In vivo characterization of a thioredoxin h target protein defines a new peroxiredoxin family.

Disruption of the two thioredoxin genes in yeast dramatically affects cell viability and growth. Expression of Arabidopsis thioredoxin AtTRX3 in the Saccharomyces thioredoxin Delta strain EMY63 restores a wild-type cell cycle, the ability to grow on methionine sulfoxide, and H2O2 tolerance. In order to isolate thioredoxin targets related to these phenotypes, we prepared a C35S (Escherichia coli numbering) thioredoxin mutant to stabilize the intermediate disulfide bridged complex and we added a polyhistidine N-terminal extension in order to purify the complex rapidly. Expression of this mutant thioredoxin in the wild-type yeast induces a reduced tolerance to H2O2, but only limited change in the cell cycle and no change in methionine sulfoxide utilization. Expression in the Delta thioredoxin strain EMY63 allowed us to isolate a complex of the thioredoxin with YLR109, an abundant yeast protein related to PMP20, a peroxisomal protein of Candida. No function has so far been attributed to this protein or to the other numerous homologues described in plants, animals, fungi, and prokaryotes. On the basis of the complementation and of low similarity with peroxiredoxins, we produced YLR109 and one of its Arabidopsis homologues in E. coli to test their peroxiredoxins activity. We demonstrate that both recombinant proteins present a thioredoxin-dependent peroxidase activity in vitro. The possible functions of this new peroxiredoxin family are discussed.

Thioredoxins are small disulfide-containing redox proteins (Ϸ13 kDa) that have been isolated from almost all organisms (reviewed in Refs. 1 and 2). Three types of function have regularly been proposed. In the first type, they act as structural components required for the activity and synthesis of some components of T7 DNA polymerase or for phage assembly. These reactions are mostly redox independent, while the two other types are directly redox dependent. In the second type, they are intermediate energy donors to some enzymes like ribonucleotide reductase, PAPS 1 reductase, methionine sulfox-ide reductase, and hydrogen peroxide reductase which accept energy through a proton transfer on cysteines. In the third type, they regulate the function of enzymes or transcription factors by modifying their disulfide bridged conformation. Examples of redox regulated enzymes are most Calvin cycle enzymes, CF1 and malate dehydrogenase of plant chloroplasts (3). Redox regulated transcription factors have been described in mammals, for example, NF-B and AP1 (4). In most studies, the suggestion that a thioredoxin is the cellular reductant of a particular protein is sustained almost exclusively by in vitro experiments. Nevertheless, other cellular reducers like glutaredoxins, NADPH reductases, or even glutathione are also able to reduce disulfide bridges. The situation is even more complex in plants which present multiple thioredoxin genes and their products. For example, the Arabidopsis thaliana genome encodes at least five cytosolic thioredoxins h (5), and seven chloroplastic thioredoxins (2). In addition, other proteins with thioredoxin domains have been described in plants (6 -10).
Mutants are useful tools for the characterization of gene function. Budding yeast presents two thioredoxin genes. While the inactivation of each of the genes does not significantly alter yeast growth, the disruption of both genes profoundly affects cell viability. This mutant strain (EMY63) is unable to use sulfate as sole a sulfur source and grows very poorly on methionine sulfoxide. Rapid growth is obtained with methionine but the cell cycle is profoundly modified, with a longer S phase and a shorter G 1 phase. Moreover, this mutant yeast shows an increased sensitivity to hydrogen peroxide (11,12). We have previously shown that the five A. thaliana thioredoxins h (AtTRX1 to AtTRX5) confer a normal cell cycle and the ability to grow on methionine sulfoxide as unique sulfur source when expressed in the yeast mutant. AtTRX3 confers H 2 O 2 tolerance but cannot restore sulfur assimilation, while AtTRX2 restores sulfur assimilation, but is unable to confer H 2 O 2 tolerance (13). These data clearly indicate that yeast thioredoxins interact with multiple targets while each A. thaliana thioredoxin h interacts only with some of them.
One way to characterize the exact function of the unique cytosolic Trx in mammals or of each Trx in plants would be the isolation of in vivo complexes between one thioredoxin and its target(s) protein(s). This approach has been unsuccessful so far, probably because the complexes have a very short half-life. At the present time, only two articles, one on human, the second on plants report indirect evidence for such complexes using the two-hybrid system (14,15). In this paper, we have developed a new in vivo approach in order to isolate biochemically complexes between thioredoxins and their cellular targets. We have used the recent knowledge on the reaction mechanism of thioredoxins to stabilize the complexes.
In vitro studies have shown that Trx reduce protein disulfide bridges through a two-step reaction involving the two cysteine residues of the conserved redox active site WCxPC (x ϭ G or P). In the first step, the more N-terminal cysteine of the Trx (equivalent to the Cys 32 of Escherichia coli TrxA) attacks the target protein disulfide bridge, reducing one cysteine of the S-S bridge and establishing a disulfide bridge with the second cysteine of the target, forming a mixed intermediate between the Trx and the target protein. The second step involves an intramolecular attack by the second cysteine of the Trx (equivalent to the Cys 35 for the E. coli thioredoxin) on the mixed disulfide intermediate, releasing the reduced target protein and the oxidized Trx (16,17). The intermediate disulfide bridged complex formed in vitro is stable if the second cysteine of the Trx is replaced by a structural analog of cysteine, like a serine or an alanine (18,17).
