Chronic Toxoplasma gondii infection enhances β-amyloid phagocytosis and clearance by recruited monocytes

All animal experiments were approved according to German and European legislation by the local authorities. Experiments were conducted with 8 weeks old C57BL/6J mice and female and male 5xFAD mice (5xFAD/Tg6799 strain (B6SJL-Tg(APPSwFlLon, PSEN1*M146L*L286V) 6799Vas/Mmjax backcrossed for >10 generations to C57BL/6J, JAX stock#006554) [59, 60]. In this model, Aβ plaques in the brain are detectable from the age of 50 days and their number increases by age [61]. At least five animals per group in up to two independent experiments were used.

T. gondii cysts of the ME49 type II strain were used for this study. Parasites were harvested from the brains of female NMRI mice infected intraperitoneally (i. p.) with T. gondii cysts 4 to 5 months earlier. Brains obtained from infected mice were mechanically homogenized in 1 mL sterile phosphate-buffered saline (PBS) and the cysts were counted using a light microscope. Two cysts were administered i. p. into 8-week-old mice in a total volume of 200 μL. This time point was chosen because at the age of 8 weeks, animals are fertile and are considered adults in this respect.

To specifically ablate CCR2⁺Ly6C^hi monocytes, 66 μg of anti-CCR2 antibody (clone MC-21, provided by M. Mack, University of Regensburg) were administered i. p. on d15, d18, d21, d24 and d27 post infection. Control mice received PBS. On d22 (12-15 h post antibody injection), blood was collected retro-orbitally to confirm depletion. Mice were sacrificed and samples were collected on d28.

At the chronic stage of infection (8 weeks post infection and 4 weeks post infection for depletion and phagocytosis assays, see below), mice were deeply anesthetized and if necessary blood was drawn from the inferior vena cava using a 26G needle and syringe. Subsequently, mice were perfused intracardially with 60 ml sterile ice-cold PBS. Brains were removed and prepared accordingly for further analysis.

Brain hemispheres were removed and immersed in 4 % paraformaldehyde (PFA) for several days. Paraffin-embedded, 4 μm thick sections were deparaffinized and conventionally stained with hematoxylin-eosin (H&E) stain. Immunohistochemical analysis was performed according to our previous publications [12, 13, 62‐66] using a BOND-MAX (Leica Microsystems GmbH/Menarini, Germany) with antibodies against Aβ (clone 4G8, Chemicon, Germany), ionized calcium-binding adapter molecule 1 (IBA1, Wako 019–19741, Germany) to label microglia, glial fibrillary acid protein (GFAP, DAKO Z033401, Germany) to label astrocytes, NeuN (Millipore MAB377, Germany) to label neurons, and anti-Toxo (Dianova DLN-16734, Germany) to label T. gondii. Slides were developed using the Bond^TM Polymer Refine Detection kit (Menarini/Leica, Germany). For the evaluation whole tissue sections were digitized at 230 nm resolution using a MiraxMidi Slide Scanner (ZeissMicroImaging GmbH, Germany) [67].

For immunofluorescent staining, coronal brain sections (16 μm) were prepared with a cryomicrotome (Leica, Germany). Immunolabeling with antibodies against Aβ (4G8, Chemicon, Germany), Iba1 (polyclonal, Wako) and Ly6C (ER-MP20, Acris Antibodies, Germany) were performed overnight at 4 °C after 2 min pretreatment with 98 % formic acid. Secondary antibodies goat anti-rat (Alexa Fluor 488, 1:200, Invitrogen, Germany), goat anti-rabbit (Alexa Fluor 488, Invitrogen, Germany) and goat anti-mouse (Alexa Fluor 594, Invitrogen, Germany) were used. Free floating sections were mounted with ProLong Gold with DAPI (life technologies, Germany). A Zeiss (Carl Zeiss, Germany) microscope equipped with an AxioCam HRc 3 digital camera and AxioVision 4 Software were used to analyze staining and obtain images.

Quantification of Iba1 and Ly6C association with plaques was performed using the ImageJ plot profile function (http://​imagej.​nih.​gov/​ij/​). For this purpose, two perpendicular fluorescence profiles spanning 400 μm and centered over the plaques were measured in immunofluorescence stainings. At least 30 plaques from at least four different tissue sections were analyzed.

For in vivo staining of amyloid plaques, mice were i. p. injected with 10 mg kg⁻¹ methoxy-X04 (Tocris Bioscience) in 5 % DMSO/95 % NaCl (0.9 %) 12 h before brain harvesting and two-photon image acquisition.

