Comparative chemical genomics in Babesia species identifies the alkaline phosphatase PhoD as a determinant of antiparasitic resistance

Significance

Babesia spp. infect a wide range of vertebrates, from humans to cattle to penguins. Babesia spp. cause significant disease, with limited treatment options. By developing a strategy of parallel in vitro evolution in two Babesia spp., we identified mutations in a predicted alkaline phosphatase (phoD) associated with resistance to a potent anti-babesial (MMV019266). Using reverse genetics in Babesia divergens, we validated a mutation in bdphoD resulting in resistance to MMV019266. BdPhoD localizes to the apicoplast and endoplasmic reticulum in B. divergens, as previously observed in diatoms. Finally, modulating expression levels of BdPhoD alters sensitivity to MMV019266. We have established a pipeline to facilitate drug target/resistance discovery in Babesia spp., revealing an uncharacterized alkaline phosphatase as a resistance mechanism.

Abstract

Babesiosis is an emerging zoonosis and widely distributed veterinary infection caused by 100+ species of Babesia parasites. The diversity of Babesia parasites and the lack of specific drugs necessitate the discovery of broadly effective antibabesials. Here, we describe a comparative chemogenomics (CCG) pipeline for the identification of conserved targets. CCG relies on parallel in vitro evolution of resistance in independent populations of Babesia spp. (B. bovis and B. divergens). We identified a potent antibabesial, MMV019266, from the Malaria Box, and selected for resistance in two species of Babesia. After sequencing of multiple independently derived lines in the two species, we identified mutations in a membrane-bound metallodependent phosphatase (phoD). In both species, the mutations were found in the phoD-like phosphatase domain. Using reverse genetics, we validated that mutations in bdphoD confer resistance to MMV019266 in B. divergens. We have also demonstrated that BdPhoD localizes to the endomembrane system and partially with the apicoplast. Finally, conditional knockdown and constitutive overexpression of BdPhoD alter the sensitivity to MMV019266 in the parasite. Overexpression of BdPhoD results in increased sensitivity to the compound, while knockdown increases resistance, suggesting BdPhoD is a pro-susceptibility factor. Together, we have generated a robust pipeline for identification of resistance loci and identified BdPhoD as a resistance mechanism in Babesia species.

Babesiosis is a febrile illness caused by infection of vertebrate hosts with the tickborne apicomplexan parasite genus Babesia. There is a vast diversity of parasites, capable of infecting nearly any vertebrate host, with at least 100 described species (1–3). Babesia spp. have long been recognized as pathogens of tremendous veterinary and agricultural importance, causing hundreds of millions of dollars of economic losses every year (4, 5). Babesiosis has steadily been gaining recognition as an important human zoonotic disease (3, 6–8). Human babesiosis is caused by a range of parasites worldwide including B. microti, B. duncani, B. venatorum, B. crassa, B. divergens, and related parasites (3, 8–12). Human infections can also be acquired through blood transfusion and solid organ transplant (13–19). Symptoms are generally most severe in splenectomized, immunocompromised, and elderly patients (3, 8, 20). Immunocompromised patients often experience treatment failure, occasionally leading to death (7, 21). Cases of human babesiosis are likely underestimated (13, 22–24). The vast diversity of Babesia parasites demands the development of a species-transcendent antibabesial.

The first-line treatment for babesiosis in humans consists of atovaquone and azithromycin, while the second-line treatment for severe, refractory, or relapsing Babesia is a combination of quinine and clindamycin (20, 25). De novo resistance has been reported in patients, resulting in life-threatening illness and even death (7, 20). Very few new treatments for Babesiosis are in development, with the exception of tafenoquine (26–30). Treatments for veterinary Babesiosis include imidocarb dipropionate, diminazene aceturate, in combination with an antibiotic. While effective, both drugs suffer from major toxic side effects (31). Additionally, variability in treatment efficacy is high across species. For example, there is a 50-fold to 100-fold range in efficacy of atovaquone between B. divergens, B. microti, and B. duncani (32–34). Further, the reported efficacy of imidocarb ranges from an effective dose of ~0.1 ng/mL in B. bovis to ~10 ng/mL in B. divergens to no measurable efficacy in B. gibsoni (35–37). While the efficacy of diminazene aceturate is similar across species, high toxicity precludes its clinical usefulness (36, 38, 39).

Leveraging antimalarials for antibabesial development is strategic and cost effective. The Malaria Box compound library has been made publicly available by the Medicines for Malaria Ventures (MMV) (40), and many of these compounds are effective against Babesia (34, 40–42). The structural diversity of the library and metabolomic studies suggest that these compounds target many distinct pathways (40, 43). Additionally, selecting a compound from the Malaria Box ensures cross-sensitivity with Plasmodium species, meaning identified targets in Babesia may be relevant to other apicomplexan parasites of importance.

Multiple target-inhibitor pairs have been identified using in vitro resistance evolution in Plasmodium falciparum (44–51). The majority of these efforts have focused on generating resistance in single species (44, 50). However, performing experiments in multiple species in parallel facilitates the identification of conserved resistance mechanisms and druggable targets by generating overlapping datasets, increasing the confidence in candidate identification (52) (Fig. 1A). Previous studies of drug resistance in Babesia have relied on in vitro evolution (37, 39, 53, 54). However, none have yet paired this technique with full genome sequencing. Both B. bovis and B. divergens are easily cultured and amenable to genetic manipulation, permitting reverse genetics to functionally validate mutations (55–57).

Fig. 1.

Comparative chemical genomics identifies a single locus associated with high-level resistance to MMV019266. (A) Schematic representation of the comparative chemical genomics pipeline (CCG). Representative light microscopy images B. bovis is outlined in teal, B. divergens is outlined in purple. Schematic made with https://Biorender.com. (B) The structure of MMV019266 and IC₅₀ values is next to a schematic of intermittent drug selection. For each species, selections were run in duplicate. (C) IC₅₀ values for B. bovis MMV019266 resistant clones. Hashed bars represent the IC₅₀ of the bulk population, solid bars represent the IC₅₀ of clonal lines derived from the bulk population. Error bars represent SD between three biological replicates in technical triplicate. (D) IC₅₀ values for B. divergens MMV019266 resistant clones (same format as panel C) (E) Cross-resistance testing for two clones of resistant to MMV019266 B. bovis parasites (selection A-B11; selection B-E1) Compound legend: MMV- MMV019266, ATV- atovaquone, AZ- azithromycin, QUI- quinine, CLIND- clindamycin, ID- imidocarb dipropionate, DA- diminazene aceturate. Black dots represent WT IC₅₀, colored dots represent resistant clone IC50. Statistical significance calculated by ordinary one-way ANOVA for each compound. Error bars represent SD. (F) Cross-resistance screening in B. divergens. Formatting is as in panel (E). Statistical significance calculated by ordinary one-way ANOVA for each compound. (G) Whole genome sequencing identified a single conserved gene associated with resistance to MMV019266 after filtering steps. The numbers within the arrows indicate the approximate range in the number of variants (SNPs/indels) identified in each filtering step across the clones/species (exact filtering numbers can be found in Dataset S2).

