1-Cys Peroxiredoxin, a Bifunctional Enzyme with Glutathione Peroxidase and Phospholipase A 2 Activities*

This report provides definitive evidence that the protein 1-Cys peroxiredoxin is a bifunctional (“moonlight-ing”) enzyme with two distinct active sites. We have previously shown that human, rat, and bovine lungs contain an acidic Ca 2 1 -independent phospholipase A 2 (aiPLA 2 ). The cDNA encoding aiPLA 2 was found to be identical to that of a non-selenium glutathione peroxidase (NSGPx). Protein expressed using a previously reported E. coli construct which has a His-tag and 50 additional amino acids at the NH 2 terminus, did not exhibit aiPLA 2 activity. A new construct which contains the His-tag plus two extra amino acids at the COOH terminus when expressed in Escherichia coli generated a protein that hydrolyzed the sn -2 acyl chain of phospholipids at pH 4, and exhibited NSGPx activity with H 2 O 2 at pH 8. The expressed 1-Cys peroxiredoxin has identical functional properties to the native lung enzyme: aiPLA 2 activity is inhibited by the serine protease inhibitor, diethyl p -nitrophenyl phosphate, by the tetra-hedral mimic 1-hexadecyl-3-trifluoroethylglycero- sn -2-phosphomethanol (MJ33), and by 1-Cys peroxiredoxin monoclonal antibody experiments were performed according to the manufacturer’s protocols. Briefly, the mutagenic oligonucleotides contain a single or double base change in the middle of the sequence and were synthesized with a 5 9 -phosphate. Oligonucleotide sequences used to generate the C47S and S32A mutations are 5 9 -CCCCAGTGTcCACCACAGAG-3 9 and 5 9 -TCT- GGGAGACgCATGGGGCATT-3 9 which are located between nucleotides 131–150 and 83 and 105, respectively, in the open reading frame of 1-Cys peroxiredoxin (1). The numbering of oligonucleotides in the sense strand of 1-Cys peroxiredoxin begins at the start codon and the mutated nucleotides are in lowercase letters. plasmid DNA (0.03 pmol) and phosphorylated mutagenic oligonucleotide (50 ng/ m l) were denatured at 100 °C for 5 min, incubated in an ice water bath for 5 min and then at room temperature for 30 min. The reaction was started by adding MORPH y synthesis buffer, T4 DNA polymerase, and T4 DNA ligase, and incubation was fo r 2 h at 37 °C tosynthesize a non-methy- lated replacement strand and to ligate it into a circular molecule. After stopping the reaction by heating to 85 °C for 15 min, Dpn I was added to the reaction at 37 °C and incubated for 30 min to destroy the original, non-mutagenized methylated target plasmid DNA. After chilling for 5 min, the reaction was transformed into 200 m l of competent E. coli mutS cells, which were plated after heat shock at 42 °C for 2 min. Colonies were screened through digestion of the corresponding plasmid DNA with the specific enzyme for the designed mutation. and radiolabeled free fatty acids were sepa-rated by a two-step TLC procedure using hexane/ether/acetic acid as a solvent system. palmitic acid was co-chromatographed. The free fatty acid were identified using I 2 scraped from the plate, and analyzed by scintillation counting. For the fluorescence assay, the liposomal substrate was bisbodipy-C 11 -PC/DPPC/phosphatidyl- glycerol/cholesterol in a molar ratio of 0.05:10:2:3. obtaining a steady baseline with buffer and liposomes, enzyme was added product was measured at 490 nm excitation and 520 nm emission. For evaluation of effects of inhibitors, MJ33 (3 mol of lipid) was added to the lipid mixture prior to preparation of the liposomal substrate. The effect of other inhibitors was evaluated following preincubation of enzyme with inhibitor for 30 min.

