Distinct subcellular localization in the cytosol and apicoplast, unexpected dimerization and inhibition of Plasmodium falciparum glyoxalases

Authors


*E-mail marcel.deponte@gmx.de; Tel. (+49) 89 2180 77122; Fax (+49)  89 2180 77093.

Summary

The ubiquitous glyoxalase system removes methylglyoxal as a harmful by-product of glycolysis. Because malaria parasites have drastically increased glycolytic fluxes, they could be highly susceptible to the inhibition of this detoxification pathway. Here we analysed the intracellular localization, oligomerization and inhibition of the glyoxalases from Plasmodium falciparum. Glyoxalase I (GloI) and one of the two glyoxalases II (cGloII) were located in the cytosol of the blood stages. The second glyoxalase II (tGloII) was detected in the apicoplast pointing to alternative metabolic pathways. Using a variety of methods, cGloII was found to exist in a monomer–dimer equilibrium that might have been overlooked for homologues from other organisms and that could be of physiological importance. The compounds methyl-gerfelin and curcumin, which were previously shown to inhibit mammalian GloI, also inhibited P. falciparum GloI. Inhibition patterns were predominantly competitive but were complicated because of the two different active sites of the enzyme. This effect was neglected in previous inhibition studies of monomeric glyoxalases I, with consequences for the interpretation of inhibition constants. In summary, the present work reveals novel general glyoxalase properties that future research can build on and provides a significant advance in characterizing the glyoxalase system from P. falciparum.

Introduction

Malaria parasites have very high glycolytic fluxes allowing rapid cell division and a drastic increase of parasitemia within several days. Parasitized red blood cells, for example, were found to consume up to 75 times more glucose than uninfected erythrocytes (Sherman, 1979). The degradation of glucose via glycolysis is not a perfect process because of the spontaneous elimination of phosphate from dihydroxyacetone-phosphate or glyceraldehyde-3-phosphate. In mammals and yeast up to 0.4% of triosephosphate isomerase substrate is converted to the toxic by-product methylglyoxal (Thornalley, 1990). Accordingly, the formation of methylglyoxal is thought to be significantly increased in Plasmodium species. The compound reacts with physiological nucleophiles resulting in modified nucleic acids and altered/inactivated proteins. Methylglyoxal and other electrophilic 2-oxoaldehydes are therefore removed by the ubiquitous glyoxalase system, yielding d-lactate and other non-toxic 2-hydroxycarboxylic acids (Thornalley, 1990; Thornalley, 1996). Owing to the increased glycolytic fluxes, malaria parasites are predicted to require an efficient detoxification system for harmful methylglyoxal concentrations and could be highly susceptible to glyoxalase inhibition (Vander Jagt et al., 1990; Thornalley, 1990; Akoachere et al., 2005). Indeed, the following studies support the theory that this metabolic pathway could be suited as a drug target.

Using parasite extracts and recombinant enzymes, the human parasite Plasmodium falciparum was shown to have a functional glyoxalase system consisting of the isomerase glyoxalase I (GloI), two isozymes of the thioester hydrolase glyoxalase II (cGloII and tGloII) and glutathione (GSH) as a coenzyme (Vander Jagt et al., 1990; Iozef et al., 2003; Akoachere et al., 2005; Deponte et al., 2007; Urscher and Deponte, 2009). Probably because of this efficient detoxification system, d-lactate formation rates from P. falciparum-infected erythrocytes were found to be up to 30 times greater than from non-parasitized red blood cells. Complete inhibition of GloI for 24 h was previously estimated to translate into a methylglyoxal concentration of 0.3 M in the infected erythrocyte (Vander Jagt et al., 1990). Various glutathione-derived glyoxalase-substrate analogues inhibited recombinant glyoxalases, and the growth of P. falciparum blood stages (Thornalley et al., 1994; Akoachere et al., 2005).

The compounds methyl-gerfelin (MGFN) and curcumin were recently shown to be non-glutathione inhibitors of small homodimeric mammalian GloI: MGFN and the natural compound gerfelin (GFN) from the fungus Beauveria felina inhibited mouse GloI in a competitive manner with a Kiapp of ∼0.2 µM (Kawatani et al., 2008). Curcumin from the plant Curcuma longa is discussed to interact with more than 30 proteins resulting in antioxidant, antimicrobial, anti-inflammatory, anti-proliferative and pro-apoptotic effects (Aggarwal and Sung, 2009). Curcumin also inhibited human GloI with a Kiapp of ∼5 µM (Santel et al., 2008). An anti-malarial activity of curcumin was previously shown for P. falciparum cell cultures and for the rodent parasite Plasmodium berghei in vivo (Reddy et al., 2005; Nandakumar et al., 2006; Cui et al., 2007), although these effects were not assigned to the glyoxalase system.

Here we studied the glyoxalases from P. falciparum in cell culture and in recombinant form. We determined the cellular localization of GloI, cGloII and tGloII and analysed the formation of homo- and hetero-oligomers for GloI and cGloII. In addition, we asked if, and to what extend, MGFN inhibits GloI and the blood stages of P. falciparum and compared the effects with curcumin. We revealed an apicoplast localization for tGloII whereas GloI and cGloII were found to form a cytosolic glyoxalase system. The cytosolic enzyme couple did not seem to interact but cGloII occurred in an unexpected monomer–dimer equilibrium. Furthermore, growth of P. falciparum blood stages was inhibited by MGFN, and curcumin and MGFN were both competitive inhibitors of GloI but interacted in a different manner with the two active sites of the monomeric enzyme.

