Fosmidomycin Uptake into Plasmodium and Babesia-Infected Erythrocytes Is Facilitated by Parasite-Induced New Permeability Pathways

Background Highly charged compounds typically suffer from low membrane permeability and thus are generally regarded as sub-optimal drug candidates. Nonetheless, the highly charged drug fosmidomycin and its more active methyl-derivative FR900098 have proven parasiticidal activity against erythrocytic stages of the malaria parasite Plasmodium falciparum. Both compounds target the isoprenoid biosynthesis pathway present in bacteria and plastid-bearing organisms, like apicomplexan parasites. Surprisingly, the compounds are inactive against a range of apicomplexans replicating in nucleated cells, including Toxoplasma gondii. Methodology/Principal Findings Since non-infected erythrocytes are impermeable for FR90098, we hypothesized that these drugs are taken up only by erythrocytes infected with Plasmodium. We provide evidence that radiolabeled FR900098 accumulates in theses cells as a consequence of parasite-induced new properties of the host cell, which coincide with an increased permeability of the erythrocyte membrane. Babesia divergens, a related parasite that also infects human erythrocytes and is also known to induce an increase in membrane permeability, displays a similar susceptibility and uptake behavior with regard to the drug. In contrast, Toxoplasma gondii-infected cells do apparently not take up the compounds, and the drugs are inactive against the liver stages of Plasmodium berghei, a mouse malaria parasite. Conclusions/Significance Our findings provide an explanation for the observed differences in activity of fosmidomycin and FR900098 against different Apicomplexa. These results have important implications for future screens aimed at finding new and safe molecular entities active against P. falciparum and related parasites. Our data provide further evidence that parasite-induced new permeability pathways may be exploited as routes for drug delivery.


Introduction
The antibiotic Fosmidomycin (Fos; 3-[formyl(hydroxy)amino]propylphosphonic acid; CID 572) and its derivative FR900098 (FR; 3-[acetyl(hydroxy)amino]propylphosphonic acid; CID 162204) were described previously as inhibitors of DOXP reductoisomerase (Dxr), the second enzyme in the biosynthesis pathway of isoprenoids in P. falciparum, inhibiting its in vitro and in vivo growth at high nanomolar concentrations [1]. Several studies have confirmed the presence of the other individual enzymatic steps in this organism and its essential nature for parasite survival [2,3]. In combination with the antibiotic drug clindamycin, Fos has already been tested in phase II clinical trials against uncomplicated malaria with good success [4,5,6,7,8]. Fos has an exceptional safety profile in humans, even when given repeatedly at a dose of 8 g/day [9]. There is an ongoing need for new, safe and affordable anti-malarials, in particular after reports of decreased sensitivity against artemisinin-based monotherapy have appeared in the literature [10].
Isoprenoids are a large and diverse group of natural compounds fulfilling a large number of diverse cellular functions in all biological systems, such as cell signaling processes, protein modifications (prenylation), synthesis of the co-factor ubiquinone and modifications of tRNAs, amongst others [11]. The basic building blocks for all these structures are isopentenyl diphosphate (IPP) and its isomeric form, dimethylallyl diphosphate (DMAPP).
Two alternative routes for their synthesis are known: most eubacteria and plants follow the so-called 1-deoxy-D-xylulose-5phosphate (DOXP) pathway (also called methylerythritol phosphate (MEP) pathway) whereas eukaryotes and archaebacteria mostly use the mevalonate (MEV) pathway [12] (see Figure S1). The two pathways are fundamentally different, starting from different compounds and employing distinct enzymes leading to specific intermediate products. Unlike humans, almost all apicomplexan parasites, including Plasmodium falciparum, the causative agent of human malaria, and Toxoplasma gondii, causing toxoplasmosis, are now known to synthesize isoprenoids exclusively via the DOXP pathway in the apicoplast, an essential, metabolically active, reduced plastid of endosymbiotic descent found in almost all Apicomplexa [13].
Bioinformatic analyses of the published genome sequences from several Apicomplexa of human and veterinary medical importance (Plasmodium, T. gondii, Neospora caninum, Theileria and Babesia bovis) have unequivocally shown the presence of all genes of the DOXP pathway in all these organisms [13,14] (see also Table S1). Notably, the Dxr protein sequences are highly similar among these organisms, and residues known from 3D-structures of bacterial Dxr to be important for Fos binding are well conserved ( Figure  S2). They can be superimposed onto the respective amino acids in a 3D-model of the T. gondii sequence [13]. Given these facts it could be assumed that Fos and FR are also active against those parasites. Surprisingly, however, several reports have shown that Fos does not kill T. gondii, Eimeria tenella and T. parva, even at very high (.100) micromolar concentrations [1,14,15,16].
There are numerous reasons why drugs may be ineffective, but the most obvious one is a failure of its uptake into the infected host cell. We therefore started to investigate whether basic differences exist in Fos uptake between T. gondii-infected fibroblasts and cells infected with susceptible parasites, namely P. falciparum-infected erythrocytes. The latter induce alterations in the permeability of the red blood cell (RBC) plasma membrane for a variety of different solutes and which are collectively called new permeability pathways (NPP) [17].