Images were processed and superimposed using the Imaris (Version 7.7., Bitplane, Zürich, Switzerland) software. Methoxy-X04-positive objects co-localizing with blood vessels (identified by different tissue autofluorescence) were manually excluded from the analysis. Plaques were automatically detected and quantified in three dimensions using the measurement package of the Imaris software.

Freshly harvested brain tissue was snap-frozen in liquid nitrogen and stored at −80 °C. Aβ₄₂ was quantified in whole brain hemispheres by ELISA against human Aβ₄₂ (Thermo Fisher Scientific, Germany). For total Aβ₄₂ quantification, brains were homogenized in 5 M guanidine hydrochloride buffer (GuHCl) (5 M GuHCl in 50 mM Tris/HCl, pH 8.0) and incubated for 4 h at room temperature on a shaker. For discrimination between small (soluble) and large (insoluble) Aβ₄₂ aggregates, brains were homogenized in carbonate buffer (100 mM Na₂CO₃ in 50 mM NaCl, pH 11.5, 20 μl/mg) and centrifuged. The supernatant was mixed with 8 M GuHCl (in 82 mM Tris/HCl, 0.6 ml GuHCl/1 ml carbonate buffer) to obtain the soluble Aβ₄₂ fraction. The pellet was dissolved in 5 M GuHCl (10 μl/mg) and incubated 3 h at RT on a shaker to obtain the insoluble Aβ₄₂ fraction. Prior to ELISA analysis, each extract (total, soluble and insoluble) was diluted with BSAT-PBS (5 % bovine serum albumin in PBS with 0.03 % Tween-20 and protease inhibitor (Roche, Germany)) according to the expected Aβ₄₂ content. ELISA was then performed according to the manufacturer’s instructions.

After removal, tissue samples from brains were immediately transferred to RNA later (QIAgen, Germany). Total RNA was isolated as previously described with the peqGOLD HP Total RNA Kit (peqlab, Germany) including an on-membrane DNase I digestion (peqlab, Germany) [18].

Relative gene expression was determined similar to previous descriptions [18, 68] using TaqMan® RNA-to-C_T ^TM 1-Step Kit (life technologies, Germany). Reactions were developed in a LightCycler® 480 Instrument II (Roche, Germany). Reverse transcription was performed for 15 min at 48 °C followed by 10 min at 95 °C. Subsequently, 45 amplification cycles were run, comprising of denaturation at 95 °C for 15 s and annealing/elongation at 60 °C for 1 min. TaqMan® Gene Expression Assays (life technologies, Germany) were used for mRNA amplification of HPRT (Mm01545399_m1), IDE (Mm00473077_m1), (IL10, Mm00439616_m1), MMP9 (Mm00442991_m1), NEP (MME, Mm00485028_m1), PSMB8/LMP7 (Mm00440207_m1), PSMB9/LMP2 (Mm00479004_m1) and PSMB10/MECL-1 (Mm00479052_g1). HPRT mRNA expression was chosen as reference for normalization and target/reference ratios were calculated with the LightCycler® 480 Software release 1.5.0 (Roche, Germany). Resulting data were further normalized to values of appropriate control groups.

Isolated brain single cell suspensions were surface stained as described below and sorted on a BD FACSAria^TM III. After sorting, cells were pelleted, any remaining liquid was removed and cells were frozen at −80 °C. Total RNA was isolated using the RNeasy® Mini Kit (QIAgen, Germany). cDNA was synthetized with the iScript™ cDNA Synthesis Kit (BIO-RAD, Germany). Relative gene expression was measured using the TaqMan® Universal PCR Master Mix (Applied Biosystems, Germany). TaqMan® Gene Expression Assays (life technologies, Germany) and data analysis was the same as for RT-PCR from whole-brain homogenates.

For mononuclear cell isolation brains were homogenized in a buffer containing 1 M Hepes pH 7.3 and 45 % glucose and then sieved through a 70 μm strainer as published previously [18]. The cell suspension was washed and re-suspended in 10 mL 75 % Percoll (GE Healthcare, Germany) in PBS and over layered with 10 mL 25 % Percoll in PBS and 5 mL PBS. The gradient was centrifuged for 45 min at 800 g without brake. Cells were recovered from the 25 %/75 % interphase, washed and used immediately for further experiments.

Single cell suspensions were incubated with an anti-FcγIII/II receptor antibody (clone 93) to block unspecific binding and Zombie Violet^TM (Biolegend, Germany), a fixable viability dye. Thereafter, cells were stained with fluorochrome conjugated antibodies against cell surface markers: CD45 (30-F11, eBioscience, Germany), CD11b (M1/70, eBioscience, Germany), Ly6G (1A8, BD Biosciences, Germany), Ly6C (HK1.4, eBioscience, Germany), CCR2 (475301, R&D, USA), TREM2 (237920, R&D, USA), CD36 (HM36, Biolegend, Germany), in FACS buffer (PBS containing 2 % FCS and 0.1 % NaN₃) for 30 min on ice and then washed and fixed in 4 % paraformaldehyde (PFA) for 10 min.

