|Home | About | Journals | Submit | Contact Us | Français|
The activities of the bifunctional folate pathway enzyme dihydrofolate synthase–folylpolyglutamate synthase from Plasmodium falciparum are characterised with respect to their kinetics, substrate specificities and responses to folate analogue inhibitors.
Unusually for a eukaryote, the malaria parasite Plasmodium falciparum expresses dihydrofolate synthase (DHFS) and folylpolyglutamate synthase (FPGS) as a single bifunctional protein. The two activities contribute to the essential pathway of folate biosynthesis and modification. The DHFS activity of recombinant PfDHFS–FPGS exhibited non-standard kinetics at high co-substrate (glutamate and ATP) concentrations, being partially inhibited by increasing concentrations of its principal substrate, dihydropteroate (DHP). Binding of DHP to the catalytic and inhibitory sites exhibited dissociation constants of 0.50 μM and 1.25 μM, respectively. DHFS activity measured under lower co-substrate concentrations, where data fitted the Michaelis–Menten equation, yielded apparent Km values of 0.88 μM for DHP, 22.8 μM for ATP and 5.97 μM for glutamate. Of the substrates tested in FPGS assays, only tetrahydrofolate (THF) was efficiently converted to polyglutamylated forms, exhibiting standard kinetics with an apparent Km of 0.96 μM; dihydrofolate, folate and the folate analogue methotrexate (MTX) were negligibly processed, emphasising the importance of the oxidation state of the pterin moiety. Moreover, MTX inhibited neither DHFS nor FPGS, even at high concentrations. Conversely, two phosphinate analogues of 7,8-dihydrofolate that mimic tetrahedral intermediates formed during DHFS- and FPGS-catalysed glutamylation were powerfully inhibitory. The Ki value of an aryl phosphinate analogue against DHFS was 0.14 μM and for an alkyl phosphinate against FPGS 0.091 μM, with each inhibitor showing a high degree of specificity. This, combined with the absence of DHFS activity in humans, suggests PfDHFS–FPGS might represent a potential new drug target in the previously validated folate pathway of P. falciparum.
The antifolates are clinically important drugs used to target both bacterial and eukaryotic pathogens, including the apicomplexan, Plasmodium falciparum (Pf), a parasite that still claims over a million lives each year and causes considerable economic loss to developing countries [1,2]. The antifolates that have been deployed against malaria target only two enzymes in the folate biosynthesis pathway of the parasite. Dihydropteroate synthetase (DHPS; EC 188.8.131.52), which catalyses the synthesis of 7,8-dihydropteroate (DHP) from 6-hydroxymethyl-7,8-dihydropterin pyrophosphate and p-aminobenzoate (pAB), is inhibited by sulfonamide and sulfone drugs, while dihydrofolate reductase (DHFR; EC 184.108.40.206), which reduces 7,8-dihydrofolate (DHF) to the active cofactor 5,6,7,8-tetrahydrofolate, is targeted principally by pyrimethamine (PYR) and cycloguanil. Inhibition of these activities, most commonly in a synergistic combination of sulfadoxine and PYR, disrupts the constant supply of tetrahydrofolate (THF) cofactors required for key 1-carbon transfer reactions, including synthesis of thymidylate, which is essential for DNA replication.
The long-established efficacy of folate metabolism as a clinical target and the widespread resistance to current antifolates and other classes of antimalarials highlights the importance of identifying other targets within the P. falciparum folate/thymidylate biosynthesis pathways, which comprise a further seven enzyme activities in addition to DHPS and DHFR . One activity of folate biosynthesis yet to be characterised in malaria parasites is dihydrofolate synthase (DHFS; EC 220.127.116.11), which adds an l-glutamate residue to the pAB component of DHP, the product of DHPS, to form DHF, the substrate of DHFR (Fig. 1a). DHFS represents a target unique to the parasite, as the human host is unable to synthesise folates and lacks this enzyme. Closely related to the activity of DHFS is that of folylpolyglutamate synthase (FPGS; EC 18.104.22.168), which adds further glutamate residues to reduced folate monoglutamates by γ-linkage (Fig. 1a), with the number of residues incorporated varying among organisms [4,5], ranging from an average of 3 in Escherichia coli  up to as many as 9 in Lactobacillus casei  and mammals . In P. falciparum, the pentaglutamate form of 5-methyltetrahydrofolate has been identified as the predominant polyglutamylated folate metabolite [8,9], demonstrating the importance of FPGS catalysis in these organisms as well. Such conjugated forms are apparently ubiquitous and the crucial role of polyglutamylation has been demonstrated by studying organisms mutated in their fpgs genes. For example, CHO cells mutant in this gene require supplementation by the end-products of folate metabolism and exhibit much reduced levels of intracellular folates, predominantly as monoglutamates [10,11]. Similarly, the MET7 gene encoding FPGS in Saccharomyces cerevisiae is essential for methionine biosynthesis and the maintenance of mitochondrial DNA . In mammals and plants several folate-dependent enzymes exhibit much higher affinity for polyglutamylated folates compared to their monoglutamylated equivalents [4,13,14]. Another function of polyglutamylation is to prevent folates from leaking through the cell membranes and sub-cellular compartments by substantially increasing the negative charge they carry [10,15,16] and in human cells, polyglutamylation by FPGS has been shown to have a critical role in the cellular retention and enzyme targeting of the major anti-cancer drug and folate analogue methotrexate (MTX) [17,18].
