Introduction

Toxoplasma gondii is an ubiquitous apicomplexan parasite that infects virtually all warm-blooded animals, including 25-30 % of humans (Montazeri et al. 2020). Of all food-borne pathogens, T. gondii causes 20% of the total disease burden in Europe (Havelaar et al. 2015). Although most acute infections caused by the fast replicating tachyzoite stage of the parasite remain asymptomatic in immunocompetent individuals, the parasite persists life-long in the form of encysted semi-dormant bradyzoites (Weiss and Dubey 2009). Currently, no medical treatment exists to eradicate these tissue cysts with their rigid cyst wall from infected individuals. T. gondii is transmitted through undercooked meat products of infected livestock but also via feline-shed oocysts that can contaminate water sources and soil (Weiss and Dubey 2009). Severe toxoplasmosis occurs after reactivation of cysts or primary infection of immunosuppressed individuals or fetuses of seronegative mothers and can lead to systemic infections, uveitis, encephalitis as well as congenital transmission with potentially fatal outcomes (Montoya and Liesenfeld 2004). Hence, tissue cysts of T. gondii constitute a long- term source for transmission and risk of acute disease. Also, other related cyst-forming coccidian parasites such as Neospora caninum and Besnoitia besnoiti pose an immense veterinary burden by causing abortion mainly in sheep and severe chronic infections in cattle and horses, respectively. Current treatment options exist only against the tachyzoite form of these parasites and include antifolates, such as pyrimethamine or trimethoprim, that inhibit the bifunctional dihydrofolate reductase-thymidylate synthetase, in combination with sulfadiazine, which blocks the dihydropteroate synthetase. These enzymes are implicated in DNA synthesis in both the parasite and side-effects often require pre-mature termination of the treatment (Katlama, De Wit, et al. 1996; Alday and Doggett 2017). Another potent drug target in coccidians and other apicomplexan parasites such as Plasmodium spp. is the cytochrome bc₁-complex in the mitochondrial respiratory chain, which is inhibited by a number of approved drugs, such as the naphthoquinones atovaquone (Meneceur et al. 2008; Nixon et al. 2013) and buparvaquone (Sharifiyazdi et al. 2012). However, both coenzyme Q analogs including bioavailable nanosuspensions (Shubar et al. 2011) fail to eradicate encysted bradyzoites of T. gondii and N. caninum from the brains of infected mice after multiple weeks of treatment (Doggett et al. 2012; Müller et al. 2015; Doggett et al. 2020). Experimental inhibitors of the bc₁-complex includes 1-hydroxy-2-dodecyl-4(1H)quinolone (HDQ) that arrests the growth of asexual blood stages of the related malaria parasite P. falciparum (Vallières et al. 2012). However, in T. gondii HDQ appears to inhibit multiple mitochondrial election donors such as the alternative NAD-dehydrogenase and the dihydroorotate dehydrogenase in tachyzoites (S. S. Lin, Gross, and Bohne 2009). It does not affect mature bradyzoites in vitro (Christiansen et al. 2022), and HDQ derivatives do not clear T. gondii cysts from the brain of infected mice (Bajohr et al. 2010). In addition, endochin-like quinolones constitute a promising series of potent bc₁-complex inhibitors that are active against T. gondii (McConnell et al. 2018) and Plasmodium parasites (Song et al. 2018). However, these compounds also do not fully eradicate the cysts in vivo (Doggett et al. 2012, 2020). Similarly, bumped kinase inhibitors that selectively inhibit parasite calcium-dependent protein kinases in various apicomplexans fail to clear tissue cysts in vitro (Christiansen et al. 2022) and in vivo (Hulverson et al. 2019). Despite these efforts, it remains unclear whether the target proteins of these drugs are rendered dispensable for the survival of bradyzoites or whether the eradication failures have other reasons, such as poor permeability across the blood brain barrier and/or the cyst wall. To date, the genetic tools available to directly characterize the function of essential genes in bradyzoites are limited to promoter replacements (Smith et al. 2021), while no genetic methods exist to directly interrogate the fitness contribution of mitochondrially encoded genes such as the cytochrome b subunit of the cytochrome bc₁-complex (J.-L. Wang et al. 2016). Consequently, reverse pharmacological approaches, that involve screening of inhibitors against target proteins, remain difficult and the underlying reasons for limited in vivo efficacies remain uncertain.

To address these shortcomings, we employed a recently developed human myotube-based culture system to generate mature T. gondii drug-tolerant bradyzoites (Christiansen et al. 2022) to screen the MMV Pathogen Box. Untargeted and stable isotope-resolved metabolomic analyses of dually active drugs identified established and unsuspected modes of action in tachyzoites and bradyzoites, respectively. Our data directly show metabolic consequences of bc₁-complex inhibition in bradyzoites and suggest its crucial role in ATP production.

Material and methods

Tachyzoite Culture

Pru-Δhxgprt tdTomato (John et al. 2009) (PruTom) and RH-Rep1/2-S9(33-159)-GFP (DeRocher et al. 2000; Thomsen-Zieger, Schachtner, and Seeber 2003) (RH-S9) parasites were maintained in HFF cells (BJ-5ta human foreskin fibroblasts ATCC, CRL-4001) with Dulbecco’s modified Eagle’s medium (DMEM) containing 25 mM glucose, 4 mM L-glutamine, 1 mM sodium pyruvate, 100 U/mL penicillin, 100 µg/mL streptomycin and 1 % heat-inactivated fetal bovine serum (FBS, Gibco) at 37°C with 10 % CO₂ (Christiansen et al. 2022).

In vitro T. gondii tissue cyst culture

The T. gondii in vitro cyst culture was performed as in (Christiansen et al. 2022; Maus et al. 2024). In short, KD3 myoblasts (Shiomi et al. 2011) were grown to 70 % confluency, followed by the induction of myotube differentiation by changing the medium to DMEM as described above but with 2% horse serum instead of FBS and additionally supplemented with 10 µg/mL human insulin, 5 µg/mL human holo-transferrin and 1.7 ng/mL Na₂SeO₃. After five days, the myotubes were infected with PruTom (MOI 0.1) in bicarbonate-free Roswell Park Memorial Institute medium (RPMI) containing 5.5 mM glucose, 50 mM HEPES, 4 mM L-glutamine, 100 U/mL penicillin, 100 µg/mL streptomycin, 2 % horse serum, 10 µg/mL human Insulin, 5 µg/mL human holo-transferrin and 1.7 ng/mL Na₂SeO₃. Cultures were incubated at 37 °C at ambient CO₂ for four weeks and the medium was changed every two to three days. A comprehensive scheme of the procedure is shown in Fig. S1.