Animals were sacrificed 4 weeks post infection and single cell suspensions were prepared as described above. Live cells remained unstained by the Zombie Violet^TM (Biolegend, Germany) viability dye and were sorted on a BD FACSAria^TM III. 150,000 cells were seeded into each well of a 96-well round bottom plate and allowed to settle down for 1 h in an incubator (37 °C, with 5 % CO₂ and 70 % humidity) before HiLyte Fluor^TM 488-labeled, HFIP monomerized Aβ₄₂ (Eurogentec, Belgium) was added to a final concentration of 500nM. Cells were incubated for 6 h, washed, surface stained and measured by conventional and imaging flow cytometry.

Cell acquisition was performed on a BD FACS Canto^TM II flow cytometer. Data were analyzed using FlowJo software (TreeStar).

Data were acquired with FlowSight^TM (EMD Millipore, USA) with a 20x objective and analyzed using IDEAS software version 6.0. At least 52,000 (control animals) or 500,000 (T. gondii infected animals) cells were acquired for each sample and gated to select images with single cells in good focus (by bright field area/aspect ratio and gradient root mean square of the bright-field image, respectively). To analyze internalization, the erode mask was used on the bright field picture of Aβ⁺ cells to remove 2 pixels from the edges of the starting mask and to define the inner part of the cell. Then, the internalization feature was used to define the ratio of intensity of the Aβ signal between the inside of the cells (defined by the erode mask) and the intensity of the total cell. While cells with little internalization have negative scores, cells with high internalization have positive scores. Here, an internalization > 0 was defined as intermediate to high internalization. Compensation matrix generated by single-color compensation controls was used to correct spectral overlap. Representative pictures from cell populations were chosen.

Results are presented as mean + standard error of the mean (SEM). Statistical analysis was performed with Prism version 6 (GraphPad Software, USA). Different tests were used to compare values, namely Mann–Whitney U test (plaque numbers and volume), Fisher’s LSD test (results with multiple comparisons) and Student’s t test (results with one comparison). p values of p ≤ 0.05 were considered statistically significant.

Infection with T. gondii led to reduced plaque burden as determined by immunohistological staining of Aβ plaques in 5xFAD mice (Fig. 1a and a’). Quantification revealed significantly lower plaque numbers in the cortex of mice infected with T. gondii as compared to transgenic control mice (control 21.9 plaques/mm², T. gondii 5.9 plaques/mm², p < 0.006, Fig. 1b). This observation was further confirmed by in situ two-photon microscopy, which was technically restricted to the observation of cortical layers I and II (Fig. 1c and c’). Here, methoxy-X04 stained plaque volumes of the remaining amyloid plaques were significantly reduced in T. gondii infected animals (controls 1548 ± 205 μm³, T. gondii 466 ± 92 μm³, p < 0.002, Fig. 1d).

Using morphological methods, we detected substantial changes in the cortex and subcortical regions of infected mice such as inflammatory lesions (Fig. 2a’), extensive ionized calcium-binding adapter molecule 1 (Iba1)-reactivity (Fig. 2b’), due to the activation of microglia and myeloid cell infiltration, as well as glial fibrillary acid protein (GFAP)-positive astrogliosis (Fig. 2c’). Non-infected 5xFAD mice presented with known brain histology (Fig. 2a) and only minor activation of resident glia cells (Iba1 and GFAP staining, Fig. 2b, c) [59]. Specific immunodetection for neurons (NeuN labeling) revealed no alterations in the number of neurons in the cortex of infected versus control mice (Fig. 2d, d’). The slightly darker shading in the infected brain is due to increased background staining as previously seen in this infection model in wildtype C57BL/6J mice [18]. Occasionally, T. gondii cysts were detectable (Fig. 2a, inset).

Noting the vast activation of microglia and astrocytes in the infected animals, we have to potentially take into account additional specific and unspecific immunological responses as further factors for the reduction of Aβ [69, 70].