Some bacteria, such as Corynebacterium species , Neisseria gonorrhoeae , Mycobacterium tuberculosis  and E. coli  express bifunctional DHFS–FPGS molecules that perform both the first and subsequent additions of glutamate. In E. coli and M. tuberculosis, recent work has shown that the active centres of the DHFS and FPGS functions co-localise closely, at least for addition of the first glutamate to THF by the latter, but are not identical [21,23,24]. However, in L. casei the FPGS has no accompanying DHFS activity  and folate must be salvaged. This is also the case in mammals, including humans [26,27], where pre-formed folate is an essential nutrient. In eukaryotes that can synthesise folate de novo, such as fungi and plants, both activities are present but are usually encoded by separate genes [28,29]. However, we have demonstrated previously that a single protein from P. falciparum carries both DHFS and FPGS activities, the first example of a bifunctional enzyme of this type from a eukaryotic organism .
The critical dual role of parasite DHFS–FPGS in both the biosynthesis and modification of folates, and the absence of DHFS activity in humans suggest the possibility that parasite-specific inhibitors targeted to this molecule might be feasible and effective. We therefore undertook a detailed study of PfDHFS–FPGS with respect to its kinetic properties, substrate specificities and susceptibility to the antifolate drug MTX, as well as to novel inhibitors based on phosphinic acid analogues of folic acid.
Specialty reagents were obtained commercially as follows: l-[U-14C] glutamic acid (238 mCi/mmol), 3H-folinic acid (30 Ci/mmol) and 3H-methotrexate (31.8 Ci/mmol) (Moravek Biochemicals, Inc., California); DE81 anion-exchange chromatography paper (Whatman International Ltd., UK); DHF, THF, folinic acid and DHP (Schircks Laboratories, Jona, Switzerland); 2-mercaptoethanol, sodium hydrosulfite (dithionite), folic acid, ATP, BSA, l-glutamic acid, and dithiothreitol (DTT) (Sigma), Overnight Express™ Instant TB Medium (Merck), Ni-NTA resin (Qiagen Ltd., UK). The E. coli expression host used was BL21(DE3) (Novagen). The isolates of P. falciparum used were K1, FCB, V1/s, Fcr3, as well as the cloned line 3D7. The aryl phosphinate folate analogue, 2-[[[4-[N-[(2-amino-3,4-dihydro-4-oxo-6-pteridinyl)methyl]amino]phenyl](hydroxy)phosphinoyl]methyl]pentane-1,5-dioic acid (compound 1)  and the alkyl phosphinate folate analogue, 2-[[[3-[[4-[[(2-amino-3,4-dihydro-4-oxo-6-pteridinyl)methyl]amino]benzoyl]amino]-3-carboxypropyl]hydroxyphosphinyl]methyl]pentane-1,5-dioic acid (compound 2) [32,33] were synthesised in the laboratory of JKC.
A verified cDNA fragment encoding the entire dhfs-fpgs gene [30,34] was cloned from the K1 isolate of P. falciparum into pET22b (Novagen, UK) using the NdeI and BamHI sites and the construct transformed into E. coli BL21(DE3) host cells. Production of DHFS–FPGS was carried out by an autoinduction procedure  in Overnight Express instant medium (Novagen, UK). Incubation was at 37 °C overnight and then at 18 °C for a further 24 h. The cell pellet was resuspended in 50 mM sodium phosphate buffer, pH 8, with 300 mM NaCl at a concentration of up to 200 g wet weight/l, and processed with a high pressure cell breaker (Constant Cell Disruption System, Constant Systems Ltd., UK) at a pressure of 25,000 psi. Cell debris was removed by centrifugation and the supernatant collected. The His-tagged DHFS–FPGS was then purified on a Ni-NTA column and bound protein eluted with 250 mM imidazole. The protein was further purified on a Q-Sepharose high performance ion-exchange column, using a loading buffer of 25 mM Tris–HCl, pH 7.5 and eluting with a gradient from 0 to 1 M NaCl in the same buffer. The activity was monitored by the DHFS assay described below and the active peak area collected. Yields were ~2 mg/l of starting E. coli culture. Glycerol (50%, v/v) was added to the protein, which was stored at concentrations of 0.5–1 mg/ml at −80 °C for future use.
