Reduction of the expression of the late-onset Alzheimer's disease (AD) risk-factor BIN1 does not affect amyloid pathology in an AD mouse model

Alzheimer's disease (AD) is pathologically characterized by the deposition of the β-amyloid (Aβ) peptide in senile plaques in the brain, leading to neuronal dysfunction and eventual decline in cognitive function. Genome-wide association studies have identified the bridging integrator 1 (BIN1) gene within the second most significant susceptibility locus for late-onset AD. BIN1 is a member of the amphiphysin family of proteins and has reported roles in the generation of membrane curvature and endocytosis. Endocytic dysfunction is a pathological feature of AD, and endocytosis of the amyloid precursor protein is an important step in its subsequent cleavage by β-secretase (BACE1). In vitro evidence implicates BIN1 in endosomal sorting of BACE1 and Aβ generation in neurons, but a role for BIN1 in this process in vivo is yet to be described. Here, using biochemical and immunohistochemistry analyses we report that a 50% global reduction of BIN1 protein levels resulting from a single Bin1 allele deletion in mice does not change BACE1 levels or localization in vivo, nor does this reduction alter the production of endogenous murine Aβ in nontransgenic mice. Furthermore, we found that reduction of BIN1 levels in the 5XFAD mouse model of amyloidosis does not alter Aβ deposition nor behavioral deficits associated with cerebral amyloid burden. Finally, a conditional BIN1 knockout in excitatory neurons did not alter BACE1, APP, C-terminal fragments derived from BACE1 cleavage of APP, or endogenous Aβ levels. These results indicate that BIN1 function does not regulate Aβ generation in vivo.

Alzheimer's disease (AD) is pathologically characterized by the deposition of the ␤-amyloid (A␤) peptide in senile plaques in the brain, leading to neuronal dysfunction and eventual decline in cognitive function. Genome-wide association studies have identified the bridging integrator 1 (BIN1) gene within the second most significant susceptibility locus for late-onset AD. BIN1 is a member of the amphiphysin family of proteins and has reported roles in the generation of membrane curvature and endocytosis. Endocytic dysfunction is a pathological feature of AD, and endocytosis of the amyloid precursor protein is an important step in its subsequent cleavage by ␤-secretase (BACE1). In vitro evidence implicates BIN1 in endosomal sorting of BACE1 and A␤ generation in neurons, but a role for BIN1 in this process in vivo is yet to be described. Here, using biochemical and immunohistochemistry analyses we report that a 50% global reduction of BIN1 protein levels resulting from a single Bin1 allele deletion in mice does not change BACE1 levels or localization in vivo, nor does this reduction alter the production of endogenous murine A␤ in nontransgenic mice. Furthermore, we found that reduction of BIN1 levels in the 5XFAD mouse model of amyloidosis does not alter A␤ deposition nor behavioral deficits associated with cerebral amyloid burden. Finally, a conditional BIN1 knockout in excitatory neurons did not alter BACE1, APP, C-terminal fragments derived from BACE1 cleavage of APP, or endogenous A␤ levels. These results indicate that BIN1 function does not regulate A␤ generation in vivo.
Overwhelming evidence places the generation and eventual deposition of the ␤-amyloid peptide in the brain parenchyma as a key pathological driver of Alzheimer's disease (AD) 2 (1). Targeting its production, which occurs through the sequential proteolysis of the amyloid precursor protein (APP) by the ␤-site APP cleaving enzyme 1 (BACE1) and ␥-secretase complex, remains an important therapeutic target in the Alzheimer's disease field. Although early onset, familial forms of AD are driven by excess A␤ accumulation as a result of mutations in the vicinity of the A␤ coding region of the APP gene or mutations in genes encoding the catalytic subunits of the ␥-secretase complex, the molecular mechanisms that contribute to aberrant A␤ production and accumulation in late-onset AD (LOAD) remain largely unknown.
Recent genome-wide association studies (GWAS) have identified a range of genetic loci that are linked to LOAD risk, implicating several molecular mechanisms through which genetic LOAD risk may be conferred. The bridging integrator 1 (BIN1) gene was identified as the second most prevalent susceptibility locus for LOAD alongside several other genes with known functions in the regulation of endocytic trafficking including SORL1, PICALM, and CD2AP (2)(3)(4). BIN1 belongs to the Bin-Amphiphysin-Rvs (BAR) family of adaptor proteins which are implicated in the production of membrane curvature and endocytosis in various cellular functions (5,6). BIN1 mRNA undergoes alternative splicing to produce multiple protein isoforms in a tissue-and cell-specific manner (7). Several BIN1 isoforms, including those expressed in the brain, contain a clathrin and AP-2 adaptor complex binding (CLAP) domain, which is involved in clathrin-mediated endocytosis (CME) (5,8). The exact consequence of the LOAD-associated SNPs identified at the BIN1 locus remains unknown. Previous studies have reported conflicting findings on BIN1 protein expression levels in AD (9 -12) and how it correlates with amyloid and tau pathology (9,10,(13)(14)(15) shown to be altered in disease and correlate with age of disease onset and disease duration (16). Endocytic dysfunction has been described in AD (17), and numerous in vitro studies highlight the importance of correct endosomal regulation in the spatial separation of APP and BACE1, which significantly impacts A␤ production. Following maturation in the constitutive secretory pathway, APP reaches the cell surface. After internalization from the surface, APP is sorted to early endosomes and can be recycled or targeted for lysosomal degradation (18,19). The acidic pH of the early endosomes provides optimal conditions for BACE1 proteolytic activity, and the majority of amyloidogenic proteolysis of APP and A␤ generation is thought to occur in this subcellular environment (20,21). Although many of these trafficking pathways have been extensively studied in immortalized cells lines, they fail to consider the additional complexities of APP and BACE1 trafficking in neurons and in vivo (22)(23)(24)(25). In vitro evidence exists for a role for BIN1 in the regulation of the membrane trafficking of BACE1 (26,27), but a role for BIN1 in this process in vivo is yet to be demonstrated. Here, we report that a 50% reduction in BIN1 expression does not alter BACE1 or APP localization or endogenous A␤ production in WT mice. Similarly, 50% reduction of BIN1 did not modify amyloid pathology and behavioral deficits in a mouse model of AD. Furthermore, neuron-specific conditional knockout (cKO) of Bin1 did not alter BACE1 or APP localization to membrane fractions generated in the mouse brain or affect A␤ production.