This report provides definitive evidence that the protein 1-Cys peroxiredoxin is a bifunctional ("moonlighting") enzyme with two distinct active sites. We have previously shown that human, rat, and bovine lungs contain an acidic Ca 2؉ -independent phospholipase A 2 (aiPLA 2 ). The cDNA encoding aiPLA 2 was found to be identical to that of a non-selenium glutathione peroxidase (NSGPx). Protein expressed using a previously reported E. coli construct which has a His-tag and 50 additional amino acids at the NH 2 terminus, did not exhibit aiPLA 2 activity. A new construct which contains the His-tag plus two extra amino acids at the COOH terminus when expressed in Escherichia coli generated a protein that hydrolyzed the sn-2 acyl chain of phospholipids at pH 4, and exhibited NSGPx activity with H 2 O 2 at pH 8. The expressed 1-Cys peroxiredoxin has identical functional properties to the native lung enzyme: aiPLA 2 activity is inhibited by the serine protease inhibitor, diethyl p-nitrophenyl phosphate, by the tetrahedral mimic 1-hexadecyl-3-trifluoroethylglycero-sn-2phosphomethanol (MJ33), and by 1-Cys peroxiredoxin monoclonal antibody (mAb) 8H11 but these agents have no effect on NSGPx activity; NSGPx activity is inhibited by mercaptosuccinate and by 1-Cys peroxiredoxin mAb 8B3 antibody which have no effect on aiPLA 2 activity. Mutation of Ser 32 to Ala abolishes aiPLA 2 activity, yet the NSGPx activity remains unaffected; a Cys 47 to Ser mutant is devoid of peroxidase activity but aiPLA 2 activity remains intact. These results suggest that Ser 32 in the GDSWG consensus sequence provides the catalytic nucleophile for the hydrolase activity of aiPLA 2 , while Cys 47 in the PVCTTE consensus sequence is at the active site for peroxidase activity. The bifunctional catalytic properties of 1-Cys peroxiredoxin are compatible with a simultaneous role for the protein in the regulation of phospholipid turnover as well as in protection against oxidative injury.
Evidence has emerged that a lysosomal-type Ca 2ϩ -independent phospholipase A 2 with acidic pH optimum (aiPLA 2 ) 1 (1,2) and a non-selenium glutathione peroxidase without glutathione S-tranferase activity (NSGPx) (3,4) are the same enzyme, based upon their identical cDNA sequence. Furthermore, based upon the cDNA sequence, this protein belongs to the thioredoxin peroxidase, or peroxiredoxin, family although with only one instead of the usual two conserved cysteine residues and thus has been called 1-Cys peroxiredoxin (5). The nomenclature is further confusing since the enzyme does not utilize thioredoxin (4,5), but rather functions as a GSH peroxidase (3,4).
Although the cDNA sequence of aiPLA 2 (1) is identical to that of NSGPx (4), the presence of both activities up to the present has not been confirmed in the same peptide. Native protein isolated from bovine eye or rat olfactory mucosa had peroxidase activity but PLA 2 activity was not tested (6,7). Native protein isolated from rat or bovine lungs demonstrated PLA 2 activity although peroxidase activity was not tested (1,2,8). Likewise, studies of recombinant 1-Cys peroxiredoxin have reported either peroxidase (3,4) or PLA 2 (1, 2) activity. One study (5) did evaluate both peroxidase and PLA 2 activity in recombinant protein with somewhat mixed results: 1) the peroxidase activity of the recombinant protein required dithiothreitol and was not supported by GSH; 2) the purified recombinant protein did not have PLA 2 activity; 3) a low level of PLA 2 activity was detected in NIH 3T3 cells transfected with the cDNA although the activity was not affected by mutagenesis of the putative active site serine of the protein. Thus, the presence of both activities in the same protein remains unconfirmed.
cDNA Constructs-The human HA0683 clone (GenBank accession number, D14662), a 1653-bp (exclusive of the poly A tract) cDNA which contains a 672-bp open reading frame of 1-Cys peroxiredoxin, was cloned from a human myeloid cell line, KG-1, into the pBluescript SK(ϩ) expression vector at EcoRV and NotI sites (Fig. 1A) as described previously (1). This is termed construct 1.
An NH 2 -terminal His-tag construct was generated by cloning the major part of HA0683 cDNA (1044 bp) including the complete coding sequence into the pET-28c vector at the HindIII site. Translation initiation in this construct begins 125 nucleotides upstream of the insert, adding 42 amino acids, including 6 histidines (the His-tag), to the amino terminus of the peptide product. The insert itself includes 43 nucleotides of 5Ј-untranslated region, adding 14 additional amino acids to the amino terminus. The insert also includes the 672-bp coding sequence and 329 nucleotides of 3Ј-untranslated region sequence which follow the termination codon and do not contribute to the final translation product (Fig. 1B). The extra nucleotides from the vector and the 5Ј-untranslated region of the insert are predicted to result in 56 extra amino acids with an extra molecular mass of 5986 daltons. This construct (Nterm 1-Cys peroxiredoxin clone) was termed construct 2 and has been reported previously (2,4). A plasmid in which the HindIII insert was inserted in reverse orientation was also prepared and used as a control (construct 2 antisense control).