MMV Pathogen Box Screening Procedure

All Pathogen Box compounds were stored at -20°C (Veale 2019). DMSO-solved compounds were diluted in DMEM or RPMI media and stored at 4°C until use. Host cells were grown in transparent, flat-bottom 96-well plates and infected as described above to cultivate either tachyzoites or in vitro tissue cysts. One column per plate was left uninfected to estimate fluorescence background. Four hours after infection with tachyzoites or after four weeks of bradyzoite maturation the culture medium was replaced by medium containing either 0.1 %

DMSO (negative control) or 10 µM of the respective Pathogen Box compound. During the seven-day treatment period the medium was renewed every two to three days. It was then changed to tachyzoite growth medium to induce re-differentiation of viable bradyzoites to tachyzoites, and then the culture was monitored over 28 days. Prior to each medium exchange, the cultures were monitored for compound crystallization. The fluorescence was measured in a fluorescence plate reader (Tecan Infinite M200 PRO) (excitation Λ = 554 nm, emission Λ = 589 nm). Fluorescence values passing a threshold of 2,000 after background subtraction indicated proliferating parasites and hence an ineffective treatment (Fig. 1).

Figure 1:

Resazurin viability assay

Host cell monolayers were washed with PBS and incubated with 500 nM resazurin in Milli-Q water at 37 °C for 4 hours. Fluorescence was measured (excitation Λ = 540 nm, emission Λ = 580 nm) in a fluorescence plate reader and normalized to the average fluorescence values of the DMSO-treated controls, which were set at 100 %.

The dose response of dually active compounds

Tachyzoite and bradyzoite cultures were treated with the MMV compounds which were pipetted in a serial dilution across the 96-well plate and treated for 7 days. Afterwards, the compound containing medium was removed and possible regrowth was observed for another 7 days (tachyzoite assay) or 21 to 28 days (bradyzoite assay). The gathered relative fluorescence units were background-subtracted. Depending on the development of the DMSO control, a dataset that optimally reflected the dose-response was chosen to calculate the half- maximal inhibitory concentration (IC₅₀) for tachyzoites and the half-maximal lethal concentration (LC₅₀) for bradyzoites. In tachyzoite datasets, this time point was between day 4 and 7. Bradyzoite cultures showed generally lower MOI, and the chosen LC₅₀ time point varied between experiments from 7 to 28 days post treatment. For an optimal LC₅₀, we chose the first time points where in one condition a bradyzoite culture completed tachyzoite regrowth, another condition showed no growth and in an intermediate condition the culture showed intermediate growth (Fig. 2). The concentration data was log-transformed and a non-linear regression (curve fit with variable slope) was performed to determine the dose response (Prism 9, GraphPad).

Figure 2:

In silico predictions

MMV provided the SMILES IDs for every compound, which were compiled with LightBBB (Shaker et al. 2020) to predict the blood-brain permeability. SwissADME (Daina, Michielin, and Zoete 2017) was used to predict the gastrointestinal absorption capability. The logP values (Fig. S2) were provided by MMV and remaining logP values were calculated on SwissADME (Daina, Michielin, and Zoete 2017) as consensus values.

Fluorescence stain of mitochondrial activity

This assay was performed on tachyzoites and bradyzoites. For the tachyzoite experiment, HFF cells were infected with RH-S9 parasites at an MOI 0.1 and treated for 24 h with 1 µM atovaquone, 25 nM buparvaquone or 5 µM MMV1028806, respectively, corresponding to five-fold of the respective IC₅₀ values, or 0.1 % DMSO as a solvent control. For the bradyzoites experiment, 4 weeks-old ME49 in vitro cysts were treated for 24h with 5 µM atovaquone, 500 nM buparvaquone or 5 µM MMV1028806, respectively, corresponding to the lowest lethal concentration, or 0.1 % DMSO as a solvent control. Cultures were then stained with 500 nM MitoTracker Deep Red FM (MT; Thermo Fisher) in DMEM for 40 min at 37 °C, followed by a 20 min de-staining with DMEM and final fixation with 1 % paraformaldehyde for 20 min at room temperature. The cyst wall was decorated with Dolichos Biflorus Agglutinin (DBA) and stained with Streptavidin-Cy2. Mowiol containing DAPI (1:3,000) was used to mount the cells. To avoid a biased image acquisition, all parasite vacuoles or cysts identified in a specimen along an horizontal axis were imaged. Tachyzoite images were taken with a Zeiss Axio Imager Z1 Microscope equipped with the ApoTome 2 at identical exposure times (DAPI 200 ms, MT 500 ms, S9-GFP 200 ms, Brightfield 10 ms) using a Zeiss Plan APOCHROMAT 63x/1.4 oil objective lens. Bradyzoite images were acquired on a Leica Stellaris 8 laser scanning confocal microscope using a Leica HC PL APO CS2 63x/1.4 oil objective lens. The pinhole diameter was set to 1 AU. All detectors were operating in photon counting mode. The acquisition parameters were set as follows. DAPI: diode laser at 405nm_exc with 2.2%, 430-550nm_em with a HyD S-detector. Cy2: WLL at 468nm_exc with 8.5%, 473-591nm_em with a HyD S-detector. MitoTracker: WLL at 641nm_exc with 2.7%, 650-784nm_em with a HyD X-detector (4x line average). The zoom was set to 7x, images were acquired in frame-sequential mode with a unidirectional scan speed of 400 Hz, the pixel dwell time for all channels was 3.18µs. For z- Stacks, z-step size was set to 0.299 µm.

Using ImageJ (Schindelin et al. 2012), the intensity threshold of individual pixels of MT (190 to 255 for tachyzoites and 125 to 65535 for bradyzoites) and DAPI (140 to 255 for tachyzoites and 363 to 65535) channels was cut off and the signal intensity was calculated per parasitophorous vacuole or cyst.