We selected a top candidate compound to establish the comparative chemical genomics (CCG) pipeline (MMV019266) (34). We used in vitro evolution to select for resistance, followed by whole genome sequencing to identify candidates associated with resistance to MMV019266. We identified a single, species transcendent locus associated with resistance—a previously uncharacterized membrane-bound metallodependent phosphatase, PhoD. We used recently developed genetic methods to validate and investigate the function of PhoD as a resistance mechanism (55). We show that perturbations to the levels of PhoD in B. divergens (BdPhoD) lead to shifts in sensitivity to MMV019266. PhoD has been studied deeply in prokaryotes (58–62), and in marine algae (63–65), which are distant relatives of apicomplexans (chromalveolates), where they play important roles in phosphate metabolism (62, 66). We demonstrate that BdPhoD localizes to the apicoplast and endoplasmic reticulum (ER), which are observed locations for PhoD in marine algae (63, 64). Together, we demonstrate that a previously uncharacterized alkaline phosphatase, BdPhoD, is involved in resistance to MMV019266, a potent antibabesial.

Results

Selection for Resistance against MMV019266.

MMV019266 was selected from the Malaria Box based on similar potency in both Babesia species, the lack of a known target in Babesia, for having drug-like properties, and efficacy in Plasmodium (34, 40, 44). Additionally, at the initiation of this study, there was also a lack of a known target in Plasmodium—which has subsequently been identified (47). In both species, the IC₅₀ was submicromolar: 0.28 ± 0.01 nM in B.bovis and 0.91 ± 0.01 µM in B. divergens. In vitro evolution experiments were performed in parallel in two independent starting cultures (labeled A and B selections) of B. divergens and B. bovis for 10 wk. Selections consisted of exposure to compound (5 × IC₅₀) until no parasites were detectable (2 to 5 d), followed by a recovery period until recrudescence (1 to 2 wk) (Fig. 1B). In both B. bovis and B. divergens, we were able to achieve at least a fivefold shift in IC₅₀ to MMV019266 after five rounds of intermittent selection (Fig. 1 C and D). Resistance was generated in a stepwise fashion, resulting in low and high resistance (Fig. 1 C and D). We observed differing levels of resistance between clones within selections (i.e. clones derived from within the A selection) and between selections (i.e., A vs. B). The IC₅₀ values for all selections can be found in Dataset S1.

To assess the specificity of the resistance generated by in vitro evolution, we tested resistant parasites against a panel of clinically relevant compounds (diminazene aceturate, imidocarb dipropionate, clindamycin, quinine, azithromycin, and atovaquone) to assess for potential cross-resistance in both species of Babesia. We observed no cross-resistance to any compounds tested, supporting a specific mechanism of resistance generated against MMV019266 (Fig. 1 E and F).

Whole Genome Sequencing of Resistant Clonal Parasites Reveals a Single Conserved Locus.

We performed whole genome sequencing (WGS) of nine resistant clones from the two independent selections (A and B) for B. bovis, and seven clones from the B. divergens A and B selections (information about sequencing and sequencing metrics can be found in Dataset S2). To identify potential genetic changes in resistant clones, we generated VCF files and filtered to select for variants in which at least one clone was different from the parental and reference strain (Datasets S3 and S4). Based on studies in P. falciparum, we reasoned that mutations of interest would be in the coding genome, and would most likely be in predicted enzymes (44). Subsequently, we filtered the resulting SNP mutations identified across clones and species to include only those within CDS. We also reasoned that antigenic surface proteins and antigens would have inherently high variability (67) and removed these genes from the analysis. Finally, we excluded hypothetical proteins with no annotated functional domains (Fig. 1G). In B. bovis and B. divergens resistant to MMV019266 across the clones, we identified 5 and 13 possible gene candidates, respectively (Fig. 1G and Dataset S2). Of these, only one gene candidate was identified as mutated in both species: Bdiv_001570c and the orthologous BBOV_I003300, hereafter referred to as bdphoD and bbphoD, respectively (Fig. 1G). It is important to note that the other candidates identified may also contribute to resistance to MMV019266, especially those which arose in multiple clones across both independent selections within a single species (Dataset S2). However, as we aimed to identify a conserved pan-babesiacidal target, we chose to focus on phoD, which was mutated in both species. BdPhoD is the only predicted PhoD-like phosphatase in B. divergens; however, there appears to be a duplication of the phoD gene (BBOV_I003305) in B. bovis. No mutations were identified in BBOV_I003305, and it shares less sequence similarity with BdPhoD than BBOV_I003300 (BbPhoD). In the WGS B. divergens samples, we identified a single amino acid resulting in a missense mutation that was mutated across five different clones: C196W or C196F (Fig. 2A and Dataset S2). C196F resulted in approximately 2.5-fold higher resistance than C196W (Fig. 1D). These mutations are located in the predicted PhoD-like phosphatase domain of the gene. In B. bovis, we identified three different missense mutations across five different clones: G229D in the PhoD-like phosphatase domain, and R418L/R418P in the C-terminal helix. We also identified two intronic mutations in two clones of resistant B. bovis (Fig. 2A and Dataset S2). Of these, the mutations R418L/P resulted in higher resistance than the mutation G229D (Fig. 1C and Datasets S1 and S2). We sequenced bbphoD in one additional resistant clone not used for WGS, and identified an additional mutation in clone B6 (G222W) (Fig. 2A and Dataset S2). The majority of mutations in bbphoD were only identified in the more highly resistant parasite clones, suggesting low-level resistance occurs through an alternative non-genetic mechanism. A single intronic mutation was identified in the low-level resistance clone D10 of B. bovis. We found no compelling evidence that copy number variation contributes to resistance to MMV019266 in either lowly or highly resistant parasites (information about CNV analysis can be found in SI Appendix, Figs. S1–S8).

Fig. 2.

Reverse genetic engineering of predicted PhoD results in resistance to MMV019266. (A) Babesia PhoD linear model schematics. mutation locations are denoted by yellow stars. B. divergens (Bdiv_001570c, Top); B. bovis (BBOV_I003300, Bottom), key to the left denotes the different predicted domains of the protein. The clone ID is colored by the selection (A or B lineage, and number of rounds of drug pressure) from which it arose. For B. divergens blue denotes the fifth round of selection in the A lineage, and purple the fifth round of the B lineage. For B. bovis: teal = third round of the A selection, red = fifth round of the A selection. (B) Alphafold structure of Bdiv_001570c (Left) and BBOV_I003300 (Right). The functional domains are highlighted: signal peptide (green), phoD domain [teal (B. divergens), purple (B. bovis)], C-terminal helix (dark gray), transmembrane domain (orange). Mutations are highlighted in yellow in each structure and shown in stick form. The active residues of the phosphatase domain are highlighted in red. (C) Alignment of the PhoD domain (InterPro). Known metal binding residues from the crystal structure of B. subtilis 2YEQ are denoted by purple boxes beneath the amino acid. Each residue where a SNP has occurred is colored- orange represents those selected after ENU mutagenesis, magenta those selected on wild-type parental parasites: BdPhoD- V82 (orange), C196 (magenta); BbPhoD G222 (magenta), G229 (magenta). BbPhoD R418 is outside the PhoD domain. Apicomplexans are arranged as follows by order and denoted by colored bar to the right of the species name: piroplasmida (purple), haemosporidia (pink), eucoccidiorida (orange). PhoDs from P. tricornutum are denoted by the green bar (PtPhos5-7). Bacteria (B. subtilis) are denoted by blue. Sequence gaps are denoted by vertical hashed lines. The legend for the percent similarity (Blosum45) coloring in the alignment is in the Upper Left corner. Gene IDs for sequences in the alignment are as follows (Top to Bottom): Bdiv_001570c, BBOV_I003300, BMR1_02g01775, PF3D7_0912400, TGME49_265830, PtPhos5_48970, PtPhos6_45757, PtPhos7_45174, 2YEQ (D) Schematic of the CRISPR/Cas9 strategy used to introduce C196W into B. divergens. (E) Sanger sequencing confirms the presence of the SNP resulting in the C196W mutation, as well as the shield mutation. (F) MMV019266 IC₅₀ of the CRISPR-engineered clone (yellow) compared to the in vitro selected lines (purple). Error bars represent SD of three biological replicates done in technical triplicate.

