Allosteric Coupling of Two Different Functional Active Sites in Monomeric Plasmodium falciparum Glyoxalase I*

Glyoxalase I (GloI) catalyzes the glutathione-dependent conversion of 2-oxoaldehydes to S-2-hydroxyacylglutathione derivatives. Studies on GloI from diverse organisms such as man, bacteria, yeast, and different parasites show striking differences among these potentially isofunctional enzymes as far as metal content and the number of active sites per subunit are concerned. So far, it is not known whether this structural variability is linked to catalytic or regulatory features in vivo. Here we show that recombinant GloI from the malaria parasite Plasmodium falciparum has a high- and a low-affinity binding site for the diastereomeric hemithioacetals formed by addition of glutathione to methylglyoxal. Both active sites of the monomeric enzyme are functional and have similar \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(k_{\mathrm{cat}}^{\mathrm{app}}\) \end{document} values. Proteolytic susceptibility studies and detailed analyses of the steady-state kinetics of active-site mutants suggest that both reaction centers can adopt two discrete conformations and are allosterically coupled. As a result of the positive homotropic allosteric coupling, P. falciparum GloI has an increased affinity at low substrate concentrations and an increased activity at higher substrate concentrations. This could also be the case for GloI from yeast and other organisms. Potential physiologically relevant differences between monomeric GloI and homodimeric GloI are discussed. Our results provide a strong basis for drug development strategies and significantly enhance our understanding of GloI kinetics and structure-function relationships. Furthermore, they extend the current knowledge on allosteric regulation of monomeric proteins in general.

The ubiquitous glyoxalase system comprises two enzymes that catalyze the sequential glutathione (or in rare cases, trypanothione)-dependent conversion of methylglyoxal and other 2-oxoaldehydes to 2-hydroxycarboxylic acids. In this reaction, rate-determining dehydration of hydrated 2-oxoaldehyde is followed by the spontaneous formation of diastereomeric hemithioacetals between GSH and the 2-oxoaldehyde ( Fig. 1A) (1,2). The first enzyme, glyoxalase I (GloI 2 ; EC 4.4.1.5), isomerizes both hemithioacetal adducts to a single diastereomeric thioester. The second enzyme, glyoxalase II (EC 3.1.2.6), hydrolyzes the thioester, releasing GSH and 2-hydroxycarboxylic acid (see Ref. 3 for review). Thus, GSH acts as a coenzyme and is not consumed in the overall reaction. Despite decades of intensive research, the physiological functions of the glyoxalase system and the sources, toxicities, and potential functions of its substrates are still a matter of debate.
To date, GloI from different organisms can be roughly subdivided into three different groups according to the type of divalent cation bound at the active site and the number of subunits forming the functional enzyme (Fig. 1B). For example, GloI from human and yeast (4) and Plasmodium falciparum (5) prefers Zn 2ϩ , whereas GloI from several bacteria such as Escherichia coli (6) and Yersinia pestis, Pseudomonas aeruginosa, and Neisseria meningitidis (7) and from the protozoan parasite Leishmania major (8,9) is optimally activated in the presence of Ni 2ϩ . GloI from human (10,11) and E. coli (12) is active as a homodimer. The crystal structures of GloI from human (11), L. major (9), and E. coli (12) show that both subunits contribute residues to both active sites of the homodimer. The N-terminal domain of one subunit interacts with the C-terminal domain of the other subunit (Fig. 1B). Homodimeric GloI consist of four ␤␣␤␤␤ domains (which appear to have arisen by a gene duplication event) forming an eight-stranded ␤-sheet (10). In contrast, GloI from Saccharomyces cerevisiae (13) and the malaria parasite P. falciparum (5,14) is approximately twice the size of homodimeric GloI and has two potential active sites per monomer (Fig. 1B). An alignment of monomeric (putative) GloI from plants, apicomplexan parasites, and Ascomycetes is shown in supplemental Fig. 9. Frickel et al. (13) demonstrated, by site-directed mutagenesis, that both active sites of GloI from yeast are functional. Studies on P. falciparum (PfGloI) support the hypothesis that a second gene duplication and fusion event occurred during evolution of monomeric GloI (5). No structure of a monomeric GloI has been solved yet, and according to sequence alignments and partial molecular models, it is quite likely that the structure of PfGloI differs significantly from the analyzed structures of dimeric GloI (14). PfGloI might be suited as a target for novel antimalarial drugs (5,14); however, it is not known which of the potential active sites should be targeted for inhibition.
Traditionally, allosteric effects are explained either by the symmetry model of Monod et al. (15) or by the sequential model of Koshland et al. (16). These classical models of allosteric effects are restricted to oligomeric proteins. However, allostery (allosteric ϭ other space) means that action in one part of a molecule causes an effect at another site (see Ref. 17 for review). According to this broader definition, allosteric systems are not restricted to oligomeric proteins because the binding of one ligand might affect another binding site of the same monomeric protein. Depending on whether the effector is identical or non-identical to the substrate, homotropic and heterotropic allosteric effectors are distinguished (17). In the case of monomeric proteins, studies on allosteric regulation are usually restricted to heterotropic allosteric interactions.
In this study, we address the questions of whether both potential active sites of PfGloI are functional, whether they are identical in terms of their kinetic parameters, and whether they act independently. Using site-directed mutagenesis and steadystate kinetic measurements, we show that monomeric PfGloI has two different functional active sites with clearly distinguishable substrate affinities. We suggest that both active sites are able to adopt two different conformations and are allosterically coupled. As a consequence, previous results on monomeric GloI from yeast can be reinterpreted. Our results significantly improve the understanding of GloI kinetics as well as structurefunction relationships and extend the knowledge of homotropic allosteric regulation of monomeric proteins in general.