A standard reaction mix based on those established for bacterial and eukaryotic DHFS/FPGS assays was adopted [28,36,37] and contained 10 mM MgCl2, 5 mM DTT, 200 mM KCl, 100 mM Tris, 50 mM glycine, 1 mg/ml BSA, 45 μg/ml DHFS–FPGS protein (0.72 μM), pH 10. For the three substrates of the reaction, up to 5 mM ATP, up to 1 mM l-[U-14C] glutamate (diluted from the commercial stock with unlabelled l-glutamate to the required specific activity) and up to 100 μM DHP (for DHPS) or THF (for FPGS) were used, depending upon the purpose of the assay, as indicated in the text. The enzyme reactions were carried out at 37 °C for 60 min in triplicate and stopped by transferring reaction tubes into an ice-cold water bath. Reaction mixes were then spotted onto DE81 ion-exchange paper (pre-treated with 0.5 M EDTA, pH 8.0 and then dried), air dried and the unreacted label washed away with 125 ml of 40 mM NaCl, 10 mM Tris–HCl, pH 8.0, four times, 20 min each wash. The filter was then exposed to a storage phosphor screen overnight and the screen scanned with a quantitative imager (Typhoon Trio Variable Mode Imager, GE Healthcare, UK). The measured counts were converted to concentrations by comparison with a standard curve derived from known amounts of the radiolabelled l-glutamate subsequently used as substrate in the assays. The enzyme activity data was analysed using the Matlab software package. Regression curves were fitted using the least squares routine.
0.5 μCi/ml of either 3H-folinic acid or 3H-methotrexate at the same specific activity was used to label parallel 50 ml synchronous cultures of P. falciparum FCB strain. Label was incubated with ring-stage (6–10 h) parasites, which were harvested at the late trophozoite stage 20 h later. The labelled products were extracted by the boiling method and analysed using HPLC on a C18 column eluted with an acetonitrile gradient as described .
The matrix used was α-cyano-4-hydroxycinnamic acid, which was dissolved in acetonitrile:water (1:1, v/v), 0.1% formic acid (v/v) at a concentration of 10 mg/ml. The reaction mix was partially purified with a C18 SPE device. One μl of the eluted material in 50% acetonitrile, 0.1% formic acid was mixed with the same volume of matrix solution and 1 μl of the mix placed onto the target and allowed to air dry. Spectra were acquired with an AXIMA-CFR MALDI-ToF spectrometer (Kratos Analytical Ltd., Manchester, UK), in positive reflectron mode with a power setting of 180. The spectra were obtained by overlapping 100 profiles. Mass to charge (m/z) calibration was performed externally using the matrix peaks and folyldiglutamate as standards.
Before evaluation as possible inhibitors, the pterin rings of compounds 1 and 2 (Section 2.1) were reduced with mild sodium hydrosulfite (dithionite) treatment to provide the corresponding 7,8-dihydro derivatives [39,40], as confirmed by MALDI-ToF spectrometry of the parent and reduced compounds (negative ion mode, 1 as [M−H]− = 604.2, H2-1 as [M−H]− = 606.2; 2 as [M−H]− = 474.4, H2-2 as [M−H]− = 476.4; negligible signals for H4X forms). For this reduction, the parent compounds were first dissolved in DMSO at 5 mM then reduced in a 50% aqueous solution containing sodium hydrosulfite (20 mg/ml) and 2-mercaptoethanol (10 mM). Mixes were kept at room temperature for 80 min, and then the compounds aliquoted, analysed and either used immediately or moved to −20 °C for short-term storage. A fresh aliquot was used for each assay.
The dhfs-fpgs open reading frame obtained as cDNA from P. falciparum isolate K1  was cloned, expressed and purified as described in Section 2. The purified His-tagged protein ran on SDS-PAGE close to its predicted molecular mass of 62.4 kDa (Fig. 1b). Identity and purity were confirmed by LC–MS–MS analysis of tryptic peptides from the purified protein (Mascot score = 468, against complete UNIPROT database; 31% sequence coverage, with no detectable peptides from any other protein, including E. coli DHFS–FPGS). Gel filtration experiments in non-denaturing conditions gave an apparent molecular mass of ~90 kDa (data not shown), which although significantly higher than the calculated mass of a single chain, was too small to represent a dimer, indicating that the plasmodial protein is most likely to be monomeric, in common with all other DHFS/FPGS enzymes thus far reported from bacterial [22,25], fungal , plant  and mammalian [36,42] sources.
