Identification and Characterization of the Functional Amino Acids at the Active Center of Pig Liver Thioltransferase by Site-directed Mutagenesis*

By using site-directed mutagenesis techniques, the essential amino acids at the catalytic center of porcine thioltransferase (glutaredoxin) were determined. Seven oligonucleotides were designed, synthesized, and used to construct mutants, ETT-C22S, ETT-C25S, K27Q, and ETT-C78S:C82S, by altering their codons in pig liver thioltransferase cDNA/M13mplS clones.

Identification and Characterization of the Functional Amino Acids at the Active Center of Pig Liver Thioltransferase by Site-directed Mutagenesis* (Received for publication, October 18, 1990) Yanfeng Yang and William W. Wells By using site-directed mutagenesis techniques, the essential amino acids at the catalytic center of porcine thioltransferase (glutaredoxin) were determined. Seven oligonucleotides were designed, synthesized, and used to construct mutants, ETT-C22S, ETT-C25S, K27Q, and ETT-C78S:C82S, by altering their codons in pig liver thioltransferase cDNA/M13mplS clones. Each of the thioltransferases was purified to homogeneity and its dithiol-disulfide exchange, and dehydroascorbate reductase activities were compared with those of the wild-type (ETT). Evidence was obtained that CysZ2 was essential for catalytic activity, and the extremely low pK, value of its sulfhydryl group was facilitated primarily by Arg". The role of LysZ7 at the active center was different from that of ArgZ6 and may be important in stabilizing the E . S intermediate by electrostatic forces. The second pair of cysteines, Cys7' and C y P , nearer the C terminus, were not directly involved in the active center, but may play a role in defining the native protein structure. The replacement of the original Cys with a Ser at position 25 increased rather than decreased the enzyme activity, suggesting that the proposed intramolecular disulfide bond between CysZ2 and CysZs is not necessary for the catalytic mechanism of the Serz5 mutant, but does not rule out such a mechanism for the wild-type enzyme.
ETT-C25A, ETT-R26V, ETT-K27Q, ETT-RS6V: Thioltransferase was originally called glutathione-homocystine transhydrogenase by Racker (1) who discovered the enzyme in beef liver in 1955. In contrast, glutaredoxin was reported as a component in an alternate electron transport system for ribonucleotide reductase in mutant Escherichia coli lacking thioredoxin (2). As more thioltransferases and glutaredoxins were purified and characterized from different sources (1-ll), the similarities between the two proteins became obvious. First, both enzymes could catalyze dithioldisulfide exchange reactions in the presence of GSH (1,3,10,12). Second, amino acid sequence comparison of pig liver thioltransferase (13,14), calf thymus glutaredoxin (15,16), and rabbit bone marrow glutaredoxin (11) demonstrated over 83% sequence identity among the three proteins, and all of them contained an active site sequence of -CysZ2-Pro-Tyr(Phe for the pig enzyme)-Cys""-. Third, polyclonal antibodies raised * This work was supported by United States Public Health Service Grants CA-51972 and GM-38634. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisernent" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$To whom correspondence and reprint requests should be addressed. Tel.: 517-353-3978. against pig liver thioltransferase can recognize calf thymus glutaredoxin and calf liver thioltransferase with the same sensitivity (17), and the antiserum against calf thymus glutaredoxin cross-reacts with human placenta thioltransferase (18). These catalytic, structural, and immunological properties lead to the conclusion that thioltransferase and glutaredoxin are identical.
Pig liver thioltransferase was extensively characterized with respect to its biochemical properties, primary structure, and immunology in this laboratory (8,13,17). The enzyme contains 105 amino acids and has a molecular weight of 11,740 (8). Its primary structure was determined by both direct amino acid sequence (13) and nucleotide sequence (14) analysis. CysZ2 was proposed as the active site of the enzyme (19) whose nucleophilic sulfhydryl group had a pK, of approximately 3.8 (19,20). But, direct evidence for the pivotal role of CysZ2 and the adjuvant roles of and LysZ7 in the reaction mechanism have not been established.
