Endoplasmic reticulum stress responses in mouse models of Alzheimer's disease: Overexpression paradigm versus knockin paradigm

Endoplasmic reticulum (ER) stress is believed to play an important role in the etiology of Alzheimer's disease (AD). The accumulation of misfolded proteins and perturbation of intracellular calcium homeostasis are thought to underlie the induction of ER stress, resulting in neuronal dysfunction and cell death. Several reports have shown an increased ER stress response in amyloid precursor protein (APP) and presenilin1 (PS1) double-transgenic (Tg) AD mouse models. However, whether the ER stress observed in these mouse models is actually caused by AD pathology remains unclear. APP and PS1 contain one and nine transmembrane domains, respectively, for which it has been postulated that overexpressed membrane proteins can become wedged in a misfolded configuration in ER membranes, thereby inducing nonspecific ER stress. Here, we used an App-knockin (KI) AD mouse model that accumulates amyloid-β (Aβ) peptide without overexpressing APP to investigate whether the ER stress response is heightened because of Aβ pathology. Thorough examinations indicated that no ER stress responses arose in App-KI or single APP-Tg mice. These results suggest that PS1 overexpression or mutation induced a nonspecific ER stress response that was independent of Aβ pathology in the double-Tg mice. Moreover, we observed no ER stress in a mouse model of tauopathy (P301S-Tau-Tg mice) at various ages, suggesting that ER stress is also not essential in tau pathology–induced neurodegeneration. We conclude that the role of ER stress in AD pathogenesis needs to be carefully addressed in future studies.

For around two decades, APP-and/or PS1-overexpressing transgenic (Tg) mice have been used widely as AD mouse models for basic and clinical studies. However, the underlying processes of A␤ overproduction in conventional mouse models differ greatly from that in AD patients. APP overexpression in animal models overproduces in an unphysiological manner fragments other than A␤ such as soluble amyloid precursor protein, C-terminal fragment of APP, and APP intracellular domain. Moreover, APP and/or PS1 overexpression can induce an artificial endoplasmic reticulum (ER) stress response due to increased cytoplasmic calcium concentrations (2). To overcome these drawbacks of the overexpression paradigm, we recently developed mouse models utilizing an App-knockin (KI) strategy. The App-KI mice, which express humanized A␤ with familial AD mutations at endogenous levels, exhibit ADassociated pathologies, including pronounced A␤ amyloidosis and gliosis (3,4). In contrast, App-KI failed to reproduce some of the observations made using conventional mouse models (3)(4)(5). For example, the early lethality of calpastatin-knockout (KO) ϫ APP23 mice, which contradicted the chronic nature of AD, was not reproduced in calpastatin-KO ϫ App-KI (3,6). Moreover, with App-KI mice, we detected no calpain-dependent conversion of p35 to p25, which up-regulates cyclin-dependent enzyme 5 (CDK5) activity. Although calpain activation is generally considered to play an important role in AD progression due to its involvement in caspase-dependent neuronal cell death and CDK5-mediated hyperphosphorylation of tau, our observations indicate that the role of calpain may have been overestimated.
In this study, we focused our attention on ER stress. The accumulation of unfolded/misfolded proteins within the ER lumen along with the disruption of calcium homeostasis leads to ER dysfunction, known as ER stress. Under ER stress conditions, cells escape from serious damage by activating adaptive response pathways known as the unfolded protein response (UPR). UPR restores proteostasis in the ER by arresting protein synthesis, degrading unfolded/misfolded proteins, and increasing molecular chaperone concentrations. Conversely, UPR induces cell death signaling upon prolonged stress or serious damage. Several reports have suggested that ER stress induced by A␤ accumulation is involved in neurodegeneration in AD (7)(8)(9). To this end, exposure of hippocampal brain slices, primary neurons, or cell lines to oligomerized or fibrilized A␤ has been shown to induce ER stress (10,11). Moreover, UPR upregulation has been detected in several AD mouse models such as APP/PS1, 5XFAD, and 3XTg-AD (10 -12). However, until the present time, it has been difficult to clarify whether ER stress is triggered by A␤ pathology in vivo. To answer this important question, i.e. which abnormally overexpressed membrane proteins or A␤ deposition triggers ER stress, we evaluated the ER stress response in several AD mouse models, including App-KI.