PfDHFS–FPGS represents a complex system where the closely related DHFS and FPGS activities each catalyse reactions in which there are three substrates, the pterin/folate moiety (DHP in the case of DHFS, THF mono- or polyglutamylated forms in the case of FPGS), ATP and glutamate. To establish initial conditions, the enzyme was tested separately for DHFS and FPGS activities using saturating concentrations of the natural substrates DHP and THF, respectively, together with ATP and glutamate. From progress curves with respect to both enzyme concentration and time, a standard reaction for both activities containing 0.72 μM protein incubated for 60 min was adopted. This permitted accumulation of maximal counts for accurate quantitation while staying well within the period of linearity (at least 90 min; data not shown). When DHP was used as substrate, no products with mass greater than that of DHF (which, because of its lability, was seen primarily as its oxidised product folic acid in the MALDI-ToF mass spectrometer, theoretical MH+ = 442.4) were detectable (Fig. 1c), demonstrating that under these reaction conditions the DHF produced by the DHFS activity is not subject to further glutamylation by the FPGS activity.
Standard kinetic experiments in which two of the three substrates were present in excess, while the concentration of the third was varied, revealed that the DHFS reaction did not follow classical Michaelis–Menten kinetics. As seen in Fig. 2a, velocity versus [S] curves where ATP was fixed at a high concentration (0.5 mM) and glutamate was also raised to ~1 mM, displayed characteristic biphasic kinetics, with fast initial rates that diminished as the concentration of DHP was increased. The activity peaked at about 1–2 μM DHP. Similar-shaped curves indicative of substrate inhibition were obtained when glutamate was fixed at concentrations >50 μM and ATP was varied, along with DHP (data not shown). However, neither of these co-substrates showed similar inhibition curves at higher concentrations when taken as the independent variable. Thus, this phenomenon is restricted to increases in DHP concentration only. Control experiments established that the observed kinetics were independent of the order of addition of enzyme and substrates to the reaction mix prior to the incubation period.
In order to understand the mechanism of the DHFS reaction, the activity data sets were fitted globally with a series of non-linear regression models, starting from the assumption that the sequence in which the substrates bind is random, rather than ordered, to form a quaternary enzyme–substrate complex (Supplementary Fig. 1), an assumption that is confirmed below. Initially, substrate inhibition with a second DHP molecule competitively binding to the enzyme was considered. However, though the modified formula (Supplementary Fig. 2, equation e) did follow the data trend, unlike the Michaelis–Menten equation (Supplementary Fig. 2, equation f), the fit was only improved slightly, mainly at very low DHP concentrations (compare Supplementary Fig. 3e and f). The experimental data showed that, although the rate of the reaction decreased with increasing DHP concentration, the inhibition was not complete, even at high levels of DHP (≥100 μM). Therefore, a partial inhibition term, V2 (Supplementary Fig. 2, equation d), was considered more appropriate. This assumes that a second binding site exists on the enzyme complex. Binding of a second molecule of DHP would change either the affinity of the enzyme complex to the normally bound substrate or the Vmax of the enzyme. Introduction of this term considerably improved the fit, especially at higher DHP concentrations (Supplementary Fig. 3d). However, all of these curves, which were derived from the above equations without considering the second and third substrates (ATP and glutamate), gave at best only imperfect fits to the data sets, suggesting that the binding of the different substrates may also exert allosteric effects on the enzyme. Of the equations that took this into account (Supplementary Fig. 2, equations a, b and c, graphed in Supplementary Fig. 3), partial substrate inhibition in the presence of all three substrates, as in equation a, gave an extremely good fit with an R2 value of 0.95 (Supplementary Fig. 3a). The next best fit (equation b), where fully competitive substrate inhibition in the presence of three substrates was modelled, gave a much poorer R2 value of ~0.6 (Supplementary Fig. 3b). Over a broad range of conditions where concentrations of ATP from 6 μM to 5 mM and of glutamate from 25 μM to 1 mM were tested in various combinations, the best-fit model, a, gave a dissociation constant, KA, for DHP of 0.50 ± 0.06 μM for the substrate binding, and a dissociation constant, KABCA, of 1.25 ± 0.37 μM for the substrate (DHP) inhibition (n = 15 with R2 > 0.90 in each case). Due to the substrate inhibition feature of the system, there is no conventional substrate Km term in this equation.
That the extent of the DHP substrate inhibition phenomenon seen above is dependent on ATP and glutamate concentrations is also clearly seen in Fig. 2b and c. When either ATP or glutamate concentrations were lowered to ≤25 μM, as in (b) and (c), respectively, such inhibition by DHP was no longer significant. Data obtained under these conditions were found to fit well to the conventional Michaelis–Menten equation, enabling the calculation of apparent Km values for the three substrates, because here the reaction was found to be first order with respect to DHP concentration. Using a non-linear fit to the classical Michaelis–Menten equation, we determined an apparent Km for DHP under these conditions to be 0.88 ± 0.26 μM, quite closely comparable to the dissociation constants calculated above for DHP binding at higher ATP/glutamate concentrations in the partial substrate inhibition model. The apparent Km values for ATP and glutamate were 22.8 ± 2.7 μM and 5.97 ± 0.50 μM, respectively.