Here we provide evidence that these pathways, which appear to be absent in non-infected erythrocytes, greatly facilitate uptake of the respective drugs into the infected cell. Likewise, Babesia divergens, another apicomplexan parasite of erythrocytes that is also known to possess NPP-like activities, was found to be susceptible to Fos and its derivative FR. Again, Fos uptake was NPP-mediated. Our results are consistent with the view that parasite-induced changes of the host erythrocyte membrane are pre-requisites for the uptake of these drugs into infected RBC. At the same time they provide a likely explanation for the failure to kill other apicomplexans where NPP appear to be absent or not required.

Identification of DOXP reductoisomerase activity in T. gondii cell lysates and its inhibition by fosmidoymcin
In initial experiments we wished to formally prove that Fosinhibitable Dxr activity is present in T. gondii since functional data on Dxr activity in T. gondii have not been reported so far. To evaluate whether native Dxr from T. gondii (TgDxr) is an active enzyme a highly specific and sensitive radiometric assay (see Methods) was performed on lysates of tachyzoite-infected host cells. In this assay, MEP formed from DOXP and NADPH by Dxr is further converted into CDP-ME by recombinant E. coli YgbP, the enzyme performing the next step in the DOXP pathway, thereby incorporating the radioactively labeled phosphorus atom from [a-32 P]CTP ( Figure 1A; Figure S1). The results show that (i) significant Dxr acitivity is present only in lysates of cells infected with tachyzoites ( Figure 1B, compare lane 1 with lane 5) and (ii) that the activity can be completely inhibited by Fos in a dosedependent manner ( Figure 1B, lanes 2-4; 1C). We conclude that TgDxr is an active enzyme that can be inhibited by low micromolar concentrations of Fos in vitro.
DOXP reductoisomerase localizes to the apicoplast in T. gondii tachyzoites and P. falciparum blood stages We next wanted to confirm that Dxr resides in the apicoplast of apicomplexan parasites. Notably, available proteomics data do not provide direct evidence for the expression of Dxr neither in P. falciparum nor in T. gondii (see Table S1), and neither has in situ localization of Dxr in P. falciparum or in T. gondii been reported so far. Previous targeting experiments had shown that the N-terminal leader peptide of P. falciparum Dxr fused to GFP transported this construct to the apicoplast of T. gondii tachyzoites [1]. To prove expression of Dxr in the apicoplast, polyclonal antibodies were raised against recombinant PfDxr (Figure 2 and Figure S11) and antibody staining was performed on intracellular T. gondii tachyzoites and P. falciparum blood stages. Discrete anti-PfDxr reactivity can be detected in T. gondii tachyzoites ( Figure 2A) as well as P. falciparum blood stages (schizonts, Figure 2B; for other stages see Figures S3, S4, S5, S6, S7). Using co-localization with the apicoplast-resident acyl carrier protein (PfACP) in the case of P. falciparum and direct staining of the apicoplast DNA for T. gondii, these structures were clearly identified as the apicoplast in both organisms, indicating that Dxr is expressed in this organelle. For P. falciparum this is in agreement with results showing that most downstream intermediates of Dxr could be detected in all blood stages [2].

T. gondii-infected fibroblasts do not take up FR
Having shown that T. gondii possesses Dxr activity that can be inhibited by Fos we directly addressed the possibility that the failure of the drug to kill tachyzoites could be due to its reduced uptake into cells infected with T. gondii. To this end we synthesized radiolabeled drug for transport studies. [ 14 C]FR is much easier to synthesize than [ 14 C]Fos, and FR differs from Fos by a single methyl group ( Figure 1D) which neither alters its hydrophilicity nor other physico-chemical properties (Table S2) nor overall shape ( Figure S8). FR is also twice as potent in inhibiting PfDxr than Fos [1].
When we followed [ 14 C]FR uptake into human foreskin fibroblasts (HFF) for 15 min no significant cell-associated radioactivity was seen for [ 14 C]FR ( Figure 3). However, [ 3 H]-Lglutamate ([ 3 H]Glu) that served as control (see below) was readily taken up, as described previously [18]. Heavy infection of cells (.50%) with tachyzoites did not result in higher [ 14 C]FR counts, in stark contrast to four-fold increased values for [ 3 H]Glu ( Figure 3). This is consistent with the observed slight up-regulation of transcripts of the epithelial high affinity glutamate transporter EAAT3 upon T. gondii infection [19]. Intracellular accumulation of [ 3 H]Glu could be significantly inhibited by pre-incubation with unlabeled L-Glu. Extending the labeling period with [ 14 C]FR to 2 hours in similar preliminary experiments did not result in significantly different dpm between these two time points for both, infected and non-infected HFF (unpublished observations). To rule out that rapid efflux of [ 14 C]FR via the P-glycoprotein efflux pump was responsible for the observed results assays were performed in the presence of the inhibitor verapamil; however, this did not lead to significantly increased [ 14 C]FR counts in those cells (data not shown). Taken together, these experiments indicate that T. gondii-infected fibroblasts do not take up FR.