Results
UPR regulates three key pathways via three ER-binding proteins (13): pancreatic ER kinase (PERK), activating transcription factor 6 (ATF6), and inositol-requiring enzyme 1 (IRE1). The first pathway, triggered by PERK phosphorylation, arrests protein synthesis via abrogating the activity of eukaryotic translation initiation factor 2␣ (eIF2␣) by phosphorylation and activates ATF4 -mediated gene expression of ER chaperones. The second pathway, initiated by ATF6, induces the expression of ER molecular chaperones such as GRP78/BiP and GRP94 and protein-folding enzymes such as protein-disulfide isomerases (PDIs) to prevent protein misfolding. In the third pathway, phosphorylated IRE1 induces the expression of genes related to protein folding, autophagy, and apoptosis (such as C/EBP-homologous protein (CHOP)) by activating the transcription factor XBP1. Under normal conditions, PERK, ATF6, and IRE1 remain in an inactive state due to GRP78 binding. In response to ER stress, however, misfolded proteins interrupt GRP78 and sensor protein interactions, thereby initiating UPR signaling.
Several reports have described activation of the ER stress response in AD mouse models. For instance, in the 5XFAD model, which overexpresses familial AD-linked APP and PS1 mutants, phosphorylated eIF2␣ (p-eIF2␣) and XBP1 mRNA levels are elevated (14,15). The APP/PS1 mouse shows age-dependent increases of GRP78, p-PERK, p-eIF2␣, and CHOP (12). Moreover, increased GRP78 is also detected in the 3XTg-AD mouse, which expresses mutant APP, PS1, and tau (16). However, in contrast to these findings, Lee et al. (17) observed no UPR signals in Tg2576 mice. Accordingly, it remains controversial whether A␤ pathology is an essential trigger of ER stress.
GRP78 acts as an important sensor of ER stress, and its expression is up-regulated by UPR to prevent protein misfolding. GRP78 also appears to be the most sensitive and earliest ER stress marker in APP/PS1-Tg mice (13). Based on this evidence, we first analyzed GRP78 as an ER stress marker. We also examined levels of several ER stress markers: CHOP, PDI, p-eIF2␣, and spliced XBP1. To examine whether A␤ deposition induces ER stress, we quantified levels of ER stress markers in the cortices of young and older App NL-G-F mice ( Fig. 1, a, b, and c). In App NL-G-F mice, A␤ accumulation begins at 2 months and occupies the entire cortex and hippocampus by around 9 months (3). Western blot analysis showed no significant upregulation in any of the ER stress markers tested at 6 and 14 months in App NL-G-F mice compared with wildtype (WT), suggesting that increased A␤ deposition is not correlated with the ER stress response ( Fig. 1, a and c). Given that we detected elevation of ER stress markers except for PDI in thapsigargintreated primary cultured cortical neuronal or Neuro2A cells ( Fig. 1, a and b), our observations are not due to failure to specifically detect ER stress markers. PDI was increased in primary cortical neuronal cells even under more severe conditions (data not shown).
To compare ER stress response between App NL-G-F and APP-Tg mice, we analyzed ER stress markers using cortical and hippocampal samples ( Fig. 1, a, b, d, and e). As APP is a membrane-binding protein, we expected that APP overexpression would induce chronic ER stress. However, we observed no significant increase of ER stress markers in two APP-overexpressing mouse models: APP23 and Tg2576 (Fig. 1, d and e). These results indicate that neither A␤ deposition nor APP overexpression induces detectable ER stress. As such, our observations contradict previous reports describing the ER stress induced in double-transgenic mice overexpressing mutant APP and PS1 (10 -12). Consistent with previous reports, we detected activation of ER stress in APP/PS1 and 3XTg-AD mice. In contrast to App NL-G-F and single APP-Tg mice, APP/PS1 mice showed an increased ER stress marker, p-eIF2␣, in hippocampus (6 and 15 months) and cortex (6 months), and 3XTg-AD showed higher levels of GRP78 and CHOP in hippocampus compared with age-matched wildtype controls (Fig. 1a). These results indicate that this effect could be a nonspecific artifact caused by the genetic modification of PS1 or double modifications of APP and PS1. We confirmed overexpression of PS1(⌬E9) in APP/PS1 mice using antibodies and protocols that had been fully validated (18,19). We must, however, indicate that we did not detect other ER stress markers in the APP/PS1 mouse brains in a manner distinct from the previous report (12), presumably due to the reasons described under "Discussion." Under prolonged ER stress conditions, cells cease to protect themselves and turn on cell death signals. In AD and other neurodegenerative diseases, tau pathology correlates well with neurodegeneration (20). We therefore hypothesized that ER stress might mediate tau-induced neuronal cell death. To inves-ER stress in AD mouse models tigate this further, we analyzed ER stress markers in cortices (3-15 months) and hippocampi (12 months) of P301S-Tau-Tg mice on a C57BL/6 background (Fig. 2). In these mice, brain atrophy associated with neuronal cell death starts from around 9 -12 months 5 ; however, we observed no changes in all stress markers between 3 and 15 months (Fig. 2, a and b). These results suggest that tau pathology does not accompany ER stress and that the ER stress response is unrelated to tau-induced neurodegeneration.