In our report, we show that the ectopic expression of the C35S mutated AtTRX3 (amino acid numbering according to the E. coli thioredoxin) in Saccharomyces cerevisiae induces H 2 O 2 hypersensitivity, thus partially mimicking the phenotype of the yeast ⌬ Trx mutant. This suggests that the dominant negative mutant protein undergoes a stable interaction with the target responsible for H 2 O 2 tolerance. We have purified this complex and shown that it results from the interaction of the mutated Trx with the product of ORF YLR109, an abundant protein to which no function has so far been attributed. We show that the E. coli recombinant YLR109 product presents all the characteristics of the peroxiredoxin family, and that its peroxidase activity is dependent of Trx activity. Moreover, we describe well conserved sequences similar to YLR109 for almost all organisms, including mammals, plants, and bacteria, suggesting an identical peroxiredoxin activity for these proteins.
General Methods-Manipulation of DNA, E. coli, and S. cerevisiae were performed using standard methods (20,21). S. cerevisiae cells were transformed after lithium chloride treatment (22). DNA synthesis by the polymerase chain reaction (PCR) was performed with the Pfu DNA Polymerase (Stratagene). DNA was purified with Qiaquick Gel extraction kit (Qiagen). All restriction enzymes were provided by Roche Molecular Biochemicals and T4 DNA ligase was provided by Promega. All constructs were checked by sequencing. Dideoxynucleotide sequencing was performed with the ABI Prism Big Dye Terminator Cycle Reaction Kit (Applied Biosystems), in our laboratory on a ABI 377A sequencer.
Mutagenesis of Trx Active Site-AtTRX3 and YTRX1 ORFs were amplified by PCR using specific oligonucleotides designed to introduce a NdeI site upstream of the start codon and a BamHI site downstream from the termination codon. Site-directed mutagenesis of the thioredoxins active site were obtained by PCR-mediated overlap extension using a pair of complementary mutated oligonucleotides (23). Single amino acid mutants of Trx were created by PCR using the following nucleotide pairs: for C35S AtTRX3, oligonucleotide AtTrx3CSdir 3Ј (5Ј-GCAACAT-GGTGCCCACCTTCACG-3Ј) and oligonucleotide AtTrx3CSdir5Ј (5Ј-AAACGTGAAGGTGGGCACCATG-3Ј); for C35S YTRX1, oligonucleotide YTrx1CSdir5Ј (5Ј-GCAATCATTTTACTTGGACCGCACC-3Ј) and oligonucleotide YTrx1CSdir3Ј (5Ј-GGTGCGGTCCAAGTAAAATGAT-TGC-3Ј) and then cloned in pMosBlue plasmid. After dideoxynucleotide sequencing, the mutated DNA fragments were digested by NdeI and BamHI restriction enzymes and subcloned into pET16b plasmid. This plasmid allows the fusion of a 10-His residue extension at the Nterminal part of the mutated thioredoxins, allowing purification on Ni 2ϩ column (see below). A second round of PCR amplification was done with an oligonucleotide (16bMluIHis, 5Ј-TTTACGCGTAAGAAG-GAGATATACCATGGGC-3Ј) designed to introduce a MluI restriction site upstream from the start codon of the polyhistidine extension, and with the same BamHI site containing oligonucleotide. To express thioredoxins and thioredoxin mutants in yeast, we cloned the corresponding ORFs by using the two unique cloning sites MluI and BamHI, present in the shuttle vector YCp2. This plasmid contains the URA3 gene as a selectable marker and ensures protein production after induction of the GAL1 promoter. Dideoxynucleotide sequencing was performed with the YCp5Ј oligonucleotide (5Ј-CCTCTATACTTTAACGT-CAAGG-3Ј), upstream from the MluI site in YCp2 plasmid.
Expression and Purification of Recombinant Thioredoxins on Ni 2ϩ Column-Recombinant site-mutated Trx were purified from E. coli and S. cerevisiae cells. Expression conditions are different for E. coli and S. cerevisiae but purification procedures are the same. BL21(DE3) containing pET16b-mutated Trx were grown in 1 liter of LB medium up to OD 600 nm ϭ 0.5, and recombinant protein expression was induced by the addition of 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside for 3 h at 37°C. Cells were pelleted and stored at Ϫ80°C for the subsequent protein extraction. 10-ml cultures of EMY60 or EMY63, containing YCp-polyHis-mutated Trx, were grown at 30°C in YNBRaf appropriate medium, and then diluted in 5 liters of YNBRGal medium with an initial OD 600 nm ϭ 0.05. Cells were collected at OD 600 nm ϭ 1, and stored at Ϫ80°C. E. coli-or S. cerevisiae-induced cells were broken by a hydraulic press (Carver, model 3968) at 3500 p.s.i., at Ϫ80°C. Broken cells were resuspended in 10 ml of chilled 1 ϫ Binding Buffer (His.Bind TM System, Novagen), containing a protease inhibitor mixture (Complete TM Mini, EDTA-Free, Roche Molecular Biochemicals) and DNase A. Soluble proteins were recovered by centrifugation at 15,000 ϫ g for 15 min. Cell fragments were frozen and submitted two more times to protein extraction. All protein-containing supernatants were pooled and applied on the His.Bind TM (Ni 2ϩ resin) equilibrated column, and proteins were purified as recommended by Novagen. Purified proteins were washed with 50 mM Tris-HCl, pH 7.5, by ultrafiltration on a Microcon column (Amicon-Millipore) and submitted to subsequent analyses.
