Poisoning Pyridoxal 5-Phosphate-Dependent Enzymes: A New Strategy to Target the Malaria Parasite Plasmodium falciparum

The human malaria parasite Plasmodium falciparum is able to synthesize de novo pyridoxal 5-phosphate (PLP), a crucial cofactor, during erythrocytic schizogony. However, the parasite possesses additionally a pyridoxine/pyridoxal kinase (PdxK) to activate B6 vitamers salvaged from the host. We describe a strategy whereby synthetic pyridoxyl-amino acid adducts are channelled into the parasite. Trapped upon phosphorylation by the plasmodial PdxK, these compounds block PLP-dependent enzymes and thus impair the growth of P. falciparum. The novel compound PT3, a cyclic pyridoxyl-tryptophan methyl ester, inhibited the proliferation of Plasmodium very efficiently (IC50-value of 14 µM) without harming human cells. The non-cyclic pyridoxyl-tryptophan methyl ester PT5 and the pyridoxyl-histidine methyl ester PHME were at least one order of magnitude less effective or completely ineffective in the case of the latter. Modeling in silico indicates that the phosphorylated forms of PT3 and PT5 fit well into the PLP-binding site of plasmodial ornithine decarboxylase (PfODC), the key enzyme of polyamine synthesis, consistent with the ability to abolish ODC activity in vitro. Furthermore, the antiplasmodial effect of PT3 is directly linked to the capability of Plasmodium to trap this pyridoxyl analog, as shown by an increased sensitivity of parasites overexpressing PfPdxK in their cytosol, as visualized by GFP fluorescence.


Introduction
Malaria is one of the most threatening diseases in the world, with more than 300 million infected people and up to 2 million fatalities annually, mainly young children and pregnant women living in Africa. The causative agent of the most severe form, tropical malaria, is Plasmodium falciparum, a protozoan parasite that sequesters in the red blood cells of its human host. As yet, there are no vaccines available and the few antimalarials are losing their efficacy due to the rapid spread of drug-resistant parasites. There is therefore an urgent need to find new strategies to interfere with the parasite's metabolism.
Pyridoxal 5-phosphate (PLP) is the active form of vitamin B6, which comprises the vitamers pyridoxine, pyridoxal and pyridoxamine as well as their related phosphate esters. It is a crucial cofactor for various enzymatic reactions, predominantly involved in the transformation of amino acids [1]. Many of the PLP-dependent enzymes are essential and therein offer the possibility to target diseases [2]. There are already promising attempts, such as the inhibition of c-aminobutyric acid (GABA) aminotransferase by vigabatrin in epilepsy therapy [3] or of alanine racemase by antibacterial agents [4,5]. The PLP-dependent enzyme ornithine decarboxylase (ODC) is a target in the fight against African trypanosomes, the causative agent of sleeping sickness, and against cancer [2,6,7]. ODC is the key enzyme in the formation of polyamines, which are essential for cellular proliferation and differentiation [8,9], and accordingly, polyamine biosynthesis in the malaria parasite is also considered to be a possible drug target [10].
The reactions of PLP-dependent enzymes are highly diverse; however the initial formation of a Schiff base of PLP with the amino group of an amino acid is a common intermediate to all of them. Reduction of the Schiff base leads to a covalent coenzyme-substrate adduct resembling a transition-state analog with high affinity to the apoprotein [11]. Over thirty years ago, Heller et al. [12] synthesized several phosphopyridoxyl-amino acids that mimic the Schiff base intermediate involved in enzymatic decarboxylation. Among these, N-(5-phosphopyridoxyl)-ornithine inhibited rat liver ODC by binding to the holoenzyme in a non-competitive manner with respect to substrate and cofactor. Recently, a new analog, the BOCprotected pyridoxyl-ornithine conjugate (POB), was shown to inhibit human ODC in cells as well as the growth of various tumor and transformed cells effectively, whereas non-tumorigenic cells were less affected [13]. Another analog, pyridoxyl-histidine methyl ester (PHME), was able to inhibit human histidine decarboxylase in human mastocytoma cells without affecting proliferation [14].
Plants, bacteria, as well as some fungi are able to synthesize PLP de novo, whereas mammals entirely depend on the dietary uptake of pyridoxal, pyridoxine and pyridoxamine. They are converted into PLP via phosphorylation by pyridoxal/pyridoxine kinase (PdxK; EC 2.7.1.35) and subsequent oxidation by pyridoxine/pyridoxamine oxidase (EC 1.4.3.5). Although PdxK is widely distributed in human tissues, oxidase activity is low or absent in most cells [15]. Human erythrocytes have both enzymes and are proposed to be the main carrier of vitamin B6. PLP in red blood cells is mostly bound to proteins. To provide tissues with pyridoxal, PLP is dephosphorylated by vitamin B6 specific phosphatase within the erythrocyte [16]. The plasmodial parasite generates vitamin B6 de novo in its erythrocytic stage [17,18]. The assembly of the plasmodial PLP-synthesis machinery follows a hierarchically defined course, which seems to be essential for activity [19]. In addition, the parasite possesses a functional PdxK [17], which in other protozoan parasites that lack de novo synthesis, such as trypanosomes, is the only enzyme activating salvaged pyridoxal [20].
In this report we present a new approach to target the malaria parasite by designing and testing novel pyridoxyl-adducts as antimalarials. In a first step these compounds are trapped upon phosphorylation in the parasite. In this state they mimic PLP and together with the substrate moiety they bind and inhibit efficiently PLP-dependent enzymes, which eventually kills the parasite. We validate their potency as substrates of PfPdxK as well as their inhibitory effect on the PLP-dependent enzyme PfODC, and finally we demonstrate that the antiplasmodial effect can be directly linked to the activity of the plasmodial PdxK as transgenic parasites overexpressing PfPdxK exhibit an increased sensitivity towards the pyridoxyl-analog.

