Article ID: 2023-003
Experimental animal models play an essential role in the study of pathogenic mechanisms and development of novel therapies for diseases. Many disease models are developed by introducing genetic factors and environmental factors of disease onset through drug administration, surgery, or genetic modification that results in disease phenotypes. In addition, higher animals such as non-human primates are useful for investigating diseases whose phenotypes cannot be produced in mice, etc. In monkeys, some strains have diseases similar to humans by spontaneous onset, as a result of repeated generational changes over time. Disease models must have high reproducibility, extrapolation to humans, and applicability. In this review, we summarize recent advances in the development of mouse models of Alzheimer’s disease and the studies that utilize them.
Since our announcement of Alzheimer’s mouse models produced by the knock-in strategy in 2014, numerous researchers have been investigating the mechanisms involved in Alzheimer’s pathogenesis using these models. Additionally, several groups, including our own, have recently introduced new “knock-in” mouse models tailored to specific research objectives. We have summarized the characteristics of knock-in mouse models in this review.
Globally, approximately 55 million people had dementia in 2020 [1]. In Japan, a super-aging society, the number of patients with dementia is rapidly increasing. As of 2020, one in six older adults had dementia, and by 2025, this number projected to be one in five [2]. Dementia has various causes, including vascular dementia, dementia with Lewy bodies, frontotemporal lobe dementia (FTLD), and Alzheimer’s disease (AD). AD is the most common form of dementia. Less than 1% of AD cases are familial, where symptoms were caused by genetic mutations, whereas the majority of cases are sporadic. AD is a progressive dementia characterized by cognitive decline, leading to memory problems, disorientation, learning disabilities, and impaired problem-solving abilities. As the disease progresses, patients experience a loss of communication abilities and increased immobility. These symptoms are caused by neuronal damage and neuronal cell death in the brain. Magnetic resonance imaging of patients with AD show atrophy in the hippocampus, entorhinal cortex, and cerebral cortex, which are involved in memory and learning.
In the brain of patients with AD, two major pathological features are observed: deposition of amyloid β (Aβ) derived from the amyloid precursor protein (APP) and formation of neurofibrillary tangles (NFTs) through the aggregation of hyperphosphorylated tau protein. The amyloid cascade hypothesis, proposed by Hardy and Higgins in 1992 [3], suggests that amyloid pathology precedes tau pathology and neurodegeneration. This hypothesis has gained support from numerous researchers, leading to the active development of therapeutic drugs based on this concept.
In recent years, advancements in imaging technologies such as amyloid positron emission tomography (PET) and tau PET have made it possible to detect the presence of amyloid and tau pathology at an early-stage [4]. Consequently, therapeutic interventions can be initiated at earlier stages, potentially improving treatment outcomes. This has opened up opportunities for the development of therapeutic drugs targeting the early AD stages.
Therapeutics for ADAt present, cholinesterase inhibitors and N-methyl-D-aspartate receptor (NMDAR) antagonists are widely used as therapeutic drugs for AD. Cholinesterase inhibitors increase the levels of acetylcholine, a neurotransmitter involved in memory and cognitive functions, whereas NMDAR antagonists regulate the activity of glutamate, an excitatory neurotransmitter. They can enhance neurotransmission and suppress the hyperactivity of neuronal cells; however, these medications do not provide a radical cure.
In recent years, there has been active development of antibody drugs that directly target amyloid and tau and demonstrate their effectiveness. Antibody therapeutics targeting Aβ, including aducanumab, which was approved in the United States in 2021, are being actively developed. Lecanemab, an antibody that targets amyloid protofibrils, has recently received approval [5]. Lecanemab demonstrated significant improvements in a phase III study involving individuals with mild cognitive impairment and early-stage AD [6]. Preliminary data from a phase III study showed that donanemab achieved greater amyloid removal than aducanumab [7]. Lecanemab selectively binds to and removes protofibrils that accumulate in the brain, whereas aducanumab targets completed amyloid fibers and senile plaques [8]. Conversely, gantenerumab and soranezumab were not effective in patients with early-stage AD [9,10,11]. In addition to passive antibody therapy, active antibody therapy (vaccine) is being developed. Meanwhile, inhibitors of beta-site amyloid precursor protein cleaving enzyme 1 (BACE1), an enzyme involved in Aβ production, have been actively developed, although their effectiveness has not been recognized. Notably, the influential drug verubecestat has been discontinued. Regarding tau pathology, antibody drugs are vigorously developed in recent years, and their research results are highly anticipated. Similar to amyloid antibodies, the focus has shifted to targeting patients at earlier stages, as these therapies have shown limited efficacy in patients with mild dementia. However, owing to the progressive nature of AD over the long term, determining the optimal timing for treatment intervention remains challenging.
Risk factors for ADAging is the greatest risk factor for AD, with the majority of diagnoses occurring in individuals aged >65 years [12]. Recently, various risk factors have been proposed, including the brain–gut interaction [13, 14]. The brain and intestines are interconnected, and disturbances in mental health can manifest as gastrointestinal discomfort, which is referred to as the brain–gut interaction. Emerging studies have implicated the involvement of the gut microbiota in the development of brain diseases [12, 14,15,16]. The composition of gut microbiota differs between individuals with and without dementia [17,18,19,20]. Moreover, α-synuclein, a protein associated with Parkinson’s disease, was found to aggregate in the intestinal tract and can be transmitted to the brain through the vagus nerve [21]. In an AD mouse model called 5XFAD, alterations in gut bacteria composition resulted in an increase in highly toxic forms of amyloid and tau proteins because of elevated asparagine endopeptidase (AEP) activity [22].
The association between sleep disorders and AD has also been investigated. Sleep deprivation and sleep apnea syndrome have been linked to the abnormal excretion of amyloid and tau proteins into the cerebrospinal fluid (CSF), leading to the development and spread of AD [23,24,25,26]. Furthermore, the relationship between AD and viral/bacterial infections has gained attention. Herpes virus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), and periodontal disease–associated bacteria have been reported to infect the brain, contributing to the progression of AD [27,28,29]. Notably, the AD risk gene APOE4-ε4 allele is associated with an increased risk of herpes and SARS-CoV-2 [30,31,32]. Periodontal disease–associated bacteria produce a protein called gingipain, which promotes the production of Aβ and neurotoxicity [29].
Additionally, numerous genome-wide association studies (GWAS) have identified various genetic polymorphisms associated with the risk of sporadic AD. The most significant gene is APOE, which is involved in lipid metabolism and linked to Aβ clearance and tau-related neurological disorders in AD. The nuclear plasma protein BIN1 is the second most relevant factor, and it is involved in intracellular transport mediated by BACE1 and tau clearance. To date, the Alzgene website (http://www.alzgene.org/) has reported 695 genes and 2973 polymorphisms associated with AD. In addition to ApoE, recent studies have focused on TREM2, which is involved in intracellular signaling in microglia [33, 34]. Although the main focus has been on different variants of ApoE (2, 3 and 4), numerous genetic mutations have been identified, including protective mutations such as Christchurch and Jacksonville mutations [35,36,37].
Mouse models are essential for understanding the pathological mechanisms of AD and developing therapeutic drugs. AD is a progressive disease, and determining the optimal timing for intervention is challenging. Ideally, AD mouse models should mimic the entire spectrum of clinical stages, including the preclinical stage, brain atrophy, and cognitive decline. However, owing to disparities in lifespan and brain properties between mice and humans, creating a single model that encompasses all these stages is currently impossible. Currently, researchers select specific mouse models based on their research objectives. These models include mice that replicate amyloid pathology, replicate tau pathology, exhibit brain atrophy caused by tau pathology, and combine various aspects of AD pathology without strictly adhering to the chronological order of disease progression. To enhance the accuracy of AD research, efforts to develop mouse models with improved reproducibility have been ongoing.
Anabolism and catabolism of AβAβ is a peptide that is generated from the APP through cleavage. APP has two main cleavage pathways. The first pathway involves α-secretase and γ-secretase, which do not result in Aβ production. The second pathway involves β-cleavage by BACE1 and γ-cleavage by γ-secretase, leading to Aβ production [38]. Mutations in APP and presenilin genes (PSEN1 and 2), which are part of the γ-secretase complex, have been observed in familial AD cases [39, 40]. While excessive Aβ production can lead to the formation of senile plaques through oligomerization and fibrillation, normally Aβ is eliminated through degradation or excretion into the CSF and blood.
In sporadic AD, Aβ-producing system is not impaired; however, the degradation system may be inhibited for several reasons. Several degrading enzymes such as neprilysin [41], insulin-degrading enzyme [42,43,44], endothelin-converting enzyme 1/2 [45], angiotensin-converting enzyme [46], cathepsin D [47, 48], urokinase-type plasminogen activator [49], and matrix metalloendopeptidase-9 [50] have been identified as involved in peptide degradation. Neprilysin is a type of metalloprotease that primarily targets peptides with a molecular weight of <5 kDa. It cleaves hydrophobic amino acids on the amino-terminal side of these peptides. Before the identification of neprilysin, we highlighted the crucial role of neutral endopeptidases in Aβ degradation in the brain [51]. Subsequently, neprilysin was found to exhibit the highest degradation activity among neutral endopeptidases present in the brain [41]. Importantly, the expression level of neprilysin decreases with aging and in AD [52, 53], suggesting that reduced neprilysin levels are linked to the onset of sporadic AD. Furthermore, somatostatin, a neuropeptide, was found to enhance neprilysin activity, and KATP channels are involved in the somatostatin-mediated regulation of neprilysin activity [54, 55]. Additionally, we have proposed gene therapy using an adeno-associated virus (AAV) vector expressing neprilysin [56]. Notably, recent GWAS analyses have revealed an association between genetic mutations in neprilysin and AD [57]. This mutation leads to an amino acid substitution in the N-terminal region of neprilysin. Thus, the effect of the mutation on neprilysin activity and Aβ degradation must be assessed.
Second-generation models of amyloid pathologyAs mentioned earlier, most of the familial cases of AD are caused by genetic mutations in APP or PSEN1/2, which are components of the γ-secretase complex. Multiple mutations have been identified in the APP gene, including mutations that are located near the β-cleavage site and affect the β-cleavage site, mutations that are located near the γ-cleavage site and affect the γ-cleavage site, mutations that affect the Aβ42 ratio, and mutations that affect the structure of Aβ. The majority of mouse models for amyloid pathology in AD are transgenic or knock-in mice expressing familial mutations of APP and/or PSEN1/2 [58].
In 2014, our laboratory announced the development of App knock-in AD mouse models [59]. Before that, APP transgenic mouse models (APP-Tg) were predominantly used in many studies [58]. Most of the APP-Tg mice overexpress mutant forms of APP (sometimes also PSENI and MAPT) under artificial promoters [58]. Some APP-Tg mouse strains exhibit significant amyloid pathology and cored plaques because of the high levels of mutant protein expression. This leads to neuronal damage associated with amyloid pathology and notable behavioral abnormalities [60, 61]. On the contrary, APP overexpression in these models results in the overproduction of non-Aβ fragments and excessive accumulation of Aβ40 compared with Aβ42, which is are not typically observed in patients with AD. These differences from the human AD brain can lead to phenotypes that are not solely derived from amyloid pathology. In addition, unintended disruption of endogenous genes may be possible through random gene insertion. To overcome these limitations, we developed App knock-in mouse models in which exons 16 and 17 of the App gene encoding the Aβ sequence are substituted with the human sequence carrying familial mutations [59]. The three mouse strains announced in 2014 are AppNL, AppNL-F, and AppNL-G-F mice, which harbor Swedish, Swedish/Iberian, and Swedish/Iberian/Arctic mutations, respectively [59]. Unlike APP-Tg models, App knock-in mice do not overexpress APP or amyloid precursor protein intracellular domain (AICD), a fragment derived from the processing of the amyloid precursor protein. Because the AICD can translocate into the nucleus and play a role in neuronal signaling by interacting with various regulatory proteins involved in gene expression, it is plausible that overexpression of AICD could result in artificial signal transduction. Moreover, they predominantly express Aβ42 and a highly toxic form called 3pEAβ, in which the third glutamate residue is pyroglutamylated [59]. AppNL-F and AppNL-G-F mice develop amyloid pathology slowly from approximately 9 months of age and rapidly from 2 months of age, respectively [59]. AppNL mice, which contain the Swedish mutation, facilitate the generation of C-terminal fragment β (CTF-β), but do not exhibit significant amyloid formation. Therefore, they serve as suitable controls for AppNL-F and AppNL-G-F mice. Researchers can choose these models based on their research objectives. In addition, microgliosis and astrocytosis are observed around the plaques in these models.
As APP or presenilin 1 overexpression in transgenic models may induce endoplasmic reticulum (ER) stress caused by artificial overexpression, we have investigated the ER stress response in these models. Some transgenic mice exhibited elevated ER stress response, whereas App knock-in mice did not show such elevations. Thus, overexpression in transgenic models may induce artificial ER stress, while knock-in mouse models are advantageous [62, 63].
Based on the discovery that deletion of the 3′-UTR of exon 18 in the App gene reduces APP protein expression (unpublished data), we proposed a gene therapy approach targeting the 3′-UTR of exon 18. When Cas9 and 3′-UTR-targeted short-guide RNA are introduced into the brain using an AAV vector, plaque formation in AppNL-G-F mice is reduced in a manner inversely correlated with the efficiency of gene editing [64]. This strategy may be a potential gene therapy method.
In recent years, significant progress was seen in the analysis of pathogenic protein structures using cryo-electron microscopy (cryo-EM). Goedert et al. from the MRC Laboratory of Molecular Biology analyzed the structure of amyloid filaments derived from patients with AD and AppNL-F knock-in mice [65]. They found that the structure of Aβ42 in human patients with AD can be classified into sporadic AD and familial AD. They also observed that the Aβ42 filament structure in AppNL-F mice closely resembled the amyloid structure seen in familial AD. Therefore, AppNL-F mice may be a valuable model for studying the amyloid structure of familial AD and the reactivity of amyloid antibodies. In addition, both their group and ours recently reported on the structure of Aβ42 filaments in AppNL-G-F mice [66]. These studies have suggested that the Arctic variant of Aβ42 has a slightly different structure from the wild-type, promoting intermolecular hydrogen bonding and potentially increasing its propensity to form filaments. This finding aligns with the rapid acceleration of amyloid pathology observed in AppNL-G-F mice.
In parallel with cryo-EM, single-cell RNA sequencing (scRNA-seq) has made remarkable progress in gene expression profiling. This technique enables the classification of microglia into distinct clusters based on their gene expression profiles, which reflect their activity levels. A study conducted in 2017 using 5XFAD mice identified a cluster known as disease-associated microglia (DAM), which plays a significant role in neurodegenerative diseases [67]. Strooper et al. performed scRNA-seq analysis using an AppNL-G-F mouse model and identified an activated response microglia (ARM) cluster similar to DAM and an interferon response microglia (IRM) cluster that expressed more interferon genes, among others [68, 69]. They also conducted spatial transcriptome analysis, which revealed that some genes expressed in ARM and DAM were highly expressed around plaques [68, 69]. Currently, the specific functions of each cluster remain to be fully elucidated, and further functional analysis is needed.
The App knock-in rat, developed by Bai Lu’s group using CRISPR-Cas9 genome editing, is another noteworthy animal model for AD [70]. This rat model carries NL-G-F mutations (Swedish, Iberian, and Arctic mutations), similar to the AppNL-G-F KI mouse. This rat model shows early amyloid plaque formation. Moreover, APN-mab005-positive tau aggregation at 12 months of age and neuronal loss were observed. According to the report, this rat model successfully replicated AD pathologies, from plaque formation to neuronal cell death. Studies using this model are highly anticipated in the future. In addition to this model, several knock-in rat strains have been generated and reported. However, most of them do not exhibit amyloid pathology [71, 72].
Third-generation models of amyloid pathologySo far, we have described second-generation App knock-in mouse models reported in 2014 [59]. Now, we will discuss third-generation mouse models.
AppNL-G-F mice are highly valuable as they exhibit the rapid development of amyloid pathology. However, this model may not be suitable for certain studies involving Aβ metabolism and development of antibody therapeutics, as Aβ with Arctic mutation has different proteolytic resistance and antibody affinity. To address this limitation, we aimed to develop a mouse model that demonstrates rapid pathology formation without Arctic mutation. We generated a crossbred mouse of the AppNL-F mouse and the Psen1 knock-in mouse, which expresses a mutated form of presenilin 1 (P117L) associated with familial AD [73]. The AppNL-F X Psen1P117L heterozygote mice exhibited amyloid plaques starting approximately at 3 months of age, indicating significantly higher Aβ levels than AppNL-F mice. The plaques observed in this model were more cored than those in the AppNL-G-F model, and neuroinflammation increased around the amyloid plaques.
Another limitation of AppNL-F and AppNL-G-F mouse models is that both harbor the Swedish mutation. While the Swedish mutation enhances Aβ production by affecting the β-cleavage site, it poses a challenge for validating BACE inhibitors. Therefore, we developed AppG-F mice by employing genome editing techniques to eliminate the Swedish mutation from the AppNL-G-F mice [74]. Although the AppG-F mice lack the Swedish mutation, plaque formation still occurs at approximately 6 months of age. We also investigated the effects of verubecestat, a BACE inhibitor, and observed its effect on AppG-F mice rather than on AppNL-G-F mice.
In Table 1, we present App knock-in strains that we have developed. Depending on the research objectives, we hope that researchers will select and utilize these mice accordingly.
| Strain name | Modified genes | Mutations | Primary Paper | RBRC No. |
|---|---|---|---|---|
| AppNLKI | App | KM670/671NL (Swedish) | Saito et al., Nat. Neurosci., 2014 | RBRC06342 |
| AppNL-F KI | App | KM670/671NL (Swedish), | Saito et al., Nat. Neurosci., 2014 | RBRC06343 |
| I716F (Iberian) | ||||
| AppNL-G-F KI | App | KM670/671NL (Swedish), | Saito et al., Nat. Neurosci., 2014 | RBRC06344 |
| E693G (Arctic), I717F (Iberian) | ||||
| AppG-F KI | App | E693G (Arctic), I717F (Iberian) | Watamura et al., Sci.Adv., 2022 | Coming soon |
| ApphuAβ KI | AApp | humanize the Aβ sequence | Watamura et al., Sci.Adv., 2022 | Coming soon |
| Psen1P117L/WT KI | Psen1 | P117L | Sato et al., J. Biol. Chem., 2021 | - |
| AppNL-F/Psen1P117L/WT double KI | App, Psen1 | App: KM670/671NL, I716F | Sato et al., J. Biol. Chem., 2021 | RBRC11518 |
| Psen1: P117L | ||||
| MAPT KI | Mapt | human MAPT sequence | Hashimoto et al., Nat. Commun., 2019 Saito et al., J. Biol. Chem., 2019 | RBRC9995 |
| AppNL /MAPT double KI | App, Mapt | App: Swedish | Hashimoto et al., Nat. Commun., 2019 Saito et al., J. Biol. Chem., 2019 | RBRC10041 |
| MAPT: human MAPT sequence | ||||
| AppNL-F /MAPT double KI | App, Mapt | App: Swedish, Iberian | Hashimoto et al., Nat. Commun., 2019 Saito et al., J. Biol. Chem., 2019 | RBRC10042 |
| MAPT: human MAPT sequence | ||||
| AppNL-G-F /MAPT double KI | App, Mapt | App: Swedish, Arctic, Iberian | Hashimoto et al., Nat. Commun., 2019 Saito et al., J. Biol. Chem., 2019 | RBRC10043 |
| MAPT: human MAPT sequence |
The term “RBRC No.” refers to the registration number assigned by the RIKEN BioResource Research Center (BRC). Strains labeled with an RBRC No. are available from the BRC. If a strain is indicated as “coming soon”, it means that it is currently being prepared for supply by the BRC.
Tau is a microtubule-binding protein primarily located in neuronal cell axons. Its physiological functions include stabilizing microtubules, bundling microtubules together, regulating axonal transport, and controlling neurite height [75]. Recent advancements in cryo-EM analysis and single-molecule imaging have provided insights into the various physiological functions of tau. These studies have revealed that tau acts as a patch to stabilize the heterodimer of tubulin, the building block of microtubule [76]. Tau also forms oligomers on tubulin, acting as a gate to control the transfer of molecules onto microtubules [77]. Tau regulates microtubule dynamics and plays a crucial role in cellular processes. However, under certain conditions, tau can undergo aggregation, leading to the formation of pathological tau aggregates and exerting toxicity. The exact triggers for tau aggregation and the mechanisms underlying its toxicity are still being actively explored.
TauopathiesTau pathology is observed not only in AD but also in other diseases, such as chronic traumatic encephalopathy, progressive supranuclear palsy (PSP), and corticobasal degeneration (CBD). These diseases are collectively known as tauopathies [78, 79]. Several familial mutations in the MAPT gene have been identified in frontotemporal dementia and parkinsonism linked to chromosome 17 (FTDP-17) [80]. Additionally, primary age-related tauopathy (PART), which is not neurodegenerative or symptomatic, has been discovered in recent years [81]. Tau isoforms accumulate, and the specific pathological features observed vary depending on the disease. The adult human brain expresses six splicing variants of tau, including three variants on the N-terminal side (0N, 1N, and 2N) and two variants of the microtubule-binding domain (3R and 4R), resulting in a total of six variants. CBD and PSP exhibit aggregates of 4R tau isoforms, whereas Pick’s disease (PiD) exhibits aggregates of 3R isoforms known as pick bodies. In AD, NFTs consisting of both 3R and 4R isoforms can be observed [78, 79].
In recent years, the structure of tau filaments derived from patients with tauopathies has been analyzed using cryo-EM [82, 83]. Researchers have reported differences in the filament structure between CBD, which consists of 4R tau, and AD, which consists of both 3R and 4R tau. In AD, the core of the filaments is composed of two identical protofilaments comprising residues 306–378 of the tau protein, which adopt a combined cross-β/β-helix structure [84]. On the contrary, in CBD, unlike AD, PiD, and CTE, the core of CBD filaments is composed of residues lysine 274 to glutamate 380 of tau, and it adopts a four-layer structure [85]. In 2021, groups led by Michel Goedert and Masato Hasegawa jointly analyzed the structures of protofilaments derived from various tauopathies [86]. They discovered that CBD and PSP, despite being pathologies associated with 4R tau, have different structures. They also found that AD has the same structure as PART and familial British dementia. These findings could potentially lead to the classification of tauopathies based on filament structures and further elucidate the mechanisms of filament formation.
Although different patterns of propagation have also been observed, in AD, tau pathology is believed to propagate from the locus ceruleus and the entorhinal area to the limbic system and the medial temporal lobe before spreading throughout the brain [87]. Indeed, since the initial reports of tau propagation around 2009, vigorous research has been focusing on the cell-to-cell transmission of tau pathology [88, 89]. Understanding the mechanisms of tau transmission between cells may provide insights for the development of targeted therapeutic drugs. Studies have reported that tau can be transmitted between cells through axons [90,91,92,93], and microglia may play a role in tau transmission [94,95,96]. In addition, viral proteins such as those associated with SARS-CoV-2 have been implicated in tau transmission [28]. Another important aspect is that amyloid pathology acts as a scaffold and accelerates the spread of tau pathology [97, 98]. Although tau pathology exists before the development of amyloid pathology, the presence of amyloid pathology can facilitate the rapid propagation of tau pathology.
MAPT knock-in mouseIn tau, most of the familial mutations have been found in FTDP-17. Transgenic mice carrying these familial mutations in the MAPT gene have been widely used as models for tau pathology. In these models, tau isoforms with or without familial mutations are overexpressed using artificial promoters. Specifically, rTg4510 and PS19 mouse models are frequently used and are characterized by severe tau pathology and associated neurodegeneration [99, 100].
Wild-type adult mice express only 4R tau isoforms. In addition to the endogenous tau, most tau transgenic mice overexpress a single isoform. However, humans express six tau isoforms, depending on the variant of the N-terminal region and repeat domain. Therefore, models that express all six isoforms without artificial promoter-driven overexpression can be considered ideal for studying tau pathology.
To address the limitations of tau transgenic mice, we developed MAPT knock-in mice in which the coding region of the MAPT gene was replaced with the human sequence (Table 1) [101, 102]. These MAPT knock-in mice express all six isoforms of human tau without any mutations. Although MAPT knock-in mice alone do not exhibit tau pathology because of the lack of familial mutations, we investigated whether crossing them with App knock-in mice accelerates tau pathology [101]. Our findings showed enhanced tau phosphorylation, as recognized by various phosphorylated antibodies, when MAPT knock-in mice were crossed with App knock-in mice. In addition, the formation of dystrophic neurites positive for the AT8 marker was observed around amyloid plaques. However, we observed no NFTs as seen in the brains of individuals with AD in the App and MAPT double knock-in mice. These results indicate that factors other than amyloid pathology are required to induce tau pathology and neurodegeneration.
As stated earlier, amyloid pathology accelerates the transmission of tau pathology. Therefore, we conducted experiments involving the injection of tau seeds into the brains of wild-type, AppNL-G-F, MAPT knock-in, and double knock-in mice [101]. Consistent with a previous report [97], tau transmission was accelerated in App knock-in mice compared with wild-type mice. Furthermore, we observed further acceleration of tau propagation when the tau isoforms were humanized, indicating that the human-like expression pattern of tau is also important for tau propagation.
The MAPT knock-in mouse, which does not exhibit tau pathology, is an effective tool for studying changes in tau under amyloid pathology. However, to investigate tau pathology, we are developing additional tauopathy model strains based on the MAPT knock-in mice by introducing FTDP-17 mutations using base editing techniques. These mutant MAPT knock-in mice should serve as valuable tools for elucidating the molecular mechanisms underlying tau pathology formation and its progression.
Furthermore, the Model AD Consortium at the Jackson Laboratory (https://www.model-ad.org/) has been actively developing various mouse models that serve as valuable research tools for AD. They have released App knock-in mice and mutant tau knock-in mice as early-onset AD models. In addition, knock-in mice carrying variants of risk genes associated with late-onset AD have also been released. As described earlier, numerous knock-in mouse models are being actively developed. Obtaining more accurate and reproducible research results using these various knock-in models is important.
Given that the crossing of App knock-in and MAPT knock-in without FTDP17 mutations did not result in tau pathology or neurodegeneration, more factors may be necessary to induce tau pathology and cell death in the presence of amyloid pathology. We hypothesized that proteins bound to tau might be crucial for the formation of tau pathology, and we conducted comprehensive investigations of tau binding proteins using immunoprecipitation-mass spectrometry analyses. Among the identified proteins, we focused on a protein called carboxy-terminal PDZ ligand of nitric oxide synthase (nNOS) (CAPON) [102].
CAPON is a partner protein of nNOS and is believed to recruit substrates to nNOS. Moreover, genetic mutations in the CAPON gene (Nos1ap) have been linked to psychiatric disorders such as schizophrenia, bipolar disorder, and anxiety disorders. CAPON is also involved in neuronal cell death induced by NMDAR activation and spine formation, suggesting that the interaction between CAPON and nNOS regulates neuronal activity through nitric oxide (NO) [103, 104]. Regarding its relevance to AD, Hashimoto et al. [105] previously reported CAPON accumulation in hippocampal pyramidal cells in AD. We also observed CAPON accumulation in hippocampal pyramidal cells of AppNL-G-F knock-in mice, suggesting that amyloid pathology leads to CAPON accumulation, which may be involved in the formation of tau pathology and neuronal cell death.
To investigate the role of CAPON in AD, we conducted experiments involving CAPON overexpression in the brains of AppNL-G-F /MAPT double knock-in mice using AAV. This overexpression resulted in significant hippocampal atrophy accompanied by neuronal death (Fig. 1). Furthermore, CAPON overexpression induced tau hyperphosphorylation and insolubilization. Conversely, crossing CAPON knockout mice with PS19 tauopathy model mice attenuated tau pathology and neuronal cell death observed in PS19 mice. These results suggest that CAPON accumulation in the cell body under amyloid pathology plays an important role in the development of tau pathology and neurodegeneration.
Hippocampal atrophy caused by CAPON overexpression. We introduced a sequence consisting of the Nos1ap (CAPON) gene connected to the syn promoter using an AAV vector. Brain magnetic resonance images were acquired 7 days and 3 months after adeno-associated virus (AAV) administration. While GFP overexpression (control) did not result in any brain atrophy, CAPON overexpression significantly induced hippocampal atrophy 3 months after AAV injection. MAPT: microtubule associated protein tau.
A recent study indicated that the expression of Nos1ap encoding CAPON is regulated by TDP43, and TDP43 proteinopathy suppresses Nos1ap expression [106]. In addition, CAPON binds to alpha-synuclein, and this interaction promotes the development of synucleinopathy [107]. Based on these findings, CAPON may be involved in various proteinopathies, and the function of nNOS-CAPON could be inhibited or enhanced by proteinopathy. Further in-depth analysis is necessary to elucidate the role of CAPON in various proteinopathies.
Oxidative stress in ADOxidative stress crucially influences AD progression. For instance, the aggregation of metal ions with amyloid-β plaques generates reactive oxygen species (ROS), and neuroinflammation and mitochondrial disorders resulting from AD pathogenesis also contribute to ROS production. In addition, the brain exhibits high levels of ROS production because of its high oxygen consumption. Endogenous antioxidant molecules typically degrade ROS and maintain an appropriate balance, whereas excessive production or a decline in degradation capacity leads to oxidative stress. Among these endogenous antioxidant molecules, glutathione plays a significant role in the brain but decreases with aging and disease progression. Glutathione is synthesized through a two-step process involving glutamyl cysteine ligase (GCL) and glutathione synthase. GCL, composed of a catalytic subunit (GCLC) and a modifier subunit (GCLM), is a rate-limiting enzyme.
Our study revealed that GCLC expression levels were significantly reduced in App knock-in mice and postmortem AD brains [108]. To investigate the effect of GCLC decline on AD progression, we analyzed AD pathologies in neuron-specific conditional knockout mice of GCLC (GCLCfloxed CaMKII-Cre; GCLC-KO). At the age of 3 months, GCLC-KO mice exhibited pronounced activation of astrocytes and microglia. Moreover, at the age of 8 months, these mice showed significant brain atrophy caused by caspase-3-mediated neuronal cell death (Fig. 2). These findings revealed that GCLC deficiency-induced neuroinflammation exerts cytotoxic effects.
Brain atrophy in neuron-specific glutamyl cysteine ligase catalytic subunit (GCLC) knockout mouse. The neuron-specific GCLC knockout (GCLCfloxed X CamKII-Cre; GCLC-KO) mice exhibited age-dependent brain atrophy (upper panels). Furthermore, in the GCLC-KO mice, NeuN-positive neurons and activation of caspase-3 decreased (lower panels). These findings suggest that glutathione deficiency leads to brain atrophy through neuronal cell death.
We also observed an elevation of gasdermin D and E, which are factors responsible for inflammatory cell death (pyroptosis), in GCLC-KO mice. Furthermore, the deficiency of gasdermin E in GCLC-KO mice mitigated hippocampal atrophy. In addition, we noted a significant increase in complement1q (C1q), a protein involved in microglial phagocytosis of neurons, and elevated markers of DAM, a cluster of microglia closely associated with neurodegeneration. Therefore, neuroinflammation induced by amyloid pathology leads to a decrease in glutathione levels, further activating neuroinflammation and neuronal cell death. The vicious cycle of neuroinflammation and oxidative stress likely plays a crucial role in AD progression.
This report introduces knock-in mouse models and the studies conducted using them in the context of AD. Selecting the appropriate mouse model based on the research objectives and obtaining consensus among multiple model mice are crucial steps for advancing AD research.
All the knock-in models developed by RIKEN, which are presented in Table 1, are available from the RIKEN BioResource Research Center (RIKEN BRC).
All animal experiments adhered to the guidelines set forth in the “Regulations for the Animal Experiments” and the “Wako Institute Animal Experiment Handbook”. The Wako Animal Experiment Committee at RIKEN granted approval for the animal experiments, with the approval number W2021-2-020 (1). Overall, this research project was conducted in accordance with relevant laws, regulations, and institutional guidelines, as well as the Biosafety Manual at RIKEN. We are committed to upholding the highest ethical standards and ensuring the integrity and validity of the research results.
The authors declare no conflicts of interest associated with this manuscript.
The authors thank the members of the Laboratory for Proteolytic Neuroscience (RIKEN Center for Brain Science) and research resource division (RRD) (RIKEN Center for Brain Science) for their valuable suggestions and technical assistance.