Thin section electron microscopy

The T. gondii in vitro cyst culture was performed as described above. After four weeks, cells were treated with 5 µM atovaquone, 500 nM buparvaquone, 5 µM MMV1028806, corresponding to the lowest lethal concentration, or 0.1% DMSO, as a solvent control, for 24h. Cells were fixed by covering the monolayer with a fixation solution of 1% paraformaldehyde and 2.5% glutaraldehyde (Electron Microscopy Services) in 0.05 M HEPES buffer (pH 7.4). After incubation at RT for 3h, samples were sealed with parafilm and stored in the fridge until further processing. Fixed cells were scraped, sedimented by centrifugation (3000 g, 10 min) using a swing-out rotor and washed twice with 0.05 M Hepes buffer. The cell sediment was embedded in low melting-point agarose and stored in 2.5% glutaraldehyde in 0.05 M HEPES buffer. Small blocks of the embedded sediment were subjected to postfixation (osmium tetroxide, tannic acid), en bloc contrasting (uranyl acetate), dehydration, and embedding in epoxy resin according to a standard protocol (Laue 2010).

Ultrathin sections were prepared with an ultramicrotome (UC7, Leica Microsystems, Germany) using a diamond knife (45°, Diatome, Switzerland). The sections were collected on bare copper grids (300 mesh, hexagonal mesh shape) or copper slot grids, contrasted with 2% uranyl acetate and 0.1% lead citrate, and coated with a thin (2-3 nm) layer of carbon. The ultrathin sections with a thickness of ∼65 nm were examined with a transmission electron microscope

(Tecnai Spirit, Thermo Fisher Scientific) operated at 120 kV. Images were recorded with a side-mounted CCD camera (Phurona, EMSIS, Germany) at 4095 x 2984 pixels.

Metabolomic analysis of drug-treated intracellular tachyzoites and bradyzoites

HFF cells were infected with PruTom at MOI 1. Approximately 64 h later, the cultures were treated with 1.5 µM MMV019189, 7 µM MMV1028806, 30 µM MMV228911, 1.2 µM MMV595321, 5 µM MMV688762, 2 µM MMV688854, respectively, and 0.1 % DMSO as a control for 3h. These concentrations correspond to the five-fold respective IC₅₀ value. Cultures were quenched in ice-cold PBS and parasites were isolated from their host cells by passaging scraped monolayers through a 23G needle, filtered through a 3 µm polycarbonate filter and after microscopic counting distributed at 1 x 10⁸ parasites per sample. Parasites were washed with ice-cold PBS via 5 min centrifugation at 21,500 x g at 0 °C. Metabolites were then extracted by sonication for 2 min in 50 µL 80 % acetonitrile in water. The insoluble particles were pelleted for 5 min at 21,500 x g at 0 °C. 5 µL of each sample was collected to generate a pooled biological quality control (PBQC) which was injected regularly within the otherwise randomized run list. 5 µL of each sample was injected onto either an ACQUITY UPLC BEH amide column (Waters) equipped with a VanGuard amide pre-column (Waters) or onto a SeQuant ZIC-pHILIC (Merck) column with an OPTI-LYNX ZIC-pHILIC guard column cartridge (Optimize Technologies) operating on a Vanquish Flex (Thermo Fisher). Analytes were chromatographically separated starting with 100 % eluent A (90 % acetonitrile in water with 10 mM ammonium carbonate) and 0 % eluent B (10 mM ammonium carbonate in water), decreasing to 40% A and 60% B over 25 min. Metabolites were detected on a Q Exactive Plus (Thermo Fisher) equipped with an electrospray ionization source operated in rapid polarity- switching mode at an MS¹ resolution of 70k and an MS² resolution of 35k (Top3 MSX1). Peak extraction, retention time alignment, and peak annotation were computed with Compound Discoverer 3.1 (Thermo Fisher) using an in-house library of authentic standards. Blank subtraction, gap-filling as well as statistical analysis were performed in Excel (Microsoft), and the data was plotted with GraphPad Prism 8.4.1. Principal component analysis was performed with ClustVis (Metsalu and Vilo 2015).

Isotopic labeling

Intracellular parasites were treated as mentioned above and incubated for three hours with DMEM containing either 25 mM U-¹³C-glucose or 4 mM ¹⁵N-amide-glutamine instead of the respective unlabeled carbon sources. Quenching, parasite isolation, metabolite extraction and liquid chromatography were performed as described. To increase coverage of the isotopologues, metabolites were measured in positive and negative mode separately without MS² scans but with an increased MS¹ scan resolution of 140k. For peak identification purposes, the unlabeled DMSO-treated sample was analyzed with intermittent MS² scans as described above.

ATP luciferase assay

PruTOM tachyzoites were cultivated in HFF cells without glucose, isolated 48 h post-infection and treated with 1 µM atovaquone or HDQ for 3h. PruTOM bradyzoites were matured for three weeks, glucose-starved for an additional week and treated for 24 h with 1 µM atovaquone or HDQ before being released by needle passage 23G, pepsin digestion for 20 min at 37 °C. The digestion reaction was stopped by neutralizing in 1.2 % sodium bicarbonate (pH ≈ 8.3) with 0.0159 g/l phenol red as pH indicator dye (Christiansen et al. 2022). They were isolated by douncing on ice and filtered through a 3 µm polycarbonate filter. 10⁵ parasites were then aliquoted in 100 µL ice-cold PBS in a 96-well plate. ATP was quantified using BacTiter-Glo™ Microbial Cell Viability Assay (Promega) according to the manufacturer’s protocol using a standard curve. Data were plotted in GraphPad Prism 8.4.1..

Metabolomic analysis of bc-complex inhibitors on in vitro cysts

Intracellular cysts were isolated from their myotube host cells and extracted as described in (Christiansen et al. 2022; Maus et al. 2024). In short, cysts were released by needle passage (23G) and captured with DBA-coated magnetic beads. PBS (with 2% BSA) washed cysts were extracted in 80% acetonitrile and analyzed as described above. One dish yielded approximately 10⁶ cysts and was distributed into 3 samples. Prior to a statistical analysis, the dataset was normalized by dividing the compound intensities through the summed intensities of the respective samples.