Discussion
In the present study, we found an absence of ER stress responses in App-KI and single APP-overexpressing mice. We thus conclude that neither A␤ nor APP overproduction triggers ER stress. Lee et al. (17) have consistently shown that ER stress does not occur in Tg2576 mice. The elevated UPR detected in several lines of APP and PS1 double-transgenic mice is thus likely to be a nonspecific artifact. As presenilins are polytopic membrane proteins containing nine transmembrane domains, we suggest that mutant PS1 overexpression specifically impacts ER membranes in which presenilins are enriched (21, 22).
A number of studies have reported that PS1 plays a role in the regulation of ER calcium homeostasis (for reviews, see Honarnejad et al. (23) and Zhang et al. (24). PS1 modulates not only the function of sarco/endoplasmic reticulum Ca 2ϩ -ATPase, which transfers calcium from the cytosol to the lumen, but also of ER-associated calcium channels such as the inositol trisphosphate receptor and ryanodine receptor (25)(26)(27)(28)(29). In addition, familial AD (FAD)-linked mutations of PS1 alter its activity in calcium transfer (for reviews, see Honarnejad et al. (23) and Zhang et al. (24). Alteration of the ER cytosolic calcium concentration is a strong inducer of ER stress as seen in cells treated with the sarco/endoplasmic reticulum Ca 2ϩ -ATPase inhibitor thapsigargin ( Fig. 1) (30). Based on these findings, the genetic modification of PS1 is very likely to affect the ER stress response. Indeed, FAD-linked PS1 mutation results in the delayed activation of UPR in fibroblasts and primary cultured neurons of mutant PS1-KI mice (31,32). Moreover, deletion or overexpression of PS1 in primary neurons also alters the ER stress response (33,34). Taken together, the ER stress responses observed in APP/PS1 double mutant mice are not causally associated with AD etiology. Artificial ER stress responses induce artificial cellular responses and cell death. We therefore suggest that the results obtained with APP/PS1 double mutant mice should be further validated.
In this study, however, we did not detect marked activation of ER stress in APP/PS1 mice even though we utilized a strain identical to that used by Barbero-Camps et al. (12) and Jankowsky et al. (35). We presume that partial reproducibility was due to reduced expression levels of APP and PS1 in the APP/PS1 mice (Fig. 1a) after a number of passages.
In addition to the above, we detected no ER stress response in a mouse model of tauopathy, suggesting that ER stress does not contribute to tau-induced neurodegeneration. It is plausible that tau overexpression will not induce ER stress because tau is basically a cytosolic protein.
Several groups have reported that the ER stress response is up-regulated in post-mortem human AD brains (14,36,37). In contrast, Katayama et al. (31) showed a significant decrease of GRP78 in the brains of AD patients. The post-mortem degradation of mRNA and protein may be different between control and AD patients because neurons in AD brain had undergone degeneration, which would accompany destruction of lysosomes and mitochondria, before sampling. We thus need to be careful when we analyze and discuss mRNA and protein levels in post-mortem samples. In addition, because calcium concentrations and calcium-related responses might be altered by post-mortem conditions, ER stresses in post-mortem samples require careful interpretation. To this end, we have shown an unphysiological activation of the calcium-dependent protease calpain in post-mortem mouse brains (5).
Our observations raise serious concerns surrounding efforts to translate basic findings obtained using APP/PS1 genemodified mice to clinical applications. If pharmacological candidates that improve the pathological and neurological parameters of the APP/PS1 gene-modified mice exert their effects via the modification of nonspecific ER stress, then these candidates may not be effective in a preclinical setting or in clinically defined AD patients. Choosing appropriate models is thus extremely important if the mechanisms underlying AD are to be fully elucidated (4).

