The zebrafish as a novel model for the in vivo study of Toxoplasma gondii replication and interaction with macrophages

Toxoplasma gondii is a successful human pathogen that often remains asymptomatic, although complications can arise in the immunocompromised and in neonates if infection is contracted during pregnancy (Pappas et al., 2009). Toxoplasma exist as invasive rapidly replicating tachyzoites in intermediate hosts (such as rodents and livestock), and convert into bradyzoite cysts in immune-privileged sites and long-lived cells (such as the brain and muscle tissue) during chronic infection (Pittman and Knoll, 2015). Once inside the host cell, parasites reside in a non-fusogenic parasitophorous vacuole (PV), where Toxoplasma asexually replicates (Clough and Frickel, 2017). Egress leads to dissemination into neighboring tissues, culminating in systemic infection. Predation of intermediate hosts by the definitive feline host completes the Toxoplasma life cycle. Control of infection by the host immune response is thus critical for host survival and for continued parasite transmission. As a result of its well-understood life cycle, Toxoplasma has emerged as a valuable model organism to understand the balance between pathogen survival and innate cellular immune control.

Three clonal lineages of Toxoplasma dominate across Europe and South America; namely, the type I, II and III strains (Howe and Sibley, 1995). These three closely related Toxoplasma strains have been characterized by the severity of infections they cause in murine models (Gazzinelli et al., 2014). Infection with type I parasites causes acute mouse mortality, whereas infection with type II and type III parasites progresses towards chronic infection (Saeij et al., 2005; Szabo and Finney, 2017). In humans, it is thought that type II strains predominate in Europe, yet strain-dependent differences in pathogenesis and host responses are poorly understood (Ajzenberg et al., 2002, 2009).

Innate immune mechanisms against Toxoplasma infection have been studied in vitro using both murine and human cell lines, and in vivo using mice. In vivo studies have shown that monocytes and neutrophils are recruited to the intestine upon oral infection, and are the major cell types infected with Toxoplasma both in vivo and ex vivo in human peripheral blood (Channon et al., 2000; Gregg et al., 2013; Coombes et al., 2013; Harker et al., 2015). The importance of neutrophils in parasite control in vivo is not fully understood, although neutrophil-specific depletion studies have suggested a minor protective role against Toxoplasma (Del Rio et al., 2001; Denkers et al., 2012). In contrast, inflammatory monocytes are the first responders to infection and are crucial for controlling acute Toxoplasma infection (Mordue and Sibley, 2003; Robben et al., 2005; Dunay et al., 2008). Pioneering work identified the ability of macrophages to kill Toxoplasma (Murray et al., 1979; Murray and Cohn, 1979), by employing both IFN-γ-dependent and -independent mechanisms to control intracellular parasite replication (Sibley et al., 1991; Andrade et al., 2005; Saeij and Frickel, 2017).

While the mouse is a natural intermediate host and remains an important model to understand Toxoplasma pathogenesis, differences are emerging between the mouse and human in mechanisms of parasite control (Gazzinelli et al., 2014; Yarovinsky et al., 2008; Haldar et al., 2015; Tosh et al., 2016; Sher et al., 2017; Safronova et al., 2019). Therefore, to complement in vivo murine studies, a novel animal model can benefit analysis of Toxoplasma control on a cellular and molecular level. Zebrafish are a well-established model for studying infection and immunity (Renshaw and Trede, 2012; Yoshida et al., 2017; Torraca and Mostowy, 2018; Gomes and Mostowy, 2020). Coupled with their optical accessibility during early development, zebrafish larvae are highly suited for non-invasive study of Toxoplasma infection and host response in real time in vivo (Torraca and Mostowy, 2018; Gomes and Mostowy, 2020).

Here, we developed a zebrafish infection model to study strain-dependent infectivity and leukocyte response to Toxoplasma infection in the hindbrain. We show that Toxoplasma invade and replicate inside brain cells including post-mitotic neurons, and that type II (Pru) and III (CEP) parasites maintain a higher infectious burden than type I (RH) parasites. We also demonstrate that macrophages are crucial in the clearance of viable parasites, and use high-resolution three-dimensional correlative light and electron microscopy (3D CLEM) techniques to reveal a discontinuous PV in brain cells and macrophages. Our zebrafish infection model can therefore be used as a novel platform to enable unprecedented discoveries in strain-dependent parasite immunity.