Results

The Pathogen Box contains dually active compounds

To initially identify compounds that simultaneously target the tachyzoite and the bradyzoite stage of T. gondii, we tested 371 of 400 compounds from the Pathogen Box at a concentration of 10 µM against both stages. We excluded previously reported active compounds against T. gondii tachyzoites (Spalenka et al. 2018) from this initial screen. HFF cells were infected with tachyzoites of the Pru strain expressing tandem Tomato fluorescent protein in its cytosol (PruTom) at an MOI of 1 (Fig. 1A). 47 compounds completely inhibited growth during 7 days of exposure while the parasitemia of DMSO-treated controls already peaked after 4 days (Fig. 1B). To also identify bradyzocidal compounds, we differentiated KD3 myoblasts for one week into multinucleated myotubes and infected them under bicarbonate-deplete conditions and at a multiplicity of infection of 0.1 for four weeks to allow maturation of tissue cysts (Fig. S1). We exposed the cultures to 10 µM of the 371 MMV compounds for one week, removed the inhibitors and stimulated re-differentiation into tachyzoites by bicarbonate supplementation (Fig. 1C). Growth of DMSO-treated cultures became detectable after 10 to 14 days. Interestingly, 31 compounds prevented any tachyzoite regrowth (Fig. 1D), indicating their bradyzocidal activity. To exclude host cytotoxicity-dependent suppression of parasite growth, we measured cell viability via the reductive potential of myotubes and HFF cells using a resazurin assay. At 10 µM, 30, 12 and 13 compounds inflicted cytotoxicity onto HFFs, myotubes or both cell types, respectively. This initial screen identified 14 compounds with activity against both tachyzoites and bradyzoites and no apparent cytotoxicity (Fig. 1E; Fig. S2).

Validation of screening hits

To validate our results, we determined half-inhibitory concentrations (IC₅₀) in PruTom tachyzoites (Fig. 2A, B) and half-maximal lethal concentration (LC₅₀) in bradyzoites. We also included 28 compounds that were previously reported to target tachyzoites (Spalenka et al. ²018⁾. We confirmed dual activity of 16 compounds that exhibited mean IC₅₀ values ranging between 70 pM and 1.1 µM for tachyzoites (Fig. 2E) and LC₅₀ values between 0.15 and 8.2 µM for bradyzoites (Fig. 2F). To allow a comparison between the tachyzoite and bradyzoite data, we also determined the respective minimal lethal concentrations (MLC) over a seven- day treatment duration of tachyzoites as well (Fig. 2G). Except for four compounds (MMV021013, MMV022478, MMV658988, MMV659004), minimal lethal concentrations were higher in bradyzoites, potentially reflecting their lower metabolic activity and/or the cyst wall as a diffusion barrier.

Dually active MMV Pathogen Box hits share high lipophilicity and are indicated to be active against other pathogens

To gain insight into general properties of active compounds from our initial screen (Fig. 1E), we tabulated indicated activities as provided by MMV (Fig. S2A, D). As expected, these compounds were mostly earmarked as active against T. gondii tachyzoites (67%), however, none of those compounds exhibited bradyzocidal activity. Instead, 27% of dually active compounds were considered active against Cryptosporidium. Also, despite the relative phylogenetic proximity of P. falciparum, only 5% of the 125 potential antimalarials blocked growth of both T. gondii stages. To characterize physicochemical properties of these dually active hits, we approximated their lipophilicity by their logP values (Supplemental File 1) and compared them to compounds that were inactive or active against a single stage of T. gondii (Fig. S2B). We found that tachyzocidal compounds possess a statistically significantly higher lipophilicity than the average Pathogen Box compounds and this trend was more accentuated for bradyzocidal and dually active compounds. We estimated permeability through the blood brain barrier (bbb) using LightBBB (Shaker et al. 2020) and gastrointestinal absorbability (GIA) using swissADME (Daina, Michielin, and Zoete 2017) (Supplemental File 1). Most of the dually active drugs were predicted to be both permeable across the bbb and absorbed in the intestine (Fig. S2C). Together, these data support phylogenetic relationship as a weak predictor of activity of Pathogen Box compounds against T. gondii bradyzoites and highlight the role of lipophilicity.

Metabolic drug responses in tachyzoites indicate conserved and novel modes of action

To investigate how dually-active compounds impacted parasites, we analyzed the metabolic responses of treated parasites. To that end, we choose five compounds based on their availability and yet unknown mode of action (Fig. 3 and Supplemental File 2). Infected monolayers were treated for three hours with a 3-fold of respective IC₅₀ concentrations of MMV compounds or the solvent DMSO. The inhibitor MMV688854 (BKI 1294) (Doggett et al. 2014) is considered to inhibit a non-metabolic target (calcium-dependent protein kinase (CDPK1)) at the employed concentration (Hayward et al. 2023) and served as an additional control to estimate the metabolic impact of growth inhibition. We filter-purified parasites from quenched monolayers and analyzed their acetonitrile extracts by LCMS using both, pHILIC and BEH- amide chromatography columns. We generated 9 samples per condition from three repeats of the experiment with triplicate samples. A principal component analysis illustrates that batch effects are the main contributors to biological variances within the dataset (Fig. S3A,B). Despite that, all compounds elicited a distinct metabolic phenotype on both LCMS systems (Fig. 3). The CDPK1 inhibitor increased levels of AMP, oxidized glutathione, pseudouridine, citrulline, glycolate and glutamyl glutamine, while reducing the abundances of GABA, carbamoyl aspartate, phosphoenolpyruvate and a hexose phosphate. Shifts in these metabolites are shared among many treatments and indicate their implication in a general stress response and thus do not signify direct interference with parasite metabolism.

Figure 3:

Inhibitor-specific responses include the accumulation of cGMP-AMP after treatment with MMV019189, accumulation of carbamoyl aspartate and dihydroorotate upon MMV1028806 treatment, and a general depletion of central carbon metabolites (succinate, citrate, glyceraldehyde 3-phosphate, ATP, MEP) in MMV228911-treated parasites. MMV595321 caused an accumulation of N-acetylated amino acids, and MMV688762 led to uracil deficiency and an accumulation of dihydroorotate and orotate. These data suggest that growth inhibitors cause a bipartite metabolic response composed of generic and compound-specific changes.

