Experiments were conducted using female, wild type (WT) C57BL/6J mice (8 weeks old; purchased from Janvier Laboratories, Cedex, France). Mice were kept under specific pathogen-free (SPF) conditions. Four to five mice were used for all experimental groups. Non-infected, non-treated mice were used as the control group. All data are representative of three independent experiments. All mice were group-housed in 12h day/night cycles at 22 °C with free access to food and water. Experiments were approved by local authorities according to German and European legislation.

T. gondii cysts of type II strain ME49 were harvested from the brains of female NMRI mice infected with T. gondii cysts 6-10 months earlier, as described previously (28). In short, isolated brains were mechanically homogenized in 1 ml sterile phosphate-buffered saline (PBS), and the number of cysts in the homogenate was determined using a light microscope. Mice were infected with two cysts intraperitoneal or via oral gavage.

First, mice were deeply anesthetized by isoflurane inhalation (Baxter, Halle/Westfallen, Germany). Subsequently, mice were transcardially perfused with sterile ice-cold PBS. Single-cell suspension of mesenteric lymph nodes and spleen were generated by mechanically passing tissue through a 40 μm strainer in PBS complemented with 2% fetal calf serum (FCS). Samples stored in RNAlater (Qiagen, Stockach, Germany) were kept at 4°C overnight and then transferred to -20°C.

To isolate brain immune cells, brains were homogenized in a buffer containing 1 M HEPES (pH 7.3) and 45% glucose and then filtered through a 70 µm strainer. Leukocytes were separated via Percoll density gradient centrifugation (GE Healthcare, Pasching, Austria) as described previously (32). Living cells were counted using a Neubauer chamber and trypan blue staining.

Single-cell suspensions were incubated with an anti-FcγIII/II receptor antibody (clone 93, eBioscience, Frankfurt, Germany) to block unspecific binding and Zombie NIR™ (BioLegend, Koblenz, Germany), a fixable viability dye. Thereafter, cells were stained with fluorochrome-conjugated antibodies against cell surface markers: CD45 (30-F11), CD11b (M1/70), Ly6C (HK1.4), MHCI (28-14-8), and MHCII (M5/114.15.2), all purchased from eBioscience; CD3 (17A2), CD4 (RM4-5), CD8α (53-6.7), and CD80 (16-10A1) all purchased from Biolegend; and Ly6G (1A8) purchased from BD Biosciences in FACS buffer (with 2% FBS, 0.1% NaN3) at 4°C for 30 min and then fixed in 4% paraformaldehyde (PFA, Affymetrix, Santa Clara, CA, USA) for 15 min. Matched FMO controls were used to assess the level of background fluorescence in the respective detection channel.

Expression of cytokines was evaluated by ex vivo restimulation with Toxoplasma lysate antigen (TLA). In short, 5x105 cells were resuspended in DMEM, supplemented with 10% FCS and 1% Pen/Strep, and stimulated with 200 µg/mL TLA for 6h at 37°C, 5% CO2. After 2h of stimulation, a combination of brefeldin A (10 µg/mL, BioLegend, Koblenz, Germany) and monensin (10 µg/mL, BioLegend, Koblenz, Germany) was added. Cells were stained for cell surface markers as described above. Stained cells were then fixed in 4% PFA for 15 min and washed twice with Permeabilization buffer eBioscience, Frankfurt, Germany). Cells were then stained against IFNγ for 45 min at 4°C, and resuspended in FACS buffer for acquisition. Matched isotype controls were used to assess the level of non-specific binding.

Flow cytometric analysis was performed on Attune NxT Flow Cytometer (Thermo Fisher) and analyzed with FlowJo (version 10, Flowjo LLC).

Samples stored in RNAlater^® were homogenized in BashingBeads tubes (Qiagen, Stockach, Germany). AllPrep DNA/RNA Mini Kit (Qiagen, Stockach, Germany) was used to isolate DNA and the peqGOLD total RNA kit (Peqlab, Erlangen, Germany) was used to isolate total RNA from the homogenate following the manufacturer’s instructions.

