C57BL/6J and RAG1^−/− mice were bred at the animal facility of the Bernhard Nocht Institute for Tropical Medicine, Hamburg, Germany. Animals were kept in individually ventilated cages, in a temperature-controlled (22  ±  2°C) and pathogen-free animal facility under biosafety level 3 laboratory conditions (BSL-3). The light/dark cycle was set to a period of 12 hours. Mice received food and water ad libitum. Age-matched (7 weeks of age) female mice were used for all experiments.

86HG39 cells were used to keep parasites alive in vitro. Cells were maintained at 37°C 5 % CO₂ in complete RPMI (cRPMI, PAN Biotech: 10 % fetal calf serum (FCS), 1% L-glutamine and 0.5 % gentamycin (Capricorn Scientific) and were sub-cultured once a week after 5 min treatment with Trypsin-EDTA (Capricorn Scientific).

To recover cryopreserved blood trypomastigotes, the vials were thawed and incubated with a monolayer of 86HG389 cells (29) overnight. Non-infective parasites were removed by collecting cell culture media, the cell monolayer was washed and fresh complete RPMI was added. T. cruzi-infected 85HG39 cell culture flasks were maintained at 37°C, 5 % CO₂ until trypomastigotes were released again. The time until 86HG39 cells burst was 6 days for T. cruzi Brazil (30).

To increase the infectivity of the parasites, cell culture-derived trypomastigotes were passaged once in RAG^−/− mice. Subsequently, fresh blood of passage mice was collected by heart puncture at the peak of infection using heparinized syringes. Fresh blood diluted in PBS was used to infect the mice groups used for experiments intraperitoneally (i.p.). The infection dose contained 2 x 10³ trypomastigotes in 200 μl PBS per mouse. Not infected controls (n.i.) only received PBS. All mice groups and controls were infected at the same time under BSL-3 conditions. Mice were controlled daily during the experiments. Endpoint analysis took place on the following days post-infection (dpi): 15, 30 (acute stage) and 60, 100, 250 (early chronic stage) and 800 (late chronic stage). For this, mice were anesthetized by CO₂/O₂ overexposure and sacrificed by cervical dislocation. Organs (heart, liver, skeletal muscle, and spleen) were harvested. Samples were stored in liquid nitrogen until DNA isolation. Spleen and muscle tissue were isolated for flow cytometry. Blood was collected via heart puncture for serum analysis.

To determinate free circulating parasites, 2 μl blood was collected from the tail vein, diluted (1:10) in erythrocyte lysis buffer (10 nM tris pH 7.2, 0.15 M ammonium chloride) and incubated for 5 min. Parasites were counted in a 0.02 mm Neubauer cell chamber.

Liquid nitrogen-stored tissue samples were thawed and equilibrated to RT. 25 mg were homogenized in 200 µl lysis buffer and proteinase K (Qiagen) in Precellys ceramic Kit tubes and using a Precellys 24 homogenizer (Peqlab, Erlangen, Germany). DNA isolation was performed according to the manufacturer’s specifications using a QIAamp DNA Mini Kit (Qiagen). To assess the purity and quantity of DNA we used a NanoDrop^® spectrophotometer (Thermofischer). qPCR reactions were prepared using the QuantiTec SYBR Green PCR Kit (Qiagen) and run on a Rotor-Gene 3000 (Corbett Research, Sydney, Australia). Each reaction was performed in a final volume of 20 µL, containing 50 ng DNA and 0.5 µM each of T. cruzi specific primer 121 and 122. Primers target the minicircle variable region from kinetoplast DNA (kDNA) and amplify a 330 bp fragment (31).

121 Fwd: 5´-AAATAATGTACGGGKGAGATGCATGA-3´ and 122 Rev: 5´-GGTTCGATTGGGGTTGGTGTAATATA- 3´. Mouse-specific primers targeting murine GAPDH were used (Fwd: GTCGGTGTGAACGGATTTGG, Rev: TTCCCATTCTCGGCCTTGAC) (32). Thermal profile as previously published (33). The parasite load was calculated automatically using the Rotor-Gene 6000 Series Software 1.7 (Corbett research/Qiagen) by plotting the Ct values against standards of known concentration. GAPDH was used to correct the initial DNA sample mount. qPCR standards were generated by spiking tissue homogenates from n.i. mice with cell-cultured T. cruzi trypomastigotes. Afterwards, spiked DNA was serially diluted with DNA isolated from n.i. mice tissue. Standards ranged from 10⁷ to 10^-3 (limit of detection; LOD) parasite equivalents per 50 ng of total DNA. A standard curve was generated for each organ.