Heterologous Expression and Protein Purification-pQE30 constructs of the wild-type and mutant PfGloI genes were expressed in E. coli strain M15 (Qiagen Inc.). Competent cells were freshly transformed with the respective plasmid before each expression. Initial expression and purification experiments with PfGloI were carried out as described previously (5,14) before Tris and phosphate buffers were replaced with MOPS buffers. Bacteria were precultured for ϳ8 h at 37°C in medium containing 10 g/liter yeast extract, 5 g/liter peptone, 10 g/liter NaCl, 2 g/liter MgSO 4 ⅐7H 2 O, and 10 mM MOPS/NaOH (pH 7.5) supplemented with 50 mg/liter kanamycin and 100 mg/liter carbenicillin. The preculture was added to fresh medium and grown at 30°C to an absorbance at 600 nm of 0.3. Then, 286 mg/liter ZnSO 4 ⅐7H 2 O were added, and the culture was grown for 30 min before induction with 0.3 mM isopropyl ␤-D-thiogalactopyranoside. Cells from 3 liters of culture were harvested by centrifugation (15 min, 6000 ϫ g) ϳ 14 h after induction and resuspended in 20 ml of buffer containing 10 mM MOPS/NaOH (pH 7.8), lysozyme, DNase I, and ϳ75 M phenylmethanesulfonyl fluoride. Resuspended bacteria were stirred for 1 h on ice, followed by sonication at 4°C and centrifugation for 30 min at 30,600 ϫ g. The supernatant was applied FIGURE 1. Schemes of the glyoxalase-catalyzed reactions and the structural variability of GloI. A, the conversion of 2-oxoaldehydes to 2-hydroxycarboxylic acids occurs in four steps and requires GSH as well as the enzymes GloI and glyoxalase II (Glo II). B, GloI from different organisms can be grouped into Zn 2ϩor Ni 2ϩ -dependent homodimeric (I and II, respectively) and monomeric (III) enzymes. Identical and similar domains are highlighted.

Mutation
Sequence

R186E
Sense 5Ј-CTTTTCTCAAACAATGATTGAAGTTAAGAACCCTG-3Ј Antisense 5Ј-CAGGGTTCTTAACTTCAATCATTGTTTGAGAAAAG-3Ј to an S-hexylglutathione-agarose column (Sigma), which was equilibrated with buffer containing 10 mM MOPS/NaOH (pH 7.8). The column was washed with 8 column volumes of 10 mM MOPS/NaOH and 200 mM NaCl (pH 7.8), and recombinant enzyme was eluted with 3 column volumes of 10 mM MOPS/ NaOH, 200 mM NaCl, and 5 mM S-hexylglutathione (pH 7.8). The eluate was loaded onto a nickel-nitrilotriacetic acid column (Qiagen Inc.), which was equilibrated with buffer containing 50 mM MOPS/NaOH, 300 mM NaCl, and 10 mM imidazole (pH 8.0). Afterward, the column was washed with 8 column volumes of the same buffer. The purified enzyme was eluted with buffer containing 50 mM MOPS/NaOH, 300 mM NaCl, and 125 mM imidazole (pH 8.0). Protein concentrations of the fractions were determined using the Bradford assay (18) with bovine serum albumin as a standard. The purity of the eluate was confirmed by reducing 10% SDS-PAGE (19). Determinations of k cat and K m values were performed right after purification, and the proteolytic susceptibility or the pH profiles were analyzed the next day.
Steady-state Kinetics-The steady-state kinetics of the recombinant enzymes were monitored spectrophotometrically using a thermostatted Hitachi U-2001 or a Jasco V-550 UVvisible spectrophotometer. To determine k cat and K m values at 30°C, experiments were performed as described previously (14). Initial measurements were carried out in assay buffer containing 100 mM K x H y PO 4 and 100 mM KCl (pH 7.0) (14), whereas the phosphate buffer was later replaced with buffer containing 50 mM MOPS/NaOH (pH 7.0) (see "Results and Discussion"). Stock solutions of 10 mM GSH (Sigma) and 100 mM methylglyoxal (Sigma) in assay buffer were freshly prepared before each experiment. For a desired concentration of hemithioacetal, the concentrations of methylglyoxal and GSH in the assay mixture were calculated and varied using the equation . The total assay volume was 1 ml. The calculated concentration of free GSH after 5 min of preincubation at 30°C was 0.1 mM in all assays (14). The calculated initial concentration of the hemithioacetal was 5-500 M. All reactions were initiated by the addition of enzyme, and the formation of S-D-lactoylglutathione was followed spectrophotometrically at 240 nm (⑀ ϭ 3.37 mM Ϫ1 cm Ϫ1 ). The final assay concentration of the wild-type enzyme and mutants was 2.5-15 nM, with the exception of PfGloI(E161Q/E345Q) (up to 0.1 M).