To develop a Toxoplasma-zebrafish infection model, we tested whether tachyzoites could replicate in zebrafish larvae. We first used Toxoplasma type I (RH) strain, because it is known to grow faster in vitro and survive longer extracellularly than type II (Pru) and type III (CEP) strains (Saeij et al., 2005; Khan et al., 2009; Yang et al., 2013). We injected zebrafish larvae 3 days post-fertilization (dpf) in the hindbrain ventricle (HBV) with ∼5×10³ type I strain tachyzoites expressing GFP and followed infection for 24 h at 33°C (Fig. S1A). We observed parasite replication in vivo using time-lapse widefield fluorescent microscopy (Fig. S1B, Movie 1). Consistent with this, confocal microscopy showed that the percentage of vacuoles containing two or more tachyzoites significantly increased with time (Fig. 1A,B). To visualize PV formation around the replicating parasites, infected larvae were fixed and stained for granule antigen 2 (GRA2), a dense granule protein that accumulates in the PV lumen (Mercier et al., 2002). Here, GRA2 accumulated around single and replicating parasites, highlighting PV formation in vivo (Fig. 1C).

To investigate parasite morphology and location at 6 h post-infection (hpi), 3D CLEM using serial blockface scanning electron microscopy (SBF SEM) was performed on the HBV of infected zebrafish. We observed 39 parasites (from n=2 larvae) located inside cells of the zebrafish hindbrain, which all displayed host mitochondrial association (Fig. 2A; Fig. S2, Movie 2; see also Fig. 6), a hallmark of intracellular type I parasites previously described in mouse and human cells (Pernas et al., 2014). Moreover, tachyzoites could be observed as singlets, replicating doublets (thus joined together) or fully replicated doublets (two tachyzoites not joined together) (Fig. 2A; Fig. S2, Movie 2).

To characterize the specific cell type within which type I parasites reside and replicate in the zebrafish hindbrain, we used Tg(elavl3:GCaMP6s)^jf4 transgenic larvae. The elavl3 promoter drives expression in post-mitotic CNS cells fated to become neurons (Park et al., 2000). Embryos harboring this transgene were infected with type I-Tomato parasites (Fig. 2B; Movie 3). In agreement with 3D CLEM data showing that tachyzoites reside within brain cells, infection of elavl3:GCaMP6s transgenic larvae revealed that ∼1/3 of tachyzoites (5/14 and 8/25 from n=2 larvae) reside within GFP⁺ cells at 4 hpi. Collectively, these results show that type I Toxoplasma tachyzoites can invade zebrafish cells and replicate in vivo, and that tachyzoites favor brain cells over myeloid cells in the zebrafish hindbrain.

To determine if parasite strain can affect parasite burden and host response in our zebrafish model, we infected larvae with ∼5×10³ type I, II or III Toxoplasma-GFP (Fig. S3A). In all cases, infected larvae showed 100% survival and no adverse effects up to 48 hpi (Fig. S3B). To quantify parasite burden in a high-throughput manner, we optimized an automated quantification pipeline using ZedMate (Yakimovich et al., 2019 preprint) for the different strain types at 6 hpi and 24 hpi. Strikingly, type II and III parasite burden was ∼3× higher than type I parasite burden at 6 hpi (Fig. 3A,B; Fig. S3C). Analysis by fluorescent stereomicroscopy showed that, from the initial parasite input of ∼5×10³ tachyzoites (Fig. S3A), parasite burden was reduced ∼95% by 6 hpi, suggesting that ∼5% of parasites (∼250 tachyzoites, Fig. 3B; Fig. S3C) successfully invade zebrafish cells and establish infection. Once established at 6 hpi, all three strain types persisted equally and decreased by ∼20% between 6 hpi and 24 hpi.

To test if host mitochondrial association was observed across the three strain types, we stained host mitochondria in the HBV of infected zebrafish larvae. In agreement with in vitro observations (Pernas et al., 2014), ∼82% and ∼70% of type I and type III parasites, respectively, showed clear host mitochondrial association, while only ∼20% of type II parasites showed clear host mitochondria association (Fig. 3C,D). These results demonstrate that strain type-dependent host mitochondrial association characteristics are conserved in zebrafish in vivo.