Results

Subcellular localization of the three glyoxalases

GloI and cGloII both lack any obvious protein targeting signals and are thought to be cytosolic proteins. tGloII was previously suggested to contain a targeting signal based on in silico analyses (Akoachere et al., 2005). Using different programmes we got rather low probability scores for predicted targeting sequences of tGloII. For example, a value of only 0.26 (out of 1.0) was obtained with PATS 1.2.1 (Zuegge et al., 2001) for the prediction of a so-called bipartite topogenic signal required for apicoplast import. To investigate the true localization of the glyoxalases we expressed GloI, cGloII and tGloII as green fluorescent protein (GFP) fusion proteins in blood stage parasites. Following transfection and selection with 2.5 nM WR99210, a drug-resistant parasite population was obtained after 19–25 days (hereafter referred to as 3D7GloIG, 3D7cGloIIG and 3D7tGloIIG). Epifluorescence microscopy of erythrocytes infected with 3D7GloIG or 3D7cGloIIG revealed GFP fluorescence only within the body of the parasite (Fig. 1A). In contrast, red blood cells infected with 3D7tGloIIG showed GFP fluorescence in a pattern reminiscent of an apicoplast localization (Fig. 1B). This was especially clear in merozoite stage parasites, where each individual merozoite contains one fluorescent ‘dot’ (Fig. 1B, lower panels). To verify this, we carried out immunolocalization studies with antibodies raised against the acyl carrier protein (ACP), an apicoplast marker (Tonkin et al., 2004). Indeed, we observed a high degree of colocalization between the GFP and ACP signal in both trophozoite (Fig. 1C, upper panels) and schizont stage parasites (Fig. 1C, lower panels), confirming that tGloII-GFP is transported to the apicoplast. In some erythrocytes infected with 3D7tGloIIG we also observed a weaker fluorescent signal surrounding the body of the parasite (white arrow in Fig. 1B), probably representing the parasitophorous vacuole. This may be the result of overexpression of the GFP fusion protein, causing overloading of the apicoplast import pathway and thus allowing a portion of our chimeric protein to follow the default secretory pathway to the lumen of the parasitophorous vacuole. Western blot analysis of protein fractions derived from 3D7GloIG, 3D7cGloIIG and 3D7tGloIIG infected erythrocytes showed that all these transgenic parasites expressed GFP-chimera proteins of the predicted molecular mass (Fig. 1D). For tGloII-GFP an additional lower band was observed that probably reflects the imported mature protein following cleavage of both signal and apicoplast transit peptide. Additionally a ∼26 kDa band could be visualized, probably representing a commonly observed degradation of GFP fusion proteins (Waller et al., 2000).

Figure 1.

Localization of GloI, cGloII and tGloII GFP fusion constructs in infected erythrocytes.
A. Live cell imaging of erythrocytes infected with transgenic parasite lines 3D7GloIG (upper panels) and 3D7cGloIIG (lower panels). GFP fluorescence was evident only within the body of the parasite (excluding the food vacuole).
B. Live cell imaging of erythrocytes infected with 3D7tGloIIG. A distinct fluorescent ‘dot’ was seen in both trophozoite (upper panels) and merozoite (lower panels) stage parasites, indicative of an apicoplast localization. The white arrow shows additional low levels in the parasitophorous vacuole.
C. Colocalization of tGloII-GFP and the apicoplast marker ACP in fixed, immunodecorated cells. Note the high degree of overlap between tGloII-GFP and ACP.
D. Detection of fusion constructs in transgenic parasites by Western blot analysis against the GFP moiety. The calculated molecular masses for GloI-GFP, cGloII-GFP and tGloII-GFP (of 69.6, 57.7 and 65.6 kDa, respectively) were in perfect agreement with the obtained signals. Additional bands at approximately 60 and 26 kDa were observed for tGloII-GFP. The star and the GFP label probably reflect successfully imported and degraded protein subpopulations respectively.