The kinetic data of the initial reaction velocities were plotted according to Michaelis-Menten, Lineweaver-Burk, Eadie-Hofstee, and Hanes and fitted using the program SigmaPlot 10.0 (Systat Software Inc.). The plots of several PfGloI mutants were biphasic, and as a consequence, two apparent values for K m (K m1 app and K m2 app ) and k cat (k cat1 app and k cat2 app ) were determined by linear regression. The K m and k cat values calculated from the different plots of a single experiment varied by Ͻ10%. Hill coefficients (n H ) of both phases were determined in a Hill plot by linear regression (using V max ϭ [E]⅐k cat2 app to calculate the maximum reaction velocity of mutants with biphasic kinetics). An equation describing the steady-state reaction velocity according to the scheme shown in Fig. 8 was computed using the program ALBASS 1.0 (20).
pH Profiles-The initial velocities of the formation of S-Dlactoylglutathione were measured at pH values ranging from 5.5 to 9.0 at 25°C. Stock solutions of 100 mM GSH and 1.0 M methylglyoxal in Millipore water were freshly prepared before each experiment. The calculated initial concentration of the hemithioacetal between methylglyoxal and GSH in the assay was 400 M, and the calculated concentration of free GSH after 5 min of preincubation was 0.1 mM. Assays were performed using three different buffers containing 50 mM MES (pH 5.5-6.75), 50 mM MOPS (pH 6.5-8.0), or 50 mM Tris (pH 7.4 -9.0).
Gel Filtration Chromatography-The apparent molecular mass of the wild-type enzyme was studied by gel filtration chromatography on a HiLoad 16/60 Superdex 200 preparation grade column, which was connected to an Á KTA FPLC system (Amersham Biosciences). The column was calibrated with a gel filtration standard (Amersham Biosciences) and equilibrated with buffer containing 50 mM MOPS/NaOH, 300 mM NaCl, and 10 M ZnCl 2 (pH 7.0). Fast protein liquid chromatography fractions were detected photometrically, and peak areas and k av values were evaluated using the software UNICORN 4.11 (Amersham Biosciences).
Proteolytic Susceptibility-The wild-type enzyme and PfGloI(E161Q/E345Q) were concentrated in a Centricon YM-10 device (Millipore Corp.) and treated with subtilisin (Sigma), and the time-dependent protein degradation in the presence or absence of ligands at 25°C was analyzed by reducing 15% SDS-PAGE. All protease assays were performed in 50 mM MOPS/NaOH (pH 7.0) and started by the addition of subtilisin (50 -200 milliunits/ml). The pH of the assay solutions containing either 20 mM S-hexylglutathione or 5 mM hemithioacetal and 0.1 mM GSH was readjusted before the addition of 10 -15 M PfGloI. GSH and methylglyoxal were preincubated for 5 min without PfGloI to allow formation of the hemithioacetal. Aliquots (20 l) were withdrawn from the digestion mixture (90 l) after 0, 1, 5, and 10 min and incubated with 4 l phenylmethanesulfonyl fluoride (final concentration of 17 mM) for 2 min on ice before the addition of 10 l of SDS-PAGE sample buffer and boiling.

RESULTS AND DISCUSSION
Generation of PfGloI Active-site Mutants-Each active site of GloI enzymes contains two highly conserved glutamate residues that are most likely involved in acid-base catalysis during substrate enediolate formation and reprotonation (21-23). According to a previously proposed structural model of PfGloI (14), the pairs Glu 91 /Glu 345 and Glu 161 /Glu 272 are these critical residues at the two putative active sites, which we will refer to as sites A and B, respectively (supplemental Fig. 10). To study their role in catalysis, we generated a set of three different mutants per active site by replacing Glu 91 and/or Glu 345 with glutamine at putative active site A and Glu 161 and/or Glu 272 at putative active site B. In a seventh mutant, both active sites were subjected to mutation by substituting glutamine for Glu 161 and Glu 345 . All of the recombinant proteins were purified to electrophoretic homogeneity (as judged by SDS-PAGE) by two consecutive affinity chromatographic steps (supplemental Fig.  11). The wild-type enzyme and all seven mutants were soluble and stable. The average final yield of the purified proteins strongly depended on the amino acid replacement and differed between 0.3 nmol (PfGloI(E161Q/E345Q)) and 25 nmol (PfGloI(E272Q)) per liter of E. coli culture.
Phosphate Buffers Inactivate PfGloI Active-site Mutants during Catalytic Turnover but Not During Storage-The standard buffers used to measure GloI activity contain (di)hydrogen phosphate for historical reasons (24,25). In previous studies, recombinant wild-type PfGloI was also purified and assayed in (di)hydrogen phosphate-containing buffers (5,14). However, our initial steady-state kinetic measurements of recombinant PfGloI(E91Q) (Fig. 2) and of other mutants (data not shown) revealed that these enzymes were rapidly inactivated during the assay when a phosphate buffer was used. Determination of reaction velocities right after the addition of enzyme was diffi-cult, whereas smaller reaction velocities after 60 s could be accurately reproduced (see, for example, first pair of error bars in Fig. 2, A and B). This observation was surprising because PfGloI(E91Q) was stable for several days with only slight loss of activity when stored at 4°C in buffer containing 50 mM sodium (di)hydrogen phosphate, 300 mM sodium chloride, and 125 mM imidazole (pH 8.0). Inactivation during the assay was decelerated, and the overall reaction velocity was increased by the addition of zinc chloride to the assay mixture or by using an alternative assay buffer containing MOPS instead of phosphate. The addition of zinc chloride to the MOPS-buffered assay mixture did not further increase the activity or slow down the inactivation (Fig. 2, A and B). The wild-type enzyme also had a slightly increased activity with MOPS buffer compared with phosphate FIGURE 2. Dependence of the enzymatic activity on salt concentration and buffer composition during catalytic turnover. Steady-state reaction velocities at 10 s (black bars) and 60 s (white bars) after the addition of enzyme are shown for comparison. All assays were performed with enzyme preparations containing imidazole and phosphate (14). The enzymes were diluted ϳ1:125 in the assay. The influence of Zn 2ϩ and (di)hydrogen phosphate on the activity of PfGloI(E91Q) during catalytic turnover at low and high substrate concentrations is shown in A and B (0.05 and 0.5 mM hemithioacetal, respectively). The formation of S-D-lactoylglutathione at 30°C was followed spectrophotometrically at 240 nm in buffer containing 100 mM K x H y PO 4 and 100 mM KCl (pH 7.0) (P i ) or 50 mM MOPS/NaOH and 100 mM KCl (pH 7.0) (MOPS) in the absence or presence of 10 M ZnCl 2 (Zn 2ϩ ). ZnCl 2 was added to the assay mixture and preincubated at 30°C for 5 min before the addition of enzyme. The activity of the wild-type enzyme with 0.05 mM substrate is shown for comparison (C). PfGloI(E91Q) had a decreased activity in the presence of sodium or potassium chloride (D). Assays were performed with 0.4 mM substrate in 50 mM MOPS/NaOH (pH 7.0). Millimolar salt concentrations are indicated. buffer (Fig. 2C), and PfGloI(E91Q) had a decreased activity at high concentrations of sodium or potassium chloride (Fig. 2D).
