A TauP301L mouse model of dementia; development of pathology, synaptic transmission, microglial response and cognition throughout life

Background Late stage Alzheimer’s disease and other dementias are associated with neurofibrillary tangles and neurodegeneration. Here we describe a mouse (TauD35) carrying human Tau with the P301L mutation that results in Tau hyperphosphorylation and tangles. Previously we have compared gene expression in TauD35 mice to mice which develop plaques but no tangles. A similar comparison of other pathological features throughout disease progression is made here between amyloidβ and Tau mice described in Parts I and II of this study. Methods In vitro CA1 patch clamp and field recordings were used to investigate synaptic transmission and plasticity. Plaque load and microglia were investigated with immunohistochemistry. Cognition, locomotor activity and anxiety-related behaviours were assessed with a forced-alternation T-maze, open field and light/dark box. Results Transgene copy number in TauD35 mice fell into two groups (HighTAU and LowTAU), allowing assessment of dose-dependent effects of overexpression and resulting in tangle load increasing 100-fold for a 2-fold change in protein levels. Tangles were first detected at 8 (HighTAU) or 13 months (LowTAU) but the effects on synaptic transmission and plasticity and behaviour were subtle. However, severe neurodegeneration occurred in HighTAU mice at around 17 months, preceded by considerable proliferation and activation of microglia at 13 months of age; both increasing further at 17 months. LowTAU mice at 24 months of age showed a comparable tangle load and microglial proliferation to that occurring at 13 months in HighTAU mice. However, LowTAU mice showed no neurodegeneration at this stage and considerable microglial activation, stressing the dependence of these effects on overexpression and/or age. Conclusions Comparison of the effects of amyloidβ and plaques without tangles in a model of preclinical Alzheimer’s disease to the effects of tangles without amyloidβ plaques in the late stage model described here may clarify the progressive stages of Alzheimer’s disease. While Tau hyperphosphorylation and neurofibrillary tangles are eventually sufficient to cause severe neurodegeneration, initial effects on synaptic transmission and the immune response are subtle. In contrast while even with a heavy plaque load little if any neurodegeneration occurs, considerable effects on synaptic transmission and the immune system result, even before plaques are detectable.


Background
The prevalence of dementia in the ageing population is becoming a major problem for the health care system. Understanding the progression of the disease and the links between pathology and cognitive outcomes is essential if we are to find ways of slowing or preventing the onset of dementia. Alzheimer's disease is the most common form of dementia and, in its later stages, shares common pathologies with frontotemporal dementia and other tauopathies. In all these dementias, the development of neurodegeneration is closely correlated with the hyperphosphorylation of the microtubule associated protein Tau and development of neurofibrillary tangles. In the case of Alzheimer's disease this development is subsequent to a rise in amyloidβ (Aβ) and deposition of plaques, whereas some of the other tauopathies occur as a direct result of mutations in the Tau gene.
To study the relative time course of synaptic and microglial changes and neurodegeneration in relation to phosphorylation of Tau and neurofibrillary tangles, we have characterised a novel strain of mice (TauD35), transgenic for MAPT, which encodes human Tau protein with a mutation that, in man, results in frontotemporal dementia with parkinsonism linked to chromosome 17 [1] . The pathology in this mouse progresses much more slowly than previously described models of Tauopathy [2,3] . Previously we have correlated gene expression and development of neurofibrillary tangles in this mouse and undertaken gene expression network analysis finding that most gene expression changes, including upregulation of microglial genes and decrease in expression of synaptic genes, occur only very late in disease progression, months after neurofibrillary tangles become evident [1] . This is in contrast to more aggressive overexpression models in which neurodegeneration occurs at a young age and synaptic changes are seen even before neurofibrillary tangles are evident [4] . A very interesting study in another Tau model showed synaptic and behavioural changes in middle age (approximately 10 months) which were reversible on turning off the transgene [5,6] .
However, further work is required to elucidate the mechanisms of this progression [7] . It is thought that neurofibrillary tangles may underlie neuronal death by interfering with axonal transport but mitochondrial dysfunction, leading to calcium homeostasis breakdown and apoptosis, likely also contributes [7] . Evidence suggests that caspases are involved, especially at earlier stages [8][9][10] in these aggressive models but in the slower progressing model TauD35 mice, while some caspases show increased expression, this again occurs only late in disease progression [1; www.mouseac.org] .
The more aggressive models, such as rTg4510, are useful in that they start to develop neurofibrillary tangles very early and show rapid neurodegeneration, which is convenient in terms of cost and various practical considerations and interesting data have emerged [2,3] . However, it makes the separation of stages of disease progression difficult and means that it occurs in mice at ages equivalent to adolescence through to early middle age rather than old age, as is most common in Alzheimer's disease. In the present study we characterised mice developed at GlaxoSmithKline (TauD35) with the MAPTP301L mutation also on the CamKIIα promoter (the same mutation and promoter as for the rTg4510 mice but with much slower progression of pathology). During this study we established that the TauD35 line in fact consisted of two lines of these mice with different copy numbers (referred to here as LowTAU and HighTAU) enabling us to assess the dependency of changes seen on the gene dosage, albeit post hoc.

Methods
Methods are the same as for [11] except adapted for neurofibrillary tangles.

Animals
All experiments were performed in agreement with the Animals (Scientific Procedures) Act 1986, with local ethical approval and in agreement with the GlaxoSmithKline statement on use of animals.
Transgenic TauD35 male mice were generated by GlaxoSmithKline (Harlow, UK) on the background mouse line C57Bl/6J (Charles River, UK) via pronuclear injection and bred and maintained at Charles River Laboratories. The mice harbour human cDNA for the 0N4R isoform of MAPT carrying the P301L mutation under the alpha isoform of the Ca2+/calmodulin dependent protein kinase II (CaMKII) promoter. Age-matched wild type littermates were used as controls.
Mice from Charles River were shipped to UCL at 3-months-old. Mice were kept in large cages (20 x 35 x 45 cm) with an enriched environment. Thus, cages containing 2-8 male mice (TauD35 and littermate wild type controls) were maintained in a 12-hour light/12-hour dark cycle with food and water ad libitum. Environmental enrichment consisted of changes of food location, bedding type (e.g. tissue, shredded paper, paper roll, paper bags) and inanimate objects (e.g. running wheels, rodent balls, tubing, houses (mostly purchased from Eli Lilly Holdings Limited, Basingstoke, UK)) within the cage at least once per week. Mice were used for experimentation at the ages stated (± 0.5 months) and, where unavoidable, were single-housed for no longer than 24 hours. Tails or ear punches were genotyped using standard PCR protocols.

Genotype confirmation using conventional PCR methods
Briefly, genomic DNA was extracted using the 'HotSHOT' lysis method. Alkaline lysis reagent (25 mM NaOH, 0.2 mM EDTA, pH12) was added to tissue samples prior to heating to 95°C for 30 minutes. The sample was then cooled to 4°C before the addition of neutralisation buffer (40 mM Tris-HCl, pH 5). The PCR reaction was performed through addition of MyTaq DNA Polymerase (Bioline) reaction buffer and primer pair: using the cycling parameters: 94°C (2 minutes), 58°C (30 s), 72°C (30 s), for 30 cycles and a final extension at 72°C for 4 minutes. PCR product size 130 bp.

Transgene copy number confirmation
Genomic DNA was extracted using the 'HotSHOT' lysis method described above. Unique TaqMan primer and probe sequences were generated using Applied Biosystems custom design sequence using the published sequence for the CaMKIIα promoter sequence (Accession# AJ222796). The qPCR assay was run with 4 μl of genomic DNA per well using the CaMKIIP_CCVI3TR primer and probe set (catalogue number 4400294, Applied Biosystems), using the Taqman genotyping master mix with each sample run in triplicate in parallel with the Taqman copy number reference assay for mouse Tfrc (catalogue number 4458370, Applied Biosystems), using the manufacturer's instructions.

Brain tissue extraction
Brains were removed from the skull on ice and hippocampus extracted within 5 minutes of death, then snap frozen on dry ice and stored at -80°C until protein extraction was performed. Sarkosyl-insoluble tau extracts were obtained as described previously [3] 90µl of homogenate was ultracentrifuged at 150,000g for 15 minutes at 4°C. The supernatant (S1) was removed for storage and the pellet re-homogenised in 4

Nissl Stain
Sections were mounted and left to dry overnight prior to staining. Sections were dipped in H20 and then submerged in 1% w/v cresyl violet (Alfa Aesar, Massachusetts) for 2 minutes and blotted to remove excess solution. Sections were de-stained in 70% ethanol containing 1% glacial acetic acid for <1 minute and then dried by submersion in 100% ethanol for 2 minutes followed by submersion in xylene (3 minutes) to completely remove water. Sections were then cover-slipped using DPX mounting medium.

Electrophysiological recordings
Acute hippocampal brain slice preparation Mice were decapitated and the brain rapidly removed and placed in ice-cold dissection

Patch-clamp recordings in brain slices
Once transferred to 1 Mg 2+ , 2 Ca 2+ ACSF, slices were allowed to return to room temperature and after at least a further 40 minutes recovery time, a single slice was transferred to a submerged chamber and superfused with recording ACSF (containing WinEDR synaptic analysis software was used for detection of spontaneous and miniature currents and WinWCP used to analyse identified spontaneous, miniature and evoked currents. Criteria for detection of spontaneous or miniature currents was to remain over a threshold of 3 pA for 2 ms. Currents were inspected by eye and only included if the rise time was <3 ms and faster than the decay.

Field potential recordings in brain slices
Slices were transferred as needed to a heated (30±1°C) submerged chamber and superfused with ACSF and allowed to recover for 1 h in the recording chamber. A glass stimulating electrode (filled with ACSF, resistance ~2 MΩ) and an identical recording electrode (connected to an AxoClamp 1B via a 1X gain headstage) were both positioned in stratum radiatum of the CA1 field to obtain a dendritic excitatory postsynaptic field potential (fEPSP). Recordings were controlled and recorded using WinWCP software (as above), filtered at 10 kHz and subsequently at 3 kHz and digitized at 10 kHz via a

T-maze forced alternation task
Previously reported methods, optimised for mouse, were used to assess hippocampusdependent learning [19] . Mice were food deprived to 90% free-feeding-weight, beginning 2 days before the start of the habituation phase and with ad libitum access to water.
Each mouse was handled at the start of food deprivation and throughout T-maze habituation for 15-20 minutes per weekday.
The T-maze was constructed from three arms, each measuring 50

Light/dark box
Analysis in the light/dark box was based on methods described by Packard et al [20] . The were recorded. An entry into a box was defined as all four paws resting inside the given box.

Statistics
All data analysis was carried out blind to genotype. All statistics were performed using Graphpad Prism 6 with appropriately designed two-tailed t-test or ANOVA. Post hoc tests were only performed if a significant interaction between the independent variables was obtained. Animals were considered as independent samples and, where multiple data were collected from an animal, these were averaged (mean) prior to pooling. Thus sample sizes represent the number of animals. Unless stated otherwise, data are presented as mean ± SEM and differences considered significant at p<0.05.

Identification of two separate TauD35 transgenic lines determined using TaqMan qPCR
Tau D35 express human Tau with the P301L mutation that causes hyperphosphylation of Tau [21] and was initially developed at GlaxoSmithKline as a single mouse line.
However it became clear, as the genotypes were progressively uncoded in this blind study, that in some but not all types of experiments, the results from the transgenic mice fell clearly into two groups. In particular, a subset of mice showed a considerably higher density of neurofibrillary tangles at around 13 months of age than others. Moreover, in mice aged further, a subset of Tau mice showed a severely neurodegenerative phenotype, rapidly developing over about 2 weeks at 16-18 months of age, consisting of a hunched posture, piloerection and akinesia. This was so severe that it required these mice to be euthanised at this age, while others remained phenotypically normal at 24 months. We thus retrospectively determined the transgene copy number from all of the TauD35 mice included in the study.
In order to ascertain the number of transgene copies harboured by TauD35 animals, qPCR techniques were employed. Due to the similarity of the human and murine Tau sequence, transgene copy number was determined solely through the development of a probe to the CaMKIIa promoter. Assuming an endogenous murine CaMKIIa copy number of two, results gathered from transgenic animals were analysed relative to wild type copy numbers. Two clear groups within the transgenic line were apparent, one with approximately twice the copy number of the other (in the order of 10 and 5 copies respectively). Unfortunately, by the time the reason for this variability was uncovered, the HighTAU copy number had largely been bred out of the colony preventing us from increasing sample sizes in this group. The finding of this variability in copy number has the disadvantage of some experiments featuring only few animals with high copy number but the advantage is that we can at least preliminarily judge whether any effects of the TauP301L transgene are dose-dependent. Consequently, all data were pooled but with the individual copy number groups indicated in overlying scatter graphs. If strong trends or significant differences were seen between these lines then they were also analysed separately, where possible.
The two lines are referred to here as LowTAU and HighTAU, respectively.

Relative protein levels are also different in the two Tau lines, especially at younger ages
Protein levels and phosphorylation states of Tau were measured by Western blot using an array of antibodies (Fig 1). This further confirmed that the two Tau lines expressed different levels of protein in relation to their transgene copy number. Firstly, using the human-Tau-specific antibody HT-7, we assessed levels of the TauP301L protein. At 4 months of age, levels of human Tau were higher in HighTAU mice than in LowTAU mice of the same age (p=0.01; Fig 1A). This difference was less marked at 13 months but was highly significant (p<0.005). As expected, there was no immunopositive bands for human Tau observed in samples from wild type mice. The less pronounced difference at the older age is presumably because of increased Tau aggregation in the HighTAU mice which would be removed during initial centrifugation for Western blot [22] . We next used the pan-Tau antibody, DA9, which, in brain tissue from wild type animals at high exposure, revealed bands at ~50kDa, ~58kDa and ~64kDa, corresponding to the 3 endogenous monomeric isoforms of mouse Tau expressed in adult mice [23][24][25] , with small increases in molecular weight due to endogenous phosphorylation, when compared to de-phosphorylated/recombinant Tau [26] . The bands at the higher molecular weights are clearer when the HT-7 antibody is used to label the Sarkosyl insoluble fraction which represents the Tau tangles ( Fig 1E). Moreover, although we cannot accurately quantify the protein levels in this fraction because the usual housekeeping genes are not present, we attempted to standardise the loading of protein between lanes to allow at least a qualitative comparison within ages and total protein assessed using amido black. The relatively strong signal of Tau in ΗighTAU mice compared to LowTAU mice supports the hypothesis that the apparently low level of Tau seen in the blots in Figs 1A-D is due to the increased deposition resulting in Tau being lost in the first stage of homogenisation which is the source of the P3 pellet from which the sarkosyl insoluble fraction (Fig 1E) is prepared.
Finally, we assessed the phosphorylation at early (S202) and late (S396/S404) sites on the Tau protein using CP13 and PHF-1 antibodies, respectively [13] . A similar pattern of age-dependent Tau phosphorylation was seen with both CP13 and PHF-1, with a low level of phosphorylation present in wild type mice at all ages observed with both antibodies. Transgenic mice display a high level of both S202 (CP13) and S396/S404 (PHF-1) phosphorylated Tau when compared to wild type mice at all ages ( Fig 1C&D). Interestingly, while at 4 months of age transgenic mice present phosphorylated Tau species between 50 and 60 kDa with little if any of the higher molecular weight component detectable, the larger species at around 64 kDa become clear in the 13-months old HighTAU mice but not until 24-months in the LowTAU mice, suggesting a dose and age dependent increase in this hyperphosphorylated component.

Neurodegeneration and neurofibrillary tangle development
There were notable differences in brain size of HighTAU mice at 18 months compared to either age-matched wild type or LowTAU mice (Fig 2). This was indicated by a reduced hippocampal area, thinning of the cortex and enlargement of the dorsal ventricle.
Furthermore, cresyl violet staining revealed that there was overt neuronal loss in CA1, indicated by lower cell counts at both 13 months (~80% of wild type, p=0.07) and 18 months of age (~40% of wild type, p<0.01; 2-way ANOVA interaction between age and genotype p=0.02; Fig 2). Similarly, there were fewer CA3 neurones in HighTAU animals, reflected by a significant main effect of genotype by 2-way ANOVA (p<0.01). In this case there was no interaction between age and genotype. We have concentrated subsequent analysis on the CA1 region but similar differences are seen throughout the hippocampus and cortex.
Immunohistochemistry was also carried out to assess the occurrence of neurofibrillary tangles and the regional specificity of different stages of Tau phosphorylation. Similar antibodies were employed to the Western blot experiments above, after testing specificity of antibodies in this context. Antibodies needed to be specific for Tau, recognise common pathological epitopes found in both human patients as well as other mouse models and also cover a wide range of pathological species (see methods and for review [13] ). To this end, a range of both commercial antibodies and non-commercial antibodies (kindly provided by Peter Davies, Albert Einstein College of Medicine, USA) were chosen, including: HT7 (human Tau), as well as AT8, CP13 (both markers for early Tau phosphorylation [14,15,27] ) and MC1 (misfolded human Tau [16,17] ) and PHF-1 (later Tau phosphorylation [15] ). A distinct pattern of immunofluorescence was observed in both CA3 and CA1 pyramidal neurones of the hippocampus, granule cells of the dentate gyrus, cells within the hilus and also the surrounding cortex (Fig   3Ai). Mossy fibre axonal projections from dentate gyrus to CA3 were particularly prominent (Fig 3Ai and Aii). The tangle-like structure of the Tau immunofluorescence were detected within the soma of neurones using confocal imaging (Fig 3Aiii).
Interestingly, once separated, it was clear that the 2.5-fold difference in Tau protein levels seen at 4 months between the HighTAU and LowTAU lines translated into considerable differences in the occurrence of neurofibrillary tangles by 13 months of age ( Fig 3B&C). The density of tangle-like structures labelled with antibodies to all the proteins listed above by 13 months was around 100-fold higher in HighTAU than in LowTAU mice in the CA1 region (Fig3Bi). This is consistent with the effects seen above in the Western blot analysis, particularly in the Sarkosyl insoluble fraction. By 24 months, by which time neurodegeneration has resulted in the culling of all the HighTAU mice, the LowTAU mice eventually reach the high density of tangle-like staining that is seen at 13 months in the HighTAU mice (Fig 3Bi). The increase in the number of tangles remained evident when normalised to the cell density (Fig 3Bii).

Microglia in Tau mice
We have previously reported that the same Tau mice used in the present study showed no difference in expression of Iba1 (Aif1) or a range of other microglial genes at young ages, even at 8 months as neurofibrillary tangles start to appear in HighTAU mice [1] .
This suggested that there was little or no proliferation or activation of microglia up to this age. Here we examine distribution and phenotype of microglia in Tau mice in more detail using immunohistochemistry (Fig 4). Firstly, at 4 months of age, we confirm that there is no proliferation or activation of microglia, as measured by counting Iba1 and CD68 positive microglia, respectively, in the CA1 region in fixed sections of hippocampus in either LowTAU or HighTAU mice. By 13 months, while the wild type mice and lowTAU mice show similar age-related increases in both proliferation and activation (see also wild type mice in [11] ), HighTau mice show much greater changes. At this age HighTAU mice show a heavy load of neurofibrillary tangles and the start of neurodegeneration and there is a significant proliferation of microglia, (two-way ANOVA genotype x region interaction p<0.0001). This was especially strong in the stratum lacunosum moleculare (SLM, 3 fold change compared to WT (p<0.0001) but also with significant increases in all other layers (SO ~2-fold increase, p<0.001; SP, SR ~65% increase p<0.05). Moreover, activation in HighTAU mice is also increased (two-way ANOVA genotype x region interaction p<0.005) and again the increase in SLM was particularly strong (~7 fold p<0.0001) and SO also showing a significant increase (2.7 fold, p<0.01). However, this result is difficult to interpret as the proportion of activated microglia in WT mice was very variable between individual mice, possibly relating to this age being the time when increases in activation and proliferation are occurring in all genotypes [11] . Moreover, the number of HighTAU mice and the number of activated microglia per section is very low, leading to potential sampling artefacts so that only large changes could be reliably detected. Similar results were obtained in other hippocampal regions.
Having established these initial changes, we proceeded to study the latest stages of pathology available in these mice. LowTAU mice were aged to 24 months but HighTAU mice could not be aged this far because the strong neurodegenerative phenotype that occurs at 17-18 months is considered end stage. As outlined above, at this stage there is considerable neurodegeneration. We thus compared these two groups of mice in detail, also referring back to the earlier stages described above.
The 24-month-old LowTAU mice have a similar density of tangles to the 13month-old HighTAU mice (Fig 3) but, at this stage, they have no measurable neurodegeneration (Fig 2). Proliferation of microglia was evident at around 2-fold in the 24-month-old LowTAU mice compared to WT mice but only in the stratum lacunosum-moleculare and not in other CA1 regions (Fig 4F). Moreover activation (measured by CD68 positive microglia), was also increased by almost 10-fold in the SLM.
Hence the SLM showed a very similar microglial response at this stage to the HighTAU mice at the equivalent stage. Interestingly there was a strong trend to show increased proliferation also in the other CA1 regions despite lack of proliferation at this stage in the 24 month old LowTAU mice suggesting that activation may precede proliferation in these mice.

Synaptic transmission in TauD35 mice
To study the early effects of the Tau Fig 5Aii).
While there was no difference in miniature EPSC amplitudes between genotypes (Fig   5Aiii), the decay time constant was significantly different (one-way ANOVA, p<0.05) and a Tukey post hoc test revealed a significant difference between LowTAU and HighTAU mice (p<0.05 ; Fig 5Aiv, possibly reflecting a change in receptor subtype and/or cell geometry). Hence, while action potential-mediated activity and the resulting release of glutamate appeared to be unchanged, miniature synaptic activity was increased in the younger age group in the HighTAU mice compared to LowTAU or wild type mice.
We next examined potential changes in probability of glutamate release by applying pairs of stimuli to analyse paired pulse ratios (amplitude of second response over first) from Schaffer collateral synapses onto CA1 pyramidal neurones (Fig 5Bi). As expected, synapses from wild type animals displayed paired-pulse facilitation (indicating a low probability of glutamate release), that decreased towards no change or paired-pulse depression at longer intervals. At 4 months of age there was no significant difference between transgenic mice and wild type mice regardless of copy number. At 13 months of age, post hoc genotyping revealed only LowTAU mice in this group. There were no differences in paired-pulse ratio at intervals between 25 and 800 ms, while at this age, the transgenic mice, despite having a LowTAU copy number, showed a significant depression at 1500ms compared to the wild type mice (p<0.001; Fig 5Biii). As expected wild type mice showed little if any interaction between the first and second response at this long interval. This suggests an increase in an inhibitory metabotropic autoreceptor or heterosynaptic connection in the transgenic animal.

Magnitude of long-term potentiation is normal in Tau mice but locus of expression changes with age
Longer term synaptic plasticity was studied using extracellular field potentials recorded from stratum radiatum of CA1 in response to stimulation applied to the Schaffer collaterals (axons from CA3 neurones) in slices prepared from mice aged 4 to 24 months. Initially, input-output relationships were determined by applying increasing voltage pulses (Fig 6A). There were no significant differences in the field EPSP slopes recorded between the genotypes at any age. Similar to the patch clamp recordings, paired-pulse ratios did not differ between the genotypes (Fig 6B). Moreover, the magnitude of LTP (mean of responses recorded over the last 10 minutes compared to baseline) induced by a moderate tetanic stimulus did not differ between the genotypes at any age ( Fig 6C&D). However, paired-pulse ratios recorded during the time course of the LTP experiment did differ. In slices from both wild type and Tau mice, the pairedpulse ratio measured at different times after induction of LTP compared to baseline showed the expected transient decrease during post tetanic potentiation (no significant difference between genotypes at any age, data not shown). While paired-pulse ratios in wild type slices returned to baseline as expected, those from Tau slices tended to remain lower than baseline (Fig 6C&E, two-way ANOVA, main effect of genotype, p<0.01). There was no significant effect of age nor an interaction between age and genotype.. This suggested that the locus of LTP induction may have a presynaptic component in the transgenic mice. As the magnitude of LTP was unchanged this may suggest a compensation between pre-and postsynaptic changes.

Mutant Tau mice do not show cognitive changes in hippocampus-dependent learning at 12 months of age
The hippocampus-dependent forced-alternation T-maze task was used to assess cognition in 12-month-old TauD35 mice (Fig 7A; mice used subsequently for electrophysiology or immunohistochemistry). As outlined in detail in the methods, the test consisted of a sample run followed immediately by a choice run. For the sample run one arm of the T was blocked off so that the mouse was forced to turn in one direction where they received a food reward. This was then followed by a choice run in which the both arms were open, giving them a free choice of turning left or right; the correct arm (where a food reward was then available) was the opposite arm from the sample run.
Two mice, determined post hoc to be HighTAU mice, are reported separately from the LowTAU mice (n=9).
TauD35 mice started training at similar levels to wild type mice, both improving to ~90% over the training days (Fig 7Ai). Transgene copy number did not show any obvious differences to the outcome of the task during the training period. When the mice were challenged with delays between sample and choice runs there were no significant differences between genotypes, although surprisingly the two HighTAU mice tended to improve their performance with increased delays while both wild type and LowTAU mice showed a decrement, as expected (Fig 7Aii). However, a clear difference between the HighTAU mice and both the wild type and LowTAU mice was the rate of response in the first (sample) run of each trial (Fig 7Aiii). The wild type and LowTAU mice explored the maze, taking 5 to 20 s from being placed in the maze until all 4 feet were within the destination arm. In contrast, both HighTAU mice ran rapidly through the maze with response times starting at about 5 s but reducing to 2 s from the third block of the training period onwards. Although the sample size is small, this was consistent for both HighTAU mice in all the subsequent trials, including the delay trials, suggesting a lack of interaction with the novel environment or a behavioural disinhibition, similar to symptoms in a subgroup of frontotemporal dementia patients [28] . In the choice trial there was no difference in response time between genotypes with all groups responding within about 5 seconds (Fig 7Aiv).
Unfortunately, the number of HighTAU mice was so low that no firm conclusions can be made about this potentially interesting behavioural difference but the consistency between the two mice flags an interesting point to be investigated in future studies. However, it is clear that, despite a substantial load of neurofibrillary tangles and, in the case of the HighTAU mice, a decrease in cell number, neither the HighTAU nor LowTAU mice appear to have any learning deficit in the forced T-maze.
We proceeded to test whether these possible behavioural changes reflected a change in locomotor activity or anxiety-like behaviours in the HighTAU mice (Fig 7B&C).
In an open field arena, while there were no differences in the total path lengths between genotypes (Fig 7Bi), the pattern of activity was again very different for HighTAU mice.
Both wild types and LowTAU mice decreased their total exploratory activity over time, while HighTAU mice maintained a constant level of exploration throughout the test period (Fig 7Bii&iii). Furthermore, the time they spent in the centre of the arena did not increase significantly over time (Fig 7Biv). Although one mouse came close to wild type levels in the final 5-minute block of the test period, the other remained largely on the edge of the arena through all trials, a behaviour that would generally be associated with anxiety.
No significant differences were seen in the light dark box (Fig 7C) when genotypes were considered as independent groups and tested by ANOVA although there was a strong trend for the Tau mice to take longer to enter the dark box (one-way ANOVA, p=0.06). Consistent with this a tendency was also observed for the transgenic mice to stay in the light for longer than the wild type mice and this was dose dependent showing a significant correlation to Tau copy number (p=0.05). In contrast to the open field result this would suggest a decrease in anxiety.
Together, these data indicate that, despite a heavy neurofibrillary tangle load and, in the case of HighTAU mice, initial neurodegeneration, there is no learning deficit measurable with the T-maze. Possible behavioural changes in habituation and anxiety particularly in the HighTAU mice would need further investigation but again would be compatible with a lack of exploratory interaction with the environment rather than specific effects of anxiety.

Discussion
TauD35 mice express human Tau harbouring the P301L mutation under the CaMKIIa promoter, which directs expression to glutamatergic cells in the forebrain (avoiding expression in the spinal cord or peripheral nervous system). Moreover, the distribution of CaMKIIa in the brain is similar to Tau making this a very suitable promoter [29] .
The P301L mutation causes frontotemporal dementia with parkinsonism linked to chromosome 17 in humans but here we were also interested in considering the role of neurofibrillary tangles in all forms of dementia, including Alzheimer's disease.
Understanding the interactions of Aβ and Tau in the development of Alzheimer's disease is complicated by the lack of animal models that display neurofibrillary tangles and neurodegeneration dependent on rising Aβ, rather than due to mutations in Tau. While expressing mutated human Tau is useful for understanding the link between neurofibrillary tangles and neurodegeneration, it bypasses the proposed link between the Aβ and Tau proteins [30] . However, this is the best approach available for studying the effects of neurofibrillary tangles on the various functional aspects of disease progression. Previous studies have described a range of mice with Tau mutations but, in general, these models have progressed rapidly, reaching neurodegeneration at relatively young ages (see Alzheimer's Disease Research Models | Alzforum. Retrieved 14 June 2018 from https://www.alzforum.org/research-models/alzheimers-disease). In fact, a popular model, rTg4510, harbours the same mutation controlled by the same promoter but including a tetracyclineoperon-responsive element allowing suppression of the transgene. This is a rapid model that has increasing tangle load from 4 months of age and substantial neurodegeneration from about 5 months onwards [3] . Moreover, spatial memory deficits are seen even earlier from around 2 months of age in these mice [3,31] .
However, in humans, particularly in Alzheimer's disease, the cognitive effects of Tau pathology and neurodegeneration are usually seen in relatively old age. Hence, the more slowly progressing model used here allows the influence of age to be considered versus the rate of pathology development. This is particularly the case when comparison of the LowTAU mice at 24 months to the HighTAU mice at 13 months at which ages their neurofibrillary tangle development is similar. What is perhaps most noticeable in these results is that, even with a heavy neurofibrillary tangle load and modest neurodegeneration seen at 13 months of age in HighTAU mice, there is little if any learning deficit, although some behavioural changes may be evident in these slowly developing models. This is similar to the human condition, in which the diagnosis of the disease only occurs after considerable brain tissue is lost [32] . However, it is interesting to note that the behavioural changes seen, seem to be a lack of adaptation and interaction with the environment which may be relevant to cognitive effects in the more complex tasks of the human experience.

Tau overexpression, phosphorylation and pathology
Phosphorylated tau is detected in both LowTAU and HighTAU animals at 4 months of age and some of human Tau is insoluble in sarkosyl. We have previously shown in TauD35 mice that rare neurofibrillary tangles could be detected in 1 out of 4 mice at 4 months and an increasing tangle load in all mice at 8 months [1] .  (Fig 1).
The relatively small increase in mutated human tau protein expressed in HighTAU mice (~2.5 fold, as assessed at 4 months, Fig 1A, before substantial development of tangles) compared to their LowTAU counterparts was seen to result in a much greater difference in the timing and extent of tau pathology development. This was particularly evident at 13 months of age, when LowTAU mice were seen to display only neurofibrillary tangles within the hippocampus, compared to the extensive spread observed in HighTAU animals. At this age the ratio of neurofibrillary tangles within the CA1 region of the hippocampus in HighTAU was ~60 times the ratio of neurofibrillary tangles calculated in LowTAU animals in the same region (Fig 2Bii). This relationship between protein level and neurofibrillary tangle development has been observed previously [3] , suggesting that not only the mutation but the overall concentration of protein may be important in aggregation initiation. It is also an important observation in terms of the relevance of different overexpression models, including normal Tau protein [25,33] , in which the concentration of Tau may be a pathological factor in itself.

Pathology and microglial proliferation
We have previously shown in mice with mutations in the Αβ pathway that the expression of microglial genes is closely related to plaque pathology, while this relationship is much weaker when compared to neurofibrillary tangle load in Tau mice [1] . Here we have used immunohistochemistry to study the phosphorylation of Tau and eventual misfolding into neurofibrillary tangles in much more detail and compared this to the proliferation and activation of microglia. In the TauD35 mice, although the detection of Tau pathology is low but clear by 8 months of age in HighTAU mice, our previous gene expression study shows no change in expression of either Iba1 or CD68 at this age [1] . Here we show with immunochemistry that also as tangles first appear in the 13-month-old LowTAU mice, that microglia are apparently unchanged but that, with the greater tangle load at this age in the HighTAU mice, proliferation is the main effect. CD68 is clearly expressed throughout the hippocampal layers once the neurodegenerative phenotype is fully developed in the HighTAU mice at 18 months of age. Increased CD68 expression is suggested to relate to an increasingly phagocytotic phenotype and appears to coincide with cell loss in these mice. This widespread microglial activation may be specifically related to the removal of damaged neurones. In contrast, at the oldest age of LowTAU mice, where no neurodegeneration is measurable, activation of microglia is also present and is particularly strong in the stratum lacunosum moleculare, although there is a tendency to activation in other regions. The SLM is the synaptic zone for the entorhinal cortex inputs via the temporoammonic pathway. Initially, as synapses are altered and particular axons become dysfunctional, removal of such dysfunctional axons could be protective to the ongoing network activity. This is, however, a one-way process and eventually the network would not continue effectively once substantial cell loss has occurred. Interestingly, although very delayed compared to rTg4510 mice, the neurodegenerative phenotype once it begins in HighTAU mice, is rapid. This rapidly developing neurodegenerative morbid phenotype makes cognitive testing impossible. It is interesting to note however that the lowTAU mice develop a substantial tangle load by 24 months (similar to 13 months highTAU mice) but feature a much more robust activation of the microglia at this stage but no detectable neurodegeneration. This may suggest that the microglia can protect against neurodegeneration more effectively if the Tau pathology develops more slowly.

Synaptic transmission and plasticity
In mice with rising Αβ, we have previously reported that at 2 months in the CA1 region, even before the first plaque deposition, there is a complete loss of spontaneous action potential mediated release [34] . Moreover, in evoked transmission in the Schaffer collateral pathway, an increase in glutamate release probability was observed in the Αβ mice compared to wild type mice at these early stages. These changes persisted throughout pathology development [11] . In contrast, the TauD35 [32,35] reported that people with familial genes for Alzheimer's disease have already lost 20% of the hippocampus before they start to show symptoms, confirming that the brain is extremely good at compensating for change.
While both Tau and Αβ mice performed at similar levels as their wild type littermates in the T-maze, there were differences in motor activity and anxiety related behaviours both compared to wild type mice and comparing the different pathologies.
This was reflected in both the T-maze, where HighTAU mice completed the sample run in under a quarter of the time of either wild types or LowTAU mice, and also the open field arena, where HighTAU mice failed to habituate to the arena but continued being highly active at a time when wild type mice had decreased their activity. In contrast, the Αβ mice [11] were slower to perform the T-maze task or failed to do so altogether and were less active in the open field than the wild type mice. Surprisingly, when plaque load was high at 12 months, they showed a tendency to retain memory for longer than the wild type mice.
When tests of anxiety were performed the Tau and Αβ mice initially both appeared to be more anxious, staying on the periphery of the open field. However, while this interpretation was confirmed in further tests on the Αβ mice, the interpretation of the behaviour of the Tau mice was less clear. In the light/dark box anxiety would be indicated by staying in the dark which was the result for the Αβ mice but the Tau mice stayed in the light more than wild type mice and moreover this seemed to be dose dependent in relation to copy number. In combination these behaviours suggest that there is certainly a behavioural change in the Tau mice which may indicate a lack of interaction with their environment.
The lack of memory deficits and overall subtlety of the behavioural changes reported here for both the Tau and Αβ mice, even at stages of advanced pathology but without gross neurodegeneration, is consistent with previous findings for the present Αβ model [36][37][38][39] and in other Αβ models [for reviews, see 40,41] . Deficits are seen early in the rTg4510 mouse as neurodegeneration begins [2,3,31] . The results in these more slowly developing models is however more consistent with the human condition, where, by the time sufficient deficit is evident for diagnosis, there is already a substantial reduction in hippocampal volume of up to 20% [32,35] . It is notable that in the rTg4510 mice the neurodegeneration occurs over several months (for example between 5 and 16 months [2] ) with behavioural testing up to 12 months [31] ), although this varies considerably between different studies [2,3,31] . In the much older TauD35 model reported here the initial tendency to neurodegeneration at 13 months in HighTAU mice does not result in any measurable cognitive deficit, whereas substantial neurodegeneration sets in relatively rapidly with a strong morbid phenotype developing over about 2 weeks at ages between 16.5 and 19 months preventing behavioural testing.
Given that Αβ mouse models are not a complete model of AD, i.e. lacking neurofibrillary tangle formation and neurodegeneration, which more faithfully correlate with cognitive decline, they should be considered as models of the preclinical disease, when Αβ is first deposited. The Tau models may make better models of end-stages of disease, when neurodegeneration and neurofibrillary tangle predominate the pathology.
However, unless bypassed by mutations in Tau, to date, no animal model reliably offers the crucial link between rising Aβ and neurofibrillary tangles that are required to understand the full progression of the disease. It is interesting to note how well the mammalian brain can compensate and, particularly in the Tau mice how little apparent phenotype is present in terms of cognition even once neurodegeneration is under way.

Conclusion:
Given that Αβ mouse models are not a complete model of AD, i.e. lacking neurofibrillary tangle formation and neurodegeneration, which more faithfully correlate with cognitive decline, they should be considered as models of the preclinical disease, when Αβ is first deposited. The Tau models may make better models of end-stages of disease, when neurodegeneration and neurofibrillary tangle predominate the pathology. However, unless bypassed by mutations in Tau, to date, no animal model reliably offers the crucial link between rising Aβ and neurofibrillary tangles that are required to understand the full progression of the disease. It is interesting to note how well the mammalian brain can compensate and, particularly in the Tau mice, how little apparent phenotype is present in terms of cognition even once neurodegeneration is under way. The comparison of Aβ and Tau models brings up some interesting contrasts, suggesting that the levels of soluble Aβ in the general neuropil, independent of the position or even presence of plaques, has substantial effects, particularly on neurotransmitter release from glutamatergic synaptic transmission. The presence of plaques does not greatly change this, although it causes a strong immune response, possibly related to the removal of dystrophic neurites. Such dystrophic neurites only occur locally in and around the plaque, with synaptic loss having been reported to decrease with distance from the plaques [42,43] . The percentage area covered by plaques is small, even when the plaque load is heavy and hence, relative to the total number of synapses, only relatively few will be directly lost due to proximity to plaques. In contrast, although the development of Tau tangles is associated with neurodegeneration, generalised effects on synaptic responses in the wider network are more subtle until a period of rapid deterioration occurs as neurodegeneration reaches a critical stage. We suggest that this may relate to the Tau pathology causing loss of whole axons which eventually will cause network disruption in contrast to the localised effects of plaques but that in the human disease, the localised synaptic damage caused by plaques may trigger the phosphorylation and eventual dissociation of Tau from the microtubules.