MMV1028806 and buparvaquone target the bc-complex in T. gondii similar to atovaquone

The metabolic impact of most tested MMV compounds indicated a novel mode of action that did not resemble known or recognizable metabolic fingerprints. However, the metabolic fingerprint of MMV1028806 was consistent with inhibition of pyrimidine biosynthesis at the bc₁- complex-dependent oxidation reaction of dihydroorotate to orotate, as has been reported for P. falciparum (Fig. 4A) (Allman et al. 2016; Creek et al. 2016; Antonova-Koch et al. 2018). A similar response would be expected to buparvaquone, which we found to be dually active against T. gondii tachyzoites and bradyzoites, targets the bc₁-complex in Theileria annulata (McHardy et al. 1985) and Neospora caninum (Müller et al. 2015) and was recently found active against T. gondii tachyzoites (Hayward et al. 2023).

Figure 4:

To test these hypotheses, we exposed intracellular parasites to MMV1028806 or buparvaquone (BPQ) while substituting glucose with ¹³C6-glucose for three hours. We then measured the incorporation of ¹³C into the central carbon metabolites as well as the intermediates of the pyrimidine synthesis pathway. As controls, we utilized the known bc₁- complex inhibitor atovaquone (ATQ) and the solvent DMSO (Supplemental File 3). Treatment with all three inhibitors led to a dramatic accumulation of carbamoyl aspartate and dihydroorotate which was accompanied by a depletion of uracil and uridine, in particular of the +5-labeled uridine isotopologue. These changes are consistent with impaired function of the bc₁-complex-dependent dihydroorotate dehydrogenase (Fig. 4B,C). Another common effect was the accumulation of succinate and depletion of fumarate, suggesting reduced recycling of FADH₂ via the bc₁-complex. Anaplerotic substrates were also affected by all three treatments, most notably a strong accumulation of aspartic acid.

To independently verify our results, we hypothesized that T. gondii also incorporates glutamine-derived nitrogen into pyrimidines. We replaced glutamine with its ¹⁵N-amide-labeled isotopologue in the tachyzoite culture media for three hours while simultaneously exposing the parasites to ATQ, BPQ or MMV1028806. As expected from our previous experiments, we observed a marked decrease of uracil and uridine levels while aspartate, carbamoyl aspartate and dihydroorotate accumulated again (Fig. 4D, Supplementary File 5). Consistently, ¹⁵N- incorporation into pyrimidines was diminished while carbamoyl aspartate and dihydroorotate were heavily labeled as M+1 isotopologues (Fig. 4E).

Together, these data show that treatments with all three compounds lead to the accumulation of metabolites upstream of the bc₁-complex-dependent reactions during pyrimidine biosynthesis and the TCA cycle and to the depletion of downstream metabolites (Fig. 4A). These metabolic signatures strongly suggest an inhibition of the T. gondii bc₁-complex in tachyzoites.

Atovaquone, buparvaquone and MMV1028806 decrease the mitochondrial membrane potential of tachyzoites and bradyzoites

The inhibition of the bc₁-complex by ATQ leads to the collapse of the mitochondrial membrane potential in T. gondii tachyzoites (Vercesi et al. 1998) and we anticipated a similar phenotype in BPQ or MMV1028806-treated parasites. We treated intracellular tachyzoites and bradyzoites for 24 h with the three inhibitors or 0.1 % DMSO alone and estimated the mitochondrial membrane potential and morphology using a fluorescent Mitotracker dye. The mitochondria of control tachyzoites appeared as extended loops but largely collapsed after treatments with all three compounds (Fig. 5A, B). Notably, also host mitochondria appear influenced. In bradyzoites, mitochondria appeared mostly as spheres or were donut-shaped (Supplementary File 6) and their relative fluorescence decreased after 24 h treatment with ATQ, BPQ or MMV1028806 at 5 µM, 500 nM or 5 µM, respectively while their morphology appeared unchanged (Fig. 5C, D). Analysis by thin section electron microscopy revealed a largely unaffected sub-mitochondrial ultrastructure but the area of mitochondrial profiles and their relative abundance were changed in comparison to control after exposure with ATQ and MMV1028806 but not with BPQ (Fig. S5). This indicates that the latter decreases Mitotracker fluorescence intensity directly, while ATQ and MMV1028806 cause both, a drop in Mitotracker fluorescence and a reduced size of the mitochondria which leaves open whether a change of the membrane potential, of the size, or even of both parameters were responsible for the drop in Mitotracker fluorescence.

Figure 5:

Inhibitors of bc₁-complex alter catabolic processes in bradyzoites

Next, we sought to confirm this mode of action directly on bradyzoites and record the metabolic impact of ATQ, BPQ or MMV1028806. To that end, we exposed intracellular four-week-old cysts to 5 µM, 500 nM or 5 µM of the respective compounds and for three hours and used 0.2% DMSO as a mock treatment. These concentrations represent their LC₅₀ against bradyzoites and were chosen to minimize off-target effects. The cysts were isolated for LCMS analysis using magnetic beads (Christiansen et al. 2022) (Fig. 6A). Common to all three metabolic phenotypes is the accumulation of AMP and NADH, while varying TCA cycle intermediates, such as malate, succinate and α-ketoglutarate and mitochondria-independent energy sources such as glucose and phosphocreatine were depleted (Fig. 6B-D). These metabolic changes suggest an energy starvation scenario marked by rising AMP levels and reduced recycling of NADH via the bc₁-complex. Interestingly, in BPQ and MMV1028806 treated cysts, we also observed prominent accumulation of several acyl-carnitines species which are mobile intermediates of ß-oxidation (Fig. 6B-D). Since this pathway is thought to be absent in T. gondii, we attribute these changes to inhibition of host mitochondria (Fig. 5A). Notably, we did not observe changes in the pyrimidine synthesis pathway, which may reflect low nucleic acid demand from largely dormant bradyzoites. These data demonstrate that bradyzoites, which were considered metabolically inactive, exhibit metabolic responses to inhibitory compounds. Although we characterized AMP accumulation as part of a general stress response in tachyzoites, the metabolic profiles of inhibited bradyzoites are consistent with inhibition of their mitochondrial bc₁-complex and suggest its role in ATP generation.