T. gondii burden was determined using the FastStart Essential DNA Green Master kit (Roche, Penzberg, Germany). T. gondii B1 gene was used as target (Tg. B1), and Mus musculus Asl gene (Mm. Asl) was used as a reference. Parasite stage was quantified using bradyzoite-specific Bag1 and tachyzoite-specific Sag1 as targets and Gapdh as reference gene (Power SYBR^® Green RNA-to-CT™ 1-Step Kit, Thermo Fisher, Dreieich, Germany) for. All primers were purchased from TIBMolbiol, Berlin, Germany.

Relative gene expression was determined similar to previous descriptions (33, 34) using the TaqMan^® RNA-to-CT™ 1-Step Kit (life technologies, Darmstadt, Germany). TaqMan^® Gene Expression Assays (life technologies) were used for mRNA amplification of Ifng (Mm00801778_m1), Tnf (Mm00443258_m1), Il12a (Mm00434165_m1), Nos2 (Mm00440485_m1). Expression of Hprt (Mm01545399_m1) was chosen as reference and target/reference ratios were calculated with the LightCycler^® 96 software version 1.1 (Roche, Penzberg, Germany). All results were further normalized to the mean of the wild type infected group.

Cytokine and chemokine profile was characterized using the LEGENDplex™ system (BioLegend, Koblenz, Germany). A more detailed protocol is published (35). Briefly, we used the Mouse Inflammation Panel (13-plex). Serum isolated from WT infected mice was incubated with fluorescence-encoded capture beads to cytokine and chemokine targets, including TNF, IL-1β, IL-6, and IFNγ. The fluorescent signal of analyte-specific bead regions was quantified using flow cytometry, and the concentrations of particular analytes were determined using provided data analysis software (BioLegend, LegendPlex™ software v8.0).

Content samples were collected from the terminal Ileum and stored at -20°C until DNA extraction. DNA was extracted using a phenol-chloroform-based method previously described (36). In brief, 500 µL of extraction buffer (200 mM Tris Roth, Dautphetal, Germany), pH 8.0), 200 µL of 20% SDS Applichem, Darmstadt, Germany), 500 µL of phenol:chloroform:isoamyl alcohol (PCI) (24:24:1) Roth, Dautphetal, Germany) were added per sample. Lysis of bacteria was performed by mechanical disruption using a Mini-BeadBeater-96 (BioSpec, Berlin, Germany) for two times 2 min. After centrifugation, aqueous phase was passed for another phenol:chloroform:isoamyl alcohol extraction before precipitation of DNA using 500 µL isopropanol (J. T. Baker, Radnor, Pennsylvania, USA) and 0.1 volume of 3 M sodium acetate (Applichem, Darmstadt, Germany). Samples were incubated at -20°C for at least several hours or overnight and centrifuged at 4°C at maximum speed for 20 min. Resulting DNA pellet was washed, dried using a speed vacuum, and resuspended in TE Buffer (Applichem) with 100 µg/ml RNase I (Applichem, Darmstadt, Germany). Crude DNA was column purified (BioBasic Inc., Braunschweig, Germany) to remove PCR inhibitors.

16S rRNA gene amplification of the V4 region (F515/R806) was performed according to an established protocol previously described (37). Briefly, DNA was normalized to 25 ng/µl and used for sequencing PCR with unique 12-base Golary barcodes incorporated via specific primers (obtained from Sigma, Steinheim am Albuch, Germany). PCR was performed using Q5 polymerase (New England Biolabs, Frankfurt/Main, Germany) in triplicates for each sample, using PCR conditions of initial denaturation for 30 s at 98°C, followed by 25 cycles (10 s at 98°C, 20 s at 55°C, and 20 s at 72°C). After pooling and normalization to 10 nM, PCR amplicons were sequenced on an Illumina MiSeq platform via 250 bp paired-end sequencing (PE250). Using Usearch8.1 software package (http://www.drive5.com/usearch/) the resulting reads were assembled, filtered, and clustered. Sequences were filtered for low quality reads and binned based on sample-specific barcodes using QIIME v1.8.0 (38). Merging was performed using -fastq_mergepairs – with fastq_maxdiffs 30. Quality filtering was conducted with fastq_filter (-fastq_maxee 1), using a minimum read length of 250 bp and a minimum number of reads per sample = 1000. Reads were clustered into 97% ID OTUs by open-reference OTU picking and representative sequences were determined by use of UPARSE algorithm (39). Abundance filtering (OTUs cluster >0.5%) and taxonomic classification were performed using the RDP Classifier executed at 80% bootstrap confidence cut off (40). Sequences without matching reference dataset were assembled as de novo using UCLUST. Phylogenetic relationships between OTUs were determined using FastTree to the PyNAST alignment (41). Resulting OTU absolute abundance table and mapping file were used for statistical analyses and data visualization in the R statistical programming environment package phyloseq (42).

