the BL-5 beamline of the Photon Factory (Tsukuba, Japan) using a Quantum 315 area detector. The data set was obtained at 2.9 Å resolution. The data were processed and scaled with HKL2000. Molecular replacement was performed with EPMR using the coordinates of the crystal structure of human decameric 2-Cys Prx II purified from erythrocytes (Protein Data Bank entry 1QMV; Ref. 22 and see also Ref. 46). The model was refined with CNS (47). The final R-factor and free R-factor were 20.5 and 28.0%, respectively.

RESULTS AND DISCUSSION
Molecular Mass of HBP23/PrxI-Although we previously determined the crystal structure of the C83S mutant of rat liver HBP23/Prx I as a dimer (PDB entry 1QQ2), oligomerized forms other than the dimer are observed in solution for many 2-Cys Prxs from bacteria, plants, and humans (27,28,(35)(36)(37)(38)(39) (supplemental Fig. 1). To examine the molecular species of HBP23/Prx I in rat liver, the homogenate was subjected to size-exclusion chromatography using a Sephacryl S-200 column pre-equilibrated with 50 mM phosphate buffer, pH 7.4, containing 0.1 M NaCl ( Fig. 1). SDS-PAGE and Western blotting showed that protein was eluted in at least two different positions. Comparison with molecular mass standards showed that HBP23/Prx I in the 20% homogenate was predominantly eluted at the position of molecular mass 44 kDa, but HBP23/Prx I in the 33% homogenate was also eluted at the position of molecular mass 230 kDa, thus being present as both dimer and decamer. HBP23/Prx I seems to form dimeric and decameric structures depending on the protein concentration or ionic strength, similarly to other 2-Cys Prxs. Therefore, we further studied the quaternary structure of the recombinant HBP23/ Prx I by means of chromatographic analysis under various conditions of protein concentration/ionic strength, as well as in the absence or presence of DTT.
When 0.5 mg of freshly purified wild-type protein (1 ml) was applied to a Sephacryl S-200 column at 0.2 M NaCl, the elution profile was more complex in the absence of DTT than in its presence (Fig. 2). In the absence of DTT, the wild-type protein tended to form the decamer at high protein concentration, but the decamer tended to dissociate into tetramer or dimer as the solution was diluted (Fig. 2, A and C) or when the ionic strength was increased (Fig. 2D), whereas in the presence of DTT, the decamer was the only form under all the conditions studied (Fig. 2B). These results suggested that Cys 83 may not be involved in decamer formation through disulfide bond formation, although it is located at the dimer-dimer interface in the decameric structure (see below). The mutants C52S and C173S, which cannot form an intermolecular disulfide bridge in the vicinity of the active site under oxidative conditions, were eluted only as decamers under all conditions studied (Fig. 2, A and B), which is consistent with the observation that reduced, DTT-treated decameric HBP23/Prx I is rather stable. The mutant C83S was eluted at the position corresponding to the dimer in the absence of DTT. Upon the addition of DTT, however, the dimer reverted to the monomer rather than forming the decamer, suggesting that the replacement of the sulfur atom with oxygen influences the dimer-dimer interaction. The double mutants, C52S/C83S and C83S/C173S, were eluted at positions corresponding to both dimer and monomer. However, in both cases, the monomer content was increased by the addition of DTT, probably because disulfide bridge formation between Cys 173 and Cys 173 or between Cys 52 and Cys 52 became less favorable (Fig. 2, A  and B).
Crystal Structure of Decameric HBP23/Prx I-To elucidate in detail the structural features of HBP23/ Prx I, the C52S mutant was successfully crystallized, and its structure was determined (PDB entry 2Z9S) (supplemental table). The final model was compared with that of the oxidized dimeric C83S mutant. The diffraction data were collected to 2.9 Å resolution. The crystal belongs to the P2 1 space group with one molecule in the asymmetric unit. The structure was determined by molecular replacement techniques and subjected to molecular dynamics refinement. The model has good overall stereochemistry, with 82.9% of all residues in the most favorable region of the Ramachandran plot. Residues in disallowed regions are defined in 2F o Ϫ F c electron density maps.
The C52S mutant was crystallized as a toroidal decamer (Fig. 3) very similar to the reported structures of other Prx species (25)(26)(27)(28)36). The toroid has a diameter of ϳ130 Å, an inside diameter of ϳ55 Å, and a thickness of ϳ45 Å. A decamer is composed of five dimers, forming an (␣ 2 )5 complex. The chain trace of each monomer is quite clear and continuous from amino acid residues 3-198. We observed clear densities of the C-terminal loops (residues 176 -198), whereas this was not the case in the C83S mutant dimer, probably because of dislocation (24). The extra C-terminal fold of the C52S mutant contains a loop region (residue 176 -187) and a short helix (residue 188 -197).