Uptake of FR into P. falciparum-infected erythrocytes depends on functional new permeability pathways (NPP) Since P. falciparum induces permeability changes of the host cell for a variety of different substrates, we next studied [ 14 C]FR uptake into P. falciparum-infected human red blood cells (Pf-iRBC). We observed a time-dependent increase of radioactivity in Pf-iRBC ( Figure 4A). Strikingly, non-infected cells (RBC) showed only minor amounts of cell-associated counts. This largely increased uptake behavior into iRBC is reminiscent of compound entry via the so-called new permeability pathways (NPP). This is illustrated by uptake of the known NPP substrate [ 14 C]L-Glu ( Figure 4B). To test whether [ 14 C]FR entry is indeed via the NPP, we inhibited these pathways pharmacologically with 50 mM 5nitro-2-(3-phenylpropylamino)-benzoic acid (NPPB), a well-known NPP inhibitor (Figure 4 A,B). Consistent with its entry via the NPP, FR uptake was considerably inhibited (.50%) by NPPB ( Figure 4A, B; iRBC+NPPB). Similar results were obtained with another NPP inhibitor, furosemide (data not shown). Both compounds decreased [ 14 C]FR uptake dose-dependently ( Figure  S9).
It has been reported previously that limited protease treatment of Pf-iRBC with chymotrypsin almost completely abrogates NPP activity [20]. Accordingly, chymotrypsinization of Pf-iRBC resulted in a drastic decrease of FR uptake to almost background levels ( Figure 4A, iRBC+chymptrypsin), as did the known NPPsubstrate L-glutamate ( Figure 4B). Together, these results clearly implicate a role of NPP in FR uptake into P. falciparum-infected erythrocytes. FR entry via NPP is also observed in Babesia-infected erythrocytes Some apicomplexan parasites of the genus Babesia also infect mammalian erythrocytes and are considered emerging human pathogens [21]. Recent data provided evidence for permeability changes in Babesia divergens-infected human RBC (Bd-iRBC) [22]. To test if FR uptake via NPP-like mechanisms is a general feature of Apicomplexa-infected RBC we used B. divergens previously adapted to in vitro culture in human RBC to measure [ 14 C]FR uptake. The results show that Bd-iRBC, like Pf-iRBC, have significantly higher cell-associated [ 14 C]FR counts than non-infected cells ( Figure 5). Uptake of [ 14 C]FR is not as prominent as in P. falciparum-infected erythrocytes, most likely because of a lower trophozoite stage parasitemia, which is due to the current lack of methods for stage-specific enrichment of Babesia-infected RBC (average parasitemia of 20% trophozoite stage Bd-iRBC, compared to 80% trophozoite stage Pf-iRBC). In good agreement with a direct role of NPP-like mechanisms in FR uptake, inhibition by furosemide leads to a statistically significant decreased [ 14 C]FR uptake (P = 0.011 by paired t-test).
Having shown that FR also enters Bd-iRBC it was of interest whether the drug would also inhibit the intracellular growth of B.  . Cell-associated radioactivity was determined by scintillation counting of an aliquot of lysed cells and dpm for each compound were determined according to the manufacturer's instructions (shown as dpm6SD). In some assays cells were preincubated for 30 min in 1 mM unlabelled FR or L-Glu to test for uptake specificity (indicated by '+FR' and '+Glu'). This is one representative of three similar experiments. doi:10.1371/journal.pone.0019334.g003 divergens at pharmacologically relevant concentrations. A sharp reduction in parasite growth (10% of untreated control after 72 h) could already be seen with 10 mM FR (Table 1). Our results are in very good agreement with recently published data showing that Fos inhibits the intraerythrocytic growth of two other Babesia species (B. bovis and B. bigemina) in bovine RBC at such low concentrations (reported IC 50 ,5 mM; [23].
Failure of FR to inhibit growth of liver stage P. berghei parasites in vitro and in vivo To our knowledge, the growth inhibiting activity of Fos and FR for Plasmodia has only been studied on blood stage parasites.
Based on published microarray data of the liver stages of the rodent parasite P. yoelii many of the genes for the DOXP pathway are assumed to be expressed at least as strongly in liver stage as in blood stage parasites ( [24]; Figure S10). We therefore tested whether the stages of P. berghei multiplying in nucleated liver cells (exo-erythrocytic forms; EEFs) are sensitive to increasing FR concentrations ( Figure 6). Human hepatoma cells infected with GFP-expressing P. berghei sporozoites were incubated with 1, 10 and 100 mM FR, respectively, for 48 h and parasite replication was recorded. These drug concentrations are between 2 to 200fold above the IC 50 for P. falciparum blood stages [1]. In no case were we able to detect differences in EEF growth ( Figure 6A, B).