Enzymatic Activities-His-tagged YTRX1 thioredoxin activity was tested in a 500-l reaction volume using the insulin-disulfide reduction assay (24). YTRX1 served as positive control and bovine serum albumin as a negative control. The assay was monitored by addition of 1 mM DTT, and measurements were performed at OD 650 nm for 45 min on a spectrophotometer (Model DU7400, Beckman). Metal-catalyzed oxidation DNA cleavage protection assays were performed as described previously (25) with the following modifications. Reactive oxygen species were generated for 30 min at room temperature by addition of 0.32 mM DTT in the presence of 3 M FeCl 3 . Reactions mixtures were incubated at room temperature with 20 M YLR109 and 1 g of plasmid DNA for 4 h. DNA degradation was checked by electrophoresis. His-tagged YLR109 and AtTPX2 peroxidase activity assays were performed as follows: the reaction was initiated by the addition of either A. thaliana NADPH Trx reductase (NTR: 180 nM) or H 2 O 2 (100 M) to 1 ml of 30 mM Tris-HCl, pH 8, reaction medium containing, 670 nM Chlamydomonas reinhardtii Trx h, 200 M NADPH, 10 -50 g (0.5-2.5 M) of Prx. The reaction was followed spectrophotometrically at 340 nm at 20°C.
Immunoblotting-Purified proteins were submitted to denaturing SDS-PAGE on a standard 15% polyacrylamide gel (20), under reducing or nonreducing conditions. Gels were subsequently electroblotted for 40 min onto a nitrocellulose membrane. Rabbit polyclonal antibodies against YTRX1 and AtTRX3 (1:10000 dilution), goat anti-rabbit horseradish peroxidase conjugate (Bio-Rad), and a colorimetric detection system (HRP color, Bio-Rad) were used to visualize protein and complexes according to the manufacturer's instructions.
Two-dimensional SDS-PAGE-Purified proteins were diluted in the standard SDS buffer without ␤-mercaptoethanol and submitted to a first SDS-PAGE on a 12% polyacrylamide gel. Gel slices were treated for 1 h with the reducing SDS buffer (containing 5% ␤-mercaptoethanol), applied on the top of 12% polyacrylamide gels and submitted to a second SDS-PAGE. Proteins were stained by Coomassie Blue, or revealed by silver nitrate detection, using the Bio-Rad Silver Stain Plus kit.
H 2 O 2 Sensitivity and Flow Cytometry-To test H 2 O 2 sensitivity, the transformed cells were first grown in YNBRGal up to a concentration of 10 7 cells per ml, and then diluted to OD 600 nm ϭ 0.2. Four 1/5 dilutions of this cell suspensions were prepared and a 15-l droplet of each were plated on YNBRGal-agar medium containing different H 2 O 2 concentrations. Plates were then incubated 3 days at 30°C. DNA content of the transformant cells was measured by a flow cytometry technique. 10 ml of cells were grown in YNBRGal medium to OD 600 nm ϭ 0.5, centrifuged, and washed with 10 ml of Tris-HCl, 50 mM, pH 8. Cells were fixed overnight at 4°C in 70% ethanol, centrifuged, and resuspended in 1 ml of Tris-HCl, 50 mM, pH 8, containing 1 mg ml Ϫ1 RNase A, and incubated for 2 h at 37°C. After centrifugation, the cell pellet was treated for 1 h at 50°C with 50 l of proteinase K at 10 g ml Ϫ1 , resuspended in 500 l of propidium iodide solution (50 g ml Ϫ1 ), and stained overnight in the dark at 4°C. Fluorescence was analyzed using a Bruker ACR 1000 flow cytometer. Data were collected with the FRIED software which was used to estimate the proportion of cells in G 1 phase.

N-terminal His-tagged Yeast Trx 1 Is Fully Active in Vitro and in
Vivo-Because our goal was the isolation of Trx targets and because Trx are not abundant proteins, we chose to use polyhistidine-tagged Trx in order to facilitate protein and complex purification. The first step of our work was to demonstrate that the addition of such His-extension does not modify thioredoxin activity in vivo. For this control, we chose the YTRX1 thioredoxin, which can restore wild-type phenotype when expressed in the Trx double mutant yeast EMY63. The YTRX1 open reading frame was then introduced into a production plasmid, pET16b, using the NdeI/BamHI restriction sites in order to fuse a polyhistidine extension at the N-terminal end. The recombinant protein was then produced in E. coli and purified on an Ni 2ϩ column. The in vitro activity of thioredoxin is usually tested by measurement of insulin reduction and the subsequent ␤-chain precipitation is followed spectrophotometrically at 650 nm (24). We checked that YTRX1 sharing an N-terminal polyhistidine extension is fully active in the reduction of human insulin (data not shown). We also had to check that the His-tagged YTRX1 is still able to complement EMY63 phenotype. For this purpose, the His-tagged ORF was transferred into the centromeric shuttle plasmid YCp2 under control of the inductive GAL1 promoter. Transformed yeast EMY63 (⌬ YTRX1, ⌬ YTRX2) were selected on YNBRaf minimal medium, and then transferred to the YNBRGal medium to ensure induction (see "Experimental Procedures"). The His-tagged YTRX1 complements EMY63 in all aspects, allowing growth on sulfate or methionine sulfoxide, re-establishing a normal cell cycle and H 2 O 2 tolerance (not shown).