We additionally carried out in vitro evolution of resistance to MMV019266 in B. divergens parasites treated with the chemical mutagen N-ethyl-N-nitrosourea (ENU). ENU has been used with success in Plasmodium (reviewed in ref. 68) and Toxoplasma (69) to induce single nucleotide polymorphisms throughout the genome. After exposure to ENU and selection with MMV019266, parasites became highly resistant (~10 × IC₅₀ to MMV019266) after ~2 wk of constant exposure to MMV019266 (Dataset S1), suggesting this method may be of use for rapid resistance selection in future experiments. Sanger sequencing of bdphoD in two resistant clones generated from the ENU mutagenized resistant clones revealed the presence of a point mutation in bdphoD, V82M.

BdPhoD and BbPhoD are annotated as putative membrane proteins with a C-terminal transmembrane helix, containing a PhoD-like phosphatase domain (Fig. 2 A and B). This configuration is observed in several PhoD phosphatases in the marine diatom Phaeodactylum tricornutum (64). BdPhod shares the highest similarity with the P. tricornutum PtPhos6, while BbPhoD is most similar to PtPhos5 and PtPhos7 (SI Appendix, Fig. S9C). The gene is homologous to alkaline phosphatase (PhoD-like) in both P. falciparum (PF3D7_0912400) and T. gondii (TGME49_265830). This gene also contains a predicted signal peptide, which is predicted by PATS (70) and ApicoAP (71) to be an apicoplast signal peptide. The phosphatase domain is highly conserved within the Babesia genus, and across apicomplexans, marine algae, and bacteria despite differences in overall gene architecture (Fig. 2C and SI Appendix, Fig. S9 A and C). Differences in gene architectures of phoD phosphatases are observed in both bacteria and diatoms, and this variation is hypothesized to determine the localization of the protein in the cell (62, 63). Furthermore, protein sequence alignments reveal high conservation in the metal binding residues and predicted active site, suggesting these PhoD phosphatases are functional in apicomplexans (Fig. 2C).

Protein Alignment of PhoD across Apicomplexa Reveals a Reduction in Paralogs.

Searching the plasmoDB, piroplasmaDB, and toxoDB databases we identified a total of 128 PhoD domain–containing genes across 53 species (Dataset S5). In the piroplasms, species either contained one or two genes with an annotated PhoD domain. The majority of Plasmodium species have a single gene containing a PhoD domain. Finally, in Eimeriidae and Sarcocystidae the majority of organisms have two genes containing a PhoD domain, with the exception being Besnoitia besnoiti which has three. No orthologs were identified through BLAST of PhoA and Phytase from P. tricornutum. P. tricornutum is annotated to contain five PhoD-like alkaline phosphatases, each with different localizations, believed to be due to differing roles within specific cellular niches (62–64). This suggests that while at least one PhoD was retained in apicomplexans to serve functions within the cell, the parasites employ PhoD phosphatases for fewer functions than their free-living algal counterparts.

Modeling Reveals Mutations Near Predicted Catalytic Domains of Babesia PhoD Enzymes.

Using AlphaFold (72), we generated a predicted protein structure of both BdPhoD and BbPhoD with and without the predicted signal peptides (Fig. 2 B and F and SI Appendix, Fig. S10 A and B). In all cases, the truncated and full-length peptide sequences yielded very similar folded structures (SI Appendix, Fig. S10 E and F). Aligning the protein structures of BdPhoD and BbPhoD shows high structural conservation (rmsd = 0.833) (SI Appendix, Fig. S10C). The majority of the mutations that were selected for are located proximal to the predicted active site in B. divergens, and around the base of the C-terminal helix in B. bovis (SI Appendix, Fig. S10D). In BdPhoD, V82M is directly next to two of the predicted conserved metal binding residues (D80, Y83). C196W/F is proximal to the predicted active site of BdPhoD (SI Appendix, Fig. S10D). Based on the locations of these mutations in both species, we hypothesize that they affect the activity of PhoD, either through altered catalytic activity, substrate engagement, substrate accessibility, or enzyme stability, subsequently leading to resistance to MMV019266.

The crystal structure of PhoD has been resolved in Bacillus subtilis, and access to the active site is controlled by a C-terminal helix (60). Unfortunately, the crystal structure of B. subtilis PhoD was resolved in the closed conformation, and despite efforts the topology of the open structure or the mechanism of movement of the C-terminal helix has not been resolved. Despite this, we attempted molecular docking to assess the ability of MMV019266 to interact with BbPhoD and BdPhoD. We did not identify any high-confidence docking sites in either species of Babesia or in B. subtilis. Together, these data suggest no direct or allosteric interaction of MMV019266 with the active site of PhoD in either Babesia species.

Reverse Genetics Validates BdPhoD Mutations in Resistance to MMV019266.

To confirm that selected mutations in BdPhoD confer resistance to MMV019266, we adapted the previously successful CRISPR/Cas9 system of gene editing to B. divergens to introduce the single point mutation of C196W/F into wild-type parasites (Fig. 2 D and E) (55). While we attempted to generate both amino acid substitutions, we were only successful in generating C196W. Clonal parasite lines containing the single point mutation C196W were resistant to MMV019266 at a level consistent with that observed from clones generated via intermittent selection (~fourfold increase in IC₅₀), showing this mutation is alone sufficient for resistance (Fig. 2F). As with the selected lines, the CRISPR edited line was not resistant to atovaquone (Dataset S1).

BdPhoD Is Localized in the Endoplasmic Reticulum and Apicoplast.

In both prokaryotes and eukaryotes, PhoD localization varies widely depending on gene architecture and functional requirements of the cell (62–64). In the diatom P. tricornutum, there are several PhoDs: PtPhos5 and PtPhos6 share the highest protein sequence similarity with BdPhoD and BbPhoD, respectively (Fig. 3A and SI Appendix, Fig. S9C). Both PtPhos5 and PtPhos6 are predicted to be membrane-integrated phosphatases with a single transmembrane helix. PtPhos5 is localized to the plastid and endomembrane system, while PtPhos6 was detected in the ER (64). In apicomplexans, PhoD was identified in the apicoplast proteome of P. falciparum, and in the endoplasmic reticulum proteome fraction of T. gondii (73–75). As BdPhoD and BbPhoD both have predicted apicoplast signal peptides [as predicted by PATS (70)], we hypothesized that BdPhoD would localize to the apicoplast. Based on homology with PtPhos5, we also anticipated BdPhoD to localize to the endoplasmic reticulum. Using the CRISPR/Cas9, we introduced a GFP tag to the 3′ end of bdphoD (BdPhoD-GFP) (SI Appendix, Fig. S11). Through live cell imaging, we found that BdPhoD-GFP was compartmentalized adjacent to the nucleus (SI Appendix, Fig. S13). We confirmed that the presence of the GFP tag was not disrupting the localization of BdPhoD (SI Appendix, Fig. S13A).