A 50% global loss of Bin1 does not alter APP or BACE1 expression or localization or A␤ production
The role of BIN1 in the brain remains largely unknown, but several lines of evidence indicate its expression in a range of brain cell subtypes (11,12,28). Furthermore, evidence exists for noncell-autonomous deposition of A␤ in the absence of PSEN1 expression in excitatory neurons (29). To explore the possible contribution of Bin1 expression in A␤ production, we sought to generate a model of global Bin1 knockdown. Because homozygous KO of Bin1 in mice results in perinatal lethality due to severe cardiomyopathy (30), we generated mice with germline deletion of a single Bin1 allele by crossing male Bin1 Fl/Fl :Syn-Cre mice to female C57BL/6 mice. Synapsin expression in the testes causes recombination of the floxed gene in the sperm (31) resulting in progeny with an inherited recombined floxed allele. The resulting progeny had a global single-allele deletion of Bin1 (Bin1ϩ/Ϫ), and pups were viable and had no obvious phenotypic abnormalities. Loss of a single Bin1 allele caused a 50% reduction in forebrain BIN1 levels in Bin1ϩ/Ϫ mice compared with WT littermate controls, but no significant differences were observed in the steady-state levels of APP, BACE1, or the BIN1 paralog amphiphysin 1 (Amph1) (Fig. 1A). Quantification showed a 45% reduction in high molecular weight BIN1 isoforms (BIN1:H) and 55% reduction in low molecular weight BIN1 isoforms (BIN1:L) (Fig. 1B). Prominent BACE1 localization in hippocampal mossy fibers has been widely reported, with its absence in BACE1 KO mice resulting in disorganized infrapyramidal mossy fiber bundles (32). The 50% reduction of BIN1 in Bin1ϩ/Ϫ mice did not cause any overt mislocalization of BACE1 or APP with similar BACE1 localization of BACE1 to hippocampal mossy fibers and APP to the CA3 neuronal cell layer in WT and Bin1ϩ/Ϫ mice (Fig. 1C). To assess whether the diminution in BIN1 expression affected amyloidogenic processing of APP, we quantified endogenous A␤ levels. Analysis of homogenates from 4-month-old WT and Bin1ϩ/Ϫ mice showed no significant differences in the steady-state levels of A␤40 or A␤42 (Fig. 1D).