To analyze Toxoplasma-macrophage interactions over time, 3 dpf transgenic larvae possessing red macrophages Tg(mpeg1:Gal4-FF)^gl25/Tg(UAS-E1b:nfsB.mCherry)^c264 (herein referred to as mpeg1:G/U:mCherry), were infected with type I, II or III Toxoplasma-GFP, and macrophage recruitment was quantified by fluorescent stereomicroscopy. Compared to mock injection, the number of macrophages recruited to the infection site was significantly increased (∼1.5 fold) for all three strain types at both 6 hpi and 24 hpi (Fig. 4A,B).

To analyze Toxoplasma-neutrophil interactions over time, 3 dpf transgenic larvae possessing red neutrophils Tg(lyz:dsRed)^nz50 (herein referred to as lyz:dsRed), were infected with type I, II or III Toxoplasma-GFP, and neutrophil recruitment was quantified by fluorescent stereomicroscopy. Here, the number of neutrophils recruited to the infection site was significantly increased (∼3 fold) compared to mock injection for all three strain types at 6 hpi (Fig. 4C,D). In contrast to macrophages, which remained at the infection site by 24 hpi, the number of neutrophils recruited to the infection site (for all three strain types) decreased significantly, reaching basal levels by 24 hpi.

To analyze the interactions between type I Toxoplasma and macrophages in depth, we imaged infected mpeg1:G/U:mCherry larvae with Toxoplasma-GFP at 6 hpi by 3D CLEM. Of the 18 tachyzoites found inside macrophages (from n=3 larvae), seven were intact parasites that had actively invaded macrophages, as identified by host mitochondrial association to the membrane surrounding the parasites (Fig. S4). Six of the seven intact tachyzoites were single tachyzoites inside PVs. This suggests that macrophages may prevent parasite replication. Conversely, a single event captured by 3D CLEM showed Toxoplasma replication within a zebrafish macrophage, as judged by host mitochondrial association to the vacuolar membrane and the presence of two replicating tachyzoites (still joined together), with one of the replicating tachyzoites harboring two nuclei (Fig. S5A,B, Movie 4). To follow the fate of type I parasites engulfed by macrophages in real time, mpeg1:G/U:mCherry larvae infected with Toxoplasma-GFP were imaged by time-lapse confocal microscopy. Here, we captured the movement of macrophages in the hindbrain harboring type I parasites that had been actively invaded, as identified by host mitochondrial association (Fig. S5C), suggesting that, in some cases, macrophages may facilitate parasite movement in the brain tissue. However, we also observed parasite engulfment by macrophages followed by loss of GFP fluorescence, suggesting active parasite degradation (Fig. 5A; Movie 5). Consistent with this, 3D CLEM showed parasite degradation inside macrophages, as identified by fragmentation of tachyzoite organelles (Fig. 5B; Fig. S6).

To investigate the structure of the PV surrounding tachyzoites inside brain cells and macrophages in the hindbrain, we imaged mpeg1:G/U:mCherry larvae infected with Toxoplasma-GFP using high-resolution correlative serial section transmission electron microscopy (ssTEM) and focused ion beam scanning electron microscopy (FIB SEM). At 6 hpi, ssTEM revealed gaps in the PV around type I tachyzoites in brain cells (Fig. 6; Movie 6). Consistent with this, and reports of vacuole breakage in human macrophages (Fisch et al., 2020 preprint), FIB SEM also captured incomplete vacuole membranes around type I tachyzoites within a macrophage (Fig. 7; Movie 7). Together, ssTEM and FIB SEM demonstrate that cell-intrinsic host immune pathways may be activated within both zebrafish brain cells and macrophages in order to control Toxoplasma infection.

To test the role of macrophages in Toxoplasma infection at the whole-organism level, we used the transgenic line Tg(mpeg1:Gal4-FF)^gl25/Tg(UAS-E1b:nfsB.mCherry)^c264, which enables the specific ablation of macrophages upon metronidazole treatment (Fig. S7A,B). In the absence of macrophages, infected larvae showed 100% survival (Fig. S7C). However, upon infection of ∼5×10³ type I tachyzoites, parasite burden was significantly increased in the absence of macrophages at both 6 hpi and 24 hpi, suggesting that macrophages are responsible for parasite clearance in vivo (Fig. 8A,B; Fig. S7D,E). Similarly, a significant increase in parasite burden in macrophage-ablated larvae (compared to control larvae) was observed upon infection of ∼5×10³ type II and III tachyzoites (Fig. S7F,G). To analyze the viability of parasites that are cleared by macrophages, we performed pixel volume quantification, and show that control and metronidazole-treated larvae contain equally replicating Toxoplasma tachyzoites (% of total vacuoles counted in the HBV; Fig. 8C). These data suggest that macrophages have a dominant role in clearing healthy, viable parasites rather than supporting their replicative niche.