Activation of resident microglia is complemented by recruitment of immune cells from the periphery to the CNS. Collectively, resident and recruited cells form a robust immune response to control T. gondii [18]. Among recruited immune cells, the myeloid compartment plays an important role in cerebral toxoplasmosis [71‐73]. In the following, we focused our analysis on three subsets of myeloid-derived CD45^hi CD11b^hi Ly6G^neg CCR2⁺ mononuclear cells, the CCR2^hi Ly6C^hi monocytes, Ly6C^int cells and Ly6C^low macrophages (gating strategy depicted in Fig. 3a, a’). These subsets of myeloid cells were chosen because age-related defects in the Ly6C^hi monocyte population have been linked to cognitive decline in a murine AD model [74]. First, we detected increased numbers of engrafted Ly6C^hi, Ly6C^int and Ly6C^low cells in the brains of T. gondii infected 5xFAD mice compared to non-infected mice (Fig. 3a, a’).

Subsequently, we determined the expression of certain phagocytosis-related surface molecules and quantified the MFI by flow cytometry: Triggering Receptor Expressed on Myeloid cells 2 (TREM2), CD36, and Scavenger Receptor A1 (SCARA1). We compared their expression on resident and myeloid-derived mononuclear cells and the pattern was similar for all three markers: We found the highest expression of TREM2, CD36 and SCARA1 on Ly6C^low F4/80^hi macrophages and activated microglia and a lower expression on Ly6C^hi CCR2^hi monocytes and resting microglia. Ly6C^int mononuclear cells expressed intermediate amounts of TREM2, CD36 and SCARA1 on their surface (TREM2: microglia 688 ± 25, activated microglia 2191 ± 179, Ly6C^hi 595 ± 48, Ly6C^int 1298 ± 13, Ly6C^low 3222 ± 511, Fig. 3d; CD36: microglia 152 ± 8, activated microglia 1616 ± 281, Ly6C^hi 749 ± 37, Ly6C^int 1175 ± 92, Ly6C^low 1879 ± 208, Fig. 3e; SCARA1: activated microglia 15977 ± 3220, Ly6C^hi 6319 ± 1395, Ly6C^int 10805 ± 2221, Ly6C^low 14964 ± 3088, Fig. 3f).

Resident microglia activation was confirmed by increased CD11c, MHC I and MHC II levels as measured by flow cytometry (data not shown). Thus, our data indicate that cerebral toxoplasmosis induces the recruitment of different myeloid-derived mononuclear cell subsets to the CNS that may contribute to the removal of Aβ by activated microglia.

Resident and recruited immune cell subsets display different phagocytic capacity [23]. Here we compared mononuclear cell subpopulations with respect to their ability to specifically phagocytose Aβ₄₂ in an ex vivo phagocytosis assay. To this end, we freshly isolated brain mononuclear cells and exposed them to Aβ₄₂. Resident surveilling microglia (CD45^int CD11b⁺), activated microglia (CD45⁺ CD11b⁺) as well as Ly6C^hi monocytes, Ly6C^int and Ly6C^low cells (CD45^hi CD11b^hi Ly6G⁻ Ly6C^hi/^int/^low) were distinguished by flow cytometric analysis. The low ability of resident microglia to take up Aβ₄₂ was reflected in their low median fluorescence intensity (MFI) for Aβ₄₂-HiLyte Fluor 488 (1749 ± 88, Fig. 4b, green bar). Upon T. gondii infection, microglia cells became activated, but their Aβ₄₂ uptake remained low (1564 ± 119, Fig. 4b, blue bar). Notably, Ly6C^hi monocytes exhibited the highest MFI (5238 ± 239, Fig. 4b, dark red bar) suggesting greater phagocytic capacity. Similarly, also Ly6C^int and Ly6C^low cells were found to exhibit significantly higher MFI compared to microglia or activated microglia (Ly6C^int 4396 ± 204, Ly6C^low 3890 ± 387, Fig. 4b, medium and light red bars, p < 0.0001). Relative MFIs measured in cells isolated from wildtype mice showed a similar pattern, confirming that Aβ42 uptake was not restricted to 5xFAD cells (Additional file 1: Figure S1).

We further verified that the detected fluorescence resulted from internalized Aβ₄₂ rather than from surface-bound signals. Therefore, we analyzed the cells using imaging flow cytometry that allows obtaining images of individual cells. Gating was performed as described in Fig. 3a and representative pictures for each population are shown in Fig. 4c. Microglia and activated microglia populations contained lower numbers of Aβ⁺ cells (Table 1, first column). Consistent with the previous measurements by conventional flow cytometry, fluorescence was low or absent in resting and activated microglia, but all monocyte populations showed an intense signal (Fig. 4c and Additional file 1: Figure S1). Fluorescence was distributed equally across the cells, sometimes with several additional bright spots inside individual cells. This further indicated the uptake of Aβ₄₂ by recruited mononuclear cells (CD45^hi CD11b^hi Ly6G⁻ Ly6C^hi/^int/^low) upon T. gondii infection. Morphologically, monocytes and the other two monocyte-derived cell subsets were more granular than (activated) microglia, represented by higher side scatter intensities (Fig. 4c and Table 1, last column). We quantified the internalization of Aβ₄₂ by calculating the ratio of the fluorescence intensity within the cell to the intensity of the entire cell. Ratios higher than 0 indicate intermediate to high internalization and were found in more than 97 % of all Aβ⁺ cells regardless of the population (Table 1, second column). The mean internalization varied between populations with activated microglia being the lowest and Ly6C^hi monocytes being the highest (Table 1, third column).