With the exception of microbiome data, all datasets were statistically analyzed using GraphPad Prism 7.02 (GraphPad software). We used either two-way or one-way ANOVA with uncorrected Fischer’s LSD tests or unpaired two-tailed Student’s t-tests. Owing to the small sample sizes, unequal variances were assumed in all t-tests. The significance level was set to *p < 0.05, **p<0.01, ***p<0.001 for all statistical comparisons, and two-tailed p-values are reported for all t-tests. Data are presented as mean ± SEM.

Statistical analyses of microbiome data were performed using R v3.6.1 and the packages phyloseq (42) and ggplot2 (43). For Mann-Whitney U tests, p values lower than 0.05 were considered as significant after multiple testing correction (Benjamini-Hochberg false discovery rate correction). The permutational multivariate ANOVA (ADONIS) were computed with 999 permutations. For ADONIS tests, an R2 > 0.1 (effect size, 10%) and p value < 0.05 were considered as significant.

The environments T. gondii encounters when entering the digestive tract versus the peritoneal cavity differ fundamentally. In order to determine if this difference in the barrier to entry affects the course of infection, we infected mice with a low-dose, two cyst (ME49) infection either orally (p.o. Tg) or intraperitoneally (i.p. Tg). All mice were weighed daily until 28 days post infection (d.p.i.) ( Figure 1A ). Compared to naïve animals, p.o. infected mice began to lose weight at 7 d.p.i. (dotted line with *), which remained significantly lower until the end of the experiment (28 d.p.i.). Mice infected i.p. began to lose weight a day later at 8 d.p.i., which also remained significantly lower until the end of the experiment. When comparing the weight change between the infection routes, we observed a higher gain of weight by p.o. infected animals starting at 14 d.p.i. that persisted until the end of the experiment (dotted line with #). The difference in weight recovery suggests an enhanced inflammatory response upon i.p. infection, which persists over time. In order to determine if the route of infection is affecting the parasites viability and overall associated immune response, we measured the parasite burden in the terminal ileum, spleen, lung, and brain of mice at 7, 14, and 28 d.p.i. These dates were chosen for two primary reasons. First, they were critical days in the weight change. Second, these time points coincide with the early, acute, and chronic phases of infection. As expected, we observed a significantly higher parasite burden in the ileum of p.o. infected mice, while the burden in spleen, lung, and brain were comparable ( Figure 1B ). Parasites were also detected in the ileum after i.p. challenge. By 14 d.p.i., the parasite was primarily found in lung and brain tissue as it penetrated the deeper tissues. In i.p. infected mice, there was a higher parasite burden in the lung and a significantly higher burden in the brain ( Figure 1B ). At 28 d.p.i., parasites were not detectable in the peripheral organs and were mainly found in the brain ( Figure 1B ). Similar to 14 d.p.i., there was a trend of higher parasite burden in the brains of i.p. infected mice.

Next, we measured the blood cytokine levels of TNF, IL-1β, IL-12, and IFNγ. We observed that there was a significantly higher level of IL-12 and IFNγ at 7 d.p.i. upon i.p. infection and nominally, but non-significantly, higher expression of TNF and IL-1β ( Figure 1C ). As the course of infection progressed, the difference in cytokine between the infection routes disappeared. This further supports that the parasite induces inflammation more quickly upon i.p. infection. Given the normalization of the immune response starting around 14 d.p.i. implies that once the adaptive immune response is initiated, the systemic cytokine production reaches a level of saturation. As T. gondii infection progresses, most of the parasites are cleared from the periphery, persisting in muscle cells or within neurons in the CNS (44). To determine if the parasite burden is also associated with the progression of T. gondii infection-induced neuroinflammation, we assessed the whole brain cytokine gene expression at 7, 14, and 28 d.p.i. in the brain. At 7 and 14 d.p.i., there was no distinguishable difference between cytokine gene expressions ( Figure 1D ). However, as the chronic phase developed at 28 d.p.i. we observed a trend toward a higher TNF expression and a significant increase in IL-1β, IFNγ and iNOS expression ( Figure 1D ).