Zebrafish infection models for studying eukaryotic parasites are beginning to emerge (Gomes and Mostowy, 2020; Dóró et al., 2019). In this study, we established a novel Toxoplasma infection model using zebrafish larvae to explore host-parasite interaction in vivo in the hindbrain. We found that the three main clonal lineages of Toxoplasma are able to establish infection in the zebrafish hindbrain and replicate within their PV, and reveal that macrophages are key in controlling viable parasites in vivo. Using the type I strain we showed that Toxoplasma invade zebrafish brain cells (including post-mitotic cells fated to become neurons, expressing GCaMP6s driven by the elavl3 promoter). Additionally, using state-of-the-art electron microscopy techniques, we observed vacuole breakage around tachyzoites in both zebrafish brain cells and macrophages in vivo.

Using time-lapse confocal microscopy and 3D CLEM, we visualized single and replicating type I tachyzoites in the zebrafish HBV exhibiting host mitochondrial association. The relatively slow replication cycle of Toxoplasma observed in tissue culture cells in vitro (>6 h) is consistent with that observed in the zebrafish HBV. GRA2 staining and 3D CLEM of replicating tachyzoites strongly suggests the formation of a PV. This is indicative of normal type I parasite behavior, as demonstrated in vitro using tissue culture cells and in vivo using other animal models (Black and Boothroyd, 2000).

The zebrafish HBV is well established to investigate host response to infection (Yoshida et al., 2017; Torraca and Mostowy, 2018; Gomes and Mostowy, 2020). We did not observe Toxoplasma dissemination from the HBV, and this allowed us to monitor leukocyte-parasite interactions within a localized area. Here, type II and III strains were more efficient than type I strains at maintaining a higher infectious burden. This suggests that type II and III strains may be more efficient at invading non-phagocytic cell types found in the HBV and/or evading clearance by host cells. Together with 3D CLEM and confocal microscopy, we conclude that Toxoplasma can invade non-phagocytic cells (with ∼1/3 of parasites within post-mitotic neurons) in the HBV of zebrafish larvae. Although mortality was not observed in our 24 h Toxoplasma infection model, it will be of great interest to explore the long-term consequences of Toxoplasma infection on parasite dissemination and zebrafish survival.

Both live-cell imaging and 3D CLEM showed type I parasite uptake and clearance by macrophages. In all cases of macrophage-parasite interaction captured by 3D CLEM, macrophages retained their fluorescence during Toxoplasma infection and had intact nuclei and mitochondria, indicative of a healthy host cell. Conversely, phagocytosed tachyzoites (i.e. those that had not entered by active invasion) exhibited no host mitochondrial association, which suggests no PV formation, and showed loss of organelle integrity. Our evidence obtained from time-lapse microscopy, 3D CLEM and macrophage ablation experiments highlight that active parasite clearance by zebrafish macrophages in vivo occurs within the first 6 hpi. Using ssTEM and FIB SEM, we observed incomplete PV membrane in zebrafish brain cells and macrophages. PV breakage has been observed in vitro after IFN-γ stimulation of murine cells and human macrophages, and is thought to be part of the cell-intrinsic host defense mechanism against Toxoplasma tachyzoites (Martens et al., 2005; Zhao et al., 2008; Yamamoto et al., 2012; Selleck et al., 2015; Fisch et al., 2020 preprint). Zebrafish possess an ortholog of IFN-γ together with a related gene (IFN-γ-rel), and studies have suggested that zebrafish IFN-γ (also known as Ifng1) expression is associated with host protection against bacterial infection, while IFN-γ-rel expression is associated with host protection against viral challenge (Igawa et al., 2006; López-Muñoz et al., 2009, 2011; Sieger et al., 2009; Yoon et al., 2016). Therefore, future work using our zebrafish infection model could explore the precise anti-parasitic mechanisms (including the role of IFN-γ and IFN-γ-rel) employed by brain cells and macrophages during Toxoplasma infection.