Figure 6:

The differential efficacy of atovaquone and HDQ against bradyzoites correlates with the inhibition of ATP levels in bradyzoites

We sought to test the importance of the mitochondrial electron transport chain in ATP generation in bradyzoites and exploited the differential targets of the coenzyme Q analogs ATQ and 1-hydroxy-2-dodecyl-4(1)quinolone (HDQ). However, while ATQ inhibits the bc₁- complex, HDQ inhibits dihydroorotate dehydrogenase (DHOD) (Hegewald, Gross, and Bohne 2013) and the alternative NADH dehydrogenase (NDH2) (Saleh et al. 2007). Both compounds block tachyzoite growth at low nanomolar concentrations in vitro with IC50s at 20 and 4 nM, respectively (Fig. 7A). However, in contrast to ATQ, HDQ does not affect the viability of mature T. gondii bradyzoites despite diminishing their mitochondrial membrane potential (Christiansen et al. 2022). Up to 20 µM HDQ did not arrest re-differentiation into proliferating tachyzoites, while bradyzoites that were treated with 5 µM ATQ (LC₅₀ 2.5 µM) did not generate proliferating tachyzoites (Fig. 7B). To investigate how these modes of actions affect the mitochondrial transport chain in bradyzoites, we tested their influence on mitochondrial ATP production by luciferase-based ATP assay. This assay shows a linear response to parasite cell equivalents between 10³ and 10⁶ parasites per 100 μL, and in order to minimize variance we decided to work with 10⁵ parasites (Fig. S6). To maximize mitochondrial ATP production, we starved three-week-old bradyzoites for one additional week of glucose. As a comparison, we utilized tachyzoites that were cultured for two days in glucose-deplete conditions. Untreated tachyzoite samples contained in average 196 ± 21 fmol ATP per 10⁵ parasites, whereas untreated bradyzoites only contained 14.1 ± 1.1 fmol. (Fig. 7C). Three hour-long treatment with 5 µM ATQ decreased ATP levels in both stages by 75 % (Fig. 7D). In contrast, HDQ reduced ATP levels by 60 % in tachyzoites but only marginally affected bradyzoites (Fig. 7D). These data indicate that bradyzoites can maintain very low ATP levels and suggest the presence of a diminished but obligate production of ATP in the mitochondria that functions independently of exogenous glucose and HDQ-target enzymes.

Figure 7:

Discussion

New treatment options for chronic T. gondii infections are needed because so far, they cannot be cured and currently available treatments of acute toxoplasmosis are associated with severe side-effects, in particular in infants after congenital transmission (Konstantinovic et al. 2019). To foster the simultaneous identification of drug targets in bradyzoites and along with respective inhibitory compounds, we utilized a recently developed in vitro culture system of pan-drug tolerant bradyzoites (Christiansen et al. 2022) to screen 400 compounds of the MMV Pathogen Box. 16 compounds exhibited confirmed cidal activity against bradyzoites while also inhibiting growth of tachyzoites. Metabolic profiling and stable isotope tracing in treated tachyzoites suggested the inhibition of the mitochondrial bc₁-complex by MMV1028806 and the reference compound BPQ.

Interestingly, of the 15 compounds in the Toxoplasmosis disease set, as earmarked by MMV, only 10 were found to be active against tachyzoites and none against bradyzoites. These differences may be attributed to strain-specific susceptibilities which are known for anti-folates and atovaquone (Meneceur et al. 2008) and differences in the design of growth assays. MMV assessed activities via plaque assays that prevent media change and rely on visual inspection of growth of a low number of parasites and might overestimate inhibitory effects.

We found that bradyzocidal compounds exhibit higher predicted logP coefficients, suggesting that the parasite entry might occur via diffusion across the cyst wall and less via solute transporter proteins. A known example were high polarity and low import rates caused poor efficacy against T. gondii tachyzoites is fosmidomycin (Nair et al. 2012; Baumeister et al. 2011).

This case, might also be made for the antifolates, pyrimethamine, sulfadiazine, trimethroprim and sulfamethoxazole which exhibit low consensus logP values of 2.09, 0.4, 1.16 and 0.92 and also fail to eradicate tissue cysts (Alday and Doggett 2017; Enshaeieh et al. 2021).

Interestingly, MMV675968 is a dually active compound with a logP value of 2.46, that has been shown to be a potent inhibitor of the dihydrofolate reductase (DHFR) in a range of pathogens (Songsungthong et al. 2019, 2021; Kim et al. 2021; Borba-Santos, Vila, and Rozental 2020; Nugraha et al. 2019; Rollin-Pinheiro et al. 2021; Vila and Lopez-Ribot 2017; Cantillon et al. 2022). Its target in T. gondii, however, remains to be identified.

Another important diffusion barrier that needs to be overcome to eradicate T. gondii tissue cysts is the blood brain barrier (BBB). Pyrimethamine and trimethroprim, both penetrate the blood brain barrier (Geils et al. 1971; Ineichen et al. 2020) but do not clear brain-resident cysts (Montazeri et al. 2018; Enshaeieh et al. 2021; Christiansen et al. 2022) highlighting the cyst wall as a potentially decisive obstacle. Consistent with this hypothesis is the tolerance of mature in vitro cysts tissue cysts against high dosis of sulfadiazine and pyrimethamine (Christiansen et al. 2022).