HisYTRX1C35S and HisAtTRX3C35S Induce a Partial Dominant Negative Phenotype in S. cerevisiae-We have constructed two shuttle plasmids, one encoding the HisYTRX1C35S mutant and the second encoding HisAtTRX3C35S, a Histagged and mutated version of A. thaliana AtTRX3. Expression of the wild-type AtTRX3 allows EMY63 to grow with a normal cell cycle on methionine and to use methionine sulfoxide efficiently as sole sulfur source. AtTRX3 also allows EMY63 to tolerate 0.8 mM H 2 O 2 , but does not allow growth on sulfate. Based on the catalytic mechanism of Trx, if the mixed disulfide intermediates are stable, C35S Trx mutants should compete with the endogenous Trx for the target proteins and induce a mutant phenotype in the wild-type yeast.
After introduction and induction of the mutant Trx in the wild-type yeast EMY60, the cells remained able to grow on sulfate and on methionine sulfoxide as sole sulfur source. This suggested that the mutated proteins do not efficiently compete with the endogenous Trx for the target proteins implicated in these aspects of the phenotype (not shown). In contrast, a small reduction of the number of cells in the G 1 phase is observed in cells expressing HisYTRX1C35S and HisAtTRX3C35S, compared with EMY60 YCp2 wild-type cells (Fig. 1). These results suggest a weak interaction between these mutated Trx and a protein involved in the S phase. But the most important phenotypic effect is a reduction of H 2 O 2 tolerance induced by HisAtTRX3C35S as is shown by the reduced growth of EMY60 expressing HisAtTRX3C35S on 0.5 mM H 2 O 2 (Fig. 2B). EMY60 expressing HisAtTRX3C35S was almost unable to grow on a medium containing 0.8 mM H 2 O 2 (Fig. 2C). This strongly suggests that HisAtTRX3C35S competes efficiently with the endogenous yeast Trx, probably by the formation of a stable complex with the protein target(s) implicated in H 2 O 2 tolerance. In contrast, expression of HisYTRX1C35S does not modify H 2 O 2 tolerance of EMY60 cells (Fig. 2, B and C), because of a weak production of HisYTRX1C35S recombinant protein (data not shown).
Purification and Characterization of Mixed Disulfide Intermediates-In order to isolate the disulfide intermediate, cellular extracts of EMY60 and EMY63 yeasts expressing AtTRX3C35S were purified on Ni 2ϩ column, under nondenaturing, nonreducing conditions (26). Purified extracts were then submitted to denaturing SDS-PAGE under nonreducing (without ␤-mercaptoethanol) conditions to preserve disulfidebonded complexes, and analyzed by Western blot with anti-AtTRX3 antibodies (Fig. 3). For comparison, extracts of E. coli expressing HisAtTRX3C35S were purified on Ni 2ϩ column and analyzed in the same way. In E. coli, most HisAtTRX3C35S is present as dimers and trimers (Fig. 3A). Although dimers could result from a disulfide bridge between C32 of the two monomers, the presence of trimers suggests that the additional cysteine situated in the N-terminal part of the protein is also implicated in disulfide bridges. In EMY60, most HisAtTRX3C35S is monomeric and some dimers are present. Only one faint band that could correspond to a complex is detectable at 50 kDa. In EMY63 (the Trx minus strain), although HisAtTRX3C35S is expressed at equivalent levels as in EMY60, a far more complex pattern is observed. Using anti-AtTRX3 antibodies, five abundant bands are detectable by Western blot. A two-dimensional electrophoresis consisting in a nonreducing SDS-PAGE as first dimension followed by a reducing SDS-PAGE was analyzed by silver staining (Fig. 3B). Two bands of the first dimension correspond to the Trx monomer and dimer, while three bands (molecular mass 34, 36, and 50 kDa) are composed of Trx associated with one protein of 20 kDa. The silver staining also reveals the presence, in the Ni 2ϩ column eluate, of a free amount of the 20-kDa protein which is not disulfide-bonded to a Trx and which was consequently not detected by immunoblotting of the first dimension. One possibility is that the mutated Trx is bridged with one subunit of a target dimer, stabilized by noncovalent interactions. In the presence of SDS, the dimeric structure is destroyed, releasing a free target monomer and a Trx-target complex. In addition, a faint band of 40 kDa shown by an arrow is detectable. This band is abundant in large scale purifications performed using EMY63 and EMY60 cells and corresponds to alcohol dehydrogenase I (ADHI P00330), a Zn 2ϩ -binding protein, and which shows affinity for Ni 2ϩ . This protein is not related to Trx. In contrast, the 20-kDa band is present only in EMY63 cells expressing the mutated Trx HisAtTRX3C35S.