In murine in vivo models, various leukocytes have been implicated in trafficking Toxoplasma from the site of infection. Examples include infected neutrophils that pass from the intestine to the lumen (Coombes et al., 2013), as well as infected macrophages and dendritic cells that pass the blood brain barrier (Courret et al., 2006). It is intriguing, therefore, to note that our study using time-lapse microscopy showed parasite-infected macrophages moving through brain tissue for possible parasite transport (Fig. S5C). Analysis of our 3D CLEM data also identified an actively replicating type I parasite inside a macrophage exhibiting association of host mitochondria with the PV (Fig. S5A,B). This observation is reminiscent of parasite replication ‘hot spots’ described in vivo in the murine intestinal villi (Coombes et al., 2013). Injection of tachyzoites directly into the hindbrain of zebrafish is likely to promote Toxoplasma infection of neurons [its favored long-term niche (Ferguson and Hutchison, 1987; Cabral et al., 2016)], and may explain the lack of parasite dissemination by macrophages observed in our model. Overall, it is remarkable that zebrafish macrophages during Toxoplasma infection in vivo have the capacity to phenocopy known behavior exhibited by murine macrophages during Toxoplasma infection in vivo. This is the case for both the type of event observed (e.g. parasite killing, trafficking, sustaining replication) and their approximate in vivo frequency.

In summary, we have established a novel animal model for studying the in vivo innate immune response to Toxoplasma infection, and for comparing host response to the three main Toxoplasma strain types in vivo. We also demonstrate a dominant role for macrophages in parasite clearance. Having established a zebrafish model of Toxoplasma infection, we have revealed a unique in vivo infection platform for CRISPR targeting and high-throughput drug screens that, together with time-lapse microscopy, can be used to identify determinants underlying Toxoplasma infection control.

Animal experiments were performed according to the Animals (Scientific Procedures) Act 1986 and approved by the Home Office (project licenses PPL P84A89400 and P4E664E3C). All experiments were conducted up to 5 dpf.

Fish were reared and maintained at 28.5°C on a 14 h light, 10 h dark cycle. Embryos obtained by natural spawning were maintained in 0.5× E2 medium supplemented with 0.3 µg/ml Methylene Blue. Larvae were anesthetized with 20 µg/ml tricaine (Sigma-Aldrich) during the injection procedures and for live in vivo imaging. All experiments were carried out on TraNac background (Krauss et al., 2013) larvae to minimize obstruction of fluorescence signal by pigmentation leading to misrepresentation in parasite dose quantification.

Toxoplasma (RH/Pru/CEP) expressing GFP/luciferase or Tomato was maintained in vitro by serial passage on human foreskin fibroblast (HFF) cultures (American Type Culture Collection, CVCL_3285). Cultures were grown in Dulbecco's modified Eagle medium high glucose (Life Technologies) supplemented with 10% fetal bovine serum (FBS; Life Technologies) at 37°C in 5% CO₂. Parasites were prepared from 25G followed by 27G syringe-lysed HFF cultures in 10% FBS. Excess HFF material was removed by centrifugation for 10 min at 50 g. After washing with PBS, Toxoplasma tachyzoites were resuspended in PBS at 2×10⁶ tachyzoites/µl. During injection, tachyzoites were maintained at room temperature and passed through a 29G myjector syringe (Terumo) to dissociate clumps and homogenize the suspension. Control infections were carried out using uninfected HFF cultures prepared as described above. Larvae at 3 dpf were anesthetized and injected with ∼2 nl parasite suspension into the HBV. HBV injections were carried out as previously described (Mazon Moya et al., 2017). Larvae are optimally maintained at 28.5°C, but develop normally between 23°C and 33°C (Westerfield, 2007). Toxoplasma invades and replicates at a minimum of 33°C. Infected larvae were therefore transferred into medium pre-warmed to 33°C to ensure normal zebrafish development and parasite replication (Fig. S1A). Progress of infection was monitored by fluorescent stereomicroscopy (Leica M205FA, Leica Microsystems).

For parasite dose quantification, z-stacks of the infected hindbrain were taken within 5-10 min using the Leica M205FA fluorescent stereomicroscope at 130× magnification using a 1× objective. Images were analyzed using the particle analysis function in Fiji software (Schindelin et al., 2012). For manual quantification of parasite burden, z-stacks were taken using the Leica M205FA. GFP-positive punctae were quantified using the multi-point tool in Fiji software. Computer vision-driven automated parasite burden quantifications were carried out using the ZedMate plugin in Fiji software to corroborate manual quantifications (Yakimovich et al., 2019 preprint). Pixel volume quantifications were carried out by 3D projecting confocal z-stacks and using the 3D objects counter tool in Fiji (Movie 8). For Movie 8, scale was increased from 124×165 pixels (11 pixels=4.8 µm) to 1240×1650 pixels for clarity.