Following oral T. gondii infection with a high dose of parasites, mice develop acute necrotizing inflammation in the terminal ileum characterized by strong secretion of pro-inflammatory cytokines IL-1β, TNF, and IFNγ (45, 46). However, around d9-10 p.i., intestinal inflammation begins to subside and parasite burden declines in peripheral organs (47), which aligns with our data. To determine how gut homeostasis is altered through the course of infection, we used qRT-PCR to determine the expression of parasite-related immune markers from terminal ileum of infected animals. When analyzing pro-inflammatory cytokine gene expression, we observed a significant increase in gene expression at d7 p.i. of IL-1β and IFNγ with no change of TNF or iNOS expression in p.o. infected mice when compared to i.p. infection ( Figures 2A–D ). This aligns with the observed difference of parasite burden at d7 p.i. Of note, there was a trend for a reduced iNOS expression at d14 and increased expression at d28 p.i. in p.o. infected mice compared to i.p. ( Figure 2D ). iNOS is able to control aberrant microbiota growth (48), and the trend of increased expression of iNOS may be a sign of disrupted microbiota in p.o. infected mice.

The effect of T. gondii infection on the populations of the ileal microbiome during the early phase of infection (d0-d8 p.i.) have been well described (30, 49–51). However, given that early T. gondii infection is associated with an acute inflammation in the ileum that disrupts the resident microbial communities, we investigated if this early gut disturbance leads to long lasting alterations of the ileal gut microbiota (49, 51). Therefore, we isolated the ileum contents of mice d28 p.i. and analyzed them via 16S rRNA amplicon sequencing. When assessing the alpha diversity, we found that p.o., and not i.p., T. gondii infected mice had a significant reduction in species diversity ( Figure 3A ). Non-metric multidimensional scaling (NMDS) ordination analysis of ileal microbiota composition revealed that the infection route had a significant effect as T. gondii infected mice clustered distinct from controls independent of the infection route ( Figure 3B ). To further assess the microbiota composition, we compared the relative abundances of bacteria at the level of phylum and family ( Figure 3C ). T. gondii-infected mice displayed a strong reduction in families of the phylum Bacteroidetes and a substantial increase in the Firmicutes phylum, greatly skewing the Bacteroidetes/Firmicutes ratio. Of note, a loss of several Firmicutes families was also observed in infected microbiomes that were replaced by members of the Lactobacillaceae family upon p.o. infection, while i.p. infection resulted in the expansion of the Erysipelotrichaceae and Lachnospiraceae families. Using the linear discriminant analysis (LDA) effect size (LEfSe) method, microbiome changes were further evaluated ( Figures 3D, E ; Supplementary Figure 1 ). In agreement to previous studies, we observed increases in the gram-negative, proteobacteria family Enterobacteriaceae (49, 51). In particular, we detected an increase of the genus Escherichia/Shigella ( Supplementary Figures 1A–C ), which is known to contain commensal species that aggravate intestinal inflammation (51). There were i.p. infection-specific changes to the microbiome with an increased abundance of the Proteobacteria Desulfovibrio and Firmicutes Turicibacter. Further family-level comparison of naïve and T. gondii-infected mice revealed a reduced abundance of many commensal species upon infection ( Supplementary Figures 1D, E ). Of note, there was a loss in the abundance of known beneficial, gram-positive families Clostridiaceae, Lachnospiraceae, Erysipelotrichaceae and Ruminococceae; as well as gram-negative Bacteriodetes families Muribaculaceae andPrevotellaceae ( Supplementary Figures 1D, E ). These changes highlight that T. gondii infection leads to a disruption of the microbial composition that persists into the chronic stage of infection. This microbial imbalance is highlighted by increases in the abundance of potentially proinflammatory proteobacteria species and a reduction of beneficial commensal species. If and how this microbial imbalance affects the later stages of infection and neuroinflammation remains to be investigated.