50% reduction in BIN1 does not modify the behavior in 5XFAD mice
We performed behavioral tests in 5XFAD and 5XFAD Bin1ϩ/Ϫ mice to determine whether the partial loss of BIN1 expression influenced the previously reported cognitive dysfunctions in the 5XFAD line in the absence of a change of cerebral amyloid burden. In the elevated-plus maze, designed to test anxiety phenotypes, no significant differences between the groups were observed in the percentage of time spent in open arms of the maze in female (Fig. 3A) or male mice (Fig. S3A). Novel object recognition, Y-maze, and contextual fear conditioning paradigms were used to test cognitive deficits in 5XFAD and 5XFAD Bin1ϩ/Ϫ mice. No significant differences were observed in the discrimination index in female (Fig. 3B) or male cohorts (Fig. S3B). Similarly, in the Y-maze test, no significant difference in the number of arm entries or spontaneous alternation was observed in female (Fig. 3, C and D) or male mice (Fig. S3, C  and D). In contextual fear conditioning, response to the initial

BIN1 does not alter ␤-amyloid pathology in vivo
shock was not significantly different between genotypes in female ( Fig. 3E) or male (Fig. S3E) mice, nor was freezing behavior 24 h later upon reintroduction to the fear conditioning chamber ( Fig. 3F and Fig. S3F). Thus, the partial loss of BIN1 expression in 5XFAD mice does not influence the anxiety or memory performance in behavior tests.

Neuronal conditional knockout of BIN1 does not alter presynaptic APP and BACE1 localization or endogenous A␤ generation
Previous reports showed a role for BIN1 in A␤ generation and BACE1 trafficking in primary neuronal cultures (26,27).
However, it was also reported that the knockdown or the overexpression of BIN1 in human neuroblastoma cells had no effect on APP processing (9). To test whether the select loss of BIN1 expression in neurons influences amyloidogenic processing of APP in vivo, we used the Synapsin1-Cre driver to generate neuronal Bin1 cKO mice (Bin1 Fl/Fl :Syn-Cre). By crossing female Bin1 Fl/Fl :Syn-Cre to male Bin1 Fl/Fl mice, we avoided germline Bin1 deletion encountered in offspring generated with male Syn-Cre breeders (31). The cKO mice showed a significant reduction in BIN1 levels in the cortex and hippocampus as compared with littermate controls (Fig. 4A). Quantification showed greater than 50% reduction in both high and low molecular weight BIN1 Figure 2. Reduction of BIN1 expression does not alter amyloid deposition in female 5XFAD mice. A, representative images of A␤ deposit core staining with thioflavin S and mAb 3D6 immunostaining in 4-month-old female 5XFAD and 5XFAD Bin1ϩ/Ϫ mice at 4 months of age. B, quantification of amyloid burden identified by thioflavin S staining (n ϭ 10 per genotype). C, quantification of amyloid burden identified by mAb 3D6 immunostaining (n ϭ 9 5XFAD and 10 5XFAD Bin1ϩ/Ϫ ). Forebrain tissue from 4-month-old female 5XFAD and 5XFAD Bin1ϩ/Ϫ mice was sequentially extracted in TBS and formic acid to generate soluble and insoluble A␤ fractions. D and E, the levels of soluble and insoluble A␤40 and A␤42 were measured by V-PLEX 6E10 immunoassay (n ϭ 10 per genotype).  arms of the elevated-plus maze was quantified for 4-month-old 5XFAD and 5XFAD Bin1ϩ/Ϫ mice. B, discrimination index score from novel object recognition for 5XFAD and 5XFAD Bin1ϩ/Ϫ mice. Box plots show the median, maximum, and minimum values. C, the number of arm entries over the 8-min Y-maze trial. D, the number of spontaneous alternations in the Y-maze over the 8-min trial period. E, freezing behavior in female 5XFAD and 5XFAD Bin1ϩ/Ϫ mice during day 1 of fear conditioning. Lightning bolts represent delivery of the shock stimulus. F, freezing behavior in female 5XFAD and 5XFAD Bin1ϩ/Ϫ mice on day 2 of fear conditioning, 24 h after shock application (n ϭ 11 for both genotypes in all analyses).