Recently, we reported the high level expression of pig liver thioltransferase in E. coli in the unfused state with all activity and kinetic behavior analogous to the native enzyme (21). The heterologous expression system is efficient and suitable for making the soluble low molecular weight protein. In this paper, we describe the construction of seven mutant thioltransferases (glutaredoxins) by site-directed mutagenesis, their expression, purification to homogeneity, and relative thiol-disulfide exchange and dehydroascorbate reductase catalytic behavior. EXPERIMENTAL PROCEDURES'

Mutagenic Oligonucleotides and Site-directed Mutagene-
sis-Six of the seven oligonucleotides were designed for exchange of the amino acids located in the alleged active center. The seventh mutant replaced the 2nd pair of cysteines, absent in procaryotic glutaredoxin, with serine residues. The oligonucleotides varied between 17-mer and 29-mer in length and are summarized in Table I. Each oligonucleotide was complementary to the nontranscribed thioltransferase cDNA strand in specific regions, except those designed mutant bases (underlined). The base changes of each mutant were confirmed by nucleotide sequencing in which each of the mutant cDNAs in M13mp18s was used as template (data not shown).

Expression and Purification of Mutant Thioltransferme-
The seven mutant pig liver thioltransferase cDNAs of con-' Portions of this paper (including "Experimental Procedures," Tables I and 11, and Figs. 1-9) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press. firmed nucleotide sequence were subcloned into plasmid pKK233-2 between the unique NcoI and Hind111 sites, separately. The seven newly constructed expression vectors, containing specific mutagenized thioltransferase cDNAs, were sequentially named pTT2 to pTT8 corresponding to the mutant protein products of ETT-C22S, ETT-C25S, ETT-C25A, ETT-R26V, ETT-K27Q, ETT-R26V:K27Q, and ETT-C78S:C82S. The expression vector for the wild-type enzyme (ETT) was named p T T l previously (21). JM105 cells were transformed with the expression vectors pTTl to pTT8 separately, and the wild-type and mutant thioltransferases were expressed during 6 h of isopropyl-1-thio-P-D-galactopyranoside induction.
Analysis of the crude extracts by immunoblotting demonstrated that the wild-type and mutant type pig liver thioltransferases were successfully expressed in E. coli with approximately equal efficiency (data not shown). All mutant enzymes were purified to homogeneity, as seen by a single band on SDS-PAGE' (Fig. 1).
Thioltransferase Activity Comparison-The thiol-disulfide exchange activities of wild-type and mutant thioltransferases were compared based on the same amount of protein (0.4 pg) (Fig. 2). The activity of the wild-type thioltransferase (ETT) was defined as loo%, and, accordingly, the relative activities of the mutants were 0% for ETT-C22S, 110% for ETT-C25S, 32% for ETT-R26V, 67% for ETT-K27Q, 5% for ETT-R26V:K27Q, 9% for C25A, and 90% for ETT-C78S:C82S. As speculated, exchange of Cys with a Ser at position 22 of pig liver thioltransferase completely eliminated the enzyme activity. This is the first direct evidence revealing that CysZ2 is the required active site amino acid residue of mammalian thioltransferase (glutaredoxin). It has been known that the sulfhydryl group of Cys" has an extremely low pK, of 3.8 (20), and this property has been speculated to be facilitated by the two neighboring amino acids, Arg2'j and LysZ7. Changing the Arg to a Val or the Lys to a Gln at the active center, we found a 68% or a 33% reduction in activity, respectively. But, if the two basic amino acids were changed simultaneously, only 5% of the wild-type enzyme activity remained. These results indicated that both basic amino acids strongly influenced the activity of the native enzyme. Interestingly, the replacement of Cys with a Ser at position 25 caused an increase rather than a decrease in enzyme activity, suggesting that the formation of an intramolecular disulfide bond between Cys" and Cysl5 is not the only possible mechanism for enzymatic catalysis. However, the substitution of the Cys with an Ala at this position caused a 91% reduction in activity, implying that an amino acid residue with a more polar side group, such as -CHZSH or -CH,OH, at this position is required for optimal enzyme activity. The second pair of cysteines, Cys7' and Cys'', near the C terminus of the enzyme, can be substituted with serines without altering thiol-disulfide exchange activity more than 10%. The thioltransferase activity of the wild-type and the mutant enzymes were assayed over the pH range of 5.5-9.5 using different amounts of protein. The general pattern of thioltransferase activity dependence on pH for each enzyme was similar with low rates a t acidic pH values and with maximum rates at pH 8.5-9.0 (Fig. 3). Thus, modifications at the active center did not change the optimum pH of the enzyme, excluding the total inactive mutant, ETT-C22S.