Live in vivo imaging was performed on anesthetized larvae immobilized in 1% low-melting-point agarose in 35 mm glass-bottomed dishes (MatTek). Widefield microscopy was performed using a 40× objective. Z-stacks were acquired at 10 min intervals. Sixty z-slices were taken at 2 µm sections. Post-acquisition, scale was increased from 155×155 pixels (100.75×100.75 µm) to 1000×1000 pixels and further cropped to a final size of 204×204 pixels around the region of interest (ROI). Confocal microscopy was performed using the Zeiss Invert LSM 710 (Carl Zeiss AG) and the LSM 880 (Carl Zeiss AG) using a 40× and 63× objective. Z-stacks were acquired at 8 min intervals. Sixty z-slices were taken at 0.9 µm sections per larva. For all time-lapse acquisitions, larvae were maintained at 33°C. For mitochondria staining, larvae were injected with 1 nl MitoTracker^® DeepRed (250 µM; Life Technologies) 40 min prior to embedding for live confocal microscopy. For characterization of the cell type that Toxoplasma tachyzoites are invading, the Tg(elavl3:GCaMP6s)^jf4 transgenic line was used. This particular transgenic line employs an enhanced GFP protein variant fused to calcium-binding proteins to read out neuronal calcium flux (Vladimirov et al., 2014); however, we did not employ the calcium sensitivity functionality for our experiments: it was used purely to identify neurons. For confocal microscopy of Tg(elavl3:GCaMP6s)^jf4 larvae, LSM 880 (Carl Zeiss AG) was used with the 63× objective. For Movie 3, a confocal z-stack of 33 z-slices (z=0.43 µm) was acquired. Post-acquisition, scale was increased from 1232×1232 pixels (104.84×104.84 µm) to 2000×2000 pixels. For Fig. 2B, a z-stack of 85 slices (covering 14.525 µm) was acquired using the AiryScan super-resolution (SR) mode and processed using the AiryScan processing tool in Zen Blue (Carl Zeiss AG).

Euthanized larvae were fixed overnight at 4°C in 4% paraformaldehyde supplemented with 0.4% Triton X-100 and washed in PBS, 0.4% Triton X-100 before staining. Briefly, after a 20 min wash in PBS 1% Triton X-100, larvae were incubated overnight at 4°C in blocking solution: PBS supplemented with 10% FBS, 1% dimethyl sulfoxide (DMSO) and 0.1% Tween 20. Primary antibody diluted 1:1000 in blocking solution was applied overnight at 4°C. Larvae were washed 4×15 min with PBS supplemented with 0.1% Tween 20. Secondary antibody diluted 1:500 in blocking solution was applied overnight at 4°C. Larvae were washed 4×15 min with PBS supplemented with 0.1% Tween 20. Hoechst 33342 staining of larvae was carried out at room temperature for 10 min, followed by 3×10 min washes with PBS, 0.1% Tween 20. Larvae were then cleared by sequential incubation in increasing glycerol concentrations: 15% glycerol for 1 h at room temperature, 30% glycerol overnight at 4°C, 60% glycerol overnight at 4°C, 80% glycerol overnight at 4°C, before imaging by AiryScan confocal microscopy (Zeiss LSM 880). Z-stacks were obtained at intervals of 0.46 µm using the AiryScan SR mode.

The primary antibody used was mouse α-GRA2 (#BIO.018.5, BIOTEM), a kind gift from Moritz Treeck, The Francis Crick Institute, London, UK. Secondary antibody used was goat α-mouse AF647 (A-21245, Invitrogen).