cGloII forms transient homodimers in solution

When we analysed the apparent molecular mass of previously described cGloII glutathione-binding site mutants (Urscher and Deponte, 2009) by gel filtration chromatography, we were surprised to find lower values for all mutant enzymes than for the wild-type enzyme (Table 1; Fig. 2A). Misfolding of the proteins would have resulted in a higher apparent molecular mass and was previously excluded by CD spectroscopy (Urscher and Deponte, 2009). Plausible causes for the different apparent molecular masses were therefore: (i) altered structural flexibilities and/or (ii) altered monomer–dimer equilibria. Depending on the association and dissociation rates, different oligomerization states are sometimes not resolved by gel filtration chromatography (e.g. Deponte and Becker, 2005; Nickel et al., 2006). In order to test this hypothesis for cGloII, we performed cross-linking experiments with glutaraldehyde and disuccinimidyl suberate (DSS) (Fig. 2B). The formation of cross-linked homodimers was observed for wild-type cGloII and the mutant cGloIIR154K. Longer incubation periods with glutaraldehyde also led to the formation of tetramers. No trimers were detected suggesting that α2 but not α3 units were formed. Comparative cross-linking kinetics of cGloIIR154K and wild-type enzyme revealed a slower appearance of cross-linked cGloIIR154K dimers (Fig. 2B, right panel). Interpreting the slower appearance as a decreased subunit interaction, the result is in accordance with the lower apparent molecular mass of the mutant enzyme as determined by gel filtration chromatography (Table 1; Fig. 2A). Dimer formation was also confirmed by analytical ultracentrifugation (data not shown) and pull-down assays (Fig. 2C): protein A-tagged cGloII co-eluted with His-tagged cGloII from a Ni-NTA column and His-tagged cGloII co-eluted with protein A-tagged cGloII from an IgG-sepharose column. In the absence of the binding partner, no significant unspecific binding was observed for the negative controls. In order to quantify the cGloII subunit interaction, we performed DSS cross-linking experiments with different protein concentrations at 37°C yielding an estimation of the apparent dissociation constant (Kdapp) of ∼5 µM (Fig. 2D). In summary, cGloII is able to form dimers in solution and exists in a rapid monomer–dimer equilibrium that cannot be resolved by gel filtration chromatography.

Table 1.  Gel filtration chromatography of wild-type cGloII and mutant enzymes.
EnzymeMapp (kDa)aMapp/31.8 kDa
  • a. 

    The apparent molecular mass (Mapp) was averaged from two to four independent gel filtration experiments (e.g. Fig. 2A). The theoretical molecular mass of recombinant MRGS(H)6GS-tagged wild-type enzyme and glutathione-binding site mutants (Urscher and Deponte, 2009) is 31.8 kDa.

cGloII WT45.3 ± 2.01.43 ± 0.06
cGloIIR154K23.2 ± 1.60.73 ± 0.05
cGloIIR154M26.9 ± 8.10.85 ± 0.26
cGloIIR257Q39.0 ± 0.21.23 ± 0.01
cGloIIR257D26.1 ± 2.10.82 ± 0.07
cGloIIK260Q32.0 ± 0.21.01 ± 0.01
cGloIIK260D37.5 ± 2.11.18 ± 0.07
Figure 2.

Gel filtration chromatography, cross-linking and pull-down experiments with cGloII.
A. The apparent molecular mass of purified His-tagged wild-type (WT) enzyme and glutathione-binding site mutants (Urscher and Deponte, 2009) was compared by gel filtration chromatography. All glutathione-binding site mutants eluted at a lower apparent molecular mass than wild-type enzyme (Table 1). Representative gel filtration chromatograms of wild-type cGloII and cGloIIR154K are shown for comparison.
B. Cross-linking kinetics with glutaraldehyde and DSS were determined with wild-type enzyme and cGloIIR154K. Samples were analysed by reducing SDS-PAGE (12%) and Western blotting against the N-terminal His-tag (left panels). M, monomer; D*, dimer; T*, tetramer. Using different exposure times, signals from the Western blots were quantified and relative intensities of the dimer bands of wild-type enzyme (black bars) and cGloIIR154K (white bars) were plotted versus time (right panel).
C. Pull-down experiments with N-terminally His-tagged (His-) and protein A-tagged (pA-) cGloII. The protein A-tagged protein was co-purified by Ni-NTA affinity chromatography (left panel) and the His-tagged protein was co-purified by IgG-sepharose affinity chromatography (right panel). Samples were analysed by reducing SDS-PAGE (15%). Controls without the binding partner are shown for comparison. m, marker; f, flow-through; w, washing step; e1–e3, eluate 1–3.
D. Estimation of the dissociation constant by cross-linking of wild-type enzyme with a constant (left panels) or variable concentration (right panels) of DSS at 37°C. Western blots were quantified as in (B) and the fraction of cross-linked dimer was plotted versus the total protein concentration. The Kdapp value determined by non-linear regression was ∼5 µM.

We also analysed the potential formation of GloI-cGloII hetero-oligomers. Although both proteins co-eluted on the gel filtration column, no interaction was observed in analytical ultracentrifugation studies as well as pull-down assays using His-tagged GloI and protein A-tagged cGloII and vice versa (data not shown). Thus, direct complex formation or even substrate channelling seem to be unlikely for the cytosolic enzyme couple.

