Functional genetic evaluation of DNA house-cleaning enzymes in the malaria parasite: dUTPase and Ap4AH are essential in Plasmodium berghei but ITPase and NDH are dispensable.

ABSTRACT Background: Cellular metabolism generates reactive oxygen species. The oxidation and deamination of the deoxynucleoside triphosphate (dNTP) pool results in the formation of non-canonical, toxic dNTPs that can cause mutations, genome instability, and cell death. House-cleaning or sanitation enzymes that break down and detoxify non-canonical nucleotides play major protective roles in nucleotide metabolism and constitute key drug targets for cancer and various pathogens. We hypothesized that owing to their protective roles in nucleotide metabolism, these house-cleaning enzymes are key drug targets in the malaria parasite. Methods: Using the rodent malaria parasite Plasmodium berghei we evaluate here, by gene targeting, a group of conserved proteins with a putative function in the detoxification of non-canonical nucleotides as potential antimalarial drug targets: they are inosine triphosphate pyrophosphatase (ITPase), deoxyuridine triphosphate pyrophosphatase (dUTPase) and two NuDiX hydroxylases, the diadenosine tetraphosphate (Ap4A) hydrolase and the nucleoside triphosphate hydrolase (NDH). Results: While all four proteins are expressed constitutively across the intraerythrocytic developmental cycle, neither ITPase nor NDH are required for parasite viability. dutpase and ap4ah null mutants, on the other hand, are not viable suggesting an essential function for these proteins for the malaria parasite. Conclusions: Plasmodium dUTPase and Ap4A could be drug targets in the malaria parasite.


Introduction
Malaria is a devastating parasitic disease killing around 400,000 people every year, mostly young children. The occurrence of drug resistance and the absence of an efficient vaccine demand a better understanding of the parasite's biology to aid the identification of suitable and novel drug and vaccine targets [1][2][3][4]. Plasmodium, the causative agent of the disease, is haploid throughout most of its life cycle. This protozoan uses multiple rounds of endomitosis followed by cellular morphogenesis to produce unicellular progeny in a process called schizogony [5]. This strategy is employed by the parasite to increase the population at specific time points during its life cycle: in the red blood cell a single P. falciparum merozoite produces up to 32 daughter cells [6]; following transmission to the mosquito an oocyst can generate hundreds of sporozoites [7] while in the liver thousands of merozoites emerge from a hepatocyte infected with a single sporozoite [8]. During sexual development, the male gametocyte produces eight motile gametes in a matter of minutes following transmission to the mosquito vector [9].
The haploid nature of the Plasmodium genome and its rapid DNA replication cycles perhaps pose an increased risk for the misincorporation of non-canonical nucleotides into the DNA; yet the enzymes preventing such DNA damage in the malaria parasite are largely unknown. Intracellular nucleotide pools are essential DNA building blocks, serve as energy storage molecules, and act as cofactors and regulators in various metabolic and signal transduction pathways. Reactive oxygen species produced during cellular redox reactions can oxidize the DNA precursor pool and produce so-called non-canonical deoxynucleoside triphosphates (dNTPs). For example oxidative deamination of the nitrogenous base produces deoxyinosine triphosphate (dITP) from deoxyadenosine triphosphate (dATP), deoxyuridine triphosphate (dUTP) from deoxycytosine triphosphate (dCTP) and deoxyxanthosine triphosphate (dXTP) from deoxyguanosine triphosphate (dGTP). Thymine lacks a free amino group and therefore is not oxidatively deaminated. Oxidation of purine bases on the other hand results in nucleotides like 2-oxo-dATP, 8-oxo-dATP, and 8-oxo-dGTP. Such modified nucleotide analogs can be incorporated into DNA during replication [10,11]. When recognized by the DNA repair system, single strand breaks are introduced which need repairing. As a consequence, cellular growth can slow down or the cell may die due to the accumulation of deleterious mutations and double strand breaks [11].