We also used metabolomics to characterize the modes of actions of four additional inhibitors. In general, metabolic phenotypes reflect direct consequences of target inhibition as well as indirect effects resulting from attenuated growth and stress responses. We chose the bumped- kinase inhibitor MMV688854 (BKI-1294) that targets the calcium-dependent protein kinase 1 (CDPK1) as a control compound to identify metabolites implicated in a general stress response and adaptation to slow growth (Winzer et al. 2015). CDPK1 is needed for invasion (Johnson et al. 2012) and also affects formation of daughter cells. An accumulation of AMP and oxidized glutathione are the most conserved responses across the six treatments and may indicate depleted energy levels and oxidative stress (Caro et al. 2012). Also, pseudouridine is elevated in many treatments. Enhanced pseudouridylation of RNA is considered to be implicated in stress responses (Cerneckis et al. 2022) and modulates RNA stability (Davis 1995). It remains to be tested which of these conserved metabolic responses to a panel of growth inhibitors signify cellular decay and which ones take part in a productive metabolic response in T. gondii. Nonetheless, metabolic profiling data on antibiotics-exposed bacterial pathogens has been used to design combination therapies, by either interfering with productive metabolic stress responses (Zampieri et al. 2017) or by simultaneously inhibiting otherwise non-essential pathways (Campos and Zampieri 2019). We found that the systematic identification of target structures and drug modes of action from profiling data alone remains challenging due to inherent non-linear relationships within the cellular metabolic network, the underlying regulatory mechanisms of enzyme functions. A systematic mapping of genetic disruptions to metabolic phenotypes, as demonstrated in E. coli, may alleviate these restrictions (Anglada- Girotto et al. 2022).

Apart from the general metabolic stress response to the treatment with MMV688854, three compounds caused apparent inhibitor-specific signatures. MMV019189, which is known to inhibit P. falciparum (Duffy et al. 2017), but also C. parvum (Hennessey et al. 2018) and L. mexicana (Berry et al. 2018), lead to the accumulation of c-GMP-AMP in the parasite. This dinucleotide is produced by the c-GMP-AMP synthase (cGAS) in mammalian cells in response to cytosolic DNA and triggers a type-I interferon response (Sun et al. 2013). While cGas is considered absent in T. gondii, host c-GMP-AMP induces STING-signaling as a response to T. gondii infection (P. Wang et al. 2019). Hence, MMV019189 may implicate the host in its antiparasitic activity.

MMV228911-treated tachyzoites showed the depletion of the DXR product 2-C- methylerythritol 4-phosphate (MEP), a metabolite that occurs in the non-mevalonate pathway in T. gondii’s apicoplast. Other metabolites upstream of DXR are either not detected or depleted, such as glyceraldehyde 3-phosphate, but also partake in other pathways. Interestingly, MMV228911 also caused a broader dysregulation of central carbon metabolism marked by a decrease of many phosphorylated metabolites such as nucleotide triphosphates and PEP but also citric acid and acetyl-CoA. This is consistent with metabolic signaling beyond the apicoplast, which also has been implicated in P. falciparum, where a cytosolic sugar phosphatase is responsible for resistance to the DXR inhibitor Fosmidomycin (Guggisberg et al. 2014).

We identified MMV1028806 as a candidate bc₁-complex inhibitor by comparing its metabolic impact with the one of BPQ with ATQ. All three compounds cause a sharp increase of pyrimidine synthesis substrates, depletion of pyrimidines and dysregulation of TCA cycle intermediates. This is consistent with previous mass spectrometry-based studies that characterized metabolic fingerprint of atovaquone on P. falciparum intraerythrocytic blood- stages (Creek et al. 2016; Allman et al. 2016; Antonova-Koch et al. 2018). Additionally, we confirmed the bc₁-complex as a target by measuring the synthesis of TCA cycle and pyrimidine biosynthesis intermediates via the monitoring of the incorporation of ¹³C and ¹⁵N stable isotopes from glucose and glutamine, respectively. Both, the isotopologue distributions and metabolite abundances strongly suggest attenuation of pyrimidine production and TCA cycle fluxes by all three inhibitors (Hortua Triana et al. 2016). TCA cycle intermediates show characteristic isotopologue distributions that are indicative of a ¹³C-label incorporation via acetyl-CoA (MacRae et al. 2012). A relatively increased M+3 isotopologue species in dicarboxylic acids in ATQ- and BPQ-treated parasites may indicate enhanced anaplerotic synthesis from glycolytic intermediates. These distinctions from MMV1028806 metabotypes may reflect a combination of off-target effects and differential timing of the metabolic inhibition.

Interestingly, six compounds from the MMV Pathogen Box have recently been shown to inhibit O₂ consumption in tachyzoites through bc₁-complex inhibition (Hayward et al. 2023). We co- identified Buparvaquone as a bc₁-complex inhibitor but we did not test the remaining five compounds for their targets. However, in our screen at 10 µM we found MMV024397 and MMV688853 to be dually active, while Trifloxystrobin (MMV688754) and Auranofin (MMV688978) lacked inhibitory activity in bradyzoites, and Azoxystrobin (MMV021057) was inactive. In addition, the previously described bc₁-complex inhibitor ELQ400 (MMV671636) (McConnell et al. 2018) was dually active. Overall, these data are consistent with the fact that the bc₁-complex is indeed needed for survival of bradyzoites, while not every tachyzocidal bc₁- inhibitor also inhibits bradyzoites.

In contrast to tachyzoites, the dihydroorotate dehydrogenase in bradyzoites does not dominate the metabolic phenotype (Fig. 6B, C, and D), as we neither observe the accumulation of carbamoyl-aspartate, dihydroorotate, aspartate, nor the decrease in pyrimidines. Instead, AMP and NADH accumulate, which in eukaryotes signifies a low cellular energy status (S.-C. Lin and Hardie 2018; X. Fu et al. 2019) that typically induces AMP-activated protein kinase and a glycolytic response. AMPK is a conserved enzyme with a pivotal role in cellular homeostasis: it responds to elevated AMP levels by repressing anabolic activity and by stimulating glucose uptake and glycolysis, β-oxidation, and ketogenesis (Winder and Hardie 1999; S.-C. Lin and Hardie 2018). T. gondii tachyzoites express AMPK which serves an important role in metabolic reprogramming during the lytic cycle of tachyzoites (Li et al. 2023). Transcriptomic and proteomic data indicate its presence in bradyzoites (Garfoot et al. 2019) but its function remains unknown. We observed decreased levels of free glucose during all three treatments that are consistent with stimulated glycolysis. We also observe depleted phosphocreatine levels in bradyzoites (Fig. 6) during treatments. In muscle cells, phosphocreatine acts as an ATP buffer system and replenishes ATP pools from ADP via creatine kinase (Kuby, Noda, and Lardy 1954; Perry 1954). It remains to be seen whether bradyzoites possess this mechanism. We also detect elevated levels of acyl-carnitines in BPQ and MMV1028806 treated bradyzoites. These molecules act as shuttles for the mitochondrial import of fatty acids for β- oxidation. However, this pathway is deemed absent in T. gondii and the presence of acyl- carnitines in bradyzoites might reflect import from the host. It is conceivable that their elevation in response to buparvaquone and MMV1028806 indicates compromised functionality of the host bc₁-complex and subsequently accumulating β-oxidation substrates. Indeed, BPQ has a very broad activity across Apicomplexa (Hudson et al. 1985) and kinetoplastids (Croft et al. 1992).