Methyl-gerfelin and curcumin as P. falciparum GloI inhibitors

The potential inhibitors curcumin and MGFN were tested with recombinant GloI. Both compounds could mimic the enediolate reaction intermediate of the hemithioacetal between methylglyoxal and GSH (Fig. 3A). We therefore analysed the steady-state kinetics of GloI at different concentrations of substrate and inhibitors. Inhibition patterns with curcumin were found to be quite complex: Lineweaver-Burk plots pointed to a competitive inhibition with a rather constant maximum reaction velocity (Fig. 3B). Indeed, curcumin inhibition became insignificant at a substrate concentration ≥ 0.25 mM (Fig. 3C). Based on the slopes of the Lineweaver-Burk plots we generated a secondary plot to determine the Ki value (Segel, 1993). In contrast to a simple competitive inhibition, the secondary plot revealed no straight line. We therefore estimated two apparent Ki values (Kiapp) of ∼25 µM and > 0.2 mM for low and high curcumin concentrations respectively (Fig. 3D). Similar values were estimated from the non-linear Dixon plots (Fig. 3E). The inhibition patterns might have been caused by different inhibitors such as curcumin and a curcumin–glutathione conjugate (Awasthi et al., 2000). We therefore compared the activity in assays where the inhibitor was either preincubated for 1 or 5 min or was added just after the enzyme. Reaction rates of the controls were identical (Fig. S1) indicating that the formation of a curcumin–glutathione conjugate is irrelevant under the chosen assay conditions. More likely, the inhibition patterns are due to the two different active sites of the monomeric enzyme (Deponte et al., 2007). Accordingly, the two Kiapp values could reflect inhibitor binding to the high- and the low-affinity binding site. The observation that the enzymatic activity was ∼50% even at low substrate and high inhibitor concentrations (Fig. 3C) can then be interpreted as preferential, almost complete inhibition of one of the active sites whereas the other reaction centre remained functional. Indeed, analysing the initial reaction velocity of the mutant enzyme GloIE345Q with 20 µM substrate in the presence of 0–100 µM curcumin did not reveal an immediate inhibition (Fig. S2). Under the same conditions the wild-type enzyme was significantly affected (Figs 3C and S2). GloIE345Q has a functional low-affinity binding site and an inactive reaction centre at the high-affinity binding site (Deponte et al., 2007). The lack of an immediate inhibition of GloIE345Q suggests that curcumin preferentially binds to the high-affinity site, which was already inactivated in the mutant but not in the wild-type enzyme (Fig. S2).

Figure 3.

GloI inhibition studies with curcumin and MGFN in vitro.
A. The chemical structures of the GloI glutathionyl-enediolate reaction intermediate, and the inhibitors curcumin and MGFN are shown for comparison.
B.–E. Summarize the steady-state kinetics of GloI with substrate in the presence of 0–100 µM curcumin.
F.–I. Summarize the kinetics with 0–10 µM MGFN. See text and Experimental procedures for further details. Data points represent averages from at least three independent transformation/expression/purification experiments.

Lineweaver-Burk plots with MGFN also suggested a competitive inhibition (Fig. 3F). The inhibitor was far more efficient than curcumin, as similar residual activities were obtained with less than a tenth of inhibitor. Moreover, a substrate concentration of 0.25 mM was insufficient to regain the maximum reaction velocity (Fig. 3G). The two Kiapp values of ∼2 and 6 µM for low and high MGFN concentrations were quite similar (Fig. 3H) suggesting that both active sites were inhibited by MGFN. The Dixon plots for MGFN were again non-linear (Fig. 3I) supporting the theory that interactions between the inhibitor and the different active sites and/or conformations of the monomeric enzyme are distinguishable. In summary, MGFN is a more potent competitive inhibitor of P. falciparum GloI than curcumin because of a more than ten times lower Kiapp value and because of the inhibition of both active sites.

We also tested MGFN in parasite cell culture (Fig. 4). Ring stage parasites seemed to be more susceptible to inhibition by MGFN than asynchronous cultures as reflected by IC50 values of ∼10 and ∼20 µM respectively (Fig. 4A). Only late trophozoites and schizonts were found in the asynchronous and in the synchronized ring stage cultures after 48 h treatment with MGFN at concentrations ≥ IC50 value (Fig. 4B, upper panels). The remaining cells in the treated cultures tended to have an enlarged food vacuole and to show morphological markers that are typical for dying parasites (Deponte and Becker, 2004). In contrast, the untreated asynchronous control contained approximately 50% ring stages/early trophozoites (Fig. 4B, lower panels) supporting the hypothesis that MGFN had a pronounced effect on ring stage parasites.

Figure 4.

MGFN inhibition studies in red blood cell cultures.
A. IC50 values of ∼10 and 20 µM MGFN were determined in an LDH assay using synchronous ring stage parasites (closed circles) and asynchronous parasites (open circles) respectively. Data points from the LDH assay represent averages from triplicate measurements (upper panel). Similar results were obtained by counting 1000–2000 erythrocytes per Giemsa-stained blood smear (lower panel).
B. Representative Giemsa-stained cells from the asynchronous culture treated with 25 µM MGFN and the control. No ring stages were detected in the treated culture after 48 h in contrast to the control.

Discussion

A functional cytosolic glyoxalase system and the first plastid GloII

Because of the high glycolytic fluxes in malaria parasites, it is not surprising to find a functional glyoxalase system in the cytosol (Fig. 1). In contrast, the physiological function of the apicoplast enzyme tGloII is puzzling. On the one hand, Fleige et al. showed that apicoplasts harbour glycolytic isozymes including triosephosphate isomerase (Fleige et al., 2007). Thus, it is likely that methylglyoxal is also formed in these organelles. On the other hand, the second glyoxalase system seems to be incomplete: a functional targeted GloI could not be identified in apicomplexan parasites yet. The hitherto only candidate, the protein GILP form P. falciparum (PlasmoDB annotation PFF0230c), was inactive in vitro, probably because some active site residues are missing (Akoachere et al., 2005). Accordingly, utilization of other substrates by cGloII cannot be excluded, also because it is well known that glyoxalases II can hydrolyse a variety of thioesters (Uotila, 1973; Wendler et al., 2009). Moreover, even the source of reduced glutathione in the apicoplast is unclear. Is there a targeted glutathione reductase or are there glutathione carriers in the apicoplast?