A third construct was made with no extra NH 2 -terminal amino acids and a His-tag on the COOH terminus (Fig. 1C). The coding region of HA0683 clone (672 bp) was obtained by polymerase chain reaction using 5Ј-CGGAATTCCATATGCCCGGAGGTCTGCTTCTCG-3Ј as the upstream primer and 5Ј-CGGAATTCCTCGAGAGGCTGGGGTGTG-TAGCGGAGG-3Ј as the downstream primer which match up to the first and last codons in the coding region (underlined). In the upstream primer, the start codon was incorporated into an NdeI site (sequence: CATATG). An XhoI codon was incorporated into the downstream polymerase chain reaction primer right after the last amino acid codon, eliminating the natural termination codon and allowing fusion with the His tag coding region. The polymerase chain reaction product was cleaved with NdeI and XhoI to remove the extra nucleotides that are not included in the coding sequence and was purified by agarose gel electrophoresis. The purified polymerase chain reaction product was cloned into pET-21b vector that had been cleaved with the same enzymes. The presence of the insert was verified by digestion of the plasmid DNA with NdeI and XhoI, and further by DNA sequencing. The resulting construct contained only two extra amino acids before the six histidine residues which constitute the His tag at the COOH terminus (Fig. 1C). The two extra amino acids and six histidine residues result in extra molecular mass of 1065 daltons. This was termed construct 3 (Cterm 1-Cys peroxiredoxin clone). Constructs 2 and 3 were expressed in Escherichia coli while construct 1 was utilized in a wheat germ expression system.
Site-directed Mutagenesis-Mutations were made using the MORPH plasmid DNA mutagenesis kit supplied by 5 Prime 3 3 Prime Inc. Cysteine at position 47 was replaced by serine (C47S), and serine at position 32 was substituted by alanine (S32A). Mutagenesis

FIG. 1. Plasmid constructs for expression of 1-Cys peroxiredoxin.
Panel A, construct 1. A 1653-bp human cDNA including the coding region was cloned into pBluescript SK(ϩ) vector. The coding region is boxed, the start codon is underlined, and the termination codon (TAA) is bold. Arrowhead ϭ T7 promoter. This construct was expressed in a wheat germ in vitro translation system. Panel B, construct 2. A 1044-bp human cDNA including the coding region and a His tag at the NH 2 terminus was cloned into pET-28c. Arrowhead ϭ T7 promoter. The insert and His tag are boxed. DNA coding for the start codon is underlined and the termination codon (TAA) is bold. This construct was expressed in E. coli. Panel C, construct 3. A 672-bp cDNA of the human coding region with His tag at the COOH terminus was cloned into pET-21b. Arrowhead ϭ T7 promoter. The insert and His tag are boxed. DNA coding for the start codon and the last amino acid codon are underlined, and the termination codon (TGA) is bold. This construct was expressed in E. coli. experiments were performed according to the manufacturer's protocols. Briefly, the mutagenic oligonucleotides contain a single or double base change in the middle of the sequence and were synthesized with a 5Ј-phosphate. Oligonucleotide sequences used to generate the C47S and S32A mutations are 5Ј-CCCCAGTGTcCACCACAGAG-3Ј and 5Ј-TCT-GGGAGACgCATGGGGCATT-3Ј which are located between nucleotides 131-150 and 83 and 105, respectively, in the open reading frame of 1-Cys peroxiredoxin (1). The numbering of oligonucleotides in the sense strand of 1-Cys peroxiredoxin begins at the start codon and the mutated nucleotides are shown in lowercase letters. Target plasmid DNA (0.03 pmol) and phosphorylated mutagenic oligonucleotide (50 ng/l) were denatured at 100°C for 5 min, incubated in an ice water bath for 5 min and then at room temperature for 30 min. The reaction was started by adding MORPH synthesis buffer, T4 DNA polymerase, and T4 DNA ligase, and incubation was for 2 h at 37°C to synthesize a non-methylated replacement strand and to ligate it into a circular molecule. After stopping the reaction by heating to 85°C for 15 min, DpnI was added to the reaction at 37°C and incubated for 30 min to destroy the original, non-mutagenized methylated target plasmid DNA. After chilling for 5 min, the reaction was transformed into 200 l of competent E. coli mutS cells, which were plated after heat shock at 42°C for 2 min. Colonies were screened through digestion of the corresponding plasmid DNA with the specific enzyme for the designed mutation.