BIN1 does not alter ␤-amyloid pathology in vivo
isoforms (Fig. 4B). To visualize BIN1 expression, we performed immunohistochemical staining using BIN1 mAb 13463 in the absence of epitope retrieval of tissue sections, which preferentially labels neuronal BIN1 (Fig. S4). As expected, a marked diminution of neuronal BIN1 immunostaining could be observed in the hippocampus of Syn-Cre Bin1 cKO mice (Fig. 4C).

BIN1 does not alter ␤-amyloid pathology in vivo
Synaptic activity has previously been shown to regulate A␤ production in vivo (35,36), and both APP and BACE1 have been shown to localize at the synapse (24,37,38). BIN1, APP, and BACE1 were readily detectable in non-post-synaptic density (PSD) membrane fractions isolated from the mouse brain alongside enrichment of the synaptic marker synaptophysin (Fig. 4D). Levels of BIN1 in non-PSD and PSD fractions were significantly reduced in Bin1 cKO mice compared with controls (Fig. 4E). Non-PSD fractions were analyzed by immunobloting for APP, BACE1, synaptophysin, and Amph1 (Fig. 4F). The levels of APP and BACE1 remained unaltered in non-PSD fractions from cKO mice as compared with littermate controls (Fig.  4G). Finally, we ascertained whether the processing of endogenous APP was altered by the loss of neuronal BIN1 expression in the cKO mice. To this end, we performed ELISA on forebrain homogenates from 4-month-old Bin1 Fl/Fl and Bin1 Fl/Fl :Syn-Cre mice to measure the levels of endogenous A␤40 or A␤42. No significant differences were observed in the steady-state levels of A␤40 or A␤42 (Fig. 4H).