Pyridoxyl-amino acids are substrates of the plasmodial PdxK
Our strategy is based on the fact that pyridoxal is phosphorylated by the plasmodial PdxK and the resulting PLP is thereby trapped within the parasite. In this respect the non-phosphorylated pyridoxyl-adducts PT3, PT5 and PHME (shown in Figure 1; their synthesis and that of their phosphorylated forms is described in the section Supporting information, Text S1) were tested as potential substrates of PfPdxK. All three pyridoxyl-adducts were phosphorylated by the plasmodial enzyme with specific activities of 38, 20 and 64 nmol min 21 mg 21 , respectively (Table 1).

Phosphorylated pyridoxyl-amino acids inhibit PfODC in vitro
Phosphorylation of pyridoxal is required in order to obtain the active cofactor for the enzymatic reaction of PLP-dependent enzymes. The activity of ODC depends on PLP and its competition by the pyridoxyl-adducts was measured as a representative for PLP-dependent enzymes. The non-phosphorylated pyridoxyl-adducts, PHME, PT5 and PT3, did not inhibit the plasmodial ODC activity in the standard in vitro assay (Table 1). Only the phosphopyridoxyl-adducts PPHME and PPT5 act as inhibitors on the plasmodial ODC with IC 50 -values of 58 mM and 64 mM, respectively, as shown in Table 1. Since PPT3 was not available its inhibitory potential was analyzed by in silico modeling.

Structural analysis of PfODC binding to PPT3 and PPT5
The active species of PT3 and PT5 in cells are presumed to be their phosphorylated forms. To study the binding mode in the active site of plasmodial ODC and to elucidate binding differences, PPT3 and PPT5 were modeled into the active site of PfODC ( Figure 2). The 5-phosphate-pyridoxyl motif forms hydrogen bonds with Ser868B, Met967B, His998B and Tyr1384B in both compounds. The side chains of both PPT3 and PPT5 occupy the pocket formed by Asp887B, Lys868B, Cys1355A, Asn1393A, Arg1117B, Asp1320B, Asp1356A and the hydrophobic residues Tyr1384B, Tyr966B, Phe1392A. The indole ring of PPT3 and PPT5 is engaged in hydrophobic interactions with Tyr1384B. The N atom of the indole ring of the cyclic PPT3 can

PT3 exhibits distinct antiplasmodial activity
To analyze the potential of the pyridoxyl-amino acid conjugates in blocking the proliferation of cultured P. falciparum, the parasites were treated with the non-phosphorylated forms. Additionally, deoxypyridoxine, known to be effective against P. lophorae [21], was tested as a control. While deoxypyridoxine and PHME had no effect on parasite growth, PT3 inhibited the proliferation of cultured Plasmodium with an IC 50 -value of 14 mM ( Figure 3). PT5 also showed antiplasmodial activity, however, the apparent IC 50value was more than one order of magnitude higher. The phosphorylated compounds did not inhibit plasmodial growth at all (data not shown). Interestingly, the effect of PT3 on the growth of mammalian cells was marginal. At a concentration of 50 mM PT3 the growth was decreased to about 75% of the untreated control. Treating the cells with 100 mM PT3 only slightly increased inhibition further ( Table 2).

Localization of the PfPdxK
In order to determine the localization of the plasmodial PdxK, the open reading frame of gfp was fused to the 39-end of pfpdxk resulting in a PfPdxK-GFP chimera. Three weeks post-transfection, cells resistant to the selection drug WR99210 were obtained and successful transfection of the construct was confirmed by plasmid rescue (data not shown). The PfPdxK-GFP chimera

P. falciparum overexpressing PfPdxK is more sensitive towards PT3
Since PfPdxK presumably activates the prodrug PT3, thereby triggering inhibition of parasite growth, another transgenic P. falciparum cell line was created. This cell line overexpressed PfPdxK fused to a C-terminal strep tag II encompassing ten amino acid residues (PfPdxK-strep), thus avoiding possible steric influences of the GFP protein. The expression of PfPdxK-strep was verified by northern blot analysis of total RNA ( Figure 5A). Additional to the endogenous mRNA of about 3 kb, present in wild-type and PfPdxK-strep parasites (a), a 4 kb band occurred only in the transgenic PfPdxK-strep expressing cell sample (b). The corresponding PfPdxK-strep protein was further verified by western blot analysis, using a monoclonal antibody against the fused strep tag ( Figure 5B). To analyze the effect of PT3 on the transgenic PfPdxK-overexpressing cells as well as on wild-type parasites, proliferation was monitored over a period of 6 days. The experiment clearly revealed that PfPdxK-overexpressing cells that were treated with 7 mM PT3, corresponding to half the IC 50value, exhibited a considerably increased doubling time (45.0 h, calculated as described in Materials and Methods), in comparison to untreated PfPdxK-strep cells (17.5 h). This parameter was only slightly altered for wild-type cells from 16.9 h for the untreated control to 20.7 h for the culture treated with PT3 ( Figure 6).

Discussion
The general lack of progress in malaria control depends not least on account of resistance to common antimalarials [22]. In this respect, not only the search for novel drug targets in the parasite's metabolism but also new approaches should be considered to fight the parasite.
We propose here a strategy to poison PLP-dependent enzymes in order to kill the human malaria parasite P. falciparum. PLP is a cofactor in more than 140 distinct enzymatic reactions, corresponding to approx. 4% of all classified activities [1]. Many of the PLP-dependent enzymes are essential; once blocked, the death of the parasite follows, as reported for the plasmodial ornithine decarboxylase (PfODC) [10,23]. The idea behind the proposed strategy is hence to target critical PLP-dependent enzymes of the parasite simultaneously by using pyridoxyl-adducts as prodrugs.
Wu et al. [14] showed that the pyridoxyl-histidine methyl ester (PHME) acts as a potent transition-state inhibitor of histidine decarboxylase (hHDC) in human cells. When PHME was applied to cultured P. falciparum no significant growth inhibition was observed; similar to the result obtained with the pyridoxine analog deoxypyridoxine, although the latter was previously reported to be an effective inhibitor in a different system [21]. Either these two compounds do not interfere sufficiently with the parasite's metabolism or have only a limited uptake. Since the erythrocyte, the host cell that harbors Plasmodium, is regarded as the main carrier of vitamin B6 in the human [15], the question arises whether the strategy of poisoning PLP-dependent enzymes in the parasite is at all applicable. However, it was shown that PLP is immediately bound by hemoglobin [24] and hence of limited availability for the parasite. It therefore remains to be clarified to what extent the parasite relies on salvage, given that the de novo synthesis of PLP seems to be functional in the blood stage, during which Plasmodium can survive without additional pyridoxine [25].
In this study we present the newly synthesized pyridoxyltryptophan methyl esters PT3 and PT5 (Figure 1). The cyclic PT3 exhibits an antimalarial effect in the low micromolar range, with an IC 50 -value of 14 mM, whereas the non-cyclic PT5 is one order of magnitude less effective (Figure 3). PT3 treatment of human cells revealed only a moderate effect (as shown in Table 2), suggesting that the impact of PT3 is parasite-specific. Moreover, effective inhibition of Plasmodium at the cellular level with this pyridoxyl-analog demonstrated that de novo synthesis of PLP is unable to compensate for the adverse effect of PT3 on parasite survival. While we assume that, in vivo, there is a dual provision to the parasite's PLP pool, further experiments concerning uptake and functional analyses on knock-out parasites would have to be performed to establish the contributions and relative importance of salvage and de novo synthesis.  As well as demonstrating their antimalarial effect the studies presented here show that both compounds, PT3 and PT5, are phosphorylated by the PfPdxK in vitro at levels comparable to those for the natural substrates pyridoxal and pyridoxamine (Table 1). From this we assume that the pyridoxyl-analogs are likely to be trapped in the parasite. Their trapping in the erythrocyte can be discounted since, beside a PdxK, the host also possesses pyridoxal phosphatase activities [16].
In order to examine whether the phosphorylated pyridoxylamino acid adducts are able to interact with their assumed target enzyme, the inhibitory effect on the plasmodial ODC activity was measured. As expected, only the phosphorylated analogs interfered with the decarboxylation of ornithine to putrescine and were thus able to displace PLP in PfODC ( Table 1). The fact that pyridoxyl-histidine and -tryptophan inhibit ODC is contrary to the generally accepted view that such transition state inhibitors should be designed only as coenzyme-substrate adducts.
As PPT3 was not available, its inhibitory action on plasmodial ODC could not be experimentally measured and thus its interaction was studied in silico. Molecular modeling of PPT3 and PPT5 in the active site of PfODC predicted that these would have different binding affinities to the enzyme. PPT3 forms hydrogen bonds with Asp887B and Cys1355A of PfODC, whilst PPT5 can form only one, with Asp1320 ( Figure 2). The concerted hydrogen bonding of the N atom of the indole ring of PT3 with Asp887B and Cys1355A found in the modelled complex, which results in an additional hydrogen bond in comparison to the complex with PPT5, could account for the better binding of PPT3. Whether this is the main contribution or whether additional energetically favourable interactions are present is beyond the scope of the present modelling studies. Notably, Cys1355A has been assumed to be the catalytic residue in PfODC [26], with a function comparable to Cys360 in mammalian [27] and trypanosomal ODC [28]. Cys360 modification totally abolishes ODC activity, as demonstrated with the well-known suicide inhibitor DFMO [27].
The use of pyridoxyl-adducts as competitive transition-state inhibitors has already been tried by many groups and their potency confirmed [12,29]. However, using such compounds as prodrugs against Plasmodium had not been tested so far. The strategy proposed here is highly dependent on the phosphorylation by PfPdxK to activate and trap the phosphorylated prodrug in the cell. The key role of PfPdxK could be demonstrated by overexpressing a transgenic copy of PfPdxK in the parasite. By means of a GFP chimera, a cytosolic localization of PfPdxKimportant for trapping PT3 -could be observed ( Figure 4). As GFP might disturb the enzyme activity of PdxK, a construct carrying a small strep tag (10 amino acids) instead of GFP (238 amino acids) was also generated. Expression of the PfPdxK-strep fusion product, previously shown to maintain PdxK activity [17], was subsequently confirmed by western and northern blot analyses ( Figure 5). Since in this case PfPdxK is not the target enzyme but the activator of the prodrug PT3, an increased sensitivity of transgenic PfPdxK overexpressing cells would be expected relative to wild-type. Consistent with the proposed strategy, monitoring parasite growth over several days clearly revealed that the impact of PT3 on PfPdxK overexpressing cells was considerably stronger. The growth rate of these cells was reduced by 61% relative to the untreated control ( Figure 6B), whereas the proliferation of the wild-type was only inhibited by 19% ( Figure 6B). From that we conclude that a time-dependent accumulation of PPT3 in the parasite's cytosol occurred, emphasizing the key role of PfPdxK.
Our results suggest that the pyridoxyl-tryptophan derivative enters the parasite, possibly directly via an amino acid transporter. After phosphorylation, PPT3 displaces PLP from the active site of ODC (and other PLP-dependent enzymes) or binds directly to newly synthesized apoenzymes which ultimately arrests growth of P. falciparum (Figure 7). The significant superiority of the cyclic form (PT3) compared to the open form (PT5) of pyridoxyltryptophan in inhibiting proliferation of the parasite can be explained by differences in binding of PfODC as indicated by the molecular modeling studies.
The selective inhibition profile for proliferation achieved with PT3 -IC 50 of 14 mM for P. falciparum ( Figure 3) and IC 50 .100 mM for human cells ( Table 2) -suggests that the approach of poisoning Plasmodium with PLP-mimicking compounds holds considerable promise. Further studies will be required to show whether, besides ODC, additional plasmodial enzymes are effectively targeted and to improve the prodrugs by appropriate structural modifications.

Enzyme assays
Recombinant expression using an anhydrotetracycline inducible system and affinity purification via strep tag of the plasmodial PdxK as well as the bifunctional P. falciparum AdoMetDC/ODC were carried out as described previously [30]. The enzyme assay of the PfPdxK was performed in 70 mM potassium phosphate buffer, pH 6.5, containing either 200 mM pyridoxine according to Wrenger et al. [17] or up to 200 mM of the non-phosphorylated pyridoxyl-adducts (PHME, PT5 and PT3), 5 mM MgCl 2 , 500 mM (2 mCi 33 P) ATP as well as 2.5 mg protein in a total volume of 100 ml. The reaction took place at 37uC for 1 h. The data were analyzed using GraphPad Prism 4.0. PfODC activity was assayed as described [26] by measuring the formation of 14

Molecular modeling
The model structure of PfODC together with bound PLP and DFMO [31] was obtained from the protein data bank (http://www. rcsb.org, PDB code: 1M9V). The active site model of PfODC with PPT3 and PPT5 was generated on the basis of the known 3D coordinates of the complex of PfODC with PLP-DFMO followed by energy minimization with the InsightII/Built module (steepest descents, up to a maximum derivative of 20 kcal/mol, consistent valence force field, CVFF). The structures were further optimized by energy minimization using the InsightII/Discover module (conjugation gradient, until a maximum derivative of 0.01 kcal/ mol was reached, CVFF) according to Wu et al. [14].
Inhibition assays on cultured P. falciparum and mammalian cells P. falciparum 3D7-strain was maintained in continuous culture according to Trager and Jensen as modified [23]. The parasites were grown in human erythrocytes (O+), RPMI 1640 medium supplemented with 25 mM HEPES, 20 mM sodium bicarbonate and 0.5% AlbuMAX II (Invitrogen, Germany) at 4% hematocrit. The cells were cultivated in 90 mm Petri dishes (Nunc, Denmark) and incubated at 37uC in presence of 90% N 2 , 5% O 2 and 5% CO 2 . The impact of deoxypyridoxine and the pyridoxyl-adducts PHME, PT5 and PT3 on the erythrocytic stages of P. falciparum was determined by using the [ 3 H]-hypoxanthine incorporation assay as described [23] and the IC 50 -value of PT3 was calculated from sigmoidal inhibition curves using GraphPad Prism 4.0. Additionally, 50 and 100 mM PT3 was tested against human cell lines RD and A 549 as well as the rhesus monkey cells FRHK 4 cultured in DMEM complete medium (PAA Laboratories, Austria) containing 7.5% FCS (Invitrogen) at 5% CO 2 for 48 h at 37uC in 96 well plates. Cell viability was determined by incorporation of [ 3 H]-hypoxanthine as described above using 10000 cells per well. To analyze the long term effect of PT3 on the PdxKoverexpressing and wild-type parasites, growth rates were monitored for several days by light microscopy of Giemsa-stained thin smears with a starting parasitaemia between 1 and 2%. Growth curves were generated with GraphPad Prism 4.0 using an exponential equation to calculate the slopes.

Generating transgenic parasites
To overexpress the plasmodial PdxK in the parasite as a strep tag II (strep) fusion protein or as a PfPdxK-GFP chimera, two constructs were generated carrying pfpdxk fused to C-terminal strep or GFP, respectively. The open reading frame of pfpdxk was amplified by PCR from gDNA using the Expand Long Template PCR System (Roche Molecular Biochemicals, Germany) and the specific oligonucleotides PfPdxK-pARL-kpn-S (sense) 59-GAGAGGTACCATGAAGAAG-GAAAATATTATCTCC-39 and PfPdxK-strep-avr-AS (antisense) 59-GAGACCTAGGCTATTTTTCGAACTGCGGGTGGCTCC AAGCGCTAAAAAAAACAGGCTCTTCTTTAATTAAAATA-TC-39 for the strep (underlined) fusion protein and PfPdxK-pARLavr-AS (antisense) 59-GAGACCTAGGAAAAAAAACAGG-CTCTTCTTTAATTAAAATATC-39 for the PfPdxK-GFP chimera to obtain a 1451 bp or 1484 bp fragment, respectively, comprising the full pfpdxK ORF without the stop codon and with strep, respectively. The fragments were cloned into pARL1a- [32] containing GFP with KpnI and Avr II restriction enzymes (bold) resulting in the constructs pARL-PfPdxK-strep and pARL-PfPdxK-GFP. Both constructs were confirmed by automated sequencing on an ABI 377 (Applied Biosystems Inc., Germany) and the plasmid DNA for transfection was obtained using a Maxi Prep Kit (Qiagen, Germany). Prior to transfection, parasites were synchronized with 5% sorbitol [33] and cultivated for at least one additional cycle of schizogony. Transfection was carried out by spontaneous uptake of DNA [34]. Before electroporation, 400 ml red blood cells were washed in cytomix [35] and mixed with 100 mg circular plasmid DNA dissolved in 30 ml Tris-EDTA. Electroporation was performed in a 0.2 cm gap cuvette (Biorad, Germany) at 310 V and 900 mF in an EasyJect Gene Pulser (EquiBio, Germany). Red blood cells preloaded with DNA in this way were transferred into prewarmed medium and provided with 50 ml red blood cells infected with 10% schizonts. The next day the culture medium was supplemented with 5 nM WR99210 (Jacobus Pharmaceutical Company, USA) as the selection drug. The PfPdxK-strep parasites obtained were maintained in 50 nM WR99210 to induce overexpression of the PfPdxK-strep fusion protein.

Analysis of transgenic parasites
Wild-type and PfPdxK-strep overexpressing parasites were separated from the host cell with saponin [36] and kept at 220uC for protein isolation or resuspended in Trizol (Invitrogen) for RNA extraction according to the manufacturer's instructions; 25 mg total RNA were analyzed by northern blotting [37] using a radiolabeled pfpdxk-probe. Total plasmodial protein extract was obtained by boiling the parasites at 95uC for 5 min in an SDS-and mercaptoethanol-containing buffer. Equal amounts (15 mg) were separated on 10% SDS-PAGE and subsequently transferred to a nitrocellulose membrane (Schleicher & Schüll, Germany). Expression of PfPdxK-GFP and of PfPdxK-strep was analyzed using either a polyclonal anti-GFP antibody (1:1000 and 1:2000; Invitrogen) or a monoclonal anti-strep antibody (1:5000; IBA) and HRP-coupled secondary anti-rabbit or anti-mouse antibodies (1:10000; Invitrogen, Germany), respectively. The blot was stripped and reprobed with a polyclonal antibody against plasmodial thiamine pyrophosphokinase (TPK) (1:10000; [38]) and the secondary anti-rabbit HRP coupled antibody (1:10000, Dianova, Germany). The detection was performed using ECL (Millipore, Germany). The green fluorescence of the PfPdxK-GFP chimera was detected and captured in live cells using a Zeiss Axioskop 2 and the OpenLab software (Improvision, Germany). The parasite's nuclei were visualized with 0.2 mg/ml Hoechst 33342 dye (Invitrogen, Germany).