We suggest that an interaction between (di)hydrogen phosphate anions and Zn 2ϩ at the active site leads to loss or shielding of the cation during catalytic turnover, resulting in decreased activity and rapid inactivation of PfGloI(E91Q) and other active-site mutants. In the absence of substrate, the zinc cations seem to be protected from an interaction with (di)hydrogen phosphate, explaining the stability of the mutants during storage and pointing to a conformational change during catalysis (see below). As a consequence, the expression and purification protocol was modified (see "Experimental Procedures"), and all following assays for determination of the K m and k cat values for wild-type PfGloI and mutants were performed in buffer containing 50 mM MOPS/NaOH (pH 7.0) without additional salt (supplemental Fig. 12).
Analyses of the Steady-state Kinetics Reveal Two Functional Active Sites with Different Substrate Binding Properties-In contrast to previous reports (5,14), we found that the steadystate kinetics of PfGloI did not follow typical Michaelis-Menten kinetics. The unusual kinetics were probably overlooked because fewer data points were determined and because the evaluation was based solely on Lineweaver-Burk plots. Our plots of the kinetic data are biphasic for the wild-type enzyme and, to a lesser degree, for the PfGloI mutants. Without postulating a mechanism and a model that results in complex equations, the data were fitted empirically by two straight lines (Fig.  3). Accordingly, two apparent values for K m (K m1 app and K m2 app ) and k cat (k cat1 app and k cat2 app ) were determined by linear regression ( Table  2). At this point, these values were not attributed to active site A or B. The K m1 app and k cat1 app values determined at lower substrate concentrations (5-50 M hemithioacetal) are smaller than the K m2 app and k cat2 app values determined at higher substrate concentrations (0.2-0.5 mM hemithioacetal). Hill plots (Fig. 3) reveal a Hill coefficient (n H1 ) Ͻ 1 for lower substrate concentrations. The wild-type enzyme and all seven mutants can be divided into four different groups (Table 2). PfGloI(E161Q/E345Q) is almost completely inactivated, whereas k cat2 app for the wild-type enzyme is ϳ2-fold larger than k cat2 app for the remaining six glutamate mutants. These six mutants can be further subdivided into two groups depending on K m1 app , K m2 app , and the maximum catalytic efficiency (k cat2 app /K m1 app ). In both groups, K m2 app is about twice as high as K m1 app ; however, the K m app values for the two groups differ by roughly an order of magnitude: K m1 app for the wild-type enzyme increases by 5-fold upon replacement of Glu 91 and/or Glu 345 , whereas K m2 app for the wild-type enzyme decreases by ϳ5-fold upon replacement of Glu 272 and/or Glu 161 . As a result, the maximum catalytic efficiency of PfGloI(E91Q), PfGloI(E345Q), and PfGloI(E91Q/E345Q) is 7% of the wild-type enzyme compared with 60% for PfGloI(E272Q), PfGloI(E161Q), and PfGloI(E161Q/E272Q). Table 2 shows that Glu 91 /Glu 345 and Glu 272 /Glu 161 are functionally paired in PfGloI. The requirement of two glutamate residues per active site is in accordance with the current model of the GloI-catalyzed reaction mechanism of both diastereomeric substrates. NMR studies on monomeric GloI from yeast indicated that both diastereomeric hemithioacetals between GSH and methylglyoxal are turned over with identical reaction rates (26). Studies on human GloI suggested that Glu 172 is required for proton transfer between C-1 and C-2 in both diastereomers, whereas Glu 99 is required only for deprotonation of C-1 of the R-diastereomer (for details, see Refs. 22 and 23). The slightly increased values of k cat app for PfGloI(E91Q) and PfGloI(E272Q) compared with k cat app for PfGloI(E345Q) and PfGloI(E161Q), respectively, might reflect a residual activity of the modified active sites of PfGloI(E91Q) and PfGloI(E272Q) because Glu 91 and Glu 272 are probably not required for C-1-C-2 proton transfer of the S-diastereomer, whereas Glu 345 and Glu 161 are essential for the turnover of both substrates. Even though the modified active sites of PfGloI(E91Q) and PfGloI(E272Q) might still have a residual activity with the S-diastereomer, they should be strongly inhibited in the assay because the R-diastereomer becomes a (competitive) inhibitor. Taking together our data and previous studies on GloI from other organisms, we suggest that wild-type PfGloI also converts both diastereomeric hemithioacetals between GSH and methylglyoxal with identical reaction rates. We conclude from the different types of mutants that PfGloI has two functional active sites and that Glu 91 /Glu 272 and Glu 345 /Glu 161 are isofunctional to Glu 99 and Glu 172 in human GloI, respectively (supplemental Fig. 10). As a consequence, Glu 91 and Glu 345 are part of active site A between the N-and C-terminal domains, and Glu 272 and Glu 161 form active site B between the intermediate domains ( Fig. 1B and supplemental Fig. 10).