We then infected groups of susceptible C57BL/6 mice with 10,000 P. berghei sporozoites each and subsequently treated them intraperitoneally with four high doses of 250 mg/kg FR at 12 h intervals. This high dose regimen is expected to result in sustained high drug levels in the liver [25,26]. However, no significant difference (P.0.05; unpaired student's t-test) of parasite load, as quantified by real-time PCR of P. berghei 18S RNA transcripts in the liver of these animals, could be observed after 42 h when compared to those of uninfected control animals ( Figure 6C). In contrast, a control group that had received 60 mg/kg primaquine at 0 h und 24 h post-infection showed a vast decline in parasite load, illustrating the susceptibility of this parasite strain to in vivo drug treatment. To further prove that the batch of FR used in the above experiments was able to kill susceptible parasites, growth inhibition of blood stages was monitored in a group of mice that had received doses of 75 mg/kg FR every 8-12 hours for 5 days starting at day 3 after iRBC inoculation. Parasitemia in treated animals was 0.22%60.38% 7 days after infection, whereas that of untreated controls was 17.1%66.8% ( Figure 6D). Taken together these results indicate that susceptibility of Plasmodium to FR treatment can only be observed when the parasites reside in erythrocytes but not when multiplying in nucleated liver cells.

Discussion
We provide evidence that only erythrocytes with an increased plasma membrane permeability for a variety of solutes upon  infection with P. falciparum or B. divergens parasites (so-called new permeability pathways, NPP [17]) are permissive for uptake of the highly charged drugs Fos and FR. In striking contrast, those parasites that do not reside in erythrocytes and do not show this property, such as T. gondii or the liver stages of P. berghei, are obviously secluded from the drug's action since these compounds seem unable to traverse the plasma membrane of the respective host cells.
The high hydrophilicity of the compounds tested presumably prevents their passage through the host cell plasma membrane in the absence of a suitable transporter or altered membrane permeabilities. This situation resembles that in a number of organisms like the cyanobacterium Synechocystis sp. PCC6803 and in Mycobacteria where Fos is known to inhibit the respective bacterial Dxr enzymes but shows no activity against the intact organisms [27,28,29]. A recent study provided evidence that Fos is unable to enter Mycobacteria, presumably because they lack the gene for a glycerol-3-P transporter (GlpT) [30], which is also absent in the genome of Synechocystis sp. PCC6803 (unpublished observation). In E. coli this transporter has been shown to be responsible for Fos uptake [31], and introduction of this gene into Brucella abortus is both required and sufficient to make these bacteria Fos-susceptible [32]. Lack of uptake is presumably also the reason why a distant relative to Apicomplexa, the dinoflagellate Perkinsus marinus, is insensitive to high concentrations of Fos [33], despite the presence of the whole DOXP pathway in its genome [34] (Figure S2).
From more than 40 compounds known to act on apicoplast targets, Fos and FR are by far the most hydrophilic compounds at physiological pH. Both drugs bear significant structural resem-  blance to DOXP (see Figure 1D) but the latter is not known to be involved in any human metabolic pathway as a precursor or product [2]. Moreover, in the Human Metabolome Database there is no indication for its presence in body fluids (http://www. hmdb.ca/metabolites/HMDB12173). It is therefore likely that no specific transporter for these compounds exist in the host cell plasma membrane. Only erythrocytes that, upon infection with either Plasmodium or Babesia, alter their membrane permeability in turn become permissive for Fos or FR uptake. To our knowledge NPP-like permeability changes have not been reported to date for any nucleated cells infected by other Apicomplexa. Of note, P. berghei sporozoites modulate the volume-regulated anion channel activity of its target cells upon infection [35]. However, the altered activity is a mere response of the host cell to the increased cell volume and not a prerequisite for parasite survival [35]. Although our data provide clear evidence that Fos or FR uptake into infected erythrocytes requires functional NPP, at present we do not know any molecular entity that is involved in this initial entry process. The existence and importance of NPP in Plasmodium (and by comparable mechanisms also in Babesia)-infected cells for the acquisition of many essential nutrients is largely undisputed, however, the exact way this is accomplished and the molecules being involved is much less so (for recent discussions see [36,37]). The recognition of NPP-specific anti-parasitic compounds in this and other studies [38,39] may aid in the identification of the molecular make-up of the NPP, for instance by exploiting differential screens of Plasmodium mutants for Fos resistance.
Fos (in combination with clindamycin) is the first anti-malarial drug that successfully completed phase II trials [8] for which specific cell entry via the NPP pathway is shown. Our findings reinforce the previously suggested concept of exploiting this phenotype for targeting drugs into iRBC [40,41,42]. In some parasite lines the development of resistance to antimalarials such as blasticidin and leupeptin has been found to be associated with alterations of a so-called plasmodial surface anion channel (PSAC) [43,44], indicating the existence of a pathway that involves a parasite-encoded transporter, which may undergo mutations under selective pressure. However, since it is likely that more than a single mechanism constitute what is collectively called NPP [37], it is also possible that in some instances the actual transporter is an activatable host cell protein. In such cases, the chances for resistance development at the point of drug transport should be minimal since there is no selective pressure on the respective host genes. So far experimentally induced resistance against Fos has only been reported to be caused by amplification of the Dxr gene locus [45].
