Longitudinal characterization of behavioral, morphological and transcriptomic changes in a tauopathy mouse model

Hyperphosphorylated tau is present since the early age in P301S mice

Hyperphosphorylation and aggregation of tau are associated with neurodegenerative disorders, and the “pathological” tau phosphorylation sites, such as Ser214, Ser262, Ser202/Thr205 and Ser396/Ser404, are highly relevant to microtubule detachment, cytoskeleton degradation and synaptic failure [11–13]. Phosphorylation at Ser202/Thr205 has been used to classify the staging of AD-related neurofibrillary pathology known as Braak stages [31]. To examine how early tau hyperphosphorylation appears in a tauopathy model expressing P301S mutant tau [16], we measured the level of ^S202/T205p-tau, ^S214p-tau and total tau in P301S mice from 1 to 6 months in both PFC and hippocampus regions. Compared to WT, levels of ^S202/T205p-tau, ^S214p-tau and total tau were significantly higher in both PFC (Figure 1A, 1B) and hippocampus (Figure 1C, 1D) of P301S mice. This increase was seen from 1 month in both regions and remained consistently increased through 6 months.

Figure 1. High expression of hyperphosphorylated tau and total tau was found in P301S Tau mice from early ages. (A, C) Representative Western blots showing ^S202/T205p-tau, ^S214p-tau and total tau in PFC (A) and Hippocampus (C) of WT vs. P301S transgenic mice (WT: n = 3, P301S: n = 4) at 1 month to 6 months of age. (B, D) Plots of quantification data of ^S202/T205p-tau, ^S214p-tau and total tau (normalized to GAPDH) in PFC (B) and Hippocampus (D) of WT vs. P301S mice at various ages. At each time point, ^*p < 0.05, ^**p < 0.01, and ^***p < 0.001, t-test.

P301S mice exhibit cognitive and social deficits in an age-dependent manner

Cognitive decline is one of the most prominent symptoms of AD. Next, we performed behavioral assays to characterize the longitudinal changes in cognitive processes in P301S mice. One of the cognitive behavioral tests is Barnes maze (BM), a spatial memory task wherein the mouse’s capability of recalling the location of an escape hole out of eight holes with spatial cues is tested. Another cognitive behavioral assay is Novel Object Recognition test (NORT), a short-term memory task wherein the mouse’s capability of distinguishing a novel object from a familiar object is examined. Two-way ANOVA was used to analyze the difference in performance between WT and P301S mice of different ages (factors: Genotype x Age).

In BM tests (Figure 2A), P301S mice showed a progressive memory impairment (F_{3,113 (age)} = 3.45, p < 0.05; F_{1,113 (genotype)} = 38.76, p < 0.0001, two-way ANOVA). The spatial memory index (calculated as Time on correct hole (T1)/Time on incorrect holes (T2)) was similar at 1–2 months, but started to show significant deficits in P301S mice at 3–4 months (WT: n = 15, P301S: n = 18, p < 0.01, two-way ANOVA), persisted through 5–6 months (WT: n = 17, P301S: n = 16, p < 0.001, two-way ANOVA), and 9 months old (WT: n = 11, P301S: n = 10, p = 0.06, two-way ANOVA). In NORT tests (Figure 2B), a similar memory decline was found (F_{3,105 (age)} = 6.84, p < 0.001; F_{1,105 (genotype)} = 8.36, p < 0.01, two-way ANOVA). The discrimination index (calculated as (T_novel − T_familiar)/(T_novel + T_familiar)) showed deficits in P301S mice at 3–4 months and reached significance at 5–6 months of age (WT: n = 17, P301S: n = 9, p < 0.05, two-way ANOVA).

Figure 2. Behavioral tests showed cognitive and social deficits in P301S Tau mice at early time points. (A) Plots of spatial memory index in Barnes Maze tests of WT vs. P301S transgenic mice at different ages (n = WT/P301S: 1–2 m, 14/21; 3–4 m, 15/18; 5–6 m, 15/16, 9 m, 11/10). (B) Plots of discrimination index in Novel Object Recognition tests of WT vs. P301S transgenic mice at different ages (n = WT/P301S: 1–2 m, 14/21; 3–4 m, 15/17; 5–6 m, 17/9, 9 m, 11/9). (C) Plots of social preference index in 3-chamber sociability tests of WT vs. P301S transgenic mice at different ages (n = WT/Tau: 1 m, 8/10; 2 m, 13/17; 3 m, 9/13; 4 m, 7/6; 5–6 m, 9/9). (D) Plots of social novelty index in social cognition tests of WT vs. P301S transgenic mice at different ages (n = WT/Tau: 1 m, 8/10; 2 m, 14/18; 3 m, 9/13; 4 m, 7/6; 5–6 m, 9/9). ^*p < 0.05, ^**p < 0.01, and ^***p < 0.001, two-way ANOVA.

Besides cognitive deficiency, another behavioral alteration of AD is social impairment [32]. Next, we characterized the longitudinal changes in sociability using 3-chamber social preference tests [33] in which the animal’s preference to interact with a social stimulus (an age and gender-matched mouse) vs. a non-social object was tested. Another assay for social cognition is the social novelty test [34], in which the animal’s capability of distinguishing a new social stimulus from an old social stimulus is examined.

In social preference tests (Figure 2C), P301S mice displayed a progressive loss of preference towards the social stimulus against the non-social (ns) stimulus (F_{4,91 (interaction)} = 2.47, p = 0.05; F_{1,91 (genotype)} = 4.94, p < 0.05, two-way ANOVA). Social preference index (calculated as (T_social-T_ns)/(T_social + T_ns)) was lower in P301S mice than WT mice starting at 4 months and reached statistical significance at 5–6 months (WT: n = 9, P301S: n = 9, p < 0.05, two-way ANOVA). In social novelty tests (Figure 2D), P301S mice also displayed a progressive loss of preference towards the new social stimulus against the old social stimulus (F_{4,93 (genotype)} = 10.52, p < 0.01, two-way ANOVA). The discrimination index (calculated as (T_new − T_old)/(T_new + T_old)) was lower in P301S mice than WT mice for 2–4 months and reached statistical significance at 5–6 months of age (WT: n = 9, P301S: n = 9, p < 0.05, two-way ANOVA).

An AD model, 5xFAD mouse, which carries five familial AD mutations on human amyloid precursor protein (K670N/M671L + I716V + V717I) and human presenilin 1 (M146L + L286V) [35], was also tested to examine the trajectory of behavioral changes. In Barnes maze tests (Supplementary Figure 1A), FAD mice didn’t exhibit memory deficits until 5–6 months old (n = 6–7/group/age, F_{2,31 (interaction)} = 4.86, p < 0.05; F_{1,31 (genotype)} = 4.2, p < 0.05, two-way ANOVA). No significant change was found in 5xFAD mice in social preference (Supplementary Figure 1B) and social novelty (Supplementary Figure 1C) tests.

Taken together, these data indicate that the tauopathy model exhibit significant cognitive and social deficits at 5–6 months. For some more challenging cognitive tasks, such as BM, the significant deficits are manifested at even earlier time points (3–4 months). Thus, P301S mice is a better model to capture early changes in various behavioral aspects of AD.

Pyramidal neurons in PFC and hippocampus of P301S mice exhibit altered dendritic morphology at different ages

To examine the longitudinal changes in neuronal structure, which may be associated with the alteration of behavioral function, we performed Golgi-Cox staining of pyramidal neurons in PFC and hippocampus of P301S mice at early (3-month) and late (9-month) time points. Neurons were traced and reconstructed from z-stack images using Neuromantic application. Axon was excluded from reconstruction as it would interfere with the analysis of dendrites. Number of branches were evaluated using Sholl analysis and compared using two-way ANOVA (factors: Genotype x Sholl Radius).

Representative traced PFC pyramidal neurons of WT vs. P301S mice at 3 months and 9 months are shown in Figure 3A, 3D. Z-project images of these neurons are shown in Supplementary Figure 2A, 2C. At 3 months old (Figure 3B, 3C), no changes in basal (n = 10 neurons per group, F_1,342 = 2.49, p = 0.12) or apical (F_1,648 = 1.49, p = 0.22) dendrite arborization were observed. At 9 months of age (Figure 3E, 3F), basal dendrites showed no significant changes (n = 11 neurons per group, F_1,460 = 2.02, p = 0.16). However, Sholl analysis of apical dendrites revealed a significant (F_1,700 = 61.98, p < 0.0001) reduction of the number of projections from proximal (240 μm, p = 0.02; 320 μm, p = 0.04; 360 μm, p = 0.04) and distal (800 μm, p = 0.04; 840 μm, p = 0.04) stems of apical dendrites in PFC pyramidal neurons from P301S mice.

Figure 3. Golgi staining revealed morphological changes in pyramidal neurons of PFC and hippocampus from P301S Tau mice at different ages. (A, D, G, J) Representative reconstructed images of layer V pyramidal neurons in PFC (A, D) or CA1 hippocampus (G, J) from WT and P301S mice at 3 months (A, G) and 9 months (D, J). Basal dendrites are traced in blue, while apical dendrites are traced in red. Scale = 100 μm. (B, C, E, F) Plots of dendritic branching as measured by Sholl analysis of basal dendrites and apical dendrites of layer V PFC pyramidal neurons from WT and P301S mice at 3 months (B, C, n = 10 neurons/group) and 9 months (E, F, n = 11 neurons/group). (H, I, K, L) Plots of dendritic branching as measured by Sholl analysis of basal dendrites and apical dendrites of CA1 pyramidal neurons from WT and P301S mice at 3 months (H, I, n = 10 neurons/group) and 9 months (K, L, n = 10 neurons/group). ^*p < 0.05, ^**p < 0.01, ^***p < 0.001, and ^****p < 0.0001, two-way ANOVA.

Representative traced hippocampal CA1 pyramidal neurons of WT vs. P301S mice at 3 months and 9 months are shown in Figure 3G, 3J. Z-project images of these neurons are shown in Supplementary Figure 2B, 2D. At 3 months (Figure 3H, 3I), Sholl analysis for arborization of basal dendrites revealed a significant reduction in CA1 pyramidal neurons of P301S mice (n = 10 neurons per group, F_1,304 = 65.62, p < 0.0001; 200 μm, p = 0.0045; 240 μm, p = 0.0004; 280 μm, p < 0.0001; 320 μm, p = 0.0026; 360 μm, p = 0.027), while apical dendrites did not show any changes (F_1,420 = 2.15, p = 0.14). At 9 months (Figure 3K, 3L), Sholl analysis revealed a significant reduction in the number of branches from basal dendrites (n = 10 neurons per group, F_1,240 = 9.17, p = 0.0027; 160 μm, p = 0.045; 200 μm, p = 0.077), as well as a decrease in projections from distal apical stems (F_1,252 = 12.43, p = 0.0005; 520 μm, p = 0.05).

Transcriptomic analysis reveals longitudinal alterations of gene expression in the hippocampus of P301S mice

To find out the molecular basis that might underlie the longitudinal changes in behavioral phenotypes and neuronal morphology, we analyzed the RNAseq data from hippocampus of WT vs. P301S (PS19) mice at various ages [36]. Using a cutoff of P_adj < 0.05 and Fold change |(FC)|>1.2, we identified differentially expressed genes (DEGs) at the late stage (9 and 12 months) of P301S tau mice (283 upregulated, 206 downregulated). The 283 late stage upregulated genes (Supplementary Figure 3A–3D) are enriched in microglia cell activation and related phagocytosis process. The protein-protein interaction network demonstrated hub genes including complement system (C1qa, C1qb). These genes had a low expression at the early stage of P301S tau mice, and gradually increased at the older age. The 206 late stage downregulated genes (Supplementary Figure 3E–3H) are enriched in synaptic transmission and action potential-related pathway, and many hub genes are ion channels (Scn2b, Kcnip2). These genes had a high expression at the early stage, and gradually decreased at the older stage. This pattern is consistent with transcriptomic changes in neurodegenerative diseases by other studies [37–39].

Next, we searched for gene changes that occur at the early stage. Among the 451 upregulated genes in the hippocampus of 3-month-old P301S mice, 389 genes were exclusively increased at 3 months, compared to 3-month-old WT mice (Figure 4A, 4B and Supplementary Table 1). GO Biological Process (BP) analyses of these upregulated genes indicated that the most prominently enriched pathways included the regulation of neuronal synaptic plasticity, modulation of excitatory postsynaptic potential, Wnt signaling, and regulation of spine development. Key genes in the top two GO pathways included Nlgn1 (encoding Neuroligin-1), Shank3, Stx1b (encoding Syntaxin 1B), Syp (encoding Synaptophysin), Rab34, and Rab5a (Figure 4C), all of which are involved in synaptic organization and synaptic vesicle trafficking.

Figure 4. Transcriptomic analysis revealed genome-wide alterations in the hippocampus of P301S mice at different ages. (A, D) Venn-diagrams showing the significantly up-regulated (A) or down-regulated (D) genes in P301S mice at 3 months and 9 months. (B, E) Heatmaps representing expression (row z-score) of genes that were up-regulated (B) or down-regulated (E) in P301S mice exclusively at 3 months, compared to age-matched WT samples (n = 5 mice/group). (C, F) GO Biological Process analysis of up-regulated genes (C) or down-regulated genes (F) in P301S mice exclusively at 3 months.

Among the 473 downregulated genes in the hippocampus of 3-month-old P301S mice, 428 genes were exclusively decreased at 3 months (Figure 4D, 4E, and Supplementary Table 2). GO BP analyses of these downregulated genes indicated that the most prominently enriched pathways included the regulation of protein folding, cellular response to stress, gene expression, and mitochondrial translational termination. Key genes in the top two GO pathways included Hspa1a, Hspa1B, Hspa5, Hspa13, Hspe1, Hsph1, Hspb1, and Hsp90b1 (Figure 4F), all of which encode the heat shock protein family members that act as chaperones to regulate protein assembly and ER homeostasis.

Gene expression is altered in the PFC and hippocampus of P301S mice at different ages

To validate the microarray data, we next performed quantitative PCR (qPCR) to examine the expression of selected genes in PFC and hippocampus of WT vs. P301S mice at 3 months and 9 months (n = 6/group). We first focused on synaptic genes, many of which were found to be exclusively upregulated in hippocampus of P301S mice at 3 months by the transcriptomic analysis. We revealed 4 patterns of gene alterations. Category 1 (UP-DOWN) genes showed the elevated expression at 3 months and the diminished expression at 9 months in both regions of P301S mice, including the genes involved in synaptic transmission and plasticity like Nlgn1, Wnt7a, Syp, and Stx1a/b (Figure 5A–5D), which is largely consistent with the microarray data. Category 2 (NO-DOWN) genes showed no change at 3 months in either PFC or hippocampus, but a significant downregulation at 9 months in both regions, which included Shank3 (encoding a postsynaptic scaffolding protein regulating dendritic spine formation and synapse maintenance) and Snap25 (encoding a SNARE complex protein regulating transmitter release) (Figure 5E, 5F). Category 3 (NO-UP) genes showed no change at 3 months and a significant upregulation at 9 months in PFC, while remained unchanged in hippocampus, which included Rab34 and Rab5a (encoding Ras superfamily protein members involved in endocytosis) (Figure 5G, 5H). Category 4 (NO CHANGE) included Vamp2, which showed no change at 3 months or 9 months in PFC or hippocampus (Figure 5I).

Figure 5. Gene expression profiling revealed various patterns of changes in PFC and hippocampus of P301S mice at different ages. (A–I) Bar graphs of mRNA levels in PFC and hippocampus of WT and P301S mice at 3 and 9 months (n: WT/Tau = 6/6 per group), which included: Category 1 (UP-DOWN) (A–D, Nlgn1, Wnt7a, Syp, Stx1b); Category 2 (NO-DOWN) (E, F, Shank3, Snap25); Category 3 (NO-UP) (G, H, Rab34, Rab5a); and Category 4 (NO CHANGE) (I, Vamp2). ^#p < 0.1, ^*p < 0.05, ^**p < 0.01, ^***p < 0.001, t-test.

We further examined a few heat shock protein genes, which were found to be exclusively downregulated in hippocampus of P301S mice at 3 months by the transcriptomic analysis. Surprisingly, we found either no change or a significant increase in most of these genes in PFC or hippocampus of 3-month-old P301S mice (n = 6), compared to 3-month-old WT mice (n = 6) (Supplementary Figure 4). The discrepancy could be due to small sample sizes or different methodological sensitivities.

While it is difficult to point out which category (UP-DOWN, NO-DOWN, NO-UP) would be the key players, we speculate that the loss of synaptic proteins resulting from less transcription (UP-DOWN, NO-DOWN) and/or more endocytosis (NO-UP) is responsible for the synaptic deficits in aged group.

Given the dynamic alteration of synaptic genes in P301S mice, we finally performed Western Blotting to detect protein changes at different time points. As shown in Supplementary Figure 5, the synaptic proteins NLGN1, STX1A, VAMP2, SYP, and SHANK3 were largely unchanged or showed a trend of reduction in total lysates of PFC and hippocampus from P301S mice at 3 months and 9 months, compared to age-matched WT mice (n = 4/group).

Animals

All experiments were performed with the approval of the State University of New York at Buffalo Animal Care Committee. The PS19 mouse line harboring the T34 isoform of microtubule-associated protein tau (MAPT) with one N-terminal insert and four microtubule binding repeats (1N4R) encoding the human P301S mutation [16] was obtained from the Jackson laboratory. The genetic background was (C57BL/6 × C3H) F1, and the breeding system was noncarrier × hemizygote. Genotyping was performed by PCR of tail DNA according to the manufacturer’s protocol. Both male and female P301S mice (1–9 months) and age-matched WT littermates were used. Since no sex-dependent effects were found in our measured parameters, data from both males and females were pooled together.

Behavioral testing

Animals were habituated to the experimental room in their home cages for at least 30 minutes before testing. The room light was adjusted to dim during all behavioral experiments. Mice were returned to the home cages between trials to rest. To mask olfactory cues, all testing apparatuses were cleaned with 75% ethanol. Operators were blind to experimental groups during testing and scoring. ANY-maze 5.1 (Stoelting) was used for animal tracking and data analysis. To avoid potential problems of repeated measurements, each animal was only tested once for each behavioral assay. At each age group, different mice were used.

Golgi-cox staining and Sholl analysis

Neurons were visualized using FD Rapid GolgiStain™ Kit (PK401A, FD Neurotechnologies, Inc.). Mouse brains were perfused with 1X PBS. After 15 days of stain impregnation, brains were sliced (100 μm) using a vibratome and mounted onto gelatin-coated slides (PO101, FD Neurotechnologies, Inc.). Allowing the slides to dry for 48 hours, slices were treated with blackening solution, dehydrated through graded ethanol, cleared with xylene, and cover-slipped with DPX mounting medium. For analysis, 4 to 8 neurons were selected per region per animal. Pyramidal neurons from the infralimbic and prelimbic regions of the PFC and the CA1 region of the hippocampus were selected. A neuron was selected if at least two intact primary basal dendrites, along with intact distal apical stem(s), were attached to the soma. Z-stack images were taken with 10× (1.6× magnification) or 20× (1× magnification) objectives using Leica DMi8 inverted microscope for bright-field imaging and Leica LAS X software for image acquisition. Neurons were semi-automatically traced and reconstructed using Neuromantic application v1.6.3. Sholl analysis was performed by uploading the reconstruction file (.swc) into ImageJ Sholl Analysis Plugin part of SNT, v4.1.12 [65]. Analyses for basal and apical dendrites were done separately at 40 μm intervals.

Both apical and basal dendrites of 84 (total) pyramidal neurons in 8 groups from two different regions (PFC and CA1) in two different genotypes (WT and P301S) of mice at two different ages (3 months and 9 months) were reconstructed and analyzed. The technical limitation of Golgi staining precludes the possibility of getting a large number of neurons with completely preserved basal and apical dendritic arborizations from each animal.

Bioinformatic analysis

RNA-seq dataset (GSE89979) for gene expression in hippocampus of PS19 Tau Transgenic mice at 3 and 9 months [36] was acquired using NCBI’s public database gene expression omnibus (GEO) and used for Differentially Expressed Gene (DEG) analysis, as what we previously described [39]. Genes with differential expression of adjusted p-values below 0.05 and fold change (FC) above 1.2 were considered as being significantly different. Gene Ontology (GO) analyses and functional classification was done using Enrichr. The p-value for each category and the combined score generated by Enrichr were used to identify top biological process enrichment for each dataset.

Quantitative real-time PCR

Total RNA was isolated with the TRIzol reagent (Invitrogen, USA), and the genomic DNA was removed by incubating with RNase-free DNase I (Invitrogen, USA). Purified mRNA was then converted to cDNA with an iScript reverse transcription kit (Bio-Rad, USA). Quantitative real-time PCR was performed on the iCycler iQ Real-Time PCR Detection System and iQ Supermix (Bio-Rad, USA) according to the manufacturer’s instructions. The average value of two replicates of each sample was expressed as the threshold cycle (Ct), at which the fluorescence signal reaches 10X the SD of the baseline. Then, the difference (ΔCt) between the Ct value for target gene and the Ct value for housekeeping gene GAPDH (ΔCt = Ct(target gene) – Ct(GAPDH)) was calculated for each sample. The relative level of target gene expression was determined by fold change (FC) = 2^−ΔΔCt, where ΔΔCt = ΔCt – mean of ΔCt (control group). Primers used are:

Western blotting

Mice were sacrificed by decapitation, and brains were quickly removed and cooled in ice-cold PBS. Next, brains were placed and sliced in a 1 mm interval mouse brain matrix (Zivic Instruments, USA). Brain sections were selected and 2 punches from PFC and whole hippocampus were taken. The tissue was homogenized in 0.2 μM filtered 1% SDS supplemented with protease inhibitor (Complete, Roche, USA). Protein concentration was quantified with Bradford protocol (Bio-Rad, USA). Samples were boiled in 4X SDS loading buffer for 5 min and the total protein extracts (10 μg) were separated on 6% or 12% SDS gels.

Nitrocellulose membranes were incubated overnight at 4°C with the following primary antibodies: tau (Tau-5) (1:500, Thermo Fisher Scientific, AHB0042), Ser202/Thr205 phospho-tau (AT8) (1:500; Thermo Fisher Scientific, MN1020), Ser214 phospho-tau (1:1000; Thermo Fisher Scientific, 44-742G), Nlgn1 (1:5000, Proteintech, 66964-1-Ig), Stx1a (1:500, Proteintech, 66437-1-Ig), Syp (1:5000, BD Biosciences, 611880), Shank3 (1:500, NeuroMab, 367-62), Vamp2 (1:500, Proteintech, 10135-1-AP), glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (1:1000, Cell Signaling, 5174). All proteins were detected by using an anti-mouse (1:2000, GE Lifesciences, NA931) or anti-rabbit (1:2000, GE Lifesciences, NA934) secondary antibody IgG coupled to peroxidase and developed by ECL (SuperSignal West-Pico, Thermo Fisher). Images and data analysis were acquired with Chemidoc XRS system (Bio-Rad).

Statistical analyses

Data were analyzed with GraphPad Prism v.7 (GraphPad). Differences between two groups were assessed with unpaired Student’s t-test with unequal variance. Experiments with more than two groups were subjected to two-way ANOVA with Bonferroni correction for multiple post hoc comparisons. Data points identified as statistically significant outliers (as determined by Grubb’s test) were removed from the analyses. All values are mean ± SEM.