These results imply that the amino acid substitutions at the active center did not cause major conformational differences, and the activity changes were the result of the active site alterations. Future three-dimensional x-ray crystallographic analysis should establish the validity of this speculation.
The kinetic property of each thioltransferase with respect to the Cys-SO: concentration is shown in Fig. 4. The u versus [SI plots of the wild-type and mutant thioltransferases showed non-Michaelis-Menten kinetics, i.e. at high substrate concentration, the enzyme activity was inhibited. The values of for these enzymes were estimated to be 0.5-0.8 mM. Our data3 clearly showed substrate (Cys-SO: or cystine) inhibition of the thioltransferases, and the products of thioltransferase action, cysteine and HSO:, were inhibitors of glutathione reductase. But the latter inhibition could be neglected in the current study because of the excess of glutathione reductase and the negligible concentration of cysteine and HSO: at the initial stages of the reactions.
The DHA Reductase Activity-In the presence of GSH, thioltransferase can catalyze the reduction of DHA to ascorbic acid (28). The intrinsic DHA reductase activity of the wildtype and each of the mutant enzymes was measured as described under "Experimental Procedures" and compared with each other based on the same amount of protein (0.4 pg) (Fig.   5). With the activity of the wild-type enzyme defined as loo%, the relative activities of the mutants were 0% for ETT-C22S, ETT-R26V:K27Q, and ETT-C25A, 194% for ETT-C25S, 30% for ETT-R26V, 73% for ETT-K27Q, and 71% for ETT-C28S:C82S. Like the thiol-disulfide exchange activity, the ETT-C22S mutant had no detectable DHA reductase activity. This result indicates that CysZ2 is very likely the catalytic site for both intrinsic activities. Compared with the thiol-disulfide exchange activity, ETT-C25S had a significantly greater DHA reductase activity, whereas there was no detectable activity for ETT-C25A. The evidence suggests that a serine is more favored than a cysteine at position 25, especially for DHA reductase activity. The DHA reductase activity of thioltransferase is less active in mutants ETT-R26V, ETT-K27Q, ETT-C78S:C82S, and ETT-R26V:K27Q than the corresponding thiol-sulfide exchange activity. The co-existence of the two basic amino acids, Arg" and LysZ7, appear to facilitate both intrinsic activities of thioltransferase.
Isoelectric Focusing Analysis-The wild-type thioltransferase (ETT) and mutant thioltransferases were analyzed both in their reduced forms (Fig. 6A) and oxidized forms (  Fig. 6). We speculate that the upper one is the unoxidized (reduced) form and the lower one is the oxidized form for both mutants.
The pK, Values of Mutant and Wild-type Thioltransferases-An example of a plot of l/[TT],,d as a function of time is shown in Fig. 7, in which 60 p~ ETT-C25S and 60 p M iodoacetamide were incubated in 100 mM sodium phosphate buffer, pH 6.8, at room temperature. The apparent rate constant, kapp, was 5.5 mM" min" and the corresponding halftime, tlIz, was 3.0 min. Similar plots at various designated pH values for each enzyme were drawn (data not shown), and their k,,, values were calculated. The pK,, values of the Cys" ' Y. Yang  sulfhydryl group for each enzyme was obtained from the midpoint of plots of k,,, uersus pH (Fig. 8). For the expressed wild-type thioltransferase (ETT), the apparent rate constant was pH dependent over the pH region of 3.0 to 4.5, whereas it was pH independent between pH 4.5-8.5. The extremely low k,,, values below pH 3 indicated that Cysz2 was in the sulfhydryl form (-SH) and not sensitive to the alkylating reagent. The increasing k,,, values between pH 3.0 and 4.5 signified that the deprotonation on the Cysz2 sulfhydryl occurred in this pH range and that more thiolated forms became exposed to iodoacetamide. The unchanged k,,, values over the pH region of 4.5 to 8.5 implied that the maximum rate of alkylation reaction was reached and all Cysz2 side chains were in the thiolate form. The pK, value of Cysz2 of the wild-type recombinant thioltransferase was about 3.8, consistent with that of the native enzyme (20). Thus, the acetylation at the N terminus of the native pig liver enzyme has no influence on the pK, of Cysz2. The pK, values of the Cyszz side chain were estimated