NPP are generally described as transporting a broad range of substrates, with a preference for anions and electroneutral compounds over cations, and with the rate of permeation influenced by size and hydrophobicity of the solute in a number of cases [17]. However, simple correlations between size, charge or hydrophobicity and permeation rates are not apparent, and a combination of different properties could play a role in other cases [46]. Fos and FR are small but very hydrophilic at physiological pH, whereas two other anti-plasmodial compounds (pentamidine; T16) taken up via the same route [38,39] are larger and much more lipophilic (Table S2). Therefore, predictions of inhibitors that enter iRBC via NPP seem unreliable and instead rigorous experimental testing is required in each case.
The concept that hydrophilic, charged molecules could be potent drugs if their structures resemble natural metabolites for which specific transport routes exist, has been recently emphasized [47]. Extending this idea to Plasmodium-infected cells we propose that future screens aimed at finding such new anti-plasmodial molecules should include more small hydrophilic metabolite-like compounds specific to 'parasite-only' pathways (e.g. from the apicoplast; [13]) in their chemical libraries (in contrast to the current situation; see [47,48] for a general discussion). This strategy would increase the chances of developing novel drug candidates with a high, built-in safety profile, provided that they can enter iRBC only via NPP. The fact that the target of Fos and FR, Dxr, is absent from the human host adds an additional safety level to its use. Interestingly, in a recent high-throughput screen of nearly 2 million compounds against blood-stage P. falciparum, among 13,500 active hits were also two purine analog phosphonates with physico-chemical properties comparable to Fos and FR [49]: CHEBI 390944 with an estimated IC 50 of 470 nM, and CHEBI 641822 with an estimated IC 50 of 340 nM ( [49]; Figure  S8 and Table S2). It would be interesting to see whether the presence of the purine moieties in these anti-plasmodial phosphonates make them substrates for specific transporters or if they can enter iRBC only via NPP. The latter would further advocate an appealing option [50], namely synthesizing chimeric drugs based on active phosphonates like Fos/FR that are still taken up via the NPP, and subsequently cleaved inside the cell to liberate two or more inhibitory entities.
Our results provide a plausible explanation for the observed differences in killing activity of Fos and FR against different apicomplexan parasites. At present it is not known if any of the other three membranes both compounds have to pass to reach their target Dxr in the apicoplast are also impermeable for these drugs in T. gondii, the liver stages of Plasmodium, or any of the other apicomplexans that are not killed by Fos or FR. This aspect is of general cell biological and biochemical interest and deserves further studies. However, with regard to Fos' and FR's intended application as drugs these potential barriers are of secondary importance since the host cell plasma membrane is the first and thus determining barrier that has to be overcome by any pharmacologically active compound. Failure to do so eliminates Fos and FR in their current formulation as potential drugs in those host-parasite systems.
Nevertheless, in pathogens that fail to take up Fos or FR the DOXP pathway remains a prime drug target. Hope to improve drug delivery to the pathogen comes from recent studies showing that lipophilic Dxr inhibitors structurally unrelated to Fos and FR can be developed [51]. The hydrophilicity problem of phosphonates is well known and might be overcome by the development of phosphonate esters [52], although cleavage of the ester linkage by cellular esterases in the host cytosol might generate again a charged Fos or FR molecule impermeable for the parasite plasma membrane. Importantly, specific inhibitors for other enzymes of the DOXP pathway in bacteria have been described recently [53,54].
In conclusion, our results show that the parasite-and life cyclespecific action of anti-infectives may be explained by differential uptake into the infected host cell, which has important implications for future screens aimed at finding safe, affordable and potent molecular entities active against P. falciparum and other related Apicomplexa.

Ethics Statement
All animal work was conducted in accordance with the current German Protection of Animals Act (BGBl. I S. 1207), which implements the directive 86/609/EEC from the European Union and the European Convention for the protection of vertebrate animals used for experimental and other scientific purposes. The protocol was approved by the ethics committee of the Max-Planck-Institute for Infection Biology and the Berlin state authorities (LAGeSo Reg# G0469/09).

Chemicals
All chemicals were from Sigma-Aldrich (except where noted). Fosmidomycin and FR900098 were synthesized as previously described [55,56]. [  Cell culture P. falciparum. The in vitro culturing and synchronization of P. falciparum (isolate FCBR) was carried out by standard protocols [57,58]. Human erythrocytes and plasma were purchased from Blood Bank Universitä tsklinikum Gieben und Marburg GmbH, Germany. Parasites were cultured in human erythrocytes (blood group A + ) at 37uC and a hematocrit of 2%. The culture medium was RPMI1640 (PAA, Germany) supplemented with 10% human plasma (A + ; heat inactivated at 56uC for 30 min). Trophozoiteinfected erythrocytes were enriched to a parasitemia of .80-90% by plasmagel floatation [59].