Animals
All animal experiments were carried out in accordance with RIKEN Brain Science Institute guidelines. We previously produced App NL-G-F/NL-G-F -knockin (App NL-G-F ) mice using genomic DNA containing introns 15-17 of mouse App with the humanized A␤ sequence into which KM670/671NL (Swedish), I716F (Iberian), or E693G (Arctic) mutations (3) had been introduced. APP23 mice (38), which overexpress Swedish mutation-containing APP751, were maintained on a C57BL/6J , which is divided by intensity of ␤-actin (for GRP78, CHOP, and PDI) or total eIF2␣ (for p-eIF2␣). As a positive control, ER stress markers in thapsigargin-treated primary cultured cortical neuronal cells or Neuro2a cells were confirmed. Arrowheads shows bands of CHOP and PS1, and the asterisk shows nonspecific bands. FL, full length; CTF, C-terminal fragment; NTF, N-terminal fragment. b, mRNA levels of unspliced/spliced XBP1 and CHOP were determined. XBP1 mRNA was detected by semiquantitative reverse transcription-PCR. Unspliced/spliced XBP1 was observed as a 152/126-bp band, respectively. c-e, expression levels of ER stress markers in cortices (c and d) and hippocampi (e) were normalized to that of ␤-actin (for GRP78, CHOP, and PDI) or total level of eIF2␣ (for p-eIF2␣) and reported as relative levels compared with expression in 6-month-old WT mice. The expression level of spliced XBP1 mRNA was divided by that of unspliced XBP1 mRNA. The positive control is thapsigargin-treated primary cultured cells. Data are shown as means Ϯ S. E. (n ϭ 3). Differences between groups were examined for statistical significance with one-way ANOVA. n.s., no significant difference. Error bars represent S.E.

