Background
Alzheimer’s disease (AD) is the most common neurodegenerative disorder and is characterized by extracellular amyloid plaque deposition, intracellular neurofibrillary tangle (NFT) development, and memory impairments [
1,
2]. Although the disease has been extensively studied over the past several decades, the exact mechanisms and pathological causes of AD remain unclear. Genetically modified mouse models that recapitulate the major features of AD pathologies are invaluable in determining the underlying disease mechanisms and evaluating new therapeutic approaches. Few transgenic mouse models, such as the triple transgenic mice, 3xTg-AD model, have been introduced to develop the concomitant manifestation of both amyloid plaques and NFT formation [
3]. Because this model showed no neuronal cell death in the hippocampus, the development of a new mouse model that fully mimics AD pathologies is still needed.
One of the most widely used models in AD research is the 5xFAD (Tg6799) model, which contains five familial AD mutations within the
APP and
PSEN1 genes. The 5xFAD model is considered to be an effective AD model because of the rapid progression of the amyloid pathology [
4]. The JNPL3 mouse model, which expresses P301L-mutant human tau, has been widely used to examine intraneuronal NFTs [
5]. To recapitulate the main features of AD pathogenesis, we herein developed a new mouse model that carries six mutations within transgenes encoding human amyloid precursor protein (APP), presenilin-1 (PSEN1), and tau. The resulting transgenic mouse model, Alzheimer’s disease-like pathology with
APP,
PSEN1, and
MAPT transgenes (ADLP
APT), exhibited Aβ accumulation, NFTs, early neuronal loss in the brain, and subsequent memory impairments. The pathological phenotypes of ADLP
APT mice, which feature both Aβ deposition and NFTs, were compared with those of ADLP
APP/PS1 (Aβ deposition, no NFTs) and ADLP
Tau (no Aβ deposition, NFTs) mice of the same genetic background. Understanding the interplay between Aβ accumulation and NFTs is imperative for elucidating the pathogenesis of AD. Thus, the ADLP
APT mouse model, which shows both robust amyloid and NFT pathologies, should be an excellent model for examining the Aβ-tau axis in vivo.
The hippocampus is known to play an important role in memory formation [
6]. Thus, understanding the pathological status of the hippocampus under AD is crucial for studying the mechanisms of AD-related memory impairments. Although some mechanisms that contribute to AD pathogenesis have been uncovered by studies on individual genes or proteins, systematic analysis of the pathological changes of the hippocampus is lacking. Mass spectrometry (MS)-based proteomics is expected to be an appropriate tool for systematic analysis [
7,
8]. Although MS-based proteomics have been limited because of the incomplete coverage of the proteome, recent technological advances allowed researchers to study comprehensively up to 10,000 proteins from a single cell line [
9]. However, this level of coverage requires extensive pre-fractionation, large samples, and several months of instrument time [
10]. In addition, the reliable MS-based quantitation under several perturbation states requires the use of biological and technical replicates, thus increasing the complexity of MS experiments. Importantly, these shortcomings can be overcome by an isobaric labeling strategy, such as the application of tandem mass tags (TMTs) [
11]. The recent expansion of multiplexing capacity up to 10 samples per MS injection has markedly increased the scope of quantitative proteomics [
12]. In TMT experiments, protein quantification is accomplished by comparing the intensities of reporter ions produced during MS/MS [
11]. Since this approach enables sensitive and precise protein quantification, many research groups have successfully used TMT-based strategies [
13‐
16].
With the aid of 10-plex TMT quantification strategy combined with high-resolution MS, we constructed a comprehensive proteome map of the newly developed mouse models. We have successfully discovered nearly 10,000 proteins and quantified 7000 proteins from the hippocampus of wild type, ADLPAPP/PS1, ADLPTau, and ADLPAPT mice. The protein abundances of ADLPAPT mice were compared with those of other single transgenic mice to discover differentially expressed proteins and characterize functional signatures of ADLPAPT mice via bioinformatics analysis. Furthermore, our network analysis could suggest the presence of interacting proteins that connect between amyloid and NFT pathologies. In conclusion, new ADLPAPT mice and their hippocampal proteome dataset may help to offer a novel insight of pathogenesis of AD in further studies targeting the concurrent molecular network of amyloid and NFT pathologies.