The designed mutations were identified by restriction analysis. The introduction of the S32A mutation creates a HgaI restriction site so plasmid DNAs prepared from colonies in the mutagenesis experiment were cut with HgaI to identify the mutant plasmid DNA that has an extra band of 387 base pairs. The C47S mutation removes an ApaLI site, so plasmid DNAs were digested with ApaLI to detect the mutant plasmid DNA lacking one band of 1124 base pairs. Subsequently, each mutant DNA was sequenced to further confirm that the DNA had the expected mutation and that this is the only mutation present. Only confirmed mutant DNA was further processed for in vitro expression.
For construct 1, the mutant plasmids were cleaved with NarI and BamHI to obtain a 372-bp fragment, and the wild-type plasmid DNA was also cleaved with the same enzymes to get a wild type vector that contained the remainder of the insert. The fragment and the wild type vector were purified by agarose gel electrophoresis and were religated in order to avoid unexpected mutations. The presence of the mutated fragment was confirmed by both DNA restriction analysis and sequencing. Construct 2 was not used for mutation analysis. For construct 3, a fragment of 291 base pairs including the mutation sites for S32A or C47S was obtained by digestion of S32A or C47S 1-Cys peroxiredoxin-pBluescript SK(ϩ) with EagI and BamHI, and was subsequently recloned into 1-Cys peroxiredoxin-pET-21b, which had been digested with the same enzymes, effectively replacing this region with the mutated version. The recloned insert was tested for the correct orientation through digestion with EcoRV. DNA sequencing was used to confirm that only the designed mutation was present in the insert.
In Vitro Transcription-cRNA was expressed from the 1-Cys peroxiredoxin-pBluescript SK(ϩ) wild type and mutant clones using the T7 mMESSAGE mMACHINE in vitro transcription kit. The template DNAs were digested with NotI to linearize the DNA and provide a transcriptional terminus, then purified by phenol/chloroform extraction and isopropyl alcohol precipitation according to standard procedures. A 20-l transcription reaction was assembled with transcription buffer, ribonucleotide mixture, linearized template DNA (1 g), and enzyme mixture including transcriptase, and was incubated at 37°C for 2 h to reach a maximal yield. The remaining template DNA was then removed by adding 2 units of RNase-free DNase I at 37°C for 15 min. The reaction was terminated by adding 30 l of nuclease-free distilled H 2 O and 25 l of 7.5 M LiCl with 75 mM EDTA, and was chilled overnight at Ϫ20°C. The solution was centrifuged at 4°C for 30 min at 14,000 ϫ g to pellet the RNA. The pellet was washed with 70% ethanol. After drying, cRNA was dissolved in nuclease-free distilled H 2 O, and was stored at Ϫ70°C. Meanwhile, cRNA concentration was measured at OD 260 /OD 280 , and was then electrophoresed onto a 1% agarose/formaldehyde gel to evaluate its quality. The transcripts for the wild type, C47S and S32A enzymes are shown in Fig. 2A. Their mobility was near their predicted size of approximately 1750 bp, near the 18 S rRNA of rat lung.
In Vitro Translation-The recombinant wild type and mutant proteins prepared with construct 1 were expressed in vitro in a wheat germ translation kit from Ambion Inc. Briefly, a 50-l reaction contained 2.5 l of 1 M potassium acetate, 2.5 l of minus leucine Master Mix (1 mM amino acids without leucine, 0.16 M creatine phosphate), 2.5 l of minus methionine Master Mix (1 mM amino acids without methionine, 0.16 M creatine phosphate), 25 l of wheat germ extract, 1 g of cRNA, and distilled H 2 O. The reaction was incubated at 27°C for 60 min, and stored at 4°C for further activity assays. Wheat germ translated protein was quantified by [ 35 S]methionine incorporation. In these experiments, 0.5 mM [ 35 S]methionine (63 Ci/mmol) was present in the translation reaction. Otherwise the method was as described previously (4). After translation, 1-Cys peroxiredoxin represented about 0.05% of the total wheat germ extract protein for both wild type and mutant proteins, indicating that wild type and mutant 1-Cys peroxiredoxin cRNA have comparable translation efficiencies. SDS-PAGE with autoradiography of the translated [ 35 S]methionine-labeled protein showed an apparent molecular mass of 26 kDa for wild type, C47S and S32A, similar to the molecular mass deduced from the open reading frame and of the PLA 2 enzyme isolated from rat lung (Fig. 2B). The 672-bp coding sequence should generate a polypeptide of 25,035 daltons.
Human recombinant proteins with constructs 2 or 3 were expressed as fusion proteins with a series of six histidine residues in BL21(DE3) cells. These cells produce T7 RNA polymerase and express pET-28c and pET21b inserts efficiently. Human wild type 1-Cys peroxiredoxin in construct 2 was grown in LB broth as described previously (4). Based on the intensity of the protein band on Coomassie Blue-stained 12% SDS-PAGE gel (not shown), wild type 1-Cys peroxiredoxin is about 13% of total protein. The proteins were purified through Ni 2ϩ columns as described previously (4). The purity of the expressed proteins after the Ni 2ϩ column was about 90% as estimated from SDS-PAGE (Fig. 3A). The purified protein migrated with a molecular mass of about 32 kDa on 12% SDS-PAGE (Fig. 3A) and immunoblot probed with the 1-Cys peroxiredoxin mAb 8H11 (Fig. 3B). The theoretical mass of 1-Cys peroxiredoxin with construct 2 is 31,021 daltons.

1-Cys Peroxiredoxin: A Bifunctional Enzyme
Wild type, S32A and C47S 1-Cys peroxiredoxin in construct 3 were expressed in M9 minimal medium instead of LB broth since the latter was found to give a better yield. After 15 min induction with 1 mM isopropyl-␤-D-thiogalactopyranoside, 100 g of rifampicin/ml was added to culture medium containing 50 g of ampicillin/ml to inhibit the bacterial RNA polymerase and block bacterial protein production while not affecting the T7 polymerase and thus allowing an enrichment of 1-Cys peroxiredoxin. The bacteria were collected by centrifugation and resuspended in binding buffer (5 mM imidazole, 0.5 M NaCl, 20 mM Tris-HCl, pH 7.9). In order to prevent degradation of the recombinant protein, protease inhibitor mixture tablets were added to the binding buffer (one tablet per 25 ml). The crude cell lysate was tested for the presence of the recombinant protein using Coomassie Brilliant Blue gel staining of 12% SDS-PAGE (not shown). Based on the intensity of the protein band on the gel, the expressed wild type, S32A, and C47S proteins represented about 20% of total protein. The protein was purified using the His tag on Ni 2ϩ columns. The purity of the recombinant proteins after the Ni 2ϩ column was about 96% as estimated from Coomassie-stained gels. This construct generated a protein with a molecular mass of about 27 kDa on 12% SDS-PAGE (Fig. 3A) and immunoblot probed with the 1-Cys peroxiredoxin mAb 8H11 (Fig. 3B). This construct is predicted to produce a protein of 26,100 daltons.
Enzymatic Assays-NSGPx activity was assayed by measuring consumption of NADPH in the presence of GSH and GSH reductase with H 2 O 2 as described previously (4). The standard reaction buffer (3 ml) was 50 mM Tris-HCl, 2 mM NaN 3 , 0.1 mM EDTA (pH 8), 0.03 mM NADPH, 0.36 mM GSH, and 0.23 units of GSH reductase/ml. Preincubations and assays were performed at room temperature. The mixture plus enzyme was preincubated for 5 min with continuous stirring. Fluorescence was continuously recorded at 460 nm (340 nm excitation) using a fluorescence spectrophotometer (Photon Technology Instruments, Bricktown, NJ). After a steady base line was achieved, the reaction was started by the addition of H 2 O 2 (250 M) and the linear change in fluorescence was recorded for 5-10 min. The change in fluorescence was corrected for the relatively small base-line nonenzymatic oxidation of NADPH and used to calculate enzyme activity based on authentic NADPH standards. A serine protease inhibitor (DENP), a PLA 2 inhibitor (MJ33), and an inhibitor of thiol-mediated reaction (mercaptosuccinate), as well as 2 monoclonal antibodies against human 1-Cys peroxiredoxin recombinant protein (8H11 and 8B3) were investigated for their effect on enzymatic activity. The enzyme was preincubated with each inhibitor for 30 min. aiPLA 2 activity of expressed protein was measured at pH 4 (40 mM sodium acetate, 5 mM EDTA buffer) using a liposome-based assay based on radiochemical detection (10) or fluorescence detection (1). For the radiochemical assay, the liposomal substrate was labeled with [ 3 H]D-PPC/egg PC/egg phosphatidylglycerol/cholesterol in a molar ratio of 10:5:2:3. The specific activity of [ 3 H]DPPC was 2 mCi/mmol. Lipids were dried under N 2 , resuspended in isotonic saline, repeatedly freeze/ thawed by alternating liquid N 2 and warm H 2 O, and then extruded through a 100-m pore size membrane to generate unilamellar liposomes. The reaction was stopped by the addition of CHCl 3 /CH 3 OH (2:1, v/v); lipids were extracted and radiolabeled free fatty acids were separated by a two-step TLC procedure using hexane/ether/acetic acid as a solvent system. Authentic palmitic acid was co-chromatographed. The free fatty acid spots were identified using I 2 vapor, scraped from the plate, and analyzed by scintillation counting. For the fluorescence assay, the liposomal substrate was bisbodipy-C 11 -PC/DPPC/phosphatidylglycerol/cholesterol in a molar ratio of 0.05:10:2:3. After obtaining a steady baseline with buffer and liposomes, enzyme was added and the fluorescent product was measured at 490 nm excitation and 520 nm emission. For evaluation of effects of inhibitors, MJ33 (3 mol % of total lipid) was added to the lipid mixture prior to preparation of the liposomal substrate. The effect of other inhibitors was evaluated following preincubation of enzyme with inhibitor for 30 min.
Monoclonal Antibodies and Immunoblotting-The production of monoclonal antibody to the recombinant protein (construct 2) and the procedure for immunoblotting have been described previously (2). Monoclonal antibodies 8H11 and 8B3 were from different clones obtained using the same antigen. Coomassie Blue-stained gels were quantified using the FluorS Imager and Quantity One Software (Bio-Rad). Protein was measured with Coomassie Blue (Bio-Rad protein dye binding kit) using bovine ␥-globulin as a standard.

Characterization of Bifunctional Enzymatic Activities-Both
NSGPx and aiPLA 2 activities were evaluated for the protein expressed with the 1-Cys peroxiredoxin clone using three different constructs (Table I). NSGPx activity was measured with H 2 O 2 and PLA 2 activity was measured with [ 3 H]DPPC. The translation product of wild type cRNA (construct 1) expressed in wheat germ showed significant NSGPx activity; it also hydrolyzed DPPC at pH 4 in the absence of Ca 2ϩ to liberate free fatty acid, thus exhibiting aiPLA 2 activity (Table I). These specific activities are based upon protein concentration estimated from incorporation of [ 35 S]methionine. In the absence of cRNA, neither NSGPx nor aiPLA 2 activity was detected from construct 1 in the wheat germ expression system (not shown), indicating that the activity indeed came from the expressed gene fragment.
The purified protein expressed from construct 2 also had significant NSGPx activity although the specific activity was only one-half of that estimated for construct 1 (Table I). In contrast, aiPLA 2 activity with this preparation was extremely low. Thus, the activity ratio of NSGPx to aiPLA 2 for construct 2 is far greater than that for construct 1 (Table I). No peroxidase or phospholipase activity was found with the antisense preparation (data not shown). The extremely low aiPLA 2 activity detected in the recombinant protein expressed from construct 2 could be due to extra amino acids and the His tag at the amino terminus that could modify protein conformation; it was not due to the purification procedure because the activity in the for purified human WT 1-Cys peroxiredoxin using 3 g of construct 2 (labeled as WT2) and WT, S32A, and C47S 1-Cys peroxiredoxin using construct 3 (8 g each). The first lane shows molecular mass standards (STD). The immunoblot utilized monoclonal antibody 8H11 raised from expressed human protein (construct 2) in E. coli. The larger size of protein for WT2 is consistent with the amino acid composition predicted from the cDNA in construct 2.

1-Cys Peroxiredoxin: A Bifunctional Enzyme
crude cell extract was also extremely low (data not shown). Therefore, we generated another construct for expression in E. coli containing only two extra amino acids before the sixhistidine residues which were placed at the COOH terminus (construct 3 shown in Fig. 1C). The protein generated from construct 3 exhibited both aiPLA 2 and NSGPx activities (Table I). The NSGPx activity for construct 3 was significantly higher than for construct 2 and somewhat higher than for construct 1. For aiPLA 2 activity, construct 3 was slightly higher than construct 1 and markedly higher than construct 2. The activity ratios of NSGPx to aiPLA 2 in enzyme generated from constructs 1 and 3 were similar and indicate that the peroxidase activity under the conditions of the assay was 2 orders of magnitude greater than the PLA 2 activity. Thus, we identified both NSGPx and aiPLA 2 activities in translated protein from both wheat germ (construct 1) and E. coli (construct 3) expression systems. Construct 2 exhibited NSGPx but little PLA 2 activity suggesting that PLA 2 activity was sensitive to protein conformation and also that the active sites for the two activities are different.