To characterize the immune response based on the infection route, we assessed how the frequency of circulating neutrophils, Ly6C^hi monocytes, CD4⁺ and CD8⁺ T, immune cell populations crucial for parasite control, changed over the course of infection using flow cytometry ( Figure 4A ). Consistent with the serum cytokine levels, there was a significant increase of the Ly6G⁺ neutrophil populations on d14 p.i. in the blood of i.p. infected mice compared to p.o. ( Figure 4B ). There were no differences to Ly6C^hi inflammatory monocytes in circulation as the infection progressed ( Figure 4B ). There was a skewing of the CD4/CD8 T cell response between the different infection routes. Orally infected mice displayed an increased CD4⁺ T cell population significantly at d14 p.i. and with a trend at d28 p.i. whereas i.p. infected mice exhibited a higher CD8⁺ T cell population at d28 p.i. ( Figure 4B ). These results support a more severe course of Toxoplasmosis upon i.p. infection.

Since we observed a skewing of the T cell response, and a stronger Th1 associated immune response upon i.p. infection, we next investigated if there was a difference in the induction and activation of T cells. At 7 d.p.i., we observed comparable parasite numbers in the spleens of both infection routes, so we decided to assess splenic DC maturation as well as IFNγ production by CD4 and CD8 T cells via flow cytometry. We characterized three DC subpopulations. After gating for CD11c⁺ cells, we identified: CD11b⁺CD8^- monocyte-derived DCs (mDCs), CD11b^-CD8⁺ DCs (CD8⁺ DCs), and CD11b^-CD8^-B220⁺PDCA-1⁺ plasmacytoid DCs (pDCs) ( Figure 4A ). We observed a non-significant increase in the total number of each DC subpopulation in i.p. infected spleen ( Figure 4C ). To assess DC maturation, we measured the surface expression of MHC II, CD86, and CD80. The expression of MHC II, CD86, and CD80 was not differentiable between infection routes for mDCs and CD8⁺ DCs, whereas there was a higher expression on pDCs in i.p. infected mice ( Figure 4D ). pDCs, known to release high levels of Type-I interferons, are likely contributing to the exacerbation of disease course in i.p. mice. To determine the IFNγ T cell response, T cells were isolated from the spleens of mice and stimulated with toxoplasma lysate antigen (TLA), and their IFNγ production was measured via intracellular flow cytometric analysis. On d7 p.i., we detected a trend for increased CD8⁺IFNγ⁺ T cells in i.p. spleens with no detectable difference in the number CD4⁺IFNγ⁺ T cells ( Figures 4E, F ). When detecting the amount of IFNγ produced (via MFI expression), we found a trend for higher IFNγ production in CD4⁺IFNγ⁺ T cells and a significant increase in CD8⁺IFNγ⁺ T cells ( Figures 4E, F ). By d14 p.i., there continued to be significantly more CD4⁺IFNγ⁺ and a trend for more CD8⁺IFNγ⁺ T cells. However, when assessing the amount of IFNγ produced, there were no detectable differences for either cell type when compared between infection routes ( Figures 4G, H ). These results suggest that there is an earlier activation of effector T cells in i.p. infected animals, which indicates a faster parasite spread upon i.p. injection.