Methods
Experimental design
The aim of this study was to construct a mouse model of Alzheimer’s disease that carries mutant human genes and to introduce its molecular and functional characteristics. The protein expressions of newly constructed mouse models were assessed by quantitative proteomics combined with LC-MS/MS and TMT isobaric labeling. A total of 36 hippocampi samples were used in the proteomic experiments. (4 mouse types * 3 age-points * biological triplicates), which were randomly divided into four 10-plex TMT experimental sets. All samples were analyzed twice via MS. Three to twelve mice were sacrificed accordingly to the type of biochemical experiment, which includes western blotting and immunostaining. Behavioral tests of the AD model mice were performed by investigators in blind with respect to genotypes. No data were excluded.
Reagents and materials
Tandem mass tag (TMT) 10-plex isobaric reagents, bicinchoninic acid (BCA) assay kit - reducing agent compatible, tris (2-carboxyethyl) phosphine (TCEP), and LC/MS-grade solvents such as acetone, acetonitrile (ACN), and water were purchased from Thermo Fisher Scientific (Waltham, MA). Other reagents and materials were purchased from the following companies: Dithiothreitol (DTT) and urea from AMRESCO (Solon, OH), Sodium dodecyl sulfate (SDS), Trizma base from USB (Cleveland, OH) and sequencing-grade modified trypsin from Promega Corporation (Madison, WI), POROS20 R2 bead from Applied Biosystems (Foster City, CA). High-purity (>97%) mass spectrometry (MS) grade ovalbumin from Protea (Morgantown, WV), HLB OASIS column from Waters (Milford, MA). All other reagents, unless noted, were purchased from Sigma-Aldrich (St. Louis, MO).
Transgenic mice
5XFAD mice (Tg6799; Jackson Laboratory, Stock#006554) express both mutant human APP with the Swedish, Florida, and London mutations and mutant human PSEN1 with the M146 L and L286 V mutations under the murine Thy1 promoter. JNPL3 mice (TauP301L-JNPL3; Taconic, Stock#2508 homozygote) carry mutant human tau with the P301L mutation under the murine prion protein promoter. Due to the mixed genetic background of JNPL3 mice, JNPL3 mice were backcrossed with B6SJL (C57BL/6 X SJL) mice. The resulting JNPL3 mice on the B6SJL genetic background were crossed with 5XFAD mice to create a novel animal model, ADLP animal model. This carries the three human mutant genes and its corresponding mutations mentioned before. Only female mice were used for pathological characterization due to earlier signs of aggravated pathologies and memory deficit than male mice.
Immunohistochemistry (IHC)
Mice were anesthetized and perfused with 4% paraformaldehyde (PFA) solution in phosphate-buffered saline (PBS). The brain tissues were fixed with 4% PFA for 20 h at 4 °C, incubated in 30% sucrose (wt/vol) for 72 h and then frozen. The frozen brains were cut into 30 μm coronal sections using a Leica CM 1850 Cryostat. Brain slices were washed with PBS and then incubated in 70% formic acid in PBS for 20 min to perform antigen retrieval when amyloid plaques and NFTs were stained. Brain slices were permeabilized and blocked with blocking solution (0.3% Triton X-100, 5% horse serum, and 0.05% BSA solution) for 1 h at 20 °C prior to incubating it with primary antibodies overnight. Amyloid plaques were stained with the biotin-4G8 antibody (1:700, COVANCE), followed by the streptavidin-488-conjugated secondary antibody (Invitrogen). Hyperphosphorylated tau was examined by using the AT8 (1:300, Thermo Scientific) and AT180 (1:300, Thermo Scientific) antibody, which recognize the Ser202/Thr205 epitopes and Thr231 of human tau, followed by biotinylated anti-mouse IgG (Vector Laboratories) and streptavidin-594-conjugated secondary antibody (Invitrogen). To visualize astrocytes and microglia, anti-GFAP (1:1000, Invitrogen) for astrocytes and anti-Iba-1 (1:500, Wako) antibodies were used respectively. Hippocampal neurons of the CA1 layer were visualized by staining with anti-NeuN (1:1000, Millipore) antibody. Stained brain slices were incubated with goat anti-rat Alexa 488, donkey anti-rabbit Alexa 488, and donkey anti-mouse Alexa 647 antibody (1:500, Life Technologies) for 1 h at 20 °C. Images were obtained using LSM 700 (Carl Zeiss). At least six serial sections of each sample were imaged to consider the volume of cells in brain slices. When the number of neuronal cells in CA1 layer stained by anti-NeuN antibody were counted, one middle region of the hippocampus tissue was imaged to avoid an overlap of CA1 pyramidal neurons. All images were quantified using ImageJ software (NIH).
