Mutations in the pantothenate kinase of Plasmodium falciparum confer diverse sensitivity profiles to antiplasmodial pantothenate analogues

The malaria-causing blood stage of Plasmodium falciparum requires extracellular pantothenate for proliferation. The parasite converts pantothenate into coenzyme A (CoA) via five enzymes, the first being a pantothenate kinase (PfPanK). Multiple antiplasmodial pantothenate analogues, including pantothenol and CJ-15,801, kill the parasite by targeting CoA biosynthesis/utilisation. Their mechanism of action, however, remains unknown. Here, we show that parasites pressured with pantothenol or CJ-15,801 become resistant to these analogues. Whole-genome sequencing revealed mutations in one of two putative PanK genes (Pfpank1) in each resistant line. These mutations significantly alter PfPanK activity, with two conferring a fitness cost, consistent with Pfpank1 coding for a functional PanK that is essential for normal growth. The mutants exhibit a different sensitivity profile to recently-described, potent, antiplasmodial pantothenate analogues, with one line being hypersensitive. We provide evidence consistent with different pantothenate analogue classes having different mechanisms of action: some inhibit CoA biosynthesis while others inhibit CoA-utilising enzymes.


Introduction
In recent years, the effort to roll back malaria has shown encouraging progress through the increased use of insecticide-treated mosquito nets, improved diagnostics and artemisininbased combination chemotherapies (ACTs) [1]. Evidence of this includes the decreasing worldwide malaria incidence (266 million cases in 2005 down to 212 million cases in 2015) and mortality (741,000 deaths in 2005 down to 429,000 deaths in 2015) over the past decade [1]. However, there is an alarming trend of ACT-resistant parasites emerging in multiple Asian countries where the disease is endemic [2]. Recently, there have also been multiple reports of patients contracting ACT-resistant Plasmodium falciparum malaria from various African countries [3,4], which exemplify the clear risk of artemisinin resistance developing in the continent. This threat to the efficacy of ACTs highlights the requirement for a new armoury of antimalarial medicines, with several compounds representing different chemotypes entering the preclinical trial stage. However, the antimalarial drug-discovery pipeline is reliant on just a few known drug targets and the probability of successfully producing a new blood-stage medicine remains low [5]. In order to manage the threat of parasite drug resistance, there needs to be a continued effort to identify new classes of antimalarials. One metabolic pathway that has garnered recent interest for drug-development is the parasite's coenzyme A (CoA) biosynthetic pathway [6,7].
Early seminal studies have shown that the asexual stage of intra-erythrocytic P. falciparum absolutely requires an exogenous supply of vitamin B 5 (pantothenate; Fig 1) for survival [7][8][9]. Pantothenate is taken up by the parasite [10,11] and converted into CoA, an essential cofactor for many metabolic processes [7]. This conversion is catalysed by a series of five enzymes, the first of which is pantothenate kinase (PfPanK), an enzyme that phosphorylates pantothenate to form 4'-phosphopantothenate [11]. By performing this step, the parasite traps pantothenate within its cytosol and commits it to the CoA biosynthetic pathway [10]. The additional four steps are, in turn, catalysed by phosphopantothenoylcysteine synthetase (PfPPCS), phosphopantothenoylcysteine decarboxylase (PfPPCDC), phosphopantetheine adenylyltransferase (PfPPAT) and dephospho-CoA kinase (PfDPCK) [6]. Putative genes coding for each of the enzymes in the pathway (with several enzymes having multiple putative candidates) have been identified in the P. falciparum genome [12,13] and have also been shown to be transcribed during the intraerythrocytic stage of the parasite's lifecycle [14]. In order to capitalise on the pathway as a potential target for drug-discovery, however, it is crucial to ascertain the exact identity of each of these putative genes. This will allow the process to become more efficient and targeted.
In this study, we have used continuous drug-pressuring with PanOH or CJ-15,801 to generate a number of P. falciparum parasite lines that are several-fold resistant to these pantothenate analogues. Whole-genome sequencing revealed mutations in one of the two putative Pfpank genes of all of the clones, Pfpank1. Complementation experiments confirmed that these mutations are responsible for the resistance phenotypes. Characterisation of the effects of the mutations on parasite growth in culture and also PfPanK function, generated data consistent with the mutated gene coding for an active PanK in P. falciparum and with the gene being essential for normal parasite development during the intraerythrocytic stage of its lifecycle. Additional characterisation of the PanOH and CJ-15,801-resistant lines revealed that antiplasmodial pantothenate analogues have at least two distinct mechanisms of action, targeting CoA biosynthesis or utilisation. Both of these mechanisms can be influenced by the Pfpank1 mutations identified here. Furthermore, our study provides genetic evidence validating the importance of the metabolic activation of pantothenate analogues in the antiplasmodial activity of these compounds.

Materials and methods
Parasite culture and lysate preparation P. falciparum parasites were maintained in RPMI 1640 media supplemented with 11 mM glucose (final concentration of 22 mM), 200 μM hypoxanthine, 24 μg/mL gentamicin and 6 g/L Albumax II (referred to as complete medium) as previously described [22]. Clonal parasite populations were generated through limiting dilution cloning as reported previously [23], with modifications. Parasite lysates were prepared from saponin-isolated mature trophozoite-stage parasites as described previously [10].

Plasmid preparation and parasite transfection
Several plasmid constructs were generated through the course of this study to be used for different lines of investigations. The strategies used to generate the Pfpank1-pGlux-1, Pfpank1stop-pGlux-1 and ΔPfpank1-pCC-1 plasmids are detailed in the SI and the primers used in these strategies are listed in S1 Table. The constructs were transfected into ring-stage parasites and positive transfectants were selected by introducing WR99210 (10 nM) [24].

SYBR Safe-based parasite proliferation assay
The effect of various compounds on the in vitro proliferation of the different parasite lines were tested using a previously-reported SYBR Safe-based fluorescence assay [17], with minor modifications (SI). In vitro pantothenate requirement experiments were performed similarly (SI), except instead of a test compound, ring stage-infected erythrocytes were incubated in pantothenate-free complete RPMI 1640 medium (made complete as described above; Athena Enzyme Systems) supplemented with 2-fold serial dilutions of pantothenate. IC 50 and SC 50 (defined in the Results section) values were determined from the sigmoidal curves fitted to each data set using nonlinear least squares regression (SI).

Drug pressuring
Two independent drug-pressuring cultures were initiated for each of the pantothenate analogues, PanOH and CJ-15,801. Pressuring was initiated by exposing synchronous ring-stage Parent line parasites (10 mL culture of 2 or 4% parasitaemia and 2% haematocrit) to either analogue at the IC 50 values obtained for the Parent line at the time (PanOH = 400 μM and CJ-15,801 = 75 μM). Parasites were then exposed to cycles of increasing drug-pressure that lasted about 2-4 weeks each (SI). When the pressured parasites became approximately 8 × less sensitive than the Parent line to the selecting analogues, they were cloned and cultured in the absence of the analogues for the remainder of the study.

Competition assay
In order to compare the fitness of the mutant clones with that of the Parent, we set up three competition cultures, each containing a mixture of one mutant line and the Parent line. Equal number of parasites (5 × 10 8 cells in the first experiment and 2.5 × 10 8 cells in the second experiment) from each line were mixed into a single culture. Aliquots (3 to 5 mL) of these cultures were immediately used for a PanOH SYBR Safe-based parasite proliferation assay (to generate Week 0 data) as described above. The cultures were then maintained under standard conditions as detailed above for a period of 6 weeks before they were used to perform another PanOH proliferation assay (to generate Week 6 data).

Whole genome sequencing and variant calling
Next generation whole genome sequencing was performed by the Biomolecular Resource Facility at the Australian Cancer Research Foundation, the Australian National University. Samples were sequenced with the Illumina MiSeq platform with version 2 chemistry (2 × 250 base pairs, paired-end reads) Nextera XT Kit (Illumina). To determine the presence of any single nucleotide polymorphisms (SNPs) in the genomes of the drug-pressured clones, the genomic sequencing data were analysed using an integrated variant calling pipeline, PlaTyPus, as previously described [27], with minor modifications to resolve operating system incompatibility. As PlaTyPus does not detect insertions-deletions polymorphisms ("indels"), the Integrated Genome Viewer (IGV) software (Broad Institute) was used to manually inspect the gene sequences of all putative enzymes in the CoA biosynthetic pathway for indels.

Quantitative polymerase chain reaction (qPCR)
To determine the proportions of parasites (i.e. Parent versus mutant clone) that make up each competition culture, we performed qPCR to measure the amount of Parent and mutant genomic DNA (gDNA) at week 0 and week 6 of the competition assay. gDNA was extracted using a DNeasy Plant Mini Kit (QIAGEN) or QIAamp DNA Blood Midi Kit (QIAGEN) following the manufacturer's instructions, unless otherwise specified (detailed in SI). Six different primer sets were designed, each to detect the specific variants (SNPs or indel) in the wild-type and mutant Pfpank1 alleles (S1 Table). The qPCRs were performed using QuantiTect 2 × SYBR Green PCR Master Mix (QIAGEN) essentially as reported previously [28] with the following specifications: The reactions (20 μL final volume) for "Parent vs PanOH-A" and "Parent vs PanOH-B" contained 2 μL of 10 ng/μL gDNA stocks while that for "Parent vs CJ-A" contained 4 μL of 10 ng/μL gDNA stocks. The PCR program for "Parent vs PanOH-A" and "Parent vs CJ-A" entailed an initial DNA polymerase activation step (95˚C for 15 min) followed by 40 cycles of denaturation (94˚C for 15 s), annealing (60 or 55˚C, respectively, for 30 s) and extension (72˚C for 30 s) with the detection step (green channel) set during the extension step of each cycle. For "Parent vs PanOH-B" the PCR was programmed as above except that it was set for 45 cycles, the annealing step was carried out at 58˚C and included an additional detection step (65˚C for 15 s, green channel) in each cycle. Following each reaction, a melt curve analysis was performed starting with a temperature that is 1˚C lower than the annealing temperature and increased to 99˚C at 1˚C/s. Control reactions that each contained defined proportions of the Parent and mutant gDNA as templates, set at the same final DNA concentration as detailed above, were included in each qPCR experiment to enable preparation of a standard curve. The standard curve was generated by plotting the threshold cycle values obtained for the control reactions against the corresponding log 10 gDNA amount and fitting a straight line (y = y0 +ax, where y denotes threshold cycle value, x denotes log 10 DNA amount and a denotes primer efficiency). The concentration of Parent and mutant gDNA in each sample was determined by comparing their threshold cycle against the appropriate standard curve. All reactions were performed in duplicate.

Generation of PfPanK1 model
The structure of PfPanK1 minus its parasite-specific inserts was predicted by homology modeling using the AMPPNP and pantothenate-bound human PanK3 structure (PDB ID: 5KPR [29]) as a template. PfPanK1 shares 28% sequence identity with human PanK3 over the parts of the protein that have been modeled. The model was generated using the one-to-one threading module of the Phyre2 webserver (available at http://www.sbg.bio.ic.ac.uk/phyre2) [30].

Confocal microscopy
Erythrocytes infected with trophozoite-stage 3D7 strain P. falciparum parasites expressing PfPanK1-GFP were observed and imaged either with a Leica TCS-SP2-UV confocal microscope (Leica Microsystems) using a 63 × water immersion lens or a Leica TCS-SP5-UV confocal microscope (Leica Microsystems) using a 63 × oil immersion lens. The parasites were imaged as fixed or live cells as described in the SI.

Attempted disruption of Pfpank1
The Pfpank1 disruption plasmid, ΔPfpank1-pCC-1 (SI), was transfected into wild-type 3D7 strain P. falciparum, and positive transfectants were selected as described above. P. falciparum parasites have previously been shown to survive equally well in a pantothenate-free complete RPMI 1640 medium supplemented with !100 μM CoA as compared to standard complete medium, consistent with them having the capacity to take up exogenous CoA, hence bypassing the need for any PfPanK activity [7]. Therefore, to support the growth of any Pfpank1 genedisrupted parasites generated with the ΔPfpank1-pCC-1 construct, parasites were continuously maintained in complete medium supplemented with 100 μM CoA following transfection. Positive and negative selection steps (with WR99210 and 5-fluorocytosine (5-FC), respectively) were performed to isolate ΔPfpank1-pCC-1-transfectant parasites in which the double crossover homologous recombination had occurred (detailed in SI).

Southern blot analysis
gDNA samples (~2 μg) extracted from ΔPfpank1-pCC-1-transfectant parasites isolated through the positive and negative selection steps were digested with the restriction enzyme AflII (New England Biolabs), before being analysed by Southern blotting using the digoxigenin (DIG) system (Roche) according to the Roche DIG Applications Manual for Filter Hybridisation. The results from the Southern blot were confirmed with PCR (the primers used are shown in S1 Table).

[ 14 C]Pantothenate phosphorylation by parasite lysate
The phosphorylation of [ 14 C]pantothenate by parasite lysates prepared from the Parent and mutant clonal lines was measured using Somogyi reagent (which precipitates phosphorylated compounds from solution) as outlined previously [31], with some modifications (detailed in SI).

Metabolism of N5-trz-C1-Pan
Cultures of predominantly trophozoite-stage P. falciparum-infected erythrocytes (Parent line) were concentrated to >95% parasitaemia using magnet-activated cell sorting as described elsewhere [32]. Following recovery, trophozoite-infected erythrocytes were treated with N5-trz-C1-Pan (1 μM) or a solvent control (0.01% v/v DMSO) before the metabolites in these samples were extracted and processed for liquid chromatography-mass spectrometry (LC-MS) analysis. Metabolite samples were analysed by LC-MS, using a Dionex RSLC U3000 LC system (Ther-moFisher) coupled with a high-resolution, Q-Exactive MS (ThermoFisher), as described previously [33] (detailed in SI). LC-MS data were analysed in a non-targeted fashion using the IDEOM workflow, as described elsewhere [34]. Unique features identified in N5-trz-C1-Pantreated samples were manually assessed by visualising high resolution accurate mass LC-MS data with Xcalibur Quanbrowser (ThermoFisher) software.

PfPanK1 protein levels
To measure PfPanK1 levels, parasite samples were prepared from Parent and mutant cultures as previously described with minor modifications [35]. Briefly, mature trophozoites were saponin-isolated from a culture of infected erythrocytes (10% parasitaemia, 2% haematocrit in 30 mL) by resuspending the pellet in 1 × phosphate buffered saline (PBS; Sigma-Aldrich) containing 0.05% (w/v) saponin, 1 × complete mini protease inhibitor cocktail (Roche), 20 mM sodium fluoride and 0.1 mM sodium orthovanadate. The isolated trophozoites were pelleted (2,000 × g, 8 min), washed (15,850 × g, 30 s, supernatant discarded after each wash) once in 1 mL of the above solution excluding saponin, and twice with 1 mL 1 × PBS. The trophozoite pellet was then stored at −80˚C until required. For sample processing, a total of 500 μg of protein, accurately determined using the Pierce BCA protein assay kit (ThermoFisher), was incubated overnight with sequencing-grade trypsin (1:50 dilution; Promega). On the following day, trypsin activity was quenched using 5% formic acid (FA) and the detergent (sodium deoxycholate) used for protein solubilisation was removed. The samples were then dried and resuspended in 20 μL of 2% (v/v) acetonitrile (ACN) and 0.1% (v/v) FA for LC-MS/MS analysis. For facilitating retention-time (RT) alignments among samples, a RT kit [36] (iRT kit, Biognosys) was spiked into all samples (1:20 dilution).
LC-MS/MS was carried out as described previously [35], with minor modifications. Briefly, samples were loaded at a flow rate of 15 μL/min onto a reversed-phase trap column (100 μm × 2 cm; Acclaim PepMap media, Dionex), which was maintained at a temperature of 40˚C. Peptides were then eluted from the trap column at a flow rate of 0.25 μL/min through a reversed-phase capillary column (75 μm × 50 cm; LC Packings, Dionex). The HPLC gradient was set to 158 min using a gradient that reached 30% ACN after 123 min, then 34% ACN after 126 min, 79.2% ACN after 131 min and 2% ACN after 138 min, at which it was maintained for a further 20 min. The mass spectrometer was operated in a data-independent acquisition (DIA) mode with a 25-fixed-window setup of 24 m/z effective precursor isolation over the m/z range of 375-975 Da.
For Spectronaut processing, raw files were loaded into Spectronaut (version 11, Biognosys) with an in-house P. falciparum-infected red blood cell library and default settings. Briefly, RT prediction type was set to dynamic iRT and the correction factor for the window was set to one. Mass calibration was set to local mass calibration. Interference correction was on MS2 level. The false discovery rate was set to 1% at peptide precursor level. For quantification, the interference correction was activated and a cross run normalisation was performed using the total peak area as the normalisation base. A significance filter of 0.01 was used. The peptides used in the identification and quantitation of PfPanK1 and the four control proteins are listed in S2 Table. Statistical analysis Statistical analysis of means was carried out with unpaired, two-tailed, Student's t tests using GraphPad 6 (GraphPad Software, Inc) from which the 95% confidence interval of the difference between the means (95% CI) was obtained. All regression analysis was done using Sigma-Plot version 11.0 for Windows (Systat Software, Inc).

Pfpank1 mutations mediate parasite resistance to PanOH and CJ-15,801
The 3D7 P. falciparum strain was cloned through limiting dilution, and a single parasite line (henceforth referred to as the Parent line) was used to generate all of the subsequent lines tested in this study (unless otherwise specified). This was done to ensure that all of the parasite lines generated during the course of this study would share a common genetic background. Using the Parent line, three independent drug-pressuring cultures were set up (two with PanOH and one with CJ-15,801). When these parasites had attained approximately 8-fold decrease in sensitivity (~11-13 weeks of continuous pressuring), they were subsequently cloned by limiting dilution and maintained in the absence of drug pressure. In this manner, three parasite clones were generated: PanOH-A and PanOH-B were generated from the two independent PanOH-pressured cultures while CJ-A was generated from the CJ-15,801-pressured culture. The clones are significantly resistant (95% confidence interval (CI) exclude 0) to the pantothenate analogues they were pressured with. The 50% inhibitory concentration (IC 50 ) values of PanOH against PanOH-A and PanOH-B, and the IC 50 value of CJ-15,801 against CJ-A are approximately 7-8-fold higher than those measured against the Parent line (Fig 2A and 2B and S3 Table). Significant cross-resistance towards the other pressuring analogue was observed for these clones, as compared to the Parent line (95% CI exclude 0). The PanOH-A and PanOH-B lines were found to be 4-6-fold less sensitive to CJ-15,801 while CJ-A was found to be 13-fold less sensitive to PanOH (Fig 2A and 2B and S3 Table). To ensure that the clones did not develop a general drug-resistance phenotype during the selection process, we tested their sensitivity to chloroquine, an antiplasmodial with a mechanism of action that is unrelated to the parasite's CoA biosynthetic pathway [37]. We found that all of the drug-pressured lines have chloroquine IC 50 values that are indistinguishable from that of the Parent line (95% CI include 0; Fig 2C and S3 Table).
The PanOH and CJ-15,801 resistance phenotypes observed in the clones were stable for several months of continuous culture in the absence of the pressuring analogue (! 3 months), consistent with a genetic alteration in these parasites. To determine the mutation(s) responsible for these phenotypes, gDNA was extracted from each clone and subjected to whole genome sequencing. All of the drug-resistant clones were found to harbour a unique mutation in the putative pantothenate kinase gene, Pfpank1 (PF3D7_1420600), as shown in Fig 3A. Other non-synonymous mutations were detected for each clone (S4 Table) but we did not find another gene that is mutated in all three clones. The mutation found in the Pfpank1 of PanOH-A results in the substitution of Asp507 for Asn. The other two drug-resistant clones have a mutation at position 95 of the protein: the Pfpank1 of PanOH-B harbours a deletion of an entire codon leading to a loss of Gly from the PfPanK1 protein, while the PfPanK1 of CJ-A has a Gly to Ala substitution. Since the structure of PfPanK1 has not yet been resolved, we generated a three-dimensional model in order to map the mutations within the enzyme. To confirm that the resistance phenotypes observed for the clones are directly caused by the mutations in Pfpank1, each clone was transfected with an episomal plasmid (Pfpank1-stop-pGlux-1) that enables the parasites to express the wild-type copy of Pfpank1 (in addition to the endogenous mutated copy). These complemented lines are indicated with a superscripted "+-WTPfPanK1". From Fig 4 (and S3 Table), it can be observed that the complemented mutant clones (grey bars) are significantly less resistant to PanOH (Fig 4A) and CJ-15,801 (Fig 4B) compared to the non-complemented mutant clones (black bars; 95% CI exclude 0). As expected, the relative sensitivity of the mutant clones to chloroquine is unchanged by the presence of the PfPanK1-encoding plasmid (95% CI include 0; Fig 4C). Transfection of the PfPan-K1-encoding plasmid into the Parent line did not alter its sensitivity to PanOH, CJ-15,801 or chloroquine (95% CI include 0; S3 Table). These data are consistent with the mutations observed in Pfpank1 being responsible for the resistance phenotype observed in the mutant clones. An alternative strategy to confirm that the mutant PfPanK1 is responsible for the

Fig 2. Percentage proliferation of parasites from the Parent (white circles), PanOH-A (black triangles), PanOH-B (black squares) and CJ-A (black diamonds) lines in the presence of (a) PanOH, (b) CJ-15,801 or (c) chloroquine.
Drug-pressured lines were generated by exposing Parent line parasites to 11 − 13 weeks of continuous drug-pressuring with either PanOH (for PanOH-A and PanOH-B) or CJ-15,801 (for CJ-A), followed by limiting dilution cloning. Values are averaged from ! 4 independent experiments, each carried out in triplicate. All error bars represent SEM. Error bars are not visible if smaller than the symbols.
https://doi.org/10.1371/journal.ppat.1006918.g002 observed phenotypes would be to express the mutant forms of PfPanK1 in the Parent line. This has not, however, been attempted.

Pfpank1 mutations impair parasite proliferation
We performed a competition assay to determine whether the Pfpank1 mutations impart a fitness cost to the parasite clones. Each of the competition cultures was set up by mixing an equal number of parasites from the Parent line and one of the mutant clonal lines, and was maintained under standard conditions for a period of 6 weeks (Fig 5A). The sensitivity of each competition culture to PanOH was tested on the day the lines were mixed (Week 0) and again at the end of the 6-week period (Week 6). As expected, each Week 0 (dashed line) PanOH doseresponse curve is between those obtained for the Parent and the respective mutant clone (dotted lines). A shift of the dose-response curve obtained at week 6 (solid line) towards the doseresponse curve of the Parent line would indicate that the mutant PfPanK1 imparts a fitness cost on the clone. As shown in Fig 5B, the Week 6 curve for the PanOH-A competition culture only exhibits a marginal leftward shift from Week 0, whereas those for PanOH-B (Fig 5C) and CJ-A (Fig 5D) exhibits a more substantial shift, almost reaching the dose-response curve of the Parent line (dotted line, white circles). These trends are supported by qPCR analysis designed to determine the proportions of mutant versus Parent gDNA at week 0 and week 6 (S1