B. divergens. Blood stage parasites of B. divergens (isolate Rouen 1987; [61] were cultured in vitro in human A + erythrocytes (Blood Bank Marburg, Germany) using RPMI1640 supplemented with 10% human serum (A + ; Blood Bank Marburg, Germany) and 0.2 mM hypoxanthine (c.c. pro, Germany) at 37uC and a hematocrit of 5%. Flasks were flushed with a gas mixture of 90% N 2 , 5% O 2 and 5% CO 2 . The medium was changed daily and cultures were diluted to 1% once they reached a parasitemia of 20-30%.

Production of anti-PfDxr antibodies
For the immunization of rabbits recombinant Dxr of P. falciparum (PfDxr) was produced by an optimized version of a previous protocol [1]. A synthetic gene for PfDxr adapted to the preferred codon usage of E. coli was inserted into the pQE31 expression vector (Quiagen, Hilden, Germany) providing an amino-terminal His 6 tag. E. coli XL1-blue pREP cells were transformed with this construct and grown in Terrific Broth medium at 37uC until an OD 600 of 1. After induction by the addition of 1 mM IPTG the culture was continued for 15 h at 30uC reaching an OD 600 of 5 to 9. The cells were harvested by centrifugation, resuspended in a 10-fold volume of IMAC buffer (100 mM NaCl, 30 mM Tris-HCl, 2 mM 2-mercaptoethanol, 14% glycerol, pH 8.0) and disintegrated by ultrasonic treatment (3 times 5 min with 5 min pause, VS70T sonotrode, 30% pulse, 60% amplitude). Insoluble material was removed by centrifugation (75,000 g, 25 min, 4uC) and filtration through a 0.22 mm filter. The soluble fraction was loaded on an immobilized cobalt column (Talon Superflow, Clontech), which was eluted with a two-step gradient of 50 and 150 mM imidazole in IMAC buffer. PfDxr was obtained in the 150 mM imidazole fraction with a purity of ca. 90% as judged by SDS-PAGE. The specific activity determined in a photometric Dxr enzyme assay was 2 U/mg [62]. The protein was concentrated by ultrafiltration and stored at 270uC. A typical purification procedure starting with 8 flasks each containing 350 ml bacteria culture resulted in ca. 1 mg PfDxr. The relatively low yield was due to the fact that most of the protein was produced as inclusion bodies.
The purified enzyme was used for custom immunization (Eurogentec, Seraing, Belgium) of 2 rabbits applying a protocol including an initial injection followed by 3 booster shots. For each injection 1.6 mg PfDxr were used. The sera were tested for reactivity against PfDxr by Western blot. Only one of the two rabbits developed a sufficiently high specific antibody titer (see Figure S11).
Immunofluorescence assays of T. gondii and P. falciparum For indirect immunofluorescence assays, T. gondii-infected HFF cells grown on 22 mm glass cover slips were processed exactly as previously described [63]). Briefly, cells were fixed in 4% paraformaldehyde, permeabilized in 0.25% Triton-X 100, and blocked in 3% BSA fraction V (Fisher). P. falciparum-infected RBCs were processed exactly as described by Tonkin et al. [64]. All primary and secondary antibody incubations were carried out at room temperature for 1 h each in blocking solution, followed by three 10 min washes in 0.1% Triton-X 100. Apicoplast and nuclear DNA were stained with 2.8 mM 49,6-diamidino-2phenylindole (DAPI, Invitrogen) for 5 min (in PBS) right after the secondary antibody step, and mounted on glass slides using Fluoromount-G (Southern Biotechnology Associates, Inc). Samples were examined on a Leica DM IRBE, 100 W Hg-vapor lamp and an Orca-ER digital camera (Hamamatsu). Images were captured and analyzed using Openlab software (Improvision). For co-localization analysis the 'co-localization' plugin of the ImageJ program (Rasband, W.S., ImageJ, U. S. National Institutes of Health, Bethesda, Maryland, USA, http://rsb.info.nih.gov/ij/, 1997-2009) was used.

Uptake assays
Human erythrocytes/P. falciparum. Non-infected and infected erythrocytes (2610 8 cells/ml with a parasitemia of ,80%) at the trophozoite stage were incubated at 37uC in RPMI1640 medium without human serum in the presence of [ 14 C]FR or [ 14 C]Glu, respectively, of the same specific activity (1 mCi/ml).
At the indicated time points, aliquots of 100 ml corresponding to 10 7 cells were removed and spun through a layer of 200 ml dibutylphtalate. After freezing the sample in liquid nitrogen the bottom of the tube containing the cell pellet was cut, cells were lysed in scintillation cocktail and subsequently radioactivity of the samples was determined in a Beckman Coulter MR4000 scintillation counter. Inhibitor studies using NPPB, furosemide and chymotrypsin were done as described previously [20], with the exception that [ 14 C]FR or [ 14 C]Glu were added and the cultures were then processed as described above.