To prevent DNA damage, cells have evolved housecleaning or DNA sanitation enzymes to hydrolyze toxic nucleotides into their non-toxic monophosphate forms, which are poor substrates for their respective nucleoside kinases and thus not phosphorylated to their toxic nucleotide forms. Based on structural features, sanitation enzymes belong to four superfamilies: ITPases (inosine triphosphate pyrophosphatase) are defined by the Ham1 (6-n-HydroxylAMinopurine sensitive) domain consisting of a long central beta-sheet forming the floor of the active site [12]; dUTPases (deoxy-uridine triphosphatase) which present a trimeric fold, each with an eight-stranded jelly-roll beta barrel [13]; the NuDiX (nucleotide diphosphate linked to some other moiety, X) superfamily is defined by the NuDiX box domain G-x(5)-E-x(5)-[UA]x-R-E-x(2)-E-E-x-G-U, where U is a hydrophobic residue and x any residue [14]; and finally, the fourth group contains all-α-helical NTPases which are active against dNTPs as well as dNDPs with the generation of dNMPs as an end product [15].
The four groups share no sequence homology and are specific for their respective substrates. ITPases hydrolyze ITP and XTP as well as their deoxy forms (dITP, dXTP) into their respective monophosphates [16,17], which prevents the incorporation of dITP and dXTP into DNA. Escherichia coli and yeast mutants that lack ITPase suffer increased sensitivity to the mutagen 6-N-hydroxyaminopurine (HAP) due to the mispairing of dCTP or dTTP leading to A:T to G:C and G:C to A:T transitions [18][19][20]. itpase null mutant mice die within 2 weeks of birth [21].
dUTPase, on the other hand, is selective for the deoxynucleotide form and specifically hydrolyzes dUTP to dUMP [22]. Certain organisms have lost the dUTPase gene [23] but the enzyme is essential for E. coli [24], Saccharomyces cerevisiae [25], Trypanosoma brucei [26], Mycobacterium smegmatis [27], Candida albicans [28] and Drosophila melanogaster [29]. The essential nature of human dUTPase has yet to be established, but Koehler et al. have recorded increased drug sensitivity upon dutpase RNA interference [30]. In Plasmodium, dUTPase also provides the sole source for dTMP and dTTP as, unlike other organisms, the parasite lacks dCMP deaminase, which catalyzes dUMP synthesis from dCMP [31] ( Figure 1). It is unknown whether the protein is essential for the intraerythrocytic developmental cycle of the malaria parasite, but selective inhibitors for P. falciparum dUTPase inhibit the enzyme in vitro as well as parasite growth in culture with a good correlation between the enzyme inhibition and cellular assays [32]. The cocrystallized dUTPase-inhibitor structure reveals an atypical trimeric enzyme suggesting the role of this enzyme as an antimalarial drug target [33].
NuDiX (nucleoside diphosphate linked to some other moiety, X) family members can hydrolyze a variety of substrates ('X') including dinucleoside polyphosphates, nucleotide-sugars, capped RNA and dinucleotide coenzymes [34]. The E. coli NuDiX enzyme MutT hydrolyzes both 8-oxo-GTP and 8-oxo-dGTP and thus performs house-cleaning at both DNA and RNA level. In mutT depleted E. coli a 100 to 10,000 times increase in occurrence of AT to CG transversions was reported [34,35] and this unidirectional mutation causes GC enrichment in chromosomal DNA [36]. The human protein NUDT1/MTH1, which catalyzes the same reactions, is key to the survival of cancer cells [37].
The haploid nature of the Plasmodium genome and its rapid DNA replication cycles pose perhaps an increased risk for DNA damage; yet the enzyme repertoire preventing the incorporation of non-canonical dNTPs into the DNA of the malaria parasite is largely unknown and its requirement for cell viability has not been explored. Here, we bioinformatically identify sanitation enzyme homologs in the rodent malaria model P. berghei andthrough reverse geneticsdefine their importance for parasite life cycle progression in the mouse host and mosquito vector.

Ethics statement
All animal experiments were performed according to the FELASA and GV-SOLAS standard guidelines. German authorities (Regierungspräsidium Karlsruhe) approved the animal experiments.

Bioinformatics analyses
Human nucleotide sanitation enzyme sequences were retrieved from the Universal Protein Resource (http://www. uniprot.org/) repository (Table 1) and searched against the P. berghei ANKA proteome at NCBI-BLAST (https://blast.ncbi. nlm.nih.gov/Blast.cgi) using the Domain Enhanced Lookup Time Accelerated-Basic Local Alignment Search Tool (DELTA-BLAST) of protein-BLAST search. The default algorithm parameters used in all searches were as follows: algorithm: DELTA-BLAST (Domain Enhanced Lookup Time Accelerated BLAST); max target sequence: 500; expected threshold: 10; word size: 3; scoring matrix: BLOSUM62; threshold: 0.05.