to be 4.9 for ETT-C25S, 4.3 for ETT-K27Q, 5.9 for ETT-C25A, and 4.4 for ETT-C78S:C82S, all more basis that that of the wild-type. Substitution of the cysteine at position 25 with either a serine or an alanine caused a more basic shift of the pK, of Cysz2 than the changes at position 27 (ETT-K27Q) and position 78 and 82 (ETT-C78S:C82S). Lysz7 as well as ArgZ6 was originally speculated to facilitate the low pK, at Cys2'. However, in the mutant, ETT-K27Q, the pK, of CySZ2 was only slightly increased (approximately 0.5 pH unit) and the activity decreased by 33%, whereas a greater pK, increase occurred in the mutant ETT-C25S, and the enzyme activity was raised 10%. Thus, for ETT-K27Q, the 33% loss of enzyme activity was not the result of the slight pK, increase, but Lys17 likely played some other role in the enzyme catalytic mechanism. Currently, the precise function of LysZ7 is unknown, although one possibility is that this residue can stabilize the enzyme-substrate intermediate by ionic interactions between its positively charged side chain and a negatively charged group of the substrates, e.g. GSH. It is interesting that replacing Cys2' with Ser2' and with Alaz5, separately, resulted in different pK, changes at Cysz2 (1.1 and 2.1 pH units, respectively) and resulted in totally different activity alterations (10% increase uersw 91% decrease, respectively). Compared with serine, the relatively more hydrophobic alanine replacing Cys at this position might disturb the local three-dimensionsal structure of the active center and cause CysZ2 to be less exposed. The exchanges of the two downstream cysteines, Cys7* and CysS2, with 2 serines had little influence on the pK, value of Cysz2. Accordinging, the low pKa of Cyszz sulfhydryl in the wild-type enzyme group is not facilitated by either Cys2', L y P , or Cys7* and CysSz. In contrast, the amino acid responsible for the low pK, at Cys" was Arg26. We could not measure the pK, value of mutants, R26V and R26V:K27Q, because the apparent rate constants, kapp, of both mutants were very low (about 0.4 m"' min") and pH-independent over the region of 2.5 to 8.5; that is, the two mutants were not sensitive to iodoacetamide. These results clearly indicated that replacing A r e with ValJ6 significantly decreased the deprotonation of the active site sulfhydryl group. We conclude that the role of Arg26 is to facilitate the low pK, of Cyszz, i.e. enhance its S-nucleophilicity, necessary for the thioltransferase catalytic reaction. Despite the apparent correlation between Cysz2SH pK,, the presence of arginine at position 26, and enzyme activity, it is still conceivable that valine at position 26 may exert its effect on activity by blocking accessibility of Cyszz.

DISCUSSION
The primary structure of thioltransferase (glutaredoxin) has been reported for E. coli (31), calf thymus (15,16), pig liver (13,14), rabbit bone marrow ( l l ) , and, recently, yeast (32). All the proteins have an active site of Cys-Pro-Tyr(Phe for pig enzyme)-Cys-, while the three mammalian enzymes contain an additional pair of cysteines near the C terminus (Fig. 9). The sequences of the two regions containing the cysteine pairs are highly conserved in thioltransferases (glutaredoxins). The first region is the active center for each enzyme, and the sequences in this region are identical except that a Phe instead of a Tyr was found in the pig enzyme, and only one basic amino acid is located in this region for the E. coli and yeast enzymes. In the second conserved region near the C terminus, the sequences are the same in the three mammalian enzymes except that a Thr is replaced by a Ser in the rabbit enzyme. Despite the lack of the extra pair of cysteines, the E. coli and the yeast enzyme still have considerably high sequence identity with the mammalian proteins in this region suggesting that the second conserved region might have a structural function. However, our data showed that the replacement of the second pair of cysteines in the pig enzyme affected its activity only slightly. It is interesting that similar cysteine pair distributions occur in thioredoxin, another low molecular weight protein catalyzing various thioldisulfide exchange reactions, i.e. there are two pair of cysteines in mammalian enzymes (33) and only one pair in the active center of the bacterial and yeast thioredoxins (34,35).