The Target Protein YLR109 Is Related to Abundant Prokaryotic and Eukaryotic Proteins-The 20-kDa protein band was purified from 1 liter of EMY63/HisAtTRX3C35S cells in exponential growth (OD 600 nm ϭ 0.5) on Ni 2ϩ column followed by a reducing SDS-PAGE. The Coomassie Blue-stained band was submitted to microsequencing. Because the N-terminal part of the protein was blocked, trypsin peptides were analyzed and showed the following sequences: ETNPGTDVTVSSVE, MEV(DV)Q(VA)I(VI)K, DQVI(VA)VTXDNPYA, IGFELAVG-DGVYXS, A(NY)(PI)(VQ)Q(IV)(TG)(SF)N(AM)FQA, and (FM)-P(GQ)TVYPDV. All these sequences are present in the yeast ORF product YLR109, a predicted 19.1-kDa cytosolic protein of unknown function.
Similarities with other proteins were searched using blastx in the NCBI nonredundant data base. It appears that YLR109 is highly related to a peroxisomal membrane protein from Candida boidini (27) and to a small number of proteins from very different sources including prokaryotes (Synechocystis, Hemophilus, and Rhyzobium) and fungi (Aspergillus, Malassezia, and Lipomyces). We have searched for similar sequences in EST data bases using tblastn. Thirty-one human, 75 mouse, and 5 Drosophila ESTs allowed the reconstruction of a unique and complete cDNA for each organism. In higher plants, 32 A. thaliana ESTs were found which can be build in to two contigs. We have fully sequenced one complete clone of each contig corresponding to two highly related cDNAs, deposited under the names AtTPX1 (GenBank AF121355) and AtTPX2 (Gen-Bank AF121356). One complete cDNA was also reconstituted from ESTs for rice and Populus and partial clones can be deduced from maize, Pinus, and Ricinus ESTs. This suggests the presence of highly conserved genes in fungi, animals, plants, and bacteria.
The multiple alignment shown in Fig. 4 indicates that the conservation is particularly high around Cys 62 of YLR109. The good conservation of the C-terminal part of the proteins for all eukaryotic members is obvious. This part of the sequence addresses C. boidini PMP20 to the peroxisome, suggesting that in eukaryotes, all these proteins could be located in the peroxisome. In contrast, the prokaryotic sequences present various C termini. The Hemophilus sequence shows an interesting particularity: it is composed of two domains, the N-terminal region, which is similar to YLR109, followed by a C-terminal domain similar to a glutaredoxin active site. This structure reinforces the idea that all these proteins interact with dithiol reducers.
No function has been attributed to YLR109 and homologous proteins so far, but they are abundantly accumulated in Saccharomyces and in Arabidopsis callus as shown by the size of the spots on the proteomes publicly available at http://www. proteome.com/graph1.html for Saccharomyces cerevisiae and http://www.rs.noda.sut.ac.jp/ϳkamom/2de/2dacallus.html for A. thaliana. The N-terminal sequence of Spot PA0022 (hypothetical protein QA100011) exactly matches the AtTPX1 sequence. YLR109 have also recently been characterized to be three times more abundant after H 2 O 2 treatment of S. cerevisiae cells (28).
YLR109 Shares Similarities with Thioredoxin-dependent Peroxidases-We further compared the YLR109 sequence to other proteins sharing weaker similarities. We found that YLR109 shares some similarity with other proteins characterized as thioredoxin-dependent peroxidases (TPx), also called TSA (thiol-specific antioxidant). The sequences around the putative catalytic active site of these proteins are more conserved. Cysteine 62 in YLR109 is always aligned with a conserved cysteine in the active site of these different thioredoxin-dependent peroxidases. We have constructed a phylogenic tree using DARWIN (29) with most TPx-related proteins (Fig. 5), including TSA homologues, alkyl hydroperoxide reductases, bacterioferritin comigratory proteins, 1-Cys peroxiredoxins (Prx), and other TPx homologues that have not been classified up to now. The tree clearly shows four distinct clusters (Fig. 5). Cluster 1 includes all classical 2-Cys TSA first discovered in yeast (30), the mammal TPx including four human sequences, the nuclear encoded chloroplastic Bas proteins from higher plants, which are close to the chloroplast encoded Bas of the red alga Porphyra and to the prokaryotic homologue of the blue alga Synechocystis. Numerous sequences from prokaryotes belong to this cluster: a subcluster associates bacterial alkyl hydroperoxide reductases, redox-dependent peroxidases which are reduced directly by a NADPH-dependent reductase, without a thioredoxin intermediate. Cluster 2 associates a subgroup of archebacterial sequences with another subgroup formed of 1-Cys peroxiredoxins (31). The sequences include rehydrins, a particular set of cytosolic proteins from plant seeds, and other human, yeast, and Synechocystis sequences. Cluster 3 is composed of sequences from prokaryotes and fungi. No biochemical information is available for these sequences which are described as bacterioferritin-associated proteins. Cluster 4 associated sequences similar to YLR109 as described previously. Sequences from fungi, animals, plants, and prokaryotes are members of this group.
YLR109 Displays Peroxidase Activity in Vivo-An important step in our work was to respond to the hypothesis that YLR109 and homologues belong to a new peroxidase family. We decided to test the ability of YLR109 to ensure a protection against H 2 O 2 in vivo. We first cloned the YLR109 sequence in the inducible shuttle plasmid YCp2, and overexpressed the corresponding protein in EMY60 wild-type cells on YNBRGal medium. These cells were plated at various dilutions on YNBRGal medium containing several dilutions of H 2 O 2 , and growth on this medium was compared with wild-type cells (Fig. 6). As indicated by the result of the experiment, YLR109 protects efficiently against H 2 O 2 in vivo, since the cells overexpressing YLR109 are able to grow at low cell concentration at 0.8 mM H 2 O 2 . Indeed, these cells overexpressing YLR109 are even able to grow on a medium containing up to 1.2 mM H 2 O 2 (not shown), a concentration which is lethal for wild-type cells.
YLR109 and Its Arabidopsis Homologue AtTPX2 Present a Thioredoxin-dependent Peroxidase Activity in Vitro-The abil-ity of HisAtTRX3C35S to reduce H 2 O 2 tolerance in EMY60 (dominant negative phenotype), the ability of YLR109 to increase H 2 O 2 tolerance in EMY60 and the similarity of YLR109 to thioredoxin peroxidases suggest that YLR109 and the other unidentified proteins of the same phylogenic group may be thioredoxin-dependent peroxidases. One characteristic of the peroxiredoxin family to which YLR109 belongs is the ability to form dimers. To test the property of YLR109 to form such a structure, recombinant His-tagged YLR109 protein was produced in E. coli and purified from the soluble fraction of the bacterial cells. One part of the purified protein was diluted in SDS sample buffer in the presence of ␤-mercaptoethanol and incubated at room temperature for 5 min (denaturing reducing condition). The second part was diluted in SDS sample buffer in the absence of ␤-mercaptoethanol (denaturing nonreducing condition). Both samples were analyzed by SDS-PAGE (Fig. 7). The His-tagged version of YLR109 is mainly present as a dimer of 42 kDa under nonreducing conditions, whereas it is present as a monomer of 23 kDa under reducing conditions. Both subunits are bridged by a disulfide bond, which can be reduced by ␤-mercaptoethanol, releasing 23-kDa monomers.
Other in vitro tests were also set up in order to characterize TPx activity further. We first used "the plasmid protection test:" in the presence of an electron donor such as DTT or ascorbate, Fe 3ϩ catalyzes the reduction of O 2 to H 2 O 2 , which is further converted by the Fenton reaction to hydroxyl radicals (HO ⅐ ) (32). These reactive oxygen species can inflict damage on various biomolecules, including proteins and DNA. In this test, TPx are known to prevent such damage by removing H 2 O 2 preventing the Fenton reaction. We therefore investigated whether YLR109 can protect DNA from damage induced by this metal-catalyzed system (Fig. 8). In our plasmid protection test, 0.32 mM DTT in the presence of 3 M FeCl 3 were able to degrade 1 g of plasmid DNA. Both FeCl 3 and DTT are necessary for plasmid degradation as shown by the smear observed in Fig. 8. 20 M YLR109 efficiently protects DNA while 20 M bovine serum albumin is not efficient. This result is in agreement with a redox-dependent peroxidase activity of this protein. The thioredoxin dependence of the peroxidase activity of YLR109 was demonstrated by constructing a complete reduction system with recombinant proteins purified from E. coli: Trx reductase (NTR) was produced from an A. thaliana clone (33), Trx h from C. reinhardtii (34) (35). Table I shows the requirements of the H 2 O 2 reduction assay. No NADPH oxidation was recorded in the absence of either NTR or thioredoxin indicating that these proteins were necessary for transmitting the reducing power. The reaction was also strictly dependent both on the presence of YLR109 and H 2 O 2 . Thus, under these conditions YLR109 is able to reduce H 2 O 2 using electrons from NADPH (see also Fig. 9A). This clearly establishes that YLR109 is a Trx-dependent peroxidase. The H 2 O 2 reducing activity was dependent on the amount of Prx added (Fig. 9A)  min Ϫ1 per mg of YLR109 Ϫ1 does not differ significantly from values reported previously for classical 2-Cys TPx (35,36). In this assay system, the polyhistidine tail of YLR109 does not seem to not prevent this protein from being an efficient catalyst for H 2 O 2 reduction.
The effect of increasing H 2 O 2 concentrations on the Prx activity was tested. The reaction exhibited a Michaelis type saturation with a K m H 2 O 2 of about 14 M (Fig. 9B). In addition, H 2 O 2 inactivates the Prx especially at high concentration (36). This is apparent in the kinetics of NADPH oxidation at 340 nm, the rate of which decreases as time increases (data not shown).
The YLR109 protein used C. reinhardtii Trx h as a substrate with a very good affinity (K m about 1 M). This type of Trx possesses a canonical active site (WCGPC). We have also tested in the same conditions AtTRX3 which possesses a WCPPC active site. The kinetics of saturation with this alternate donor were extremely similar to those obtained with the WCGPC type with a K m nearly identical (data not shown).
We have produced and purified a His-tagged version of the protein encoded by clone 149H22T7, which encodes AtTPx2 from Arabidopsis. In the same in vitro test, AtTPx2 shows a typical Trx-dependent peroxidase activity, but the protein is far less stable that its Saccharomyces counterpart.