To confirm that it is indeed the recruited monocytes and their progeny contributing to plaque removal, we ablated Ly6C^hi monocytes using a monoclonal anti-CCR2 antibody. We chose a lower antibody concentration to only reduce monocyte numbers, because the complete elimination would highly increase the susceptibility of infected mice as recently described by us [23]. One week after initiating the ablation, we detected significantly reduced Ly6C^hi monocyte levels in the blood (Fig. 5a, b, c). After 2 weeks of anti-CCR2 antibody administration (28 days after infection), Ly6C^hi monocyte numbers were still significantly reduced in the blood (Fig. 5b, c). This peripheral depletion led to a trending reduction of Ly6C^hi monocytes and Ly6C^low monocyte-derived macrophages in the brain on day 28 after T. gondii infection (Fig. 5d). From the isolated brains, we quantified Aβ₄₂ by ELISA. Importantly, diminished Ly6C^hi monocyte numbers were associated with an increased total amount of Aβ₄₂ in the brain of T. gondii infected 5xFAD mice (T. gondii 148 ± 30 ng/ml, T. gondii + anti-CCR2 414 ± 75 ng/ml, Fig. 5e).

While immunohistological methods (Fig. 1) can only identify Aβ plaques, ELISA measurement allows the quantification of total Aβ. Using a sequential protocol, we separated monomeric and small oligomeric Aβ (carbonate-soluble) from larger aggregates (guanidine-soluble) in whole-brain homogenates. Next, we quantified the amount of Aβ₄₂ and found that it was significantly reduced in both fractions (carbonate soluble: control 109.3 ± 21.6 ng/ml, T. gondii 31.1 ± 12.5 ng/ml, p < 0.05; guanidine fraction: control 406.9 ± 79.9 ng/ml, T. gondii 132.3 ± 35.2 ng/ml, p < 0.05, Fig. 6a).

Along with uptake of Aβ, its proteolytic processing is equally important. Therefore, we measured the expression of three Aβ-degrading enzymes in the brain, namely insulin-degrading enzyme (IDE), neprilysin (NEP), and matrix metalloproteinase 9 (MMP9), using RT-PCR, and we observed a significant increase in the expression of MMP9 and IDE mRNA following infection with T. gondii (MMP9: 1.3 ± 0.1 fold-change over 5xFAD controls, p < 0.01, Fig. 6b’; IDE: 1.7 ± 0.1 fold-change over 5xFAD controls, p < 0.01, Fig. 6b”). The expression of NEP remained unaltered (Fig. 6b, 0.94 ± 0.1 fold-change over 5xFAD controls, p > 0.6).

Another suggested degradation pathway is the ubiquitin proteasome system (UPS). The UPS is the major intracellular protein degradation machinery and includes the proteolytic subunits β1i/LMP2, β2i/MECL-1, and β5i/LMP7. We detected a strong increase of mRNA expression for the immunoproteasome subunits β1i/LMP2, β2i/MECL-1, and β5i/LMP7 in whole-brain RNA from T. gondii infected 5xFAD mice (β1i/LMP2: 108 ± 9 fold-change over 5xFAD controls, p < 0.0001, Fig. 6c; β2i/MECL-1: 8 ± 0.7 fold-change over 5xFAD controls, p < 0.0001, Fig. 6c’; β5i/LMP7: 59 ± 2 fold-change over 5xFAD controls, p < 0.0001, Fig. 6c”). Interestingly, detailed investigation of brain resident as well as recruited immune cells in T. gondii infected mice revealed a prominent mRNA expression of β5i/LMP7 in recruited mononuclear cells (Fig. 6d”) in contrast to other Aβ degrading mechanisms. MMP9 mRNA was upregulated in activated compared to surveilling microglia, but almost undetectable in recruited mononuclear cells (Fig. 6d). IDE mRNA was similarly expressed across all innate immune cell populations we investigated (Fig. 6d’). Notably, LMP7 mRNA expression was more than twofold higher in the sorted recruited mononuclear cells in the brain when compared to surveilling or activated microglia, indicating a new mechanism of Aβ clearance. Of note, in the chronic stage of infection the intracellular parasites form cysts within neurons hiding from the immune system, thus the analyzed cells were not directly infected with tachyzoites.