Percentage identity matrices were performed at https://www. ebi.ac.uk/Tools/msa/clustalo/. The information for ortholog groups for dUTPase and ITPase was retrieved from the PlasmoDB database (www.PlasmoDB.org). Both ortholog groups were individually searched in the OrthoMCL (http://orthomcl. org/orthomcl/) database to find similar sequences in the genomes of human, yeast and of various Plasmodium species. The orthologs in Plasmodium and human were retrieved in fasta format and submitted online to Clustal Omega (http://www.ebi. ac.uk/Tools/msa/clustalo/) to perform multiple sequence alignments. Clustal Omega uses seeded guide trees and hidden Markov model (HMM) profile-profile techniques to perform a global alignment of sequences. The input protein sequences were submitted with the following parameters: dealign input sequences: no, mbed-like clustering guide-tree: yes, mbed-like clustering iteration: yes, number of combined iterations: 0, max guide tree iterations: −1, max hmm iterations: −1, order: input. The percentage identity matrices and globally aligned sequences were downloaded.

GFP-tagging of Plasmodium berghei house-cleaning genes
We generated transgenic parasite lines expressing full-length C-terminally GFP-tagged proteins from episomal plasmids (dUTPase and ITPase) or through the integration of linearized vectors into the corresponding endogenous loci (Ap4AH and NDH) as outlined in (Figure 2(d-g)). P. berghei WT parasites were transfected as described [40]. Protein expression was monitored by microscopy and Western blotting. For fluorescence microscopy, a drop of tail blood was mixed with Hoechst 33,342 on a slide that was immediately covered with the coverslip and visualized under Zeiss 200 M Axiovert widefield (63x)  microscope. Image processing was performed with ImageJ. For Western blotting blood-stage parasites were released from red blood cells with saponin and resuspended in RIPA buffer (5 M NaCl, 0.5 M EDTA pH 8, 1 M Tris pH 8, 1% NP-40, 10% Nadeoxycholate, 10% SDS, dH 2 O) containing protease inhibitor (complete, Mini, EDTA-free, Roche). The mixture was loaded on to a Mini-Protean TGX Gel, 4-15% (BioRad) for electrophoresis and transferred onto a nitrocellulose membrane using the Trans-Blot Turbo Transfer System (BioRad). The membrane was blocked with 5% milk in Tris (pH 7.5) supplemented with 1% Tween 20 (d-g). GFP-tagging approach of house-cleaning enzymes using circular plasmids (d, e) or linear DNA (f, g) for genome integration (left panel). Middle panel: live fluorescence expression images in the indicated blood stage parasites depicting merged Hoechst DNA and GFP signal; scale bar = 5 μm. Right panels: western blot analyses of blood stage protein samples. Predicted sizes of GFP-fusion proteins are given below each blot in kDa. The expected size corresponds to individual protein +27 kDa (GFP-tag). We observed no protein degradation in the two essential enzymes i.e. dUTPase and AP4AH. We suspect this degradation (also the smear in ITPase) appeared due to the in vitro preparatory steps for the western-blots. Importantly we find full-length expression for each fusion protein. Please note that control blots (WT parasites and GFP alone) are not shown in figure but were used in all experiments.
(TBST) for 2 h at RT and probed with an anti-GFP antibody (1:5000 dilution, Clontech Laboratories) diluted in 5% milk in TBST overnight at 4°C. The following day, the membrane was washed 3 times with TBST and incubated with anti-mouse (1:10,000 dilution, GE Healthcare) secondary antibody in 5% milk in TBST for 1 h at RT. Proteins were detected using ECL substrate (Pierce).

RT-PCR analyses
Total RNA was extracted from different stages of the Plasmodium life cycle and cDNA was prepared according to manufacturer's instructions (Invitrogen). Transcripts were detected with the oligonucleotide primers that bind specifically to the gene. All primer details are given in Supplementary Table 1.

