Dissection of glutathionylspermidine synthetase/amidase from Escherichia coli into autonomously folding and functional synthetase and amidase domains.

The bifunctional glutathionylspermidine synthetase/amidase from Escherichia coli catalyzes both the ATP-dependent formation of an amide bond between N1 of spermidine (N-(3-amino)propyl-1,4-diaminobutane) and the glycine carboxylate of glutathione (γ-Glu-Cys-Gly) and the opposing hydrolysis of this amide bond (Bollinger, J. M., Jr., Kwon, D. S., Huisman, G. W., Kolter, R., and Walsh, C. T. (1995) J. Biol. Chem. 270, 14031-14041). In our previous work describing its initial characterization, we proposed that the 619-amino acid (70 kDa) protein might possess separate amidase (N-terminal) and synthetase (C-terminal) domains. In the present study, we have confirmed this hypothesis by expression of independently folding and functional amidase and synthetase modules. A fragment containing the C-terminal 431 amino acids (50 kDa) has synthetase activity only, with steady-state kinetic parameters similar to the full-length protein. A fragment containing the N-terminal 225 amino acids (25 kDa) has amidase activity only and is significantly activated relative to the full-length protein for hydrolysis of glutathionylspermidine analogs. This observation suggests that the amidase activity in the full-length protein is negatively autoregulated. The amidase active site catalyzes hydrolysis of amide and ester derivatives of glutathione (e.g. glutathione ethyl ester and glutathione amide) but lacks activity toward acetylspermidine (N1 and N8) and acetylspermine (N1), indicating that glutathione provides the primary recognition determinants for glutathionylspermidine amide bond cleavage. No metal ion is required for the amidase activity. A tetrahedral phosphonate analogue of glutathionylspermidine, designed as a mimic of the proposed tetrahedral intermediate for either reaction, inhibits the synthetase activity (Ki ∼ 10 μM) but does not inhibit the amidase activity.

The polyamine, spermidine (N-(3-amino)propyl-1,4-diaminobutane), and the tripeptide, glutathione (␥-Glu-Cys-Gly, ab-breviated GSH), are present at high concentrations (0.1-10 mM) in most cells (see, for example, Refs. 1 and 2 for reviews). With its redox active cysteine thiol, GSH is the primary smallmolecule antioxidant in many cells, serving to maintain redox poise and reductively scavenge reactive oxygen species. It is catalytic in these roles by virtue of glutathione reductase, which maintains GSH in active, reduced form. As a polycation, spermidine complexes with nucleic acids, proteins, and phospholipids, thereby influencing their structures and biological properties (3).
An intriguing link between GSH and spermidine metabolism is found in the protozoal parasites of genera Trypanosoma and Leishmania, including those that cause African sleeping sickness, South American Chagas' disease, and various afflictions known collectively as leishmaniases. In these parasites, it appears that the bis(glutathionyl)spermidine conjugate, trypanothione (see Scheme 1 for structure), has appropriated the important antioxidant roles normally played by GSH (4 -10). Thus, these parasites lack GSH reductase and GSH peroxidase activities, but have analogous enzymes that use trypanothione (5, 8 -10). Because trypanothione appears to have important roles in these pathogens and is not present in their hosts, its metabolism is an obvious target for design of new antiparasitic drugs.
The synthesis of trypanothione from glutathione and spermidine is catalyzed by glutathionylspermidine (GSP) 1 synthetase and trypanothione synthetase (11,12). Each couples hydrolysis of ATP (to ADP and P i ) with formation of an amide bond (Scheme 1). The intermediate in this pathway, glutathionylspermidine, was first identified in Escherichia coli more than 3 decades before the discovery of trypanothione itself, and a GSP synthetase activity was partially purified (13,14). In spite of its early discovery, the physiological role in E. coli of the glutathione-spermidine conjugate is not yet known.
As part of an ongoing investigation of the enzymology and physiology associated with these glutathione-spermidine conjugates, we recently characterized GSP synthetase from E. coli (15). We purified the enzyme, isolated and sequenced its gene, overproduced it, and characterized the recombinant protein.
We were surprised to discover that the 70-kDa protein possesses a second catalytic activity: hydrolysis of glutathionylspermidine back to glutathione and spermidine. As the net of these two activities is hydrolysis of ATP (i.e. futile cycling), we proposed that reciprocal regulation of its activities might be an important feature of the bifunctional enzyme. In addition, because the synthetase activity was selectively abrogated by proteolytic cleavage after Arg-538 of the 619-amino acid protein, we proposed that it might possess separate domains for its two activities (N-terminal amidase and C-terminal synthetase) (15).
In the present study, we have confirmed the hypothesis of separate domains by genetic dissection of the bifunctional protein into independently folding and functional amidase and synthetase fragments, and have characterized the fragments with respect to their steady-state kinetic constants. We have evaluated, as an inhibitor of the synthetase activity, a phosphonate analog of glutathionylspermidine designed to mimic the proposed tetrahedral intermediate. Finally, we have begun to address the question of substrate specificity and mechanism of the amidase reaction by 1) evaluating analogs of GSH, spermidine, and glutathionylspermidine as possible substrates and/or inhibitors of the reaction and 2) testing for the presence of a catalytic metal ion.

Preparation of GSP Synthetase/Amidase and Fragments
Materials-Oligonucleotides were purchased from the Harvard Medical School Biological Chemistry and Molecular Pharmacology departmental biopolymer facility or from Integrated DNA Technologies (Coralville, IA). Restriction endonucleases and T4 DNA ligase were purchased from New England Biolabs (Beverly, MA). The vector pET22b and E. coli strain BL21(DE3) were purchased from Novagen (Madison, WI). Construction of the vectors pJMB1 and pGSP has been described previously (15).
Construction of pAMID Vector to Overexpress 25-kDa GSP Amidase Fragment-We had previously (15) subjected the plasmid pJMB1, which contains a 5.8-kilobase pair insert spanning gsp in the vector pBluescript (Stratagene), to the transposon mutagenesis procedure of Berg et al. (16). A mutant plasmid from the resulting set, with the transposon inserted after nucleotide 675, was used as template to amplify a 5Ј fragment of gsp (containing the amidase domain) using the polymerase chain reaction (PCR). The primers were 5Ј-AAGGTAAA-CATATGAGCAAAGGAACGACCAG-3Ј (15), which primes for synthesis of the sense strand beginning just 5Ј of gsp and the "Res" sequencing primer of Berg et al. (16), which primes in the transposon for synthesis of the antisense strand of gsp. The resulting 751-base pair PCR frag-ment was digested with NdeI (which cuts in primer 1) and BamHI (which cuts in the transposon just 3Ј of gsp) and ligated into vector pET22b to give pAMID, which encodes an N-terminal fragment of GSP synthetase/amidase consisting of amino acids 1-225 fused to the transposon-encoded dipeptide GV. Sequencing of the plasmid by the Dana Farber Cancer Institute Core Facility confirmed that no mutations were introduced during PCR amplification.
Construction of pSYN Vector to Overexpress 50-kDa GSP Synthetase Fragment-The vector pSYN was constructed to overproduce a C-terminal fragment of GSP synthetase/amidase comprising amino acids 189 -619. PCR was used to amplify a 1362-base pair, 3Ј fragment of the gsp gene. The template for amplification was the vector pGSP (15), which contains gsp inserted in the plasmid pET22-b (Novagen) via NdeI (5Ј) and EcoRI (3Ј) restriction sites. Primer 1 (5Ј-ACCATTCTGGGC-CATATGATCCAGACGGAAGAT-3Ј) corresponds to nucleotides 550 -582 of gsp but introduces an NdeI site (CATATG) for cloning. Primer 2 (5Ј-ATATTGAATTCTTTGATTAATCCCCGTACTGATTATTC-3Ј) primes immediately 3Ј of gsp for synthesis of the antisense strand. The amplified fragment was digested with NdeI, which cuts both in primer 1 and at an internal site in gsp after nucleotide 932. The resulting 367-base pair fragment (nucleotides 565-932) was ligated with NdeIdigested pGSP (two sites resulting in excision of nucleotides 1-931 of gsp). Transformants of E. coli strain DH5␣ containing pSYN (with the NdeI insert in the desired orientation) were identified by restriction analysis of plasmid DNA. Sequencing of the plasmid by the Dana Farber Cancer Institute Core Facility confirmed that no mutations were introduced during PCR amplification.
Growth of Overexpressing Strains-E. coli strain BL21(DE3) transformed with pGSP, pAMID, or pSYN was grown aerobically in LB medium containing 150 g/ml ampicillin at 37°C to an A 600 of 0.6 -0.8, when protein expression was induced by addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 500 M for full-length protein and amidase fragment or 100 M for synthetase fragment. Cultures were incubated an additional 2-4 h and were harvested by centrifugation. A typical yield was 2.5 g of wet cell paste/liter of culture.
In a typical purification of the amidase fragment, cells from 2 liters of culture were resuspended and lysed as described previously (15). Treatment with streptomycin sulfate was as described (15). Ammonium sulfate fractionation steps were as described (15), but the concentrations were 30% of saturation for the first cut and 70% for the second cut. The pellet from the 70% ammonium sulfate step was redissolved in 50 mM Tris-HCl, pH 7.5, 5 mM DTT, 1 mM EDTA (buffer A) and dialyzed against buffer A. The desalted solution was loaded on a DEAE-Sepharose (Pharmacia Biotech Inc.) column (2.5 ϫ 25 cm) equilibrated in buffer A. The column was washed with 100 ml of buffer A, and then SCHEME 1 developed with a gradient of NaCl in buffer A (50 ml of 0 -100 mM, then 450 ml of 100 -400 mM). GSP amidase activity of column fractions was determined by a previously described (15) qualitative assay (thin-layer electrophoresis with detection by ninhydrin staining). Fractions with greatest specific activity were combined (78 ml eluting at 230 -280 mM NaCl). A 10-ml aliquot of this pool was made 1 M in (NH 4 ) 2 SO 4 by addition of the solid. This solution was chromatographed in two 5-ml portions on a Phenyl-Superose HR 10/10 column (Pharmacia) equilibrated in 20 mM potassium phosphate (pH 7.25), 5 mM DTT, 1 mM EDTA (buffer B) containing 1 M (NH 4 ) 2 SO 4 . After loading, the column was washed with 5 ml of buffer A containing 1 M (NH 4 ) 2 SO 4 and then developed with a gradient of decreasing (NH 4 ) 2 SO 4 concentration (10 ml of 1.0 -0.75 M, then 80 ml of 0.75-0.25 M) in buffer B. Fractions comprising the major protein peak were combined (20 ml from two injections, eluting at 550 -500 mM (NH 4 ) 2 SO 4 ). The pool was frozen in liquid N 2 and stored at Ϫ80°C. SDS-PAGE analysis of the combined fractions showed the fragment to be Ͼ 90% pure (Fig. 1, lane 4).
For purification of the synthetase fragment, lysis of cells (17 g of wet cell paste from 6 liters of culture), streptomycin sulfate and ammonium sulfate fractionation steps, and desalting of redissolved ammonium sulfate pellet were carried out as described for the full-length protein (15). The desalted protein was then loaded on a 2.5 ϫ 25-cm DEAE-Sepharose (Pharmacia) column equilibrated in 50 mM Bis-Tris propane⅐HCl, pH 7.15, 5 mM DTT, 1 mM EDTA (buffer C). The column was developed with a gradient of NaCl in buffer C (70 ml of 0 -100 mM, 610 ml of 100 -300 mM), and fractions containing synthetase activity were pooled (81 ml eluting at 120 -180 mM NaCl). This solution was made 1.2 M in (NH 4 ) 2 SO 4 by addition of the solid and was chromatographed in 10-ml aliquots on a Phenyl-Superose HR 10/10 column (Pharmacia) equilibrated in 20 mM potassium phosphate, pH 6.8, 5 mM DTT, 1 mM EDTA (buffer D) containing 1 M (NH 4 ) 2 SO 4 . The column was developed with a gradient of decreasing (NH 4 ) 2 SO 4 concentration (1-0 M) in buffer D. A fraction of the synthetase activity eluted at the end of this gradient (0 M (NH 4 ) 2 SO 4 ), and the remainder eluted with H 2 O (free of buffer). The fractions were combined, frozen in liquid N 2 , and stored at Ϫ80°C. SDS-PAGE analysis of the combined fractions showed the fragment to be Ͼ90% pure (Fig. 1, lane 3).

Synthesis of Substrates and Inhibitor
General Synthetic Methods-These methods were as described previously (17). 1 H NMR spectra were recorded at 300 and 360 MHz and are reported in the following manner: chemical shift in ppm downfield from internal tetramethyl silane (multiplicity, integrated intensity, coupling constant in hertz).
Materials-Commercially available amino acid and dipeptide precursors were purchased from Bachem or Fisher Scientific. Boc-alanylglycyl benzyl ester was prepared by a literature procedure (18) General Procedures for Coupling Z-Glutamic Acid ␣-Benzylester with C-terminal Protected Dipeptides (Scheme 2)-For the DCC method, a solution of Z-Glu-OBn (3, 1.0 eq), DCC (Fluka, 1.1 eq) and hy-droxybenzotriazole⅐H 2 O (1.2 eq) in DMF (5 ml/mmol 3) was stirred at 0°C for 10 min. A mixture of the C-terminal protected dipeptide ( ⅐ HCl or ⅐ TsOH) (1.0 eq) and NMM (1.0 eq) in DMF (5 ml/mmol 3) was then added to the above solution and the reaction mixture was stirred at 0°C for 1 h, then at room temperature overnight. The precipitated dicyclohexylurea was removed by filtration and the filtrate was concentrated in vacuo. For the MCCA method, to a stirred solution of Z-Glu-OBn (3, 1.0 eq) in dry tetrahydrofuran (10 ml/mmol 3) was added NMM (1.0 eq), followed by i-butylchloroformate (1.1 eq) at Ϫ25°C under dry nitrogen. Stirring was continued at Ϫ25°C for 15 min, then a mixture of NMM (1.1 eq) and the C-terminal protected dipeptide ( ⅐ HCl or ⅐ TsOH) (1.0 eq) in dry tetrahydrofuran (20 ml/mmol 3) was added. The reaction mixture was allowed to stir at Ϫ20°C for 1 h and at room temperature for 24 h. Volatile components were removed under reduced pressure.
Z-␥-Glu(␣-OBn)-Ala-Gly-OBn (6)-6 was synthesized from 3 and 4 by both the DCC and MCCA method (yields 93% and 82%, respectively). After concentration of the reaction mixture, the residue was suspended in EtOAc and washed successively with 5% citric acid, H 2 O, 5% NaHCO 3 , and brine. Removal of the solvent afforded a solid product, ␣-O-Benzyl-Z-␥-Glutamylanalylglycylamide (7)-7 was synthesized from Z-Glu-OBn and Ala-Gly-NH 2 ⅐HCl by both the MCCA and DCC methods with yields of 22% and 78%, respectively. The resultant syrupy residue was triturated with EtOAc. A yellow solid was obtained, which was triturated successively with 5% citric acid, H 2 O, 5% NaHCO 3  ␥-Glu-Ala-Gly-NH 2 (2)-2 was prepared by two methods. In method 1, a mixture of 7 (0.1 g, 0.2 mmol) and 10% Pd/C (50 mg) in MeOH (25 ml) was shaken on a Parr hydrogenator (40 p.s.i.) for 20 h. The catalyst was removed by filtration and the filtrate was concentrated to afford a syrupy material, which was triturated with acetone and a minimum amount of MeOH. The resultant white solid product was recrystallized from MeOH-Et 2 O to afford a white crystalline material (0.05 g, 87%); dryness. The residue was chromatographed on silica gel with CHCl 3 -MeOH (12:1) to give Boc-␥-Glu(␣-OtBu)-Ala-Gly-NH 2 (2.04 g) in 85% yield. Treatment of the protected tripeptide with 50% trifluoroacetic acid in CH 2 Cl 2 under N 2 for 30 min, evaporation of the solvent, and chromatography on Sephadex G-10 afforded ␥-Glu-Ala-Gly-NH 2 in 88% yield. TLC analysis of the product (n-PrOH/AcOH/H 2 O:10/1/5) on silica gave one spot which comigrated with the material synthesized and characterized by Method 1 above, and the 1 H NMR spectra of the products from the two methods were identical.
␥-Glu-Cys-Gly-NH 2 (GSH Amide)-GSH amide was prepared by ammonolysis of glutathione ethyl ester. In a typical reaction, 10 ml of 2 M NH 3 in methanol (Aldrich) was cannulated under N 2 into a sealed, N 2 -purged vessel containing 0.10 g of glutathione ethyl ester, 0.052 g of dithiothreitol, and a single crystal of dimethylaminopyridine (Sigma) as catalyst. The solution was incubated at 42°C for 72 h. The reaction was monitored by thin-layer chromatography on silica with CH 3 Cl:MeOH: 33% acetic acid (5:3:1) in water mobile phase (R F values: glutathione ethyl ester, 0.6; glutathione methyl ester, 0.5; product, 0.2). After 72 h, a 0.2-ml aliquot of 1 M dithiothreitol in methanol was added to the reaction, and the solvent was evaporated in vacuo. The solid was redissolved in 5 ml of 0.1% trifluoroacetic acid in H 2 O. The solution was filtered through a column of 1.5 ml of Dowex AG-50 (Na ϩ form, Bio-Rad) to remove remaining NH 4 ϩ . Fractions containing product (flow-through) were then filtered through a C 18 spice cartridge (Rainin) to remove oxidized and reduced dithiothreitol (product elutes in the void volume, while dithiothreitol is retarded). Fractions containing product were evaporated to dryness in vacuo, and product was redissolved in 10 mM dithiothreitol in H 2 O for use. The product with R F ϭ 0.2 was quantitatively converted to a species that comigrated with GSH by treatment with GSP synthetase/amidase. In this conversion, ammonia was produced, as detected by the glutamate dehydrogenase spectrophotometric assay described below. The quantity of NH 3 released following exhaustive hydrolysis was assumed to be equal to the quantity of glutathione amide originally present.
Phosphonate Analog (8) of Glutathionylspermidine-The hydroxyspermidine-containing phosphopeptide, 8 (see "Results" for structure), was prepared in a convergent synthesis based on the retrosynthetic pathway outlined previously (17). The detailed syntheses of 8 and several related phosphopeptides, together with more extensive inhibition studies, will be published elsewhere. 2
Protein Assays-Protein concentrations were determined by the method of Bradford as supplied by Bio-Rad. Bovine serum albumin (Pierce) was used as standard.
Enzyme Activity Assays-GSP synthetase activity was measured by a previously described coupled assay (15). Amidase activity toward glutathionylspermidine was assayed radiometrically by monitoring conversion of [ 35 S]GSP to [ 35 S]GSH by a previously described thin-layer electrophoresis assay (15).
GSP amidase-catalyzed hydrolysis of glutathione ethyl ester was assayed by coupling production of ethanol to reduction of NAD through the activities of alcohol dehydrogenase and aldehyde dehydrogenase. In a final volume of 400 l, the assay contained 50 mM Tris-HCl, pH 8.2, 1 mM NAD, 500 M DTT, 0.5 mg (170 activity units) of alcohol dehydrogenase, 0.07 mg (4 activity units) of aldehyde dehydrogenase, either 6.6 ng of GSP amidase fragment or 66 ng of full-length GSP synthetase/ amidase, and varying amounts of GSH ethyl ester. GSH ethyl ester has a significant background rate of hydrolysis at pH 8.2 (k obs ϳ 3 ϫ 10 Ϫ4 min Ϫ1 ), so this rate was measured at each concentration prior to initiation of the reaction with the amidase fragment or full-length protein.
The (increasing) absorbance at 340 nm was monitored. By subjecting limiting quantities of GSH ethyl ester to complete hydrolysis, the assay stoichiometry was verified to be 2 mol of NADH produced (in oxidation of ethanol to acetate)/mol of GSH ethyl ester hydrolyzed.
GSP amidase-catalyzed hydrolysis of glutathione amide and ␥-Glu-Ala-Gly-NH 2 was assayed by coupling production of NH 3 to oxidation of NADPH through the activity of glutamate dehydrogenase. In a final volume of 400 l, the assay contained 50 mM Tris-HCl, pH 8.2, 1 mM 2-oxoglutarate, 200 M NADPH, 64 g of glutamate dehydrogenase, either 50 ng of GSP amidase fragment or 800 ng of full-length protein, and varying amounts of GSH amide. The reaction was initiated by addition of substrate, and the (decreasing) absorbance at 340 nm was monitored.
Analysis for Metal Ions in GSP Synthetase/Amidase and the Amidase Fragment-Solutions containing 320 M GSP synthetase/amidase or GSP amidase fragment were dialyzed exhaustively against 50 mM HEPES (pH 7.5), 5 mM DTT, 1 mM EDTA. Catalytic activity was verified after dialysis. A 0.5-ml aliquot of each protein sample and of the dialysis buffer was mixed with 0.5 ml of 4 N HCl (J.T. Baker Ultrex Ultrapure Reagent), and the samples were incubated in sealed tubes at Ͼ 90°C for 12 min with periodic vigorous mixing. Insoluble protein residue was removed by centrifugation, and the supernatant of each sample was submitted to the Chemical Analysis Laboratory of the University of Georgia (Athens, GA) for inductively coupled plasma emission spectroscopy.

Dissection of Bifunctional Protein into Separate Domains
Preparation of Amidase Fragment and Comparison to Fulllength Protein-Our previous finding that trypsin cleavage of GSP synthetase/amidase after Arg-538 gives a 61.6-kDa Nterminal fragment with only amidase activity suggested that the protein might possess independent amidase and synthetase domains (15), and led us to test whether a smaller N-terminal fragment might independently fold into a functional amidase domain. We used a previously constructed set of transposon insertional mutants of gsp in the high copy number plasmid pBluescript (15). Several plasmids with the transposon disrupting gsp still gave rise to ϳ 50-fold greater amidase activity in crude lysates than a pUC19 control plasmid. Of these, one with the transposon inserted after nucleotide 675 encoded the smallest N-terminal fragment, amino acids 1-225 of GSP synthetase/amidase fused to the transposon-encoded dipeptide GV. A plasmid with the transposon inserted after nucleotide 207 did not give rise to increased activity, suggesting that the minimal N-terminal fragment needed for amidase activity is between 69 and 225 amino acids. The 225-amino acid N-terminal fragment was overproduced, and was purified to ϳ 95% homogeneity (Fig. 1, lane 4) by the same two-column procedure employed for the full-length protein. The fragment showed no evidence of insolubility or instability during this procedure. The yield corresponded to 66 mg/liter culture.
Crude estimates of the steady-state kinetic parameters (k cat and K m ) for hydrolysis of glutathionylspermidine by the ami- dase fragment were obtained by the thin-layer electrophoresis assay, and these data suggested that liberation from the Cterminal 394 amino acids significantly activates the amidase domain (severalfold greater k cat /K m than for the full-length protein). Kinetic parameters were subsequently measured for several analogues of glutathionylspermidine (␥-Glu-Cys-Gly-OEt, ␥-Glu-Cys-Gly-NH 2 , ␥-Glu-Ala-Gly-NH 2 , ␥-Glu-Ala-Glyp-nitroanilide) for which colorimetric assays allowed more precise determinations to be made (see section below on substrate specificity studies).
Preparation of Synthetase Fragment and Comparison to Full-length Protein-The recent discovery that the tertiary folds of D-alanine:D-alanine ligase and glutathione synthetase from E. coli are closely related despite minimal similarity in their primary structures (20) suggested that this fold might be characteristic of a family of bacterial ATP-cleaving (ADP-forming), amide bond-forming enzymes. D-ala:D-Ala ligase and GSH synthetase are composed of ϳ 310 amino acids each, and we considered this as an estimate for the extent of the synthetase domain of GSP synthetase/amidase. A second consideration in designing a synthetase-only construct was the sequence similarity among the C-terminal ϳ 370 amino acids of GSP synthetase/amidase and the hypothetical protein products of three other bacterial open reading frames, ygiC and yjfC from E. coli and ygiC from Haemophilus influenzae (Fig. 2). These hypothetical proteins may be related ATP-cleaving, amide bondforming enzymes, and the region of homology may therefore delimit the synthetase domain. On the basis of these considerations, two potential synthetase constructs were prepared. A fragment containing the C-terminal 318 amino acids (beginning with Met-312), which does not span the entire homology region, was found to be insoluble upon overexpression and to lack detectable synthetase activity. In contrast, a fragment of 431 amino acids (beginning at Met-189) is soluble and active for ATP-dependent glutathionylspermidine synthesis. This fragment was purified to ϳ 90% homogeneity (Fig. 1, lane 3)  same two-column procedure employed for the recombinant, full-length protein. The protein eluted from the hydrophobic interaction matrix, Phenyl-Superose, in two fractions, the first eluting with column buffer containing no (NH 4 ) 2 SO 4 and the second eluting with H 2 O (requiring removal of even the buffer salt). This strong interaction suggests that the synthetase fragment has surface-exposed hydrophobic residues not displayed in the full-length protein.
Steady-state kinetic parameters for the synthetase fragment were determined (Table I). The fragment has only slightly (ϳ3-fold) reduced k cat relative to the full-length protein and similar specificity constants for its three substrates. Thus, the C-terminal 50-kDa fragment folds autonomously into a fully functional amide-forming domain.

Substrate Specificity Studies
Specificity of Synthetase Activity-A main objective of our ongoing investigation of GSP synthetases from E. coli and Trypanosomatidae has been to design GSP analogs as inhibitors that might be useful as mechanistic and physiological probes. In design of such analogs, we hoped to substitute a redox inert residue for Cys in the GSH portion of the molecule, in order to avoid synthetic problems involving thiol redox chemistry. To assess whether such a substitution would diminish substrate/inhibitor specificity, we examined ␥-Glu-Ala-Gly as a substrate of the synthetase (Table I). The GSH analog is a reasonable substrate, with k cat (6.4 s Ϫ1 ) identical with that of GSH and K m (10 mM) elevated by ϳ 14-fold.
Specificity of Amidase Activity-In order to define recognition determinants for the amidase activity, potentially hydrolyzable derivatives of glutathione and spermidine were tested as substrates of both the full-length protein and the amidase fragment (Table II). Both the simple amide and the ethyl ester of GSH are good substrates, while no hydrolysis of N 1 -acetylspermidine, N 8 -acetylspermidine, or N 1 -acetylspermine was detected by the qualitative electrophoresis assay following a 100-min incubation under conditions (10 mM substrate, 3.2 M amidase fragment, pH 8.2, 37°C) which gave Ͼ 90% hydrolysis of glutathionylspermidine in 5 min. The cysteine sulfhydryl group is not a crucial recognition determinant, as its substitution by a proton in ␥-Glu-Ala-Gly-NH 2 results in only 10-fold loss in k cat /K m for the amidase fragment and a 2-3-fold increase for the full-length protein. As is often true of amidases, the ester derivative is more efficiently cleaved than the two amide substrates. For example, k cat /K m for hydrolysis of ␥-Glu-Cys-Gly-OEt by full-length protein is ϳ 40-fold greater than for hydrolysis of glutathionylspermidine (15), arising from a greater k cat . Somewhat surprisingly, the simple amide of glutathione is hydrolyzed with a much greater (ϳ 20-fold) k cat than glutathionylspermidine, though its K m also increases somewhat. These data indicate that the amidase active site recognizes predominantly the glutathione portion of glutathionylspermidine.
The kinetic parameters for the various GSH analogues in Table II also demonstrate that the amidase fragment is indeed activated relative to the full-length protein. Depending on the substrate, k cat /K m for the fragment is between 3-fold (␥-Glu-Ala-Gly-NH 2 ) and 74-fold (␥-Glu-Ala-Gly-p-nitroanilide) greater than for the full-length protein. These data suggest that the amidase domain is negatively autoregulated in the context of the full-length protein.

Inhibitor Design and Testing
Phosphonate Analog of GSP as Inhibitor of Synthetase Activity-Those ATP-cleaving (ADP-forming), amide bond-forming enzymes that have been well characterized are believed to employ a mechanism (Scheme 3, A) involving phosphoryl transfer from ATP to the carboxylate oxygen to form an acyl phosphate (9), attack on this intermediate by the amine, and decomposition of the resulting tetrahedral adduct (10) by elimination of phosphate (21)(22)(23). There is extensive precedent for potent inhibition of several of these, including D-alanine:Dalanine ligase (24 -26), glutamine synthetase (27)(28)(29), and glutathione synthetase (30), by phosphonate (11) and phosphinate (12) analogs of the corresponding substrates. Some of these analogs can undergo enzyme-mediated phosphoryl transfer from ATP to the phosphon(phin)ate O (a step that is typically slow, leading to slow-binding inhibition kinetics), to produce species akin to 13 that (presumably) closely mimic the normal tetrahedral intermediates (Scheme 3, B) and inhibit with high affinity (nanomolar K D ) due to their extremely slow dissociation (24,27,28,31). On the basis of these precedents, we designed the hydroxyspermidine-containing phosphopeptide, 8, as a potential inhibitor of GSP synthetase, with the afore- a Details on the synthesis of this chromogenic GSH analog will be reported in a future manuscript describing pre-steady-state kinetic analysis of the GSP amidase reaction (C. H. Lin, D. S. Kwon, J. M. Bollinger, Jr., and C. T. Walsh, manuscript in preparation). mentioned Ala 3 Cys substitution incorporated for synthetic convenience (see Scheme 3). Phosphonate 8 is a potent inhibitor of GSP synthetase activity with respect to the substrate GSH (Fig. 3). Analysis of Fig.  3 according to non-competitive or mixed-type inhibition (Scheme 4) gave inhibition constants of 6 M for inhibitor binding to free enzyme (K I ) and 14 M for binding to the enzyme-GSH complex (KЈ I ). No time-dependent (slow-binding) inhibition was observed (Fig. 4), nor was 8 found to stimulate uncoupled ATP hydrolysis by the synthetase. These results suggest that phosphorylation of 8 is not occurring, and, therefore, that the inhibitor may be acting as a simple bisubstrate analog rather than as an intermediate or transition state mimic.
Phosphonate Analog of GSP as Inhibitor of Amidase Activity-In addition to their use as slow-binding inhibitors of amide bond-forming enzymes, phosphapeptides have also been used as amidase inhibitors (32,33). Thus, for amidases that facilitate direct attack of H 2 O (e.g. zinc or aspartic proteases), a tetrahedral species akin to 14 is an intermediate (Scheme 5), and phosphapeptides analogous to 8 have been observed to bind tightly as mimics (33,34). Phosphonate 8 was therefore tested for inhibition of amidase-catalyzed ␥-Glu-Cys-Gly-NH 2 hydrolysis. In the presence of 100 M substrate, 1.5 mM 8 had no discernible effect on the amidase activity of either the fulllength protein or the amidase fragment. Thus, 8 is at best a poor (high mM K i ) inhibitor of GSP amidase.
Analysis for Metal Ions in GSP Synthetase/Amidase and the Amidase Fragment-In order to test for the presence of a catalytic metal ion in the amidase active site, the metal ion contents of GSP synthetase/amidase and amidase fragment were determined. Neither contains significant quantities of zinc, iron, manganese, cobalt, or nickel, though stoichiometric Ca 2ϩ was found in both. As Ca 2ϩ is not known to act as a cofactor for amide hydrolysis, these observations suggest that a metal ion is not required for GSP amidase catalysis. DISCUSSION The above results demonstrate that the two activities of glutathionylspermidine synthetase/amidase reside in independently folding and functional domains, suggesting that the protein evolved by fusion of amidase and synthetase fragments. By their fusion, any possibility for differential regulation of the component activities of this potential futile cycle at the level of transcription or translation is seemingly eliminated (although production of alternative mRNAs is formally possible). Assuming that the physiological function of the protein does not derive from futile ATP consumption, the two activities are probably differentially regulated post-translationally, either by an allosteric mechanism or by covalent modification. The observation by Tabor and Tabor (35) that glutathionylspermidine accumulates in saturated, anaerobic cultures of E. coli B grown in glucose-rich medium and is rapidly (in less than 5 min) and completely hydrolyzed following dilution into fresh medium lends credence to the suggestion of physiological regulation. The increased amidase activity of the N-terminal fragment relative to the full-length protein suggests one possible mechanism for differential regulation; if this activity is inhibited in the context of the full-length protein, relief of inhibition either by proteolysis or by a conformational change upon ligand binding could selectively enhance amidase activity.
The selective advantage (if any) conferred to E. coli by this unique bifunctional enzyme remains an enigma. We previously proposed that the enzyme might serve to modulate levels of free spermidine or glutathione (15). If so, the presence of synthetase and amidase activities would allow for a bidirectional response (increase or decrease in concentration) without requirement for new protein synthesis or for synthesis or degradation of the relevant metabolite. In addition, in regulating levels of free spermidine, the specificity of the amidase for glutathione-containing amides would render this system orthogonal with the spermidine acetyltransferase system (36).
The amino acid sequence of the amidase domain provides no clue as to catalytic mechanism, as it lacks similarity with any known protein sequence. The amidase may therefore function either by acid/base-assisted direct attack of H 2 O (as do the aspartic and zinc proteases) or by covalent catalysis (as do the serine and cysteine proteases). Two observations suggest that the latter is more likely. First, the domain lacks a catalytic metal ion. Second, the phosphonate 8, which should closely mimic the tetrahedral intermediate for direct attack by H 2 O, does not inhibit the amidase. We are currently using chromogenic ester derivatives of glutathione and rapid kinetic methods to search for a glutathionyl enzyme intermediate.