To study the binding of the three substrates to DHFS, this activity was further investigated in a series of assays where each of the three substrates in turn was taken as the variable substrate (x-axis), maintaining the concentration of one of the other two constant and increasing the third in steps to yield double-reciprocal plots that are indicative of the enzyme mechanism. These experiments were done under the above-mentioned conditions of low substrate concentrations, where the reaction data fit to the Michaelis–Menten equation. All combinations of conditions gave rise to convergent plots rather than parallel lines, as exemplified in Supplementary Fig. 4. These data confirm that the binding of the substrates DHP, ATP and glutamate occurs sequentially as depicted in Supplementary Fig. 1, giving rise to the quaternary complex anticipated in the modelling exercise, and exclude the possibility of a ping-pong mechanism.
Similar experiments to the above were conducted using THF as substrate to assay the FPGS activity of PfDHFS–FPGS. Unlike the case for the DHFS activity, no unusual kinetics were observed in parallel assays on the same enzyme preparation, either under low or high substrate concentrations (Fig. 3). Velocity versus [S] curves fitted well to the Michaelis–Menten equation, yielding an apparent Km value of 0.96 ± 0.12 μM for THF, very similar to the KA and apparent Km values for DHP in the DHFS reactions above. This lack of substrate inhibition of the FPGS activity of PfDHFS–FPGS is in contrast to the marked substrate inhibition noted previously in studies with human FPGS [43,44].
Given the close similarity in the reactions catalysed by DHFS and FPGS, the similarity in the affinities of their substrates and data from studies of bacterial DHFS–FPGS proteins suggesting that the catalytic sites of these activities are at the very least closely linked [21,24], we measured the rate of glutamylation in the presence of equimolar amounts of both DHP and THF substrates at two concentrations differing by an order of magnitude (Fig. 4a and b). In both cases, no significant additive effect was seen in the rate of glutamate incorporation measured relative to those seen for the same concentration of each substrate alone, consistent with there being only a single site of glutamylation available for either substrate at any given time.
To establish the specificity of PfDHFS–FPGS, equal concentrations of substrates varying in the oxidation level of the pterin moiety were compared, i.e. DHP, DHF, THF and folic acid. The activities were studied under both low and high concentration conditions of the test substrates and of the co-substrates ATP and glutamate (Fig. 5a and b). Under both of these conditions, the highest level of glutamylation activity was seen on DHP, the natural substrate of DHFS, followed by THF, the natural substrate of FPGS. In contrast, both folic acid and DHF were very poor substrates for the polyglutamylation activity of the latter. Especially under low substrate conditions, incorporation of glutamate onto folic acid was insignificant. Similarly, although DHF, the product of the DHFS activity, is in the active centre after the DHFS reaction, it seemingly cannot be efficiently polyglutamylated by the FPGS activity. Thus, using the same concentrations of DHF and THF as substrates for FPGS, the enzyme produced only 21% under low substrate concentrations (25 μM of either) and 12% under high substrate concentrations (100 μM of either) of polyglutamylated forms of folate from DHF compared with that produced from THF. Although some incorporation onto DHF does occur in the above experiments, the concentrations used are artificially high in both cases—about two orders of magnitude higher than the concentration of DHF that could be produced in our assays from the same concentration of DHP via DHFS over a 1 h incubation period. This is consistent with the MALDI-ToF mass spectrometry analysis of the reaction products (Fig. 1c), where no trace of polyglutamylated products was seen when DHP was used as substrate under these conditions. These experiments confirm that the reduction status of the folate moiety is critical for the polyglutamylation reaction, with the fully reduced tetrahydrofolate much the preferred form for the plasmodial FPGS activity. This is in marked contrast to the properties of human FPGS, which is able to polyglutamylate the dihydro-form of folic acid with efficiencies comparable to that for the tetrahydrofolate form, and can also modify the fully oxidised form to a significant, albeit lesser, degree [36,42].