Human erythrocytes/B. divergens. To determine the influx of [ 14 C]FR into B. divergens infected erythrocytes 10 8 infected cells with a parasitemia of 15-20% were washed three times in Dulbecco's phosphate buffered saline (DPBS) and subsequently incubated in DPBS for 30 minutes at 37uC, to reach a point of zero membrane transport. Thereafter, cells were incubated in 1 ml DPBS containing 1.7 mCi/ml at 37uC for 20 min. The incubation was carried out in the presence or absence of 100 mM furosemide. To separate the cells from the radioactive medium, aliquots of 100 ml were centrifuged through a cushion of 600 ml dibutylphtalate at 13,000 rpm for 2 min. The resulting pellet was carefully harvested and lysed in 2 ml scintillation cocktail (Roth) and analyzed in a scintillation counter (Beckman LS 6000SC).
Human fibroblasts/T. gondii. HFF (hTERT-BJ1) cells were cultured in 6-well plates (Corning) to confluency (1610 6 cells) in DMEM incl. 10% fetal bovine serum and penicillin/ streptomycin. Where appropriate they were infected in duplicate with T. gondii tachyzoites (RHb1) resulting in an infection rate .50% of cells 24 h post infection. Depending on the assay the cells were pre-treated with 1 mM unlabelled FR or Glu for 30 min before they were incubated with either [ 14 C]FR or [ 3 H]Glu, respectively, of the same specific activity (1.45 mCi/ml, equaling 25 mM drug concentration per assay) for 15 min at 37uC. To rule out a possible competition between phosphate present in DMEM and the phosphonate FR for uptake into cells, labeling was performed in phosphate-free ''extracellular buffer'' which mimics the extracellular milieu [65]. Labelling was terminated by 3 washes of the monolayers with ice-cold phosphate buffer (50 mM sodium phosphate pH 8.2) containing 1 mM unlabelled FR and Glu each. At this stage microscopic examination of the plates showed that all monolayers were still intact and evenly infected. Cell lysis was performed by incubating the plates on a shaker with 500 ml lysis solution (0.1 M NaOH, 0.1% SDS) for 1 h at 4uC, before an aliquot (200 ml) was counted in a PerkinElmer MicroBeta Trilux scintillation counter. Dpm were determined according to the manufacturer's instructions.

Coupled radiometric Dxr activity assay
The assay was basically performed as described previously [66]. For preparation of the cell lysate, 10 ml of packed purified T. gondii tachyzoites (ca. 10 7 cells) were suspended in 600 ml assay buffer (100 mM Tris-HCl, 20 mM NaF, 10 mM MgCl 2 , 1 mM MnCl 2 , pH 8.0) supplemented with a protease inhibitor cocktail (4 mM Pefabloc SC, 2 mg/ml aprotinin, 2 mg/ml leupeptin, 2 mg/ml pepstatin A, 2 mg/ml antipain), disintegrated by sonication for 2 min (Sonoplus HD 70, Bandelin, Berlin, Germany; 30% pulse, 70% amplitude) and centrifuged (22,000 g, 20 min, 1uC). For the inhibition test, 45 ml aliquots of the lysate were combined with 5 ml of a 20-fold concentrated Fos solution in water and pre-incubated for 5 min on ice. Then, the activity test was started by combining 10 ml of these aliquots with 10 ml of reaction mixture consisting of 8 mM DOXP, 8 mM NADPH, 0.5 mCi/ml [a-32 P]CTP (400 Ci/mmol) and 0.1 mg/ml E. coli YgbP (IspD) in assay buffer. After incubation at 37uC for 5 min, 0.7 ml samples were spotted onto 10620 cm silica gel 60 HPTLC plates (Merck), which were developed longitudinally for 320 min with a mixture of npropanol/ethyl acetate/H 2 O (6:1:3, v/v). For autoradiography, the plates were exposed for 3 h to a Kodak X-Omat AR film. P. berghei blood stage growth inhibition assay in mice 10 C57BL/6 mice were intravenously infected with 30,000 P. berghei ANKA infected red blood cells from a NMRI donor mouse. At day three post-infection (p.i.), parasitemia was determined by means of Giemsa-stained thin blood smears at different time intervals. The baseline parasitemia before addition of the first dose of 75 mg/kg FR900098 was 4.9%60.66% (mean6standard deviation). 5 mice were treated with 75 mg/kg FR900098 dissolved in PBS, which was given intraperitoneally every 8-12 hours for 5 days. Control mice received PBS only.
In vitro effects of FR900098 on exoerythrocytic forms of P. berghei Labtek slides with 30,000 HuH7 human hepatoma cells [67] were infected with 10,000 sporozoites per well (cl. 507, constitutively expressing GFP; [68]) and treated in triplicate with 100 mM, 10 mM and 1 mM FR900098, respectively. 48 hours after infection cells were fixed with cold methanol. Parasites were visualized with a mouse monoclonal antibody to HSP70, followed by a goat Alexa488-labelled antibody directed against mouse IgG, and Hoechst 33342 to stain the nuclei. EEFs were counted per well.