Generation of Plasmodium berghei knockout parasites
In order to delete individual P. berghei house-cleaning genes (dUTPase: PBANKA_0921300, ITPase: PBANKA_0618300, Ap4AH: PBANKA_1235300, NDH: PBANKA_1361900), 5ʹUTR and 3ʹUTR regions were amplified from genomic DNA of mixed blood stages of P. berghei strain ANKA by polymerase chain reaction (PCR) using Phusion High-Fidelity DNA Polymerase (New England Biolabs, USA) with primer sets as shown in Table S3; fragments were digested with restriction endonucleases as indicated in Table S3 and ligated either side of the human dihydrofolate reductase or a yeast bifunctional enzyme cytosine deaminase fused to an uridyl phosphoribosyl transferase (hdhfr-yFCU) as selection cassettes [41]. Linearized vectors were transfected into purified schizonts of P. berghei ANKA using standard protocols [40]. Transfected parasites were selected with pyrimethamine dissolved in drinking water (conc. 0.07 mg/mL) starting at 24 h post transfection. Parasitemia was monitored from day 7 onwards. Parasites were harvested above 2% parasitemia and cloned by limiting dilution for further genotype and phenotype analysis. The integration of the vector into the genome was confirmed by PCR using primer sets as shown in Supplementary Table 1.

Analyses of parasite viability
Pyrimethamine-selected mutant parasites were cloned and genotyped as described [40]. To test for blood stage development, an NMRI mouse infected with cloned parasites was sacrificed to collect blood at a parasitemia of 0.2-0.5%. The blood was diluted with PBS and 100 µL of the diluted blood containing 5000 infected red blood cells was injected into a C57BL/6 mouse (in a group of 4). The parasitemia was measured daily by counting Giemsa stained thin blood smears. The experiment was repeated on three separate occasions.
To quantify mosquito stage development, naïve Anopheles stephensi mosquitoes were allowed to blood-feed on infected mice. Mosquitoes were dissected on day 16-post infection to examine the midgut oocyst loads. The midguts were stained with 1% mercurochrome and the oocysts were counted per infected mosquitoes. On the same day, mosquitoes were dissected to isolate salivary glands from which in turn sporozoites were isolated [42].
Gliding motility was quantified as follows: freshly isolated salivary gland sporozoites were re-suspended in RPMI medium (PAA Laboratories GmbH, Germany) supplemented with 3% bovine serum albumin (BSA, Roth) and 50 µL of re-suspended sporozoites transferred to glass bottom 96 well plates, and spun at 209 g for 5 min at RT. Next, the sporozoites were imaged with an inverted Axiovert 200M Zeiss microscope using the Axiovision 4.6 software and a 25x Apoplan objective (NA = 0.25) in differential interference contrast illumination. Movies were recorded at an interval of 3 s per frame for the period of 3 min.
To examine liver stage development we conducted the following assays: 1) In vitro liver stage development and indirect immunofluorescence assay (IFA): For IFAs, HepG2 cells were infected with salivary gland sporozoites, incubated for 24 and 48 h and fixed with 4% paraformaldehyde. Fixed slides were washed twice with PBS, permeabilized with 0.5% Triton X-100 and blocked with phosphate buffered saline (PBS) containing 3% BSA. Mouse anti-HSP70 antibody (obtained through the MR4 as part of the BEI Resources Repository, NIAID, NIH: Mus musculus b5, MRA-662, deposited by MF Wiser) and Alexa-Fluor® 594conjugated anti-mouse IgG (Molecular Probes) were used as primary and secondary antibodies, respectively. Host cell and parasite nuclei were stained with the DNA stain Hoechst 33,342. 2) In vivo liver stage development: female C56BL/6 mice were used in groups of four animals. Freshly isolated salivary gland sporozoites were suspended in PBS and 100 µL solution containing 10,000 sporozoites were injected into the tail vein of each mouse. The parasitemia was monitored daily on Giemsa stained slides from a drop of blood obtained from the tail vein [43].

Results
3.1. Bioinformatic identification of the house-cleaning enzyme repertoire of the rodent malaria model Plasmodium berghei, the human P. falciparum parasite and related apicomplexan lineages Nucleotide sanitation enzymes are key to cell survival and promising drug targets. In order to identify such proteins in the malaria model, Plasmodium berghei we performed basic local alignment search tool (BLAST) searches using human DNA sanitation enzymes as query sequences. Our approach returned single homologs for ITPase and dUTPase, as well as two NuDiX domain-containing proteins with similarities to NUDT1/MTH1 and NUDT2/AP4AH (Supplementary figure 1-4); the latter two are annotated as nucleoside diphosphate hydrolase (NDH from here on) and bis(5ʹ-nucleosyl)-tetraphosphatase [asymmetrical] (Ap4AH from here on). Each protein contains a well-supported Pfam protein domain (Table 1 and Figure 2(a)) [44]. A multiple sequence alignment of the orthologs from selected Plasmodium species revealed a high degree of conservation across experimental rodent (P. berghei and P. yoelii) and human malaria parasites (P. falciparum and P. vivax) (Supplementary Figures 5-8). Among the four enzymes, dUTPase is the most conserved with a minimum amino acid sequence identity of 81% between P. vivax and P. berghei dUTPase. P. vivax dUTPase contains an additional loop region, which is absent in other Plasmodium species, and therefore is thus least similar to other Plasmodium dUTPases as well. The other enzymes are slightly less conserved with the lowest sequence identity of 59% occurring between PfITPase and PbITPase (Figure 2(b,c)).