Single-cell suspensions of splenic cells were obtained as described in (34). Briefly, spleens were mashed through a 70 μm strainer into cRPMI medium. Erythrocyte lysis was subsequently performed for 5 min at room temperature (RT). Cells were washed with PBS and filtered through a 50 μm strainer. Muscle cells were isolated as described in (35) with some changes. The tissue was dissected from both hindlimbs. Connective and fatty tissue was removed, samples were washed in cold PBS and minced into fine pieces, then digested in 5 ml of digestion medium (0.2% Collagenase type II (2 mg/ml) and 0.05 % Dispase (0.5 mg/ml) in DMEM at 37°C for 40 min. Finally, cells were resuspended in an appropriate volume of PBS and counted.

Antibodies used for flow cytometric analyses are listed in ( Table 1). Single-cell suspensions were first stained with H-2kb-ANYKFTLV Dextramer (Immundex) in PBS + 2 % FCS for 15 min at RT. Subsequently, surface staining antibodies mixed in Fc-block for surface epitopes were added for 30 min at 4°C. For intracellular cytokine staining, cells were fixed and permeabilized with Foxp3/Transcription Factor Buffer Set (ThermoFisher) for 45 min at RT. After fixation and permeabilization, intracellular staining was performed for 60 min at 4°C. Measurements were performed on a BD Fortessa, a BD LSR II (Beckton Dickinson) Cytometer, and a Cytek Aurora. Flow data were analyzed using the FlowJo 10.8.1 software. Gates were set according to fluorescence minus one (FMO) controls. The gating strategy is depicted in the corresponding Supplementary Figure.

Stimulation of T. cruzi-specific CD8⁺ T cells was performed with T. cruzi protein lysate from a mix of T. cruzi trypomastigotes and amastigotes as described in (26) with some modifications. In brief, cell culture-derived trypomastigotes from the Brazil strain were cultured overnight in Dulbecco’s modified Eagle medium (Mediatech; pH 5.0) to transform them into amastigotes. After that, the amastigote and trypomastigote mix was frozen at −196°C and thawed at 56°C four times. After that, the sample was subjected to sonication. The supernatant was collected after centrifugation at 12,000 x g and sterile-filtered, then the protein concentration was determined. To stimulate immune cells from the muscle, C57BL/6-Ly5.1 splenocytes were incubated in cRPMI (10 % FCS, 1 % L-Glutamine, 0.5 % Gentamycine) with T. cruzi lysate for 5 h. Subsequently, muscle single-cell suspension was added and co-cultured with the antigen-pulsed splenocytes for 5 h at 37°C with 5 % CO₂. To detect the cytokines IFN-γ, TNF-α and GrB, intracellular staining was performed.

Serum samples were obtained by agglutination for 15-20 minutes at RT in Eppendorf tubes followed by centrifugation for 15 min at 5650 × g and stored at −20°C. A custom bead-based Immunoassay LEGENDPLEX™ Murine Cytokine Panel (Customized 13-plex from Biolegend) was performed according to the manufacturer’s instructions. Samples were read on a BD Accuri™ C6 Flow Cytometer.

TIM-3 neutralization was performed in vivo using 100 μg Ultra-LEAF™ (Low Endotoxin, Azide-Free) α-TIM-3 (Biolegend, monoclonal rat IgG) per mouse, injected intra-peritoneally (i.p.) at different time points. Control mice were injected with the corresponding Rat IgG2 a,k (Biolegend).

