Reactive microglia drive tau pathology and contribute to the spreading of pathological tau in the brain

Mice

The hTau (Andorfer et al., 2003) (expressing human MAPT and deficient for endogenous mouse Mapt) and Cx3cr1^−/− (Jung et al., 2000) [targeted deletion of Cx3cr1 via insertion of green fluorescence protein (GFP) gene into the Cx3cr1 locus] mice were crossed and maintained in the C57BL/6J genetic background and were obtained from the Jackson Laboratory and Dan Littman (HHMI, New York University School of Medicine). The hTauCx3cr1^−/− mice were generated as previously described (Bhaskar et al., 2010). All experimental protocols involving animals were performed in accordance with US National Institutes of Health guidelines on animal care and were approved by the University of New Mexico and Cleveland Clinic Institutional Animal Care and Use Committees.

Antibodies and reagents

MAPT antibodies: mouse monoclonal antibodies AT8, AT180 (Pierce) and Tau5 (Invitrogen); PHF-1 (provided by Peter Davies, Albert Einstein College of Medicine). Inflammatory markers: rabbit polyclonal antibody against ionized calcium binding adaptor molecule 1 (Iba1) (Wako), rat monoclonal antibodies against CD68 and CD45 (Serotec). Other antibodies: mouse monoclonal antibody against GAPDH; mouse monoclonal NeuN (Millipore), total p38 MAPK and activated p38 MAPK (phospho-T180/Y182) (Cell Signaling for western blotting; Invitrogen for Immunohistochemistry); mouse monoclonal antibody against SNAP25 (a kind gift from the late Dr Michael Wilson, University of New Mexico). Other reagents: interleukin-1 receptor antagonist (IL-1Ra) (Kineret®, Amgen).

Tissue preparation and measurement of hippocampal wet weight

Prior to biochemical and neuropathological analysis, the mice were anaesthetized and transcardially perfused with phosphate buffer and the brains were removed. The left hemisphere was immerse fixed in 4% paraformaldehyde and the hippocampi from the right hemisphere were micro-dissected and the wet weights were recorded before freezing in liquid nitrogen for subsequent biochemical analysis.

SDS-PAGE and western immunoblotting

Proteins were homogenized in the Tissue Protein Extraction Reagent (TPER®, Thermo Scientific) and soluble hippocampal lysates were resolved via SDS-PAGE and immunoblotted as previously described (Bhaskar et al., 2010). The dilutions of primary antibodies used were as follows: AT8 1:5000; AT180 1:2500; PHF-1 1:5000; Tau5 1:5000; GAPDH 1:10 000; phospho-p38 MAPK and total p38 MAPK 1:1000; SNAP25 1:2500. See the online Supplementary material for additional details on SDS-PAGE/western blotting.

Immunohistochemistry, quantitative morphometry and Gallyas silver staining

Free-floating random sections (30 µm thick) derived from multiple mouse brains per group were utilized for all of the immunohistochemical and immunofluorescence analysis. The antibody dilutions were: AT8, AT180, NeuN, p-p38 MAPK (T180/Y182) and CD45 at 1:250; Iba1 at 1:500 incubated overnight at 4°C. Secondary antibodies (1:250) conjugated to either biotin (for immunohistochemistry from Vector Laboratories) or Alexa Fluor® dyes (for immunofluorescence from Life Technologies) were used. Sections were then either mounted in DAPI Hardset Reagent (for immunofluorescence) or incubated with Avidin/Biotinylated enzyme Complex (ABC reagent, Vector Laboratories; for immunohistochemistry) reagent for 1 h at room temperature. The immunoreactive signals were revealed by developing sections in SigmaFast® 3,3′-diaminobenzidine (DAB) either with (SigmaFast® DAB with CoCl₂ metal enhancer) or without metal enhancer and urea tablets (Sigma-Aldrich). Bright field and epifluorescence images were acquired using Leica DMR upright fluorescence/bright field microscope. Confocal images were acquired and analysed with Leica TCS-SP and SP-AOBS upright confocal microscopes with Leica confocal software or a Zeiss inverted Meta confocal machine and composed the images and videos via Zeiss Zen software.

For the quantitative morphometry, NIH ImageJ was used to quantify the percentage of CD45 and p-p38 MAPK immunoreactive areas. The number of AT180⁺ and NeuN⁺ neurons, AT8 immunoreactivity and EGFP⁺ microglia in the adoptive transfer experiments were manually scored in four random sections/mouse. See the Supplementary material for additional details on immunohistochemical analysis and quantitative morphometry. The Gallyas silver staining on 30 µm free-floating sections was performed as described (Braak and Braak, 1991; Bhaskar et al., 2010).

Morris Water Maze

The mice were tested in the visible (for 3 days) and hidden (for 4 days) platform task of the Morris Water Maze (Morris, 1984). See Supplementary material for a detailed description.

Isolation and adoptive transfer of microglial cells

Mononuclear cells were isolated from a pool of two brains per group as previously described (Bergmann et al., 1999) (Fig. 5C). Briefly, the mice were anaesthetized, transcardially perfused with phosphate buffer, brains removed and dissociated in 0.25% trypsin/RPMI media. Mononuclear cells were separated via a 30%/70% discontinuous isotonic Percoll® gradient followed by magnetic assisted cell sorting (Dynabead FlowComp™ Flexi kit, Cat # 110-61D; Life Technologies; DSB-X™ Biotin Protein Labeling Kit, Life Technologies; Cat # D-20655) using the CD11b antibody (Millipore) and elution method as per the manufacturer’s protocol. Purified microglia (0.5 × 10⁵ cells/ml, in 50 µl) from donor mice with described genotypes or vehicle (RPMI media) were injected with or without IL-1Ra (Kineret®, Amgen; 3 ng) into the brains of 2-month-old C57BL/6J non-transgenic recipient mice with the following stereotaxic co-ordinates: from Bregma: −1.94 mm posterior; 1.5 mm lateral; 1.5 mm dorsoventral, as previously described by Cardona et al. (2006). After 72 h, the recipient mice were perfused with 4% paraformaldehyde and the brains were processed for immunofluorescence analysis for AT8 and GFP⁺ microglia and quantitative morphometry as described in the Supplementary material.

Gene expression analysis

RNA from the hemi-brain was extracted using TRI Reagent® Solution as described by the manufacturer (Invitrogen). Total RNA (50 ng/µl) was converted to cDNA using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems Inc., ABI) and amplified using specific TaqMan® probes [for CD45 (now known as Ptprc) and Il1B] and Gadph was used as a housekeeping gene for normalization, on the StepOnePlus® Real-Time PCR System (Life Technologies).

Sarkosyl insoluble assay

The Sarkosyl-insoluble fraction of MAPT was isolated from hippocampal tissues as described previously (Greenberg and Davies, 1990) with minor modifications, which have been previously described (Bhaskar et al., 2010).

Primary microglial culture

Microglial cultures were prepared from post-natal Day 3 (P3) pups from Cx3cr1^+/− or Cx3cr1^−/−genotypes as previously described (Saura et al., 2003; Bhaskar et al., 2010). See Supplementary material for a description of the primary microglial culture.

IL1B assay in HEK-Blue™ reporter cells

HEK-Blue™ IL1B cells (InvivoGen) specifically respond to IL1B. Once IL1B binds to its receptor IL-1R on the surface of HEK-Blue™ IL1B cells, this triggers a signalling cascade leading to the activation NF-κB and the subsequent production of secreted embryonic alkaline phosphatase (SEAP). Induced SEAP levels in the supernatant of HEK-Blue™ IL1B cells can be assessed using QUANTI-Blue™, a SEAP detection medium. QUANTI-Blue™ turns blue in the presence of SEAP (Fig. 4C), which can be easily quantified using a spectrophotometer. The intensity of the colour reaction is proportional to the amount of IL1B in the supernatant from hippocampal lysates measured. IL1B levels in the samples can be determined by making a standard curve using standards (recombinant murine IL1B).

Hippocampal lysates were prepared by homogenizing in tissue-protein extraction reagent (T-PER, 78510; Pierce) with protease (p8340; Sigma-Aldrich) and phosphatase (p5726; Sigma-Aldrich) inhibitor cocktails. Lysates were spun down and the supernatant was saved. Supernatant (50 µl) and different concentrations of standards (recombinant murine IL1B) were then transferred onto 150 µl (∼50 000 cells) of HEK-Blue™ IL1B reporter cells (InvivoGen), which respond to mouse IL1B by expressing a reporter gene (SEAP). HEK-Blue™ IL1B reporter cells were incubated overnight and the supernatant of HEK-Blue™ IL1B reporter cells was then incubated with QUANTI-Blue™ for 1 h at 37°C. Induced SEAP levels were measured by spectrophotometry at 620 nm. The intensity of the colour reaction is proportional to the amount of IL1B in the supernatant from hippocampal lysates. IL1B levels in the supernatant were determined by the standard curve.

Deficiency of Cx3cr1 in hTau mice lead to epitope-specific tau hyperphosphorylation and aggregation at 2 months of age

We performed an age-based study to determine how early tau pathology begins and how it progresses at different stages of disease in hTauCx3cr1^−/− mice. Two groups of mice were analysed: (i) hTau mice (Andorfer et al., 2003), which are deficient for endogenous mouse tau and express all six isoforms of non-mutant human tau; and (ii) hTau mice lacking the microglial receptor, CX3CR1 (hTauCx3cr1^−/− mice). Separate groups of mice were aged to 2, 6, 12 or 24 months and the hippocampi were processed for western blot and immunohistochemical analysis to determine tau phosphorylation and activation of p38 MAPK.

Tau phosphorylation on the AT8 (pS202), AT180 (pT231) and PHF-1 (pS396/S404) sites were elevated in hTauCx3cr1^−/− mice as early as 2 months of age compared to hTau mice. Consistent with our prior observations (Bhaskar et al., 2010), 6-month-old hTauCx3cr1^−/− mice showed elevated levels of tau phosphorylated at AT8, AT180 and PHF-1 sites (Fig. 1A) compared to other groups. Upon quantification, we observed AT8-, AT180- and PHF-1-site tau phosphorylation was significantly higher in hTauCx3cr1^−/− mice compared to hTau at 2- and 6-months of age (Fig. 1B–D). By 12 months of age, the hTau mice showed AT8 and PHF-1 site tau phosphorylation that were indistinguishable from that of hTauCx3cr1^−/− mice (Fig. 1A–D). Interestingly, there was an overall reduction in AT8, AT180 and PHF-1 site phosphorylation at older age groups of both genotypes compared to younger groups.

In a previous study, we identified significantly increased activation of p38 MAPK in hTauCx3cr1^−/− mice at 6 months of age (Bhaskar et al., 2010). A biphasic response was observed for active p38 MAPK (phosphorylated at T180/Y182) in the hippocampus of hTau and hTauCx3cr1^−/− mice (Fig. 1E). Notably, active p38 MAPK was already present in 2-month-old hTau and hTauCx3cr1^−/− mice. The level of active p38 MAPK was significantly reduced to undetectable levels in hTau mice at 6 months of age, but remained elevated in hTauCx3cr1^−/− mice (Fig. 1A and E). At later stages (12- and 24-months of age), a modest but stable increase in the level of phospho-p38 MAPK was observed in hTau mice (Fig. 1A and E). Unlike p38 MAPK, we did not detect any significant alterations in the levels of activated ERK/JNK kinases across ages or between the two groups.

Consistent with the biochemical studies, immunohistochemical analysis revealed the presence of AT8⁺ neurons in the dentate gyrus of hTau mice at 6 months of age and continued to exist until 24 months of age, but with a small dip at 12 months (Fig. 2A and C). A significant increase in the AT8⁺ cells in the hilus of the dentate gyrus in hTauCx3cr1^−/− mice was apparent as early as 2 months of age (Fig. 2A and C) and stayed elevated through to 24 months of age (Fig. 2A). A similar pattern of immunoreactivity was also observed for AT180 (Fig. 2B) with elevation of AT180 positive neurons in the hilus of dentate gyrus in 2-month-old hTauCx3cr1^−/− mice compared to hTau mice (Fig. 2B and 2C; lower magnification images are in Supplementary Fig. 2). Notably, the CA3 region of the hippocampus also displayed a similar increase in AT8 and AT180 immunoreactive neurons in hTauCx3cr1^−/− mice as early 2 months of age (Supplementary Fig. 1A and B). Overall, we observed a consistent immunoreactive pattern between biochemical (western blot) and immunohistochemical analysis for AT8 and AT180. For example, there was a small drop in the overall AT8/AT180 immunoreactive area at older ages compared to younger groups (Fig. 2C). However, correlating the tau immunoreactivity in different subregions of the hippocampus (dentate gyrus and CA3) with those observed in the entire hippocampus via western blot analysis suggest subregion-specific variability in tau phosphorylation in hTau and hTauCx3cr1^−/− mice with increasing age.

To determine whether 2-month-old hTauCx3cr1^−/− mice display any signs of pathological aggregation of tau, we performed Gallyas silver staining—a standard method for the detection of aggregated tau, pre- and mature tangles (Braak and Braak, 1991). In comparison to our previous study where we observed silver positive mature tangles in 6-month-old hTauCx3cr1^−/− mice (Bhaskar et al., 2010), a majority of the neurons in the CA1 region (Fig. 2D) displayed granules of silver aggregates within their cell bodies in 2-month-old hTauCx3cr1^−/−, but not in age-matched hTau mice (Fig. 2D). No silver positive granulations were observed in 2-month-old non-transgenic and Cx3cr1^−/− mice (data not shown). To confirm the pre-tangle aggregation of tau, we performed Sarkosyl insoluble assay in the hippocampus of 2-month-old hTau and hTauCx3cr1^−/− mice. Notably, while the total tau (Tau5⁺) was present in the Sarkosyl soluble fractions of both hTau and hTauCx3cr1^−/− mice (Fig. 2E), Sarkosyl-insoluble tau was present only in hTauCx3cr1^−/− mice (Fig. 2E). Taken together, these results suggest that deficiency of microglial CX3CR1 in hTau mice leads to enhanced tau phosphorylation and pre-tangle aggregation as early as 2 months of age.

Reduced SNAP25 levels, abnormal CA1 neuronal architecture and spatial memory impairment and neuronal loss in the hTauCx3cr1^−/− mice

To determine whether build-up of pathological tau in the hTauCx3cr1^−/− mice results in altered synaptic and neuronal structure, we assessed the expression of various synaptic proteins (synaptophysin, syntaxin, synapsin and SNAP25) and the neuronal marker NeuN within the hippocampus. Although we did not detect significant alterations in the levels of synaptophysin, syntaxin and synapsin, the level of SNAP25 was significantly reduced from 6 months to 12 months only in the hTauCx3cr1^−/− mice compared to hTau mice (Fig. 3A and B). Notably, the thickness of NeuN⁺ cells in a stereologically identical region of the CA1 subfield was markedly thinner in hTauCx3cr1^−/− mice at 2 months of age compared to age-matched hTau and non-transgenic mice (Fig. 3C). Finally, as the hTauCx3cr1^−/− mice aged, the wet weight of the hippocampus was significantly reduced compared to hTau mice (Fig. 3D). Together, these data suggest that the hippocampus is affected in hTauCx3cr1^−/− mice as early as 2 months of age and this worsens with age.

We have previously demonstrated that Cx3cr1 deficiency leads to working memory impairment in the 6-month-old hTau mice (Bhaskar et al., 2010). To further assess hippocampal-dependent learning and memory, hTau and hTauCx3cr1^−/− mice were compared with non-transgenic mice for spatial memory in the Morris Water Maze at 6 and 12 months of age. First, the mice were trained for 3 days to escape to a visible platform in a cued paradigm (the mice use visual cues to seek the platform) and the escape latency (time to reach the platform) was calculated. Both hTau and hTauCx3cr1^−/− mice displayed longer escape latencies on the second day of training (Fig. 3E, ‘Visible Platform’). Notably, statistical analysis with group-wise comparisons revealed that on Day 2, both hTau and hTauCx3cr1^−/− mice took a significantly longer time to learn to reach the platform as compared to non-transgenic mice (Fig. 3E). By the third day, there was no difference observed in the escape latency among different groups. For the next four consecutive days, we tested these mice in the hidden-platform paradigm of Morris Water Maze. While non-transgenic animals showed a decrease in escape latency (time required to reach the hidden platform), hTau and hTauCx3cr1^−/− mice took much longer to reach the hidden platform (Fig. 3E). Two-way ANOVA followed by Tukey’s multiple comparisons revealed statistically significant delays in finding the hidden platform for hTau (on Day 4) and hTauCx3cr1^−/− mice (on Day 7) compared to non-transgenic control groups (Fig. 3E). Interestingly, two-way ANOVA with repeated measures comparing the mean latency between different testing days within each genotype suggested that non-transgenic and hTau mice showed differences in latency with highest statistical significance (***P < 0.0001) compared to hTauCx3cr1^−/− mice (*P < 0.05) between Days 4 and 7, suggesting that the latency in hTauCx3cr1^−/− mice plateaus from Day 5 to day 7 (Fig. 3E). Furthermore, no significant differences were observed with respect to swim speed, total distance travelled, proximity to platform and probe test among different groups of mice tested. In summary, these results suggest that while hTau and hTauCx3cr1^−/− mice display variations in their learning abilities compared to non-transgenic mice; their latencies appear to plateau (primarily for hTauCx3cr1^−/− mice on Day 7), suggesting a possible impairment in retention or consolidation in hTauCx3cr1^−/− mice at 6 months of age.

Next, we tested 12-month-old cohorts of mice on both cued and hidden versions of the Morris Water Maze. Notably, no significant difference was observed in the escape latency when two-way ANOVA with multiple comparisons were performed. However, both hTau and hTauCx3cr1^−/− mice displayed significant impairment (Fig. 3F) in proximity to platform [referred to as ‘Proximity Measure’ or ‘Gallagher’s measure’ (Gallagher et al., 1993)], which is one of the most sensitive tests to assess cognitive function (Maei et al., 2009). Finally, in the probe test, only hTauCx3cr1^−/− mice spent less time in the target quadrant (Fig. 3G) and displayed a statistically significant impairment compared to age-matched non-transgenic controls (Fig. 3H). Taken together, these results suggest that deficiency of Cx3cr1 in hTau mice leads to notable alterations in the spatial memory function as early as 6 months of age, which is more pronounced at 12 months of age.

To assess if the behavioural impairment observed in hTauCx3cr1^−/− mice was related to neuronal cell loss, we performed immunohistochemistry and quantitative morphometry for the neuron-specific antigen NeuN in 12-month-old non-transgenic, hTau and hTauCx3cr1^−/− groups of mice (Supplementary Fig. 3). Earlier studies have suggested that hTau mice show significant reductions in the number of neurons in the piriform cortex at 17 months of age (Andorfer et al., 2005). Based on this, we performed immunohistochemical and quantitative analysis in layer II of the piriform cortex. The NeuN⁺ neurons were dense with several layers of cells in 12-month-old non-transgenic brains (Supplementary Fig. 3A), wereas the NeuN⁺ neurons in hTau mice appeared dysmorphic, but the density was only slightly lower compared to non-transgenic piriform cortex (Supplementary Fig. 3B). There was a reduced density of NeuN⁺ cells in hTauCx3cr1^−/− mice compared to non-transgenic and hTau mice (Supplementary Fig. 3C). Upon quantification of the NeuN⁺ area fraction, there was a statistically significant decrease in the 12-month-old hTauCx3cr1^−/− mice compared to non-transgenic controls (Supplementary Fig. 3D). These results, along with thinner CA1 layer in the 2-month-old hTauCx3cr1^−/− mice (Fig. 3C), suggest that deficiency of microglial CX3CR1 in hTauCx3cr1^−/− mice induces neurodegeneration, which occurs several months prior to significant neuronal loss demonstrated in the hTau mice (12 months versus 17 months of age) (Andorfer et al., 2005).

Microglial activation in pre- pathological hTauCx3cr1^−/− mice

To determine whether microglial activation precedes or follows tau hyperphosphorylation and aggregation, we performed immunohistochemical and biochemical analysis in 2-month-old hTau and hTauCx3cr1^−/− mice. Iba1⁺ microglia displayed activated morphology in the CA3 regions of hTauCx3cr1^−/− mice at 2 months of age compared to hTau mice (Fig. 4A). Based on our previous observation that IL1R1 knockout (Il1r1^−/−) mice were resistant against lipopolysaccharide-induced tau hyperphosphorylation and that blocking IL1R1 prevented microglia-induced tau phosphorylation in vitro (Bhaskar et al., 2010), we next studied the mRNA and protein levels of active IL1B. Notably, quantitative real-time PCR analysis revealed a significant increase in Il1b mRNA levels in hTauCx3cr1^−/− mice compared to non-transgenic control mice at 2 months of age (Fig. 3B). Three distinct steps tightly control the production of IL1B (Burns et al., 2003). The first step involves the production of the pro-IL1B protein (p35); this is followed by cleavage of pro-IL1b by caspase 1 to produce the active IL1b protein (p17); which then is released into the extracellular space (Burns et al., 2003). The processing of pro-IL1b involves the activation of a caspase 1-activating complex, known as the ‘inflammasome’ (Lamkanfi and Dixit, 2009). Similar to IL1b, caspase 1 itself undergoes cleavage to produce p20 (20 kDa) and p10 (10 kDa) activation forms that are recruited to the inflammasome complex for the processing of pro-IL1b. Elevated levels of pro-IL1b were observed in both hTau and hTauCx3cr1^−/− mice at 2 months of age (Supplementary Fig. 4A). Upon quantification, an increase in the mean ratio of pro-IL1b/GAPDH was observed in both hTau and hTauCx3cr1^−/− mice compared to non-transgenic mice (Supplementary Fig. 4A and B), but the hTauCx3cr1^−/− mice differed with higher statistical strength than hTau mice (Supplementary Fig. 4B). While the levels of cleaved (active form) IL1b were indistinguishable from 2-month-old hTauCx3cr1^−/− mice compared to hTau mice on a western blot analysis, the levels of p10 cleaved fragment of active caspase 1 were significantly higher in 2-month-old hTauCx3cr1^−/− mice compared to age-matched hTau mice (Supplementary Fig. 4A and C). To confirm that activation of caspase 1 is leading to maturation of IL1b, we performed an active IL1b assay using HEK-Blue™ reporter cells. The HEK-Blue™ reporter cell assay (Invivogen) USES a cell line that expresses IL1R1 (interleukin 1 receptor, type 1), a cognate receptor exclusively for mature IL1B (Fig. 4C). The activation of IL1R1 leads to engagement of MYD88 adapter protein and activation of NF-κB complex to lead to the transcription of SEAP reporter (Fig. 4C) in the media. Levels of SEAP in the media reflect the absolute level of IL-1b in the test samples. This assay is considerably more sensitive than western blot analysis. Soluble lysate from 2-month-old hTauCx3cr1^−/− mice displayed a significant increase in the levels of active IL1B HEK-Blue™ reporter cell assay compared to age-matched non-transgenic mice (Fig. 4D), suggesting the activation of the IL1B pathway in hTauCx3cr1^−/− mice as early as 2 months of age.

Next we assessed the expression of CD68—a surface marker of phagocytic cells. There was no difference in the CD68 immunoreactivity in 2-month-old hTauCx3cr1^−/− mice compared to 2-month-old non-transgenic via immunohistochemical analysis (Supplementary Fig. 4D and E). The 24-month-old hTauCx3cr1^−/− mice showed a marked increase in CD68 in the hippocampus compared to age-matched non-transgenic mice (Supplementary Fig. 4D and E). Upon quantification, the CD68 immunoreactive area was significantly higher in 24-month-old hTauCx3cr1^−/− mice compared to age-matched non-transgenic mice (Supplementary Fig. 4E). A statistically significant increase in the CD68 mRNA levels was observed in 2-month-old hTauCx3cr1^−/− mice compared to age-matched non-transgenic mice in the real-time quantitative PCR analysis (Supplementary Fig. 4F). Together, these results suggest that Iba1⁺ microglial activation, caspase-1 (inflammasome) activation, maturation of IL1B and CD68 expression, correlate with early stages of tau pathology in the hTauCx3cr1^−/− mice.

Microglial activation correlates with the spread of tau pathology in the anatomically connected regions of the hippocampus

Recent evidence suggests that pathologically misfolded and hyperphosphorylated tau can spread like ‘prions’ via anatomically connected regions of the brain (Frost et al., 2009; Liu et al., 2012). To determine if microglial activation correlates with the spread of tau pathology we examined the subicular region of the hippocampus in 2- and 24-month-old hTauCx3cr1^−/− mice. The subiculum (Sb) acts as a relay centre between the projections from the CA1 and the entorhinal cortex (Fig. 4E). The axonal projections from CA1 neurons to the subiculum and onward to the entorhinal cortex form the principal output for the hippocampus and follow a strict anatomical layout. The distal end of the CA1 neurons projects to the proximal end (dendrites) of subiculum neurons (Fig. 4E) thus it provides an opportunity to study the anterograde-based spread of tau pathology. First, only the CA1 neurons of the hippocampus were positive for AT180 in the 2-month-old hTauCx3cr1^−/− mice, while the subiculum showed very few neurons positive for AT180 (Fig. 4F and Supplementary Fig. 2A). Adjacent sections stained with CD45, a cell surface marker for activated microglia (Ho et al., 2005), showed a substantial number of CD45⁺ microglia within the identical subregions of the subiculum (Fig. 4F). By 24 months of age, a majority of subiculum neurons displayed AT180 immunoreactivity (Fig. 4F) in the hTauCx3cr1^−/− mice and this was spatiotemporally correlated with substantial CD45 immunoreactivity in the identical region (Fig. 4F, CD45 and AT180). Notably, a majority of CD45⁺ microglia displayed activated morphology (Fig. 4F). Finally, quantitative analysis revealed that the per cent area immunoreactive for CD45⁺ in the subiculum of hTauCx3cr1^−/− mice was significantly higher as early as 2 months of age than those from age-matched non-transgenic mice (Fig. 4G). Interestingly, while a very small portion of AT180 positive neurons were present in the subiculum of 2-month-old hTauCx3cr1^−/− mice, their number was not significantly higher than those in the non-transgenic group (Fig. 4H). We observed a similar trend in AT8⁺ neurons in the subiculum of 2- versus 24-month-old hTauCx3cr1^−/− mice. Notably, the subiculum of hTauCx3cr1^−/− mice was devoid of any AT8⁺ neurons at 2 months of age (Supplementary Fig. 5)—a time point at which numerous CD45⁺ cells are already present in the subiculum (Fig. 4F and G). Together, these results suggest that microglial activation may precede the spreading of pathological AT180⁺ and AT8⁺ tau in anatomically connected regions of the hTauCx3cr1^−/− mouse hippocampus. Therefore, microglia may participate in ‘seeding’ and/or ‘spreading’ of pathological (specifically AT8⁺ and AT180⁺) tau species.

Activated microglia induce tau pathology

If CD45⁺ microglial activation precedes the induction of AT8/AT180 site tau phosphorylation, then isolation and administration of such reactive microglia (independent of other CNS cell types) may also be able to induce tau hyperphosphorylation and aggregation inside the brains of naïve, non-transgenic mice. To test this, a series of adoptive transfer experiments were performed. First, Cx3cr1^−/− microglia (GFP positive because of targeted insertion of GFP into Cx3cr1 locus—Fig. 5A) derived from primary microglial cultures were directly injected in to the brains of 2-month-old non-transgenic mice. Seventy-two hours following microglial administration, recipient brains were examined for the presence of GFP⁺ donor microglia. Numerous AT8⁺ dystrophic neurites/processes and cell bodies were detected around the needle track (Fig. 5B). No significant alterations in the AT8 immunoreactivity were observed in the contralateral side (data not shown).

In the next set of experiments, microglia from 6-month-old non-transgenic, hTau and hTauCx3cr1^−/− mice were purified using an established protocol (Bergmann et al., 1999) with minor modifications that included CD11b antibody mediated enrichment of reactive microglia/macrophages (Bhaskar et al., 2014) (Fig. 5C). Purified microglia were microinjected (50 000 cells/injection) into the cortical layer VI of 6-month-old non-transgenic (isogenic) recipient mouse brain (Fig. 5C). Seventy-two hours after adoptive transfer, recipient brains were analysed for the presence of donor microglia within 1–2 mm radius of the needle track. Recipient mice that received hTauCx3cr1^−/− microglia displayed numerous GFP⁺ donor microglia within several millimetres surrounding the tip of the needle track (small white square in Fig. 5D) and they appear activated (Fig. 5D, square inset). Interestingly, the majority of injected microglia have localized to layer VI of the cortex, dorsal caudate-putamen and several of them tangentially migrated via the corpus callosum (Fig. 5D, rectangle inset). Immunohistochemical analysis in the caudate-putamen area of the non-transgenic mice that received vehicle revealed very few AT8⁺ dystrophic neurites (due to needle-injury induced dystrophy) AT8 and no obvious cellular labelling of AT8 (Fig. 5E). On the other hand, non-transgenic recipients that received microglia derived from non-transgenic donors displayed some AT8⁺ dystrophic neurites and some AT8 immunoreactivity appeared intracellular (Fig. 5F). Notably, an increase in AT8⁺ dystrophic neurites and cellular staining was observed in non-transgenic recipients that received microglia from hTau donors (Fig. 5G). This trend was more robust in the recipients that received microglia from hTauCx3cr1^−/− donor mice (Fig. 5H). Notably, some of the pyramidal neurons around the needle track in the cortical region also showed intracellular AT8 labelling (Fig. 5H, inset). To determine if adoptive transfer of reactive microglia also induced AT180-site phosphorylation in the mouse tau within the brains of non-transgenic mice, adjacent brain sections of non-transgenic recipients that received hTauCx3cr1^−/− microglia were stained for AT180. Robust AT180 reactivity was present inside striatal and layer VI of cortical neurons (Supplementary Fig. 6A and B) of recipient mice. Furthermore, similar to AT8⁺ dystrophic neurites, numerous AT180⁺ dystrophic neurites were observed at the site of injection within the striatum (Supplementary Fig. 6C).

Next, to test whether AT8 immunoreactivity spatially overlaps with the presence of reactive microglia from hTauCx3cr1^−/− donors within the brains of recipients, we performed confocal double immunofluorescence analysis for GFP and AT8 (Fig. 5I–K). Interestingly, AT8⁺ dystrophic neurites and cellular structures spatially co-existed near GFP⁺ donor microglia (Fig. 5I–K). Notably, many of the AT8⁺ cells were neurons in the layer VI of the cortex (Supplementary Fig. 7A and B). Confocal 3D reconstruction of 15-µm thick brain slice of the field shown in Supplementary Fig. 7B suggested a clear induction of AT8 site hyperphosphorylation of endogenous mouse tau within several neurons that surround GFP⁺ microglia from hTauCx3cr1^−/− donor mice (Supplementary Video 1). Based on our previous work (Bhaskar et al., 2010) and present observation (Fig. 4B–D) that IL1B secreted from activated microglia may drive inflammation-induced tau hyperphosphorylation, we tested the effect of intereleukin-1 receptor antagonist (IL-1Ra or Kineret®), which is known to compete with IL1B and block IL-1R activation (Clark et al., 2008). Inclusion of IL-1Ra with the inoculum of hTauCx3cr1^−/− microglia resulted in complete elimination of AT8 immunoreactive structures within the recipient mouse brain (Fig. 5L). Next, to determine whether hyperphosphorylation of endogenous tau within the recipient mouse brain leads to the formation of silver positive pre- or mature-tangles, we performed Gallyas silver staining. Non-transgenic mice receiving either vehicle or microglia from non-transgenic donors did not display any silver positive structures/cells (Supplementary Fig. 7C and D). While 72 h post-inoculation is too early to expect any silver positive neurons/dystrophic neurites within the recipient mouse brain, we did observe diffuse peri-nuclear staining in a small population of striatal neurons of the mice receiving microglia from hTau donors (Supplementary Fig. 7E). The intensity of silver staining was more obvious in the striatal and cortical neurons of the recipient mice that received microglia from hTauCx3cr1^−/− donors (Supplementary Fig. 7F and inset). However, these silver positive neurons were very few in number (in the striatum right around the area of needle tip) and appeared to contain early stages of tau aggregates compared to typical tangle-bearing neurons found in human Alzheimer’s disease brain (Supplementary Fig. 7G and inset). Notably, the mice inoculated with hTauCx3cr1^−/− donor microglia along with IL-1Ra showed no silver positive cells/dystrophic neurites (Supplementary Fig. 7H). Next, we performed quantitative analysis of multiple coronal free-floating brain sections from the recipient mice to examine AT8⁺ neurons/intracellular immunoreactivities. We scored four random fields 1–2 mm area on either side of the needle track in four sections per mice (n = 3–5 recipient mice per group). Recipients that received Cx3cr1^−/− microglia derived from primary cultures showed 20–40 AT8⁺ neurons per mm², while those that received vehicle alone or microglia from non-transgenic donors displayed <10 AT8⁺ neurons within the same defined area (Fig. 5M). Remarkably, microglia derived from hTau mice resulted in a notable increase in the number of AT8⁺ neurons. The most significant increase was observed in the recipients of hTauCx3cr1^−/− microglia that show a statistically significant increase in the number of AT8⁺ neurons compared to those receiving vehicle or non-transgenic microglia (Fig. 5M). Inclusion of IL1-Ra into the hTauCx3cr1^−/− microglial inoculum reduced the number of AT8⁺ neurons by nearly 50% (Fig. 5M). Finally, to determine if the effect of IL-1Ra on reduced AT8 and AT180 site tau phosphorylation is mediated via downregulation of p-p38 MAPK, we performed immunohistochemical and triple immunofluorescence analysis for p-p38 MAPK alone or with DAPI and GFP⁺ donor microglial cells. Notably, numerous p-p38 MAPK positive cells (majority of them non-neuronal) were apparent in the area around the needle track (Fig. 6A) and surrounding GFP⁺ donor microglia (Fig. 6C) in the non-transgenic mice receiving hTauCx3cr1^−/− microglia. In contrast, mice receiving hTauCx3cr1^−/− microglia+IL-1Ra displayed markedly reduced number of p-p38 MAPK positive cells in the identical region around the needle track (Fig. 6B). Notably, such reduction in p-p38 MAPK immunoreactivity was also observed around GFP⁺ donor microglia (Fig. 6D). Quantification of the percent area positive for p-p38 MAPK (from the bright field immunohistochemical images) reveal a statistically significant reduction in the mice receiving hTauCx3cr1^−/− microglia+IL-1Ra compared to hTauCx3cr1^−/− microglia alone (Fig. 6E). Taken together, these results suggest that reactive microglia derived from either an in vitro source or from hTauCx3cr1^−/− mice are sufficient and can directly induce AT8 and AT180 site hyperphosphorylation of endogenous mouse tau that can be blocked by an IL-1Ra via reduced activation of p38 MAPK.