At saturating substrate concentrations, both active sites have a similar catalytic turnover as reflected by similar k cat2 app values. Interpreting K m1 app and K m2 app as a measure of substrate affinity, we suggest that active site A, including Glu 91 and Glu 345 , is a highaffinity binding site for the diastereomers. This reaction center is half-saturated in the lower micromolar concentration range, whereas active site B (formed by Glu 272 and Glu 161 ) is a lowaffinity binding site that is half-saturated in the higher micromolar concentration range. On the one hand, these differences might facilitate the development of relatively specific glyoxalase inhibitors as novel antimalarial drugs. On the other hand, such specific inhibitors will probably inactivate only one of the two active sites. Provided that methylglyoxal is the physiologically most relevant substrate of PfGloI, it is more promising to target the high-affinity binding site because of its significantly higher catalytic efficiency. However, the range of physiological substrates and their stage-dependent concentrations, functions, and toxicities in P. falciparum are not known (see below).
The pH Optima of Both Reaction Centers Are Similar-PfGloI has a very broad pH optimum with two small local maxima at pH 7.0 and 7.5 and a third local maximum at pH ϳ5.8. We wanted to know whether both active sites have similar pH profiles or different overlapping pH optima and therefore compared the pH dependence of the enzymatic activities of wildtype PfGloI and the glutamate mutants (Fig. 4). The activity of all enzymes decreased significantly above pH 8.0 and showed a very similar insensitivity to pH changes between 6.5 and 8.0. In addition, the functional active sites of PfGloI(E91Q) and PfGloI(E345Q) were sensitive to pH values Ͻ6.5. Because both active sites of the wild-type enzyme were not strikingly inactivated at pH values ranging from 5.5 to 6.5, the sensitivity of PfGloI(E91Q) and PfGloI(E345Q) seems to be due to rapid inacti-P. falciparum Glyoxalase I SEPTEMBER 28, 2007 • VOLUME 282 • NUMBER 39

JOURNAL OF BIOLOGICAL CHEMISTRY 28423
vation of the mutants in 50 mM MES buffers and does not indicate a general property of the low-affinity active site B. We conclude that both active sites have very similar pH profiles.
Both Actives Sites Are Able to Adopt Two Different Conformations and Are Allosterically Coupled-What might be the reason for the observed biphasic kinetics of PfGloI (Fig. 3)? In the case of the wild-type enzyme, the two phases might simply reflect the existence of the two different active sites, which do not necessarily have to interact with each other. The reaction velocities of wild-type PfGloI could indeed be fitted as an overlay of two Michaelis-Menten hyperbolas (reflecting active sites A and B) according to Equation 1 (Fig. 5).
Values obtained from the fit (Fig. 5) qualitatively confirmed the theory of a high-and a low-affinity binding site; however, K mB is too large to explain the activity of PfGloI(E91Q), PfGloI(E345Q), and PfGloI(E91Q/E345Q) at lower substrate concentrations (Fig. 3). Calculated S.D. values of K mB , despite the relatively accurate measurements, also suggest that Equation 1 does not perfectly explain the kinetics of PfGloI. Other, even more important evidence contradicts the hypothesis that two independent active sites are sufficient to explain the biphasic kinetics: inactivation of one of the two reaction centers does not result in monophasic kinetics of the mutants. One might argue that the glutamate replacements are not sufficient and that a residual activity of the modified active site causes the biphasic kinetics of the six active mutants. Indeed, this cannot be absolutely excluded for PfGloI(E91Q) and PfGloI(E272Q) because they might be slightly active with the S-diastereomer (see above). The hypothesis that glutamate replacements are not sufficient to inactivate PfGloI in general, however, is disproved by the almost complete inactivation of PfGloI(E161Q/ E345Q). In addition, similar studies on active-site mutants of GloI from human (21) and yeast (13) showed that replacements of the conserved glutamate residues with glutamine are sufficient to decrease activity by several orders of magnitude. We therefore conclude that the modified reaction centers of at least

PfGloI(E161Q), PfGloI(E161Q/E272Q), PfGloI(E345Q), and
PfGloI(E91Q/E345Q) are inactivated. Thus, a residual turnover at the modified active sites cannot be the reason for the observed biphasic kinetics of all functional PfGloI mutants. Biphasic kinetics and a Hill coefficient Ͻ1 are indicative of enzymes with negative subunit cooperativity (27,28). Such enzymes have to be able to oligomerize and to adopt at least two different conformations. We analyzed PfGloI by gel filtration chromatography and exclusively detected monomeric protein (data not shown), suggesting that there is no subunit cooperativity. Allosteric regulation is also possible for monomeric proteins in case they occur with different conformations and have more than one binding site. What is known about conformational changes in GloI from different organisms? NMR studies on yeast GloI showed a very low incorporation of solvent protons into product (29), suggesting that proton transfers during catalysis occur at a highly protected active site. Fluorescence quenching experiments (30,31) and proteolytic susceptibility studies (31) revealed that human GloI adopts different conformations in a substrate-dependent manner. A comparative analysis of the crystal structures of human GloI suggested that, during catalysis, several side chains at the active site adopt alternative positions and that a flexible loop is closing over the active site (11). Conformational changes during catalysis are also in agreement with the observed