Euthanized larvae were fixed overnight at 4°C in 4% formaldehyde (Taab Laboratories Equipment). Hoechst 33342 staining of larvae was carried out at room temperature for 10 min without permeabilization, and larvae were subsequently washed 3×10 min with 0.1 M phosphate buffer (PB). Larvae were embedded in 3% low-melt agarose in 35 mm glass-bottomed dishes. Larvae were covered in 0.1 M PB for high-resolution confocal microscopy (Zeiss LSM 710). Larvae were maintained in 1% formaldehyde in 0.1 M PB until further processing. The embedded larvae were sectioned using a Leica VT1000 S vibrating blade microtome (Leica Biosystems). Fifty-micrometer sections were collected and stored in 0.1 M PB in a 24-well glass-bottomed plate (MatTek). The sections were imaged again using a Zeiss Invert 710 LSM confocal (Carl Zeiss AG) and a 20× Ph2 objective. The sections containing Toxoplasma were then processed following the method of the National Center for Microscopy and Imaging Research (Deerink et al., 2010). In brief, they were post-fixed in 2.5% (v/v) glutaraldehyde/4% (v/v) formaldehyde in 0.1 M PB for 30 min at room temperature, stained in 2% osmium tetroxide/1.5% potassium ferricyanide for 1 h on ice and incubated in 1% w/v thiocarbohydrazide for 20 min, before a second staining with 2% osmium tetroxide and incubation overnight in 1% aqueous uranyl acetate at 4°C. Sections were stained with Walton's lead aspartate for 30 min at 60°C and dehydrated stepwise through an ethanol series on ice, incubated in a 1:1 propylene oxide/Durcupan resin mixture and embedded in Durcupan ACM^® resin according to the manufacturer's instructions (Sigma-Aldrich). Blocks were trimmed to a small trapezoid, excised from the resin block and attached to a SBF SEM specimen pin using conductive epoxy resin (Circuitworks CW2400).

Prior to commencement of a SBF SEM imaging run, the sample was coated with a 2 nm layer of platinum to further enhance conductivity using a Q150R S sputter coater (Quorum Technologies). SBF SEM data were collected using a 3View2XP (Gatan) attached to a Sigma VP SEM (Zeiss). Inverted backscattered electron images were acquired through the entire extent of the ROI. For each 50 nm slice, a low-resolution overview image (pixel size of ∼50 nm using a 1.5 µs dwell time) and several high-resolution images of the different regions of interest (indicated magnification ∼5000×, pixel size of 6-7 nm using a 1.5 µs dwell time) were acquired. The overview image was used to relocate the ROI defined by the confocal images of the sections. The SEM was operated in variable pressure mode at 5 Pa. The 30 µm aperture was used, at an accelerating voltage of 2 kV. Typically, between 300 and 1000 slices were necessary for an entire ROI. As data were collected in variable pressure mode, only minor adjustments in image alignment were needed, particularly where the field of view was altered in order to track the cell of interest.

Prior to ssTEM, the sample was imaged by SBF SEM using low-resolution overview images (100 nm slices, 2 kV, pixel size of 33 nm, 2 µs dwell time, high vacuum with focal charge compensation on at 50%) to relocate the ROI defined by the confocal images. After SBF SEM, the sample was serial sectioned using a UC7 ultramicrotome (Leica Microsystems) and consecutive 70 nm sections were picked up on Formvar-coated 2 mm slot copper grids (Gilder Grids). Then, 143 consecutive sections containing 11 Toxoplasma were viewed at 440×, 890× and 6800× (plus some selected areas at 18,500×) using a 120 kV Tecnai G2 Spirit transmission electron microscope (Thermo Fisher Scientific), and images were captured using a OneView UltraScan^® 4000 camera and GMS3 software (Gatan). The three Toxoplasma shown in Fig. 6 were imaged in their full volume (∼75 sections).

After sectioning for ssTEM, the sample was imaged again in the SBF SEM at low and high resolution (pixel size of 35 nm and 6 nm, 50 nm slices, 2 kV, 2 µs dwell time, high vacuum with focal charge compensation on at 50%), until the macrophage to be imaged by FIB SEM appeared.

All the ssTEM and SBF SEM images were converted to tiff in Digital Micrograph or GMS3 (Gatan), and tiff stacks were automatically aligned using TrakEM2, a Fiji framework plug-in (Cardona et al., 2012). For ssTEM stacks of images, misaligned images were in addition manually aligned to the neighboring sections. Manual segmentations were performed in TrakEM2. For Figs 2A and 6A, labels were exported as tiff for visualization in 3D in ClearVolume, a Fiji plug-in (Royer et al., 2015). For Fig. S5B, they were exported as Amira labels for visualization in 3D in Amira Software (Thermo Fisher Scientific).