Discussion
In this study, we investigated a role for the Alzheimer's disease risk gene BIN1, in the development of amyloid pathology in a mouse model of Alzheimer's disease. Deletion of a single Bin1 allele resulted in the loss of 50% of endogenous BIN1 protein expression but did not alter BACE1 or APP levels and their localization or endogenous A␤ levels in the brain. In addition, 50% loss of BIN1 neither altered the steady-state A␤ levels and cerebral A␤ deposition nor influenced the behavioral outcomes in a widely used mouse model of Alzheimer's disease amyloid pathogenesis. Similar to the global reduction of BIN1 in heterozygous Bin1 KO mice, pan-neuronal conditional knockout of BIN1 resulted in 50% reduction of BIN1 levels in the cortex and hippocampus but did not alter BACE1 or APP levels in pre-synaptic fractions of mouse brains or endogenous A␤ levels. These results indicate the LOAD genetic risk associated with the BIN1 loci is unlikely mediated through a direct role in amyloid production in vivo.
GWAS over the past decade have advanced the field by the identification of several risk loci for LOAD. However, defining the molecular mechanisms through which GWAS genes contribute to LOAD risk remains a daunting task. Notably, two decades after the discovery that APOE is a major genetic risk factor, novel insights on how APOE modifies molecular and cellular phenotypes relevant to AD pathophysiology continue to emerge. Nevertheless, the investigations on APOE and, more recently LOAD-associated TREM2 mutations, have given invaluable insights into how genetic factors can contribute to disease risk and reveal potential pathways for disease modification or intervention (39). Despite its being the second most prevalent susceptibility locus for LOAD, little is known about the function of BIN1 in disease pathogenesis. Recent evidence implicates a reduction in BIN1 in the increased trans-synaptic transfer of extracellular tau in vitro through a mechanism involving endocytosis (40). A physical interaction between BIN1 and tau in neuronal cytosol has also been demonstrated (13,41), although it is not clear whether this interaction in any way affects the endocytic function of BIN1.
Emerging evidence has also implicated BIN1 in the process of A␤ generation, namely through a role in endocytic trafficking of BACE1. The loss of Bin1 expression was shown in one study to increase BACE1 levels and A␤ production in neurons, with the subsequent analysis in immortalized cells lines implicating loss of endosomal trafficking and lysosomal degradation of BACE1 as the cause (26). Although we saw a small increase in BACE1 levels in the forebrain lysates of 5XFAD Bin1ϩ/Ϫ male mice, this was not the case in female mice. Nevertheless, the loss of one Bin1 allele did not cause a measurable change in A␤ deposition in males or females. A second study showed a role for BIN1 in tubule scission and BACE1 exit from early endosomes, with a loss of BIN1 resulting in increased intracellular A␤ but no change in A␤ secretion (27). A lack of an effect of BIN1 knockdown or overexpression on APP processing and A␤ secretion was also reported in a study that used human neuroblastoma cells (9). In our study, we analyzed the pathology in 5XFAD mice at 4 months of age in both males and females to quantify pathogenesis before the brain gets overloaded with amyloid burden. It would be interesting to determine whether the loss of BIN1 expression influences BACE1 levels in older animals.
In addition to BIN1, other genes including SORL1, PICALM, and CD2AP have implicated endosomal dysfunction in AD. Indeed, studies have shown a role for PICALM in CME of ␥-secretase and subsequent A␤ production (42) and SORLA in the intracellular trafficking and processing of APP (43). Although alterations in SORLA levels translated into changes in A␤ in both WT mice and mouse models of AD amyloidosis (44 -46), our results using heterozygous KO and cKO models do not validate a similar translation from in vitro to in vivo data with the reduction of BIN1 expression. Nevertheless, the effect of BIN1 loss on intracellular A␤ accumulation identified by Ubelmann et al. (27) was rescued by the re-introduction of the neuronal BIN1 isoform but not the ubiquitous isoform, implicating potential isoform-specific differences in BIN1 function in the brain which may require further investigation in light of differential BIN1 isoform expression in AD (9 -11). Moreover, given reports of cell-specific BIN1 isoform expression, more in-depth analysis of cell-autonomous roles for BIN1 in the context of amyloid deposition would also seem pertinent.
Mice with targeted deletion of both Bin1 alleles are not viable as the pups do not survive beyond a day or two after the birth  (n ϭ 4). F, non-PSD membranes were analyzed for APP, BACE1, Amph1, and synaptophysin by immunoblotting. G, quantitative analysis of APP and BACE1 levels in pre-synaptic fractions (n ϭ 4). H, forebrain lysates were analyzed for steady-state levels of A␤40 and A␤42 using V-PLEX 4G8 immunoassay (A␤40, n ϭ 10 per genotype; A␤42, n ϭ 9 Bin1 Fl/Fl and 10 Bin1 Fl/Fl :Syn-Cre). (30). Our choice to use a global BIN1 heterozygous knockout, and a neuronal cKO circumvents the neonatal lethality of Bin1 KO mice and still models the reduction of neuronal BIN1 isoforms reported in patients with sporadic AD (9 -11). The presence of the CLAP domain in longer brain-specific BIN1 isoforms predicts an important role in CME. Indeed, a role for BIN1 in endocytosis at the synapse has been proposed based on the documented role for amphiphysin 1 in synaptic vesicle recycling and dynamics (5,8). However, previous studies did not reveal alterations in the density of synaptic vesicles in cultures of neurons from Bin1 knockout mice (30). BIN1 has also been identified among the 50 most highly expressed genes in oligodendrocytes and shows prominent expression in oligodendrocytes and white matter tracts in mouse and human brain tissue (11,12,47). White matter dysfunction is emerging as an important element of Alzheimer's disease pathogenesis (48), with evidence of a direct relationship between A␤ levels and white matter hyperintensities (49). Cause and effect relationship between A␤ and white matter dysfunction is yet to be fully understood, but evidence exists for white matter dysfunction in autosomal dominantly inherited forms of AD which precedes symptoms by 20 years (48). Demyelination of axons in the central nervous system is thought to leave them more vulnerable to A␤ toxicity. Interestingly, axonal damage in close proximity to amyloid deposits is a key feature of the 5XFAD mouse model (50). Furthermore, myelin deficits have been identified in the 5XFAD mouse model and some evidence exists for a role of myelin in amyloid deposition (51,52). Despite high expression in oligodendrocytes, a definitive role for BIN1 in myelination is yet to be established, but our data indicate 50% loss of BIN1 expression does not recapitulate previously reported data for a role of myelin in amyloid deposition.