Cell culture
Primary cultured cells were prepared as below. Cortices and hippocampi were separated from E16 -18 embryos of WT mice and moved to Neurobasal medium (Thermo Fisher Scientific, Waltham, MA). Tissues were chopped with scalpels and treated with 5 ml of 0.25% trypsin at 37°C for 15 min with rotation. Then 0.125 ml of 1% DNase I was added and mixed by pipetting. After centrifugation of the tissues at 1500 rpm for 3 min, 5 ml of Hanks' balanced salt solution containing 0.125 ml of 1% DNase I was added to the pellet and incubated at 37°C for 5 min while moving slightly in a water bath. Tissues were again centrifuged at 1500 rpm for 3 min, and the resulting pellets were suspended in 15 ml of Neurobasal medium containing 2% B27 and 0.5 mM glutamate. Cells were filtrated using a Falcon 2360 Cell Strainer (100-m nylon) and seeded in cell culture plates with Neurobasal medium containing B27 and glutamate. Prepared cells at DIV7 were used for experiments. Neuro2A cells were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum and incubated at 37°C in a humidified atmosphere of 5% CO 2 . To induce ER stress, cells were treated with thapsigargin (final 2 M for 8 h for primary cells; final 5 M for 18 h for Neuro2A cells).

Western blotting
Extirpated brains were immediately frozen in liquid nitrogen and stored at Ϫ80°C. The cortices were homogenized in 400 l of Tris-HCl buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 1% Triton X-100) containing a protease inhibitor mixture and a phosphatase inhibitor mixture (Sigma-Aldrich). The homogenates were centrifuged at 15,000 ϫ g for 20 min at 4°C. Resulting supernatants were used for subsequent analyses. Protein concentrations were determined using a BCA protein assay kit (Pierce). An equivalent amount of protein from each animal was mixed with 4ϫ sample buffer with 2-mercaptoethanol, separated by SDS-polyacrylamide gel electrophoresis, and transferred electrophoretically to a 0.22-m PVDF membrane (Merck Millipore). The membrane was treated with the ECL Prime blocking agent (GE Healthcare) and incubated with each primary antibody (Table 1) diluted in Tris-buffered saline containing 0.05% Tween 20 (TBST), pH 7.5, overnight at 4°C. The membrane was washed three times in TBST for 5 min and treated with horseradish peroxidase-conjugated anti-rabbit or -mouse IgG (GE Healthcare) for 1 h. Immunoreactive bands on the membrane were visualized with ECL Select (GE Healthcare) and scanned with a LAS-3000mini LuminoImage analyzer (Fujifilm, Tokyo, Japan). The Neuro2A lysates were analyzed in a similar manner at an identical protein concentration.

RNA isolation and polymerase chain reactions (PCRs)
The cortex samples were homogenized in 1 ml of RNAiso Plus total RNA extraction reagent (Takara). Neuro2A cells and primary cultured cortical neuronal cells (1 ϫ 10 7 cells/sample) were dissolved in 500 M RNAiso Plus. Total RNA from each sample was isolated according to the manufacturer's instruc-tions. To obtain complementary DNA, a reaction mixture containing 2 g of RNA and Primescript reverse transcriptase (Takara) was incubated according to the manufacturer's directions for 60 min at 42°C and then 10 min at 70°C to stop the reaction. Semiquantitative PCR was performed using KOD FX Neo (Toyobo, Osaka, Japan) for XBP1 or Ex-Taq (Takara) for CHOP and ␤-actin. PCR was conducted at 94°C for 2 min followed by 40 cycles of 98°C for 10 s, 50°C for 30 s, and 68°C for 1 min using primers 5Ј-acacgcttgggaatggacac-3Ј (sense) and 5Јccatgggaagatgttctggg-3Ј (antisense) for XBP1; 95°C for 2 min followed by 40 cycles of 95°C for 30 s, 50°C for 30 s, and 72°C for 1 min using primers 5Ј-agaggagccagggccaacagaggtcacacg-3Ј (sense) and 5Ј-tccggagagacagacaggaggtgatgccca-3Ј (antisense) for CHOP; and 95°C for 2 min followed by 40 cycles of 95°C for 30 s, 50°C for 30 s, and 72°C for 1 min using primers 5Ј-gggtcagaaggattcctatg-3Ј (sense) and 5Ј-ggtctcaaacatgatctggg-3Ј (antisense) for ␤-actin.

Statistical analysis
All data are shown as means Ϯ S.E. Differences between groups were examined for statistical significance with one-way or two-way ANOVA.