Sarkosyl-insoluble tau fractionation
One side of the hippocampus was homogenized in 8 volumes of Tris buffer solution (TBS) including phosphatase inhibitors and protease by tissue grinder [TBS solution; 25 mM Tris/HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, phenylmethylsulfonyl fluoride (PMSF), protease 1, and 2]. The homogenates were centrifuged at 14,000 x g at 4 °C for 15 min. The supernatant was collected for further fractionation. The supernatant was incubated with N-Lauroylsarcosine sodium salt solution 20% solution (1% final concentration) at 37 °C rotator for 1 h and then ultracentrifuged at 150,000 x g at 25 °C for 1 h. The resulting sarkosyl-insoluble pellets were resuspended in TBS solution for washing pellets. The mixture was concentrated by ultracentrifuge again at 150,000 x g at 25 °C for 1 h. The resulting pellets containing tau aggregates were suspended with 5xSample buffer (Serva Blue G) and heated at 70 °C for 10 min to prevent further aggregation.
Behavioral test
For Y-maze test, after introduction to the middle of the maze, the mouse was allowed to freely explore new environments for 8 min. Spatial memory function was measured as the percent of spontaneous alteration [
17]. The number of total arm entries and the sequence of the Y-maze arm into which mice entered were recorded in order to calculate the percentage of spontaneous alteration. The number of alternations was counted when the mouse entered into the three different maze arms consecutively. The percent of spontaneous alteration was calculated as the number of alterations divided by the total entry number multiplied by 100.
RT-PCR analysis
To verify the amount of mouse or human tau mRNA levels in ADLPTau and ADLPAPT mice, RT-PCR analysis was carried out with 10-month-old ADLP model mice. Total RNA was extracted from the hippocampus with the RNeasy Mini kit (QIAGEN). All RNA samples were converted into cDNA using Maxime RT PreMix Kit (iNtRON BIOTECH). Quantitative RT-PCR was carried out in triplicates using KAPA SYBR FAST ABI Prism qPCR kit (KAPA Biosystems). For mouse Tau, the primers 5’-AGCCCTAAGACTCCTCCA-3′ and 5’-TGCTGTAGCCGCTTCGTTCT-3′ were used. Human tau was amplified with the primers 5’-CTCCAAAATCAGGGGATCGC-3′ and 5’-CCTTGCTCAGGTC AACTGGT-3′. The mRNA levels of mouse Tau and human Tau were normalized with GAPDH which was amplified with the primers 5’-GGCCTTGACTGTGCCGTTGAATTT-3′ and 5’-ACAGCCGCATC TTCTTGTGCAGTG-3′. Once the reaction was completed, the RT-PCR products were evaluated/analyzed via gel-electrophoresis to measure/calculate the specificity of human tau primers.
Western blot analysis
Western blot analysis was used to confirm tau aggregates from sarkosyl-insoluble tau fractionation and the validation of proteomic analysis results. After isolated tau aggregates were heated at 70 °C for 10 min to prevent further tau protein aggregation, the same volume of sarkosyl-insoluble tau aggregate samples was loaded per lane of 4–12% Bis-Tris polyacrylamide precast gels (NuPAGE system, Invitrogen). Following electrophoresis, proteins were transferred to a PVDF membrane. Membranes were blocked with 5% skim milk solution and then incubated with primary antibodies against human tau (Tau13, Abcam, 1:1000) and total tau (endogenous tau and human tau) (Tau5, Abcam, 1:1000). Primary antibodies against ABCA1 (Abcam, 1:500), Ptprc (CD45, Abcam, 1:1000) and Hcls1 (HS1, CST, 1:1000) were used for the validation of proteomic analysis. For confirmation of kinase expression levels, CK1δ (Abcam, 1:5000), RSK1 (CST, 1:2000) and GSK-3β (CST, 1:2000) were used as primary antibodies. Anti-mouse or rabbit IgG conjugated HRP was used to detect primary antibodies and West Save Gold (Ab frontier) was used for their visualization. Since sarkosyl-insoluble fractionation only extracts protein aggregates, certain proteins generally used for normalization such as GAPDH and β-actin were not detected in the sarkosyl-insoluble pellets. Thus, total antibody signals of tau aggregates or each signal for distinct sizes of tau aggregates were quantified for quantification.
Mass spectrometry-based proteomics
Mouse hippocampus tissues were resected and subjected to the previously described sample preparation methods with some modifications [
18‐
20]. Detailed procedures including protein digestion, peptide labeling, fractionation, and MS analysis are described in Additional file
1: Supplementary Methods.