Additional evidence for two different active sites was obtained by determining the response of the two catalytic activities to inhibitors (Table II). Mercaptosuccinate, an inhibitor of cysteine-and selenocysteine-mediated reactions, inhibited NS-GPx activity by 98% at 20 M, but had no effect on aiPLA 2 activity (Table II). Both DENP, a serine protease inhibitor and MJ33, a transition state phospholipid analogue, inhibited aiPLA 2 activity by approximately 80 -90% but had no effect on NSGPx activity (Table II). Monoclonal antibodies raised against E. coli-expressed human 1-Cys peroxiredoxin also showed differential effects. mAb 8B3 significantly inhibited NSGPx by 88% but aiPLA 2 activity was unaffected whereas mAb 8H11 inhibited aiPLA 2 by 80% but had no effect on NS-GPx activity (Table II).
Identification of Protein Active Sites-Based on the results with DENP and the presence of a putative "lipase" motif (1), we mutated Ser 32 to Ala (S32A) in constructs 1 and 3 to study its role in aiPLA 2 activity. The S32A mutation abolished PLA 2 activity with both constructs 1 and 3 but did not affect NSGPx activity (Table III). The results obtained with construct 3 are illustrated graphically in Fig. 4 using bisbodipy-C 11 -PC liposomal substrate to assay aiPLA 2 and NADPH fluorescence to assay NSGPx. These results suggest that Ser 32 in the GDSWG motif is an active site residue for the phospholipase function of 1-Cys peroxiredoxin.
Since the peroxidase activity is inhibited by mercaptosuccinate (Table II), we mutated Cys 47 , the only conserved Cys in 1-Cys peroxiredoxin, to Ser (C47S). Previously, mutation of this Cys abolished peroxidase activity (5) although the results of this latter study are clouded by the inability of GSH to function as co-factor with the wild type protein. Mutagenesis of Cys 47 totally abolished NSGPx activity while PLA 2 activity in the mutant remained at the wild type level (Table III and Fig. 4). These results confirm that the enzyme has a Cys-active site at position 47 for NSGPx activity and provide additional evidence for two distinct active sites for the two enzymatic activities of this protein.

DISCUSSION
Although one gene-one protein-one function has been a paradigm of biochemistry, an increasing number of exceptions are being reported (11). The term "moonlighting proteins" has been used to designate proteins that have multiple functions (11). The present results with constructs 1 and 3 clearly show the presence of both glutathione peroxidase and phospholipase activities in the same recombinant protein. The activities of the recombinant human protein using construct 3 (5460 for NSGPx and 50 for aiPLA 2 in mol/min/mg protein from Table I) were similar to values for native protein isolated from bovine eye (NSGPx, 5070 nmol/min/mg of protein) (6) and bovine lung (aiPLA 2 , 65 nmol/min/mg of protein) (8). Activities with construct 1 in wheat germ-translated protein and for NSGPx in construct 2 expressed in E. coli were slightly less. aiPLA 2 activity with construct 2 was very low which may have been due to an altered serine hydrolase site, possibly caused by misfolding due to the His tag at the NH 2 terminus. An analogous effect might account for the lack of aiPLA 2 activity in the E. coli expressed protein of Kang et al. (5).
The above evidence indicates that this protein has two separate activities with apparently two distinct active sites, since only one activity was lost with construct 2. Results with the S32A and C47S mutants provide additional evidence for this observation which is further supported by the differential response of the two activities to the inhibitors mercaptosuccinate (cysteine active) and DENP (serine active). Finally, mAb 8B3 and mAb 8H11 each inhibit only one of the two activities.
Our S32A mutation results show that Ser 32 is critical for the hydrolase activity of the enzyme, compatible with the inhibi-  a ND, none detected; the lower limit of detection is 10 nmol/min/mg for NSGPx and 0.020 nmol/min/mg for aiPLA 2 .