BIN1 does not alter ␤-amyloid pathology in vivo
Despite an emerging body of data regarding the possible role of BIN1 in LOAD risk, a consensus on the mechanism through which its risk is conferred is yet to be reached. Our data do not support a role for BIN1 in A␤ generation in vivo despite evidence from some in vitro studies to suggest the opposite. Future investigation using Bin1 neuronal cKO in the 5XFAD model may provide additional information regarding a definitive role for BIN1 in neuronal amyloid production and deposition. Further understanding of the biological role of BIN1 in the brain, understanding the consequences of BIN1 SNPs on cellular BIN1 protein isoform abundance or subsequent function in select neuronal cell types remain essential pieces of information in understanding the role of BIN1 in LOAD disease progression.

Mice
Bin1 Fl/Fl mice were generously provided by Dr. George C. Prendergast (Lankenau Institute for Medical Research) (53). Synapsin1-Cre line (Syn-Cre; stock no. 003966) was obtained from The Jackson Laboratory (Bar Harbor, ME). Bin1 Fl/Fl :Syn-Cre and Bin1Ϯ mice were generated in-house and maintained in the C57BL/6 background. Male Bin1 Fl/Fl :Syn-Cre mice were bred to female C57BL/6 mice to generate progeny with germline deletion of a single Bin1 allele. Bin1ϩ/Ϫ mice were intercrossed, and pups sacrificed at postnatal day zero (P0) to con-firm the generation of Bin1ϩ/ϩ, Bin1ϩ/Ϫ, and Bin1Ϫ/Ϫ using PCR genotyping and immunoblot analyses. Bin1ϩ/Ϫ mice were bred to 5XFAD animals (C57BL/6-SJL mixed background). All procedures related to animal care and treatment in this study have been approved by the Institutional Animal Care and Use Committee at the University of Chicago.

Immunoblot analysis
Detergent lysates of mouse brain were prepared as described previously (55). The samples were resolved by SDS-PAGE on 4 -20% Tris-glycine gels (Bio-Rad). APP-derived C-terminal fragments were resolved on NuPAGE 4 -12% Bis-Tris protein gels (Thermo Fisher). Proteins were transferred to PVDF membranes and developed with IR680 anti-rabbit and IR800 antimouse secondary antibodies and imaged using the Odyssey IR Imaging System (LI-COR Biosciences). Densitometric analysis was performed using ImageJ.

Membrane fractionations
Post-synaptic and non-post-synaptic membrane fractions were purified from neuronal cKO mice using an adapted methodology (56). Briefly, forebrains were homogenized in 0.32 M sucrose and 10 mM HEPES, pH 7.4, and centrifuged at 800 ϫ g for 10 min to pellet nuclei and large debris. The supernatant was centrifuged at 12,000 ϫ g for 20 min to pellet a crude membrane fraction. The pellet was washed twice in wash buffer (4 mM HEPES and 1 mM EDTA, pH 7.4), resuspended in 20 mM HEPES, 100 mM NaCl, 0.5% Triton X-100, pH 7.2, and held on ice for 30 min. The suspension was centrifuged at 12,000 ϫ g for 20 min to separate pre-synaptic proteins. The pellet was resuspended in 20 mM HEPES, 0.15 mM NaCl, 1% Triton X-100, 1% deoxycholic acid, and 1% SDS, pH 7.5, and held on ice for 1 h. The suspension was centrifuged at 800 ϫ g for 10 min, and the supernatant saved as the post-synaptic fraction.