Fig. 3.

PhoD is compartmentalized to the ER and apicoplast in the cell, is dynamic depending on cell state. For significance values: NS P > 0.05, *P < 0.05, **P < 0.005, ***P < 0.0005. (A) Similarity matrix of apicomplexan and P. tricornutum PhoD domains based on protein alignments (similarity by % similarity Blosum45, threshold 0): BbPhoD1 = BBOV_I003300, BbPhoD2 = BBOV_I003305, BdPhoD = Bdiv_001570c, PfPhoD = PF3D7_0912400. The five PhoD proteins from P. tricornutum are included (Dataset S5). The type of InterPro predicted phosphatase domains are included in the first right column, localization in the last right column. Localization abbreviations: PM = plasma membrane, ER = endoplasmic reticulum, NE = nuclear envelope, EMS = endomembrane system (membrane surrounding ER, plastid). (B) Fixed and permeabilized transgenic cells (endogenous tagged phoD-GFP) expressing ectopic lytB-mCherry (pBd- SP+TP_lytB-mCherry) stained using α-GFP (BdPhoD) and α-mCherry (magenta, BdLytB – apicoplast marker). Top row of images are 1N parasites, bottom row are 2N parasites. (Scale bar represents 2 µm.) (C) Fixed and permeabilized transgenic cells (endogenous tagged phoD-GFP) stained using α-GFP (green, BdPhoD) and α-BIP (magenta, ER marker). Top row of images are 1N parasites; bottom row are 2N parasites. (Scale bar represents 2 µm.) (D) A schematic of the grading metric used to quantify the levels of colocalization between various organelles. Magenta represents the organelle of interest, green represents BdPhoD. I) shows the grading scheme for colocalization with the mitochondrion and apicoplast and II) colocalization with the ER. (E) Quantification of (B), number of cells counted and graded–1N = 20, 2N = 30. (F) Quantification of (C), number of cells counted–1N = 21, 2N = 27. (G) Dose–response curve of fosmidomycin (Left, positive control) and MMV019266 (Right) in B. divergens using either geranylgeraniol (G-ol, black line) or IPP (orange line). N = 3 biological replicates done in technical triplicate. Error bars represent SD between three independent dose–response curves.

We next investigated the potential for apicoplast localization. We fused the signal peptide and transit peptide of LytB from B. divergens to GFP (SP+TP_lytB-GFP) and observe that its localization was consistent with that observed previously in B. bovis (76) (SI Appendix, Fig. S13B), showing that this is a valid marker of the apicoplast. We then co-transfected our endogenously tagged line BdPhoD-GFP with SP+TP_lytB-mCherry (apicoplast marker) (SI Appendix, Fig. S13C). In the 2N parasites, BdPhoD-GFP forms a distinct punctate structure, surrounded by a diffuse horse-shoe. In the 1N parasites, the punctate structure is lost, and BdPhoD-GFP appears only in the horse-shoe pattern (Fig. 3 B and C). The amount of overlap between BdPhoD-GFP and the apicoplast was assessed using a qualitative metric for high, mid, and low level of overlap between the fluorescent signals for at least 20 1N and 2N parasites, which is described in Fig. 3D. The level of overlap with the apicoplast is higher in 2N parasites than in 1N parasites, suggesting the localization of BdPhoD is stage dependent in the parasite (Fig. 3E).

The remaining signal is horseshoe shaped excluded from the nucleus, typical of the endoplasmic reticulum. We demonstrate that BdPhoD-GFP overlaps with the ER specific marker BIP. The level of overlap between BdPhoD-GFP and the ER in 1N and 2N parasite is similar (Fig. 3 C and F). Interestingly, the ER and apicoplast have been shown to be in direct contact in T. gondii (77), and there may be a direct flow of nutrients, metabolites, or messengers between the organelles. Further, apicoplast proteins are trafficked through the endoplasmic reticulum (78–80), so this localization may be indicative of different patterns of protein trafficking between 1N and 2N parasites. Together, these results suggest that BdPhoD is localized to the apicoplast and ER and suggests BdPhoD has functions in multiple cellular compartments.

Finally, using the mitochondrial stain MitoTracker Red CM-H2XRos, we show that BdPhoD does not colocalize with the mitochondrion (SI Appendix, Fig. S12). Together, we have demonstrated that BdPhoD localizes to the apicoplast and ER in B. divergens. Whether the catalytic domain of BdPhoD is exposed to the cytoplasm or to the ER/apicoplast lumen is unknown.

Treatment with MMV019266 Does Not Result in a Delayed Death Phenotype and Cannot Be Rescued with Isopentenyl Pyrophosphate (IPP) or Geranylgeraniol (G-ol).

We next investigated the possibility of a delayed death phenotype with MMV019266—something characteristic of many apicoplast-targeting compounds (81, 82), with some exceptions (83, 84). We measured the nuclear content of synchronized parasites treated with MMV019266 (10 × IC₅₀ = approx. IC₉₀) over the course of 12 h (one intraerythrocytic development cycle) as a proxy for understanding killing speed (85). The synchronization protocol does not prevent multiple infections, which likely explains the 2 N parasites recorded at 0 hours post invasion (hpi) (SI Appendix, Fig. S14 A and B). Regardless, over the course of 12 h in both species, we observed no increase in the number of 2N parasites, and 1N parasites decrease, suggesting cell death (SI Appendix, Fig. S14 A and B). Together, the phenotypic data suggest MMV019266 kills B. bovis and B. divergens within the first replication cycle.

We attempted to rescue MMV019266 treatment through supplementation of media with isopentenyl pyrophosphate (IPP) and Geranylgeraniol (G-ol). Rescue of compounds using IPP is a well-established method for confirming apicoplast targeting of compounds (82) and has been demonstrated in B. orientalis and B. microti (86, 87). Using the apicoplast-targeting compound fosmidomycin (inhibitor of isoprenoid biosynthesis), we were able to confirm the ability to rescue with both G-ol and IPP in B. divergens. Treatment with MMV019266 could not be rescued in B. divergens (Fig. 3G). Additionally, clindamycin targets the apicoplast (88), and we observed no cross-resistance in MMV019266 resistant selected parasite lines (Fig. 1 E and F). This suggests that the target of MMV019266 is not in the apicoplast and that mutations in BbPhoD and BdPhoD do not broadly protect against apicoplast-targeting compounds.

BdPhoD Levels Regulate Chemical Sensitivity.

We generated a line of parasites which episomally overexpress BdPhoD-HA on its endogenous promoter (Fig. 4A). Overexpression of BdPhoD-HA did not induce a proliferation defect (Fig. 4B). Fivefold overexpression of BdPhoD-HA was demonstrated by qRT-PCR (Fig. 4C). Protein expression of BdPhoD-HA was confirmed by immunoblotting (SI Appendix, Fig. S15A). We observed a threefold decrease in IC₅₀ in the parasites overexpressing BdPhoD-HA as compared to wild type, revealing that an increase in the protein level leads to increased sensitivity to MMV019266 (Fig. 4D). While direct targets may have complex mechanisms of action resulting in similar phenotypes observed here [i.e. DNA gyrase and ciprofloxacin (89)], these results suggest that BdPhoD is likely involved indirectly as a resistance mechanism to MMV019266.

