The Glutathione Synthetase of Schizosaccharomyces pombe Is Synthesized as a Homodimer but Retains Full Activity When Present as a Heterotetramer*

Glutathione synthetase was overexpressed as a histidine-tagged protein in Schizosaccharomyces pombe and purified by two-step affinity chromatography. The recovered enzyme occurred in two different forms: a homodimeric protein consisting of two identical 56-kDa subunits and a heterotetrameric protein composed of two 32-kDa and two 24-kDa subfragments. Both forms are encoded by the GSH2 gene. The 56-Da protein corresponds to the complete GSH2 open reading frame, while the subfragments are produced following the cleavage of this larger protein by a metalloprotease. A stable homodimer was obtained by site-directed mutagenesis to remove the protease cleavage site, and this showed normal activity. A structural model of the fission yeast glutathione synthetase was produced, based on the x-ray coordinates of the human enzyme. According to this model the interacting domains of the proteolytic subfragments are strongly entangled. The subfragments were therefore coexpressed as independent proteins. These subfragments assembled correctly to yield functional heterotetramers with equivalent activity to the wild type enzyme. Furthermore, a permuted version of the protein was created. This also showed normal levels of glutathione synthetase activity. These data provide novel insight into the mechanisms of protein folding and the structure and evolution of the glutathione synthetase family.

Glutathione (GSH, 1 ␥-Glu-Cys-Gly) is a biologically important thiol found in almost all procaryotic and eucaryotic cells. It has been assigned several cellular functions, including protection against oxidative damage, maintenance of a reducing cellular thiol-disulfide balance, electron donation for a number of enzymes, protection of protein sulfhydryls from irreversible oxidation, and detoxification of foreign compounds (1)(2)(3). Glutathione is synthesized enzymatically from its constituent amino acids in two consecutive reactions. Glutathione synthetase (EC 6.3.2.3) catalyzes the second step, the addition of glycine to ␥-glutamylcysteine.
In most eucaryotic organisms, including Saccharomyces cerevisiae, Arabidopsis thaliana, Homo sapiens, and Rattus norvegicus, glutathione synthetase is a 104 -112-kDa homodimer composed of two identical 52-56-kDa subunits (4 -10). In contrast, the structure of the Schizosaccharomyces pombe enzyme has not been completely elucidated. The enzyme was purified as a 120-kDa heterotetramer consisting of two kinds of subunits, called the "small" and "large" subunits, with apparent molecular masses of 26 and 33 kDa (11). While the gene encoding the 26-kDa subunit was unknown, the gene encoding the 33-kDa subunit, GSH2, had been cloned, and the coding region had been assigned to the 3Ј end (12). When the GSH2 gene was re-sequenced at a later date, however, frameshift errors in the original sequence were identified (13,14). These errors resulted in a shortened open reading frame, downstream from the correct start codon. A protein derived from the full-length GSH2 gene would have a molecular mass of 56 kDa, and BLAST searches showed that this protein is homologous to the glutathione synthetase proteins of other eucaryotic organisms. Wang and Oliver (13) proposed that the 56-kDa protein was the primary translation product, from which the subunits could be derived by proteolysis.
Here we report the homologous expression and purification of a histidine-tagged form of the fission yeast glutathione synthetase. We show that a 56-kDa protein corresponding to the whole open reading frame is indeed formed. Proteolytic cleavage then yields the small subunit, the 24-kDa N-terminal fragment, and the large subunit, the 32-kDa C-terminal fragment of this protein. Moreover, a model of the tertiary and quaternary structure of the fission yeast glutathione synthetase was generated, based on the structure of the human ortholog. On the basis of these results, in vitro mutagenesis experiments were performed to shed light on the subunit structure of the enzyme. The mutated recombinant proteins were expressed and affinity purified; in vivo as well as in vitro activities were determined. Our experiments provided novel data relating to folding and evolution of glutathione synthetase.

EXPERIMENTAL PROCEDURES
Materials-The chemicals used were of molecular biology grade or pro analysi grade.
DNA Methods-Standard molecular biology techniques were used for DNA isolation, analysis, and cloning (15,16). The identity of all clones was tested by DNA sequencing, which was carried out by SequiServe (Vaterstetten, Germany) following the method of Sanger.
Histidine Tagging of the Gene and Plasmid Construction- Table I shows the oligonucleotides (part A) and the amplified fragments and vectors used for cloning and expression of specific regions of the fission yeast GSH2 gene (part B); part C provides an overview over the resulting plasmids.
The oligonucleotide PNMT1F, which is specific for the nmt1 promoter, was used in combination with one of the GSH2 oligonucleotides to confirm the presence of the expression vectors in the fission yeast.