The above analysis showed that plasmodial FPGS is very poor at adding further glutamate residues to folic acid and thus differs importantly from human FPGS in this respect. This observation led us to test whether methotrexate (MTX) could act as a substrate for the plasmodial enzyme. MTX is a fully oxidised folic acid analogue extensively used in anti-cancer antifolate therapy, whose efficacy depends not only on its binding to the DHFR of rapidly dividing cancer cells, but also on its polyglutamylated forms binding to other enzymes in the folate pathway . The FPGS in human cells can convert MTX into such forms, which, as in the case of folates themselves, are more efficiently retained in the cell and also show markedly increased affinity to certain folate-dependent enzymes , such as thymidylate synthase , 5-amino-4-imidazolecarboxamide ribotide transformylase (AICAR formyltransferase), involved in purine biosynthesis , or the enzymes of methionine biosynthesis . The relevance to malaria is that the use of MTX has been advocated as a cheap antimalarial that could be safely used in doses far lower than necessary in cancer chemotherapy, and it has thus been envisaged that the parasite could also convert MTX into polyglutamylated forms and therefore enhance its efficacy by broadening its range of targets, as in the case of cancer cells . We therefore tested MTX in two ways. First, using radiolabelled MTX in parasite cultures in parallel with a near-equimolar concentration of radiolabelled folinic acid (5-formylTHF), we found by HPLC analysis that no polyglutamylated versions of MTX could be detected after overnight incubation (Fig. 6a). In contrast, cultures supplemented with 3H-folinic acid generated a large amount of labelled polyglutamylated folates up to the pentaglutamate level (Fig. 6b), as seen in previous reports [8,9,38]. Indeed, the conversion of 3H-folinic acid to such forms over this period was almost complete. Moreover, when the efficiency of uptake and retention of the two radiolabelled compounds was compared, MTX was found to have accumulated in the parasites to somewhat higher levels than folinic acid (Fig. 6c), showing that it is not the transport into the cell that limits the polyglutamylation of MTX, but the lack of subsequent processing. This result clearly indicated that MTX is not a substrate for the plasmodial FPGS activity.
To further investigate any possible interaction between MTX and PfDHFS–FPGS, MTX was tested as a potential inhibitor in individual DHFS and FPGS assays. However, neither the DHFS nor FPGS activities showed any sensitivity to this drug, even at concentrations as high as 1.0 mM (Supplementary Fig. 5a and b). In contrast, growth of P. falciparum cultures was highly sensitive to MTX, with average IC50 values of ~40 nM for four different strains of the parasite (Fig. 6d), as seen previously with other strains , i.e. about 25,000 times lower, presumably as a result of its known, very tight binding to DHFR (Ki < 1 nM), regardless of whether this enzyme carries the mutations that confer PYR resistance . This indicates a complete absence of either allosteric or competitive effects of MTX on the parasite DHFS–FPGS enzyme. We conclude that if MTX has more than one target in P. falciparum, as has been suggested, it is not mediated via polyglutamylated forms of the drug, unlike in mammalian systems, and further that the DHFS and FPGS activities are unaffected by this antifolate agent.
To date, the reactions catalysed by PfDHFS–FPGS have not been exploited, nor even explored, as possible targets for antimalarial inhibitors. Given that DHFS has no counterpart in mammals and that the parasite FPGS activity shows properties significantly different from those of human FPGS, identifying effective and selective inhibitors of these activities would be an attractive goal. We therefore investigated two novel compounds that had been designed as folate analogue inhibitors by mimicking the unstable tetrahedral intermediates derived from the acyl phosphate intermediates formed during DHFS- and FPGS-catalysed ligation of glutamic acid to DHP and THF, respectively . Thus, two tetrahedral phosphinic acids (1  and 2 ; Fig. 7) were evaluated as predicted inhibitors of PfDHFS (1) and PfFPGS (2), respectively. In exploratory experiments, the 7,8-dihydro derivative of 1 exhibited an IC50 value of 0.41 μM for the DHFS activity of the recombinant protein while the value for the 7,8-dihydro derivative of 2 against the FPGS activity was 0.39 μM (Supplementary Fig. 6). The specificity of these two inhibitors was more rigorously studied by using Vmax values in secondary plots to calculate the Ki values for both compounds against both activities. The data showed that reduced 1 was about 5-fold more potent in inhibiting the DHFS activity than the FPGS activity. Thus, it had a Ki value of 0.14 μM for the former and one of 0.63 μM for the latter. Conversely, reduced 2 was a much more potent inhibitor of FPGS activity (almost 20-fold), with a Ki value of 0.091 μM for the FPGS activity and of 1.69 μM for the DHFS activity (Fig. 8). These figures were derived under low substrate concentration conditions where data fit to the Michaelis–Menten equation was good (see above), but were little changed at high concentrations of ATP and glutamate (data not shown), where Vmax values were determined using the full equation of model a in Supplementary Fig. 2.
To assess the mode of binding of the two drugs to PfDHFS–FPGS, either DHP was used as the variable substrate for the DHFS activity or THF as that for the FPGS activity with the inhibitors used in a series of constant concentration steps in each data set to yield the double-reciprocal plots of Supplementary Fig. 7. Despite converging to points in all of the plots, these lie off both the x- and y-axes, as both inhibitors induced changes in both the Km and the Vmax values of the reaction. This indicates that the effects of the drugs are not uncompetitive, but neither are they fully competitive nor fully non-competitive, with both apparently exerting their effect through a mixed mode of competitive action on the bifunctional DHFS–FPGS enzyme of the parasite.
Folate biosynthesis and metabolism is a long-established clinical target for antimalarial intervention. We have demonstrated previously that the malarial genome encodes a bifunctional DHFS–FPGS as a component of its folate pathway and verified its predicted roles by complementation experiments in mutants of yeast and E. coli defective in these activities [30,34]. However, a thorough kinetic study has not yet been carried out on this protein, whose functions are critical to both folate synthesis and folate modification, from this or any other protozoan parasite. We first studied the binding of the substrates of the PfDHFS activity by systematically varying the concentrations of the three compounds involved, DHP, ATP and glutamate. Our data are consistent with their binding sequentially to form a quaternary complex, with DHP having the highest affinity among the three substrates. This would be consistent with a low level of synthesis of DHP in the parasites relative to the availability of ATP and glutamate. However, we demonstrate that the plasmodial DHFS activity is an unusual enzyme type in that accurate modelling of its kinetics required the concept of a second DHP binding site to be introduced together with the assumption that the binding of this second site by DHP inhibits the enzyme's activity, but that such inhibition is only partial, even at high DHP concentrations. A similar phenomenon has been observed for DHPS from Staphylococcus aureus with respect to pAB  and Cryptosporidium parvum inosine monophosphate dehydrogenase with respect to NAD . In our experiments, substrate inhibition became significant at DHP concentrations higher than ~2 μM and was most pronounced when the co-substrates ATP and glutamate were readily available at the higher concentrations expected in vivo. Thus in general, ATP concentrations are in the mM range in most cell types investigated , and this is also true of P. falciparum blood stages, where an average level of ~2.5 mM has been measured for parasites cultured in physiological glucose . Similarly, studies on P. lophurae indicate that glutamate levels are normally at least 1 mM and probably higher in the cytosol . The unusual kinetics we observe suggest that under normal in vivo conditions, production of the folate moiety might be a step in the biosynthetic pathway at which a certain degree of negative feedback is operational, depending upon the (unknown) concentrations of DHP. In contrast, no evidence was seen for substrate inhibition in the FPGS reaction with THF as substrate under any conditions, quite different to what is observed with human FPGS [43,44,58]. The apparent Km values of ~1 μM that we determined for the parasite DHFS and FPGS activities with respect to their principal substrates are comparable to those seen with other homologues, e.g. ~0.4–6 μM for DHP with bacterial DHFS [22,24,37] and ~2–5 μM for THF with bacterial and mammalian FPGS [37,42,59,60]. Values of Vmax were very variable, depending upon the enzyme preparation and its age, reflecting significant instability, as also reported for a range of other DHFS/FPGS molecules, both prokaryotic and eukaryotic [22,25,28,36,59]. This, together with the DHP inhibition phenomenon observed here for the PfDHFS activity made meaningful comparisons with values reported for other systems difficult. However, the highest turnover numbers we calculated were 0.04 s−1 (uninhibited) and 0.01 s−1 (substrate inhibited), significantly slower than reported for human and bacterial enzymes (0.6–0.8 s−1 [25,37,42]), but similar to figures of 0.02–0.03 s−1 reported for the P. falciparum HPPK–DHPS bifunctional enzyme that precedes DHFS in the folate biosynthetic pathway [61,62]. Another enzyme in the thymidylate cycle component of the P. falciparum folate pathway, serine hydroxymethyltransferase (SHMT), exhibits faster turnover numbers (0.16–0.74 s−1 [63,64]), but are similarly calculated as being, on average, about 30 times slower than their human or bacterial equivalents .
Our studies of substrate specificity also indicate that the product of the DHFS step, DHF, despite its location at the catalytic site of the DHFS activity, almost certainly must dissociate and undergo reduction by DHFR to THF, before rebinding in this form to the site of FPGS activity for conversion into polyglutamylated forms. Whether the two active sites are discrete or overlapping in bacteria with a bifunctional DHFS–FPGS has been the subject of some controversy [21,23,24], but our data on the parasite protein (Fig. 4) are more consistent with the latter view, that DHP and THF must occupy the same or extensively overlapping catalytic sites. This in turn raises the question as to how the relative levels of folate production and subsequent polyglutamylation are controlled, if at all, in this organism.