Fig. 4.

Perturbations to the level of BdPhoD lead to changes in resistance level. For significance values: NS P > 0.05, *P < 0.05, **P < 0.005, ***P < 0.0005. All error bars represent SD. (A) Schematic of the plasmid used to overexpress PhoD-HA. (B) Proliferation of wild-type (WT) and overexpressing parasites (pBdPhoD) measured in fold change in parasitemia after 48 h of growth, counted on thin smears by light microscopy (n = 3). Significance determined by the paired t test. (C) qPCR of bdphoD in WT and BdPhoD-HA overexpressing parasites (n = 3). Significance determined by the paired t test. (D) IC₅₀ for MMV019266 in WT and BdPhoD-HA overexpressing parasites. (E) Schematic of the plasmid used to tag BdPhoD with an HA-DD-glmS tag for inducible knockdown. (F) Proliferation of wild type (WT) and BdPhoD-HA-DD-glms inducible knockdown parasites measured in fold change in parasitemia after 48 h of growth, counted on thin smears by light microscopy (n = 3). Significance determined by one-way ANOVA. (G) Immunoblot for α-HA to detect BdPhoD-HA-DD-glmS (~54 kDa). Knockdown (−Shld) was followed for 72 h. Loading control is α-Histone H3. (H) Quantification of the immunoblots following knockdown of BdPhoD-HA-DD-glmS (−Shld), normalized to loading control. Significance determined by one-way ANOVA. (I) IC₅₀ for MMV019266 in WT and DD (+Shld1) and (−Shld1) parasites. Significance determined by one-way ANOVA. KD was initiated at the start of the dose response assay (0 h) and parasites were maintained off of Shld1 for the duration. Graph represents the average of three biological replicates done in technical triplicate.

We next generated an inducible knockdown parasite line by tagging the gene with a destabilization domain (DD), which has been used with success previously in B. divergens (55) (Fig. 4E and SI Appendix, Fig. S11). In this system, the protein is stabilized by the small molecule Shield1 (Shld1), and knock down is induced upon washout of Shld1 which leads to protein degradation (55). Assessing parasite proliferation over 72 h showed no significant difference between the knockdown and wild-type parasites (Fig. 4F). After removing Shld1, a 3.5-fold reduction in protein is observed via immunoblot by 24 h post wash out, and the protein is nearly undetectable by 72 h post Shld1 wash out (Fig. 4 G and H and SI Appendix, Fig. S15B). We observed a 2.3-fold increase in IC₅₀ in Shld1 (−) knock-down parasites, showing that depletion of the protein results in resistance to MMV019266 (Fig. 4I). This result is consistent with our hypothesis that the selected mutations in BbPhoD and BdPhoD lead to lower activity of the enzyme, leading to resistance. Neither overexpression nor knockdown of PhoD leads to resistance to atovaquone, supporting a mechanism of resistance specific to MMV019266 (Dataset S1).

Finally, we endeavored to generate a knockout of BdPhoD. Very few essential genes have been validated in Babesia, and to date, no large screen of essentiality has been done. In P. falciparum, the homologous gene (PF3D7_0912400) did not display a growth defect in the piggyBac transposon–based screen (90). No data are reported on essentiality in P. bergei in the plasmoGEM screen (91). In a genome-wide CRISPR screen in T. gondii, the homologous gene (TGME49_265830) is nonessential (92). Here, we have shown there is no significant defect on growth rate upon knockdown of BdPhoD (Fig. 4F). Yet, we were unable to generate a knockout of BdPhoD despite multiple attempts. Amplification of the bdphoD locus (primers designed outside of the homology regions used) in clonal lines from knock-out transfections revealed that parasites always retained the wild-type bdphoD locus, in addition to integrating the knockout cassette, suggesting a genome rearrangement occurring to prevent knock-out (SI Appendix, Fig. S16). We also observed no loss of function mutations in phoD in either species of our selected lines (Dataset S2). These data suggest BdPhoD may perform an essential function in B. divergens which can be maintained at very low levels of expression. Interestingly, knock out of the PhoD phosphatase PtPhos6 (PhoD_45757) in the diatom P. tricornutum results in a significant growth phenotype, suggesting a more essential role for the gene may be an ancestral trait (65).

Perturbations in PhoD Do Not Affect Sensitivity to Reveromycin A or Mupirocin.

Cytoplasmic isoleucyl tRNA synthetase (cIRS) was recently identified as a target of MMV019266 in P. falciparum (47). Reveromycin A is a known, specific inhibitor of cIRS in P. falciparum and bacteria (47, 93, 94). To investigate the role PhoD may play as a resistance mechanism for cIRS targeting compounds in B. divergens we performed dose–response assays using wild type, PhoD* (CRISPR, C196W), PhoD-DD, and overexpressing PhoD-HA parasite lines. We show no significant difference in IC₅₀ for reveromycin A in any PhoD mutant line vs. wild type (SI Appendix, Fig. S17B and Dataset S1). We also show no significant difference in sensitivity to mupirocin, an inhibitor of apicoplast IRS (aIRS) (SI Appendix, Fig. S17A and Dataset S1). This suggests mutations in PhoD do not affect sensitivity to either cIRS or aIRS.

Discussion

Current treatments for babesiosis in both humans and animals are suboptimal and the efficacy of the same compound can vary significantly across species, hindering treatment efforts. In this study, we have developed a comparative chemical genomics pipeline, allowing for the rapid identification of PhoD as a resistance determinant against the small molecular inhibitor MMV019266. PhoD family phosphatases have been extensively characterized in B. subtilis, as well as in the cyanobacteria Aphanothece halophytica and Synechococcus elongatus (58, 60–62). In these prokaryotes, PhoD phosphatases are predicted to localize to different subcellular compartments depending on the species and sequence architecture (58, 59, 61, 62). In eukaryotic phytoplankton such as dinoflagellates and diatoms, APs have also been shown to have highly variable localizations and demonstrate rapid evolution which may be driven by different strategies for accessing dissolved organic phosphate (63, 64). In both marine prokaryotes and eukaryotes, there are secreted PhoD phosphatases (62–64). We found no evidence for secretion of BdPhoD, and it appears that these secreted PhoD phosphatases have been lost in apicomplexa, and that there has been generally a loss in PhoD diversity in the parasites— a possible adaptation to a parasitic lifestyle and a relatively stable environment.

PhoD plays important roles in phosphate starvation in both bacteria and marine algae, and is often upregulated during this condition (58, 61, 62, 64, 66). While much remains unknown about responses to phosphate starvation in apicomplexan parasites, nearly all key metabolic processes depend on the availability of both organic and inorganic phosphorous, including metabolic processes occurring in the ER (77, 95, 96). Further, phosphate translocators are essential to apicoplast function in T. gondii (97, 98). Recently, phosphate starvation in Toxoplasma gondii was shown to restrict growth. Yet, there was no upregulation of phosphate transporters, suggesting that parasites are able to access internal stores of phosphate, such as phospholipids on the plasma membrane, by an unknown mechanism (99).