A comparative sequence and structural analyses of these four house-cleaning enzymes with their respective human orthologs revealed that the overall architecture, as well as the key catalytic residues (dUTPase: D190; ITPase: E44; Ap4AH and NDH: NuDiX domain residues), are conserved between the parasite and host proteins suggesting the conserved mechanism of action (Supplementary Figure 9A-B). On the other hand, we also observed residues in the active site cavity that differ between the host and the parasite enzymes in each case (Supplementary Figure 9C). Such differences in active site residues have been explored to design [49] and test [50] targetspecific inhibitors. Therefore, these parasite-specific enzymes may be selectively inhibited. Indeed, P. falciparum dUTPase has been explored to design selective parasite-specific inhibitors [33] but whether this or other enzymes are essential for the parasite is unknown. Low sequence identities of the P. berghei enzymes to their human orthologs (Supplementary Figure 9D) suggest these enzymes could be selectively inhibited.
P. berghei and P. falciparum belong to the Apicomplexa phylum which includes parasites of humans and livestock. They include Toxoplasma gondii which causes infections in immunocompromised individuals, Cryptosporidium (causing cryptosporidiosis), Theileria (theileriosis and East Coast fever in cattle) and Eimeria (coccidiosis in poultry). Using BLASTP we explored the repertoire of house-cleaning enzymes in these related protozoans using P. berghei as query sequences. Nucleotide sanitation enzymes are present in the three apicomplexan lineages Piroplasmida, Hemosporida, and Coccidia. Specifically dUTPase, ITPase and Ap4AH were conserved across the two Plasmodium species, Toxoplasma gondii, Eimeria spp., Theileria annulata, and Cryptosporidium spp; NDH on the other hand was only found in Plasmodium and Theileria ( 3.2. House-cleaning enzymes are constitutively expressed throughout the Plasmodium berghei intraerythrocytic developmental cycle A recent RNAseq study of blood stage and ookinete P. berghei life cycle stages indicated that all four genes are constitutively transcribed [51]; dutpase several folds higher than the remaining three candidates, which all share similar expression levels (PlasmoDB, http://plasmodb.org/plasmo/). Our independent RT-PCR analyses in blood and mosquito stage parasites confirmed the whole genome data and expression of these genes throughout the life cycle with transcription evident in oocysts and salivary gland sporozoites as well (Supplementary Figure 14A). Additional support for constitutive protein expression throughout the parasite life cycle comes from P. falciparum 3D7, P. vivax Sal-1 and P. yoelii 17X proteomics (PlasmoDB, http://plas modb.org/plasmo/) (Supplementary Figure 14B). For example, PfITPase is expressed in different erythrocytic stages, in gametocytes (male and female both), as well as in the salivary gland sporozoites. In absence of any liver stage expression data available for P. falciparum, we checked the expression level of the PfITPase ortholog in P. yoelii (Py), a rodent parasite, which confirmed the liver-stage expression of PyITPase (supplementary Figure 14B). The details of protein expression of the remaining three enzymes are also given in supplementary Figure 14B. Using data from these large-scale proteomics studies, we confirmed that ITPase, dUTPase, and NDH are expressed in all three stages of life cycle, i.e. blood-, mosquito-and liver-stage. Although these studies report no blood-stage expression for PfAp4AH, a recent publication confirmed that PfAp4AH is indeed expressed during P. falciparum blood stage [52].