PfPanK1 is a functional pantothenate kinase located within the parasite cytosol
The phosphorylation of radiolabelled pantothenate by lysates prepared from each of the mutant clones and the Parent line was measured to determine if the mutations in the putative Pfpank1 gene affect PanK activity, and hence whether the gene codes for a functional PanK. As shown in Fig 6, at the end of the 75 min time-course, the lysate prepared from PanOH-A phosphorylated approximately 3 × more [ 14 C]pantothenate than the lysate prepared from the Parent line, while the lysate of PanOH-B generated about 3 × less phosphorylated [ 14 C]pantothenate compared to the Parent line. By comparison, the lysate prepared from CJ-A only produced a small amount of phosphorylated [ 14 C]pantothenate in the same time period. The inset in Fig 6 demonstrates that PanK activity could be detected in CJ-A lysates when the experiment was carried out in the presence of a 100-fold higher pantothenate concentration (200 μM) and for an extended time (420 min). These observations provide strong evidence that the Pfpank1 gene codes for a functional PanK. Further, we generated from the wild-type 3D7 strain a transgenic parasite line that episomally expresses a GFP-tagged copy of PfPanK1 in order to localise the protein within the parasite. We found that PfPanK1 is largely localised throughout the cytosol of trophozoite-stage parasites, and is not excluded from the nucleus (Fig 7). To investigate further the effects of the PfPanK1 mutations present in the PanOH-and CJ-15,801-resistant clones, we analysed the PanK activity profiles using lysates prepared from each mutant and the Parent, and determined their kinetic parameters from the Michaelis-Menten equation (Fig 8). The apparent pantothenate K m values of the mutant clones are 26-609-fold higher (95% CI exclude 0) than that of the Parent line. The maximal velocity (V max ) of pantothenate phosphorylation by lysates prepared from PanOH-A, PanOH-B and CJ-A are also significantly higher (95% CI exclude 0) than that of the Parent. To eliminate the possibility that the elevated V max values are due to increased PfPanK1 levels in the mutant parasites, we used DIA-MS to compare the PfPanK1 level in each of the mutant lines with that of the Parent parasite. We found that PfPanK1 levels are unchanged in the CJ-A line (95% CI include 0) and are only elevated by 26 ± 5% and 27 ± 4% (mean ± SEM) in PanOH-A and PanOH-B, respectively (S3 Fig). This modest (or lack of an) increase in PfPanK1 levels is insufficient to explain the larger increases in V max values observed in lysates prepared from the mutant lines, as the increase in protein levels would need to be~20-fold for PanOH-A and~2-fold for PanOH-B to account for the altered V max values. We have therefore not adjusted the observed V max values to account for these small increases in the PfPanK1 levels. We also calculated the PfPanK relative specificity constant for each parasite line, which indicates the catalytic efficiency of each variant of PfPanK relative to that of the Parent line. The relative specificity constant obtained for PanOH-A (0.74 ± 0.04, mean ± SEM) is not significantly different (95% CI include 0) from that of the Parent. However, those of PanOH-B (0.058 ± 0.004, mean ± SEM) and CJ-A (0.019 ± 0.005, mean ± SEM) are significantly lower (95% CI exclude 0). These data are consistent with all three PfPanK1 mutations observed reducing the enzyme's affinity for pantothenate, although the associated increase in the enzyme V max compensates for the reduced affinity: fully in the PanOH-A clone, but to a much lesser extent in PanOH-B and CJ-A (resulting in a 17-fold and 52-fold reduction in the enzyme's catalytic efficiency, respectively).

CJ-A requires a higher extracellular pantothenate concentration
An extracellular supply of pantothenate is essential for the in vitro proliferation of the intraerythrocytic stage of P. falciparum [8]. Given the impact that the PfPanK1 mutations have on PanK activity (Fig 8), we set out to determine whether a higher extracellular concentration of pantothenate is required to support the proliferation of the different mutant clones relative to that required by the Parent line. As observed in Fig 9A, the proliferation of the Parent line (white circles) increased as the extracellular pantothenate concentration was increased, reaching the 100% control level (parasites maintained in the presence of 1 μM pantothenate, the concentration of pantothenate in the complete RPMI medium used to maintain all of the parasite cultures, which is within the physiologically relevant range in human blood [38,39]) at approximately 100 nM. In order to compare the extracellular pantothenate requirement between the different lines, we determined the SC 50 (50% stimulatory concentration; i.e. the concentration of pantothenate required to support parasite proliferation to a level equivalent to 50% of the control level) values for the mutants (with and without complementation) and Parent. From Fig 9B, it can be seen that the SC 50 values of PanOH-A and PanOH-B are not different from that of the Parent line (95% CI include 0). Conversely, as illustrated by the rightward shift in its dose-response curve (black diamonds, Fig 9A), the pantothenate SC 50 of CJ-A is approximately 3-fold higher than that of the Parent (Fig 9B; 95% CI = 2.7 to 30.9). Furthermore, consistent with the data from the complementation experiments (Fig 4), the SC 50 value of CJ-A +WTPfPanK1 (grey diamonds, Fig 9A) is comparable to that of the Parent line and also the control line, Parent +WTPfPanK1 (Fig 9B; 95% CI include 0), indicating that the episomal expression of wild-type PfPanK1 is sufficient to reverse the phenotypic effects of the mutation.

N5-trz-C1-Pan and N-PE-αMe-PanAm have a different mechanism of action to PanOH and CJ-15,801
To determine whether the resistance of the clones to PanOH and CJ-15,801 extends to other pantothenate analogues, we tested the mutant clones against the two recently-described,

Fig 8. PfPanK activity profiles derived from the initial rates of pantothenate phosphorylation by lysates generated from Parent (white circles), PanOH-A (black triangles), PanOH-B (black squares) and CJ-A (black diamonds) parasites at various pantothenate concentrations.
The grey diamond indicates a data point that was outside the 95% confidence band when it was included in the non-linear regression. This data point was therefore deemed an outlier and was omitted from the data set used to generate the fitted curve shown. Values are averaged from 3 independent experiments, each performed with a different batch of lysate and carried out in duplicate. Error bars represent SEM and are not visible if smaller than the symbols. Relative specificity constant (rsc) values express the specificity constants obtained for each cell line relative to that of the Parent. The V max (μmol/(10 12 cells.h)), K m (μM) and rsc values determined from the curves are shown within each box. Errors represent SEM (n = 3). An asterisk indicates that the value is significantly different compared to the Parent line (V max 95% CI compared to Parent: PanOH-A = 173 to 344, PanOH-B = 5. modified pantothenamides with potent antiplasmodial activities, namely N5-trz-C1-Pan [20] and N-PE-αMe-PanAm [21]. We found that PanOH-A is 3-fold more sensitive to N5-trz-C1-Pan, PanOH-B is 2-fold more resistant and CJ-A is 9-fold more resistant when compared to the Parent line (95% CI exclude 0; Fig 10A and S5 Table, left side). Similarly, we found that relative to the Parent line, PanOH-A was more sensitive to N-PE-αMe-PanAm (~2-fold), while CJ-A is 2-fold more resistant (95% CI exclude 0; Fig 10B and S5 Table, left side). The sensitivity of PanOH-B to N-PE-αMe-PanAm was statistically indistinguishable from that of the Parent line (95% CI = -0.046 to 0.005; Fig 10B and S5 Table, left side). These results indicate that PfPanK1 can influence the sensitivity of the parasite to multiple antiplasmodial pantothenate analogues. Remarkably, the mutation at position 507 of the PfPanK1 in PanOH-A makes the parasite resistant to the antiplasmodial activity of certain pantothenate analogues (PanOH and CJ-15,801) whilst at the same time hyper-sensitises the parasite to pantothenate analogues of a different class (modified pantothenamides, N5-trz-C1-Pan and N-PE-αMe-PanAm).
Previous work has shown that, in bacteria, pantothenamides are metabolised by the CoA biosynthetic pathway to form CoA antimetabolites [40], consistent with PanK activity being important for metabolic activation of pantothenamides. Additionally, it has been reported recently that pantothenamides are also phosphorylated by the PanK in P. falciparum [41], in line with their metabolic activation in bacteria. We therefore set out to determine whether the modified, pantetheinase-resistant, pantothenamides are metabolised and to what extent. In order to do so, we exposed intact Parent line P. falciparum-infected erythrocytes to N5-trz-C1-Pan (at~10 × the IC 50 for 4 h) and subjected lysates from the treated (and control) samples Lastly, we investigated the ability of the different pantothenate analogues to inhibit the phosphorylation of [ 14 C]pantothenate by parasite lysates prepared from the various mutant clones. These data are shown in Fig 10C-10F and S5 Table, right side. All of the analogues tested are significantly less effective (95% CI exclude 0) at inhibiting pantothenate phosphorylation by lysates generated from the mutant clones compared to their ability to inhibit pantothenate phosphorylation by lysates prepared from the Parent line. The exception being N-PE-αMe-PanAm, which did not reach statistical significance when tested against lysate prepared from PanOH-A. Additionally, the effectiveness of the analogues at inhibiting pantothenate phosphorylation by lysates prepared from the mutant lines all observe the following order:

The putative Pfpank1 gene codes for a functional pantothenate kinase
When PfPanK1 is compared to type II PanKs from other organisms in a multiple protein sequence alignment, the three nucleotide-binding motifs characteristic of this superfamily can be seen to be conserved in PfPanK1, consistent with it being a functional PanK [7]. However, biochemical confirmation of its putative function as a PanK has not been demonstrated. In this study, we show that mutations in the putative PfPanK1 lead to substantial changes in pantothenate kinase activity (Fig 8) providing, for the first time, biochemical evidence that the cytosolic PfPanK1 (Fig 7) is a functional pantothenate kinase. Furthermore, the fact that multiple independent experiments aimed at generating parasites resistant to PanOH and CJ-15,801 always selected for mutations in PfPanK1 (Fig 3A and S4 Table) is consistent with the kinase being the primary PanK involved in the metabolic activation of pantothenate analogues, at least during the intraerythrocytic stage. Although it is intriguing that PfPanK1 is not excluded from the nucleus, this phenomenon has been reported for some PanKs in other organisms [42].
Although it is clear that the PfPanK1 residues at position 95 and 507 are required for normal PfPanK1 activity (Fig 8), and the PfPanK1 model structure (Fig 3B) shows that both residues are situated adjacent to the enzyme active site, their exact role(s) in modifying the activity of the protein is less obvious. The Gly residue at position 95 is conserved in eukaryotic PanKs [7] and is the residue at the cap of the α2-helix in the inactive conformation of the protein (S5A Fig). One possibility is that the change to Ala at this position could affect the structure of the helix and consequently the overall stability of the protein (at least when the protein is in the inactive state), as Gly has been shown to be much better than Ala at conferring structural

PanAm. Symbols represent Parent (white circles), PanOH-A (black triangles), PanOH-B (black squares) and CJ-A (black diamonds) parasite lines. The Y-axes of c−f indicate the percentage of total pantothenate phosphorylation.
Values are averaged from ! 3 independent experiments, each carried out in triplicate for the parasite proliferation assays and duplicate for the phosphorylation assays. Error bars represent SEM and are not visible if smaller than the symbols.
https://doi.org/10.1371/journal.ppat.1006918.g010 stability when located at the caps of helices [43]. Alternatively, Gly residues have been shown to be present at a higher frequency in the active sites of some enzymes where they likely confer the flexibility to alternate between open and closed conformations [44]. Although the Gly95 residue is not part of the PfPanK1 active site per se, it is within close proximity to the site (Fig  3B), and the α2-helix certainly undergoes a conformational change when PanK switches from the inactive conformation to the active one [29]. This conformational change is demonstrated in S5A Fig, which shows an overlay of the human PanK3 crystal structures in the active [29] and inactive [45] state (Gly95 in PfPanK1 is equivalent to Gly117 in the human enzyme). It is worth noting that acetyl-CoA (an inhibitor of the enzyme) can only be accommodated in the binding site when the enzyme is in the inactive state, as when the enzyme is in the active state, the α2-helix encroaches on the space occupied by acetyl- CoA (S5A Fig). This lends further support to the importance of this helix for the enzyme to transition from the inactive to the active state (and vice versa) and, therefore, to the role that Gly95 could play in this process. Either way, these suggestions may explain why a mutation at this position has a greater impact on PfPanK1 function than the mutation at position 507. The Asp residue at position 507 is replaced by a different amino acid (Glu) in most other eukaryotic PanKs, although they are both negatively-charged [7]. A substitution to the uncharged Asn could disrupt any important salt-bridges or hydrogen bonds with the residue at this position. In human PanK3, the amino acid equivalent to Asp507 is Glu354. In the enzyme's inactive state, Glu354 is within ionic bonding distance of Arg325, which in turn interacts with the 3'-phosphate of acetyl-CoA (S5B Fig) [45] and may therefore stabilise the inactive state of the enzyme. Conversely, in the enzyme's active state, Glu354 and Arg325 are not within ionic bonding distance (S5B Fig). If Asp507 in PfPanK1 plays a similar role to that proposed for Glu354 in human PanK3, the change at position 507 to an Asn (abolishing the negative charge) could prevent stabilisation of the inactive state, providing an explanation for the increased activity observed in PanOH-A (Fig 6). Determining the crystal structure of PfPanK1 bound with pantothenate may provide a better understanding of the roles these residues play in PanK function.