To our knowledge all non-cytosolic glyoxalases II described so far were found in mitochondria (Bito et al., 1997; Cordell et al., 2004; Marasinghe et al., 2005; Limphong et al., 2009). Thus, despite multiple studied plant isozymes (e.g. in Arabidopsis thaliana), tGloII is the first experimentally confirmed plastid GloII. Because of the evolution of the apicoplast (McFadden et al., 1996; Moore et al., 2008) it is quite likely that in the near future further plastid glyoxalases II will be identified in algae and maybe plants.

Uncovering of cGloII dimer formation and potential physiological functions

A monomer–dimer equilibrium (Fig. 2) explains several seemingly contradicting observations of previous studies and point to a physiological relevance of cGloII dimer formation: although glyoxalases II from different organisms were reported to be monomeric (Irsch and Krauth-Siegel, 2004; Marasinghe et al., 2005; Campos-Bermudez et al., 2007), subunit contact sites were observed in the crystal structures of human GloII and mitochondrial Glx2–5 from A. thaliana (Cameron et al., 1999; Marasinghe et al., 2005). The two subunits in the crystallized human enzyme had completely accessible active sites and interacted solely along helices α8 and α8′ (including residues Glu238, Arg250, Gln254 and Lys256 from both subunits) (see PDB entry 1QH5). The dimerization interface of Glx2–5 was much larger and included both substrate-binding sites. Thus, the active sites of crystallized Glx2–5 were not accessible to the substrate (see PDB entry 1XM8). Several of the contact-forming residues of Glx2–5 are conserved in P. falciparum cGloII, which might explain why the mutations have altered the monomer–dimer equilibrium (Table 1; Fig. 2A and B).

Could dimer formation also occur in vivo? The cellular concentration of GloII from most organisms is unknown. In haploid yeast, having a volume of 30–80 fl (µm3), ∼13 700 molecules of a cytosolic GloII were measured per cell (Ghaemmaghami et al., 2003; http://www.yeastgenome.org/). Dividing the molecules per volume by the Avogadro constant yields a rough estimation of a concentration around 0.3–0.8 µM for yeast GloII. Even if the concentration of cGloII in P. falciparum was lower, a Kdapp of ∼5 µM (Fig. 2D) could be sufficient for partial but significant dimerization: for example, at a protein concentration of 0.1 µM about 4% of cGloII could form a dimer in vivo. Such a concentration corresponds to only 1500 molecules of cGloII per P. falciparum schizont (having a volume of approximately 30 fl ∼ 35 h after red blood cell invasion, Saliba et al., 1998). Obviously, analysing the cellular concentration of P. falciparum cGloII and the Kdapp of cytosolic yeast GloII is required to quantify dimer formation in vivo. What might be the function of a monomer–dimer equilibrium? Kinetic studies on cGloII and its glutathione-binding site mutants did not reveal deviations from typical Michaelis-Menten kinetics (Urscher and Deponte, 2009). Subunit cooperativity during catalysis therefore seems less likely. However, considering that the active site could be completely blocked as observed for crystallized Glx2–5, a regulatory or signalling function because of dimer formation in vivo is possible.

Interpretation and misinterpretation of GloI inhibition studies

Inhibition of P. falciparum GloI by curcumin and MGFN resulted in complex kinetic patterns (Fig. 3). The in vitro formation of curcumin–glutathione conjugates at pH 7.0 is very slow (Awasthi et al., 2000) and did not influence the reaction rates (Fig. S1). Furthermore, inhibition of homodimeric mammalian GloI by curcumin or MGFN resulted in linear Dixon plots (Kawatani et al., 2008; Santel et al., 2008). Thus, the most likely explanation of the complex patterns is that curcumin and MGFN bind with different affinities to the two different active sites of the monomeric enzyme (see also Fig. S2). In addition, inhibitor binding to one reaction centre could alter the conformation and activity of the other active site (Deponte et al., 2007). These mechanistic aspects were overlooked in previous studies on monomeric glyoxalases I. Thus, Kiapp values obtained, e.g. for P. falciparum GloI (Akoachere et al., 2005) or the commercially available monomeric enzyme from yeast (e.g. Hamilton and Creighton, 1992), have to be interpreted with caution because they represent unweighted averages from different active sites and/or conformations. The true Ki value for one of the active sites could be much higher/lower than the Kiapp value depending on: (i) the inhibitor (e.g. Fig. 3B–E vs. Fig. 3F–I), (ii) the way the data were analysed (e.g. Fig. 3D and H vs. Fig. 3E and I), (iii) the range of substrate and inhibitor concentrations tested and (iv) the differences between the active sites.