Statistical significances were analyzed using Graphpad Prism V9 (Graphpad Software, San Diego, USA). Data were analyzed for normal distribution before analysis. All data are shown as mean or mean ± SEM. The statistical significance was determined using a Kruskal–Wallis test with a posthoc Dunn’s test. Correlation analysis was conducted using a Pearson Correlation test. Statistical assumptions were made at 95% confidence level and *p < 0.05, ** p < 0.01, ***p < 0.001 and ****p < 0.0001. If another test was used, it is indicated in the figure legend.

To directly evaluate how reliable our experimental infection was and how successful the parasite control is in our mouse model, we analyzed the parasite burden in different organs: Spleen, liver, heart, and skeletal muscle ( Figure 1). In line with the results of other groups such as Lewis et. al, obtained using visual techniques (33, 36), our infection resulted in high amounts of parasites disseminated throughout the whole organism during acute infection. Initially, during the acute phase, a high parasite load was recorded in all tissues (15-30 dpi). Subsequently, a strong reduction in parasite load was observed over time, with the lowest parasite burden at 60 and 100 dpi, where it could barely be detected in spleen, liver, heart and muscle. However, the parasite burden increased again at later time points. At 800 dpi, several mice displayed a high parasite burden in the muscle. This suggests that although T. cruzi can successfully infect all the organs studied, the skeletal muscle is strongly affected and may serve as a niche for the persistence of the parasites ( Figure 6A).

In terms of immune activation, our infection model is characterized by two completely different phases: The acute and the chronic phase. Accordingly, it is possible to analyze the dynamic nature of cytokine expression during the course of infection in more detail. To this end, we performed a longitudinal analysis of mouse sera from infected mice ( Figure 2). As controls, n.i. mice of the same age were analyzed. It is important to emphasize that there was no age-related increase in cytokine expression, as n.i. mice showed no increase in cytokine levels in the sera with increasing age. The pro-inflammatory cytokines IFN-γ, TNF-α and IL-6 were measured as markers for inflammatory processes. During T. cruzi infection, IFN-γ was strongly increased over the entire course of the infection. It strongly increased at 15 dpi, during the acute phase, compared to n.i. controls and remained high. The highest levels of IFN-γ were measured in the late chronic phase, at 800 dpi. ( Figure 2A). The analysis of TNF-α showed a very similar pattern. It strongly increased during the acute phase of infection, at 15 dpi and 30 dpi, and remained elevated. Like IFN-γ, TNF-α showed a trend to decrease at 60 dpi, 100 dpi and 250 dpi. But, like IFN-γ, the highest levels of TNF-α were measured at 800 dpi, and the differences to n.i. controls were also highly significant at this time point ( Figure 2B). TNF-α is a classical marker of chronic inflammatory processes and is associated with the development of heart disease in Chagas patients (37). IL-6 is released after activation of macrophages, but also by neutrophil granulocytes (38). IL-6 is also released by endothelial cells and somatic cells such as fibroblasts after damage (e.g. tissue destruction). In our model, it was markedly elevated during infection ( Figure 2C). In the acute phase 15 dpi, it was statistically significantly elevated, but not constantly found at this high level, as the concentration fluctuated between different time points. At 60 dpi a scattering was observed: Some animals showed very high levels, while others showed IL-6 levels similar to n.i. control mice. Thus, the difference to the n.i. group is not statistically significant. IL-15 was also analyzed, since it is a pleiotropic cytokine with a broad spectrum of biological functions that can act in an inflammatory or anti-inflammatory manner, depending on the context (39, 40). Interestingly, IL-15 has been reported to be secreted by muscle cells under homeostatic as well as pathological conditions (41, 42) and little is known about its role in T. cruzi infection. Examination of serum IL-15 showed a slight increase at 15 dpi compared to n.i. controls, with a trend towards increased levels as the chronic infection progressed. At 100 dpi, the concentration of IL-15 was significantly higher than in the n.i. controls, and IL-15 was also significantly increased in the late chronic phase ( Figure 2D). In addition, IL-2 levels were analyzed due to its importance for T cell function by promoting T cell proliferation and differentiation into effector T cells (T_eff) as well as ensuring survival and optimal function of T memory and regulatory cells. At 15 dpi, IL-2 was elevated but not significantly different from n.i. controls. It was significantly elevated at 30 dpi, at the end of the acute phase, and remained elevated, though with wide fluctuations, throughout the rest of the chronic infection ( Figure 2E). Lastly, IL-10 still plays a controversial role during T. cruzi infection; it is an anti-inflammatory cytokine, yet an elevated serum IL- 10 concentration is associated with fibrotic changes and cardiovascular disease (43). Analysis of IL-10 showed a strong and significant increase at 15 dpi compared to n.i. controls ( Figure 2F). During the course of the infection, IL-10 levels remained elevated until 30 dpi, but from the onset of the chronic phase (60 dpi), there was a decrease in IL-10. The IL-10 concentration continued to decrease in the further course of the chronic phase. However, at 800 dpi a strong increase was measured. In conclusion, these data show that the levels of pro-inflammatory cytokines in particular remained high throughout the whole course of infection in sera of infected mice, despite a strongly reduced parasite burden in the chronic stage.