Furthermore, we investigated which cell types are located around the Aβ plaques in the cortex of T. gondii infected 5xFAD mice by performing immunofluorescence stainings against Iba1, Ly6C, and Aβ. Due to their general distribution in the parenchyma, Iba1⁺ microglia and monocyte-derived macrophages were closely associated with Aβ plaques (Fig. 7a, b). Interestingly, Ly6C⁺ cells were not located directly in the vicinity of Aβ plaques (Fig. 7c, d). As to that, it is important not to confuse Ly6C⁺ monocytes with Ly6C expressing endothelial cells [75], which can be recognized by their elongated shape. This qualitative finding was confirmed by quantification of fluorescence intensity around plaques (Fig. 7e, f).

Taken together, our results show that Ly6C^hi, Ly6C^int, and Ly6C^low mononuclear cells are highly capable of removing soluble Aβ peptides. While Ly6C^hi monocytes contribute significantly to this clearance, they cannot be found in the direct vicinity of established plaques and thus, may rather lower the general amount of Aβ to prevent higher molecular weight Aβ aggregates and plaque formation.

Aβ and hyperphosphorylated tau form disease promoting aggregates in AD that trigger chronic cerebral inflammatory processes [76, 77]. Further modulation of the chronic inflammation may occur following several infectious diseases, which are known to induce inflammatory cascades in the CNS. It is well established that certain pathogenic components modulate the course of disease in murine β-amyloidosis models [78]. Although the impact of particular infections during the pathogenesis of AD has been discussed, the underlying mechanisms are especially challenging to untangle (reviewed in [38]). In general, previous studies have found chronic or latent infections to be associated with an increased risk for AD (reviewed in [38]), although evidence regarding T. gondii specifically is inconclusive [54, 55]. On the other hand, a recent study using a mouse model of cerebral β-amyloidosis, T. gondii infection has been associated with a reduced risk of developing AD-like pathology [58]. Our study provides further evidence for this notion, as we detected a significant reduction in the number and volume of β-amyloid plaques in T. gondii infected compared to non-infected 5xFAD mice.

Despite this observed reduction in plaque deposition, one should be cautious in concluding that T. gondii infection may protect against developing AD. In fact, while evidence regarding actual disease risk is sparse and conflicting [54, 55], latent toxoplasmosis in healthy individuals has recently been associated with subtle reductions of cognitive performance in various tasks [79‐81]. Directly translating this finding to the situation of human AD patients is rather difficult, considering the limitations of the applied experimental model. Low dose infection of mice with the parasites is a broadly used model to mimic human chronic Toxoplasma infection [18, 23, 44, 46, 49, 82‐85], although the immune cell recruitment in mice is most likely more pronounced than during latent chronic infection of humans. Considering that we propose the recruited immune cells as major mediators of the beneficial effect, it is unclear which effects are applicable to the human CNS.

As mentioned above, we measured decreased levels of Aβ plaques in the brains of T. gondii infected 5xFAD mice and the remaining plaques were also smaller in volume. These findings are consistent with the aforementioned study by Jung and colleagues, where they further described an improved performance in behavioral tests compared to non-infected transgenic mice [58].

We performed a more detailed investigation of the Aβ load and detected that the general reduction of Aβ plaques was paralleled by a decrease in small soluble Aβ species. Because higher levels of soluble Aβ has previously been linked to decreased cognitive performance [86, 87], this finding could suggest that our experimental setup at least partially protects from cognitive decline.

Controlling the infection with T. gondii requires the collaboration of innate and adaptive immunity [48]. Even though there are inflammatory diseases of the CNS which involve adaptive immune cells, such as multiple sclerosis, the inflammation observed in AD seems to be restricted to innate immune cells [88]. Thus, we focused our further analysis on the contribution of the innate compartment.

The detailed analysis of infiltrating immune cells confirmed the strong recruitment of innate immune cells, particularly CD45^hi CD11b^hi myeloid cells, to the brains of T. gondii infected 5xFAD mice, similar to that seen in wildtype mice. It has been described previously by our group that in mice chronically infected with T. gondii, Ly6C^hi monocytes migrate to the CNS and further differentiate into Ly6C^int mononuclear cells and Ly6C^low macrophages in order to carry out specific tasks in host defense, such as cytokine production and Fc receptor mediated cellular phagocytosis [18, 23].

We analyzed the contribution of these myeloid-derived mononuclear cell subsets in the process of accumulating Aβ plaques that, according to the widely accepted view on AD pathophysiology, ultimately promote neurodegeneration.