In vitro effects of FR900098 on B. divergens
Erythrocytes infected with B. divergens were cultured in 24-well plates in triplicates for 72 h in the presence or absence of FR at different concentrations as described above. The parasitemia at time point zero was 1% in each well. The culture medium was changed daily, and parasitemia was monitored by counting Giemsa-stained blood films. From each blood film three different areas were counted. Parasitemia is expressed as the average of infected cells (from triplicates) per 100 cells.
Assessment of the in vivo activity of FR900098 on P. berghei liver stage development Five C57BL/6 mice were infected with 10,000 sporozoites and treated via i.p. injection with 250 mg/kg FR at 0, 12, 24 and 36 hours p.i. 5 control mice received PBS instead of FR but were otherwise treated identical. As a control for successful treatment of infection four infected mice received 60 mg/kg primaquine given 0 and 24 h p.i.
For determination of parasite load total [69,70] Figure S1 Comparison of the mevalonate and DOXP pathway for the biosynthesis of the isoprenoid precursors IPP/DMAPP. The pathways shown are based on MetaCyc [71] and were drawn using the Pathway Tools software [72]. Numbers drawn in blue are enzyme EC numbers. (TIF) Figure S2 Sequence alignment of Dxr proteins from select bacteria and plastid or apicoplast-containing organisms. Sequences were taken from NCBI and aligned using MUSCLE at http://www.phylogeny.fr/. Residues colored in red, green and black in the E. coli sequence are also highly conserved in all other Dxr proteins and have been implicated in binding/interaction with the substrate DOXP and/or NADPH (see [73]). Amino acids colored brown in the T. gondii sequence (aa 23 and 67) correspond to the aa following the predicted cleavage site of either the signal sequence (determined with SignalP 3.0) or the apicoplast targeting sequence (taken the first aa of the E. coli sequence as reference point), respectively. (TIF) . Aligned calculated three-dimensional conformer coordinates for Fos and FR were retrieved from PubChem (http://pubchem.ncbi.nlm.nih.gov) and visualized using Chimera [74]. (C) Structures of the two anti-plasmodial phosphonates described in [49]. See http://www.ebi.ac.uk/ chemblntd/ and also the main text for details.  Figure S10 Expression levels of the DOXP pathway genes in P. yoelii. The values are based on microarray data of the rodent parasite P. yoelii [24] and were extracted from the P. yoelii gene entries (accessible via the respective cross-references from the P. falciparum gene entries; see Table S1) at PlasmoDB (http://www.plasmodb.org/). M values denote the relative expression level between pairs of conditions, expressed as base-2 logarithm (M = 61 means a 2-fold difference in expression between the compared samples). BS: mixed erythrocytic stages when parasitemia was at 5-10%. LS24, 40, 50: Isolated liver stage-infected hepatocytes 24, 40 or 50 hrs, respectively, after in vivo infection. The data indicate that some of the genes for the DOXP pathway are even stronger expressed in liver than in blood stages (negative M values). (TIF) Figure S11 Western blot analysis of rabbit anti-PfDxr antisera. The western blot of a T. gondii cell lysate with preimmune sera from two rabbits (lanes 1 & 2) and the respective hyper-immune sera after PfDxr immunization (lanes 3 & 4) is shown. It clearly shows in lane 3 a very prominent band,50 kDa. This size correlates very well with a predicted molecular weight of 48.8 kDa of the mature protein (i.e. without a cleaved bipartite apicoplast targeting sequence; see also Fig. S2). The other rabbit serum did not contain specific antibodies upon PfDxr immunization. (TIF)

Supporting Information
Table S1 Enzymes of the DOXP pathway of isoprenoid biosynthesis in five Apicomplexa. Given are the EC numbers, enzyme names and respective accession numbers in either EuPathDB (http://eupathdb.org/eupathdb) for T. gondii, N. caninum, P. falciparum and T. parva, or NCBI (http://www.ncbi.nlm. nih.gov) for B. bovis. Designations printed in blue mean that MassSpec data have been deposited in EuPathDB for this protein (such data are available only for Plasmodium and T. gondii), indicating that this protein is present in the respective parasite stage. (DOC) Table S2 Physico-chemical properties of select compounds known to act on apicoplast targets in Plasmodium and/or T. gondii. Compounds were compiled from the literature [72,73]. L-glutamic acid and pantothenic acid as physiological NPP substrates (indicated in yellow) and five other anti-plasmodials (blue) are included for comparison (see main text for details). Corresponding data were retrieved from PubChem (http://pubchem.ncbi.nlm.nih.gov/). LogD were calculated using the service ADME Boxes (http://www.pharma-algorithms.com/ webboxes/). Compounds up to hexachlorophene are sorted according to their decreasing LogD. CID, compound ID at PubChem; MW, molecular weight; XlogP3, calculated Log 10 of the partition coefficient in octanol-water [74] (http://www.siocccbg.ac.cn/software/xlogp3); LogD, Log 10 of the apparent octanol-water partition coefficient D at various pH; TPSA, polar surface area of substance [75]. * known to enter iRBC via NPP [38,39]. (DOC)