Only weak interactions appear to exist between dimers. The dimerdimer interfaces involve mainly van der Waals contacts of hydrophobic residues, including Thr 48 , Phe 81 , Phe 82 , Ala 86 , and Trp 87 . All 10 Cys 83 residues are located at the interfaces. Among them, eight are in pairs without any electron density between SG atoms (Cys 83 of chains B and F, D and G, E and J, and H and I). The distances between the SG atoms of these four cysteine pairs are obtained as 3.4 -3.6 Å, which are too far for a disulfide bond but Elution of protein was monitored at 280 nm, and protein-containing fractions were subjected to SDS-PAGE (15% gel). The separated protein was visualized with Coomassie Blue (CBB) (A) and transferred by electroblotting to a nitrocellulose membrane (PerkinElmer Life Sciences). The blot was blocked with 1% skim milk buffer at room temperature for 60 min and incubated overnight with antibody against HBP23/Prx I. Binding to HBP23/Prx I was visualized with an Immun-Blot assay kit (Bio-Rad) (B). Calibration of the column was performed using a standard as described under "Experimental Procedures" (C). consistent with van der Waals contacts between the sulfur atoms. However, one Cys 83 pair (Cys 83 of chains A and C) has clear electron density between the SG atoms, with a separation of ϳ2.1 Å, which is consistent with a disulfide bond (Fig. 4). This disulfide bond is buried in the protein and is not exposed to the solvent, suggesting that it may not easily dissociate even in the presence of DTT. The structural changes accompanying the disulfide formation appear to be restricted. In the chain A and C pair, C␣ of Cys 83 and Ala 84 move ϳ0.7 Å toward their counterparts of the monomer pair, but this structural change does not extend to the rest of the protein. The root mean square deviations of the traces of chains A and B are 0.37 Å, whereas the root mean square deviation of chains A and C is 0.41 Å. Fig. 5A compares the dimer structures of the C52S and C83S mutants. The core structures are very similar to each other, but some differences can be seen. The most significant difference between the two mutants is the position of the two cysteines (or corresponding Ser 52 residue) in the active site. In the C83S mutant, Cys 52 and Cys 173 form a disulfide bond that is exposed to the solvent (24). On the other hand, Ser 52 is positioned in the N terminus of helix ␣2 and is located in a hydrophobic pocket in the C52S mutant. The distance between the ␣-carbons of Ser 52 and Cys 173 is ϳ13 Å, suggesting that this structure is in a catalytically inactive form since the two cysteine residues would be too far apart to form a disulfide bond efficiently in the catalytic cycle. This structural change in the active site extends to the dimer-dimer interface region. Formation of the disulfide bridge between Cys 52 and Cys 173 causes partial unfolding of the ␣2 helix, and Thr 54 moves ϳ7 Å. This change causes positional changes of residues Trp 87 and Phe 48 , which are located at the dimer-dimer interface (Fig. 5B). The new locations of these residues may result in a steric clash with Ala 86 from the adjacent monomer.
Influence of Oligomerization on the Peroxidase Activity-The molecular mechanism of the reduction of target proteins by reduced Trx is well understood in the E. coli system (29 -32). E. coli Trx Cys 33 forms a mixed disulfide intermediate with the target protein during the reduction process, and then attack of Cys 36 results in release from the reduced protein. In the case of rat HBP23/Prx I, it was suggested that Cys 173 is also a target sulfur for reduced Trx since Cys 173 is located close to the surface of the molecule in the crystal structure of the oxidized dimer (PDB code 1QQ2). To confirm this, the interaction between HBP23/Prx I and Trx was examined by Ni 2ϩ column chromatography with a His-tagged E. coli Trx variant as the bait. Mutant HBP23/Prx I was applied to a Ni 2ϩ column with the mutant His-tagged Trx C36S, and the fractions eluted with DTT were subjected to complex formation analysis and then resolved by SDS-PAGE (Fig. 6). The wild type and the HBP23/ Prx I mutants C83S and C52S/C83S each formed a stable complex with the Trx mutant C36S. However, the HBP23/Prx I mutant C83S/C173S was not trapped on the column (Fig. 6, panel  c). It should be noted that none of the HBP23/Prx I mutants studied was trapped by a Ni 2ϩ column with the His-tagged Trx mutant C33S. In parallel, it was found that none of the HBP23/Prx I mutants studied showed peroxidase activity in the reaction system containing the Trx mutant C33S, instead of the wild-type Trx (not shown). These results clearly demonstrate that HBP23/Prx I Cys 173 is indeed the target sulfur of reduced Trx. The reduced Cys 52 residue is likely attacked by hydrogen peroxide, and then a water molecule is released by heterolytic cleavage. Finally, another water molecule is released by attack of the sulfur base of Cys 173 to form Cys 52 -Cys 173 disulfide-bridge.