We first determined whether co-inhibitory receptors were induced during experimental T. cruzi infection and analyzed the expression of PD-1, LAG-3, and TIM-3 on splenic CD8⁺ T cells at 15 and 30 dpi using flow cytometry. These time points represent the peak of acute infection, with the highest parasitic load in all analyzed tissues. We focused on the activated effector CD8⁺ T cell population, characterized as CD44⁺ CD62L^-, since polyfunctionality of this subset is necessary to limit further parasite dissemination. The gating strategy and representative dot plots are depicted in Supplementary Figures 1A–E. Splenic CD8⁺ T cells showed a significant increase in the expression of all three markers compared to not infected controls (n.i.). Our results revealed a strong and statistically significant increase in the percentage of TIM-3⁺ T cells ( Figure 3A). This molecule is an important regulator of T cell-mediated immune responses and its expression correlates with the strength of activation by the T cell receptor. The n.i. controls showed very low expression of this molecule, with on average 0.65 % of CD8⁺ T cells in the spleen being TIM-3⁺. The TIM-3⁺ population decreased towards the end of the acute phase at 30 dpi, but was still statistically significantly higher than in n.i. controls. The frequency of the PD-1⁺ population also increased in comparison to n.i. controls, but this increase was lower than the increase in TIM-3 ( Figure 3B). These results are in contrast to earlier findings in other infection models, in which strong T cell activation leads to a high transient PD-1 expression (44). LAG-3 also showed increased expression compared to not infected controls (n.i.), which was significant at 15 dpi and also 30 dpi ( Figure 3C). The question arose whether there was co-expression between TIM-3, PD-1 and LAG-3. The proportion of co-expression of these markers within effector CD8⁺ T cells is depicted for 15 dpi in Figure 3D and for 30 dpi in Figure 3E. In summary, the analysis of CD8⁺ T cells showed that strongly activated CD8⁺ T cells in the spleen transiently expressed the co-inhibitory receptors TIM-3, PD-1 and LAG-3. After the establishment of the early chronic T. cruzi infection, when the parasitemia was largely controlled, none of these co-inhibitory receptors could be detected as phenotypic markers for T-cell exhaustion until 250 dpi (Supplementary Figures 1F–H). Interestingly, in the late chronic stage at 800 dpi, PD-1⁺ T cells were once again present in spleen and muscle (Supplementary Figure 1G).

To investigate the influence of the infection on the CD8⁺ T cells in the muscle, tissue isolated at 15 and 30 dpi was enzymatically digested. The immune infiltrate in the muscle was characterized by flow cytometry (Figures 4A–D). In infected mice, the muscle contained a high number of CD3⁺CD45⁺CD8⁺ T cells compared to the muscle of n.i. mice. CD8⁺ T cells in the muscle of infected mice were highly activated and CD44⁺CD62L^- ( Supplementary Figure 2), similar to the results for the spleen. They also expressed the co-inhibitory receptor TIM-3 to a higher extent. The highest frequency of TIM-3⁺ CD8⁺ T cells was found at 15 dpi and 30 dpi, decreasing at later time points (Supplementary Figure 1F). Since only a small percentage of CD8⁺ T cells were PD-1⁺ and the T. cruzi infection did not affect the frequency of PD-1⁺ CD8⁺ T cells (Figure 4A), we focused on the analysis of TIM-3⁺CD8⁺ T cells (Figure 4B). To determine the cytotoxic potential of effector CD8⁺ T cells from muscle tissue, the intracellular expression of GrB and CD107a was determined (Figures 4C, D). CD107a, which is exposed on the cell surface as a result of the degranulation of CD8⁺ T cells, serves as a marker for cytotoxicity. While the CD8⁺ T cells of the n.i. mice expressed neither GrB nor CD107a, the frequency of cells expressing these markers was highly increased in infected mice at 15 dpi. It should be emphasized that GrB could no longer be detected at 30 dpi. This loss of GrB production at 30 dpi was also recently described by Mateus et al. in spleen cells from T. cruzi-infected mice (28). In contrast, analysis of CD107a showed consistently higher levels than the n.i. control without being statistically significant at 30 dpi. It should be noted that the strongest CD107a expression was found in TIM-3⁺CD8⁺ T cells at 15 dpi (not shown). Even if no GrB could be detected, the CD107a expression suggests that degranulation had occurred at 30 dpi.

Based on the results of the previous experiments, we next aimed to investigate the effect of a therapeutic blockade of TIM-3 using α-TIM-3 during the peak of TIM-3 expression. The expected outcome, as shown in other infection models (45, 46), was an enhanced T cell response, thus more effective elimination of the parasite (47). For this reason, reduction of the parasite load in the tissue was used as a parameter to evaluate the effect. A scheme for the administration of antibodies is shown in (Figure 5A). Tissue samples from the spleen, muscle, heart and liver were collected to determine the parasite load. The results of the qPCR for the spleen and muscle are shown (Figure 5B). The TIM-3 blockade only resulted in a trend towards a reduction of parasitic burden in muscle tissue, but this difference was not statistically significant compared to the control isotype antibody-treated group. Thus, we could not demonstrate that blocking TIM-3 has an effect on eliminating the parasite in the muscle reservoir. Furthermore, the parasitic load in the spleen was highly increased. Heart and liver tissue did not show differences in the parasitic burden as a consequence of the α-TIM-3 treatment (data not shown).

Muscle and heart samples were isolated and the parasitic burden was determined by qPCR. In the heart tissue, many samples exhibited a parasite burden below the quantification limit of the qPCR. The mean parasite burden determined for the heart was 3.5 parasites/50 ng heart tissue, while in muscle 8800 parasites/50 ng muscle tissue were found. Comparing the parasite burden in the heart and muscle tissue for each mouse ( Figure 6A), the parasite burden is significantly higher in muscle tissue than heart tissue, and there is no correlation between the number of parasites in muscle and the number of parasites in the heart of the same mouse. Furthermore, a large scattering of the values can be seen, ranging from 1 to 1x10⁶ parasites/50 ng DNA in muscle tissue. This number of parasite equivalents per nanogram tissue corresponded to the observed tissue parasite burden in the acute phase ( Figure 1). This strongly suggests that, although all mice were in good health, some mice in the infection group were less able to control tissue parasite burden in the skeletal muscle over a longer period. The parasite load and the number of activated CD8⁺ T cells present in the muscle were correlated and the r-value of 0.54 indicates a moderate positive correlation between the parasite burden and the number of activated CD8⁺ T cells in the muscle ( Figure 6B).