phosphate-dependent inactivation of our PfGloI glutamate mutants (see above). Furthermore, wild-type PfGloI and PfGloI(E161Q/E345Q) were treated with subtilisin in the absence or presence of S-hexylglutathione or substrate, resulting in different time-dependent degradation patterns (Fig. 6). Binding of S-hexylglutathione slowed the degradation of the wild-type enzyme and PfGloI(E161Q/E345Q) in comparison with uncomplexed protein. The time-dependent degradation patterns of PfGloI and PfGloI(E161Q/E345Q) in the presence of substrate were similar to samples with uncomplexed protein as far as the intensities of the ϳ45-kDa band are concerned, whereas the intensities of three bands at ϳ30 kDa seemed to depend on whether active enzyme or the almost completely inactive mutant was used. A plausible interpretation of our proteolytic susceptibility analy- FIGURE 3. The steady-state kinetics of wild-type PfGloI and glutamate mutants are biphasic. All assays were performed with enzymes that were purified in MOPS buffers. The formation of S-D-lactoylglutathione at 30°C was followed spectrophotometrically at 240 nm in buffer containing 50 mM MOPS/NaOH (pH 7.0). Results from representative single experiments are shown as Eadie-Hofstee (left panels), Hanes (middle panels), and Hill (right panels) plots (28). f and Ⅺ, data points at lower and higher substrate concentrations, respectively; E, data points that were not used for linear regression analysis. PfGloI(E91Q)/ PfGloI(E345Q)/PfGloI(E91Q/E345Q) and PfGloI(E272Q)/PfGloI(E161Q)/PfGloI(E161Q/E272Q) each have very similar kinetics and form two distinguishable groups. WT, wild-type PfGloI.

TABLE 2 Mutation of glutamate residues at both active sites of PfGloI affects the kinetics
The apparent steady-state kinetic constants of PfGloI were determined from Hanes plots (for example, see Fig. 3). Very similar constants were obtained from Lineweaver-Burk and Eadie-Hofstee plots. All values given were averaged from at least three independent transformation/expression/purification experiments. ND, not determined. Even if both active sites of monomeric GloI are able to adopt different conformations, this is not sufficient to explain the biphasic kinetics; for this, the conformational changes have to be triggered in a concentration-dependent manner. Do the inactivated reaction centers still bind the substrate? This is probably the case because PfGloI(E161Q/E345Q) (with no functional active site) can be purified by an S-hexylglutathioneagarose column, and the proteolytic susceptibility of PfGloI(E161Q/E345Q) also depends on the absence or pres- FIGURE 4. pH dependence of wild-type PfGloI and glutamate mutants. All assays were performed at 25°C with enzymes that were purified in MOPS buffers, and all data points were averaged from at least three independent transformation/expression/purification experiments. Assays were performed using three different buffers containing 50 mM MES (pH 5.5-6.75) (F), 50 mM MOPS (pH 6.5-8.0) (ƒ), or 50 mM Tris (pH 7.4 -9.0) (f). WT, wild-type PfGloI.  . Altered proteolytic susceptibility of complexed PfGloI supports the existence of conformational changes upon ligand or substrate binding. Proteolytic susceptibility assays of PfGloI with subtilisin were performed at 25°C in the presence or absence of either 20 mM S-hexylglutathione as a stable ligand (L) or 5 mM hemithioacetal as a substrate (S). Samples were withdrawn from the assay mixture at the indicated time points, and the reaction was stopped before analysis by reducing SDS-PAGE. Equal sample volumes were loaded onto each lane. A, wild-type PfGloI (ϳ15 M) treated with ϳ200 milliunits/ml subtilisin; B, PfGloI(E161Q/E345Q) (ϳ10 M) treated with ϳ50 milliunits/ml subtilisin. ence of S-hexylglutathione or substrate (Fig. 6B). Substrate binding alone (without catalytic turnover) could therefore have a concentration-dependent effect on the conformation of the other active site. To prove this hypothesis, a mutant with a significantly decreased affinity for the substrate was required. Such a mutant should then possess monophasic kinetics. The two arginine couples Arg 22 /Arg 295 and Arg 186 /Arg 111 of PfGloI align with Arg 37 /Arg 122 in human GloI (14). Crystallographic studies suggested that both guanidino groups of Arg 37 and Arg 122 in human GloI interact with the carboxylate group of the ␥-glutamyl moiety of glutathione (supplemental Fig. 10) (11). We therefore replaced the conserved Arg 22 or Arg 186 residue (both of which are probably localized at the high-and lowaffinity binding sites of PfGloI, respectively) with glutamate and subsequently analyzed the steady-state kinetics of the recombinant mutants PfGloI(R22E), PfGloI(R22E/E91Q/E345Q), and PfGloI(E161Q/R186E/E272Q). The kinetic constants of PfGloI(R22E) support the theory that Arg 22 is involved in substrate binding because k cat2 app is similar to the wild-type enzyme k cat2 app , whereas the K m1 app and K m2 app values are significantly increased (Fig. 7A and Table 3). The kinetics of PfGloI(R22E) can be fitted using Equation 1; however, as is the case for the wild-type enzyme (Fig. 5), the constants obtained are in poor agreement with our data on the glutamate mutants: k catA (17 s Ϫ1 ) is too small, and k catB (455 s Ϫ1 ) is much too large for a single active site (Fig. 7A) compared with k cat2 app values (Table 2), suggesting again that a model of two independent functional active sites is too simple. Replacement of Arg 22 or one of the active-site glutamate residues led to an increase in n H1 (Fig. 7A and Tables 2 and 3), which is in agreement with a coupling of both active sites. Finally, the kinetics of PfGloI(R22E/E91Q/ E345Q) are indeed monophasic, and Eadie-Hofstee and Hanes plots could be accurately fitted with a single straight line ( Fig.  7B and Table 3). Substrate binding at the high-affinity binding site A is probably abrogated in PfGloI(R22E/E91Q/E345Q), and the mutant seems to be trapped in the conformation that predominates at lower substrate concentrations. In contrast, the kinetics of PfGloI(E161Q/R186E/E272Q) are still biphasic (data not shown), demonstrating again that active sites A and B are different. An additional replacement of Arg 111 might be necessary to significantly reduce substrate binding to the low-affinity binding site B. We conclude from our proteolytic susceptibility studies and from the steady-state kinetics of glutamate mutants and PfGloI(R22E/E91Q/E345Q) that both active sites of PfGloI FIGURE 7. Alteration of the potential glutathione-binding site results in decreased substrate affinity of PfGloI(R22E) and monophasic kinetics of PfGloI(R22E/E91Q/E345Q). All assays were performed with enzymes that were purified in MOPS buffers. The formation of S-D-lactoylglutathione at 30°C was followed spectrophotometrically at 240 nm in buffer containing 50 mM MOPS/NaOH (pH 7.0). Measurements were averaged from two independent transformation/ expression/purification experiments, and the data are shown in Michaelis-Menten, Eadie-Hofstee, Hanes, and Hill plots (from left to right) (28). f and Ⅺ, data points at lower and higher substrate concentrations, respectively. Results from PfGloI(R22E) and PfGloI(R22E/E91Q/E345Q) are shown in A and B, respectively. The k cat1 app , k cat2 app , K m1 app , K m2 app , n H1 , and n H2 values for both mutants are listed in Table 3. Furthermore, the Michaelis-Menten plot of PfGloI(R22E) was fitted using Equation 1. The values obtained for k catA and k catB are 17 and 455 s Ϫ1 , respectively. The values obtained for K mA and K mB are 60 and 670 M, respectively.

TABLE 3 Mutation of Arg 22 affects substrate binding
The apparent steady-state kinetic constants of PfGloI were determined from Hanes plots (for example, see Fig. 7). Similar constants were obtained from Lineweaver-Burk and Eadie-Hofstee plots. All values given were averaged from two independent transformation/expression/purification experiments. are able to adopt two discrete conformations and are allosterically coupled.

Model of Positive Homotropic Allosteric Coupling in PfGloI
Mutants-On the basis of our findings that PfGloI adopts different conformations and that the biphasic kinetics of activesite mutants are influenced by binding of the substrate to the modified active site, we suggest the reaction scheme shown in Fig. 8A. Equation 2 is derived from this scheme.
The biphasic reaction velocities of PfGloI(E345Q) (Fig. 8B) and all other functional glutamate mutants (data not shown) can indeed be fitted according to Equation 2. The values obtained for k 8 ϭ k 9 and k 10 are very similar to k cat1 app and k cat2 app , respectively. The association constants K 1 ϭ K 4 and K 7 are in good agreement with K m1 app and K m2 app for the functional active site. In addition, K 3 is in the same concentration range as K m1 app for the modified inactivated reaction center. K 6 was estimated to be ϳ1 for active site A formed between the N-and C-terminal domains and much smaller than 1 for the low-affinity active site B. According to the calculated parameters, E in Fig. 8 symbolizes a conformation with a higher substrate affinity, and the substrate (S) is a positive homotropic allosteric regulator stabilizing or inducing the more active conformation (EЈ) of PfGloI (Fig. 8, A and C).
Biphasic Kinetics Might Be a General Feature of Monomeric GloI-We hypothesize that biphasic kinetics and allosteric coupling are not restricted to PfGloI, but are a common feature of monomeric GloI that has been overlooked for a very long time. For example, Frickel et al. showed that the catalytic efficiencies of both active sites of monomeric yeast GloI are significantly different (see Table 2 Fig. 1 of Ref. 32). The authors' interpretations are not in accordance with later models favoring the one-substrate reaction (Fig. 1A), and the cause of the biphasic kinetics remained unclear. We think that two different active sites of monomeric yeast GloI are a plausible explanation for the observed biphasic kinetics. Another remaining question about yeast GloI is why the k cat for the wild-type enzyme (1700 s Ϫ1 ) is much higher than the sum of the k cat app values for both active-site mutants (1100 s Ϫ1 ) (13). Our model of an allosteric FIGURE 8. Model of the allosteric coupling between the inactivated reaction center and the functional active site of PfGloI mutants. A, PfGloI adopts two conformations, E and EЈ. Both active sites are able to bind substrate (S); however, only the functional active site converts the substrate to the product (P). Equation 2 mathematically describes the scheme. Substrate binding at the inactivated reaction center allows or stabilizes the conformational change. Association constants given in Equation 2 are defined as K 1 ϭ k Ϫ1 /k 1 , K 3 ϭ k Ϫ3 /k 3 , etc. B, the kinetics of PfGloI(E345Q) were averaged from three independent transformation/expression/purification experiments and fitted according to Equation 2. The resulting parameters are in good agreement with the apparent steady-state kinetic constants (K 1 ϭ K 4 Ϸ 98 M, K 3 Ϸ 9 M, K 6 Ϸ 0.01, K 7 Ϸ 199 M, k 8 ϭ k 9 Ϸ 101 s Ϫ1 , and k 10 Ϸ 116 s Ϫ1 ). C, shown is the assignment of the kinetic constants (Table 2) to the two different active sites of PfGloI. The black and white symbols represent functional and inactivated reaction centers, respectively, and conformations E and EЈ are shown as angular and rounded symbols, respectively. coupling between both active sites of monomeric GloI might be the correct explanation for this observation. We therefore suggest a re-evaluation of old data on yeast GloI and a reinvestigation of yeast active-site mutants.