Quantification of protein abundance and statistics
The quantification and statistical processing methods described below are related to proteomic data. First of all, among the 9814 identified proteins, only 6964 proteins satisfying the following criteria were used for subsequent quantitative analysis; identification in all channels (7022 proteins), high protein confidence (6970 proteins, assessed by Proteome Discoverer), and possessing 1 or more unique peptides (6964 proteins). The protein abundance ratio of individual samples to pooled sample (named “normalized protein abundance”) was generated by dividing the reporter ion intensity of each channel by the intensity of the pooled sample channel in its corresponding experimental TMT set (Additional file
2: Figure S3B). There was no significant difference between the values of the pooled sample channels (Additional file
2: Figure S5E). Thus, the denominators were considered to be common and eliminated. The fold-change values used in the bioinformatics analysis were generated by dividing the normalized protein abundance of each transgenic mouse by the value of the age-matched wild type mouse. The distribution of ratiometric data was almost normal (Additional file
2: Figure S5G) but this was not thoroughly tested.
Statistical processes for the proteomic data were performed based on the normalized protein abundance using Perseus [
21]. Initially, total identified proteins were filtered based on the 6964 proteins that quantified in all mouse samples. The statistical cut-off value for significance was set to
p-value <0.05 for the Student’s t-test, while Benjamini-Hochberg FDR adjusted p-value cut-off [
22] of 0.05 was applied for the ANOVA test. The normalized protein abundances were subjected to z-normalization followed by hierarchical clustering. The statistical tests for the other biochemical experiments were described in each figure legend.
The Gene Ontology (GO) of the proteins was classified using DAVID bioinformatics tool (version 6.8) [
23]. The GO classification was evaluated by Fisher’s exact test to obtain a set of
P-values, which were then filtered at a cut-off value of 0.01. Canonical pathways, downstream biological functions, and upstream regulators were enriched using Ingenuity Pathway Analysis (IPA, QIAGEN) [
24]. The analytical algorithms embedded in IPA uses input protein list (here differentially expressed proteins) to predict putative upstream regulators such as transcription factors and growth factors, as well as downstream effects on known biological pathways. IPA derives these protein set-pathway (or regulator) relationships from their own large-scale causal network database, named Ingenuity Knowledge Base. Because the algorithm cannot determine with certainty which causalities in its database can explain our experimental results, the tool performs statistical tests (i.e. Fisher’s exact test) to assess the reliability of predicted upstream genes and pathways. Finally, IPA also assigns activation states (activated or inhibited) to putative regulators or pathways based on the quantitative values of protein members. The user will be given a confidence in the
P-value obtained from the Fisher’s exact test and the magnitude of activation as a Z-score, respectively. In this study, the
P-value cut-off criteria for the enrichment was 0.01 for Fig.
6a and b and the predictive activation Z-score cut-off was 1. For the Fig.
4b, the cut-off value was 0.05 and the predictive activation Z-score cut-off was 1. The initial pathway diagrams were obtained by IPA but were manually modified. Protein-protein interactions (PPIs) for the network analysis was interrogated from STRING database (
http://www.string-db.org) [
25]. The PPIs in network model were visualized using Cytoscape [
26].
Discussion
The amyloid cascade hypothesis claims that Aβ is the initial trigger for further pathological changes, including tau hyper-phosphorylation and NFT formation, which accelerate AD progression [
51]. To investigate the prominent role of Aβ in causing AD pathogenesis, we developed ADLP
APT mice, a novel animal model of AD that develops more robust amyloid and NFT pathologies than do other available animal models. This model showed no aggravation of the amyloid pathology beyond that seen in ADLP
APP/PS1 mice, indicating that the development of the NFT pathology does not affect the amyloid pathology. However, based on sarkosyl-insoluble tau fractionation for the aggregated form of tau and immunostaining for phosphorylated tau, the accelerated NFT pathology appeared in the hippocampus of ADLP
APT mice starting from 7 months of age in the absence of any observable difference in the transcription/expression of the human
MAPT gene. Abnormal tau phosphorylation is a crucial event that triggers tau aggregation in AD brains [
52]. Various kinases have been suggested to be involved in tau phosphorylation process. Among external stimuli to activate kinases, the role of Aβ for activating tau kinases has been reported in a downstream of Aβ toxicity [
53]. In order to investigate whether increased kinase expression levels are causative factors for accelerating NFT pathology, we have confirmed the expression levels of three kinases (
Csnk1d,
Gsk3-beta and
Rps6ka1) involved in tau phosphorylation by western blot analysis [
54‐
56]. Notably, no differences were observed in the kinase levels between wild-type and ADLP
APT mice with the disease progression. In addition to western blot analysis, the MS results also indicated no changes in kinase expression levels. As for the kinases not checked by western blot analysis (e.g.