FIB SEM data were collected using a Crossbeam 540 FIB SEM with Atlas 5 for 3D tomography acquisition (Zeiss, Cambridge). After image acquisition for SBF SEM and ssTEM was completed, the sample was further sputter coated with a 10 nm layer of platinum.

The ROI was relocated by briefly imaging through the platinum coating at an accelerating voltage of 20 kV and correlating to previously acquired SBF SEM and fluorescence microscopy images. The final ROI and milling orientation were targeted in order to preserve the majority of the sample whilst enclosing the entire cell of interest. On completion of preparation for milling and tracking, images were acquired at 5 nm isotropic resolution throughout the ROI, using a 4.2 µs dwell time. During acquisition, the SEM was operated at an accelerating voltage of 1.5 kV with 1.5 nA current. The EsB detector was used with a grid voltage of 1200 V. Ion beam milling was performed at an accelerating voltage of 30 kV and current of 700 pA.

The final dataset was acquired in two sessions; in order to correct for uneven milled slice thickness in the initial phase of the second session, the first 100 slices were aligned and interpolated to output a per-slice thickness of 5 nm (Atlas 5), prior to reinsertion into the dataset of ∼24.1 µm×17.8 µm×20.2 µm (8651 µm³).

After initial registration (template matching by normalized cross correlation; Fiji, https://sites.google.com/site/qingzongtseng/template-matching-ij-plugin), the images were batch processed to suppress noise, and enhance sharpness and contrast [(1) Gaussian blur 0.8 pixel radius; (2) smart sharpening with highlights suppressed: radius 10 pixels, strength 60%, then radius 1.2 pixels, strength 150%; (3) application of a medium contrast curve; (4) 8-bit grayscale conversion; Adobe Photoshop 2020]. Finally, to fine tune image registration, the alignment to median smoothed template method was applied (AMST; Hennies et al., 2020). Subregions comprising the macrophage and individual Toxoplasma were then cropped out from the final volume.

For Fig. 7A, image stacks from confocal microscopy and FIB SEM were manually aligned to each other using the BigWarp plugin of the Fiji framework (Bogovic et al., 2016), with the FIB SEM stack set as ‘target’ and the confocal stack as ‘moving’ dataset. ‘Landmark mode’ was used to add 19 pairs of corresponding points within both datasets throughout the whole volume of the cell. An affine transformation was applied to the confocal dataset.

Movies 2, 6 and 7 were generated in Fiji, Movie 4 was generated in Amira, and all were compressed in Quick Time Pro with the H.264 encoder.

Anesthetized larvae were imaged at 0, 6 and 24 hpi by fluorescent stereomicroscopy (Leica M205FA). Fifteen z-slices at 8.55 µm intervals covering 128 µm were taken at 130× magnification. Images were further analyzed using Fiji software.

For macrophage ablation, the nitroreductase/metronidazole ablation system was utilized (Curado et al., 2008). Metronidazole is converted by nitroreductase into a cytotoxic metabolite. Therefore, expression of nitroreductase-mCherry (nfsB.mCherry) in the macrophage population in the transgenic line Tg(mpeg1:Gal4-FF)^gl25/Tg(UAS-E1b:nfsB.mCherry)^c264 allows for the specific ablation of macrophages upon metronidazole treatment. Dechorionated 2 dpf TraNac-Tg(mpeg1:Gal4-FF)^gl25/Tg(UAS-E1b:nfsB.mCherry)^c264 larvae were placed in embryo medium supplemented with metronidazole (10 mM; Sigma-Aldrich), 1% DMSO. Larvae were then placed in fresh 10 mM metronidazole solution at 33°C post-infection. Control-treated larvae were maintained in embryo medium supplemented with 1% DMSO.

Significance testing was performed using unpaired Student's t-test, one-way ANOVA or two-way ANOVA (repeated measures with Sidak's/Tukey's multiple comparisons test). For data that do not conform to the assumptions of parametric statistics, Chi-square or Kruskal–Wallis (with Dunn's multiple comparisons) tests were used.

We thank Vincenzo Torraca, Gina Duggan, Margarida Castro Gomes, Joseph Wright and Barbara Clough for scientific discussions; Moritz Treeck for providing the α-GRA2 antibody; Damian Rivett for advice on statistical analysis; and Isaac Bianco and the Bianco laboratory (CDB; UCL) for providing Tg(elavl3:GCaMP6s)^jf4 embryos.