Bacterial Strain and Cultivation-Escherichia coli DH5␣ was used for all DNA cloning experiments. Bacteria were grown in LB medium or LB medium containing 125 g/ml ampicillin. Transformation and plasmid isolation were carried out according to standard methods (16).
Yeast Strains and Growth Conditions-The different forms of glutathione synthetase were expressed in a mutant strain of S. pombe (⌬GSH2), in which a central 479-bp region of the GSH2 gene (nucleotides 282-761) was deleted (17). The mutation was considered stable, because no reversions were observed. No functional glutathione synthetase was produced in this strain, so functional testing of the different proteins was possible.
Cells were grown at 28°C in Edinburgh minimal medium (18). Microbiological Methods-Standard techniques were applied for yeast transformation, crosses, the plasmid loss test, and other microbiological methods (18,19).
Physiological Tests-For heavy metal tolerance tests, the minimal inhibitory concentration of cadmium was determined using Edinburgh minimal medium containing different concentrations of CdCl 2 . GSH and phytochelatin concentrations were determined by RP-HPLC as described previously (20). Phytochelatin standards were purchased from AppliChem (Darmstadt, Germany).
Protein Expression and Purification-Cells were grown in 3 ϫ 500 ml minimal medium with gentle shaking at 28°C for 24 h, harvested by centrifugation (10 min, 1500 ϫ g, GSA rotor, 4°C), washed twice with distilled water, and lyophilized. Cell disruption was performed in a cell mill using glass beads (3 ϫ 3 min dry; 6 ϫ 3 min with 10 ml of immobilized metal ion affinity chromatography (IMAC) binding buffer/g of lyophilized cells). After centrifugation (30 min, 1500 ϫ g, SS34 rotor, 4°C) the supernatant was utilized for the following purification  GSH2-Fwd  5Ј-AAA GGA TCC ATG GAA ATT GAG AAG TAT ACA CCG GAG-3Ј  BamHI  GSH2-Rev  5Ј-AAA CCC GGG TTA ATG ATG ATG ATG ATG ATG TTC AGA AAG TTC AAT ACT AG-3Ј  SmaI  GSH2M214F  5Ј-AAA GGA TCC ATG AAT ATT GCT TCT GAT AAC-3Ј  BamHI  GSH2K213R  5Ј-AAA CCC GGG TTA TTT GCT TGT AAT GTT TTT AAC GTA ATC ACG-3Ј  SmaI  CWGLU44A  5Ј-AAA GAA TTC TCA AAT ACA TGT ATA ATT TT-3Ј  EcoRI  GSH2Irev  5Ј-AAA GGT ACC GTA GGC ATC TAC AGC ATT-3Ј  KpnI  GSH2IIFwd  5Ј-AAA GGT ACC GAT AAC ACG AAA CCC ATT-3Ј  KpnI  CW8DDA  5Ј-AAA GGA TCC AAT ATC TTC TTT GAG GCA C-3Ј  BamHI  PermutGr  5Ј-AAA GGT ACC TTC AGA AAG TTC AAT ACT AG- C-terminal pREP2-Sp-GSH2-perm pREP2 Permuted S. pombe glutathione synthetase consisting of amino acid residues 214-498 followed by a 5-amino acid linker (GTPSG) and the amino acid residues 1-213 C-terminal procedure, with all steps carried out at 4°C. For IMAC chelating Sepharose fast flow was loaded with nickel as indicated by the manufacturer (Amersham Biosciences). After applying the cell extract to the column and washing with IMAC binding buffer (20 mM Na 2 HPO 4 / NaH 2 PO 4 , 500 mM NaCl, 10 mM imidazole, pH 7.4), the bound proteins were eluted by increasing the imidazole concentration stepwise to 50 mM and then to 100 mM. Glutathione synthetase eluted at an imidazole concentration of 100 mM. For reactive dye affinity chromatography, the sample was adjusted to the appropriate buffer (10 mM Tris-HCl, 20 mM NaCl, pH 7.8) by Sephadex G-25 gel filtration chromatography and applied to the column packed with Cibacron blue 3GA-agarose. After a 10-ml wash, elution was carried out by increasing the NaCl concentration to 250 mM. Assay for Glutathione Synthetase-Purified glutathione synthetase was equilibrated to 100 mM Tris-HCl buffer, pH 8.2, by Sephadex G-25 gel filtration chromatography. Glutathione synthetase activity was assayed according to Huang et al. (10) using ␥-glutamylcysteine trifluoroacetate salt (Sigma) as a substrate. One unit of enzyme is defined as the amount that catalyzes the formation of 1 mol of product/min. Protein concentrations were determined according to Bradford (21) using bovine serum albumin as a standard.
Protease Inhibition Experiments-For protease inhibition experiments the strain 46a14h Ϫ pREP2-gsh2:hft expressing the histidinetagged wild type glutathione synthetase was used. Cell disruption was performed in the presence of different protease inhibitors at the following concentrations (aprotinin, 50 g/ml; benzamidine HCl, 10 mM; EDTA-Na 2 , 10 mM; E-64, 5 g/ml; leupeptin-hemisulfate, 5 g/ml; pepstatin A, 7 g/ml; PMSF, 0.5 mM). After cell disruption, the cell extracts were incubated at 4°C. Samples were taken immediately and after 1, 5, and 21 h of incubation. Histidine-tagged glutathione synthetase was detected by immunoassay after SDS-PAGE and Western blot.
SDS-PAGE, Western Blot, and Immunodetection of the Histidine Tag-SDS-PAGE was performed following the method of Laemmli (22) using the M12 molecular mass standard (Novex, San Diego, CA). For Western blotting the proteins were electrotransferred, after SDS-PAGE, to a PVDF membrane in CAPS buffer (10 mM CAPS, pH 11.0, 15% (v/v) methanol) at 870 mA for 2 h. The prestained marker 7B (Sigma) served as a molecular mass standard. Proteins containing a histidine tag were localized on the PVDF membrane by applying an anti-His 6 monoclonal mouse antibody coupled to horseradish peroxidase and BM Blue POD substrate (Roche Applied Science, Mannheim, Germany) according to the protocol of the manufacturer.
Native PAGE-For native PAGE, precast gels with a 4 -20% acrylamide gradient (Bio-Rad, Munich, Germany) were used. Native PAGE was performed as described above for SDS-PAGE except SDS was omitted from all buffers. Glycerol, at 25% (v/v), was added to the protein samples before application to the gel. Bovine serum albumin was used as a molecular mass marker.
Gel Filtration Chromatography-Gel filtration chromatography was carried out using a Superdex 200 HR 10/30 column connected to a fast performance liquid chromatography system (Amersham Biosciences). After equilibrating the column with 15 mM Tris-HCl, pH 6.8, 20 mM NaCl, the protein sample was applied. The flow rate was 0.5 ml/min. Proteins were detected by measuring the absorbance at 280 nm. Alcohol dehydrogenase from S. cerevisiae (141 kDa) (Sigma) was used for column calibration.
N-terminal Sequencing of the Proteins-After SDS-PAGE, Western blot, and staining with Amido Black, protein bands with molecular masses of 56, 32, and 24 kDa were cut from the membrane and sequenced by the automated Edman degradation method. Sequencing was carried out by the Institute of Enzyme Technology, Forschungszentrum Jü lich (Germany).
Protein and Peptide Analysis by MALDI-TOF-MS-For the determination of exact molecular masses a purified sample of glutathione synthetase was, after partial cleavage of the 56-kDa protein, dialyzed against water and lyophilized. The residue was dissolved in 0.5% (v/v) trifluoroacetic acid and deposited on a MALDI probe. After evaporation of the solvent the matrix ␣-cyano-4-hydroxycinnamic acid (20 mg/ml in a solution of 0.1% (v/v) trifluoroacetic acid and acetonitrile (2:1 v/v) was added. For peptide analysis, the method described by Jeno et al. (23) was followed. Analysis and sequencing by MALDI-TOF-MS was performed at the German Wool Research Institute, Aachen, Germany. A BRUKER BIFLEX TM III MALDI-TOF-MS equipped with a nitrogen laser was used. Pulses of 3 ns at 337 nm were used at a voltage of 20 kV. About 100 shots were averaged for each acquired spectrum. External calibration was performed with a protein standard (angiotensin II, substance P, ACTH fragment 1-17, ACTH fragment 18 -39; Sigma).

Molecular Modeling of the Fission Yeast Glutathione Synthetase-
The alignment of the amino acid sequences from S. pombe, S. cerevisiae, Pichia angusta, A. thaliana, and H. sapiens was performed with the GCG package (Accelrys Inc., San Diego, CA). As the crystal structure of the fission yeast glutathione synthetase has not yet been determined, homology modeling was used to create a structural model. The crystal structure of the human glutathione synthetase (24) was used as the template. The modeling was performed using the Modeler program under the homology module of Insight II. To simplify the procedure, the full-length 56-kDa form of the S. pombe monomer was analyzed. The template structure of the human protein 2 was adopted for continuously aligned sequence stretches between the human and the S. pombe proteins. Sequence insertions and deletions were only allowed outside secondary structure elements. Insertions larger than three amino acids were modeled in an extended conformation.

Cloning and Expression of the Fission Yeast Glutathione
Synthetase-An 18-bp sequence encoding six histidine residues was attached to the C terminus of glutathione synthetase by PCR. The protein was expressed in a ⌬GSH2 fission yeast strain under control of the nmt1 promoter using the expression vector pREP2 (25). The presence of the plasmid was confirmed by PCR. Plasmid loss tests were performed as a control.
Purification of the Fission Yeast Glutathione Synthetase and Subunit Structure of the Enzyme-The histidine-tagged glutathione synthetase was purified to apparent homogeneity in a two-step affinity chromatography procedure. In the first step (IMAC), glutathione synthetase coeluted with other contami- 2 The atomic coordinates for the crystal structure of this protein are available in the Research Collaboratory for Structural Bioinformatics Protein Data Bank under accession code PDB 2HGS (www. rcsb.org/pdb). nating proteins. These were removed in the second step by reactive dye affinity chromatography (Fig. 1).
The purified enzyme produced a single band in a native PAGE gel (Fig. 2). The gel did not allow the accurate determination of the molecular mass. In gel filtration chromatography one protein peak was observed. The observed elution volume was slightly lower than that of S. cerevisiae alcohol dehydrogenase (141 kDa), the protein used for calibration (data not shown). These results are consistent with the value of 120 kDa reported for the molecular mass of the fission yeast glutathione synthetase by Nakagawa et al. (11). In SDS-PAGE a strong band of about 56 kDa and two faint bands of about 32 and 24 kDa were visible immediately after purification (Fig. 1). The 56-and 32-kDa proteins contained the C-terminal His 6 -tag, which was identified by Western blot and immunodetection using a monoclonal anti-His 6 antibody (data not shown). Upon prolonged storage of the protein (10 mM Tris-HCl, pH 7.8, 20 mM NaCl, 4°C); however, the 56-kDa band faded, while the other two bands became more intense. The 56-kDa protein was completely cleaved to the 32-and 24-kDa proteins when partially purified by IMAC and kept at 4°C for 4 weeks (Fig. 1). However, the 56-kDa protein was still present in fractions of the protein purified by IMAC and reactive dye affinity chromatography in which no contaminating proteins were detectable by SDS-PAGE after storage for more than 6 months under the conditions given above (data not shown).
The N-terminal sequence of the 32-kDa protein was identified as "SDNTKPIVLF" by Edman degradation. This truncated protein starts with the serine residue at position 218 and not with the methionine residue at position 214, as previously assumed (12). The 56-kDa protein and the 24-kDa protein were N-terminally blocked.
Tryptic Digestion and Peptide Analysis of the 24-kDa Protein-Because the N terminus of the 24-kDa protein was blocked, this protein was further investigated by analysis of the fragments obtained after tryptic digestion by MALDI-TOF-MS. Molecular masses between 400.0 and 2900.0 Da could be detected by the parameter file employed for MALDI-TOF-MS. Nineteen fragments theoretically arising from tryptic digestion of the N-terminal region (residues 1-218) of the 56-kDa protein are within this molecular mass range. Eleven of these fragments were recorded (Table II).
The 1140.5-Da fragment (residues 76 -84) was sequenced and the expected sequence "IANDYEFLR" was found. Thus, MALDI-TOF-MS analysis of the tryptic peptides derived from the 24-kDa subfragment provided further evidence that this protein was derived from the 5Ј region of the GSH2 gene. (Fig.  3) shows a schematic diagram of the regions of the GSH2 gene.
Amino Acid Sequence of the Fission Yeast Glutathione Syn-thetase in Comparison with Other Eucaryotic Organisms and Molecular Modeling of the Structure of the Fission Yeast Enzyme-The sequences of the fission yeast glutathione synthetase and homologous enzymes from other eukaryotes show evidence for homology over the whole alignment (Fig. 4). S. pombe glutathione synthetase shows 57, 57, 53, and 50% similarity, and 36, 37, 33, and 35% identity to the homologous enzymes from S. cerevisiae, P. angusta, A. thaliana, and H. sapiens, respectively. There is no detectable sequence similarity between the bacterial and eucaryotic enzymes. The most striking difference among the eucaryotic enzymes is a 15-amino acid segment (residues 204 -218) between helix ␣ 8 and strand ␤ 6, which occurs only in the S. pombe enzyme, although a slightly smaller insert of nine amino acids occurs in the yeasts P. angusta and S. cerevisiae (Fig. 4).
Since the crystal structure of the human enzyme has been solved (24), the structure of the S. pombe enzyme can be modeled according to these data. Although the crystal structure of glutathione synthetase from E. coli has also been determined (26), the data collected for this enzyme were not taken into consideration due to the lack of sequence homology. Fig. 5 shows a model of the tertiary structure of one 56-kDa subunit of the fission yeast enzyme. The structure was determined assuming that the fission yeast enzyme is a homodimer composed of two identical 56-kDa subunits, each subunit derived from the whole GSH2 open reading frame. The calculations were based on the x-ray coordinates of the human enzyme and the sequence alignment between the proteins. The regions cor-   With the exception of the allocations discussed above, the modeled structure of the fission yeast glutathione synthetase is similar to the crystallographic structure of the human enzyme. The amino acids essential for glutathione, cofactor, and Mg 2ϩ binding in the human enzyme have been determined (24). In the modeled structure of the fission yeast enzyme a similar, or in most cases identical, counterpart of all these amino acids can be found (Fig. 4).
The cleavage point between the large and the small subfragment of the S. pombe glutathione synthetase lies within the 15 amino acids that are unique to the S. pombe enzyme. To find out whether these 15 amino acids are within a defined domain of the enzyme or at the border between two domains, the x-ray structure of the H. sapiens glutathione synthetase (24) was superimposed on the structure of the E. coli protein (27) (Fig.  6). Although no significant sequence homology is found between the two enzymes, their three-dimensional structures show clear similarities, as already mentioned by Polekhina et al. (24). Both enzymes are members of the ATP-grasp superfamily. The 15-amino acid insert of the S. pombe enzyme (colored in yellow in Fig. 6) is located in between the C-and N-terminal domains of the representative fold of the ATP-grasp superfamily. The N and C terminus of the E. coli enzyme are in close vicinity to this region. The original termini of the ATPgrasp superfamily were fused in the eucaryotic glutathione synthetases after a gene permutation. These findings prompted further research into the role of the 15 amino acid residues.
Deletion of 15 Amino Acids within the Protein-To evaluate the functional role of the 15-amino acid residue insertion, these residues (204 -218; RDYVKNITSKMNIAS) ( Figs. 3 and 4) were deleted in the fission yeast enzyme. This was achieved by amplifying the upstream and downstream regions of the GSH2 gene and adding a His 6 -tag sequence to the 3Ј end of the downstream fragment. The two fragments were combined and subcloned in the expression vector pREP2. Two additional amino acid residues (G and T) were inserted due to the incorporation of the KpnI restriction site used for cloning. The mutated protein was expressed in a ⌬GSH2 fission yeast strain, isolated, and purified to apparent homogeneity by IMAC and reactive dye affinity chromatography. In SDS-PAGE (Fig. 1) and Western blot experiments, only the 56-kDa protein was detected. In native PAGE experiments, one band approximately equivalent in size to the wild type enzyme was identified ( Fig. 2), indicating that this enzyme is a homodimer composed of two 56-kDa subunits. The 56-kDa protein was very stable as no cleavage was observed upon storage for more than six months in a buffer containing 10 mM Tris-HCl, pH 7.8, 20 mM NaCl at 4°C.

Molecular Masses of the Subunits and Localization of the Protease Cleavage Site-
The increased stability of the protein following deletion of amino acid residues 204 -218 suggested that this region contains a protease cleavage site. N-terminal sequencing revealed that the 32-kDa protein starts with the serine residue at position 218 (Fig. 3). MALDI-TOF-MS confirmed the molecular masses of the 56-, 32-, and 24-kDa proteins to be 56.981 kDa (His 6 -tagged), 32.630 kDa (His 6 -tagged), and 24.352 kDa, respectively (Table III). The difference between the calculated masses and the experimentally determined masses is consistent with the accuracy of MALDI-TOF mass spectrometry of proteins in this size range. Taking this into consideration, the MALDI-TOF data support the protein sequencing data, showing that the cleavage takes place between the amino acid residues Ala and Ser at positions 217 and 218. structure elements are marked as follows: ffff, ␣-helix; ‹ ‹ ‹ ‹, ␤-sheet. Amino acids forming the glutathione, ATP, and magnesium binding sites are colored yellow, orange, and blue, respectively. Amino acid residues identical in all the proteins are shown as a consensus. The numbering of the amino acid residues refers to the fission yeast protein.  Identification of the Protease That Cleaves the 56-kDa Protein-Cell disruption of strain 46a14h Ϫ pREP2-gsh2:hft was performed in the presence of different protease inhibitors, and glutathione synthetase was detected by SDS-PAGE, Western blot, and immunoassay for the histidine tag. Strain D18h Ϫ served as a negative control. This strain does not contain any proteins detectable by immunoassay using the anti-His 6 antibody, and no signal was observed. In all the samples of strain 46a14h Ϫ pREP2-gsh2:hft, the intensity of the 56-kDa band decreased, while the intensity of the 32-kDa band increased in a time-dependent manner (Fig. 7). The 24-kDa band could not be detected using the anti-His 6 antibody, because the tag was attached to the C terminus of the 56-kDa protein.
The addition of EDTA-Na 2 strongly inhibited the cleavage of the 56-kDa protein (Fig. 7). The other protease inhibitors tested (aprotinin, benzamidine HCl, E-64, leupeptin-hemisulfate, pepstatin A, and PMSF) had no inhibitory effect on the cleavage of the 56-kDa protein at that position (data not shown). Therefore, we suggest that a metalloprotease is responsible for cleaving the 56-kDa protein yielding the 32-kDa and 24-kDa subfragments.
In addition to the 56-and 32-kDa bands, one more band of about 25 kDa was detected (Fig. 7). This represents a C-terminal fragment of glutathione synthetase, since the histidine tag is located at the C terminus of the protein. This additional product might arise through the digestion of glutathione synthetase by another protease present in the cell extract. The cleavage was inhibited by benzamidine HCl and PMSF, which suggests that a serine protease might be involved (data not shown). The protease is effectively separated from glutathione synthetase by the purification procedure as the additional degradation product is not observed when glutathione synthetase is purified by IMAC immediately after cell disruption.
Independent Expression of the 24-and 32-kDa Enzyme Subfragments-The coding regions corresponding to the 24-and 32-kDa subfragments of the fission yeast glutathione synthetase were cloned and expressed separately. Fragments of the subunits were amplified, and a His 6 -tag encoding sequence was attached to the N terminus of the 24-kDa subfragment (residues 1-213) and to the C terminus of the large subfragment (residues 214 -498) by PCR. The two proteins were coexpressed in the ⌬GSH2 fission yeast strain 46leu42 using the vectors shown in (Fig. 8). The protein was isolated and purified as stated above for the other forms of the enzyme. In SDS-PAGE experiments, the 24-and 32-kDa proteins were visible (Fig. 1); both proteins produced signals in Western blot and immunodetection assays. Native PAGE showed one band equivalent in size to the wild type enzyme (Fig. 2). Therefore, the enzyme produced by coexpressing the small and large subfragments must be a heterotetramer of two 24-kDa and two 32-kDa proteins.
Permutation of S. pombe Glutathione Synthetase-According to the structural model generated for the S. pombe glutathione synthetase, the N and C termini of the protein are close to each other (Fig. 5). To analyze the calculated structure experimentally, these termini were linked. The protein was then cleaved at a different position, on the N-terminal side of amino acid residue 214, between the 24-and 32-kDa subfragments, to generate new termini. As a result the fission yeast glutathione synthetase was artificially permuted by interchanging the positions of the subfragments within the protein. Five additional amino acid residues (GTPSG) were introduced between the two subfragments due to the oligonucleotides used for amplification and subcloning. The linker was introduced to prevent the formation of additional secondary structural elements and to ensure flexibility and solubility. A His 6 -tag was added to the new C terminus of the protein, i.e. the C terminus of the 24-kDa subfragment, by PCR. (Fig. 9) provides a schematic view of the permuted gene. The permuted protein was expressed in a ⌬GSH2 fission yeast strain. The purified recombinant protein was revealed as a single 56-kDa band in SDS-PAGE and following immunodetection of the histidine tag (data not shown).
In Vivo Activities-S. pombe cannot grow without glutathione (27). To investigate whether the three different forms of glutathione synthetase are active in vivo, the proteins were expressed in a fission yeast ⌬GSH2 strain. All the forms of the protein, i.e. the wild type glutathione synthetase, the recombinant protein with the 15-amino acid deletion, the protein comprising separately encoded and coexpressed 24-and 32-kDa subfragments, and the permuted protein, restored growth of the ⌬GSH2 strain on minimal medium without glutathione (Fig. 10).
As demonstrated by RP-HPLC measurements, all strains produced glutathione and, upon induction with cadmium, phytochelatins. The glutathione contents of all strains were similar to that of the wild type (Table IV).  Moreover, all recombinant yeast strains expressing the different glutathione synthetase mutants showed heavy metal tolerances similar to the wild type. They all grew on minimal medium containing up to 750 M CdCl 2 . On the contrary, growth of the ⌬GSH2 strain on minimal medium supplemented with 10 mg/liter GSH was inhibited when only 10 M CdCl 2 was added.