Possible Strategies to Characterize Proteins Interacting in
Vivo with Thioredoxins and Glutaredoxins in the Thiol-mediated Redox Cascade-In this report, we describe an improved biochemical system for purifying the target proteins of Trx implicated in thiol reduction. Most disulfide-regulated proteins can be activated in vitro by Trx in an almost unspecific way. On the other hand, genetic evidence suggests that Trx undergo specific interactions with a limited number of proteins in vivo. In order to identify unambiguously the function of Trx, characterization of in vivo thioredoxin-protein complexes is needed. Presently the most popular method to characterize proteinprotein interactions is the two-hybrid system which was very efficient in defining kinase/phosphatase cascades. Up to now, only three articles report on two-hybrid characterization of Trx complexes (14,15,37). But in these cases, it is not clear whether Trx participates in a redox cascade. In our group, we were unable to isolate putative AtTRX3 targets from an Arabidopsis two-hybrid library. Immunoprecipitations using anti-TRX antibodies were no more efficient. The most probable cause is that the half-life of the TRX-target complexes is very short. The second difficulty is the relatively low abundance of Trx in vivo.
The biochemical approach that we have developed in this study solves both aspects. The mutation of the second cysteine in the catalytic site of the Trx stabilizes the mixed disulfide intermediate which can be efficiently isolated by Ni 2ϩ chromatography, involving the N-terminal polyhistidine extension added to the Trx. The isolation of YLR109 shows that at least this mixed disulfide intermediate is sufficiently stable in vivo to allow the isolation of the complex. Surprisingly, this complex could be isolated only in the ⌬ Trx yeast presumably because the wild-type Trx attack the disulfide bridge of the mixed disulfide. This suggests that the two-hybrid approach could be efficiently improved using a C35S Trx mutant as bait and, if that was not sufficient, by using a ⌬ Trx yeast as reporter strain. We have recently screened a yeast two-hybrid library with a C35S mutant of YTRX1, one of the yeast Trx, fused to the GAL4-binding domain and isolated several putative Trx targets which are now under study. Nevertheless YLR109 was not among these clones. More surprisingly, a binary two-hybrid system with YTRX1C35S fused to the activation domain of GAL4 and YLR109 fused to the DNA-binding domain of GAL4 failed to show interaction. We have no interpretation of this result, but the large amount of free YLR109 in yeast probably competes with the hybrid GAL4-YLR109 protein in the interaction with the hybrid GAL4-Trx. Despite our success in isolating YLR109, no other protein target has yet been isolated by this method, although AtTRX3 is not only able to confer H 2 O 2 tolerance but also induces a normal cell cycle and rapid growth on methionine sulfoxide as sole sulfur source. The present failure to detect additional complexes may be due to their low concentration, or possibly to an interaction that limits the efficiency of the retention on the Ni 2ϩ affinity column. A lower stability of these mixed disulfides in the ⌬ Trx yeast cannot be excluded because it remains able to synthesize other reducers, like glutathione and Grx. Our results clearly show that both methods are complementary and suggest modifications of the two-hybrid approach that could be necessary to detect low abundance targets. In addition, the recent characterization of stable mixed disulfides between the E. coli Grx1 (C14S) and a peptide from the ribonucleotide reductase B1 suggests that the same methods could be used for the characterization of Grx targets (38). Furthermore, mixed disulfide intermediates have been obtained from E. coli with glutathione, by mutating the Grx in the more N-terminal cysteine of it active site C14S (39,40), and also between E. coli TrxA C32S and its Trx reductase (41). Thus, a similar approach would be likely to help to discriminate between the possible reducers of Trx and Grx in the case of multiple thioredoxin reductase genes, as is the case for A. thaliana, or to identify the reducer when this one is not known, as for chloroplastic APS reductase from A. thaliana. This protein shows homology to Trx but is reduced by glutathione in vitro and in E. coli (6).
YLR109 Defines a New Group of Peroxiredoxins-Our biochemical method to identify Trx targets led us to isolate and to characterize a new target in yeast (YLR109), of unknown function up to now. In contrast to most peroxidases which use cofactors to reduce H 2 O 2 , YLR109 and its related proteins belong to the recently characterized family of Prx, a set of enzymes which transfer their reducing power by means of a cysteine.
The first member which defines the first group of this large family was discovered in yeast and first named thiol-specific antioxidant (30), but characterized later as a true peroxidase and renamed 2-Cys Prx. This protein catalyzes the reduction of H 2 O 2 and alkyl hydroperoxides in vitro with the use of electrons from the Trx system (35). Yeast Prx exists as a homodimer and contains two essential Cys residues in each subunit. The Cys 47 -SH group is the primary site of oxidation by H 2 O 2 , and the oxidized Cys (probably a sulfenic acid form, Cys-SOH) rapidly reacts with the Cys 170 -SH of the other subunit to form an intermolecular disulfide. This disulfide is subsequently reduced by a Trx, and mutant TPx proteins lacking either Cys 47 or Cys 170 therefore do not exhibit Trx-coupled peroxidase activity (42). 2-Cys Prx corresponds to cluster 1 of the tree on Fig. 5. S. cerevisiae presents two very similar 2-Cys Prx genes, humans have at least four different genes. In higher plants, all 2-Cys Prx described so far are nuclear-encoded chloroplastic proteins.