Histology and immunohistochemistry
For free-floating sections, hemibrains were processed as described previously and serially cut into 40-M-thick sections (57). Sections were stained with primary antibodies described above and with secondary antibodies conjugated to Alexa Fluor 488, 555, or 647 (Life Technologies). Fibrillar amyloid deposits were stained using thioflavin S as described previously (58).

Image analysis
For amyloid burden quantification (thioflavin S and mAb 3D6), five to eight sections per mouse were analyzed using ImageJ as described previously (58).

A␤ ELISA
For endogenous A␤ ELISA, forebrain tissue from 4-monthold Bin1 Fl/Fl , Bin1 Fl/Fl :Syn-Cre, WT or Bin1ϩ/Ϫ mice were homogenized in 10 volumes (w/v) of 1% SDS in PBS containing cOmplete™, EDTA-free Protease Inhibitor mixture (Sigma-Aldrich) and clarified by centrifugation. A␤ levels were measured using a V-PLEX Plus A␤ Peptide Panel 1 (4G8) Kit (Meso Scale Discovery). For human A␤, brains from 4-month-old 5XFAD and 5XFAD Bin1ϩ/Ϫ mice were homogenized in 20 volumes (w/v) of TBS containing protease inhibitor mixture and clarified by centrifugation at 100,000 ϫ g for 1 h at 4°C. The pellet was resuspended 700 l of 70% formic acid and centrifuged at 100,000 ϫ g for 1 h at 4°C. The supernatant was neutralized with 9 volumes (v/v) of 1 M Tris-HCl, pH 11. TBS-soluble and formic acid-extracted A␤ was measured using a V-PLEX Plus A␤ Peptide Panel 1 (6E10) Kit (Meso Scale Discovery) according to the manufacturer's instructions. Samples were subjected to the bicinchoninic acid assay to determine protein concentration and A␤ concentrations adjusted to total protein concentrations.

Elevated-plus maze
Mice were placed in the closed arm of an elevated-plus maze suspended 40 cm above the ground and allowed to explore freely for 8 min. The two opposite closed arms of the Greek cross-shaped maze had 15 cm high walls surrounding a 5-cmwide platform and open arms had the same 5-cm-wide platform with no surrounding walls. Time spent in closed and open arms was monitored over the duration of the exploratory period.

Novel object recognition
Mice were introduced to the empty novel object recognition (NOR) arena for an 8-min session on day 1 of testing for habituation to the novel environment. After 24 h, mice were returned to the arena which contained two identical novel objects and allowed to explore freely for 8 min. After a further 24 h, mice were returned to the arena containing one familiar object from the previous day and one novel object. Time spent interacting with the novel and familiar objects was recorded and used to generate a discrimination index using the calculation (novel object(s) Ϫ familiar object(s))/(novel object(s) ϩ familiar object(s)) (59).

Y-maze
4-month-old mice were allowed to freely explore the arms of a Y-maze for an 8-min session. The total number of arms entries (counted when all four limbs had entered the arm of the maze) and the sequence in which arms were entered were recorded. The percentage of spontaneous alternations was calculated and analyzed as described previously (58).

Contextual fear conditioning
Contextual fear conditioning was performed on 4-monthold mice in a Coulbourn HABITEST isolation cubicle with shocks applied using a Coulbourn Precision Animal Shocker (Harvard Apparatus) as described (58). Analysis of recordings from the video-based system was performed using Freez-eFrame 4 software (Actimetrics) as described previously (58).

Statistical analysis
All statistical analyses were performed using GraphPad Prism Version 7 (GraphPad Software) with data represented as the mean Ϯ S.E. unless described otherwise in the figure legend. Comparisons between two groups were performed using unpaired t tests. Cumulative time frozen in the contextual fear conditioning paradigm was analyzed by two-way analysis of variance (ANOVA).