In order to address expression timing and localization, we generated transgenic P. berghei parasite lines expressing each protein fused to green fluorescent protein (GFP) at its C-terminus (Figure 2(d-g)). ap4ah and ndh were tagged at the endogenous loci, while we expressed dUTPase-GFP and ITPase-GFP episomally under the control of the native promoter using 986 and 865 basepairs of genomic sequence preceding each open reading frame, respectively. Protein expression in transfected parasites was followed by live microscopy and revealed a clear cytoplasmic GFP signal for all four proteins in asexual blood stages (ring, trophozoite, and schizont) as well as gametocytes. Western blot analyses showed GFP-tagged dUTPase and ITPase to run higher than their expected molecular weights, while Ap4AH::GFP and NDH::GFP were detected at their predicted positions; NDH with a clear doublet and a smaller 25 kDa band correlating with the size of the cleaved GFP tag (Figure 2(d-g)).  PF3D7_1127100  TGGT1_233140  ETH_00008300  TA20975  cgd7_5170  ITPase  PBANKA_0618300  PF3D7_0720800  TGGT1_202300  EBH_0024040-t26_1  TA17785  cgd4_4150  NDH  PBANKA_1361900  PF3D7_1349100  None  None  TA18135  None  Ap4AH  PBANKA_1235300  PF3D7_0520600  TGGT1_214780  ETH_00005960  TA07740  ChTU502y2012_414g0115 BlastP at default settings performed at www.eupathdb.org (Release 37); Cryptosporidium parvum (gene id cgd); Cryptosporidium hominis (gene id ChTU); Eimeria tenella (gene id ETH); Eimeria brunetti (gene id EBH) Taken together, all four house-keeping enzymes show constitutive expression across the intraerythrocytic developmental cycle; they represent thus suitable antimalarial drug targets.

Functional genetic analyses of the P. berghei house-cleaning enzyme repertoire
We next determined which of the four putative enzymes is required for P. berghei life cycle progression by attempting the generation of gene deletion mutants and defining each mutant's fitness during rodent and mosquito infection. Gene knockout plasmids, designed to replace each open reading frame with a gene encoding a pyrimethamine resistance maker (Toxoplasma gondii dhfr/ts, or human dhfr fused to yeast fcu), were transfected into wild type schizonts; mutant parasites were selected with pyrimethamine following standard protocols [40].
Deletion of dutpase failed in three independent attempts. This suggests a likely essential role for the enzyme during blood stage development; most likely due to its dual role in detoxification of dUTP as well as being the sole source for dUMP for dTTP synthesis in the absence of a dCMP deaminase in Plasmodium. Together with previous studies in which P. falciparum dUTPase has been reported as a potential therapeutic target [32,53], our result confirms the conserved essential role of this enzyme in different Plasmodium species.
Establishing a clonal ap4ah(-) mutant also failed although one transfection attempt out of four had resulted in the integration of the plasmid into the genome (Supplementary Figure   15A). The isolation of an isogenic, mutant clone however proved unsuccessful in two attempts (n = 16); limiting dilution cloning produced either non-infected mice or animals infected with wild type parasites only. In a third cloning attempt, we injected 5 mice with 10 instead of the standard 0.9 parasites from the mixed population of wild type and ap4ah(-) parasites. All mice became infected, but genotyping by PCR revealed that only wild type had survived (Supplementary Figure 15A). These results suggest that lack of Ap4AH causes too severe a growth defect and thus, like dUTPase, the protein is required for rodent malaria infection.
Unlike dutpase and ap4ah, we readily established mutant lines lacking itpase or ndh (Supplementary Figure 15B and C). In mammals both ITPase as well as the NUDIX domaincontaining protein NUDT16 take part in ITP metabolism [54]. Therefore, we proceeded to generating an itpase(-);ndh(-) double mutant in order to determine whether P. berghei NDH and ITPase could compensate each other's function in the single knockout lines. To this end we first removed the yFCU-hDHFR marker by negative 5-fluorocytosine selection from the itpase (-) clone and then transfected the ndh deletion plasmid into the selection marker-free itpase(-) mutant; limiting dilution cloning was used to generate an isogenic itpase(-);ndh(-) line ( Supplementary Fig. 15D). We first monitored blood stage development and parasitemia curves after injection of 5,000 wild type or itpase(-);ndh(-) infected red blood cells in C57BL/6 mice. (Figure 3(a)). All mice infected with the wild type control showed health deterioration and were therefore euthanized. Although the two genes are not essential for blood-stage Prepatent period and parasitemia (mean ± SEM) at day 6 following injection of 10,000 salivary gland sporozoites. (f). Four C57BL/6 mice were infected with 10,000 salivary gland sporozoites of P. berghei wild type (WT), ndh (-) or ndh (-);itpase (-) by intravenous injection; parasitemias were monitored daily. All mice developed lethal blood stage infections. Data represent mean ± s.d. development, mutant parasites were less-virulent with prolonged host survival rates. A similar blood-stage characteristic has been reported for cysteine-protease inhibitor, PbICP knockout parasites [55].
