Formation of Tau Inclusions in Knock-in Mice with Familial Alzheimer Disease (FAD) Mutation of Presenilin 1 (PS1)*

Mutations in the presenilin 1 (PS1) gene are responsible for the early onset of familial Alzheimer disease (FAD). Accumulating evidence shows that PS1 is involved in γ-secretase activity and that FAD-associated mutations of PS1 commonly accelerate Aβ1–42 production, which causes Alzheimer disease (AD). Recent studies suggest, however, that PS1 is involved not only in Aβ production but also in other processes that lead to neurodegeneration. To better understand the causes of neurodegeneration linked to the PS1 mutation, we analyzed the development of tau pathology, another key feature of AD, in PS1 knock-in mice. Hippocampal samples taken from FAD mutant (I213T) PS1 knock-in mice contained hyperphosphorylated tau that reacted with various phosphodependent tau antibodies and with Alz50, which recognizes the conformational change of PHF tau. Some neurons exhibited Congo red birefringence and Thioflavin T reactivity, both of which are histological criteria for neurofibrillary tangles (NFTs). Biochemical analysis of the samples revealed SDS-insoluble tau, which under electron microscopy examination, resembled tau fibrils. These results indicate that our mutant PS1 knock-in mice exhibited NFT-like tau pathology in the absence of Aβ deposition, suggesting that PS1 mutations contribute to the onset of AD not only by enhancing Aβ1–42 production but by also accelerating the formation and accumulation of filamentous tau.

Alzheimer disease (AD) 3 is characterized pathologically by neurofibrillary tangles (NFTs), which are composed of highly phosphorylated tau, and by neuronal loss and A␤ deposition. AD is manifested symptomatically by dementia. Presenilin 1 (PS1), a gene identified to be responsible, in part, for early onset familial Alzheimer disease (FAD), has been cloned (1,2). To date more than 70 mutations of the PS1 gene have been reported (3,4). In each mutation, early onset of AD develops with 100% penetration (3,4). PS1 is required for ␥-secretase to cleave amyloid precursor protein into A␤ species such as A␤  and A␤  .
Improper cleavage of amyloid precursor protein because of a PS1 mutation increases the production of A␤ 1-42 (5-7), a highly aggregative, neurotoxic species of A␤ that is longer than the less toxic A␤  . One hypothesis for the neurodegeneration observed in AD, therefore, is that PS1 mutation leads to increasing amounts of extracellular, neurotoxic A␤ 1-42 , thereby inducing neurodegeneration (8 -11).
Accumulating data suggest that, in addition to its role in A␤ 1-42 production, PS1 mutation also contributes to NFT formation. For example, PS1 conditional knock-out mice display phosphorylated tau, synaptic dysfunction, and memory impairment, even in the absence of A␤ production and deposition (12). In a related line of research, some patients clinically diagnosed with fronto-temporal dementia (FTD) have been shown to harbor PS1 mutations (13)(14)(15)(16). Interestingly, FTD is characterized by the appearance of NFTs without A␤ deposition (17). A patient harboring the G183V PS1 mutation displayed the clinical manifestations of FTD and exhibited phospho-tau-positive Pick body pathology throughout the cortex and limbic region, without A␤ deposition. Other reports have also shown that PS1 mutations accelerate NFT formation and neuronal loss without affecting the rate of A␤ deposition (18). Thus, these data lead to the hypothesis that PS1 mutations might contribute to NFT formation as well as increase A␤ 1-42 production in FAD.
To investigate this hypothesis, we examined tau pathology in mutant PS1 I213T knock-in mice (19). This line of mice was generated with a targeted insertion of the I213T missense mutation into exon 7 of the mouse PS1 gene using homologous recombination. Therefore, these PS1 mutant knock-in mice harbor the FAD mutation in the mouse PS1 gene and produce PS1 I213T. Heterozygote mutant PS1 I213T mice showed no change in A␤ 1-40 levels but had increased levels of A␤ 1-42 , a 1.3-fold increase when compared with wild-type mice. This A␤  increase is comparable to that observed in human cases. Because this mouse strain shares the same PS1 genotype and related pathology as that of patients harboring the PS1 mutation, we initially expected these mice to develop the AD phenotype. The increase in murine A␤ 1-42, however, failed to lead to a corresponding A␤ deposition, possibly because murine A␤ has a different amino acid sequence that reduces its tendency to aggregate. Using this dissociation to our advantage, we investigated the A␤ deposition-independent effects of the PS1 mutation in this mouse model. We found that GSK-3␤ activation was followed by the accumulation of hyperphosphorylated tau in the hippocampal region, which fulfills the histological criteria for the presence of NFTs.

EXPERIMENTAL PROCEDURES
Animals-Mutant PS1 I213T knock-in mice (mPS1 mice) were maintained at the RIKEN BSI animal facilities according to the Institute guidelines for the treatment of experimental animals. * This work was supported in part by a grant from the Ministry of Education, Science, Sports, and Culture of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1  Antibodies-The following antibodies were used: mouse monoclonal anti-tubulin (DM1A, Sigma); anti-ubiquitin (Santa Cruz Biotechnology); anti-GSK3␤ (Transduction Laboratory); anti-MAP2 (HM2, Sigma); anti-tau Alz50, which recognizes the conformational epitope of paired helical filaments (PHFs), component of NFT (a generous gift from Dr. P. Davies, Albert Einstein College of Medicine, Bronx, NY); anti-phosphorylated tau AT8 (Innogenetics Zwijndrecht); anti-dephosphorylated tau, Tau-1 (Chemicon); rabbit polyclonal anti-tau JM (20); anti-phosphorylated tau PS199, PS262, PS396, PS404, and PS422 (BIO-SOURCE), which recognize tau phosphorylated at the indicated sites; and anti-GSK3␤ Ser-9 (Cell Signaling).
Western Blot Analysis-Brains were homogenized in modified radioimmunoprecipitation assay (RIPA) buffer (50 mM Tris, 150 mM NaCl, 1% Nonidet P-40, 5 mM EDTA, 0.5% sodium deoxycholate, and 0.1% SDS, pH 8.0), and the suspension was centrifuged at 100,000 ϫ g for 20 min at 4°C in an Optima TL ultracentrifuge (Beckman). The pellet was washed five times with 1% SDS-Tris-buffered saline (TBS) (50 mM Tris, 150 mM NaCl, and 1% SDS, pH 8.0) followed each time by centrifugation. The SDS-insoluble pellet was solubilized in 70% formic acid, lyophilized, reconstituted in Laemmli SDS-PAGE sample buffer, and subjected to SDS-PAGE. Separated proteins were blotted onto Immobilon-P membranes (Millipore). The membranes were incubated with primary antibody then with the species-appropriate horseradish peroxidase-conjugated secondary antibody. Immunoreactivity was visualized with a chemiluminescent detection system (ECL, Amersham Biosciences). Quantitation and visual analysis of immunoreactivity were performed with a computer-linked LAS-1000 Bio-Imaging Analyzer System (Fujifilm) using the software program Image Gauge 3.0 (Fujifilm).
Glycogen Synthase Kinase (GSK)-3␤ Activity-Brains were homogenized in TBS (pH 7.4) and centrifuged at 100,000 ϫ g for 20 min at 4°C in an Optima TL ultracentrifuge (Beckman). Protein concentration in the supernatant was determined with a Bradford protein assay, and 10-g samples were assayed for GSK-3␤ activity with an immunoprecipitation assay (21).
Ultrastructural Studies-For electron microscopy studies, SDS-insoluble materials were prepared from the brains of mPS1 mice as described above in the Western blot analysis section. The materials were mildly sonicated and dispersed in phosphate-buffered saline. The dispersed solution was absorbed onto glow-discharged supporting membranes on 400-mesh grids and prefixed by floating the grids on drops of 4% paraformaldehyde in 0.1 M phosphate buffer for 5 min. After washing, the grids were incubated with primary antibody (JM, anti-tau antibody), followed with a 5-nm colloidal gold-conjugated secondary antibody. The grids were then negatively stained with 2% sodium phosphotungstic acid, dried, and observed with a LEO 912AB electron microscope at 100 kV.
Immunohistochemical and Histopathological Studies-Brains were immersion-fixed in 10%-buffered formalin, and paraffin-embedded sections (4 m) were prepared. PS199, Alz50, anti-MAP2, and AT8 were used as primary antibodies. After reacting the sections with speciesappropriate secondary antibodies, we visualized for light microscopy analyses immunoreactive elements by treating the sections with ABC followed by DAB using Peroxidase Stain DAB kits (Nacalai Tesque Japan). PS199 and anti-␣-tubulin were used as primary antibodies for confocal laser microscopy analyses. Immunoreactive elements were visualized with Alexa568-conjugated anti-mouse IgG and Alexa488conjugated anti-rabbit IgG, and then examined with a Radiance 2000 KR3 confocal microscope (Bio-Rad). We stained some sections with Congo red and Thioflavin T, which recognize the ␤-sheet structure of tau fibrils, and then examined them with a light microscope equipped with crossed polarizing filters (Nikon). NFTs were identified using a standard Gallyas silver-impregnation method, which is used to assess structural changes of the brain in AD (22).

RESULTS
The A␤ levels in the brains of mPS1 knock-in mice were quantified by sandwich enzyme-linked immunoassay (23) and Western blot analysis. Similar to a previous report (19), the level of A␤  was elevated in the brains of mPS1 mice compared with that in the brains of wild-type mice (wPS1 mice). Most of the A␤ 1-40 and A␤  was recovered in the Triton X-100-soluble fraction, and very little was recovered in the 1% SDS-insoluble fraction, suggesting that neither A␤ 1-40 nor A␤ 1-42 aggregated and deposited within the brains of the mPS1 mice. Moreover, A␤ immunostaining in tissue sections was absent, suggesting again that neither extracellular nor intracellular A␤ accumulated in vivo (data not shown). Thus, murine A␤ failed to deposit in the brains of mPS1 mice, even though A␤ 1-42 levels in these brains increased.
Characterization of NFT-like Pathology in Mutant PS1 Mice-Fivemonth-old heterozygous mPS1 mice exhibited no pathological changes (Fig. 1A); however, in 7-month-old or older mice, we detected phosphotau accumulation (PS199) in neurons in the hippocampal region (Fig.  1B). The prevalence and distribution of these PS199-positive neurons gradually increased and widened, respectively, with age, and in 14 -16month-old mice, we observed phospho-tau immunoreactivity in hippocampal CA3 neurons (Fig. 1D). By contrast, wPS1 mice did not show this pattern of phospho-tau immunoreactivity (Fig. 1, C and D). Alz50, an antibody that recognizes the conformational change of tau in PHFtau, also labeled CA3 neurons (Fig. 1, E-G). In summary, these findings indicate that heterozygous mPS1 mice, whose alleles most precisely reflect the genotype of humans bearing this mutation, exhibited phospho-tau accumulation with PHF-tau epitopes, whereas wPS1 mice of the same age showed no sign of tau accumulation.
Data related to the other histological criteria for NFTs confirmed these findings. In heterozygous mPS1 mice, we observed Congo red birefringence (Fig. 1, H-J) and Thioflavin T reactivity (Fig. 1K) in hippocampal neurons. The Gallyas silver staining method revealed argyrophillic neurons in the hippocampus of mPS1 mice (Fig. 1L). Argyrophillic and Congo red-positive neurons were less numerous than phospho-tau-positive neurons (less than 5% of phospho-tau-positive neurons). The tau-accumulating neurons of these mice also exhibited reduced ␣-tubulin immunoreactivity (Fig. 1, M and N) similar to that displayed by NFT-bearing neurons. As we previously showed in tau Tg mice (24,25), weaker ␣-tubulin immunoreactivity might indicate destruction of microtubules in tau-accumulating neurons in the mPS1 mice. TUNEL staining, however, revealed no signs of apoptosis in these neurons (Fig. 1, O-R). Taken together, these results suggest that mPS1 affects the cytoskeleton of hippocampal neurons and induces NFT-like accumulation of hyperphosphorylated tau.
Biochemical and Ultrastructural Analysis of Tau in Mutant PS1 Mice-We confirmed the accumulation of NFT-like tau in mPS1 mice with biochemical and electron microscopy analyses. Because tau becomes detergent insoluble when aggregated, we assessed the amount of tau in the SDSinsoluble fraction derived from the brains of mPS1 mice. As shown in Fig.  2A, small amounts of tau were recovered in the SDS-insoluble fraction from 2-month-old mPS1 mice. The amount of tau recovered in the SDS-insoluble fraction increased with aging, and a large amount of tau was recovered from 14-month-old heterozygous mPS1 mice compared with age-matched wPS1 mice. (The amount of SDS-insoluble tau in mPS1 mice was ϳ2% of the total tau in these mice).
We also investigated the amount of SDS-insoluble tau in different brain regions of mPS1 mice. Tau was recovered from the hippocampus, and small amounts were recovered from the cerebral cortex, striatum, and cerebellum. This might be explained by the inverse correlation of NFT with pin1 expression (26).
The SDS-insoluble material recovered from mPS1 mice were further investigated with electron microscopy. As shown in Fig. 2C, tau-positive fibrils were detected in the SDS-insoluble fraction. These fibrils appeared to be straight tubules, about 10 nm in diameter. The biochem-  Ser-9)). FEBRUARY 24, 2006 • VOLUME 281 • NUMBER 8 ical and ultrastructural analyses strongly suggest that NFT-like tau aggregation formed primarily in hippocampal neurons of mPS1 mice.

Tau Inclusions in FAD Mutant PS1 Knock-in Mice
The NFTs found in AD brains contain highly phosphorylated tau (27). Hyperphosphorylation of tau leads to the formation of fibrillar tau (28). To determine whether tau hyperphosphorylation also occurs in mPS1 mice, we examined the extent of tau phosphorylation in 14-month-old wPS1, heterozygous mPS1, and homozygous mPS1 mice (Fig. 2D). Immunoblotting with various phosphorylation-dependent anti-tau antibodies revealed that the amount of SDS-insoluble tau was nearly the same in heterozygous and homozygous mPS1 mice (Fig. 2D,  tau(ins.)). This amount, however, was greater than that in wPS1 mice. Although the total amounts of tau in soluble fractions from the three types of mice were similar (i.e. bands had slower mobility than the unphosphorylated Tau-1-immunoreactive band) (Fig. 2D, tau(sol.)), the amounts of phosphorylated tau immunostained by PS199, PS262, PS396, PS404, PS422, and AT8 were elevated in the heterozygous and homozygous mPS1 mice compared with those in the wPS1 mice. The extent of tau phosphorylation at the AT8, PS422, and PS199 epitopes appeared to depend on the number of mPS1 alleles present in the mice, as shown by the comparatively greater immunosignal intensity of samples derived from homozygous than in heterozygous mPS1 mice, suggesting that mPS1 expression affects the hyperphosphorylation of tau. The immunostaining intensity of Tau-1, an antibody that recognizes unphosphorylated tau, also correlated with genotype. Tau-1 immunoreactivity was greater in samples derived from wPS1 mice than in heterozygous and homozygous mPS1 mice (wPS1 Ͼ heterozygous mPS1 Ͼ homozygous mPS1), confirming that mPS1 induced the hyperphosphorylation of tau.
We also investigated how other kinases and phosphatases may contribute to tau phosphorylation in mPS1 mice. As shown in Fig. 3A, levels of the active forms of phosphorylated JNK; phosphorylated MAP kinase; CDK5; the CDK5 activators p35 and p25; and PP2a were similar among wPS1, heterozygous mPS1, and homozygous mPS1 mice, indicating that the tau phosphorylation mediated by these enzymes are not affected by the PS1 mutation. Nonetheless, other mechanisms are expected to be involved in the mPS1-induced hyperphosphorylation of tau, because GSK-3␤ alone cannot phosphorylate all of the phosphorylation sites in tau.

DISCUSSION
In the present study, we demonstrated that the brains of mice harboring a PS1 mutation accumulated NFT-like phospho-tau. Biochemical analysis of SDS-insoluble tau revealed tau fibrils. These NFT-like tau inclusions were similar to those observed in FTDP-17 mutant tau transgenic mice (24,25,28), in mice overexpressing p25, a CDK5 activator (29), and in Pin1 knock-out mice (26). The neurons of other PS1 knock-in and mutant PS1 transgenic mice; however, failed to show cytoskeletal changes (30,31). To create their mutant PS1 knock-in mice, Guo et al. (32) used a hybrid mouse composed of 129SV and C57BL6 strains, whereas in the present study we used mice resulting from 10 generations of crossbreeding with C57BL6J mice. The genetic background of our mice, which is most likely to be different from the backgrounds of mice used in previously studies, could have influenced how the PS1 mutation contributed to NFT formation and cytoskeletal changes. Another possible explanation for the apparent discrepancy between our findings and others is that our PS1 knock-in mice harbored a different PS1 mutation from that harbored by mutant PS1 knock-in mice developed by other laboratories.
Previously, we found certain PS1 mutations that increase A␤ 1-42 levels are poor predictors for the onset of FAD (5). Our present results, however, suggest that the accumulation of NFT-like tau could determine the onset of AD. These two differing outcomes would also explain why some PS1 mutations accelerate NFT formation and neuronal loss without accelerating A␤ deposition (18). Thus, PS1 mutations that accelerate both NFT formation and A␤ 1-42 production may further accelerate related neuropathologies, suggesting that the cause of early onset AD may be related to a PS1 mutation.
The mechanism underlying the mutant PS1-associated accumulation of NFTs may involve the activation of GSK-3␤. Our results indicate that GSK-3␤ is activated in our mPS1 knock-in mice; this is consistent with other mPS1 transgenic mice. Recently, wild-type PS1 has been shown to activate PI3 kinase/Akt signaling by promoting the association of cadherin and PI 3-kinase, whereas mutant PS1 was unable to do so (32). Thus, mutant PS1 appears to impair PI 3-kinase/Akt signaling by affecting selected signaling receptors (33) or by reducing cadherin/PI3 kinase association (32), which eventually leads to the activation of GSK-3␤. Whereas the mutant PS1-associated activation of GSK-3␤ occurred in young mice, tau accumulation occurred only later on in older mice. This led us to hypothesize that, by some mechanism, phosphorylated tau degrades before it aggregates. This unknown mechanism is then inactivated during aging, leading to the accumulation and aggregation of tau that occurs in aged individuals.
Patients harboring the FAD mutation of PS1 develop AD with 100% penetration. Based on our results, we propose that the PS1 mutation in FAD leads to the early onset of AD through the activation of GSK-3␤, which leads to NFT formation and the loss of neurons and synapses. Moreover, we believe that the rate of GSK-3␤ activation is accelerated by extracellular A␤ oligomers. The exact molecular mechanism mediating the mutant PS1-induced activation of GSK-3␤ requires clarification to further our understanding of how AD develops.