The IC50 value for curcumin in previous P. falciparum cell culture experiments was 5 or 25 µM depending on the study (Reddy et al., 2005; Cui et al., 2007). The first/lower Kiapp value for GloI with curcumin (Fig. 3D) is in this concentration range. However, this Ki only holds true at low substrate concentrations (Fig. 3C). Inhibition of GloI by curcumin in vivo will therefore highly depend on the unknown cellular concentration of the physiological substrate and on whether the targeted active site is required for the turnover of this substrate. Because curcumin is known to have many effects in the micromolar concentration range (e.g. Cui et al., 2007), it seems unlikely that the IC50 value in cell culture is solely due to a direct inhibition of GloI. For example, the homodimeric human enzyme was inhibited with a Ki of ∼5 µM (Santel et al., 2008), which might be also detrimental for P. falciparum in the erythrocyte/parasite unit. MGFN has a similar IC50 value in cell culture (Fig. 4) but is a more potent inhibitor of GloI because: (i) it competes with both active sites and (ii) has a more than 10-fold lower Kiapp value in vitro (Fig. 3H). Both compounds might be a good starting point for drug development of non-glutathione inhibitors because of their similarity with the enediolate transition state. Hydrolysis of the methoxy groups and reducing the size of curcumin (Fig. 3A) might generate a more efficient inhibitor competing with both active sites; a prerequisite that is already met by MGFN.

Experimental procedures

Cloning of GFP- and protein A-tagged glyoxalase constructs

All primers used for the following cloning procedures are given in Table S1. Vectors pHH2 and pARL were a kind gift by Professor A Cowman and colleagues (a summary on the vectors is given in Crabb et al., 2004). The glyoxalase-encoding sequences were identical to accession numbers AF486284, AY494055 and AF486285 (Akoachere et al., 2005). For cloning of constructs encoding C-terminally GFP-tagged full-length glyoxalases the genes GLOI, cGLOII and tGLOII were PCR-amplified using the templates GLOI/pQE30 (Deponte et al., 2007), cGLOII/pQE30 (Urscher and Deponte, 2009) and P. falciparum genomic DNA respectively. PCR products were digested with BamHI and AvrII and cloned into GFP/pHH2 using the BglII and AvrII restriction sites. GFP fusion constructs were subcloned into pARL-1a with XhoI. Constructs encoding N-terminally protein A-tagged full-length GloI and cGloII without His-tag were cloned as follows. The protein A gene (pA) was PCR-amplified and cloned into the NcoI/BamHI restriction sites of pET28a (Novagen) replacing the N-terminal His-tag and the T7-tag. Untagged GLOI and cGLOII were then subcloned from pQE30 into pA/pET28 using BamHI/HindIII and BamHI/SacI, respectively, yielding pA-GLOI/pET28 and pA-cGLOII/pET28. The orientation and correct sequences of all constructs were confirmed by DNA sequencing both strands.

P. falciparum cell culture and transfection

The P. falciparum 3D7 line was cultured in human 0+ erythrocytes according to standard protocols (Trager and Jensen, 1976), except cultures were incubated in gassed flasks. Transfection with the pARL constructs (see above) was carried out by electroporation of infected human 0+ erythrocytes as previously described (Fidock and Wellems, 1997). GFP-transfectants were selected with 2.5 nM WR99210 (kindly supplied by D Jacobus).

Immunofluorescence assays and live cell imaging

Immunofluorescence assays were carried out following fixation using 4% paraformaldehyde/0.00075% glutaraldehyde as previously described (Tonkin et al., 2004) except fixation was carried out at 37°C for 30 min, and quenching was performed with 125 mM glycine in phosphate-buffered saline (PBS). Primary antibodies used: rabbit anti-ACP (1:500, kindly provided by Professor G McFadden) and anti-rabbit-Cy3 (1:2000, DAKO) were both diluted in 3% BSA/PBS. Hoechst 33258 (Molecular probes) was used in a concentration of 50 ng ml−1 for fixed parasites or 10 µg·ml−1 for live parasites. All images were acquired at either 37°C (live cells) or room temperature (fixed cells) on a Zeiss Cell Observer using appropriate filter sets. Individual images were imported into Image J64 (version 1.43b, available at http://rsb.info.nih.gov/ij), converted to eight-bit greyscale, subjected to background subtraction and overlaid. To create figures, TIF files were imported into Powerpoint (Microsoft), assembled and slides exported as TIFs. No gamma adjustments were applied to any image, and all data are presented in accordance with the recommendations of Rossner and Yamada (Rossner and Yamada 2004).

Heterologous expression and protein purification

His-tagged GloI, GloIE345Q and cGloII were expressed in Escherichia coli strain XL1-Blue and purified as previously described (Deponte et al., 2007; Urscher and Deponte, 2009). Plasmids pA-GLOI/pET28 and pA-cGLOII/pET28 were freshly transformed into E. coli strain BL21(DE3) (Novagen) before each expression. Bacteria were precultured over night at 37°C in Luria Bertani medium supplemented with 25 mg l−1 kanamycin. The preculture was diluted (ratio 1:20) and grown at 37°C to an optical density at 600 nm of 0.3. Then ZnSO4*7H2O (286 mg l−1) was added and the culture was grown for 30 min before induction with 0.5 mM isopropyl-β-d-1-thiogalactopyranoside. Cells were harvested 4 h after induction by centrifugation (15 min, 4000 g, 4°C), washed once and stored at −20°C.