The finding in this study that the distance between Ser 52 and Cys 173 in the crystal structure of the C52S mutant (PDB code 2Z9S) is as large as 13 Å suggests that the structure of mutant C52S is catalytically inactive. However, the results of the gel filtration study indicate that even wild-type oxidized enzyme can form the decamer under certain conditions, such as high protein concentration. Thus, as proposed previously (24 -28), the decameric structure of the oxidized wild-type enzyme may have a different conformation from that determined in this study. As it was of interest to examine the relationship between decamerization of the wild-type enzyme and peroxidase activity, we compared the reductive half-reaction by reduced Trx at high concentrations of wild-type enzyme (mainly in decameric form) and the C83S mutant (exclusively in dimeric form). It should be noted, however, that the determination of activity using the standard Trx reductase/Trx system is not practically easy since large amounts of Trx reductase/Trx are necessary to ensure that the rate-limiting step of the overall reaction is the Prx peroxidase reaction under conditions of a high Prx concentration where the main species of Prx is the decamer. It is likely that the Trx-dependent peroxidase activity of HBP23/Prx I is influenced by Trx binding efficiency. It is noteworthy that the C83S mutant showed similar activity to that of the wild type in the Trx system under diluted conditions (Table 1). To estimate the influence of the quaternary structure of HBP23/ Prx I, the apparent reduction rate was estimated by mixing the wildtype HBP/Prx I with a higher concentration of reduced Trx under ice cooling followed by incorporation of [ 14 C]IAA, as described under "Experimental Procedures." After mixing a high concentration of HBP/Prx I with reduced Trx, significant incorporation of [ 14 C]IAA was observed, as shown in Fig. 7, whereas only a small amount of [ 14 C]IAA was incorporated into HBP/Prx I alone; ϳ7% of free cysteine residues of HBP23/Prx I was modified by [ 14 C]IAA after incubation for 1 h on ice (data not shown). These results indicated that HBP23/Prx I is oxidized even in decameric form, and the active site disulfide was reduced by reduced Trx followed by incorporation of [ 14 C]IAA. The efficiency of time-dependent incorporation of [ 14 C]IAA into the mutant C83S, however, was 1.4-fold higher than that into the wild type during 8 -24 min (13 nmol of Trx and 11 nmol of HBP23/Prx I), and then the incorporation reached a plateau level because of consumption of reduced Trx because of simultaneous competitive incorporation of [ 14 C]IAA into reduced Trx. The incorporation of radioactivity into Trx was confirmed to be 1.3 times greater with the wild type than with mutant C83S after separation of Trx by means of SDS-PAGE. A similar value, 1.4-fold, was obtained when 35 nmol of Trx and 26 nmol of HBP23/Prx I were used (data not shown). These results suggest that decamerization of HBP23/ PrxI decreases the binding efficiency of Trx.
Influence of the Mutation on Peroxidase Activity-To clarify the role of amino acid residues in the active site of HBP23/ Prx I, the activity of variants was examined in two ways, i.e. toward H 2 O 2 in the presence of the Trx system, NADPH, Trx, and Trx reductase and toward t-butyl hydroperoxide in the presence of DTT. The former was determined by measuring the consumption of NADPH in terms of decrease of absorption at 340 nm, and the latter was determined by estimating the amount of oxidized DTT, as described under "Experimental Procedures." For HBP23/Prx I, the systemcomposed E. coli Trx and Trx reductase were found to be effective (supplemental Fig. 2), so the following analysis was done with E. coli Trx and Trx reductase. The activity of all mutants determined in the Trx system was not significantly different from that in the DTT system except in the case of mutation at position 173 (Table 1). As expected, the C52S mutant exhibited no activity in either of the assay systems. The C173S mutant was fully active using DTT as an electron transfer partner, although it was inactive in the Trx system. This is consistent with the above conclusion that Cys 173 plays an essential role in the interaction with Trx as the immediate electron donor. It should be noted that C83S exhibited activity similar to that of the wild type in the Trx system, and it is present exclusively as the dimer, as determined with size-exclusion chromatography. In addition, the double mutant C83S/C173S exhibited the highest DTT-dependent activity among the variants studied, like the plant enzyme (39), in accordance with the idea that DTT can act in place of Cys 173 and that Cys 52 is involved in normal peroxidase activity. The substitution of the arginine residues at positions 128 and 151 dramatically affected the activity. When either of these was replaced with a lysine, alanine, or glutamate residue in the C83S mutant, the activity decreased to 2-7% of that of the wild type in both systems. Thus, Arg 128 appears to be important for the reactivity of the active site Cys 52 , as discussed previously (24 -28). Substitution to lysine at these positions did not restore the activity, probably due to its lower strength as a base. Substitution to alanine or glutamine at these residues does not allow hydrogen bond formation with Cys 52 .