Next, we quantified the number of oocysts following a mosquito blood meal. At day 16, the number of cysts per midgut was similar between the two groups (Figure 3(b)) indicating that neither ITPase nor NDH are required during DNA replication and sporozoite formation in the Anopheles mosquito. We did detect a significant reduction in sporozoite motility speed of salivary gland sporozoites in an in vitro assay on glass microscope slides (Figure 3(c)). The double mutant itpase(-);ndh(-) showed a speed of 1.2 ± 0.4 µm/s (n = 78) versus 1.5 ± 0.5 µm/s (n = 107) for wild type sporozoites (p < 0.001, Mann-Whitney U-test). However, this in vitro difference did not translate into a change of in vivo infectivity (Figure 3(d)), which requires larger differences in sporozoite motility [42]. DNA replication is particularly strong during liver development of the parasite, where a single sporozoite produces several thousand daughter progenies that are eventually released into the bloodstream. Imaging of liver cell development in HepG2 cells (Figure 3(d)) nor the initial establishment of blood stage infections identified a loss in fitness in the double mutant. The prepatent periodthe time it takes for blood stage parasites to appear in thin blood smearsvaried only slightly between 3.75 and 3.25 days (Figure 3(e)) and parasitemias quantified at day six following the infection were also within the normal range. Finally, similar blood stage infections were caused by intravenous injection of 10,000 salivary gland sporozoites (Figure 3(f)) and each of the parasite lines caused lethal blood stage infections with dynamics mirroring the injection of 5000 iRBC [56,57].
Although we did not observe a strong phenotype for itpase (-);ndh(-) parasite as the parasite could complete the entire life cycle, yet we observed a survival of the mice, which often succumb to experimental cerebral malaria around day 7 postinfection. This suggests that the itpase(-);ndh(-) parasite line is less virulent that the wild-type controls. Whether this is a phenotype of the deletion of the two genes or a clonal effect would require the generation of several more clones. We think that the subtlety of the effect does not justify the additional animal experiments.
Taken together our classical reverse genetics approach has uncovered the essentiality of dutpase and ap4ah and reported itpase and ndh as two redundant genes for the rodent malaria model P. berghei allowing life cycle progression between the mammalian host and mosquito vector. Further work needs to be done to study the functional details of these essential enzymes to support their potential role as the therapeutic targets. In addition, the role of these essential enzymes in the other life-stages could be explored via generating the conditional knockouts in different stages.

Discussion
House-cleaning enzymes play critical roles in maintaining genomic DNA integrity. Helical structure and histone-based packaging that protect DNA are absent for nucleotide precursors and they are thus prone to chemical modifications. However, unlike artificial nucleotide analogs, which are widely used in cancer therapy and as antiviral agents (abacavir and didanosine against HIV; lamivudine and entecavir against hepatitis B) [58], endogenous modifications of nucleotide pools by house-cleaning enzymes is poorly understood. The characterization of such proteins can provide a fundamental contribution to our understanding of nucleotide pool maintenance and sanitation. In pathogens, determining the requirements for certain enzymes for cell viability is an essential part during the identification and selection of suitable and novel targets for selective inhibitors that can be developed as drugs for parasites of humans as well as livestock.
Here, we identified four proteins in the rodent malaria parasite P. berghei with homology to evolutionarily conserved housecleaning enzymes. While all proteins are constitutively expressed, reverse-genetics identified an essential nature of P. berghei dUTPase and Ap4AH; ITPase and NDH were dispensable for the establishment and maintenance of an infection in the experimental mouse host and the Anopheles mosquito.
