Identification and Characterization of the Functional Amino Acids at the Active Site of the Large Thioredoxin Reductase fromPlasmodium falciparum *

The thioredoxin system, composed of the pyridine nucleotide-disulfide oxidoreductase thioredoxin reductase, the small peptide thioredoxin, and NADPH as a reducing cofactor, is one of the major thiol-reducing systems of the cell. Recent studies revealed thatPlasmodium falciparum and human thioredoxin reductase represent a novel class of enzymes, called large thioredoxin reductases. The large thioredoxin reductases are substantially different from the isofunctional prokaryotic Escherichia coli enzyme. The putative essential amino acids at the catalytic center of large thioredoxin reductase from P. falciparumwere determined by using site-directed mutagenesis techniques. To analyze the putative active site cysteines (Cys88 and Cys93) three mutant proteins were constructed substituting alanine or serine residues for cysteine residues. Further, to evaluate the function of His509 as a putative proton donor/acceptor of large thioredoxin reductase this residue was replaced by either glutamine or alanine. All mutants were expressed in the E. coli system and characterized. Steady state kinetic analysis revealed that the replacement of Cys88 by either alanine or serine and Cys93 by alanine resulted in a total loss of enzymatic activity. These results clearly identify Cys88and Cys93 as the active site thiols of large thioredoxin reductase. The replacement of His509 by glutamine yielded in a 95% loss of thioredoxin reductase activity; replacement by alanine provoked a loss of 97% of enzymatic activity. These results identify His509 as active site base, but imply that its function can be substituted, although inefficiently, by an alternative proton donor, similar to glutathione reductase. Spectral analysis of wild-type P. falciparum thioredoxin reductase revealed a 550-nm absorption band upon reduction which resembles the EH2 form of glutathione reductase and lipoamide dehydrogenase. This spectral feature, recently also reported for the human placenta protein (Arscott, L. D., Gromer, S., Schirmer, R. H., Becker K., and Williams, C. H., Jr. (1997) Proc. Natl. Acad. Sci. U. S. A. 94, 3621–3626), further illustrates the similarity between large thioredoxin reductases and glutathione reductases and stresses the profound differences to smallE. coli thioredoxin reductase.

The thioredoxin system composed of the small peptide thi-oredoxin (12 kDa), thioredoxin reductase (TrxR), 1 and NADPH is involved in the maintenance of the crucial intracellular redox state (1). TrxR catalyzes the transfer of electrons from NADPH to thioredoxin, which itself acts as a reductant of disulfidecontaining proteins, such as ribonucleotide reductase (2). TrxR is a member of the pyridine nucleotide-disulfide oxidoreductase family, which includes lipoamide dehydrogenase, glutathione reductase, mercuric ion reductase, and trypanothione reductase (3). Recently, TrxR from the human malaria parasite Plasmodium falciparum was cloned and recombinantly expressed in a prokaryotic expression system (4,5). The amino acid sequence of the parasite protein shows a high degree of similarity to human TrxR, whereas both proteins have only moderate similarity to the well characterized Escherichia coli TrxR (5,6). These data imply that two classes of clearly distinguishable TrxRs exist, large TrxRs, represented by human and P. falciparum TrxRs, with a subunit size of 58 and 64 kDa, respectively, and small TrxRs such as the prokaryotic E. coli protein, with a subunit size of 35 kDa (4 -8). Both classes show local regions of strong similarity, such as the FAD binding site, but their primary structures are only distantly related. Based upon their amino acid sequences, large TrxRs are much more related to glutathione reductases than to their small counterparts (5,6,8). Spectral analyses of the human placenta TrxR revealed that the catalytic mechanism of large TrxR is similar to that of glutathione reductase and lipoamide dehydrogenase (8). The evolutionary background may be a convergent development of the small TrxR class and the large TrxR class, including glutathione reductase from one ancestor gene (9). The reaction sequence of the small E. coli enzyme was extensively examined, including investigations by site-directed mutagenesis (10,11). These investigations led to the hypothesis that small TrxRs have to undergo a conformational change before they are able to exhibit their catalytic activity (11)(12)(13). To identify the active-site residues and to aid to the understanding of the catalytic mechanism of large TrxRs, we generated several mutations of P. falciparum TrxR (Pf TrxR) and characterized the kinetic and spectral features of the wild-type and mutated recombinant proteins.

EXPERIMENTAL PROCEDURES
Mutagenic Oligonucleotides and Site-directed Mutagenesis-Ten oligonucleotides were designed to replace amino acid residues potentially involved in Pf TrxR activity (Table I). The putative active site cysteines, Cys 88 and Cys 93 , were changed to alanine. Residue Cys 88 was also * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  changed to serine. The putative active site base His 509 was replaced by either alanine or glutamine. The replacement of the original amino acids with the desired ones was performed by an in vitro site-directed mutagenesis method according to Papworth et al. (14). By using high fidelity thermostable Pfu DNA polymerase, low cycle number, and primers designed only to copy the parental strand in a linear fashion, this method minimized unwanted second site mutations and generated mutants in a 1-day procedure. Briefly, the mutagenic oligonucleotides, complementary to the opposite strand of the double-stranded DNA template, were extended by the Pfu DNA polymerase during temperature cycling. 35 ng of the double-stranded, supercoiled expression plasmid pJC40Pf TrxR (5) and 100 ng of mutagenic sense and antisense primers were used in a 50-l reaction mixture containing deoxyribonucleotides, reaction buffer, and Pfu DNA polymerase according to the manufacturer's recommendations (Stratagene). The cycling parameters were 95°C for 30 s, 55°C for 1 min, and 68°C for 8 min, 12 cycles. The linear amplification product was treated with endonuclease DpnI (10 units/l, Boehringer Mannheim) for 1 h to eliminate the parental template. Subsequently, an aliqout of 6 l of this reaction mixture containing the double-nicked mutated plasmid was used for the transformation of competent E. coli BL 21 (DE3) cells (Stratagene). All mutants were analyzed performing the Sanger dideoxy chain termination reaction for double-stranded DNA (15). 100% of the analyzed colonies contained the desired mutation in the absence of other unwanted mutations.
Expression and Purification of Mutant Thioredoxin Reductase-The expression plasmid pJC40 (16), which contains a T7 promotor and encodes 10 histidine residues preceding the N terminus of the recombinant protein, was used for overexpression of the wild-type Pf TrxR and all mutants in E. coli BL 21 (DE3) cells. E. coli BL 21 (DE3) cells carrying the expression plasmid containing the mutated cDNA were grown overnight at 37°C in Luria-Bertani medium (containing 50 g/ml ampicillin). The cells were diluted 100-fold into the same medium and grown until an A 600 of 0.5 was reached. To induce the expression, 1 mM isopropyl-␤-D-thiogalactoside was added to the culture. The cells were harvested 4 h after induction by centrifugation (4000 ϫ g; 10 min) and sonicated (Branson Sonifier 250) in binding buffer (20 mM Tris/HCl, 5 mM imidazole, 0.5 mM NaCl, pH 7.9), and the cell lysate was centrifugated at 100,000 ϫ g for 1 h (Centrikon 1065, Kontron). The resulting supernatant was used for purification on Ni-nitrilotriacetic acid resin (Qiagen) in a batch procedure according to the manufacturer's recommendations. The purity of the recombinant proteins was assessed by SDS-polyacrylamide gel electrophoresis. Protein concentration was determined by the method of Bradford (17), with bovine serum albumin as a standard, and for spectral analyses the respective extinction coefficient was used for protein determination. To exchange the buffer system, recombinant proteins were dialyzed overnight against two changes of 2 liters of 20 mM Tris/HCl buffer containing 0.5 M NaCl and 1 mM EDTA, pH 7.9. For spectral analyses of Pf TrxRC88A at various pH values, the buffer was exchanged by using Centricon 50 (Amicon) filters.
FAD Contents and Extinction Coefficients-The extinction coefficients of the protein bound FAD in the 450-nm region were determined for all Pf TrxRs. Enzyme-bound flavin was liberated by thermal denaturation at 100°C for 30 min in the presence of 10 mM MgCl 2 . The denatured protein was precipitated by centrifugation (19). The concentration of free flavin was determined from its absorption coefficient at 450 nm of 11.3 ϫ 10 3 M Ϫ1 cm Ϫ1 .
Spectra-Spectral analyses of the oxidized form of the proteins were performed under aerobic conditions using 9 -14 M of wild-type and mutated Pf TrxRs. Anaerobic spectra were performed by gently flushing oxygen-free nitrogen through the protein solution for 20 min followed by sealing the cuvette with a screw cap containing a septum adequate for anaerobic work or by overlaying the solution with mineral oil (19). The spectra reported (between 250 and 750 nm) were obtained using a thermostated spectrophotometer (Uvikon 932, Kontron) with a scan speed of 200 nm/min at 22°C. Subsequently, an excess of NADPH was added to the cuvette and the spectral changes during the reduction process of the protein were determined. For anaerobic reduced spectra, NADPH was treated with oxygen-free nitrogen as described above and added to the oxidized protein using a gas-tight Hamilton syringe (19).

Expression of Thioredoxin Reductase Mutants in E. coli-It
was previously reported that Pf TrxR can be expressed as an active homodimer in the E. coli system (5). Expression and subsequent purification of wild-type Pf TrxR, Pf TrxRC88A, Pf TrxRC88S, Pf TrxRC93A, Pf TrxRH509Q, and Pf TrxRH509A yielded in 1-3.5 mg of pure protein liter Ϫ1 of bacterial culture. The purity of the proteins was determined by SDS-polyacrylamide gel electrophoresis (data not shown).
Kinetic Parameters of Wild-type and Mutated Pf TrxR-The thiol reducing activities of wild-type and mutant Pf TrxRs were a The true K m value for NADPH in this mutant could not be determined since the discrimination in rate could not be achieved even at a concentration of NADPH as low as 2 M.

TABLE I
Mutagenic oligonucleotides for site-directed mutagenesis of Pf TrxR Nucleotides that were exchanged in order to obtain the desired mutation are underlined and in bold letters. Numbers in superscript indicate the position of nucleotides in the coding region of the gene.
compared based upon their kinetic properties (Table II). To evaluate the role of Cys 88 , Cys 93 , and His 509 as putative catalytic active residues, the DTNB reduction assay was performed with wild-type and mutant proteins. Mutation of either Cys 88 or Cys 93 to alanine or serine resulted in a total loss of DTNB reducing activity, indicating that these residues are essential for enzymatic activity and most likely represent the active site thiols of the large Pf TrxR. Replacement of His 509 by either glutamine or alanine drastically decreased the reduction capacity of Pf TrxR which implies that His 509 represents the active site base of Pf TrxR. Pf TrxRH509Q retained about 5% of residual activity toward DTNB compared with the wild-type protein.
To eliminate the possibility of a partial compensation of the imidazole side chain by the amido group of glutamine, we substituted His 509 by alanine. However, the H509A mutant still retained 3% of the wild-type enzyme activity with DTNB.
Although the turnover number of the histidine mutants decreased drastically, these mutants had a lower K m value for the model substrate DTNB (Table II). Additionally, the K m for NADPH of the H509A mutant decreased. A similar effect on the K m value for NADPH is well known for the same point mutation of the functionally homologous histidine residue in the E. coli glutathione reductase (20). Spectral Characterization-Spectral analysis of oxidized wild-type Pf TrxR under aerobic conditions shows a characteristic flavin spectrum with peaks around 460 nm and 380 nm (Fig. 1, line a). The max for the oxidized enzyme is at 462 nm (⑀ ϭ 11.46 ϫ 10 3 M Ϫ1 cm Ϫ1 ) (Table III) with a shoulder at 486 nm. During reduction of the enzyme under anaerobic conditions using 10 equivalents of NADPH/FAD, the 462-nm peak bleaches and a shoulder in the 550-nm region is formed. The max of the reduced protein shifts with 26 nm toward the blue end of the spectrum, forming a peak at 436 nm (Fig. 1, line b).
A similar spectral change was observed for other members of this protein family like glutathione reductase during the redox process (21,22) and was recently reported for the human placenta TrxR (8).
The absorption spectra of Pf TrxRC88A ( Fig. 2A) and Pf TrxRC93A (Fig. 3) in the absence of reductant differ from wild-type Pf TrxR by a shift of max to the blue; Pf TrxRC88A to 451 nm with an absorption coefficient of 12.80 ϫ 10 3 M Ϫ1 cm Ϫ1 , and Pf TrxRC93A to 456 nm with an absorption coefficient of 12.62 ϫ 10 3 M Ϫ1 cm Ϫ1 (Table III). Pf TrxRC93A shows a normal oxidized spectrum, and the reduction leads to a decrease of the absorbance at 456 nm and a slight increase of the absorption at the 550-nm band. The exchange of the Cys 88 with serine has drastic effects on the enzyme as expected if the charge transfer band is due to interaction of Cys 93 with the protein-bound flavin. The purified protein has a brick-red color instead of being yellow as all other proteins. Spectral analysis of nonreduced Pf TrxRC88S shows the presence of a charge transfer complex with a max at 442 nm and a shoulder around 550 nm (Fig. 4), resembling the charge transfer produced in the EH 2 species of glutathione reductase and trypanothione reductase (19,21). Remarkably, Pf TrxRC88A is yellow and shows only a slight increase of the absorbance at 550 nm compared with the oxidized wild-type protein, and addition of NADPH does not change the spectral features apart from bleaching the absorbance at 451 nm due to FAD reduction. Determination of the oxidized spectra of Pf TrxRC88A at higher pH values resulted in spectral changes of the nonreduced protein. Fig. 2B shows the absorption spectrum at pH 7.9 (original conditions) and pH 10.6. Apart from the induction of a long wavelength band around 580 nm, the pH change from 7.9 to 10.6 resulted in a blue shift of the flavin peak by 21 nm. The appearance of the 580-nm band at higher pH values suggests that the alanine mutation of Cys 88 affects the pK of Cys 93 , so that this residue is not in its thiolate form at pH values lower than 8.5 (Fig. 2C) and can thus not interact with the protein-bound flavin of  Pf TrxRC88A. Further, these data imply that Cys 93 represents the charge transfer thiol of Pf TrxR.

DISCUSSION
Large thioredoxin-reductases represented by the human, bovine and P. falciparum proteins differ considerably from the small E. coli TrxR (5,6,8,11,18). These two types of isofunctional enzymes are distinct in molecular mass, location of their active site residues, and most likely by their structure and catalytic mechanism (5,6,8). Molecular analyses of Pf TrxR and the human enzyme revealed a high degree of similarity of both proteins to glutathione reductases rather than to E. coli TrxR, which is further supported by the results of Arscott et al. (8) for human TrxR and the data presented here for Pf TrxR.
A series of Pf TrxR mutants altered at the potential redoxactive disulfide bridge (Cys 88 and Cys 93 ) and the potential active site base (His 509 ) were generated to obtain more information about their role for the catalytic reaction of large TrxRs. The conversion of either one of the putative active site thiols, Cys 88 and Cys 93 , into alanine or serine caused a complete inactivation of the thiol-reducing ability of Pf TrxR. These data clearly identify Cys 88 and Cys 93 as essential amino acid residues involved in the catalysis of Pf TrxR. Similar results were obtained when isofunctional residues were mutated in trypanothione reductase and glutathione reductase (19,22). In contrast the replacement of the isofunctional residues in E. coli TrxR resulted in a residual activity of the mutant proteins of about 10%, which was explained by the different topology of the protein-bound flavin and the disulfide axis in comparison to other members of this protein family (11).
Further, it was demonstrated that His 509 is involved in the catalytic mechanism of Pf TrxR. The catalytic activity toward DTNB was drastically reduced but not completely eliminated with the substitution of the His 509 by either glutamine or alanine. The residual activity of 5% after substitution of His 509 by glutamine might be due to the hydrogen-bonding capacity of the glutamine residue or to the participation of another protonable side chain in the enzyme. To completely remove the potential hydrogen-bonding capacity, His 509 was replaced by alanine. Pf TrxRH509A still maintained 3% of catalytic activity with DTNB. This result implies that the function of His 509 can be substituted, although inefficiently, by a nearby side chain that functions as an alternative proton acceptor/donor. These data are consistent with investigations on the E. coli glutathione reductase, where the analogous mutants H439Q and H439A retained 1 and 0.3%, respectively, of the activity of the wild-type enzyme (20). In contrast, in E. coli TrxR Asp 139 rather than His 245 functions as the acid catalyst for the dithioldisulfide redox interconversion (23).
The most characteristic spectral feature of wild-type Pf TrxR is the formation of a new 550-nm absorbance band upon NADPH reduction (Fig. 1). This long wavelength absorbance resembles other EH 2 forms of related proteins (19,21,24). It is indicative of a thiolate flavin charge transfer complex as a stable intermediate of the two electron reduced form of the enzyme (25). Recently, the formation of a thiolate flavin charge transfer complex upon reduction with NADPH and dithionite was described for human TrxR (8) but does not occur in the E. coli enzyme (11,26). The absorption spectrum of EH 2 is influenced by complex formation with NADP ϩ and/or NADPH (27). Such a modulation might also occur in the spectrum reported here for the NADPH-reduced wild-type Pf TrxR slightly increasing the absorbance around 550 nm.
To delineate the role of the active site cysteines and to identify the residue responsible for the thiolate flavin charge transfer interaction of EH 2 , we exchanged the putative active site cysteines of Pf TrxR. In related flavoproteins such as glutathione reductase, trypanothione reductase, and mercuric ion reductase, the carboxyl-terminal cysteine residue of the redox active site is responsible for the formation of the thiolate flavin charge transfer complex (19,24,28). The replacement of Cys 93 by alanine (Pf TrxRC93A) does not change the spectral features of the oxidized form of the protein. Upon reduction with an excess of NADPH a slight increase of the 550-nm band occurs (Fig. 3). This might be due to the binding of excess NADPH to the reduced protein. The replacement of the other active site cysteine Cys 88 resulted in conspicuous spectral properties. While a serine at this position leads to the expected distinctive absorption spectrum of the oxidized protein resembling a permanent charge transfer complex even under nonreductive conditions (Fig. 4), the alanine mutant protein reveals just a slight increase at the 550 nm band (Fig. 2A). The absence of the long wavelength band in Pf TrxRC88A in the absence of the reducing agent was unanticipated, providing that the carboxyl-terminal Cys 93 represents the charge transfer thiol within the Pf TrxR active site. At higher pH values between 8.5 and 10.6 the thiolate charge transfer complex in the oxidized Pf TrxRC88A was induced (Fig. 2, B and C), indicating that in this mutant protein Cys 93 is not in its thiolate form at pH values lower than 8.5. These data imply that the pK of Cys 93 was altered within the protein due to the C88A mutation. The long wavelength band in monoalkylated lipoamide dehydrogenase (EHR) is also missing as shown by Thorpe and Williams (29). Probably the alkylation of the amino-terminal cysteine residue sufficiently perturbs the environment of the charge transfer donor and might increase its pK (21). However, in this case binding of 3-aminopyridine adenine dinucleotide to pig heart lipoamide dehydrogenase induces the appearance of the long wavelength band and is assigned to a charge transfer complex between the carboxyl-terminal thiolate and the protein-bound flavin (30). In glutathione reductase monoalkylated at the amino-terminal cysteine residue, the long wavelength band is fully retained (21). Decreasing the pH value to 3.5 results in almost a complete loss of the charge transfer absorb-ance; increasing the pH to 10.3 yields a blue shift of the FAD band and increases the long wavelength absorbance (31), also seen in Pf TrxRC88A (Fig. 2B).
In conclusion, the data presented here suggest fundamental differences between small TrxRs and large TrxRs. Cys 88 , Cys 93 , and His 509 were identified as essential active site residues of Pf TrxR. Equivalent residues are responsible for the enzymatic activity of glutathione reductase, whereas in E. coli TrxR the active site residues show a completely different topology (9,11). Further, the formation of a stable thiolate flavin charge transfer complex implies that the reaction mechanism of large TrxRs is similar to glutathione reductase and lipoamide dehydrogenase as it has also been suggested by Arscott et al. (8). However, the fact that Pf TrxRC88A does not show the expected charge transfer band at pH values lower than 8.5 hints toward distinctive differences in the architecture of the catalytic center of glutathione reductase and Pf TrxR, that demand more detailed studies dealing with the acid-base chemistry of the Pf TrxR active site. In addition, crystallographic analysis of the threedimensional structure will be necessary to conclusively discuss the active site geometry of the protein.