A third-generation mouse model of Alzheimer's disease shows early and increased cored plaque pathology composed of wild-type human amyloid β peptide

We previously developed single App knock-in mouse models of Alzheimer's disease (AD) harboring the Swedish and Beyreuther/Iberian mutations with or without the Arctic mutation (AppNL-G-F and AppNL-F mice, respectively). These models showed Aβ pathology, neuroinflammation, and cognitive impairment in an age-dependent manner. The former model exhibits extensive pathology as early as 6 months, but is unsuitable for investigating Aβ metabolism and clearance because the Arctic mutation renders Aβ resistant to proteolytic degradation and prone to aggregation. In particular, it is inapplicable to preclinical immunotherapy studies due to its discrete affinity for anti-Aβ antibodies. The latter model may take as long as 18 months for the pathology to become prominent, which leaves an unfulfilled need for an Alzheimer's disease animal model that is both swift to show pathology and useful for antibody therapy. We thus utilized mutant Psen1 knock-in mice into which a pathogenic mutation (P117L) had been introduced to generate a new model that exhibits early deposition of wild-type human Aβ by crossbreeding the AppNL-F line with the Psen1P117L/WT line. We show that the effects of the pathogenic mutations in the App and Psen1 genes are additive or synergistic. This new third-generation mouse model showed more cored plaque pathology and neuroinflammation than AppNL-G-F mice and will help accelerate the development of disease-modifying therapies to treat preclinical AD.

We previously developed single App knock-in mouse models of Alzheimer's disease (AD) harboring the Swedish and Beyreuther/Iberian mutations with or without the Arctic mutation (App NL-G-F and App NL-F mice, respectively). These models showed Aβ pathology, neuroinflammation, and cognitive impairment in an age-dependent manner. The former model exhibits extensive pathology as early as 6 months, but is unsuitable for investigating Aβ metabolism and clearance because the Arctic mutation renders Aβ resistant to proteolytic degradation and prone to aggregation. In particular, it is inapplicable to preclinical immunotherapy studies due to its discrete affinity for anti-Aβ antibodies. The latter model may take as long as 18 months for the pathology to become prominent, which leaves an unfulfilled need for an Alzheimer's disease animal model that is both swift to show pathology and useful for antibody therapy. We thus utilized mutant Psen1 knock-in mice into which a pathogenic mutation (P117L) had been introduced to generate a new model that exhibits early deposition of wild-type human Aβ by crossbreeding the App NL-F line with the Psen1 P117L/WT line. We show that the effects of the pathogenic mutations in the App and Psen1 genes are additive or synergistic. This new third-generation mouse model showed more cored plaque pathology and neuroinflammation than App NL-G-F mice and will help accelerate the development of disease-modifying therapies to treat preclinical AD.
The major pathological hallmark of Alzheimer's disease (AD), the most common type of dementia, is deposition of amyloid β peptide (Aβ) in the brain (1,2). Over 300 mutations in the presenilin 1 (PSEN1) and presenilin 2 (PSEN2) genes and more than 50 mutations in the amyloid precursor protein (APP) gene have been identified as disease-associated mutations (Alzforum, http://www.alzforum.org). These findings have led to the development of transgenic mice overexpressing mutant APP or APP/PSEN1 cDNAs (first-generation models) (3). Such mouse models, however, often suffer from experimental limitations caused by overproduction of APP fragments such as C-terminal fragment of APP generated by β-secretase (β-CTF) and APP intracellular domain (AICD), both of which do not appear to accumulate in AD brains and may induce artificial endosomal abnormalities (4) and transcriptional malfunctions (5), respectively. Other overexpression artifacts include calpain activation (6), calpastatin-deficiency-induced early lethality (7), and endoplasmic reticulum stresses (8). In addition, Gamache et al. (9) demonstrated that the random insertion of transgene(s) destroyed unexpectedly large regions in endogenous gene loci of the host animal. We suggest that all transgenic models overexpressing APP or APP/PSEN1 that are being used in research should be subjected to whole genome sequencing to identify the destroyed loci that may have affected their phenotypes.
However, the App NL-G-F line is unsuitable for investigating Aβ metabolism, clearance, and deposition because the Arctic mutation present in the middle of Aβ sequence renders Aβ resistant to proteolytic degradation (25) and prone to aggregation (12). In particular, it is unsuitable for use in preclinical studies of immunotherapy due to its discrete affinity for anti-Aβ antibodies even in the presence of guanidine hydrochloride (GuHCl) (13). The Arctic mutation may also interfere with the direct or indirect interactions between Aβ deposition and apolipoprotein E genotype (26) although there is no experimental evidence. In contrast, the App NL-F line does accumulate wild-type human Aβ, but it takes as long as approximately 18 months for the pathology to become prominent (13): 18 months are too long for researchers to wait in pragmatic terms. The aim of the present study was thus to generate a new mouse model that accumulates wild-type human Aβ as quickly as the App NL-G-F model, but without depending on the Arctic mutation.
We devised the strategy to utilize the heterozygous Psen1 P117L/WT mutant line (Psen1 P117L ) that exhibited the largest increase in Aβ 42 /Aβ 40 ratio in the brain among several Psen1 mutants that we generated (27). In the present study, we attempted to crossbreed App NL-F mice with Psen1 P117L mice despite it being unclear whether their pathogenic effects, both of which act on the γ-cleavage of β-CTF, are additive or not in vivo. We demonstrate here that the Psen1 P117L mutation markedly enhances the pathological phenotypes of App NL-F mice additively or synergistically. We anticipate that these double mutant mice (third-generation model) will become highly relevant tools for examining the mechanisms upstream of Aβ deposition and for preclinical screening of diseasemodifying therapy candidates, which, for instance, promote Aβ degradation or disaggregation, without any concern regarding the artificial effect of the Arctic mutation.

Results
App NL-F Psen1 P117L double-mutant mice produce higher levels of Aβ 42 than App NL-F mice To analyze the combined effect of App and Psen1 mutations on amyloid pathology in vivo, we first prepared App × Psen1 doublemutant mice carrying mutations in the endogenous genes. We crossbred Psen1 P117L mice (27), produced by using cytosine base editors (28), with App NL-F mice (13) to generate App NL-F/NL-F × Psen1 P117L/WT double-mutant mice (App NL-F Psen1 P117L mice). It should be noted that the double-mutant mice used in our experiments were heterozygous for the Psen1 mutation.
App NL-F and App NL-F Psen1 P117L mice expressed indistinguishable quantities of APP and α/β-CTFs (Fig. 1A), suggesting that the P117L mutation does not alter processing of APP by α and β secretases. Consistent with our previous report (13), the Swedish mutations increased the ratio of β/α-CTFs to an identical extent in both lines. We then quantified Aβ 40 and Aβ 42 levels in the cortices of App NL-F and App NL-F Psen1 P117L mice by Enzyme-Linked Immunosorbent Assay (ELISA). At 3 months of age, male App NL-F Psen1 P117L mice produced 22.5fold GuHCl-soluble (Tris-insoluble) Aβ 42 compared with App NL-F mice (Fig. 1B): female samples showed a similar (26.2fold) increase. The increase of Aβ 40 was much smaller, resulting in approximately 11-fold elevation in the Aβ 42 /Aβ 40 ratio of male App NL-F Psen1 P117L mice compared with App NL-F mice (Fig. 1B). Female mice showed a similar tendency. In 12month-old App NL-F Psen1 P117L mice, the quantity of Aβ 42 increased considerably in both Tris-soluble and GuHClsoluble fractions (Fig. 1, C and D). Given that the 3-monthold single Psen1 P117L mice showed only a 2-to 3-fold increase in Aβ 42 production compared with wild-type controls (27), our data indicate that the combination of the App NL-F and Psen1 P117L mutations acts on the γ-secretase activity in an additive or synergistic manner.
The Psen1 P117L mutation also influences Aβ 43 production We previously reported that Aβ 43 is as pathogenic as Aβ 42 (29). We thus performed Aβ 43 ELISA on cortices from 3-and 12month-old App NL-F and App NL-F Psen1 P117L mice. The Trissoluble, but not insoluble, Aβ 43 increased more than 2-fold in the brains of App NL-F Psen1 P117L mice at 3 months of age compared with App NL-F mice (Fig. 2, A and B). Because we treat the "soluble" fractions with GuHCl before the ELISA measurement (30), soluble oligomers are likely included in these fractions. Aβ 43 levels in the GuHCl fractions increased with aging in both App NL-F and App NL-F Psen P117L mice (Fig. 2C). Some Psen1 mutations such as I213T and R278I result in the overproduction of Aβ 43 in vivo (29,31). It is possible that P117L alone or combination with Swedish/Iberian mutations in the App gene may lead to an increase in Aβ 43 by modifying the carboxypeptidase-like activity of γ-secretase in the brain (32,33). Intriguingly, the Aβ 43 pathology became more prominent with aging. (See below.) Aβ deposition starts as early as 3 months of age in App NL-F Psen1 P117L mice We next examined Aβ pathology in the brains of App NL-F Psen1 P117L mice. Immunofluorescence analyses detected Aβ plaques in the cortices of App NL-F Psen1 P117L mice at 3 months of age (Fig. 3, A and B), whereas App NL-F mice took as long as 6 months to reach an initial and minimal deposition of Aβ (13). At 12 months, App NL-F Psen1 P117L mice displayed prominent amyloidosis in the cortex and hippocampus comparable with that of App NL-G-F mice, while significantly fewer Aβ plaques were observed in App NL-F mice (Fig. 3, A-C). Of note, the number of subcortical plaques in App NL-F Psen1 P117L mice was significantly less than that in App NL-G-F mice, implying that App NL-F Psen1 P117L mice may recapitulate the human pathology in a more faithful manner (34). App NL-F Psen1 P117L mice produced dominant deposition of Aβ 42 with minimal Aβ 40 (Fig. 3D), which is consistent with observations made on App NL-F and App NL-G-F mice and human samples (13). Remarkably, we detected a significantly larger number of Aβ 43positive plaques in the cortical, hippocampus, and subcortical regions of App NL-F Psen1 P117L mice than in those regions of App NL-F and App NL-G-F mice (Fig. 3, D and E). These results imply that the Swedish/Iberian and P117L mutations together may accelerate the generation of longer Aβ species including Aβ 42 and Aβ 43 , resulting in Aβ pathology at younger ages.  Figure S1 (CTF) and Figure S2 (APP). B, Aβ 40 and Aβ 42 detected by ELISA from the cortices of 3-month-old App NL-F and App NL-F Psen1 P117L mice. App NL-F (n = 4) and App NL-F Psen1 P117L/WT (n = 4) (Student's t test). C and D, Aβ 40 and Aβ 42 using Tris-HCl (C) and GuHCl (D) soluble fractions from 3-and 12-month-old mice. App NL-F (n = 4) and App NL-F Psen1 P117L/WT (n = 4) (two-way ANOVA followed by Sidak's multiple comparisons test). Each bar represents the mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001.

Combination of the Swedish/Iberian and P117L mutations is associated with cored Aβ plaque formation
Several lines of evidence support the notion that diversity in Aβ species correlates with plaque morphology such as typical cored plaques (35)(36)(37). We therefore performed costaining with N1D antibody raised against Aβ 1-5 peptide (38) and 1-Fluoro-2,5bis(3-carboxy-4-hydroxystyryl)benzene (FSB), which recognizes the β-sheet structure within amyloid fibrils and displays higher fluorescence intensity than 1-bromo-2, 5-bis-(3hydroxycarbonyl-4-hydroxystyryl)benzene (BSB) and Congo red (39,40). We observed that FSB-positive signals were positioned at the center of plaques (Fig. 4A) and that the N1D/FSB double-positive plaques were significantly increased in the cortex and hippocampus of App NL-F Psen1 P117L mice compared with those of App NL-F and App NL-G-F mice (Fig. 4B). No significant difference was observed in the subcortical region between App NL-F Psen1 P117L and App NL-G-F mice. The frequent presence of classic dense-cored plaques in the cortex of double-mutant mice was confirmed by 3,3 0 -diaminobenzidine (DAB) staining (Fig. 4C).
Neuroinflammation is elevated in App NL-F Psen1 P117L mice, particularly in the hippocampus Neuroinflammation surrounding Aβ plaques manifests as one of the pathological features in AD patients (13,41,42), and genome-wide association studies (GWAS) have suggested etiological involvement of neuroinflammation in AD development (43)(44)(45). We thus analyzed the neuroinflammatory status of three mouse lines (App NL-F , App NL-F Psen1 P117L , and App NL-G-F ) by immunofluorescence using antibodies against Aβ (82E1), microglia (Iba1), and astrocytes (anti-GFAP). We confirmed the presence of glial cells surrounding Aβ plaques in App NL-F Psen1 P117L mice (Fig. 5, A and B). Consistent with our previous reports, single App NL-G-F rather than App NL-F mice exhibit robust microgliosis and astrocytosis accompanying progressive amyloidosis (13,14). Quantification of immunofluorescence images indicated that more neuroinflammation was evident in App NL-F Psen1 P117L and App NL-G-F than App NL-F mice (Fig. 5C). This was somewhat predictable because App NL-F Psen1 P117L and App NL-G-F mice accumulate more pathological Aβ than App NL-F mice (Fig. 3, A-C). A unique observation is that App NL-F Psen1 P117L mice exhibited significantly greater microgliosis and astrocytosis than App NL-G-F mice in the hippocampus despite the indistinguishable levels of Aβ amyloidosis therein. This phenomenon may be associated with increased Aβ 43 deposition (Fig. 3, D and E) and cored plaques in App NL-F Psen1 P117L mice (Fig. 4, A and B), but these speculations alone cannot fully account for the neuroinflammation that took place selectively in the hippocampus. In any case, our findings indicate that App NL-F Psen1 P117L mice may be suitable for investigating hippocampal neuroinflammation.  's t test). C, quantity of Aβ 43 from Tris-HCl and GuHCl soluble fractions from 3-and 12-month-old mice. App NL-F (n = 4) and App NL-F Psen1 P117L/WT (n = 4) (two-way ANOVA followed by Sidak's multiple comparisons test). Each bar represents mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001.

Discussion
The primary aim of the present study was to generate App knock-in mice that pathologically accumulate wild-type human Aβ devoid of the Arctic mutation. App NL-G-F mice carrying the Arctic mutation exhibit the most rapid and aggressive pathology among all the App knock-in mice that had been created. Our motivation was based on the undesirable nature of the Arctic mutation that impedes the physiological metabolism and and App NL-G-F (n = 5) (one-way ANOVA followed by Tukey's multiple comparisons test). Each bar represents the mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001. clearance of Aβ (12,25), thus making it difficult to study etiological processes upstream of Aβ deposition. Arctic Aβ also binds to anti-Aβ antibodies raised against Aβ peptide in a distinct manner (13), making App NL-G-F mice unsuitable for preclinical immunotherapy studies. The generation of App knock-in mice that accumulate wild-type human Aβ without the Arctic mutation as quickly as App NL-G-F mice is therefore a prerequisite before disease-modifying strategies targeting mechanisms upstream of Aβ deposition can be developed.
Introduction of the Psen1 P117L mutation into App NL-F mice resulted in an unexpected acceleration of Aβ 42 and Aβ 43 deposition. Although the numbers of cortical and hippocampal plaques visualized by an antibody specific to the N-terminus of Aβ (N1D) were indistinguishable between App NL-G-F and App NL-F Psen1 P117L mice, App NL-F Psen1 P117L mice showed a larger number of cored plaques in the cortex and hippocampus and more gliosis in the hippocampus than was seen in App NL-G-F mice. Mechanisms accounting for these observations are unclear but may be associated with the prion-like nature of Aβ 43 (46). In any case, the App NL-F Psen1 P117L mice may become a useful tool for examining the roles of hippocampal neuroinflammation in the etiology of AD.
Other groups have also attempted to combine the App and Psen1 mutations.  (48). We observed only a few CAA-like structures (less than ten per hemisphere) mainly in meninges in the 12-month-old App NL-F Psen1 P117L mice, indicating that this is parenchymal amyloidosis-dominant model. Typical images of CAA are shown in Figure S3. Unlike the models generated by Flood et al. and Li et al., our third-generation model mice are heterozygous for the Psen1 mutation, making them easier to breed.
We must however point out that App NL-F Psen1 P117L mice are probably inadequate for studying βand γ-secretases and their modifiers because the cleavages catalyzed by these secretases are artificially altered by the mutations. We thus expect the mutant mice to become more suitable for the examining catabolism and clearance of Aβ than its anabolism. This makes agonist(s) for somatostatin receptor subtype(s) a medication candidate for treatment of preclinical AD (49). We will share these mice with the AD research community to accelerate the fight against a disease that deprives patients of their human dignity.

Animals
App NL-F and App NL-G-F mice were described previously (13). Psen1 P117L mice (27) were crossed with the App NL-F mice to generate App/Psen1 double-mutant mice. All double-mutant . Aβ plaques with a cored structure in App NL-F Psen1 P117L mice. A, brain sections from 12-month-old App NL-F Psen1 P117L/WT mice were costained with FSB and N1D antibody. B, FSB/N1D double-positive areas were quantified in the brains of App NL-F , App NL-F Psen1 P117L/WT , and App NL-G-F mice. App NL-F (n = 5), App NL-F Psen1 P117L/WT (n = 5), and App NL-G-F (n = 6) (one-way ANOVA followed by Tukey's multiple comparisons test). Each bar represents mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001. C, representative images of dense-core plaques surrounded by a halo effect detected by DAB staining. Scale bars represent 25 μm. mice used in this study were homozygous for the App mutations and heterozygous for the Psen1 mutation (App NL-F Psen1 P117L ). C57BL/6J mice were used as controls. Male mice were used for biochemical analyses and both male and female mice were used for immunohistochemical studies. All mice were bred and maintained in accordance with regulations for animal experiments promulgated by the RIKEN Center for Brain Science.

Brain sample preparation
Mice were anesthetized with isoflurane, transcardially perfused, and fixed with 4% paraformaldehyde in PBS. The brains were dissected on ice into two halves at the midline. One hemisphere was divided into several parts and stored at -80 C for biochemical analysis, while the other was incubated at 4 C for 24 h and rinsed with PBS until paraffin processing for histochemical analysis.

Western blotting
Mice brain tissues were homogenized in lysis buffer (50 mM Tris pH 7.6, 0.15 M NaCl, and Complete protease inhibitor cocktail [Roche]). Homogenates were incubated at 4 C for 1 h and centrifuged at 15,000 rpm for 30 min, and the supernatants were collected as loading samples. Equal amounts of proteins per lane were subjected to SDS-PAGE and transferred to PVDF or nitrocellulose membranes (Invitrogen). To detect APP-CTFs, delipidated samples were loaded and the membrane was boiled for 5 min in PBS before the next step. After washing and blocking at room temperature, the membranes were incubated at 4 C overnight with primary antibodies Figure 5. Glial responses in App NL-F Psen1 P117L/WT mouse brain. A, inflammatory signals were detected with GFAP and Iba1 antibodies using brain sections from 12-month-old App NL-F Psen1 P117L/WT mice. Aβ pathology was detected by immunostaining with 82E1 antibody. Scale bar represents 500 μm. B, higher magnification image of the area marked in white in (A). Scale bar represents 50 μm. C, immunoreactive areas of GFAP or Iba1 were quantified in the brains of App NL-F , App NL-F Psen1 P117L/WT , and App NL-G-F mice. Each bar represents the mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, App NL-F (n = 5), App NL-F Psen1 P117L/WT (n = 5), and App NL-G-F (n = 4) (one-way ANOVA followed by Tukey's multiple comparisons test).

Immunostaining
Paraffin-embedded mouse brain sections were subjected to deparaffinization and then antigen retrieval was performed by autoclave processing at 121 C for 5 min. After inactivation of endogenous peroxidases using 0.3% , and α-smooth muscle actin (1:1000, Sigma-Aldrich). Amyloid pathology was detected using biotinylated secondary antibody and tyramide signal amplification as described previously (50). For detection of glial activation and CAA, secondary antibodies conjugated with Alexa Fluor 488 or 555 were used. Before mounting, the sections were treated when necessary with Hoechst33342 diluted in PBS. Data images were obtained using a NanoZoomer Digital Pathology C9600 (Hamamatsu Photonics) and EVOS M5000 Imaging System (Thermo Fisher scientific). Immunoreactive signals were quantified by Definiens Tissue Studio (Definiens).

DAB staining
Targeted signals were detected and visualized using VEC-TASTAIN Elite ABC Rabbit IgG kit (Funakoshi) and DAB・ TRIS tablets (Mutokagaku). After deparaffinization and antigen retrieval treatment of mouse brain sections, endogenous peroxidases were inactivated using 0.3% H 2 O 2 solution for 30 min. The sections were blocked with three drops of goat serum in PBS for 30 min at room temperature and incubated with N1D antibody at 4 C overnight. The sections were washed with PBS and incubated with the Elite ABC solution for 30 min and subsequently stained with DAB solution following the manufacturer's instructions. Before mounting, dehydration treatment was performed.

FSB staining
The PFA-fixed tissue sections were deparaffinized, incubated in 0.01% FSB solution in EtOH for 30 min, and then rinsed in saturated Li 2 Co 3 in water for 15-20 s at room temperature. The sections were differentiated in EtOH for 3 min followed by immersion in water for 5 min to stop the reaction. Readers should refer to the Immunostaining section concerning methods for subsequent treatments following antigen retrieval.

ELISA
Mouse cortical samples were homogenized in buffer A (50 mM Tris-HCl, pH 7.6, 150 mM NaCl, and protease inhibitor cocktail) using a medical beads shocker. The homogenized samples were directed to centrifugation at 70,000 rpm for 20 min at 4 C, and the supernatant was measured and collected as a Tris-soluble (TS) fraction in 6 M guanidine-HCl (Gu-HCl) solution containing 50 mM Tris and protease inhibitors. The pellet was loosened with the buffer A and centrifuged at 70,000 rpm for 5 min at 4 C, and then dissolved in 6 M Gu-HCl buffer. After incubation at room temperature for 1 h, the sample was sonicated at 25 C for 1 min. Subsequently, the sample was centrifuged at 70,000 rpm for 20 min at 25 C and the supernatant collected as a Gu-HCl fraction. 100 μl of TS and Gu-HCl fractions were loaded onto 96-well plates and incubated at 4 C overnight using the Aβ 40 , Aβ 42 and Aβ 43 ELISA kit (Wako) according to the manufacturer's instructions.

Statistics
All data are presented as the mean ± SD within each figure.
For comparisons between two groups, data were analyzed by Student's t test. For comparisons among more than three groups, we used one-way analysis of variance (ANOVA) followed by Tukey's multiple comparisons test. Two-way ANOVA followed by Sidak's multiple comparisons test were used when two data sets were analyzed. Statistical analyses were performed using GraphPad Prism version 7.00 software (GraphPad software). The levels of statistical significance were shown as p-values: *p < 0.05, **p < 0.01, ***p < 0.001.

Data availability
Data will be shared upon request at takaomi.saido@riken.jp or hiroki.sasaguri@riken.jp.
Supporting information-This article contains supporting information.
Acknowledgments-We thank Yukiko Nagai-Watanabe for secretarial work. Conflicts of interest-The authors declare that they have no conflicts of interest with the contents of this article.