Bacteria expressing protein A-tagged GloI were thawed on ice and resuspended in 5 ml of 10 mM MOPS/NaOH, pH 7.8 per litre of culture. The suspension was stirred for 1 h on ice after addition of lysozyme and DNase I, followed by sonication at 4°C and centrifugation for 30 min at 30 500 g, 4°C. The supernatant was applied to an S-hexylglutathione agarose column (Sigma), which had been equilibrated with the same buffer. The column was washed with 10 column volumes of 10 mM MOPS/NaOH, 200 mM NaCl, pH 7.8, and recombinant enzyme was eluted with 4.5 column volumes of 10 mM MOPS/NaOH, 200 mM NaCl, 5 mM S-hexylglutathione, pH 7.8. S-hexylglutathione was subsequently removed by gel filtration chromatography.

Bacteria expressing protein A-tagged cGloII were thawed on ice and resuspended in 5 ml of 50 mM MOPS/NaOH, 300 mM NaCl, 0.05% (v/v) Tween 20, pH 7.8 (MST-buffer) per litre of culture. The suspension was stirred for 1 h on ice after addition of lysoszyme and DNase I, followed by sonication at 4°C and centrifugation for 30 min at 30 500 g, 4°C. The supernatant was applied to an IgG-sepharose 6 Fast Flow column (GE Healthcare), which had been equilibrated with MST-buffer. The column was then washed with 10 column volumes of MST-buffer. Afterwards, the column was incubated for 30 min at 4°C with 12 column volumes of 50 mM MOPS, 300 mM NaCl, 125 mM imidazole, pH 7.8 containing an excess of purified His-tagged protein A in order to elute the glyoxalase in a competitive manner. The eluate was diluted 1:10 and applied on a Ni-NTA column (Quiagen) to remove His-tagged protein A. Protein A-tagged cGloII was collected from the flow through and His-tagged protein A was reused after elution with 50 mM MOPS, 300 mM NaCl, 125 mM imidazole, pH 7.8.

Gel filtration chromatography

The apparent molecular mass of the purified proteins was analysed by gel filtration chromatography on a HiLoad 16/60 Superdex 200 prep grade column, which was connected to an ÄKTA-FPLC system (GE Healthcare). The column was calibrated at 4°C with a gel filtration standard (GE Healthcare) and equilibrated with buffer containing 50 mM MOPS/NaOH, 300 mM NaCl, pH 7.2. The same buffer was used for elution. FPLC-fractions were detected photometrically, and peak areas and kAV values were evaluated using the software UNICORN 3.21 (GE Healthcare). Controls with alternative pH values, protein and salt concentrations were performed to validate the obtained results.

Cross-linking experiments

Comparative cross-linking kinetics were carried out with 5–8 µM purified cGloII or cGloIIR154K (Urscher and Deponte, 2009) in 50 mM MOPS/NaOH, 300 mM NaCl, 125 mM imidazole, pH 7.8 at 4°C. Equal concentrations of both proteins were incubated in parallel with a 103-fold molar excess of glutaraldehyde (Sigma) or a 130-fold molar excess of DSS (Thermo Scientific) in a final volume of 50 µl for the indicated time points. Cross-linking was stopped by adding 5 µl of 1 M Tris-HCl, pH 8.0 to the reaction mixture. After addition of Laemmli-buffer, samples were analysed by reducing 12% SDS-polyacrylamide gel electrophoresis (SDS-PAGE) (Laemmli, 1970) and Western blotting against the N-terminal His-tag using a mouse penta-His primary antibody (Qiagen) and an anti-mouse IgG secondary antibody conjugated with horseradish peroxidase (Bio-Rad). Experiments with glutaraldehyde were repeated twice yielding similar results. Signals from three independent DSS cross-linking/Western blotting experiments were quantified using the software ImageMaster 1D Elite (Amersham Biosciences). The maximum wild-type signal was set to 100%, and results were expressed as a percentage of this signal. The data shown in Fig. 2B were averaged from the three experiments.

In analogy, a Kdapp was estimated by cross-linking with DSS. Variable concentrations of purified cGloII were incubated for 10 min at 37°C with a 130-fold molar excess or a constant concentration (130 µM) of DSS in a final volume of 50 µl. A simplified reaction scheme for dimer (D) cross-linking with DSS is:

image

Although the exact kinetic law for the formation of the cross-linked dimer (D*) with the velocity V2 is unknown, and although D* instead of D is measured by SDS-PAGE and western-blotting, an estimation of Kd = k−1/k1 = [M]2/[D] is possible (Graziano et al., 2006). The signal intensities for total protein and D* were determined from the blots, and the data were analysed as described in Graziano et al. 2006 by non-linear regression analysis using SigmaPlot 10.0 (Systat Software). Additional attempts to confirm the Kd by surface plasmon resonance failed because of unspecific binding of His- and protein A-tagged cGloII to a variety of surfaces.