CONCLUSIONS
The present study of rat HBP23/Prx I demonstrates that the redox state of the active site cysteine residues and small conformational changes of the amino acid residues located at the dimer-dimer interface have an important influence on the quaternary structure of HBP23/Prx I. Although the Cys 83 residue, which is replaced with other residues in other Prx species, is located at the dimer-dimer interface, only one out of five cysteine pairs in the decamer was found to be disulfide-bonded on the basis of x-ray diffraction analysis. This result was also confirmed by other diffraction data sets and is in contrast to a recent report on human Prx I, which has a very similar amino acid sequence to rat HBP23/Prx I; both rat and human Prx I contain conserved Cys 83 residues. It was reported that human Prx I is mainly in a decamer form having disulfide bridges at the dimer-dimer interfaces, and it was therefore suggested that the main function of Prx I might be to work as a chaperone, rather than as a peroxidase (40). Our present crystallographic analysis of the rat Prx I decamer indicated that only one dimer-dimer interface involves a disulfide bond, and the others involve weak interactions, such as hydrophobic interaction and van der Waals contacts (including the sulfur atom of Cys 83 ). This is consistent with the results of gel filtration experiments showing that the decameric form of the wild-type enzyme dissociates into dimer and probably tetramer forms under conditions of low protein concentration or high ionic strength solvent. The addition of DTT to the wild-type enzyme favored the decameric form rather than dimeric form, possibly due to reduction of the disulfide bond in the active site. The fact that only one of the cysteine pairs was disulfide-bonded suggests that the decamer may be assembled from one disulfide-bonded tetramer and three dimers, or alternatively, one disulfide bond may be formed during preparation. Although the fact that the mutant C83S cannot form the decamer might support the former possibility, it is more likely that replacement of the sulfur atom by the smaller oxygen atom favors dissociation to the dimer under circumstances where the dimer-dimer interfaces involve only weak interactions. Other types of peroxiredoxin, such as AhpC (28) and Prx II (25), exist in decamer form even when the residue corresponding to Cys 83 is replaced by threonine. Thus, the cysteine residue at the dimer-dimer interface is not essential for decamer formation. The effect of assembly on the peroxidase activity and the binding efficiency with the direct electron donor protein seems to vary from species to species in the Prx family (20,28,35,38). The dimeric form of Salmonella typhimurium 2-Cys Prx exhibited lower binding efficiency with AhpF, a direct electron donor, and lower peroxidase activity than the decameric form (28). However, the binding efficiency of reduced Trx with dimeric HBP23/Prx I was rather higher than that of the decamer, possibly because of a higher collision probability. It was also reported that the high molecular weight 2-Cys Prxs of human or yeast formed by phosphorylation show lower peroxidase activity (20). Thus, it is possible that oligomer formation is one of the regulatory systems for the functions of 2-Cys Prx. Various other modulation systems for the peroxidase activity of 2-Cys Prx have been reported, including C-terminal truncation, phosphorylation, peroxidation, and sulfiredoxin (19 -23). The regulation mechanism based on protein-protein interaction may modulate not only the functions of 2-Cys Prx but also the function of Trx.
The interaction between HBP23/Prx I and Trx appears to be regulated by redox state and by the quaternary structure of HBP23/Prx I, as described above. Trx, which is functionally involved in a variety of cellular processes as a potent endogenous reducing reagent, is estimated to be present at a concen-

TABLE 1 Specific activities of HBP23/PrxI and its variants
The proteins were assayed for specific activity as described under "Experimental Procedures." Data are the mean of at least three independent measurements. ND, not detected. tration of 0.1-1 M, which is much lower than that of 2-Cys Prx (1,23,34). The 2-Cys Prxs are generally abundant proteins in cells (7,23,39), e.g. HBP23/Prx I was estimated to be present at a concentration of about 22 M in rat liver soluble protein (1). Human 2-Cys Prx I Cys 173 forms a disulfide bond with a cysteine residue of cytokine MIF, inhibiting the MIF D-dopachrome tautomerase and Prx peroxidase activities (16). Furthermore, although 2-Cys Prx I is located mainly in the cytoplasm, it has also been found in the nucleus (4 -6, 15, 17). Binding of the nuclear proteins c-Abl and c-Myc to 2-Cys Prx I was reported to modulate their functions in the cell cycle (15,17,18). For translocation into the nuclear compartment, 2-Cys Prx would have to form a complex with a nuclear protein because it lacks a nuclear localization signal sequence. Thus, the dimeric form of Prx I seems to be functionally very important, as illustrated in Fig. 8. Since the interconversion between the oxidized dimeric and decameric forms is influenced by environmental factors, such as redox state or protein concentration, this may be one of the regulatory mechanisms for the interactions of this protein with various other proteins and for translocation into the nucleus.