The most obvious difference between parasite and host with regard to these aspects of folate metabolism is that, as in the case of E. coli and other bacteria, the plasmodial protein has two functions whereas the human enzyme has only the FPGS activity and entirely lacks a DHFS, along with other enzymes of de novo folate synthesis. However, there is a significant level of similarity between human FPGS, DHFS–FPGS from the malarial parasite and DHFS or FPGS enzymes from other organisms in their primary protein sequences . They all have an active site located in the same area of the enzyme and the binding of ATP and glutamate is conserved among all of them. Importantly though, the substrate specificity studies reported here demonstrate that, in addition, the parasite FPGS has properties that differ significantly from those reported for the orthologue of its host, particularly with respect to the oxidation state of the folate to be glutamylated, further justifying the bifunctional parasite molecule as a putative novel drug target. Aryl phosphinic acid analogues of folic acid and several antifolates have been synthesised as potential inhibitors of DHFS . With targeting of bifunctional DHFS–FPGS in mind, including that of P. falciparum, we tested for the first time such inhibitors against the plasmodial target activities. Thus, using the aryl phosphinic analogue of folic acid (1), we demonstrated that this compound in its 7,8-dihydro reduced form (but not its oxidised form) strongly inhibited the DHFS activity of the parasite with a Ki of 140 nM, but was about 5-fold less effective against the FPGS activity. We also tested an alkyl phosphinic acid analogue of pteroylglutamyl-γ-glutamate (2), an inhibitor originally designed to target the human FPGS activity [32,33,65] against both of the activities of PfDHFS–FPGS. Consistent with its design, this analogue, again in its reduced form, showed a marked preference for the plasmodial FPGS activity with a Ki of 91 nM, almost 20-fold more potent than that measured for the DHFS activity (1.69 μM). Given how poor we found DHF itself to be as a substrate for PfFPGS, the striking effectiveness of the dihydro-form of 2 against this activity is perhaps surprising, and suggests that in this case the nature of the heterocycle is less important for binding than is the phosphinate mimic of the tetrahedral intermediate. These in vitro inhibitor studies on purified enzyme serve as an important proof of principle, but in preliminary experiments, the reduced compounds were unable to significantly inhibit parasite growth in culture, probably due to instability over long periods at 37 °C and/or insufficient cell permeability. Such analogues would therefore require further appropriate structural modifications to overcome these current limitations.
The study using MTX as a candidate inhibitor yielded several interesting results. This major anti-cancer antifolate is reported to inhibit multiple enzymes in mammalian cells in addition to its traditional target of DHFR . This secondary mechanism of MTX inhibition is due to its polyglutamylation by cellular FPGS activity, increasing its affinity for other key molecules, including TS and other folate-dependent enzymes . Moreover, in cells that can potentially develop resistance to MTX by pumping the unaltered drug outside of the cell membrane, polyglutamylation enhances its accumulation by virtue of the increases in negative charge. Our study of the ability of plasmodial FPGS to catalyse MTX polyglutamylation, both in vitro and in vivo, again highlighted major differences between parasite and human FPGS activities. No polyglutamylated forms of MTX could be detected after overnight cultivation with the 3H-MTX label whereas the conversion of a near-equimolar concentration of 3H-folinic acid, a folate with a fully reduced (tetrahydro-) pterin ring, was almost complete under identical conditions, a phenomenon that we showed was not due to poor transport and accumulation of the drug in the parasite, which actually exceeded that of folinic acid. In fact, we could find no evidence for any interaction between PfDHFS–FPGS and MTX, nor more than a very low level of activity on folic acid, which MTX resembles closely, especially in the fully oxidised nature of the pterin ring. Whether MTX inhibits any enzyme in the parasite folate pathway other than its classical target DHFR, or indeed any other non-folate related activities, remains to be established, but we can confidently exclude MTX inhibition of PfDHFS–FPGS and any role of polyglutamylated forms of this drug in its powerful antimalarial effect.
Taking the above studies on phosphinic acid folate analogues and MTX together, our results emphasise that, despite the close similarities in their function and the likely extensive overlap in their catalytic site, the DHFS and FPGS activities of the parasite can be differentiated from each other at the inhibitor level. Moreover, the kinetic and substrate specificity experiments reveal that the parasite FPGS shows significant differences relative to its human orthologue. These observations, together with the absence of DHFS activity in humans, open up the possibility of developing inhibitors against the bifunctional DHFS–FPGS protein that are highly specific for the parasite.
We thank Abraham Louw and Esmare Human (University of Pretoria) for help with preliminary experiments towards expression of the pfdhfs-fpgs gene and Jeremy Derrick (University of Manchester) for helpful discussions. We also thank David Bartley and John Tomsho (University of Michigan) for the sample of compound 2 used in these experiments. We acknowledge the Wellcome Trust, UK [grant number 073896] for support of this work.
Appendix ASupplementary data associated with this article can be found, in the online version, at doi:10.1016/j.molbiopara.2010.03.012.