In Vitro Glutathione Synthetase Activities-The wild type form and the three different recombinant forms of the fission yeast glutathione synthetase were purified and their specific activities determined. All proteins exhibited similar activities of about 17 to 22 units/mg of protein (Table V).
The presence of six extra histidines at the C terminus did not affect the catalytic rate, as the specific activity of 20.1 units/mg observed for the wild type enzyme purified in this manner was higher than the value of 14.1 units/mg given for the native enzyme (without a histidine tag) in the literature (11). The lower activity reported for the native enzyme is probably due to the extended purification protocol. DISCUSSION The fission yeast glutathione synthetase is, like the homologous enzymes of other eucaryotic organisms, a homodimer composed of two identical subunits encoded by the GSH2 gene. The subunit structure of two small and two large subunits observed by Nakagawa et al. (11) results from proteolytic cleavage of the 56-kDa protein, encoded by the complete GSH2 open reading frame. Proteolysis probably takes place after cell disruption. The cleavage is catalyzed by a S. pombe protease, which has not been characterized thus far. It is highly unlikely that the cleavage is an autocatalytic process, because the rate of cleavage is inversely proportional to the purity of the protein.
The resulting fragments of 24 and 32 kDa are stable; further degradation is not observed. The protease responsible for the cleavage was assigned to the metalloprotease family by protease inhibition experiments. Residual activities have been reported for several metalloproteases after EDTA treatment (28 -30). The metal ions cannot always be effectively removed because the catalytic sites of metalloenzymes can bind metal ions with a very high affinity. A very small amount of metalloprotease might be copurified with glutathione synthetase by Ni 2ϩ -IMAC because of the affinity of metalloproteases to metal ions, particularly since IMAC has been described as a method to purify metalloenzymes (31).
The cleavage site of the metalloprotease that generates the 24-and 32-kDa subfragments is localized on the N-terminal side of serine 218, as determined by N-terminal sequencing and MALDI-TOF mass spectrometry. Site-directed mutagenesis, to delete amino acid residues 204 -218 situated around the cleavage site, yields a stable homodimer composed of two 56-kDa subunits. This region is not essential, since the recombinant enzyme shows full in vivo and in vitro activity.
Glutathione synthetase still functions in vitro upon complete cleavage of the 56-kDa protein yielding the 24-and 32-kDa subfragments. Therefore an enzyme composed of the subfragments encoded on different plasmids was expressed in a ⌬GSH2 strain. The activity of this enzyme is very surprising, since the model of the structure of the fission yeast glutathione synthetase (Fig. 5) indicates that the domains of the small and large subunit form a complex quaternary structure. Instead of a classical double domain architecture, the two subunits are strongly entangled. This has several interesting implications when the two major concepts of protein folding are considered. According to the cotranslational theory, proteins fold while the polypeptide chain is synthesized at the ribosome. The posttranslational theory suggests that proteins start to fold after the polypeptide chain has emerged from the ribosome and that folding requires the assistance of chaperones (32). For the heterotetramer to function, the subfragments must interact and assemble in the correct manner even though they are synthesized at discrete sites on separate ribosomes. This cannot be explained by the separate folding of polypeptide chains and subsequent joining of the proteins. On the contrary, according to the structure modeled for the fission yeast enzyme, the subfragments must contact each other while they are in an at least partially unfolded state. The nascent polypeptide chains of the subfragments might be stabilized by ribosomeassociated chaperones and transferred to further downstream chaperones where they interact and the folding of the fulllength protein takes place. The assistance of molecular chaperones in protein folding is a commonly accepted principle occurring in procaryotic as well as eucaryotic cells (for review see Refs. [33][34][35]. Moreover, a chaperone-mediated transport of  (6)) are able to grow, while the ⌬GSH2 strains 46a14h Ϫ pREP2 (2) and 46leu42 pREP2 ϩ pREP1 (7) are unable to grow. polypeptide chains between different chaperones (36) and a processivity of chaperone action have been described.
One should consider, however, that the structure of the fission yeast glutathione synthetase presented here is a model relying on the x-ray structure of the human enzyme. Although the two structures are very similar, there is no final proof that the fission yeast enzyme in fact folds in this way. However, major differences in the structures of the two enzymes are very unlikely. Essential and highly conserved residues of the catalytic center (24) are located in both subfragments of the S. pombe enzyme. A structure in which the subfragments were folded significantly differently would not allow to bring the essential residues into the correct position. The spatial arrangement of the domains in the ATP-grasp superfamily is highly conserved (24).