T. gondii infection-induced neuroinflammation is characterized by an early activation of the resident glia and subsequent recruitment of circulating monocytes and effector T cells (49, 51).To determine whether the route of infection affects the development of the neuroinflammatory reponse, we compared the activation status of microglia and recruited Ly6C^hi inflammatory monocytes as well as the IFNγ production by CD4 and CD8 T cells ( Figure 5A ). Microglia and monocyte activation was assessed via the surface expression on days 14 or 28 p.i. of MHC II, MHC I, CD80, F4/80 and CD68. For both microglia and ly6C^hi monocytes, we did not observe any significant differences between p.o. or i.p. infected mice on days 7 or 14 p.i. via Flow cytometric analysis of signature activation markers ( Figures 5B, C ). To control of parasite replication and reactivation of toxoplasmosis, the combined role of effector CD4 and CD8 T cells is required. When looking at the initial recruitment and effector function of both T cells at d14 p.i., we observed no difference in the total number of recruited CD4⁺IFNγ⁺ and CD8⁺IFNγ⁺ T cells or their production of IFNγ ( Figures 5D, E ). As the infection progresses to d28 p.i., there is comparable total numbers of IFNγ⁺ CD4 and CD8 T cells between infection routes, whereas both CD4⁺IFNγ⁺ and CD8⁺IFNγ⁺ cells are producing significantly higher amounts of IFNγ upon i.p. infection compared to p.o. ( Figures 5F, G ). An increased IFNγ production aligns with the worsened disease course observed in i.p. infected animals. Whether this increased severity is due to an increased parasite number or a less effective immune control of the parasite is still to be determined. Together, these results highlight a stronger disease course of Toxoplasmosis with increased numbers of neutrophils, pDC activity, and effector T cell IFNγ production, all known to strongly influence the severity of diseases (52, 53).

We have previously demonstrated that T. gondii-induced neuroinflammation and its inflammatory milieu affect the synaptic protein composition (28, 29), and particularly glutamatergic neurotransmission (54) through chronic infection, we next set out to investigate if the altered inflammatory milieu, especially the enhanced IFNγ production, influenced the gene expression of synaptic markers between infected mice. We previously described and IFNγ-dependent alteration of synaptic composition, regardless of large increases in parasite burden (28). One distinct mechanism for alterations of synaptic compositions is synaptic pruning (55). To determine if there is an alteration in synaptic pruning activity between infection routes, we assessed the gene expression of key synaptic pruning markers C1qa and CD68, markers linked with ‘eat-me signals’ (56–58). When measuring gene expression at days 7, 14, and 28 p.i., we observed that there was no difference in the gene expression at d7 and 14 p.i. whereas there was significantly higher expression of both C1qa and CD68 in i.p. infected mice at d28 p.i. ( Figure 6A ). To determine if there is a brain region-specific difference between infection routes, we investigated the parasite burden and IFNγ expression in the whole brain, cortex, and hippocampus of mice at d28 p.i. Assessing parasite burden, we observed a non-significant increase in the parasite burden of whole brain, no change in the cortex, and a significantly higher parasite burden in the hippocampus of i.p. vs. p.o. infected mice ( Figures 6B–D ). IFNγ gene expression, was significantly higher in whole brain as well as in both cortex and hippocampus ( Figures 6B–D ). To assess changes in synaptic composition, we assessed expression of the pre-synaptic vesicle protein synaptophysin (Syp), the post-synaptic scaffolding protein PSD-95 encoded by the Dlg4 gene, the glutamate transporter EAAT2 encoded by the Slc1a2 gene, the vesicular glutamate transporter 1 (VGLUT1), and the α1 subunit of the GABA-A receptor (GABAAα1) encoded by the Gabra1. In whole brain, there was a significant reduction of all synapse-associated genes in both p.o. and i.p. infected mice compared to healthy controls and a trend toward a stronger reduction of the expression of VGLUT1 and EAAT2 upon i.p. compared to p.o. infection ( Figure 6E ). These results align with previous studies showing a similar influence of infection of synaptic gene expression (28, 29). The cortex similarly displayed a substantial reduction in gene expression for all markers upon infection with either p.o. or i.p. injections. However, when comparing p.o. vs. i.p. infection, there was a stronger reduction of gene expression for all genes upon i.p. infection ( Figure 6F ). In contrast, in the hippocampus, there were no detectable differences upon p.o. or i.p. infection in gene expression of Syn and VGLUT1 ( Figure 6G ). The gene expression of PSD-95, EAAT2, and GABAa1 was significantly reduced upon both p.o. and i.p. infection compared to uninfected controls, with no changes between the infection routes ( Figure 6G ). These results highlight that there is a more severe influence on the synaptic composition of the cortex upon i.p. infection, which aligns with the enhanced disease course observed in these animals.