Ex vivo phagocytosis assay most likely by an antibody-independent mechanism revealed that all recruited mononuclear subpopulations were able to take up significantly more Aβ₄₂ than microglia from non-infected or activated microglia from T. gondii infected 5xFAD mice. We found that Ly6C^hi monocytes displayed an even higher uptake compared to Ly6C^low cells. Comparing the ability to take up Aβ₄₂ with the ability to take up latex spheres as previously published [23], we noted a difference with respect to the Ly6C^hi and Ly6C^int subsets. While their uptake of latex spheres is very low [23], they displayed a prominent uptake of Aβ₄₂. Even though our results are against the general view that Ly6C^low cells are the most macrophage-like subset, other publications have attributed phagocytic activity to Ly6C⁺ monocytes against parasites [89, 90]. Thus, we conclude that uptake of latex beads and Aβ₄₂ is mediated by different mechanisms with diverse appearance in Ly6C^hi monocytes and Ly6C^low macrophages.

While neuroinflammation has been conventionally reported to be detrimental and associated with several neurological diseases [91‐93], emerging research promotes a more differentiated view on the roles of recruited immune cells in homeostatic and repair mechanisms [12, 13, 27, 94, 95]. Consistent with this concept, there are a growing number of reports indicating the beneficial effect of recruited immune cells in AD and vascular amyloidosis [28‐30, 96‐98].

Performing the ex vivo phagocytosis assay with cells obtained from both wildtype and 5xFAD mice, we also observed that the relative contributions were independent of the genotype, despite absolute values being different. As these differences were most likely caused by the different quantification methods, we conclude that wildtype cells are potent Aβ clearing cells as well. Importantly, this finding suggests that cells probably do not have to be pre-exposed to Aβ to efficiently phagocytose Aβ in a possible treatment strategy. It is somehow challenging that human macrophages were found to be ineffective at Aβ phagocytosis when derived from AD patients [99]. Nevertheless, modulating the route of entry may provide a tool to skew recruited monocytes towards an inflammation resolving phenotype [22] and the capacity of these manipulated monocytes to remove Aβ remains to be investigated, as two studies have found that the replacement of brain resident microglia with peripheral myeloid cells does not reduce the Aβ burden [100, 101]. Both studies used a similar approach to replace microglia with peripheral cells, i. e. depletion of brain resident CD11b-expressing cells during a 10 to 14 days intracerebral ganciclovir treatment of CD11b-HSVTK (herpes simplex virus thymidine kinase) transgenic mice [100, 101]. This treatment leads to a one-time replacement with bone marrow-derived myeloid cells, as opposed to the continuous influx observed in our model of chronic cerebral T. gondii infection. Additionally, Prokop and colleagues point out the lack of an activating stimulus in their model, which would be able to induce the uptake of Aβ by myeloid cells [100]. Even though this lack of stimulation is resolved in our experimental model, finding appropriate stimuli to manipulate the cells is a complex task, as we have to keep in mind that beneficial and detrimental effects of monocytes and macrophages can occur at the same time [102].

The increased amount of Aβ₄₂ detected following CCR2^hi Ly6C^hi monocyte ablation in infected 5xFAD mice points to a causal role of these cells to Aβ clearance. Our findings are supported by a report from Naert and Rivest, who have linked the lack of Ly6C^hi (CX3CR1^low CCR2⁺ Gr1⁺) monocytes to cognitive decline in APP_Swe/PS1 mice [65]. This hypothesis is further strengthened by two very recent studies where myeloid cell recruitment to the CNS was correlated with Aβ plaque reduction [30, 103]. In a very recently published study, Baruch and colleagues proposed a novel treatment strategy to target AD via programmed death-1 (PD-1) inhibition and thereby increasing the recruitment of Ly6C^hi monocytes to the CNS in an IFN-γ dependent manner [104]. The proposed mechanisms included enhanced cellular uptake and degradation. Furthermore, Savage et al. detected phagocytic cells directly associated with plaques, and the CD45^hi status of these cells suggested their myeloid origin [105]. Only short-term recruitment of monocytes did not alter plaque deposition as seen in a mouse model of traumatic brain injury [106].

Having confirmed Ly6C^hi monocytes as key contributors to Aβ removal in our model, we were interested if they migrate into the parenchyma to “attack” Aβ plaques like previous reports have shown [28, 103, 107]. However, CCR2⁺ Ly6C⁺ monocytes were not located in the vicinity of plaques in our experiments. Therefore, we propose that the low plaque burden in the applied experimental model is due to Ly6C^hi monocytes’ increased capacity to remove soluble Aβ rather than due to direct removal of established plaques. In addition, monocyte-derived Ly6C^low macrophages upregulate F4/80 and Iba1, and can be located adjacent to the plaques, similarly to resident microglia.