The second group of Prx possesses only one conserved cysteine residue and is consequently designated as 1-Cys Prx. The first member was characterized as an antioxidant in barley seeds (43), then homologues were found in most plants, archebacteria, and animals (cluster 2 in our phylogenetic analysis on Fig. 5). The human 1-Cys does not form a disulfide and DTT acts in vitro as an efficient reducer of the 1-Cys Prx, but the natural electron donor remains unidentified, glutathione and Trx being inefficient (44). Recent advances on crystal structure of this human Prx reveals that the C-terminal domain of this protein is used for dimerization, and that the active site cysteine (Cys 47 ) exists as cysteine-sulfenic acid in the crystal (45).
In this work, we have demonstrated the peroxidase activity for YLR109 and AtTPX2, two distant members of the group. This suggests that all the members that we have identified in our phylogenic analysis (group 4 on Fig. 5) should be Prx. Our work shows that the NADPH/Trx reductase/Trx system is a very efficient electron donor for these Prx in vitro. Thus, despite the slightly higher sequence similarity of YLR109 homologues with 1-Cys Prx and the presence of only one conserved Cys (Cys 62 for YLR109), these proteins seem to be functionally closer to 2-Cys TPx than to 1-Cys TPx. This is supported by our experiment showing that YLR109, like TPx, can adopt a disulfide-bonded dimeric structure (Fig. 6A). Furthermore, we show that YLR109 is dependent on a functional NADPH/Trx reductase/Trx system to reduce H 2 O 2 . Finally, our study is the first evidence for an in vivo interaction between a Prx and the Trx system.
Physiological Function of YLR109 and the Thioredoxin Reduction System-We have previously shown that AtTRX3 can restore H 2 O 2 tolerance to the budding yeast EMY63, lacking the two Trx genes (13). This means that AtTRX3 interacts specifically with a protein involved in H 2 O 2 tolerance. Our results are in good agreement with recent two-dimensional analysis, since after H 2 O 2 treatment of S. cerevisiae cells, the amount of Trx and YLR109 increases 11 and 3 times, respectively (28). These authors suggest that YLR109 may be an antioxidant protein. We demonstrate that AtTRX3C35S interacts strongly with YLR109 and, at the same time, induces a partial dominant negative phenotype essentially limited to H 2 O 2 tolerance. In addition, YLR109 presents a Trx-dependent peroxidase activity in vitro and overexpression of this ORF in EMY60 increases H 2 O 2 tolerance. All these facts suggest that in vivo, Trx transfers reducing equivalents from NADPH to YLR109 through a thiol-mediated cascade allowing the degradation of hydrogen peroxide. Thus, in this interaction, the NADPH/Trx reductase/AtTRX3 cascade appears to transfer an energetic flux rather than to modify the structure of the targeted protein.
Nevertheless, the external application of H 2 O 2 is an artificial situation which is probably experienced by yeast and other organisms only in the laboratory. This poses the question of the real function of YLR109 and its Trx-mediated reduction. YLR109 and AtTPX2 are abundant cytosolic proteins as shown on the proteomes, and the number of ESTs in plants and mammals indicates that the corresponding genes are also very actively transcribed in these organisms. The simplest hypothesis is that YLR109 homologues eliminate the excess H 2 O 2 or other peroxides, like alkyl hydroperoxides, produced by metabolism. In relation to this hypothesis it is important to remember that PMP20a and PMP20b, two YLR109 homologues from C. boidini, are peroxisomal proteins. Furthermore, the conservation of the C-terminal end of all eukaryotic YLR109 homologues suggests a peroxisomal location. Peroxisomes are a major source of H 2 O 2 production due to fatty acid degradation in all eukaryotes and to photorespiration in plants. Thus, these Prx may help catalase in the elimination of H 2 O 2 from peroxisomes, to prevent its diffusion into the cytosol, and/or may reduce membrane bound alkyl-hydroperoxides, for which catalases are inefficient. Control of cytoplasmic H 2 O 2 concentration is crucial for cells, since diverse stimuli have been shown to use reactive oxygen species (e.g. H 2 O 2 ) as transduction signals for regulating transcription factors like NF-B, AP1, and OxyR, via the formation of an internal disulfide bridge (46 -48). In mammalian cells, the tumor necrosis factor ␣ and growth factors (epidermal growth factor and platelet-derived growth factor) are known to induce a transient increase in intracellular concentration of H 2 O 2 (49). It was recently shown that the overproduction of the mammalian Prx II blocks the NF-B activation induced by exogenous H 2 O 2 or tumor necrosis factor ␣ (44). Moreover, the activation of NF-B was also prevented by a rapid removal of H 2 O 2 by catalases (50). These data reinforce a possible function of Prx in H 2 O 2 removal. Human 2-Cys Prx (TPx II) was also characterized as a potent inhibitor of cytochrome c release from mitochondria to cytosol, and of lipid peroxidation in cells (51). In all these cases, this TPx II could protect cells from apoptosis. In higher plants, H 2 O 2 is a well established signal in response to wounding (52) and pathogen interactions (53)(54)(55). Furthermore, recent evidence shows that sulfhydryl blockers induce an H 2 O 2 burst (56). Thus, Trx-dependent peroxidases could play a central role in signal transduction and in response to pathogens. Isolation of YLR109 mutants in yeast and of the homologues in other organisms will probably be necessary to define the implication of these proteins and their Trx-reduction dependence in a general antioxidant mechanism, and/or in a more subtle function in signaling pathways.