Membrane-bound PhoD has been shown to interact with the cell wall in bacteria to release inorganic phosphate during times of phosphate starvation (58, 100). The diatoms P. tricornutum and Thalassiosira pseudonana are known to replace phospholipids with non-phospholipids in membranes during times of phosphate starvation, and PhoD is hypothesized to play a role in this process (64). Altering the balance between phosphorylated and non-phosphorylated lipids in a membrane can alter the permeability of the membrane. We hypothesize that the catalytic domain of PhoD in Babesia functions to dephosphorylate phospholipids within the membranes surrounding the endoplasmic reticulum (ER) and apicoplast. While we do not know the orientation of the catalytic domain, we hypothesize that the catalytic domain points toward the compartment containing the target of MMV019266 (i.e., cytoplasmic target, catalytic domain in cytoplasm). In doing so, the enzyme can alter membrane permeability, sequestering the compound away from the target (either the cytoplasm or ER/apicoplast lumen) (SI Appendix, Fig. S18). Alternatively, perturbations to PhoD may broadly impact general cellular metabolism, leading to resistance. In P. tricornutum, PtPhos6 (PhoD_45757) was shown to play an important role in global cellular function and cell cycle progression (65, 101). Studying the global metabolic changes in the cell after phosphate starvation or MMV019266 treatment at the transcriptomic, proteomic, and metabolomic levels would address whether PhoD is involved in phosphate metabolism related to these environmental perturbations.

At the outset of this study, no known target of MMV019266 had been identified. Recently, mutations in PfcIRS were identified after resistance selections using thienopyrimides, including to MMV019266 (47). We did not identify mutations in the B. divergens or B. bovis ortholog of PfcIRS, nor were phoD mutations identified in the P. falciparum selections. We also show that perturbations of BdPhoD do not affect sensitivity to reveromycin A (cIRS) or mupirocin (aIRS) in B. divergens. Of note, B. divergens is ~46-fold less sensitive to reveromycin A than P. falciparum (IC₅₀ of 9.2 µM and 200 nM, respectively). Further work using alternative inhibitors and genetic manipulation will be needed to determine whether cIRS is the target of MMV019266 in Babesia spp. Of interest, it is possible that mutations in PhoD may confer resistance specifically to thienopyrimidines in P. falciparum and would serve as an interesting avenue of study as these compounds progress in the therapeutic development pipeline.

In summary, we have presented MMV019266 as an effective pan-babesiacidal. We have identified and validated that amino acid substitutions in PhoD, a highly conserved alkaline phosphatase, are a resistance mechanism to MMV019266 in Babesia parasites. By leveraging the power of in vitro evolution and chemical genomics performed in multiple species in parallel, we have built and validated a robust comparative chemical genomics pipeline for high-confidence identification of resistance loci in Babesia spp. which can be used for future drug target discovery, validation, and characterization.

Materials and Methods

Parasite Culture.

The Babesia bovis strain MO7 provided by David Allred of the University of Florida was maintained in purified bovine RBCs (hemostat) at 4% hematocrit in RPMI-1640 medium supplemented with 25 mM HEPES, 11.50 mg/L hypoxanthine, 2.42 mM sodium bicarbonate, and 4.31 mg/mL AlbuMAX II (Invitrogen). Before addition of AlbuMAX-II and sodium bicarbonate, we adjusted the pH of the medium to 6.75. Babesia divergens strain Rouen 1987, kindly provided by Kirk Deitsch and Laura Kirkman (Weill Cornell Medical College), was maintained under the same conditions in purified Caucasian male O+ human RBCs (Research Blood components). All cultures were maintained at 37 °C in a hypoxic environment (1% O₂, 5% CO₂). Clonal lines of parasites were used for all selections and were derived from the provided strains via limiting dilution—these will be referenced as BdC9 (B. divergens) and BOV2C (B. bovis).

Compounds and Reagents.

MMV019266 (Vitascreen, LLC), Atovaquone (Sigma Aldrich Cat. No. PHR1591), Imidocarb dipropionate (Sigma Aldrich Cat. No. 33441), WR99210 (Jacobus Pharmaceuticals), reveromycin A (Santa Cruz Biotechnology, Cat No. sc-202314) mupirocin (Sigma Aldrich, Cat. No. M7694) were prepared in DMSO. Clindamycin (Sigma Aldrich Cat. No. 1136002), azithromycin (Sigma Aldrich Cat. No. 1046056), diminazene aceturate (Sigma Aldrich Cat. No. D7770), puromycin (Sigma Aldrich Cat. No. P8833), and blasticidin S (Invivogen Cat. No. ant-bl-10p) were prepared in water. Isopentenyl pyrophosphate (IPP, Isoprenoids LLC), and geranylgeraniol (Sigma Aldrich, Cat. No. G3278) were prepared in ethanol. Compound-1 (DMSO-based) was a gift from Jeffrey Dvorin (Boston Children’s Hospital). Shld1 was synthesized and prepared as previously described (102, 103). Sources and dilutions for antibodies used in immunoblotting (WB) and immunofluorescence analysis (IFA) are as follows: chicken α-GFP (IFA- 1:500, abcam ab13970), rabbit α-BIP was a gift from Jeffrey Dvorin (IFA-1:500, Boston Children’s Hospital), rat α-HA 3F10 (WB 1:1000, Roche Cat. No. 11867423001), rabbit α-H3 (WB 1:8000, abcam Cat No. ab1791), rabbit α-mCherry (IFA 1:1000, abcam Cat. No. ab183628). The secondary antibodies used for IFA were Alexa-Fluor 488 and 594-conjugated antibodies against chicken or rabbit IgG diluted as recommended by the manufacturer (Invitrogen Cat. Nos. chicken 488 A21207, rabbit 594 A11039). Secondary antibodies for immunoblotting were IRDye® 680 LT against rat, or rabbit IgG diluted per manufacturer’s instructions (LICOR/Odyssey Cat. Nos. rat 926-68029, rabbit 925-68023).

Plasmids.

Primers for PCR amplification and verification of genetic manipulation in B. divergens are shown in SI Appendix, Table S1. Detailed information about the generation of these plasmids, including specific reagents and cloning sites, can be found in SI Appendix, Extended Methods. The resulting plasmids will be referred to as pBdEF-SP+TP_phoD-GFP-BSD, pBdEF-SP+TP_lytB-GFP-BSD, pBdEF-SP+TP_lytB-mCherry-BSD, pBdEF- SP_lytB-mCherry-BSD.

Synchronous Phenotyping Assays.

Parasites were synchronized by mechanical release from the red blood cell as previously described (104, 105). In brief: 30 mL of culture (4% HCT, 1.2 mL packed RBCs) was grown to high parasitemia (>20%), centrifuged and resuspended in 10 mL of complete medium. For B. divergens, the suspension was passed through a 1.2-μm syringe filter; B. bovis was passed first through a 5-μm syringe filter followed by a 2-μm syringe filter. After mechanical release, the parasite suspension was centrifuged at 3,000 × g for 5 min to pellet the merozoites, then added to 1 mL of 20% hematocrit blood in complete medium and allowed to reinvade for 30 min at 37 °C, shaking at 600 rpm. Reinvaded RBCs were then washed and finally put in culture supplemented with 50 μg/mL of heparin sulfate to prevent reinvasion. Immediately after reinvasion, parasites were put on MMV019266 continuously at 10 × IC₅₀. Samples were taken every 4 h for 12 h and fixed in 4% PFA + 0.025% glutaraldehyde in 1× PBS overnight at 4 °C. Samples were then prepared for flow cytometry to access nuclear content. To assess nuclear content, iRBCs were prepared for flow cytometry via fixation in a solution of paraformaldehyde (4%) and glutaraldehyde (0.0075%) in a 1× PBS. Fixed cells were maintained at 4 °C until the end of the time course experiment, then washed with 1× PBS, permeabilized with triton 100 (0.1%), treated with RNase A (0.5 mg/mL), and finally stained with SYBR green (1: 10,000). Data were processed using FlowJo, and downstream analysis was performed using GraphPad Prism v9 (GraphPad Software, Inc.).