Next, the T cells from the muscle were further examined by flow cytometry. Representative plots and the gating strategy are depicted in Supplementary Figures 3A, B. In infected mice, a higher infiltration of CD3⁺ T cells could be observed ( Figure 7A). The biggest fraction of infiltrating cells was CD8⁺ T cells ( Figures 7B–D). Using MHC-I dextramer staining against the immunodominant TSKB20⁺ epitope of the T. cruzi trans-sialidase, we also found a strong enrichment of these cells ( Figure 7E). CD44⁺CD62L^- effector CD8⁺ T cells were examined for the expression of activation markers and co-inhibitory receptors. The frequency of effector cells was significatively increased in infected mice ( Supplementary Figure 4A). At this late timepoint, neither TIM-3, LAG-3 or GrB could be detected on CD8⁺ T cells from the muscle of infected mice anymore ( Supplementary Figures 4B–D). CD69 is a marker for early T cell activation, but it also marks non-circulating tissue-resident cells in non-lymphoid tissues (48–50). The percentage of CD69⁺ T cells was significantly higher than in n.i. controls ( Figure 7F). Additionally, comparing T_eff cells from the same mice between the tissues, although the absolute numbers of CD8⁺ T cells in the spleen were comparable to the numbers found in the muscle ( Supplementary Figure 4E), the frequency of CD69⁺CD8⁺ T cells was significantly increased only in T_eff cells from muscle tissue, whereas T_eff cells from the spleen expressed much less CD69 ( Supplementary Figure 4F). In addition, there is a moderate negative correlation between the absolute number of CD8⁺ T cells between the spleen and the muscle, suggesting an accumulation of CD8⁺ T cells in the muscle over time ( Supplementary Figure 4G). While analyzing the PD-1 expression on the isolated CD8⁺ T cells, it became apparent that a high percentage of the very few CD8⁺ T cells found in the muscle of n.i. mice already expressed PD-1 ( Figure 7G). However, upon infection, the absolute numbers of CD8⁺ T cells strongly increased ( Figure 7H). Thus, upon infection, an increased absolute number of PD-1⁺CD8⁺ T cells was found in the muscle.

In conclusion, a strong accumulation of PD-1⁺CD8⁺ T_eff in the muscle was only found after a very long infection period of two years. The high frequency of CD69⁺ CD8⁺ T cells within the muscle, compared to the very low frequency of these cells in the spleen, presents an argument against CD69 induction only as a consequence of an ongoing T cell activation, and supports a tissue-resident phenotype in the muscle. Strikingly, CD8⁺ T cells in the muscle directed against the immunodominant TSKB20 epitope derived from the trans-sialidase of T. cruzi remained PD-1^- (Supplementary Figure 5A).

To further explore whether the PD-1⁺ T cells were exhausted, the expression of the transcription factor TOX was analyzed. The frequency of PD-1⁺ TOX⁺ CD8⁺ T_eff was significantly increased in infected mice ( Figure 8A). A representative staining of PD-1 and TOX gated on CD8⁺ T_eff cells showing clear co-expression is depicted in Supplementary Figure 5B. To test the capacity of PD-1⁺ CD8⁺ T_eff cells to produce cytokines, these cells were stimulated ex vivo. The gating strategy and representative dot plots are depicted in Supplementary Figure 6. Lymphocytes from the muscle of infected mice were co-cultivated with spleen cells isolated from n.i. Thy1.1-congenic C57BL/6 mice. These cells were pre-incubated with T. cruzi lysate, since cells from the spleen contain an appropriate number of antigen-presenting cells (APC) to better stimulate the T cells isolated from the muscle. This experimental set-up takes advantage of the fact that the cells from Thy1.1 C57BL/6 mice express the congenic marker CD45.1 ( Supplementary Figure 6B), whereas cells from a wild-type C57BL/6 mouse express the CD45.2 allele ( Supplementary Figure 6C). Thus, n.i. splenic cells used as APC were CD45.1⁺ and could be discriminated from CD45.2⁺CD8⁺ T cells from muscle tissue of infected mice. The ability of PD-1⁺ TOX⁺ T cells to produce pro-inflammatory cytokines and cytotoxic granules after stimulation was examined. The gating strategy and representative dot plots for this staining are depicted in Supplementary Figure 6D. Production of IFN-γ, TNF-α and GrB was found to be impaired ( Figures 8B–D).

In summary, the experiments in the late chronic phase showed that a very long infection period can indeed lead to an accumulation of dysfunctional CD8⁺ T cells, which are characterized by a high degree of co-expression of the co-inhibitory molecule PD-1 and the transcription factor TOX. TOX, therefore, appears to play a crucial role in generating and maintaining the exhausted phenotype of CD8⁺ T cells in this model. Nevertheless, the CD44⁺CD62L^-CD69⁺PD-1⁺TOX⁺ T cells infiltrating the muscle still appear to exert control over the parasite burden in the tissue. However, whether this represents an optimal equilibrium that occurs in the course of chronic infection or represents the beginning of a complete loss of pathogen control must be clarified in further studies.