Pull-down assays with His- and protein A-tagged glyoxalases

Co-purification on IgG-sepharose: frozen bacterial cell pellets containing the His- or protein A-tagged glyoxalases were solubilized in 5 ml of MST-buffer per litre of culture. The suspension was stirred for 1 h on ice after addition of lysozyme and DNase I, followed by sonication at 4°C and centrifugation for 30 min at 30 500 g, 4°C. The supernatants with each fusion protein were subsequently mixed, incubated for 10 min on ice and applied to an IgG-sepharose column that had been equilibrated twice with 2–3 column volumes each of 0.5 M acetic acid, pH 3.4, and MST-buffer. The column was then washed with 10 volumes of MST-buffer, followed by a washing step with two volumes of 5 mM ammonium acetate, pH 5.0. The protein A-tagged glyoxalase was eluted with 0.5 M acetic acid, pH 3.4, and the fractions were analysed for co-purified His-tagged protein by reducing 15% SDS-PAGE.

Co-purification on Ni-NTA: frozen bacterial cell pellets containing the His- or protein A-tagged glyoxalases were solubilized in 5 ml of 50 mM MOPS, 300 mM NaCl, 10 mM imidazole, pH 7.8 per litre of culture. The suspension was stirred for 1 h on ice after addition of lysozyme and DNase I, followed by sonication at 4°C and centrifugation for 30 min at 30 500 g, 4°C. The supernatants were subsequently mixed, incubated for 10 min on ice and applied to a Ni-NTA column that had been equilibrated with the same buffer. The column was washed with 10 volumes of 50 mM MOPS, 300 mM NaCl, 25 mM imidazole, pH 7.8. The His-tagged protein was eluted with 50 mM MOPS/NaOH, 300 mM NaCl, 125 mM imidazole, pH 7.8, and the fractions were analysed for co-purified protein A-tagged glyoxalase by reducing 15% SDS-PAGE.

MGFN and curcumin inhibition assays

High-performance liquid chromatography-purified curcumin (> 98%) was obtained from Roth. GFN was purified as described previously (Zenitani et al., 2003a,b). Trimethylsilyl-diazomethane (2 M) was added to a solution of GFN (0.34 mmol) in methylene chloride (3 ml) and methanol (1 ml), and the mixture was subsequently stirred for 30 min at 0°C. The reaction mixture was extracted with ethyl acetate and washed with brine. The resulting extracts were further purified by reverse-phase high-performance liquid chromatography to yield MGFN as a brown powder.

Enzymatic activity of recombinant GloI was monitored at 30°C using a thermostatted Jasco V-550 UV-visual spectrophotometer (see also Deponte et al., 2007). GloI activity was determined at 240 nm by measuring the conversion of the hemithioacetal (formed between methylglyoxal and GSH) into S-d-lactoylglutathione (ε240 nm = 3.37 mM−1 cm−1). Stock solutions of 10 mM GSH (Sigma) and 100 mM methylglyoxal (Sigma) in assay buffer (50 mM MOPS/NaOH pH 7.0), and ≤ 10 mM curcumin and 25 mM MGFN in methanol were freshly prepared before each experiment. For a desired concentration of hemithioacetal, the concentrations of methylglyoxal and GSH in the equilibrated assay mixture were calculated and varied using the equation Kd = 3 mM = ([methylglyoxal][GSH]/[hemithioacetal]). The calculated concentration of free GSH at equilibrium was 0.1 mM in all assays and the calculated initial concentration of the hemithioacetal at equilibrium was 5–500 µM. Accordingly, the concentrations of methylglyoxal and GSH before equilibration were 0.155–15.5 mM and 0.105–0.6 mM respectively. The hemithioacetal was formed during a preincubation period of 5 min at 30°C (longer preincubation for 10 min led to very similar reaction rates). The inhibitor was added after 4 min during the 5 min preincubation period, and the reaction was initiated by the addition of enzyme (5 nM in all assays). The highest final concentration of methanol in the assay was 1% for curcumin and 0.5% for MGFN. Controls confirmed that the activity of GloI was not significantly affected by methanol under the chosen assay conditions.

Inhibition of parasite growth by MGFN was determined by counting Giemsa-stained blood smears and by a photometric lactate dehydrogenase (LDH) assay as previously described (Makler et al., 1993). MGFN (20 mM, dissolved in methanol) was diluted stepwise in the cell culture medium. Afterwards, asynchronous parasite cultures or synchronized ring stage parasite cultures were added to the medium resulting in a final MGFN concentration of 0.05–100 µM (0.0002–0.5% methanol). Cultures were grown for 48 h before preparation of blood smears and for 72 h before readout of the LDH assay. Pictures of Giemsa-stained parasites were taken on a Zeiss Axiovert 200 and assembled in CorelDraw 12. The overall brightness and contrast of the whole image was adjusted in order to highlight the Giemsa-stained parasites. Results from the LDH assay were corrected for background (erythrocytes only) and the data from mock treated parasite cultures (containing the corresponding concentration of methanol) were set to 100%. Results were expressed as a percentage of the mock treated control.

Acknowledgements

This work was supported by the Deutsche Forschungsgemeinschaft (DFG) grant De1431/1 and in part by a travel allowance of the Boehringer Ingelheim fonds (to MD). JMP is supported by the DFG collaborative research centre (SFB) TR1. We wish to thank Stephan Uebel for analytical ultracentrifugation, Nadja Braun for help with the LDH assays and Petra Heckmeyer for her excellent technical assistance. MD is especially grateful to Professor Alan Cowman and colleagues for support and discussions during a vocational training at the Walter and Eliza Hall Institute.

Ancillary