Asynchronous Parasite Proliferation Assays.

Proliferation assays were initiated at 0.1% parasitemia, diluted from asynchronous cultures. Parasitemia was monitored by thin blood smear and flow cytometry (SYBR green staining) over 72 h.

Dose–Response Assays in B. bovis and B. divergens.

The assays to determine the IC₅₀ of all compounds were done in 96-well plates as previously described in B. divergens (106, 107), with several modifications. Asynchronous cultures were prepared at 0.5% parasitemia in 3% hematocrit, 50 μL were added to wells of a black, clear bottom (Costar) 96-well plate. For apicoplast rescue experiments, the medium was supplemented with either 300 μM IPP or 5 μM geranylgeraniol. Compounds were prepared by serial dilution. Negative controls included no drug and 1% DMSO. Puromycin was included as killing (positive) control. Parasites were then incubated with compounds for 72 h in standard culture conditions. Subsequently, cultures were lysed with a solution containing SYBR green (1:5,000, Invitrogen Cat. No. S7567) (Lysis buffer: 0.16% saponin (Calbiochem/EMD #558255), 0.05 M TRIS-HCl (Roche Cat. No. 10812846001), 0.8 mM EDTA (Sigma Aldrich Cat No. E7889), 1.6% Triton X-100 (Sigma Aldrich Cat. No. T8787). SYBR green fluorescence was measured on a SpectraMax® iD5 (Molecular Devices) and DNA content was used as a proxy for parasite growth.

In Vitro Resistance Generation.

Rounds of intermittent drug selection were performed to attain resistance to MMV019266 in both B. divergens and B. bovis. Parasites were exposed to 5 × IC₅₀ for 3 to 5 d until growth was halted and parasites appeared punctate. Parasites were allowed to recover until normal growth resumed (between 2 and 3 wk for each round of selection). Parasites were subjected to rounds of selection until a greater than or equal to fivefold shift in IC₅₀ was achieved. Clonal lines of resistant parasites were then generated by limiting dilution.

Whole Genome Sequencing.

Genomic DNA of parasites was isolated using the Qiagen DNeasy blood and tissue kit, following manufacturer’s guidelines (Qiagen Cat. No. 69504). Clonal lines of B. bovis and B. divergens resistant to MMV019266, as well as the starting parental clonal lines, were prepared for sequencing using the Illumina Nextera XT workflow. Libraries were pooled (8 pM) and sequenced on the Illumina MiSeq using v2 reagents (500 cycle, paired end). FastQ files were assessed for quality using FastQC and low-quality bases were trimmed used trimmomatic (108). All data are publicly available and have been deposited in the SRA under the accession number: PRJNA924801.

Identification and Confirmation of Single Nucleotide Polymorphisms and Copy Number Variation (CNV) Analysis.

Cleaned, paired fastq files were aligned to the respective reference genome [B. divergens strain 1802A (67), B. bovis T2Bo (109, 110) both accessed via piroplasmaDB] using BWA (111). Alignments were sorted and subsequently merged into VCF files using SAMtools (112), BEDtools (113), and VCFtools (114). SNPs and indels were filtered first by comparison to the sequenced parental clone (B. divergens: BdC9, B. bovis: BOV2C)—rows were retained if any clone contained a variant in comparison to the parental strain. Variants were then filtered by quality score (>30), and heterozygous calls were removed (Babesia is a haploid organism). Next, variants were filtered based on location in coding sequences, and further refined based on functional annotation—all variants in antigen superfamilies were removed, as well as most hypothetical or unannotated genes, retaining variants in only genes with predicted enzymatic function. SNPs were then mapped to their respective sequences and assessed for functional significance (missense, nonsense, etc). Finally, the candidate gene sets were overlapped in B. bovis and B. divergens to identify possible conserved resistance determinants. To evaluate potential copy number variation (CNV), we counted aligned reads per 5-kb window within each BAM file. We excluded windows for which coverage across all samples was less than 20% of the median genomic coverage, and we normalized coverage per sample by dividing by mean coverage. We then identified putative CNVs as windows with at least a twofold difference in normalized coverage between an offspring sample and the parental sample (extended information about CNV analysis can be found in SI Appendix).

N-Ethyl-N-Nitrosourea Mutagenesis.

Parasites were grown to 5% parasitemia in 10 mL of culture at 4% hematocrit. Parasites were then exposed to either 1 mM or 1.5 mM N-ethyl-N-nitrosourea (ENU) for 4 h. ENU was then washed off and parasites were allowed to recover, until growth returned to 5% parasitemia after death induced by ENU. Upon recovery, parasites were placed on 10 × IC₅₀ of MMV019266 or atovaquone (control). Parasitemia was monitored daily until robust growth was observed. Parasites were assessed for mutations in the gene of interest by PCR.

Transfection of Parasites.

Free merozoites were used for transfection as previously described (55). Briefly, the pellet containing free merozoites and cell debris was resuspended in 110 µL P3+DNA solution (Lonza Cat. No. V4Xp-3024). Transfection of free merozoites carried out in a 4D-Nucleofector System (Lonza), using the FP158 electroporation settings. After electroporation of free merozoites, the parasite and buffer mixture was immediately transferred to 1 mL of RPMI containing 200 µL packed RBCs and pre-heated to 37 °C. Parasites were allowed to invade at 37 °C shaking at 600 rpm (BioShake XP tube and plate mixer) for 30 min before being washed with 10 mL RPMI to remove the P3 solution and returned to culture. Parasites were returned to culture at a final volume of 10 mL RPMI at 2% hematocrit. Parasites were allowed to recover for 24 h prior to addition of blasticidin S (20 µg per 10 mL).

Protein Sequence Alignments.

All apicomplexan sequences were downloaded from piroplasmaDB, toxoDB, and plasmoDB (Dataset S5). To identify genes containing PhoD domains in apicomplexa, we performed a search of these databases for the InterPro domain IPR018946 (“Alk_phosphatase_PhoD-like Alkaline phosphatase D-related” and “PhoD-like_MPP PhoD-like phosphatase, metallophosphatase domain”) as well as BLAST of the PhoD domain of BdPhoD. B. subtilis sequence was accessed from NCBI genbank (Bsubtilis_AAB47803). P. tricornutum PhoD sequences were acquired from the Phatr2 database (PtPhos3_45959, PtPhos4_39432, PtPhos5_48970, PtPhos6_45757, PhPhos7_45174). PhoD domains were extracted from full sequences based on the PFAM predictions for each protein. Alignments of whole protein sequences and domain sequences were performed using Geneious v9.1.5, with standard alignment parameters (Neighbor joining clustering method, ClustalW alignment). Phylogenetic trees were constructed using Geneious v9.1.5, using the Jukes–Cantor genetic distance model, the UPGMA tree build model, based on a global alignment with a Blosum62 cost matrix.

