Longitudinal Microglial Activation in Tau Transgenic P301S Mice Predicts Increased Tau Accumulation and Deteriorated Spatial Learning

Background: P301S tau transgenic mice show age-dependent accumulation of neurofibrillary tangles in brainstem, hippocampus, and neocortex, leading to neuronal loss and cognitive deterioration. However, there is hitherto only sparse documentation of the role of neuroinflammation in tau mouse models. Thus, we analyzed longitudinal microglial activation by small animal 18kDa translocator protein positron-emission-tomography (TSPO µPET) imaging in vivo , in conjunction with terminal assessment of tau pathology, spatial learning, and cerebral glucose metabolism. Methods: Transgenic P301S (n=33) and wild-type (n=18) female mice were imaged by 18 F-GE-180 TSPO µPET at the ages of 1.9, 3.9 and 6.4 months. We conducted behavioral testing in the Morris water maze, 18 F-fluordesoxyglucose ( 18 F-FDG) µPET and AT8 tau immunohistochemistry at 6.3-6.7 months. Terminal microglial immunohistochemistry served for validation of TSPO µPET results in vivo, applying target regions in brainstem, cortex, cerebellum and hippocampus. We compared the results with our historical data in amyloid -β mouse models. Results: TSPO expression in all target regions of P301S mice increased exponentially from 1.9 to 6.4 months, leading to significant differences in the contrasts with wild-type mice at 6.4 months (+11-23%, all p<0.001), but the apparent microgliosis proceeded more slowly than in our experience in amyloid-β mouse models. Spatial learning and glucose metabolism of AT8-positive P301S mice were significantly impaired at 6.3/6.5 months compared to the wild-type group. Longitudinal increases in TSPO expression predicted greater tau accumulation and lesser spatial learning performance at 6.7/6.3 months. Conclusions: Monitoring of microglial activation in P301S tau transgenic mice by TSPO µPET indicates a delayed time course when compared to amyloid-β mouse models. Detrimental associations of microglial activation with outcome parameters are opposite to earlier data in amyloid-β mouse models. The contribution of microglial response to pathology accompanying amyloid-β and tau over-expression merits further investigation.


Introduction
Along with features such as extracellular accumulation of amyloid-β plaques and neuroinflammation, the intracellular aggregation of misfolded tau protein as neurofibrillary tangles (NFT) constitutes one of the neuropathological hallmarks of Alzheimer disease (AD) [1]. Under physiological circumstances, the microtubule-associated protein tau (MAPT) plays an important role in binding and stabilizing microtubules, regulating axonal transport, interacting with filaments of the cellular cytoskeleton, and probably also contributes to DNA/RNA protection in the nucleus [2]. In AD and non-AD tauopathies, though, the natively soluble and unfolded tau protein undergoes a conformational change via mechanisms such as hyperphosphorylation and misfolding, leading to diminished physiological functions of tau and its accumulation as NFT [3][4][5].
Deposition of hyperphosphorylated tau in brain is associated with neuroinflammation [5], which may exacerbate the ongoing tauopathy and amyloid-β accumulation, while aggravating neuronal degeneration [4,5]. Indeed, the neuroinflammation in AD shows spatial overlap with deposition of amyloid-β and NFT accumulation [6]. Furthermore, particular components of the neuroinflammatory cascade promote the development of NFT [7]. Importantly, the onset of neuroinflammation occurs early in tauopathies, suggesting that biomarkers of neuroinflammation might serve as a tool to predict the individual disease course [8].
The transgenic P301S mouse model accumulates tau in the brainstem [9][10][11], hippocampus [10,12] and cerebral cortex [10], and this accumulation is accompanied by a decline in spatial learning [12]. Immunohistochemical (IHC) analysis revealed increased microglial activation in transgenic P301S mice at five months of age, compared to findings in wildtype mice [13]. Others studies demonstrate the capacity of wildtype mouse microglia to phagocytize NFTs accumulating in brain of P301S mice [14] and likewise in cultured neurons from P301S mice [15], consistent with a dual role of microglial activation in exerting neuroprotective [14] and neurodegenerative effects [15]. However, the time course of microglial neuroinflammation and its net effect on neurodegeneration is not yet established in this mouse model of tauopathy.
Because understanding the role of neuroinflammation in AD and non-AD tauopathies is of crucial importance, we undertook longitudinal monitoring of microglial activation in P301S mice by means of 18 F-GE-180 µPET in vivo, extending a technique we have established in amyloid-β mouse models. The tracer 18 F-GE-180 binds to the 18 kDa translocator protein (TSPO) expressed in activated microglial cells in living mouse brain, showing excellent correlation with ex vivo validation in several different amyloid-β mouse models [16][17][18][19].
We now aimed to test the predictive value of early microglial activation in this tau mouse model by undertaking serial 18 F-GE-180 µPET until 6.4 months of age, augmented by analyses of spatial learning with the Morris water maze test (MWM) and glucose metabolism with 18 F-fluorodesoxyglucose ( 18 F-FDG) µPET. Finally, we made an IHC examination of tau and microglia by AT8, IBA1 and CD68 antisera. Moreover, we compared the temporal kinetics of microglial activation of the tau mouse model with corresponding findings retrieved from our historical amyloid-β mouse model studies.

Animals And Study Design
All experiments were performed in compliance with the National Guidelines for Animal Protection, Germany and with the approval of the regional animal committee (Regierung Oberbayern) and were overseen by a veterinarian. Animals were housed in a temperature-and humidity-controlled environment with 12 h light-dark cycle, with free access to food (Sniff, Soest, Germany) and water. µPET experiments were carried out in homozygous female human tau P301S mice (n = 33), a mouse line expressing the human 0N4R tau isoform with the P301S mutation in exon 10 of the MAPT gene under control of the murine thy1 promoter [10], whereas control studies were conducted in age and sex matched wildtype (WT, n = 18) mice. TSPO µPET examinations were performed in a longitudinal design at baseline (1.9 months of age) and two follow-up measurements (3.9 and 6.4 months of age) (Fig. 1A). 18 F-FDG µPET scans were conducted at the age of 6.4-6.5 months. The MWM test was administered at 12 ± 7 days before the final TSPO µPET scan in P301S (n = 22) and WT (n = 18) mice. After recovery of 2-6 days following the final µPET scans, randomly selected brains from P301S (n = 13) mice and WT (n = 5) mice were processed for IBA1, CD68 and AT8 IHC in the brainstem and cortex. Additional IHC analyses were conducted in small subgroups (n = 3) of P301S mice at 2.7 and 4.8 months of age. Mice intended for IHC were deeply anaesthetized prior to transcardial perfusion with saline followed by 4% paraformaldehyde and subsequent brain extraction. We reprocessed historical µPET 18 F-GE-180 scans from amyloid-β APP/PS1 [19] and App NL−G−F mice [20] for comparison of their longitudinal microglial activation with present findings associated with tau accumulation in P301S mice.

Radiochemistry And µPET Imaging
Radiosynthesis of 18 F-GE-180 was performed as previously described [21], and 18 F-FDG was purchased commercially. µPET imaging was described as reported previously [21]. In brief, all mice were anesthetized with isoflurane (1.5%, delivered at 3.5 L/min) and were placed in the aperture of the Siemens Inveon DPET. 18 F-GE-180 TSPO µPET with an emission window of 60-90 min p.i. was used to measure cerebral TSPO expression, and (on another day) 18 F-FDG µPET with an emission window of 30-60 min p.i. was used for assessment of cerebral glucose metabolism.
Voxel-based comparisons of SUV maps between groups of P301S and WT mice were performed using statistical parametric mapping (SPM, described below) to identify first a suitable reference tissue for µPET quantification. Here, this criterion is met by any brain region in which tracer uptake did not differ with genotype or age, either for 18 F-FDG or for 18 F-GE-180. Our analysis showed that the bilateral nucleus accumbens (NAC, 10 mm ³ ) served adequately as a pseudo reference region for calculation of SUV-ratio (SUVR) values for both µPET tracers (Fig. 1C). We next calculated target-to-reference tissue SUVRs, i.e. SUVR CTX/NAC , SUVR HIP/NAC, SUVR CBL/NAC and SUVR BRST/NAC for 18 F-GE-180 and 18 F-FDG µPET. To quantify longitudinal changes in microglia activation, the percentage change between SUVR at baseline and the last follow-up scan was calculated. To allow longitudinal analysis of combined regions of interest, the area under the curve (AUC) of 8 calculated as previously described [23].

SPM Analysis
For both tracers, whole-brain voxel-wise comparisons of SUV and SUVR images between groups of P301S and WT mice were performed by SPM using SPM8 routines (Wellcome Department of Cognitive Neurology, London, UK) as previously established in our group [9]. This analysis was implemented in MATLAB (version 7.1), as adapted from Sawiak et al. [24] for mouse data. We performed two-sample t-tests, setting a significance threshold of p < 0.05, uncorrected for multiple comparisons.
Behavioral Testing n = 22 P301S and n = 18 WT mice were subjected to a MWM test for spatial learning and memory deficits, which was performed according to a standard protocol [20]. On training days one through five each mouse had to perform four trials per day in the test basin, with maximum time set to 70 seconds. The test trial was performed on day six. For analyses of escape latency and distance during MWM testing, we used the video tracking software EthoVision® XT 13 (Noldus).
Thermofisher: MN1020) and rat monoclonal CD68 (1:500. Bio-rad: MCA1857). After washing in PBS, sections were then incubated in a combination of three secondary antibodies (Alexa 488 goat anti-rabbit, Alexa 594 goat anti-mouse and Alexa 647 goat anti-rat IgG). For long-term preservation, the labelled slices were mounted in DAKO fluorescence mounting medium. After observation with a fluorescence microscopy, the overall density (OD) per region was calculated for IBA1 and CD68 whereas the area-% was calculated for AT8.

Statistics And Calculations
Statistical analyses were performed in SPSS (Version 25, IBM Deutschland GmbH, Ehningen, Germany). The Kolmogorov-Smirnov test served to evaluate normal distribution of all data. Coefficients of variation (CoVs) were calculated as a measure of robustness in group data.
Two-tailed unpaired t-tests were used to compare intergroup readouts (µPET, MWM and IHC) of P301S and age-matched WT mice for all normally distributed readouts. For intergroup comparison of non-normally distributed readouts Mann-Whitney-U-Tests were calculated. Effect sizes between P301S and WT were determined as Cohen's d.
For correlation analyses, Pearson's coefficients of correlation (R) were calculated for normally distributed readouts. For non-normally distributed readouts, Spearman's coefficients of correlation (rS) were calculated.
Cortical TSPO µPET values in P301S were transformed into z-scores relative to WT findings as described previously, and then plotted as a function of age. Our historical cortical TSPO µPET data of APP/PS1 (n = 17) [19] and App NL−G−F (n = 21) [20] mice were reprocessed in the same way. Staining intensity as a function of age was likewise calculated for the current AT8 data in P301S mice and earlier findings of 18 F-florbetaben z-score data in two historical amyloid-β mouse models. For TSPO µPET, we calculated the AUC of z-scores for all mice with successful completion of three or more serial µPET scans. AUC values were adjusted for the maximum z-score in the studied period, normalized to the observation time, and compared between tau and amyloid-β mice by ANOVA with Tukey post hoc correction.
A threshold of p < 0.05 was considered to be significant for rejection of the null hypothesis.

TSPO µPET facilitates monitoring of microglial activation in P301S mice
We first established that microglial activation can be monitored by longitudinal TSPO µPET imaging in P301S mice. To obtain robust TSPO µPET measures, we validated a suitable pseudo reference tissue: among the various possible regions, SUVR scaled TSPO µPET values in the nucleus accumbens (SUVR NAC ) showed more robust group results when compared to conventional SUV scaling, as indicated by lower CoVs. For instance, SUVR NAC CoVs in the brainstem were 6 ± 1% (range, 4 to 7%) whereas the corresponding SUV CoVs were 21 ± 3%; (range, 16-27%; Fig. 1D).

Discussion
We report the first longitudinal in vivo µPET imaging study of microglial activation together with assessment of multiple outcome parameters in a tau mouse model. Our data clearly indicate that µPET with the TSPO tracer 18 F-GE-180 gives reliable assessment of microglial activation in living P301S mice, as proven by the high correlation with specific IHC markers. Analysis of individual TSPO µPET time courses to 6.4 months of age revealed that microglial activation in the tau model mice is temporally delayed relative to comparable findings in two commonly used amyloid-β mouse models. Importantly, longitudinal elevations of TSPO expression in P301S mice predicted aggravated tau accumulation and worse performance in spatial learning; this is opposite to our findings of preserved spatial learning in amyloid-β model mice with early microglial activation model [23]. Levels of glucose metabolism at the late stage were positively associated with longitudinal TSPO µPET increases, but early elevations of TSPO expression predicted stronger hypometabolism in P301S mice. These findings may be reconciled by consideration of the ambivalent role of microgliosis in neurodegeneration, in some circumstances imparting neuroprotection, and in other circumstances marking a more aggressive pathology. This dual role is seemingly decided by the type of pathology (i.e. tau or amyloid-β over-expression), and the time course. Thus, early microglial activation bodes ill in tau mice, but may impart some protection in amyloid-β mice.
We show that transfer of TSPO µPET technology from different amyloid-β mouse models [19,23,25] to the present tau mouse model is feasible without major caveats. As in some former studies [16,20], we successfully validated a suitable pseudo-reference region for TSPO µPET in P301S mice and we were again able to show that this methodology reduces variance at the group level. This SUVR approach supported the detection of robust Another study of PS19 mice found microglial activation especially in the hippocampus to occur ahead of discernible tau accumulation and brain atrophy [27].
As there have been no direct comparisons of the time courses of microglial activation between amyloid-β and tau mouse models, we put a special focus on contrasting longitudinal in vivo TSPO expression of P301S mice against existing data in two common amyloid-β mouse models. To account for natural progression of microglial activation in the aging brain of rodents [17] we compared standardized differences (z-scores) in relation to age among the different mouse models. By this approach, we are able to show for the first time that temporal kinetics of microglial activation differ depending on whether it is driven by tau or amyloid-β pathology. In particular, TSPO expression in response to tau pathology showed attenuated and delayed development when compared to TSPO expression in response to Aβ overexpression. Importantly, there were similar increases of the amounts of AT8 positive tau in P301S mice or fibrillar Aβ in APP/PS1 and App NL−G−F mice with age, indicating that the observed differences were not driven by variant time courses of protein accumulation. In the translational aspect, the onset of Aβ and tau aggregation may precede the start of clinical symptoms in human AD [28] and associations of both proteins with microglial activation have already been shown in human PET studies [6]. Thus, the presence of amyloid-β and tau should be considered (i.e. by PET or CSF) when interpreting time-courses of microglial activation in human neurodegenerative disease, to avoid bias arising from differences in their temporal associations. Furthermore, more detailed studies comparing tau and amyloid-β mouse models employing next generation sequencing or proteomics should resolve possible differences of microglia phenotypes in relation to the two abnormal protein aggregates.
Details of the role and time dependence of neuroinflammation in AD remains a matter of controversy and debate, given the ambivalence of protective and detrimental aspects [3,29,30]. This also accounts for some earlier findings on microglial function in the P301S mouse model. Luo et al. [14] showed in vitro and ex vivo the capability of isolated wildtype microglia to phagocytize tau in brain tissue of P301S mice, implicating a possible protective effect of fully functional microglia. On the other hand, a recent study of Brelstaff et al. [15] indicated that activated microglia can phagocytize neurons of P301S mice, and therefore seemingly mediate deleterious effects. The strength of our data lies in its longitudinal in vivo design, covering a large fraction of the nine-month lifespan of P301S model mice. The compilation of data indicates that higher longitudinal TSPO expression in P301S mice predict higher tau accumulation and worse spatial learning at 6.4 months of age. Interestingly, the corresponding associations with glucose metabolism to 18 F-FDG µPET gave different predictions for baseline and longitudinal measures. While early elevation in TSPO expression was associated with hypometabolism at 6.4 months, we observed higher terminal glucose metabolism in tau mice with TSPO increasing over time.
While the first result suggests an overall deleterious effect of high early TSPO expression on the outcome of P301S mice, the second observation is more consistent with a coupling of microglial activation and glucose metabolism, as observed previously in PS2APP mice [16]. With regard to spatial learning deficits, our earlier study with congruent methodology and design in PS2APP amyloid-β mice showed that early microglial activation in the forebrain strongly correlated with better cognitive performance in MWM [31].
Speculatively, this could indicate different predictive capability of TSPO µPET depending on whether tau or amyloid-β accumulation is the primary driver of microgliosis. Regarding tau mouse models, our observation of higher tau accumulation in mice with early microglial activation is in line with findings of associated tau and neuroinflammation in the forebrain of rTg4510 tau mice [32]. Our data also fit with the observations of attenuated NFT accumulation, reduced neuronal degeneration, and averted cognitive deterioration after pharmacological ablation of senescent microglial and astroglial cells in PS19 mice [33], as well as fitting with the increased tau pathology occurring along with NLRP3 inflammasome activation [7]. In summary, tau-associated microglial activation seems more detrimental than amyloid-β-associated effects. Importantly, our compilation