In accordance with our in vivo experimental data, ndh had recently been found to be dispensable in a global phenotype screen for P. berghei [59]. No information is yet available for the remaining three P. berghei genes dutpase, ap4ah and itpase (20 May 2018; http://plasmogem.sanger.ac.uk/). Here we compared our functional genetics studies with two global screens addressing the fitness of mutants in the human malaria parasite P. falciparum and the related apicomplexan parasite Toxoplasma gondii (Table 3). Like P. berghei dUTPase the Toxoplasma homolog was found essential in a genomewide loss-of-function CRISPR screen for growth in infected human fibroblasts [60]. P. falciparum dUTPase, surprisingly, on the other hand, was identified as a dispensable gene in a genome-wide saturation mutagenesis screen by piggyBac transposon insertion [61]. This is in contrast to our current finding and the previously reported studies on P. falciparum dUTPase [32,53]. The T. gondii data again mirrored our findings for ITPase, identifying the protein as a dispensable component of the sanitation enzyme repertoire, while falciparum ITPase appeared to be essential for cell viability. We suspect that the host environment is a plausible factor that contributes to the different phenotypes in our in vivo study and previously reported in vitro, genome-wide saturation mutagenesis screen. ap4ah-deficient cells were neither viable in P. falciparum nor P. berghei; in Toxoplasma lack of the gene resulted in a minor growth defect. In vitro studies have shown that P. falciparum dUTPase catalyzes the hydrolysis of dUTP to dUMP through a conserved nucleophilic substation (SN2) mechanism in which a water molecule initiates the in-line nucleophilic attack [31]. The homology between the berghei and falciparum proteins strongly suggests that dUTPase in P. berghei performs the same reaction. That will prevent the incorporation of uracil into genomic DNA but also provide dUMP for dTTP formation in the absence of dCMP to dUMP deamination in Plasmodium.
Whether one or both of these functions are the true causes for its essential nature in P. berghei is, however, not certain. Some cells can tolerate high concentrations of dUTP in genomic DNA. For example Warner et al. [62] reported that E. coli lacking functional dUTPase were viable even though 20% of DNA thymine residues had been replaced with uracil. Also, Drosophila can tolerate up to 1097 ± 55 uracil per million bases [29]. Secondly, there is also demonstrated thymidine uptake by P. falciparum mediated by the essential plasma membrane equilibrative nucleoside transporter 1 (PfENT1) [61,63], a protein that is conserved in P. berghei. The protein may thus be required for an as yet unknown function within the cell [24,27]. The P. falciparum ortholog of Ap4AH (bis(5ʹ-nucleosyl)tetraphosphatase) hydrolyzes diadenosine 5′,5″-P1,P4tetraphosphate, which is a by-product catalyzed by tRNA synthetase and ligases, to GTP and GMP [45]. At low concentrations, Ap4A activates microphthalmia transcription factor (MITF) dependent gene expression in mast cells [64] and also regulates ATP sensitive K+ channels in cardiac myocytes which otherwise get inhibited at higher concentration [65]. High level of Ap4A is linked with DNA damage, stress responses, and apoptosis. Enzymes hydrolyzing Ap4A prevent these effects [66]. E. coli NDX-4, an enzyme that hydrolyzes Ap4A to ATP has also been reported to hydrolyze toxic 8-oxo-dGTP [67]. Our failed attempts to delete and maintain an ap4ah(-) parasite and recent structural and functional studies of the P. falciparum homolog Ap4A suggest that Ap4AH is a potentially druggable enzyme [52].
Unlike E. coli, yeast and mice, the deletion of itpase in P. berghei had no deleterious effect on life cycle progression in the scope of a rodent infection. Although the parasite does not possess an obvious homolog to human NUDT16, a functional analogue of ITPase [54,68], we cannot rule out the presence of a compensatory mechanism in Plasmodium. It is unlikely that NDH performs this function; deletion of both itpase and ndh did not affect parasite growth. This may again be due to the compensatory house-cleaning functions of multiple enzymes in the Plasmodium. This is interesting to note that the number of NUDIX domain proteins varies from 0 to 30 in different prokaryotes and eukaryotes that hints towards the metabolic complexity and adaptability of different organisms [14].

Conclusions
We report here that Plasmodium deoxy-uridine triphosphatase and Ap4AH are essential house-cleaning enzymes while inosine triphosphate pyrophosphatase and NDH are redundant for the malaria parasite. Most importantly, our failed attempts to delete dUTPase in P. berghei in combination with previous drug-inhibitory assays of P. falciparum dUTPase [69] suggest that this enzyme is indeed essential for blood-stage growth and like Ap4AH is a potential malaria drug target.

Declaration of interest
F Frischknecht was a member of the EU FP7 Network of Excellence EVIMalaR. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewer disclosures
Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.