The present work provides the first direct evidence for the identification of the essential amino acids in the active center of a mammalian thioltransferase (glutaredoxin). Since the substitution of certain original amino acids in the mutants caused large PI shifts, the pH of the buffers used in the purifications had to be modified (Table 11). The pK, value of the sulfhydryl group of CySL2 (pKa = 3.8) is much lower than that of normal cysteine (pK, = 8.5 f 0.5) (31), and CysZ2 was speculated to be the active site of thioltransferase (19). This was directly confirmed by results of changing CysZ2 to S e P , which totally eliminated the enzyme activity. These data also showed that the amino acids Arg'fi and LysZ7 are required for optimal enzyme activity since exchange at any of these positions generally decreased the activity with the exception that replacement of Cys2' with Ser" increased rather than decreased the enzyme activity. This discovery necessitated a reevaluation of the enzyme mechanism for the mutant enzymes not capable of forming an intramolecular disulfide. Individual replacement of ArgZ6 or Lysz7 with Valz6 or Glnz7 led to a relative 32% or 67% enzyme activity, respectively, but altering the 2 basic residues together with two neutral amino acids caused a cooperative loss in activity.
We did not change the two amino acids, Pro and Phe, between the 2 cysteines at the active center. However, two such studies in T4 (36) and E. coli thioredoxin (37) were reported recently. Joelson et al. (36) constructed three mutants, CGPC, CVPC, and CGYC, at the active site of T4 thioredoxin which has the sequence, CVYC, and found no significant changes in the enzyme activity. Gleason et al. (37) constructed two mutants, CCGRPC and CAC, at the native protein active site, CGPC, of E. coli thioredoxin by altering the size and demonstrated that the longer mutant lost 85% of its activity, whereas the one with a shorter chain had no activity. The mutation studies implied that the distance rather than the specific amino acids substituted between the two active site cysteines is important, and the 14-atom disulfide loop at the active site of the oxidized enzyme seems to have been the preferred choice during evolution. However, as we Active Site of Pig Liver Thioltransferase demonstrate here, a serine at position 25 might have been expected to enjoy an evolutionary advantage.
The present results indicated that exchange of some amino acids, especially the charged residues (e.g. Cys, Arg, and Lys) at the active sites caused the PI shifts of the proteins. Normally, when a protein is in its native folded form, the PI value of the protein is the sum of the net charges on the surface of the molecule (38). In the present case, all thioltransferases, analyzed on isoelectric focusing gel, were in their native folded forms. Thus, the PI values of these proteins should reflect the total surface charges of the native molecules, and the PI shift should follow the changes of the surface charged groups. The substitution of Cys with a Ser at position 22 caused a PI change in both reduced (0.5 pH unit) and oxidized (0.4 pH unit) forms, whereas, the same substitutions at positions 25, 78, and 82 had no influence on the PI values (Fig. 6). These results provide additional evidence that CysZ2 is the only cysteine residue in its thiolated form (-S-) at a physiological pH and that it is exposed or partially exposed to the molecular surface, whereas C y P , C Y S~~, a n d Cysa' are either in their sulfhydryl forms (-SH) or buried in the protein core. The three-dimensional structures of E. coli (39) and T4 (40) thioredoxins have shown that their active sites are parts of a protrusion of the molecule. So, it is possible that the active site of thioltransferase (glutaredoxin) is on the surface of the molecule, and only this location of the active center can explain how the low molecular weight enzyme interacts with its protein substrates, such as pyruvate kinase (41) and ribonucleotide reductase (42). Certainly, this proposal needs further support from the three-dimensional structure of thioltransferase, and such studies are now in progress.
Striking PI shifts were observed in both reduced forms and oxidized forms when the 2 basic amino acid residues were replaced by neutral amino acids either individually or together. Compared with the wild-type enzyme (ETT), mutants ETT-R26V, ETT-K27Q, and ETT-R26V:K27Q have more acidic PI values and the double mutant has the most acidic PI. These PI changes apparently resulted from the loss of the positively charged Arg or Lys or both in the active center possibly located on the molecular surface. We showed that thiol-disulfide exchange and DHA reductase activity of the mutants ETT-R26V, ETT-K27Q, and ETT-R26V:K27Q are significantly decreased, and, thus, the two basic amino acids, Arg2' and LysZ7, must be involved in each catalytic activity.