It was not possible to express individual subfragments of the enzyme. Neither the small nor the large subunit could be isolated from or detected in the cell extract even though the presence of the expression vector in the cell was demonstrated. Different fission yeast strains and expression conditions were tested without success. Wang and Oliver (13) reported that mutagenized forms of the A. thaliana glutathione synthetase expressed in the fission yeast were extremely unstable. Either the cell does not synthesize the protein or degrades it after synthesis. As a transfer of non-functional or misfolded proteins from chaperones to the ubiquitin-proteasome machinery has been observed (37,38), the fission yeast probably synthesizes each of the subfragments, but since there is no second subfragment present as an interaction partner to form the native enzyme, the single subfragment is transferred to the protea-some and degraded. This indicates that the cell can distinguish between functional and non-functional proteins.
Another important issue is the physiological relevance of the cleavage reaction. Although cleavage of glutathione synthetase appears to occur only during cell disruption, the specificity of the cleavage and the stability of the cleavage products suggests that the reaction may fulfill a specific functional or regulatory role. To resolve this issue further studies will be necessary.
The protein composed of the separately encoded subunits as well as the permuted version of the protein illustrate the evolution of glutathione synthetase. While there is no detectable sequence similarity between the bacterial and eucaryotic enzymes, structure-based alignments between the E. coli and eucaryotic glutathione synthetases reveal common conserved structural motifs. This is due to a gene permutation, which led to a circular shift of the conserved secondary structure elements in the eucaryotic protein (24). Because of its crystal structure the human enzyme was assigned to the ATP-grasp superfamily, which is characterized by three typical domains, each centered around a four-to six-stranded ␤-sheet. The ␤-sheet of the central and the C-terminal domain form an ATP binding site. Two flexible loops allow entry of the substrates and ATP and also protect the reaction intermediates during catalysis (39).
Permutation events play an important role in molecular evolution (40). Circular permutation of a protein can occur by gene duplication, in-frame fusion and partial deletion or mutation of the resulting tandem protein. Point mutations can produce new start and stop codons at the appropriate positions and the outer regions are deleted. As a result, N-and Cterminal parts of the protein become exchanged (41,42). Naturally occurring permutations have been observed in many protein families (for review, see Ref. 43). There are also numerous examples of proteins that have been artificially permuted (44 -47). Prerequisites that are necessary for protein permutation in most cases include close proximity between the N and C termini in the tertiary structure of the protein and surface loops that can be cleaved to generate new termini (40,48). These conditions should apply to glutathione synthetase (Figs. 5 and 6). The bacterial protein probably represents the original form of glutathione synthetase. The permutation event occurred in the early eucaryotic lineage (24), after the separation of the eucaryotic branch for this enzyme. According to the structure-based alignment and the overlay of the human and E. coli enzymes (Fig. 6), the original N-and C-terminal domains of the ATP-grasp superfamily, and as the prototype of this family, the E. coli glutathione synthetase, are situated in the middle region of the eucaryotic protein, between helix ␣ 8 and strand ␤ 6. This region also contains the 15-amino acid insertion unique to the fission yeast enzyme (residues 204 -218). The N-and C-terminal domains of the ATP-grasp superfamily were linked in the eucaryotic enzymes by gene duplication. It is likely that the two genes were fused in-frame, but not directly connected, i.e. there was some distance between them. Over evolutionary time, the tandem gene accumulated further mutations, and the external and internal parts were probably deleted. Amino acid alignments of the different eucaryotic glutathione synthetases (Fig. 4) demonstrate that, in addition to the S. pombe insert, a much shorter insertion is present in the enzymes of the yeasts S. cerevisiae and P. angusta, but not in the enzymes of A. thaliana and H. sapiens. The presence of an exon sequence in the human glutathione synthetase gene, C-terminal to the position of the 15-amino acid residue insertion in the S. pombe enzyme, might indicate the mechanism by which this insertion was eliminated during the evolution of metazoans. As shown in this study, the additional region of the fission yeast enzyme is not essential, so there was no selection pressure to maintain it. The large size of the insert in S. pombe, together with the other structural variations near this site, might explain the proteolytic degradation of glutathione synthetase specifically in S. pombe.
The permuted fission yeast enzyme is functional. This supports that the permutation event described by Polekhina et al. (24) has taken place during the evolution of the protein and illustrates that the model calculated for the fission yeast protein (Fig. 5) is a good approximation of the natural structure.
Our findings show that the fission yeast glutathione synthetase can serve as a model for protein folding. Experiments on the heterotetrameric enzyme consisting of the separately encoded subfragments might provide further insight into the mechanisms of protein folding within the cell. Moreover, calorimetric studies of the different versions of the recombinant protein could provide valuable data on protein stability.