PfPanK1 is essential for normal intraerythrocytic proliferation of P. falciparum
It has been established that almost all of the 4'-phosphopantothenate found in P. falciparuminfected erythrocytes is generated within the parasite by PfPanK as part of its metabolism into CoA [10], which is in line with PfPanK activity being essential for the parasite's survival. In the present study, our inability to knock out PfPanK1 (S2 Fig) is consistent with the protein being essential for the intraerythrocytic stage of P. falciparum, although we cannot exclude the unlikely possibility that the regions we targeted for the required double-crossover recombination event are genetically intractable. Our observation that clones PanOH-B and CJ-A, which harbour mutations at position 95 of PfPanK1, can be outcompeted by the Parent parasites in competition assays over approximately 20 intraerythrocytic cycles (Fig 5) is consistent with the mutations incurring a fitness cost. This, in turn, indicates that PfPanK1 is essential for normal parasite development, at least during the blood stage of its lifecycle. It was also found here that clone CJ-A requires an approximately 3-fold higher extracellular pantothenate concentration in order to survive (Fig 9), coinciding with this clone having the PfPanK with the highest K m . This is not surprising given the reported importance for the substrate concentration to exceed the enzyme K m for optimal enzyme efficiency [46,47]. More importantly, the requirement by this clone for a higher concentration of extracellular pantothenate is also congruent with PfPanK1 being essential for the normal development of P. falciparum during its asexual blood stage.
Our observation that PfPanK1 is essential for normal parasite growth during the blood stage is at odds with recent reports that show both PanK1 and PanK2 from Plasmodium yoelli and Plasmodium berghei are dispensable during the blood stage of those parasites [48][49][50]. However, both of these murine malaria parasite species preferentially infect reticulocytes [51,52]. Unlike the mature erythrocytes preferred by P. falciparum, reticulocytes have been shown to provide a rich pool of nutrients for the parasite, allowing the murine parasites to survive metabolic or genetic changes that would have been deleterious in P. falciparum [53]. It is therefore conceivable that unlike the condition faced by P. falciparum, the reticulocyte-residing parasites are able to salvage sufficient CoA or CoA intermediates from the host cell for their survival, rendering the two PanK proteins dispensable during their intraerythrocytic stage, a possibility acknowledged by the authors of the P. berghei study [49].

PanOH and CJ-15,801 share a common mechanism of action
We have presented data consistent with the observed PfPanK1 mutations being the genetic basis for the PanOH and CJ-15,801 resistance phenotypes observed in all of the drug-pressured clones we generated (Fig 4). As shown by the data presented in Fig 10, both PanOH and CJ-15,801 inhibited pantothenate phosphorylation by the mutated PfPanK1 proteins less effectively than their inhibition of pantothenate phosphorylation by the wild-type PfPanK1 (the order of their IC 50 values is Parent < PanOH-A < PanOH-B < CJ-A). Importantly, this order is also reflected in the level of resistance of the mutant clones to these two analogues, although the magnitude is not preserved (Fig 2 and S3 Table). These data are consistent with PfPanK1 being involved in the antiplasmodial activity of these analogues either as a target or a metabolic activator. Previous studies have demonstrated that the antiplasmodial activity of both PanOH and CJ-15,801 involves the inhibition of pantothenate phosphorylation by PfPanK [9,15]. More specifically, PanOH has been shown to inhibit PfPanK-mediated pantothenate phosphorylation by serving as its substrate [54]. CJ-15,801 is also likely to be a substrate of PfPanK, especially since it has been shown to be phosphorylated by the S. aureus PanK [55], another type II PanK [56]. In addition, the same study showed that the second enzyme in the CoA biosynthetic pathway, PPCS, subsequently accepts phosphorylated CJ-15,801 as a substrate, and performs the first step of the PPCS reaction (cytidylylation) on it. However, unlike what happens with 4'-phosphopantothenate, this cytidylylated phospho-CJ-15,801 acts as a tightbinding, dead-end inhibitor of the enzyme [55]. A separate study also concluded that PanOH targets the PPCS enzyme in Escherichia coli and Mycobacterium tuberculosis [57]. Based on the data generated in this study and the published reports that PanOH and CJ-15,801 both inhibit PPCS in other systems, we propose a similar mechanism of action for these compounds in P. falciparum, whereby they are phosphorylated by PfPanK1 and subsequently block PfPPCS as dead-end inhibitors (Fig 11). The overlapping pattern of crossresistance between the two compounds (Fig 2) is also in line with them having a similar mechanism of action. We propose that the observed resistance to PanOH and CJ-15,801 is due to the mutated PfPanK1 having a reduced capacity to phosphorylate these analogues relative to pantothenate. This would have the effect of reducing the amount of phosphorylated PanOH or CJ-15,801 generated relative to 4'-phosphopantothenate, thereby allowing the parasites to survive at higher concentrations of the drugs. Furthermore, our observation that the mutant clones have comparable levels of PanOH and CJ-15,801 resistance (Fig 2  and S3 Table), despite having PfPanK1 proteins of vastly different efficiency (Fig 8), is likely due to the pathway flux control at the PfPPCS-mediated step in the CoA biosynthetic pathway of P. falciparum, as shown in a previous study [32].