1-Cys Peroxiredoxin: A Bifunctional Enzyme
tion of PLA 2 activity by DENP. DENP inhibits aiPLA 2 activity of purified rat and bovine lung enzymes as well (2,8). Serine is also the active site in the 85-kDa cytosolic Ca 2ϩ -independent phospholipase A 2 (12), the 44-kDa platelet activating factor hydrolase (13), and the 25-kDa lysophospholipase I (14). GD-SWG is a conserved sequence in 1-Cys peroxiredoxin that fits the consensus sequence (GXSXG) found in serine hydrolases of diverse substrate specificity, such as proteases and lipases (12)(13)(14)(15)(16). We conclude that the GDSWG sequence is part of the catalytic site for aiPLA 2. The locations of other residues that make up the putative catalytic triad, Asp-Ser-His (15), remain to be identified. The results also show that Ser 32 does not participate in H 2 O 2 peroxidase activity compatible with results obtained with the serine protease inhibitor (Table II). The peroxidase activity of the expressed protein is several orders of magnitude greater than the phospholipase activity ( Table I). The enzyme can reduce H 2 O 2 and short chain organic, fatty acid, and phospholipid hydroperoxides in the presence of GSH (4). Note that classical cytosolic GSHPx has no activity toward phospholipid hydroperoxides (17) although the selenium-dependent phospholipid hydroperoxide glutathione peroxidase (PHGPx) does have this activity (18). Despite the similarity of peroxidase function, 1-Cys peroxiredoxin has no significant amino acid homology with GSHPx or PHGPx. Unlike 1-Cys peroxiredoxin, GSHPx and PHGPx are both seleno enzymes that require dietary selenium for their synthesis and activity (19). Glutathione S-transferase is also a non-seleno enzyme that can "repair" oxidized fatty acids by thiol transfer and shows a low level of activity toward phospholipid hydroperoxides but is not a true peroxidase and has no activity toward H 2 O 2 (20).
Inhibition of GSHPx activity by mercaptosuccinate, confirming the importance of the cysteine residue, has been reported before for the purified protein from bovine eye (6) and for recombinant human protein (4). The conserved Cys 47 in 1-Cys peroxiredoxin occurs in a PVCTTE cassette that represents a consensus sequence for peroxidase activity and has been confirmed to be a critical site for removing H 2 O 2 in some enzymes (21-23) including 1-Cys peroxiredoxin (5). The rat and bovine sequences show a single Cys at position 47 (2,4). The human enzyme has a second Cys at position 91 but this is not conserved and its mutation has no effect on the peroxidase activity of the enzyme (5). Cys 91 3 Ala mutant human 1-Cys peroxire-doxin has been crystallized as a dimer (24). The x-ray crystal structure study demonstrated the location of the active site cysteine at the bottom of a narrow pocket and indicated that Cys 47 exists as Cys-SOH (cysteine-sulfenic acid) in the oxidized native 1-Cys peroxiredoxin (24). For wild type enzyme, the reaction mechanism could be partially due to formation of Cys-SOH from reaction between enzyme and H 2 O 2 (5). From the kinetic data, it is evident that the mutant enzyme, C47S, is catalytically disabled for peroxidase. Therefore, cysteine in 1-Cys peroxiredoxin performs the role reserved for selenocysteine in GSHPx and PHGPx.
Maximal peroxidase activity was shown to occur between pH 7 and 8, concordant with a cytosolic localization for 1-Cys peroxiredoxin. In contrast to the pH 7-8 optimum for GSHPx activity, aiPLA 2 activity is maximal at pH 4 and essentially non-existent at pH 6 and above. By subcellular fractionation, 1-Cys peroxiredoxin protein has been localized to cytosol and also to lysosomes and lung secretory organelles (8,25) where the pH is in the appropriate range for aiPLA 2 activity (26). We have provided evidence previously that an enzyme inhibited by MJ33, presumably aiPLA 2 , functions in the metabolism of phospholipids in lung surfactant (25,27). In addition, aiPLA 2 could function synergistically with GSHPx during oxidative stress. In the cytosolic compartment, the enzyme, through GSHPx activity, could directly reduce peroxidized plasma membrane phospholipids. For phospholipid hydroperoxides that might be transferred to the lysosomal compartment, this enzyme could release peroxidized fatty acids from the sn-2 position of phospholipids for their subsequent reduction in the cytoplasm. Thus, 1-Cys peroxiredoxin may function ubiquitously in the repair of oxidized (peroxidized) membranes and could be considered a general enzyme for antioxidant defense.