Cdk5,
Src,
Fyn and
Abl1), they were not changed among all groups, as confirmed by our hippocampal proteome data. These results support that accelerated NFT pathology in ADLP
APT mice hippocampus is not accompanied by changes of tau kinases (Additional file
2: Figure S7). Previous studies have hypothesized that Aβ is the initiator of the NFT pathology [
53,
57,
58]. Indeed, Aβ affects tau kinase activity or localization resulting in tau toxicity in neurons. Moreover, the amyloid pathology in the cortex accelerates tau propagation process throughout the entorhinal cortex and aggravates the NFT pathology [
57]. Because the underlying molecular mechanisms of the pathophysiological interaction between Aβ and tau are unclear, ADLP
APT mice may enable further exploration of the Aβ-tau axis hypothesis on the progression of AD pathogenesis.
To further investigate the biological functions altered in the hippocampus of ADLP
APT mice, we first constructed a hippocampal proteomic database of ADLP animal models using LC-MS. Notably, as the disease progressed, the protein expression profiles of ADLP
Tau mice resembled those of wild type mice, indicating that the relatively slow NFT pathology had a small impact on the proteomic changes during the early stage of the disease. In contrast, ADLP
APT mice showed dramatic proteome-wide changes with a pattern similar to that of ADLP
APP/PS1 mice. One of the main explanations for the high similarity between ADLP
APP/PS1 and ADLP
APT mice is the presence of inflammatory responses in the hippocampus. Almost all immune-related pathways exhibited similar changes over time in ADLP
APP/PS1 and ADLP
APT mice, indicating that Aβ is one of the main factors for active inflammatory responses in the hippocampus (Fig.
4b). In both mice, the immune response categories exhibited changes at 4 months of age, even before the first deposition of amyloid plaques in the hippocampus. Immunohistochemical analysis confirmed similarities in the inflammation status between ADLP
APP/PS1 and ADLP
APT mice, as evidenced by the staining with astrocytes and microglia markers in the hippocampus (Additional file
2: Figure S2). These findings indicate that the immune responses in the hippocampus were mainly influenced by the amyloid pathology regardless of the NFT pathology.
Moreover, GO analysis of the gradually up-regulated DEPs in ADLP
APT mice identified many glial cell type-related categories, such as phagocytosis, cell chemotaxis, and macrophage infiltration (cluster 5, Fig.
5a). One of the main dysregulated pathways in ADLP
APT mice was the leukocyte extravasation signaling pathway. Various adhesion-related molecules (
Icam-1,
Vcam-1,
Itgb-1/2, and
Cd44) were up-regulated in the hippocampus of ADLP
APT mice (Fig.
6a and b), indicating that AD pathologies can modulate endothelial cells in the brain parenchyma and recruit peripheral immune [
59]. Both amyloid plaques and neurofibrillary tangles can activate inflammatory responses, thereby inducing the secretion of cytokines from microglia and astrocytes. Moreover, these serial pathologies also affect endothelial cells and disrupt the blood–brain barrier (BBB) integrity in AD [
60]. A disrupted BBB with increased inflammatory cytokines finally results in the recruitment of peripheral immune cells toward brain parenchyma in AD. Several previous reports have suggested communication between peripheral immune cells and brain parenchyma as a pathological phenotype of AD, even if their functional roles remain unclear. Laurent C et al. have reported that CD8-positive T cells infiltrate the hippocampus. They have also demonstrated that the infiltration of CD8-positive lymphocytes in the cortical region of patients with frontotemporal dementia who develop tauopathy resulted from P301L mutation within the
tau gene [
61]. Previously, using two-photon microscopy, we have observed that neutrophils can extravasate from blood vessels into the brain parenchyma where amyloid plaques are present. In the brain parenchyma, the neutrophils adhere to amyloid plaques and phagocyte them [
62]. Moreover, Zenaro E et al. have reported that the number of neutrophils adhere both inside blood vessels and in the brain parenchyma of patients with AD [
63]. Depleting temporarily neutrophils at the early stage of disease progression in a transgenic mouse model of AD prevented memory impairments, suggesting that the infiltration of neutrophils contributes to the pathogenesis of AD. The up-regulation of the proteins involved in the leukocyte extravasation signaling in ADLP
APT mice suggested that peripheral immune cells may interact with capillary endothelial cells for infiltration into the brain parenchyma.