Protein Structure Modeling and Molecular Docking.

The structure of BdPhoD and BbPhoD were predicted using AlphaFold (72). Protein structure alignments were performed in PyMol v2.3.2. Docking studies were conducted with Schrodinger Maestro Release 2021-1 and 2022-3. The crystal structure for B. subtilis PhoD (2YEQ) was imported from the Protein Data Bank (PDB). Both the model BdPhoD and B. subtilis PhoD structures were prepared using Protein Preparation Wizard and aligned. Glide docking grids were generated for the active site in these structures, placing no restrictions or constraints on the ligand binding protocol. The ligand MMV019266 was prepared using LigPrep. Glide docking was performed using these Glide docking grids and MMV019266 under the XP precision mode and all other defaults with the exception of increasing the number of output structures to three per docked compound. SiteMap was utilized to identify five potential binding sites in the BdPhoD models.

Immunoblotting of Parasite Lysates.

Samples were taken for comparison at the same parasitemia. Infected red blood cells were pelleted, then lysed with a solution of 0.015% saponin in 1× PBS with protease inhibitors for 2 min at room temperature (Millipore Sigma Cat. No. 11873580001). The lysis step was repeated until the majority of the RBCs were lysed and a small pellet remained (2 to 3 lyse/wash cycles). Parasites were pelleted and dissolved in sample buffer (Cell Signaling Technologies Cat. No. 7722) prepared via the manufacturer’s instructions. Samples were denatured at 95 °C for 3 min prior to loading onto a 16.5% Mini-PROTEAN® Tris-Tricine Gel (Biorad Cat. No. 4563066). Gels were transferred to a nitrocellulose membrane (Millipore Sigma Cat. No. GE10600002) in cold wet transfer buffer overnight at 15 V at 4 °C, followed by 1 h at 40 V. Membranes were blocked in Intercept (PBS) Blocking Buffer (LI-COR Cat. No. 927-70001) then treated with primary followed by secondary antibodies. Bound antibodies were detected using IR Dyes on the LiCor Odyssey Clx imager. Band intensity was quantified using Fiji version 2.3.0.

Validation of Causal Mutations by Reverse Genetics.

To introduce the single nucleotide polymorphism to generate a C196W mutation in bdphoD in B. divergens, parasites were transfected with pBdEF-Cas9-BSD-phodR. Parasites were exposed to blasticidin S 24 h post transfection and were kept on blasticidin S for 3 d. Parasites were allowed to recover until recrudescence. Gene editing was validated by PCR of bdphoD with primers that were designed outside of the homology region in the vector (SI Appendix, Table S1). Parasites were cloned by limiting dilution, and assessed for gene editing. Edited clones were then tested for resistance to MMV019266 as described.

Localization of PhoD.

To observe the location of BdPhoD, we tagged the endogenous locus of bdphoD at the 5′ end with GFP. Immuno fluorescence assays (IFA) on B. divergens were performed as previously described (55, 115–117). Briefly, thin smears of parasitized erythrocytes were air dried and fixed in ice-cold methanol, then permeabilized. For endoplasmic reticulum colocalization: chicken α-GFP was used to detect BdPhoD-GFP and was co-stained with rabbit α-BIP (ER). After antibody treatment, slides were stained with Hoechst 33342 (1:10,000 Thermo Fisher Scientific Cat. No. H3570) for 2 min. For localization of the mitochondrion, live parasites (WT or GFP) were incubated with 400 nM MitoTracker™ Red CM-H2XRos (Invitrogen, #M7513) for 20 min at 37 °C then immobilized on slides coated with poly-L lysine. Parasites were pelleted and washed once with 1× PBS. Parasites were then incubated with Hoechst 33342 (Thermo Fisher Scientific) for 5 min at RT. Parasites were dispensed onto poly-L-lysine coated slides to immobilize them for live cell imaging. Localization of the apicoplast was done by transfecting endogenously tagged parasites, BdPhoD-GFP with SP+TP_lytB-mCherry-BSD. All images were acquired on Zeiss Axio Observer using a 100× oil immersion lens. Analysis and quantification were all performed in FIJI version 2.3.0. To provide a qualitative assessment of colocalization, we visually assessed the level of overlap between the BdPhoD-GFP and the other colocalization markers (ER: α-BIP, apicoplast: SP+TP_lytB-mCherry-BSD). This was done by classifying the overlap into three categories: low, mid, and high overlap which is described in a schematic in Fig. 3D. Slides were blinded prior to evaluation. This is not presented as a quantitative assessment of colocalization, rather a subjective metric for approximate evaluation.

Overexpression of BdPhoD.

Overexpression of BdPhoD was achieved via transfection of pBd- phoD_gDNAHA-BSD. Parasites were maintained on blasticidin S 24 h after transfection. Expression of BdPhoD-HA was assessed by immunoblotting. Proliferation of wild-type and transfected parasite lines was assessed.

Conditional Knockdown of BdPhoD.

Conditional knockdown of BdPhoD-DD was achieved by transfecting parasites with pBdEF-Cas9-BSD-phod-DD. Parasites were maintained on blasticidin S 24 h after transfection until no parasites were observed- at which time blasticidin S was removed. Parasites were then maintained on 500 nM Shld1. Tagging of bdphoD was confirmed by PCR (SI Appendix, Fig. S10 and Table S1). Proliferation of wild-type and DD tagged lines +/− Shld1 was assessed. Effect of Shld1 titration on tagged parasite lines was assessed using the fluorescence-based IC₅₀ assay previously described. Degree of knockdown was assessed by immunoblotting on a nitrocellulose membrane using α-HA, normalized to α-H3. To determine IC₅₀ with MMV019266, Shld1 was washed (4×) off the parasites immediately prior to set up of the assay. Parasites were exposed to compound for 72 h prior to lysis as previously described.

Knock Out Strategy for phoD.

To attempt to knock out bdphoD in B. divergens, we used a CRISPR/Cas9 approach. Parasites were co-transfected with linear PCR product (repair template synthesized by Twist Bioscience, San Fransisco, CA) containing the BSD resistance cassette flanked by 500 bases of homology in the 5′ and 3′ ends of phoD, and two plasmids containing gRNA targeting the 5′ and 3′ ends of bdphoD. The gRNA plasmids were generated from the base plasmid pBdEF-Cas9-BSD. The resulting plasmid was further modified with the guide RNA via insertion at the BbsI restriction site, resulting in pBdEF-Cas9-BSD-phodKO-g1 and pBdEF-Cas9-BSD-phodKO-g2. Parasites were continuously maintained on blasticidin S 24 h after transfection. Clonal lines were generated by limiting dilution, and assessed for integration of the knock-out cassette and loss of bdphoD by PCR. All primers used for generating constructs can be found in SI Appendix, Table S1.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Data, Materials, and Software Availability

All data are publicly available (118) and have been deposited in the SRA under the accession number: PRJNA924801. Plasmids and other tools generated are available upon request. Genome filtering data, IC₅₀ values, sequences of interest, and primers are all available in SI Appendix All other data are included in the manuscript and/or supporting information.