N5-trz-C1-Pan is converted into an antiplasmodial CoA analogue -N-PE-αMe-PanAm likely shares the same mechanism of action
N5-trz-C1-Pan and N-PE-αMe-PanAm are pantothenamide-mimics that harbour modifications designed to prevent them from being substrates of pantetheinase, thereby preventing their degradation: N-PE-αMe-PanAm is methylated at the α-carbon [21] while N5-trz-C1-Pan harbours a triazole instead of the labile amide [20]. Our LC-MS data clearly show that N5-trz-C1-Pan is converted into a CoA antimetabolite (S4 Fig), and it is therefore likely to go on to inhibit/inactivate CoA-utilising enzymes, killing the parasite. Such a mechanism has previously been put forward to explain the antibiotic activity of two prototypical pantothenamides (N5-Pan and N7-Pan) whereby the compounds are phosphorylated by PanK and subsequently metabolised by PPAT and DPCK to generate analogues of CoA (ethyldethia-CoA and butyldethia-CoA) [40,58,59]. These CoA analogues then mediate their antibacterial effect/s primarily by inhibiting/inactivating CoA-requiring enzymes and acyl carrier proteins [40,58,59]. Similarly, phospho-N5-trz-C1-Pan is not expected to interact with PfPPCS because it lacks the carboxyl group (Fig 1) required for the nucleotide activation by nucleotide transfer [55]. It is therefore expected to bypass the PfPPCS and PfPPCDC steps of the CoA biosynthetic pathway on its way to being converted into the CoA antimetabolite version of N5-trz-C1-Pan (Fig 11).
Furthermore, the antiplasmodial activity rank order of N5-trz-C1-Pan against the various mutant clones is very similar to that of N-PE-αMe-PanAm -with PanOH-A being hypersensitive to both compounds, CJ-A resistant to both and PanOH-B, by comparison, exhibiting only small changes in sensitivity to the two compounds (Fig 10A and 10B) -and is starkly different to those of PanOH and CJ-15,801 (Fig 2). These are congruent with (i) the antiplasmodial mechanism of action of N-PE-αMe-PanAm being similar to that of N5-trz-C1-Pan and (ii) the antiplasmodial mechanism of action of N-PE-αMe-PanAm and N5-trz-C1-Pan being different to that of PanOH and CJ-15,801.
The order of antiplasmodial activity of N5-trz-C1-Pan and N-PE-αMe-PanAm against the mutant clones can be explained on the basis of (i) the difference in the rate of PfPanK1 activity in the various clones at the concentrations of pantothenate and N5-trz-C1-Pan / N-PE-αMe-PanAm used (Fig 6), (ii) the fact that the PfPPCS-mediated step imposes pathway flux control [32] and (iii) the fact that this pathway flux control is bypassed by N5-trz-C1-Pan (and almost certainly also by N-PE-αMe-PanAm) en route to its conversion into a CoA antimetabolite. As seen in Fig 6, in the presence of 2 μM pantothenate (a similar concentration to the 1 μM present in the antiplasmodial assay), the pantothenate phosphorylation rate of the different clones has the following rank order: PanOH-A > Parent > PanOH-B > CJ-A, approximately the inverse of the antiplasmodial IC 50 values of N5-trz-C1-Pan and N-PE-αMe-PanAm (described above). Therefore, PanOH-A, for example, would be expected to generate more 4'-phosphopantothenate and phosphorylated N5-trz-C1-Pan (or N-PE-αMe-PanAm), based on the assumption that the mutation also leads to increased phosphorylation activity towards the pantothenate analogues (point i). Whilst the increased levels of 4'-phosphopantothenate would not be expected to result in a concomitant increase in CoA (due to the pathway flux control at PfPPCS; point ii), the increased production of phospho-N5-trz-C1-Pan would result in increased levels of the N5-trz-C1-Pan CoA antimetabolite as the flux control step is bypassed (point iii). This would explain the increased sensitivity of PanOH-A to both N5-trz-C1-Pan and N-PE-αMe-PanAm and also the sensitivity rank order of the other parasite lines.
In conclusion, our study confirms for the first time that PfPanK1 functions as the active pantothenate kinase in the asexual blood stage of P. falciparum. Our data show that the sites of mutation in PfPanK1 reported here are important residues for normal PfPanK function and are essential for normal intraerythrocytic parasite growth, although further structural and functional studies are required to elucidate their exact role(s). Furthermore, we propose that following phosphorylation by PfPanK1, PanOH and CJ-15,801 compete with 4'-phosphopantothenate and serve as dead-end inhibitors of PfPPCS (depriving the parasite of CoA). In contrast, N-PE-αMe-PanAm and N5-trz-C1-Pan are further metabolised (by PfPPAT and PfDPCK) into CoA analogues that kill the parasite by inhibiting CoA-utilising metabolic processes. Finally, we provide the first genetic evidence consistent with pantothenate analogue activation being a critical step in their antiplasmodial activity.

S1 Materials and methods. Additional information about materials and methods used in the study.
(DOCX) S1  Table. List of non-synonymous single nucleotide polymorphisms found in the PanOH-A, PanOH-B and CJ-A parasite lines as determined by whole genome sequencing and variant calling using the PlaTyPus integrated pipeline. Uppercase letters in the "Codon Change" column denotes the base within the codon that has been altered in the coding sequence before/after the mutation. The deletion in PanOH-B at position 95 is not included here because PlaTyPus is unable to detect insertions-deletions polymorphisms. (a) Schematic representations of the ΔPfpank1-pCC-1 construct and the wild-type Pfpank1 gene locus before and after either a homologous double crossover integration of the hdhfr cassette of ΔPfpank1-pCC-1 or a 5'/3' single crossover homologous integration of the ΔPfpank1-pCC-1 construct. The positions of AflII restriction sites are indicated. The 5' Pfpank1 and 3' Pfpank1 homologous flanks are indicated by translucent blue boxes. (b) Southern blot of AflII-digested gDNA extracted from wild-type parasites and from ΔPfpank1-pCC-1-transfectants resistant to WR99210 or both WR99210 and 5-fluorocytosine (5-FC) after one or two rounds of WR99210 cycling. (i) and (ii) represent independently-selected drug-resistant cultures. The blot was probed with the 3' Pfpank1 flank (as indicated by the black bar in (a)). The probe hybridised to fragments that correspond to the 3D7 wild-type (3.3 kb) and the plasmid (8.9 kb). The fragment that is consistent with a homologous double crossover-disrupted locus (3.8 kb) was not detected in either independent culture. hdhfr: human dihydrofolate reductase-conveys resistance to WR99210. Scfcu: Saccharomyces cerevisiae cytosine deaminase/phosphoribosyl transferase-conveys sensitivity to 5-FC. (c) PCR confirmation that a Pfpank1 knockout event cannot be detected, even in a small sub-population of parasites. Samples from the same parasite populations shown in (b) were used to generate DNA templates that were then used with primers selected to amplify a product of 1.9 kb from the wild-type sequence and a 3.6 kb product from parasites with the Pfpank1 gene knocked out (and replaced with the hdhfr gene). A PCR product consistent with the size expected for amplification from the wild-type sequence was observed (black arrow). The identity of this product was confirmed by sequencing. All the reactions generated an approximately 1.1 kb PCR product and some reactions, under certain conditions, produced fainter PCR products likely due to non-specific primer binding. Importantly, no product of a size consistent with that expected for a Pfpank1 knockout (3.6 kb) was observed in any of the samples. These data, therefore, are consistent with those presented in the Southern blot (b).