When glial or peripheral immune cells detect extracellular amyloid plaques, they clear the amyloid deposits through phagocytosis [
64]. In addition to extracellular amyloid plaques, when neurons host NFTs in the cytosol, they activate the degradation system including the autophagy-lysosome system to eliminate them [
65,
66]. Consistent with these, the GO analysis revealed that the gradually upregulated DEPs
APT included many components of the degradation system (cluster 5, Fig.
5c). In addition, constituents of the endosome GO category were upregulated, demonstrating that the endocytic pathway via endocytosis was also activated in ADLP
APT mice. Thus, although we could not clarify which cell types exhibited these protein changes, the results of the GO analysis of upregulated proteins collectively suggested that AD pathologies trigger a sequential response that moves from phagocytosis to degradation via the endocytic pathway.
Several nervous system signaling pathways appear to differ between ADLP
APP/PS1 and ADLP
APT mice. The two canonical pathways, cAMP response element-binding protein (CREB) signaling in neurons and synaptic LTP, were inactivated only in the 10-month-old ADLP
APT mice (Fig.
6c). It is established that CREB signaling contributes to cognitive functions by modulating synaptic plasticity [
67]. Proteins associated with LTP processes are increased by CREB signaling in the hippocampus; in AD, alterations in Ca
2+ signaling lead to decreased CREB signaling and altered LTP [
68]. Similarly, our ADLP
APT mice showed downregulated CREB signaling and LTP compared with ADLP
Tau and ADLP
APP/PS1 mice. Thus, these two pathways are predicted to be disturbed simultaneously by both Aβ accumulation and NFTs. This suggests that ADLP
APT mice have pathological symptoms similar to those of patients with AD [
69].
To investigate the putative mechanisms underlying the molecular pathogenesis of ADLP
APT mice, we sorted exclusive DEPs in ADLP
APT to identify unique molecular alterations. The protein–protein interaction map of exclusive DEPs
APT showed their direct or indirect interactions with each other and categories that each protein belongs to. We further investigated the molecular interactions of exclusive DEPs
APT with
App and
Mapt (Fig.
7c). The generated
App–
Mapt network contained 15 components that have been reported to interact among each other, although little is known about their relevance to the Aβ-tau axis. For example,
Ptprc is a well-known microglia marker that has a strong association with Aβ oligomerization [
70]. However, to our knowledge, the regulation of this protein in the hippocampus of AD mouse models is not well studied. In contrast,
Abca1, a risk factor for AD, has been reported to play a significant role in Aβ clearance [
71]. In addition,
Hcls1 is a leukocyte-specific actin-binding protein involved in immune response mechanisms [
72]. We found that these proteins are involved in immune responses in the hippocampus, forming an interaction bridge on our
App-
Mapt network. Furthermore, western blot analysis clearly confirmed the regulation of these proteins in ADLP
APT mice. Since our
App-
Mapt protein network provides information on DEPs affected by both Aβ and tau, it may help researchers to identify novel molecular mechanisms in the Aβ-tau axis.
To verify whether ADLP
APT mice and its proteome data could be applied to further AD researches, we performed a comparative analysis with existing AD proteome datasets. First, the proteome data of ADLP
APT mice were compared with that of human AD brain generated by Seyfried and Levey [
50]. As a result, approximately 92% of the identified proteins were overlapped with the human AD proteome and approximately 30% of significantly regulated proteins in human AD data showed identical expression patterns with those of ADLP
APT mice (
p-value <0.05 in both studies, respectively). We also assessed the commonality between our mouse model and the previously established 3xTg AD mouse model [
73‐
75]. Notably, the comparative analysis indicated that our DEPs were largely different from those of other studies of 3xTg AD mice (data not shown). Most of our DEPs were neither detected nor showed consistent expression patterns in other studies. This may be attributed to the fact that the results of other studies were generated from 2-DIGE analysis [
73] or the vesicular proteome isolated from the forebrain [
74]. To the best of our knowledge, this study is the first to perform proteomic analysis in hippocampi developing both amyloid and NFT pathologies. These results indicate that the pathological relevance of our mouse model for AD research is valid at the protein level and in severe hippocampal pathologies.