We investigated the functionality of the mitochondrial electron transport chain of tachyzoites and bradyzoites in response to ATQ, BPQ, and MMV1028806 by Mitotracker staining and thin section electron microscopy. All three treatments decrease overall Mitotracker fluorescence, consistent with decrease mitochondrial electron potential. ATQ and MMV1028806 but not BPQ also shrink areas of mitochondrial cross-sections indicating a potential contribution to the observed change in overall fluorescence. These differences might reflect In agreement with previous observations (Place et al. 2023), we found bradyzoite mitochondria to differ in morphology compared to tachyzoites, as they predominantly occur as “blobs” and rarely as tubes or loops and our observations align (Supplementary File 6).

Our metabolomic and inhibition data suggest that the mitochondrial electron transport is essential for bradyzoite survival. Previously, we showed that the mitochondrial electron transport chain inhibitor, HDQ, inhibits tachyzoite growth at low nano molar concentrations while failing to target bradyzoites. A decreased mitochondrial membrane potential in HDQ- treated bradyzoites indicates its uptake (Christiansen et al. 2022) and implies differences in modes of action between bradyzocidal and non-bradyzocidal Coenzyme Q analogs, which might be responsible for the distinct efficacies. A number of metabolic reactions involve the mitochondrial electron transport chain in T. gondii (Maclean et al. 2021). To estimate which of those are essential, we tested the requirement of mitochondrial ATP production by measuring ATP concentrations in glucose-starved bradyzoites. ATP levels collapse in the presence of bradyzocidal ATQ but not HDQ, while both compounds act cidal in tachyzoites and deplete ATP (Fig. 7). This suggests that mitochondrial ATP production is tightly linked to the survival of mature bradyzoites and that HDQ does impair active but non-essential processes in bradyzoites. Interestingly, while HDQ inhibits the bc₁-complex directly in P. falciparum (Vallières et al. 2012), in T. gondii it blocks pyrimidine synthesis and mitochondrial NADH oxidation via DHOD and NDH2, respectively (S. S. Lin, Gross, and Bohne 2009), and ATQ directly inhibits the ubiquinone reductase center of the bc₁-complex in T. gondii (McFadden et al. 2000). These results also agree with the notion that the canonical TCA cycle in bradyzoites is a non-essential source of mitochondrial reduction equivalents (Christiansen et al. 2022). The electron sources feeding into electron transport chain and the bc₁-complex in bradyzoites remain to be established and may include HDQ-insensitive glycerol 3-phosphate dehydrogenase, succinate dehydrogenase (complex 2) and the malate:quinone oxidoreductase (Maclean et al. 2021).

Bradyzoites can also utilize glycolysis to generate ATP. In relatively immature, seven-day-old bradyzoites, glycolysis and mitochondria generate ATP as shown by 2-deoxyglucose and oligomycin A dependent inhibition (Y. Fu et al. 2021; Shukla et al. 2018). While we did not directly test the production of glycolytic ATP, the role of hexokinase on cyst formation and the remodeling of glycolytic enzymes (Pan et al. 2022) emphasize its importance. It remains to be investigated how mature bradyzoites rely on glycolysis to produce ATP and how this is integrated with mitochondrial ATP production.

Interestingly, despite its profound effect on ATP levels and viability of bradyzoites, ATQ fails to eradicate all T. gondii cysts from mouse brains and does not reliably prevent toxoplasmosis relapse in AIDS patients and mice (Araujo, Huskinson, and Remington 1991; Kovacs 1992; Ferguson et al. 1994; Katlama, Mouthon, et al. 1996). Underlying reasons for this discrepancy need to be established but may include insufficient bioavailability. Quinolones are known to be poorly soluble in aqueous environments and the predicted permeability of the blood-brain barriers to either ATQ, BPQ and HDQ is low, suggesting that insufficient exposure to the drug may play a role in treatment failure. Indeed, improved bioavailability of ATQ in nanosuspensions improved efficacies (Shubar et al. 2011). Also, BPQ treatment of mice that were experimentally infected with closely related Neospora caninum parasites did target parasites in muscle but not brain tissue (Müller et al. 2015). A similar conjecture arose when comparing endochin-like quinolones. Here, favorable pharmacokinetics of a variant with lower intrinsic potency likely conferred higher in vivo efficacy against cysts (Doggett et al. 2012).

Together, our data illustrate the functional importance of the bc₁-complex for the viability of bradyzoites and encourage the pursuit of bc₁-complex inhibitors as candidates for eradicating toxoplasmosis treatment. Both, efforts to develop multi-site-directed inhibitors (Doggett et al. 2020; Alday, Nilsen, and Doggett 2022) and re-purposing of respective drugs are promising avenues. However, our data also underline the importance of favorable pharmacokinetics to ensure passage through the blood-brain barrier and the cyst wall.

Acknowledgements

We thank Naohiro Hashimoto for sharing the KD3 cells, Frank Seeber for the T. gondii strains and the Medicine for Malaria Venture for the Pathogen Box and Silvio Bürge for his imaging support.

MB, DM, EP are funded by the Federal Ministry of Education and Research (BMBF) under project number 01KI1715 as part of the "Research Network Zoonotic Infectious Diseases". MB, DM, FS, TH, ML are funded by the Robert Koch Institute.

Supplements

Figure S1:

Figure S2:

Figure S3:

Figure S4:

Figure S5:

Figure S6: