Role of Electrostatic Interactions on the Affinity of Thioredoxin for Target Proteins

Chloroplast thioredoxin-f functions efficiently in the light-dependent activation of chloroplast fructose-1,6-bisphosphatase by reducing a specific disulfide bond located at the negatively charged domain of the enzyme. Around the nucleophile cysteine of the active site (-W-C-G-P-C-), chloroplast thioredoxin-f shows lower density of negative charges than the inefficient modulator Escherichia coli thioredoxin. To examine the contribution of long range electrostatic interactions to the thiol/disulfide exchange between protein-disulfide oxidoreductases and target proteins, we constructed three variants of E. coli thioredoxin in which an acidic (Glu-30) and a neutral residue (Leu-94) were replaced by lysines. After purification to homogeneity, the reduction of the unique disulfide bond by NADPH via NADP-thioredoxin reductase proceeded at similar rates for all variants. However, the conversion of cysteine residues back to cystine depended on the target protein. Insulin and difluoresceinthiocarbamyl-insulin oxidized the sulfhydryl groups of E30K and E30K/L94K mutants more effectively than those of wild type and L94K counterparts. Moreover, the affinity of E30K, L94K, and E30K/L94K E. colithioredoxin for chloroplast fructose-1,6-bisphosphatase (A 0.5 = 9, 7, and 3 μm, respectively) increased with the number of positive charges, and was higher than wild type thioredoxin (A 0.5 = 33 μm), though still lower than that of thioredoxin-f (A 0.5 = 0.9 μm). We also demonstrated that shielding of electrostatic interactions with high salt concentrations not only brings the A 0.5for all bacterial variants to a limiting value of ∼9 μmbut also increases the A 0.5 of chloroplast thioredoxin-f. While negatively charged chloroplast fructose-1,6-bisphosphatase (pI = 4.9) readily interacted with mutant thioredoxins, the reduction rate of rapeseed napin (pI = 11.2) diminished with the number of novel lysine residues. These findings suggest that the electrostatic interactions between thioredoxin and (some of) its target proteins controls the formation of the binary noncovalent complex needed for the subsequent thiol/disulfide exchange.

The superfamily of Trx 1 comprises small proteins (ϳ12 kDa) whose distinctive feature is the amino acid sequence (-W-C-G-P-C-) functional in thiol/disulfide exchange with other proteins (1). Their widespread occurrence and structural stability are matched by a range of biological properties from metabolic regulation (2) to virus replication (3) and cell proliferation (4). The ability of the redox-active site to reduce a broad spectrum of disulfide bonds in proteins is intimately linked to the regulation of carbon, nitrogen, and sulfur assimilation (2,5,6), mRNA translation (7), and the biosynthesis of deoxyribonucleotides (8). The conversion of the active disulfide bond back to sulfhydryl groups proceeds via NADPH and the flavoprotein NADP-thioredoxin reductase in most cellular compartments (9) and by reduced ferredoxin and the iron-sulfur protein ferredoxin-Trx reductase in oxygen-evolving photosynthetic organisms (10).
In higher plant chloroplasts, the presence of two distinct Trxs, Trx-m and Trx-f, is a highly relevant but still poorly understood issue (11). The comparison of their primary structures with counterparts from other sources revealed that the former bears a significant identity with a variety of prokaryotic Trxs whereas the latter groups with eukaryotic ones (12). Although the relationship between the phylogenetic origin and the metabolism of chloroplasts remains unknown, it was found that the reduced forms of both Trxs participate in the light-dependent modulation of enzymes. Moreover, these studies led to the view that chloroplast Trx-m preferentially activates and deactivates NADP-malate dehydrogenase and glucose-6-phosphate dehydrogenase, respectively, whereas chloroplast Trx-f is highly efficient in the stimulation of key enzymes of photosynthetic CO 2 assimilation (CFBPase, sedoheptulose-1,7-bisphosphatase, phosphoribulokinase, NADP-glyceraldehyde-3-P dehydrogenase). However, evidence suggests the existence of other, perhaps complementary, mechanisms that may regulate the specificity of Trx for target proteins (13)(14)(15)(16). For example, in vitro studies have shown that spinach chloroplast Trx-m and Escherichia coli Trx are indistinguishable from chloroplast Trx-f when the activation of CFBPase is performed in the presence of fructose 1,6-bisphosphate and Ca 2ϩ (13,15). Native (oxidized) CFBPase serves as excellent substrate for the analysis of the interaction between Trx and target proteins because the enzyme activity is effectively modulated not only by the formation of sulfhydryl groups but also by subtle alter-ations of the tertiary structure (14,15). The reductive activation of CFBPase, but not the catalytic step, depends on three cysteines clustered in a solvent-exposed structure encompassing ϳ15 amino acid residues (named the 170's loop by Villeret et al.) (17)(18)(19). In this context, the presence of a highly negative potential around the 170's loop was met with considerable interest because electrostatic attractions constitute a conceptually interesting framework for analyzing the affinity of Trxs for CFBPase (18). Congruent with this view, the introduction and replacement of negatively charged amino acid residues in Trxs was detrimental and beneficial, respectively, for the modulation of CFBPase. For example, K44(58)D, N60(74)D, and Q61(75)D chloroplast Trx-f as well as K69(70)E chloroplast Trx-m were less efficient than wild type counterparts, whereas the substitution of Asp-61 by asparagine in E. coli Trx enhanced 2-fold the affinity for CFBPase (henceforth, the numbering of amino acid residues refers to the location in E. coli Trx and those enclosed between parentheses to the homologous position in the particular Trx) (16,20,21). Nevertheless, previous studies lacked two relevant features related to the functional significance of electrostatic interactions. First, the creation of novel positively charged groups on the surface of Trx should not only enhance the affinity for the negatively charged domain of CFBPase but also impair that for basic proteins. Second, the interaction of Trx with target proteins should be sensitive to the shielding of electrostatic attractions by neutral salts.
Our goal in the present study was to establish whether novel positive charges will make inefficient Trxs much more effective in the activation of CFBPase and concurrently sensitive to high ionic strength. Therefore, the charge distribution on the surface of E. coli Trx was approached to that of chloroplast Trx-f by site-directed mutagenesis. The data reported herein show that mutants of E. coli Trx, in which lysine residues are located at positions 30 and 94, evoke chloroplast Trx-f in the modulation of CFBPase, even if the capacity of the active site for thiol/ disulfide exchange with reductants and oxidants only changes slightly. However, more importantly, the affinity of all variants for CFBPase approaches to a common limiting value when high concentrations of neutral salts shield the attraction between these interacting proteins. Consistent with the predominant role of electrostatics in the interaction between variants of E. coli Trx and the target protein, the higher the number of lysine residues, the lower the reduction rate of a positively charged protein, rapeseed napin (22).

EXPERIMENTAL PROCEDURES
Bacterial Strains and Cloning Vectors-The gene coding for the entire sequence of the E. coli Trx flanked by NdeI and EcoRI restriction sites was amplified by PCR using as template the plasmid pFP1 (21). Subsequently, this fragment of DNA was subcloned into an IPTGinducible vector, pET22b(ϩ), containing a T7 RNA polymerase promoter and a high-affinity ribosome binding site. Detailed properties of this overexpression vector have been described elsewhere (23). The overexpression strain JM109/DE3 (FЈ traD36 proAB lacI q ⌬{lacZ}M15/ recA1 endA1 gyrA96 thi hsdR17 (r K Ϫ ,mK Ϫ ) supE44 ⌬(lac-proAB) relA1) was used as a host strain for transformation in all experiments (Promega, Madison, WI). Bacteria were grown at 37°C in LB medium containing ampicillin (200 g/ml), and cell number was determined by light attenuation at 600 nm.
Materials-Enzymes and solutions for DNA manipulations were used according to either manufacturer's instructions (New England Biolabs, Beverly, MA), Sambrook et al. (24), or Ausubel et al. (25). Deoxiribonucleotides were procured from Amersham Pharmacia Biotech (Uppsala, Sweden) and oligonucleotides from NBI (Plymouth, MN) or GenSet (La Jolla, CA). Bovine pancreas insulin (henceforth insulin) and biochemical reagents were purchased from Sigma-Aldrich. E. coli NADP-thioredoxin reductase was purchased from IMCO (Stockholm, Sweden). CFBPase and chloroplast Trx-f were purified from fresh spinach leaves, according to Stein and Wolosiuk (26).
Construction of Expression Plasmids-Site-directed mutagenesis was accomplished using the "megaprimer" method described by Sarkar and Sommer (27). Briefly, two successive PCRs required a pair of common oligonucleotides containing either the NdeI or the EcoRI restriction sites flanking the gene of E. coli Trx at the 5Ј and 3Ј positions, respectively (primers A and B) and a particular mismatch-bearing oligonucleotide that complemented the internal segment to be mutagenized. The first round of amplification on the wild type gene as template yielded a truncated Trx sequence that spanned from either the 5Ј-or the 3Ј-end to a short sequence beyond the mutagenized codon.
Using the same template, the novel (large) oligonucleotide and the other common primer, the second PCR restored the entire, but mutated, DNA sequence of Trx. The primers used for the specific mutation are listed below (mismatches in bold type). A:

5Ј-GGAGTTGCATATGAGCGATAAAATTATTC
Thirty-three cycles of amplification (1 min at 94°C, 45 s at variable temperatures, 45 s at 72°C) were performed in 0.05 ml of a solution that contained 5 ng of the pET22b(ϩ) plasmid harboring the complete coding sequence of E. coli Trx, 5 nmol of each dNTP, 25 pmol of flanking oligonucleotides, 1 unit of Vent polymerase, and the respective commercial buffer (New England Biolabs). Alternatively, this outline used the single mutant E30K E. coli Trx as template for the construction of the double mutant E30K/L94K E. coli Trx. After the predicted size of the PCR product was confirmed by electrophoresis in 1% agarose, the excess of dNTP and oligonucleotides was removed by passage through Wizard DNA Clean Up (Promega).
Expression and Assay of Recombinant Wild Type and Mutant E. coli Trx-Amplified DNAs and the expression vector pET22b(ϩ) were digested with NdeI and EcoRI, gel-purified, and ligated with T4 DNA ligase. Expression of recombinant wild type and mutant Trx was accomplished by electroporation of the E. coli strain JM109/DE3 and subsequent induction with IPTG. Given that the replacement of amino acids altered the net charge of proteins, we screened the appearance of mutants by nondenaturing PAGE of whole cell lysates. Plasmids were isolated from overnight cultures of selected colonies and, to ensure that only desired mutations were present, the entire coding region was verified by the dideoxynucleotide sequencing on both strands of DNA with Sequenase version 2.0 (Amersham Pharmacia Biotech, Buckinghamshire, UK).
Expression and Purification of Trx-After an overnight incubation, a 20-ml culture was inoculated into 1 liter of LB broth containing ampicillin and grown at 37°C to mid-exponential phase (OD 600 nm ϳ 0.6). IPTG was added to a final concentration of 0.2 mM, and the bacterial culture was incubated at 37°C for additional 4 h. Cells were harvested by centrifugation at 10,000 ϫ g for 10 min, washed off from medium with 30 mM Tris-HCl (pH 7.9), 1 mM EDTA, 150 mM NaCl, and resuspended in the same buffer devoid of NaCl.
After two rounds of freezing and thawing, bacterial cells were subjected to two passages through a French press (100 megapascals). The lysate was clarified by centrifugation at 10,000 ϫ g for 10 min before fractionation of the supernatant fraction with ammonium sulfate (20% to 90% of saturation). The final precipitate was dissolved in 30 mM Tris-HCl (pH 7.9) and 1 mM EDTA and exhaustively dialyzed against the same solution containing 100 mM NaCl. The dialyzate was applied onto a Sephadex G-50 column equilibrated and eluted with 30 mM Tris-HCl (pH 7.9) containing 100 mM NaCl. Fractions having protein-disulfide reductase activity (see below) were pooled and loaded on a DEAE-Sepharose Fast-Flow column (Amersham Pharmacia Biotech) equilibrated in the latter buffer. After desorption of proteins with a linear gradient between 0.1 and 0.6 M NaCl, fractions containing the protein-disulfide reductase activity were concentrated by ultrafiltration or lyophilization and dialyzed against 30 mM Tris-HCl (pH 7.9).
Protein-Disulfide Reductase (Trx) Activity-The activity of Trx as protein-disulfide reductase was assessed in the presence of a dithiol (DTT) and an oxidized protein (insulin, di-FTC-insulin). In the classical turbidimetric assay (28), we followed at 25°C the increase of light attenuation at 650 nm after DTT (0.5 mM) was added to a solution of 100 mM potassium phosphate buffer (pH 7.0), 1 mM EDTA, 100 M insulin, and Trx (ϳ1.5 M). In the novel fluorometric assay (29, 30), the reaction, carried out in 50 mM potassium phosphate buffer (pH 7.0), 1 mM EDTA, 0.7 M di-FTC-insulin and Trx (typically 0.1 M), was started adding DTT (0.1 mM). The reduction of di-FTC-insulin was monitored in a Jasco FP-770 spectrofluorometer by following the emission intensity at 519 nm when the fluorophore was excited at 495 nm. Given the appearance of lag phases in both procedures, we used the maximum rate of the measurement for the calculation of the proteindisulfide reductase activity.
Alternatively, the couple NADPH and NADP-thioredoxin reductase can be used as a reductant of Trx instead of DTT. To this end, the reaction was carried out at 25°C with Trx (ϳ1.5 M) in 100 mM potassium phosphate buffer (pH 7.0), 1 mM EDTA, 0.3 mM NADPH, and either 0.1 mM insulin or 0.026 mM napin. The reaction was started by the addition of NADP-thioredoxin reductase (25 nM) and followed by the decrease of absorbance at 340 nm.
NADP-Trx Reductase Activity-To evaluate the interaction of Trx with NADP-thioredoxin reductase, we analyzed the oxidation of NADPH linked to the reduction of DTNB (Ellman's reagent). The reaction was performed at 25°C in 100 mM potassium phosphate buffer (pH 7.6), 1 mM EDTA, 0.25 mM NADPH, 1 mM DTNB, and Trx (from 0.3 to 10 M). The generation of 2-nitro-5-thiobenzoate was started by the addition of NADP-thioredoxin reductase (25 nM) and followed spectrophotometrically at 412 nm. The molar absorptivity of 2-nitro-5-thiobenzoate at 412 nm was 14,140 M Ϫ1 ⅐cm Ϫ1 for all calculations, and data were fitted to a hyperbolic function by a non-linear least squares method.
Determination of Trx Redox Potentials-The reversibility of the reaction was followed at 25°C according to Moore et al. (33).
Briefly, the reaction cuvette contained initially 100 mM potassium phosphate buffer (pH 7.0), 1 mM EDTA, 50 M NADPH (⑀ 340 nm ϭ 6, 200 , 20 nM NADPthioredoxin reductase, and variable amounts of Trx (from 8 to 20 M). Absorbance at 340 nm was followed spectrophotometrically until it remained constant for 2 min. At this stage, NADP ϩ was added and the time-progress variation of A 340 was measured again. On the basis of (i) changes of NADPH concentration and (ii) the stoichiometry of the above reaction, concentrations of NADP ϩ and oxidized and reduced Trx were calculated. From these data, the calculation of the equilibrium constant CFBPase Activity-To assess the reductive activation of CFBPase, the activity was analyzed by the two-stage assay (31). The enzyme (typically 0.8 g) was incubated at 24°C for 30 min in 100 mM Tris-HCl buffer (pH 7.9), 5 mM DTT, and Trx. Following the incubation, an aliquot was withdrawn and injected into the solution for the assay of activity (100 mM Tris-HCl (pH 7.9), 2 mM MgCl 2 , 1 mM fructose 1,6bisphosphate, and 0.1 mM EGTA). The hydrolysis of fructose 1,6bisphosphate was halted after 3 min at 24°C by adding the reagent for the quantification of the P i released (32).
Protein Determination-Protein concentrations were determined by the method of Lowry et al. using bovine serum albumin as standard (34). A molar extinction coefficient of 13,700 M Ϫ1 ⅐cm Ϫ1 at 280 nm was used to quantitate pure Trx preparations (35).
Gel Electrophoresis-One-dimensional PAGE was run using the Mini-PROTEAN II equipment from Bio-Rad. SDS-PAGE and nondenaturing PAGE were performed according to Schägger and von Jagow (36) and Laemmli (37), respectively. Gels were stained with Coomassie Brilliant Blue and destained in methanol:acetic acid:H 2 O (5:1: Computational Methods-The coordinates of the oxidized wild type E. coli Trx were derived from the high resolution NMR studies of Jeng et al. (38). Model building of spinach chloroplast Trx-f by the automated SWISS-MODEL program (41) used as template the coordinates of human Trx (39), whereas the tertiary structures of spinach chloroplast Trx-m and E. coli variants were based on the solved structures of E. coli (38) and Anabaena (40) counterparts. Electrostatic potential maps on the surface of proteins were obtained by running the GRASP program (42) on a Silicon Graphics Iris 4.X computer. To this end, the electrostatic potential was calculated using the following parameters: internal dielectric of the protein, 2.0; solvent dielectric, 80.0; ionic strength, 150 mM; probe radius, 1.4 Å; ionic radius, 2 Å; net charge of all arginine and lysine residues, ϩ1; and net charge of all aspartic and glutamic residues, Ϫ1.

RESULTS
Computational Analysis of Trx Structure-Any rational attempt to determine the amino acid residues of chloroplast Trx-f that contribute to the formation of the binary complex with CFBPase should be based on the atomic model of both partners. At the onset of the project, there had been no published three-dimensional structures of the former (43), but the structures of the oxidized and reduced forms of E. coli and human Trx were available (44). Therefore, we performed a comparative study among computer-simulated models of site-directed mutants aimed at specifically defining the distribution of charges on the surface that surrounds the redoxactive site. We initially relied on the atomic coordinates of human Trx to simulate the tertiary structure of plastidic Trx-f by the SWISS-MODEL program (41). Similarly, the tertiary structure assignment of spinach chloroplast Trx-m was based on the atomic coordinates of E. coli and Anabaena Trxs (38,40). As expected, modeled chloroplast Trxs retained the general features of these proteins, i.e. they consisted of a ␤-pleated sheet of three parallel and two antiparallel strands surrounded by four helical segments (45).
We subsequently calculated by the GRASP program the electrostatic potential at the level of solvent-accessible surface and represented the isopotential contours at Ϯ 2 kT (Fig. 1). Although modeled proteins had almost the same size (E. coli Trx, 13,300 Å 3 ; spinach chloroplast Trx-f, 13,600 Å 3 ; spinach chloroplast Trx-m, 13,700 Å 3 ), the volume of positively charged isopotential shells of chloroplast Trx-f (3,200 Å 3 ) exceeded that of E. coli Trx (2,500 Å 3 ) and chloroplast Trx-m (1,200 Å 3 ). Mirroring this distribution of charges, the volume of negative contours was 12,700 Å 3 for chloroplast Trx-m, 9,200 Å 3 for E. coli, and 7,800 Å 3 for chloroplast Trx-f. These calculations, however approximate, disclosed that the surface of chloroplast Trx-f lacked large negatively charged areas which were present in E. coli and chloroplast m-type Trxs. Moreover, these data were consistent with a higher pI value for spinach chloroplast Trx-f (6.1) than for E. coli (4.67) and spinach chloroplast Trx-m (4.5) (46). In light of these results, we sought in the primary structure of E. coli Trx information on amino acid residues that (i) interacted with the solvent in the vicinity of the active site, (ii) lay on somewhat distorted regions, and (iii) differed significantly with chloroplast Trx-f. Specifically, the segment between residues 29 -32, which connects the C terminus of ␤-2 strand to the N terminus of ␣-2 helix, contains a variable residue interposed between two highly conserved ones, Ala-29 (except in chloroplast Trx-f) and Trp-31. This particular amino acid is acidic in E. coli and related purple photosynthetic bacteria, serine in Chlorobium, proline in cyanobacteria and chloroplast Trx-m, threonine in eukaryotic Trxs, but only glutamine (spinach, rapeseed, Arabidopsis) or lysine (pea) in chloroplast Trx-f. A highly variable amino acid sequence also occurs in the 94 -96 segment that links the last ␤-strand with ␣-4 helix. Next to conserved Gly-92 and Ala-93, the side chain of the residue 94 is aliphatic and nonpolar in most eubacterial Trx and chloroplast Trx-m, asparagine in eukaryotes, but lysine in Trx-h and chloroplast Trx-f. Mutations in these particular residues led to structures in which, as exemplified by E30K/L94K Trx (Fig. 1c), the surface charge distribution departed from wild type E. coli Trx and approached that of chloroplast Trx-f.
Site-directed Mutagenesis, Expression, and Purification of Variants of E. coli Trx-DNA fragments encoding the entire sequence of the wild type and the E30K, L94K, and E30K/L94K mutants of E. coli Trx were cloned between NdeI/EcoRI sites of the bacterial expression vector pET22b(ϩ), and the nucleotide sequence was confirmed by the dideoxynucleotide termination method. Next, we employed the E. coli strain JM109 (DE3) for the expression of novel plasmids in order to avoid any homologous recombination with the chromosomal gene coding for the wild type Trx. Upon induction of transformed cells and disruption by French press, high speed centrifugation of bacterial lysates yielded all variants of Trx in the supernatant fraction. The purification process started with an ammonium sulfate fractionation of the soluble fraction followed successively by size-exclusion and ion-exchange chromatography over Sephadex G-50 and DEAE-Sepharose Fast Flow, respectively. The purified proteins have been shown to be pure at greater than 95%, as observed on SDS-PAGE (Fig. 2a). Moreover, the migration of novel Trxs in non-denaturing PAGE not only was congruent with an increase of one (L94K), two (E30K), and three (E30K/L94K) positive charges in the molecule of E. coli Trx, it also revealed the absence of any contamination with the wild type counterpart (Fig. 2b). Following the above procedure, the average yield of all Trx variants ranged from 40 to 60 mg/liter of culture media.
In order to obtain an insight on structural differences between mutant Trxs and the wild type counterpart, we analyzed the fourth-derivative of the near ultraviolet absorption spectra (14). All variants showed two large peaks (284.2 and 290.8 nm) and two large troughs (287.6 and 294.2 nm). When the intensity spanning from the maximum at 290.8 nm to the minimum at 294.2 nm was made relative to the intensity between the maximum at 284.2 nm and the minimum at 287.7 nm (i.e. the geometrical factor R ϭ h 2 /h 1 ; Ref. 47), all values ranged from 0.99 to 1.15. Moreover, the fourth-derivative spectra nearly coincided in shape and in wavelength positions when all Trxs were reduced with 10 mM DTT. Intrinsic fluorescence was also monitored as indicator of conformational differences. When the oxidized form was excited at 280 nm, the emission maximum at 339 nm and the fluorescence intensity of all variants were similar to those in the wild type protein. All reduced forms showed an emission maximum at 344 nm and a 3-fold enhancement of the quantum yield at pH 7.0. Taken together, these data demonstrate that the mutation of residues at peripheral loops did not elicit gross alterations around tryptophan and tyrosine residues of oxidized and reduced E. coli Trx.
Further evidence for the lack of large conformational changes came from a different approach. In protein-disulfide oxidoreductases, the redox potential of the catalytic disulfide bond is extremely sensitive to the substitution of amino acids that constitute or are close to the 14-atom ring (48). To determine this particular feature for all variants of E. coli Trx, we analyzed the reversible equilibrium catalyzed by NADP-thioredoxin reductase as outlined by Moore et al. (33).  Given that our primary goal was to study the affinity of reduced Trx for protein substrates, it was necessary to exclude kinetic changes due to the interaction between oxidized Trx and its reductants. Therefore, the oxidation of NADPH by NADP-Trx reductase for the reduction of Trx was assayed in the presence of DTNB. The large excess of the final hydrogen acceptor circumvented the slowness of DTNB reduction (k ϭ 10 3 M Ϫ1 ⅐cm Ϫ1 ) relative to the reduction of Trx by NADP-Trx reductase (k ϭ 10 5 M Ϫ1 ⅐cm Ϫ1 ) and in so doing kept most of Trx in the oxidized state (29). The variants were all equally reactive toward NTR; they maintained the maximum velocity (20 mol of NADPH oxidized⅐min Ϫ1 ) and At variance, when insulin was the final hydrogen acceptor, the rate of NADPH oxidation for E30K and E30K/L94K mutants (5.6 mol⅐min Ϫ1 ) was persistently higher than for the wild type and L94K Trx (4.6 mol⅐min Ϫ1 ). That this effect reflected differences in the interaction between Trx and insulin was confirmed in the reaction driven by DTT, as proteins carrying the E30K mutation were again 30% more active than other variants (Fig. 3). All these results demonstrated that novel positive charges near the active site of E. coli Trx did not affect the interaction with NADP-Trx reductase but elicited slight modifications in the affinity of the reduced form for other target proteins (49).
Trx-mediated Activation of CFBPase-Previous studies have shown that the two-stage assay is very convenient to analyze the activity of chloroplast enzymes because it separates the conversion of the enzyme to a form with different kinetic properties (modulation) from the transformation of substrates to products (catalysis) (2). Accordingly, two well defined kinetic constants account for these events, i.e. the A 0.5 is the concentration of a modulator that yields half of the maximum specific activity and the S 0.5 constitutes the concentration of substrate, cofactor or effector that yields half of the maximum velocity. On this basis, we determined the A 0.5 of Trx by incubating CFBPase with 5 mM DTT at various concentrations of Trx and subsequently measuring the catalytic capacity at 1 mM fructose 1,6-bisphosphate and 2 mM Mg 2ϩ . As shown in Fig. 4, wild type E. coli Trx (A 0.5 ϭ 33 M) was much less efficient than chloroplast Trx-f (A 0.5 ϭ 0.9 M). But more important, mutants derived from the former approached the affinity of the latter, i.e. the A 0.5 of E30K, L94K and E30K/L94K were 9, 7, and 3 M, respectively. In fact, the increase of positive charges on the surface of E. coli Trx enhanced not only the affinity for CFBPase but also the enzyme maximum specific activity (130 mol of P i released⅐min Ϫ1 ⅐mg of protein Ϫ1 ). Given that these novel mutants departed significantly from wild type E. coli Trx in their capacity for reducing other disulfide bearing substrates and became akin to chloroplast Trx-f in the stimulation of CFBPase activity, we concluded that electrostatic components play a crucial role in the interaction with the target protein.
Effect of the Ionic Strength on the Activation of CFBPase-If charge attraction is the main driving force in the recognition of CFBPase by Trx, screening of electrostatic forces by high concentrations of neutral salts should restrict the activation of CFBPase to the thiol/disulfide exchange reaction. As a consequence, the A 0.5 would converge to an unique value because all Trxs share identical active site. Fig. 5a shows that high concentrations of KCl drastically lowered the A 0.5 of wild type E. coli Trx, increased that of L94K and E30K/L94K mutants and did not modify that of the E30K variant. More importantly, however, the convergence of the A 0.5 for all variants of E. coli Trx to the limiting value of ϳ9 M confirmed the prediction that the elimination of electrostatic interactions circumscribes the activation of CFBPase to the thiol/ disulfide exchange event. Moreover, electrostatic screening not only affected variants of E. coli Trx but also increased the A 0.5 of the highly efficient chloroplast Trx-f up to 5 M. The difference between the limiting value of A 0.5 for chloroplast Trx-f and bacterial Trxs may be attributed to other structural factors that also contribute to docking Trx to CFBPase.
When the association between two proteins is dictated mainly by the electrostatic potential, the log of the association rate constant (k 1 ) correlates linearly with the log of the electrostatic contribution to the mean rational activity coefficient for a neutral salt (f*) (50). In other words, log k 1 ϭ log k 1(Iϭ0) ϩ a log f*, where I is the ionic strength of the solution and a is the proportionality constant for a specific pair of proteins. Log f* is related to I by the equation log f* ϭ Ϫ(N⅐M Ϫ1 ϩ M Ϫ1 ⅐I Ϫ1/2 ), where M and N are constants comprising the valency numbers of anions and cations, the minimal distance that separates an anion from a cation, the temperature, and the dielectric constant of the solution (cf. Equation 1 in Ref. 50). We analyzed whether kinetic constants of Trx followed quantitatively the relationship predicted for the electrostatic association with CFBPase. Fig. 5b illustrates that the log A 0.5 of all variants of E. coli Trx and chloroplast Trx-f was linearly dependent on the inverse of the square root of the ionic strength. The best fit of data to log A 0.5 versus N⅐M Ϫ1 ϩ M Ϫ1 ⅐I Ϫ1/2 was obtained when the ionic strength of a solution containing a single electrolyte of symmetrical valency included the contribution of the cationic form of Tris (0.061 M at pH 7.9). Irrespective of whether the FIG. 3. Protein-disulfide reductase activity of E. coli Trxs. The ability of E. coli Trxs to catalyze the reduction of intermolecular disulfide bonds of insulin and di-FTC-insulin with DTT was followed by turbidimetry (empty bars) and fluorometry (filled bars), respectively. The activity of Trx variants relative to the wild type protein is the mean value of at least three determinations. Control rates are 0.061 ⌬A 650 ⅐min Ϫ1 and 0.029 M di-FTC-insulin reduced⅐min Ϫ1 for the precipitation and the fluorometric assay, respectively. slope was positive or negative, the extrapolation of log A 0.5 to I Ϫ1/2 ϭ 0 pointed to a unique value (A 0.5 ϳ 9 M) for all Trxs. This result showed clearly the predominant role of electrostatics in the affinity of Trx for CFBPase and concurrently quantified the contribution of the thiol/disulfide exchange reaction to the modulation process.
Reduction of Rapeseed Napin by Mutant Trx-At this stage, we advanced the hypothesis that positive charges in the target protein would be an obstacle for the protein-disulfide reductase activity of mutant Trxs. If this assumption proved to be correct, the reactivity with cystines of basic proteins would decrease with the number of novel lysyl residues. To test this hypothesis we used napin, a small (12-15 kDa), highly basic (pI ϭ 11.2), and disulfide-rich protein found in seeds of Brassica napus (rapeseed). When the protein-disulfide reductase activity of 1.5 M Trx was assayed in the presence of NADPH and NADPthioredoxin reductase using 26 M napin as oxidant, the rate of disulfide reduction was 1, 1.02, 0.66, and 0.38 mol of NADPH oxidized⅐min Ϫ1 for the wild type, E30K, L94K, and E30K/L94K Trx, respectively. The increase of positive charges on the surface of Trx thus caused a substantial decrease in the rate of napin reduction. DISCUSSION A cascade of thiol/disulfide exchanges links the generation of reducing power in the photosynthetic electron transport system with the modulation of enzymes. When the illumination of chloroplasts triggers the functioning of this process, two proteins approach to the active site of Trx: ferredoxin-Trx reductase to the disulfide bridge of the 14-atoms ring of the oxidized form, and the protein substrate to sulfhydryl groups of the reduced form. In this context, the aim of the present study is to define structural features that modulate the affinity of reduced Trx for CFBPase. Determinants that contribute to the specific recognition of chloroplast enzymes should be located at residues that do not participate in redox reactions, as the active site of all Trxs contains identical amino acids. Indeed, Trxs seem to differ in the distribution of surface charges, while maintaining a common tertiary structure of five ␤-strands and four flanking helices (45,51). On the other hand, cysteines essential for the reductive activation of CFBPase reside in a region surrounded by high density of negative charges, i.e. the 170's loop (17,18). These premises led us to consider electrostatic interactions between the protein-disulfide reductase and the target protein as one of the mechanisms controlling the rate of thiol/disulfide exchange. Significantly, residues 30 and 94 are located in loops connecting secondary structure elements and exhibit high variability among Trxs from different species, suggesting that they may be available for interacting with target proteins. When positively charged residues are placed at these positions, the capacity to interact with hydrogen donors remains relatively constant, whereas the reduction of hydrogen acceptors differs from the wild type counterpart. Kinetic studies summarized in Table I are congruent with this view. Substitution of amino acids does not seemingly bring about perturbations of the tertiary structure, which somewhat potentiate the protein-disulfide reductase activity. In fact, spectroscopic and redox studies clearly marked similarities in the properties of the wild type E. coli Trx with those of mutants. Neither the fourth-derivative analysis of the ultraviolet spectra nor the emission of intrinsic fluorophores showed wavelength shifts or intensity changes. Moreover, redox potentials of mutants, deduced from enzyme-mediated equilibria, are in reasonable agreement with values of wild type E. coli Trx, determined by direct electrochemistry (52).
These results attest to the importance of intermolecular noncovalent interactions in controlling the reduction of disulfide bonds whereby the target protein acquires a functional state. Although A 0.5 values of E30K, L94K, and E30K/L94K E. coli Trx are somewhat higher than that of chloroplast Trx-f, lower values relative to the wild type counterpart indicate an improved affinity due primarily to the better formation of a productive complex with CFBPase. In line with these observations, the effect of charged modulators on the activation of CFBPase provide a cogent argument for the participation of electrostatic interactions in this process. First, negatively  charged dithiols are less efficient than monothiols bearing positive charges in the enhancement of CFBPase activity (26). Second, polycationic spermidine and spermine prevent the stimulation of CFBPase activity by chloroplast Trx-f, but the activation proceeds when these polyamines are removed (53). Finally, we recently observed that anionic Tris(carboxyethyl)phosphine does not activate (reduce) CFBPase, whereas neutral tributylphosphine does so very efficiently (54). In addition, we presented a direct evidence for the functional role of charge interactions in reductive processes catalyzed by Trx, i.e. the incorporation of novel lysine residues to E. coli Trx results in lower reduction rates of disulfides in a positively charged protein, such as napin (Table I).
Recent studies by Schreiber and Fersht (50) have shown that the association rate constant of barnase with barstar correlates with the electrostatic contribution to the mean rational activity coefficient for neutral salts. On this basis, the most direct way to further substantiate the involvement of surface charges in the Trx-mediated activation of CFBPase was to analyze the susceptibility to the ionic strength of the milieu. The relevant finding was that high concentrations of KCl impair the affinity of highly efficient modulators, like E30K/L94K E. coli Trx and chloroplast Trx-f, but improve that of the inefficient wild type E. coli Trx. leading to a gradual approximation of A 0.5 values to a unique value. But more important, the quantitative relationship between the log A 0.5 and the I Ϫ1/2 is congruent with the theoretical analysis that links the rate of protein interaction with the ionic composition of the solution (50).
Our data can be rationalized by a mechanism in which the electrostatic attraction between interacting proteins precedes the thiol/disulfide exchange that has been established for CFBPase (Scheme II) (55).
In this model, long range electrostatic attraction leads Trx-f and CFBPase to the formation of a noncovalent complex. The non-charged patch surrounding the nucleophile Cys-32 of Trx might provide the proper complementarity of both surfaces (reaction 1 in Scheme II). The correct docking drives the reactive thiolate anion of Trx to cleave the disulfide bond of CFBPase with the formation of a transient mixed disulfide between both proteins in which the nucleophilic catalysis dictates the appearance of a thiolate in the CFBPase moiety (reaction 2 in Scheme II) (55). The intramolecular nature of the reaction leading to oxidized Trx and reduced CFBPase could account for the rapidity and, as a consequence, the unstability of the Trx-CFBPase mixed disulfide complex (reaction 3 in Scheme II). The catalytic cycle is then completed by the splitting of the noncovalent complex and release of the products (reaction 4 in Scheme II) (see Ref. 56 for discussion on this point). Sitedirected mutagenesis on chloroplast Trx-f (57) and studies on the pK a values of active site cysteines in E. coli Trx (58) make it likely that Cys-32 is involved in mediating hydrogen transfer to the disulfide of CFBPase, but there is as yet no direct evidence for the formation of a mixed disulfide bridge. In this scheme, Trx-f would exhibit higher affinity for CFBPase than Trx-m because of their peculiar distribution of surface charges, which would favor and prevent, respectively, the formation of the noncovalent complex. Electrostatic shielding with high concentrations of KCl diminishes the contribution of the docking step to the overall reaction rate (step1 in the scheme) and, as a consequence, the activation of CFBPase becomes less sensitive to the source of Trx. Although the convergence of A 0.5 values to a common value for all Trxs is in line with this view, the marginal difference between bacterial Trx and chloroplast Trx-f in the limiting A 0.5 would point out to the role of hitherto uncharacterized structural factors in the formation of the binary complex. Conformation of non-charged residues around the active site of chloroplast Trx-f might be relevant in this aspect, i.e. Trp-31 (45), the additional Cys-60(73), and the amino acid sequence around Pro-76(90).
The involvement of other amino acid residues is claimed in studies where the affinity for CFBPase is impaired by replacing either lysine with glutamic in Trx-m (16) or lysine and polar non-charged residues with negatively charged ones in Trx-f (20). de Lamotte-Guéry et al. (21) also found a 2-fold enhancement of the A 0.5 for CFBPase when an asparagine replaces Asp-61 of E. coli Trx. Modifications we introduced on the primary structure of E. coli Trx increase from 4-to 11-fold the affinity for CFBPase and, in so doing, provide a novel role for two highly variable residues of Trxs. Nevertheless, caution must be exercised in attributing the observed variations in protein-disulfide reductase activity solely to specific amino acid residues. Several lines of evidence suggest that an approach for improving the activity of protein-disulfide reductases is likely to remain dependent on the intrinsic properties not only of the reductant, but also of the oxidant. We have shown previously that the A 0.5 of wild type E. coli Trx is indistinguishable from that of chloroplast Trx-f when fructose 1,6-bisphosphate, Ca 2ϩ , and either chaotropic anions or cosolvents modify the conformation of CFBPase (15). Moreover, results herein showed that the protein-disulfide reductase activity of mutants is sensitive to the target disulfide. For example, while the E30K mutation slightly accelerates the reduction rate of insulin and di-FTCinsulin, it does not impair the cleavage of cystines in napin but assists synergistically the L94K mutation in this process. These observations indicate that the above model is correct but probably not complete, given the contribution of structural factors of the target protein to the docking process.
The the respective reductases remains to be established; FTR has not been analyzed (although it readily reduces both Trx-m and Trx-f), and data on NTR are contradictory. Present results suggest that charged components do not significantly participate in the interaction with the reductase, insofar as kinetic parameters remained unchanged. At variance, the K m for E. coli and human Trx bearing a K36E mutation (which could be considered complementary to E30K) was higher than that for the wild type counterpart (59,60). This discrepancy raises the question whether charge distribution elicited by the K36E mutation modify intramolecular or intermolecular interactions. In brief, the specificity of Trx for other proteins will exploit different structural determinants; Trx-reductases do not seem to be specially sensitive to the distribution of charges around the active site, whereas some target proteins are, e.g. CFBPase.
Given that different forms of Trx seem to coexist in many cellular compartments (61), the proposed mechanism could indeed discriminate between target proteins.