Does homodimeric GloI also possess biphasic kinetics and an allosteric coupling? So far, there are no data supporting this hypothesis, although this might theoretically be the case provided that both identical active sites adopt different conformations in a concentration-dependent manner. Further studies are required to answer this question.
Speculations on the Physiological Functions of Monomeric GloI-Is the structural variability of GloI linked to special functional or regulatory features in vivo? The glyoxalase system has been known for almost a century since its discovery in 1913 (33). To date, we know a lot about the structural variability of glyoxalases, whereas our understanding of the physiological functions is far from complete. Studies on GloI from different organisms support the theory that one main function of the glyoxalase system is the detoxification of methylglyoxal formed, for example, during glycolysis (3). In this regard, positive homotropic allosteric coupling of PfGloI might be especially advantageous to maintain homeostasis because, at very low substrate concentrations, the enzyme has a higher affinity, decreasing the concentration of free 2-oxoaldehydes, whereas at high substrate concentrations, a conformational change leads to the more rapid turnover of the substrate. Indeed, malaria parasites have to adapt to a wide variety of different environments and glucose concentrations (for example, intracellular developmental stages in liver cells or erythrocytes and extracellular stages in human blood or the midgut of the mosquito). The situation is in some ways comparable with yeast and other Ascomycetes because they have to cope with environmental glucose concentrations that can temporarily change over several orders of magnitude as well. Shifts in glycolytic fluxes might therefore be one important element that favors the development of allosteric coupling of (monomeric) GloI.
Both K m app values for PfGloI (Fig. 8C) are much smaller than those for yeast GloI (0.30 and 0.24 mM) (13), resulting in a 5-fold higher catalytic efficiency despite slightly smaller k cat values. These differences in k cat and K m values might reflect different physiological substrate concentrations. Another possibility is that both enzymes are optimized for different substrates. So far, the stage-dependent intracellular concentration of methylglyoxal and even the existence of different 2-oxoaldehydes in P. falciparum are unclear. In yeast, the addition of external methylglyoxal under different growth conditions (34,35) but also the kinetics (36,37) and the effects (38) of the glycolysis-dependent internal formation of methylglyoxal were studied. Despite such great advances in understanding the metabolism of methylglyoxal, it has to be kept in mind that, depending on the type of organism investigated, the physiological sources of 2-oxoaldehydes can differ (for review, see Ref. 3), and novel sources, as well as the existence of alternative 2-oxoaldehydes, cannot be ruled out for P. falciparum and other eukaryotes. The activity of GloI from yeast is quite insensitive to structural variations of the 2-oxoaldehyde (1,39), and it was therefore hypothesized that glyoxalases in general have a very broad substrate spectrum. Yet, we assume that this is not necessarily the case because the broad substrate spectrum of yeast GloI might be due to the two different active sites reflecting a metabolic adaptation to different classes of 2-oxoaldehydes. For example, a significantly lower catalytic efficiency with a certain substrate might be compensated by monomeric GloI having a second, structurally different active site. Because GloI from yeast and P. falciparum is monomeric, it would be worth comparing the catalytic efficiencies of all active sites with different substrates to confirm or disprove these hypotheses.
Although it is accepted that methylglyoxal leads to harmful advanced glycation end products (see Ref. 40 for review), which certainly play a role in diseases such as diabetes (41), very recent findings showed that methylglyoxal also causes protein modifications that can lead to signal transduction and regulation of gene expression (42). Thus, it is possible that allosteric regulation of PfGloI might play a role in the conversion of 2-oxoaldehydes as signal molecules. Another possibility is that the conformational changes in PfGloI may directly influence signal transduction pathways.
Conclusion-Previous studies on GloI from human (21), yeast (13), and other organisms were usually performed in phosphate buffers. We would like to recommend the use of MOPS buffers for future analyses (especially of active-site mutants) because of possible negative and disturbing effects due to an interaction between phosphate and the metal ion at the active site. We have shown that PfGloI has two functional active sites with similar catalytic activities and pH profiles but different substrate affinities. A plausible model of our kinetic data suggests that the hemithioacetal substrate acts as a positive homotropic allosteric regulator of the monomeric enzyme. Substrate binding to the first reaction center induces or stabilizes a different conformation of the second active site. As a consequence, k cat is increased, and substrate affinity is lowered in comparison with the conformation in the absence of the allosteric regulator. Similar and potentially physiologically relevant effects for monomeric (and maybe dimeric) GloI from other organisms could have been overlooked and should be considered in future studies.