It has to be carefully investigated, at which stages of disease the recruitment of monocytes and subsequent removal of Aβ is beneficial and can delay the onset of disease, and at which stages the cascade triggered by Aβ is already on its way and additional cell recruitment potentially worsens neuroinflammation [108]. Our data from old animals provides evidence that the ability of freshly recruited immune cells to remove Aβ persists at later stages of experimental amyloidosis.

Searching for a mechanism mediating Aβ uptake, we analyzed the expression of cell surface markers related to phagocytosis on CD11b^hi Ly6G⁻ myeloid-derived cells. The measurements revealed intermediate levels of TREM2, CD36 and SCARA1 on Ly6C^hi monocytes and high levels on Ly6C^low monocyte-derived macrophages. Recent reports of a correlation between genetic TREM2 mutation and AD [109, 110], along with experiments pointing out the anti-inflammatory and phagocytosis-enhancing role of TREM2, have drawn the attention towards this molecule [111‐114]. We detected that monocyte-derived Ly6C^low macrophages expressed high levels of TREM2, in contrast to Ly6C^hi monocytes. This result underlines that, besides TREM2, other factors may determine the capacity of immune cells such as monocytes to phagocytose Aβ. Several studies have suggested that CD36 expression is associated with Aβ uptake [76, 115, 116], consistent with the CD36 expression of Ly6C^low monocyte-derived macrophages. Moreover, the lower expression of CD36 detected on Ly6C^hi monocytes may be beneficial because of less harmful pro-inflammatory CD36-Aβ interaction [117‐120]. Frenkel and colleagues had shown that SCARA1 (and not CD36) mediates phagocytosis of Aβ [118]. However, similar to TREM2 and CD36, SCARA1 expression was not a reliable predictor of the Aβ phagocytic capacity of each myeloid-derived mononuclear cell subset in our model.

Recent research highlights the importance of proteolytic Aβ degradation. Several enzymes are known to digest Aβ, including MMP9 and IDE, but the contribution of each enzyme is crucial. Removal of only one can result in significantly increased cerebral Aβ levels [121], and overexpression leads to decreased Aβ loads [121‐123]. We found both MMP9 and IDE upregulated significantly upon T. gondii infection in 5xFAD mice.

MMP9 is one of several matrix metalloproteinases that have been implicated in Aβ degradation and administration of an MMP inhibitor resulted in increased Aβ loads [121]. Activation of MMPs has to be regarded carefully as well, as a recent study has shown that Aβ increases the permeability of the BCSFB by activation of MMPs [124]. Even though Brkic and colleagues found the biggest changes for MMP3 expression, the contribution of other MMPs cannot be ruled out. IDE hydrolytically cleaves Aβ [125], and the significantly increased expression of IDE, particularly in conjunction with the simultaneously increased MMP9 expression, therefore most likely promoted the enhanced degradation of Aβ in T. gondii infected 5xFAD mice, when compared to non-infected controls. Upregulation of IDE may be a compensatory mechanism, since insulin has been shown to stimulate the growth of T. gondii in vitro [126], and increased Aβ degradation could be a beneficial secondary effect.

In addition, intracellular control of protein homeostasis is mediated by the ubiquitin-proteasome system, whereby proteasomes represent the proteolytically active part. Dysfunction of the UPS has been shown to be an early event in Alzheimer’s disease, suggesting that proteasomes may be unable to properly degrade ubiquitin-tagged proteins [127]. Immunoproteasomes are specific proteasome isoforms that have incorporated the immunosubunits β1i/LMP2, β2i/MECL-1, and β5i/LMP7. Previous data point to an important role of immunoproteasomes in the rapid degradation of oxidant-damaged proteins, thus expression of immunoproteasomes may be beneficial in Aβ clearance [128]. Indeed, recently published data indicate that reactive glia exhibit induced immunoproteasome expression and activation in the cortex of a plaque pathology mouse model [129]. On the other hand, β5i/LMP7-deficiency has been shown to result in attenuation of lymphocytic choriomeningitis virus (LCMV)-induced meningitis [130, 131]. Notably, our data presented here demonstrate the most prominent β5i/LMP7 expression in recruited mononuclear cell subsets, indicating that high immunoproteasome expression in monocytes and their progeny are associated with their enhanced capacity in Aβ degradation and suggesting a novel mechanism of Aβ elimination. However, the exact engagement of the UPS in the mentioned processes requires detailed investigation in forthcoming experiments.