Progressive accumulation of amyloid-β oligomers in Alzheimer’s disease and in amyloid precursor protein transgenic mice is accompanied by selective alterations in synaptic scaffold proteins

Authors


E. Masliah, Department of Neurosciences, University of California, San Diego, La Jolla, CA 92093-0624, USA
Fax: (858) 534 6232
Tel: (858) 534 8992
E-mail: emasliah@UCSD.edu

Abstract

The cognitive impairment in patients with Alzheimer’s disease is closely associated with synaptic loss in the neocortex and limbic system. Although the neurotoxic effects of aggregated amyloid-β oligomers in Alzheimer’s disease have been studied extensively in experimental models, less is known about the characteristics of these aggregates across the spectrum of Alzheimer’s disease. In this study, postmortem frontal cortex samples from controls and patients with Alzheimer’s disease were fractionated and analyzed for levels of oligomers and synaptic proteins. We found that the levels of oligomers correlated with the severity of cognitive impairment (blessed information-memory-concentration score and mini-mental state examination) and with the loss of synaptic markers. Reduced levels of the synaptic vesicle protein, vesicle-associated membrane protein-2, and the postsynaptic protein, postsynaptic density-95, correlated with the levels of oligomers in the various fractions analyzed. The strongest associations were found with amyloid-β dimers and pentamers. Co-immunoprecipitation and double-labeling experiments supported the possibility that amyloid-β and postsynaptic density-95 interact at synaptic sites. Similarly, in transgenic mice expressing high levels of neuronal amyloid precursor protein, amyloid-β co-immunoprecipitated with postsynaptic density-95. This was accompanied by a decrease in the levels of the postsynaptic proteins Shank1 and Shank3 in patients with Alzheimer’s disease and in the brains of amyloid precursor protein transgenic mice. In conclusion, this study suggests that the presence of a subpopulation of amyloid-β oligomers in the brains of patients with Alzheimer’s disease might be related to alterations in selected synaptic proteins and cognitive impairment.

Structured digital abstract

Abbreviations

amyloid-β 1–42

AD

Alzheimer’s disease

ANOVA

analysis of variance

APP

amyloid precursor protein

BIMC

blessed information-memory-concentration

CHO cells

Chinese hamster ovary cells

MAP2

microtubule-associated protein 2

MCI

mild cognitive impairment

MMSE

mini-mental state examination

NMDA-R

N-methyl-d-aspartic acid receptor

nontg

nontransgenic

PDZ

PSD95, discs large, and zona occludens-1 complex

PSD95

postsynaptic density-95

SAPAP

SAP90/PSD95-associated protein

SNAP25

synaptosomal-associated protein 25

tg

transgenic

UCSD

University of California, San Diego

VAMP2

vesicle-associated membrane protein-2

Introduction

The cognitive impairment in patients with Alzheimer’s disease (AD) is closely associated with synaptic loss in the neocortex and limbic system [1–3]. Several lines of investigation support the view that increasing levels of amyloid-β 1–42 (Aβ), the proteolytic product of amyloid precursor protein (APP) metabolism, might be centrally involved in the pathogenesis of AD [4–7]. The mechanisms through which the accumulation of Aβ monomers, oligomers and other APP metabolites might lead to synaptic damage and neurodegeneration are under investigation. More specifically, the potential role of neurotoxic Aβ oligomers has emerged as a topic of considerable interest in recent years [8–11].

Under pathological conditions, monomeric forms of Aβ can aggregate to form several different species, including amyloid fibrils, protofibrils, annular structures, Aβ-derived diffusible ligands [12] and smaller order oligomeric species (for reviews, see refs. [13–15]). Smaller oligomeric species of synthetic Aβ are different from protofibrils depending on how synthetic Aβ is prepared. Oligomers of Aβ peptides can organize into dimers, trimers, tetramers and higher order arrays that can form annular structures. Smaller oligomers are divided into those generated from synthetic peptides and those purified from cells, transgenic (tg) mice or AD human brains [13].

Naturally occurring Aβ oligomers can be resistant to SDS, guanidine hydrochloride and Aβ-degrading proteases [16]. An example of a naturally occurring oligomeric species is Aβ*56, derived from the brains of APPtg mice, which has been shown to promote age-dependent memory deficits [17]. Aβ*56 and Aβ trimers secreted by cultured cells may be found to share common synaptotoxic properties [13]. The Aβ dimers, trimers and higher order oligomers secreted by cultured neurons inhibit long-term potentiation, damage spines and interfere with activity-regulated cytoskeleton-associated protein location [9,10,13,18]. Additional studies have shown that Aβ dimers extracted from human cerebrospinal fluid disrupt synaptic plasticity and inhibit hippocampal long-term potentiation in vivo [19]. Together, these studies indicate that Aβ oligomers, ranging in size from two to 12 subunits, might be responsible for synaptic damage and memory deficits [20]. The mechanisms through which Aβ aggregates might lead to synaptic damage are unclear. A number of recent studies have begun to investigate the possibility that Aβ oligomers might interfere with synaptic function by altering synaptic proteins, such as postsynaptic density-95 (PSD95) [21–24] and glutamate receptors [25]. In addition to the role of oligomers, Aβ monomers also accumulate in high levels in the brains of patients with AD and may also contribute to the neurodegenerative process.

Although the neurotoxic effects of Aβ oligomers have been studied extensively in experimental models, less is known about the characteristics of oligomers across the spectrum of AD and how this correlates with cognition and synaptic proteins. For this purpose, we utilized immunoblot analysis to investigate the relationship between levels of Aβ oligomers and synaptic proteins in fractions from the brains of patients with AD and in APPtg mice. Our studies show that Aβ oligomers, in particular dimers and pentamers, progressively accumulate in the brains of patients with AD as well as in APPtg mice. This was accompanied by reductions in the levels of synaptic scaffold proteins, such as PSD95, Shank1 and Shank3.

Results

Levels of Aβ oligomers are associated with cognitive impairment and alterations in synaptic proteins in AD

To analyze the levels of Aβ monomers and oligomers in controls versus patients with mild cognitive impairment (MCI) and AD, high-resolution immunoblot assays were performed on the cytosolic and membrane fractions obtained by ultracentrifugation using samples extracted with either Buffer A [9] or Buffer B [26,27] and probed with antibodies against Aβ (clones 82E1 and 6E10, 4G8). When the fractionation procedure was performed with Buffer A (Fig. 1A,B) or Buffer B (Fig. 1C,D) and immunoblots were probed with Aβ antibodies 82E1 (Fig. 1A,C), 6E10 (Fig. 1B,D) or 4G8 (not shown), we found that the clearest banding pattern, consistent with the estimated weight of Aβ monomers and multimers, was detected using the membrane fractions of samples prepared with Buffer A and probed with the 82E1 antibody (Fig. 1A). With this approach, bands ranging in approximate molecular weight from 4 to 28 kDa were detected, with the 4 kDa band corresponding to monomers and the higher order bands (8, 12, 16, 20, 24 and 28 kDa) probably corresponding to dimers, trimers, tetramers, pentamers, hexamers and heptamers, respectively (Fig. 1A). In brain samples from cases with MCI and AD, prepared with Buffer A and probed with the 82E1 antibody (Fig. 2A), there was a significant increase in the levels of the bands corresponding to monomers (Fig. 2B), dimers (Fig. 2C) and higher order oligomers (Fig. 2D, Table 1) when compared with controls. The greatest difference between controls and cases with MCI and AD was in the levels of monomeric Aβ (Fig. 2B, Table 1). Further analysis of the human samples by Aβ ELISA confirmed the undetectable levels of Aβ in the controls and comparable higher levels in cases with MCI and AD (Table 1). Moreover, the levels of the synaptic proteins vesicle-associated membrane protein-2 (VAMP2) and PSD95, and, to a lesser extent, syntaxin (Fig. 2E–G) and synaptosomal-associated protein 25 (SNAP25) were reduced in cases with MCI and AD when compared with neurologically unimpaired controls (Table 1).

Figure 1.

 Comparative immunoblot analysis for APP/Aβ in the frontal cortex of controls and patients with AD. Samples were fractionated into membrane and cytosolic fractions and probed with Aβ antibodies (82E1 and 6E10). (A, B) In samples homogenized using Buffer A, compared with nondemented controls, in AD samples multiple bands representing Aβ monomers and multimers were identified at molecular weights ranging from 4 to 28 kDa in the membrane fraction. (C, D) In samples homogenized using Buffer B, compared with nondemented controls, in AD samples the majority of Aβ was identified as a 4 kDa band in the membrane fraction.

Figure 2.

 Analysis of the Aβ and synaptic protein bands detected by immunoblot in control and AD brain samples. Samples were homogenized using Buffer A and probed with antibodies against Aβ (82E1) and synaptic proteins (VAMP2, syntaxin, PSD95). (A) Representative western blot with the membrane fractions from controls and cases with MCI and AD displaying bands corresponding to the Aβ monomer (4 kDa) and multimers (8–28 kDa) and PSD95 (95 kDa). (B–D) Semi-quantitative analysis of the bands representing Aβ monomer (4 kDa) (B), dimer (8 kDa) (C) and higher order oligomers (12–28 kDa) (D), showing a progressive increase in cases with AD. (E–G) Semi-quantitative analysis of immunoblots for VAMP2 (E), syntaxin (F) and PSD95 (G), showing a reduction in immunoreactivity in cases with AD. n = 5 cases per group. *P < 0.05 compared with nondemented control by one-way ANOVA with post hoc Dunnett’s test.

Table 1.   Summary of immunoblot and ELISA of levels of Aβ and synaptic proteins in controls and cases with MCI and AD. Expressed as ratio of specific signal to actin, mean ± SEM. *P < 0.05 compared with nondemented control by one-way anova with post hoc Dunnett’s test.
GroupnMonomerDimerTrimerTetramerPentamerOligomers sumSynapsin IVAMP2Syntaxin 1ASNAP25PSD95Aβ (pg·mL−1)
Control50.051 ± 0.010.146 ± 0.0270.115 ± 0.0130.126 ± 0.0150.156 ± 0.0320.543 ± 0.043.690 ± 0.4153.26 ± 0.41.51 ± 0.2911.625 ± 0.181.765 ± 0.343  0
MCI50.134 ± 0.01*0.188 ± 0.01*0.208 ± 0.01*0.226 ± 0.029*0.260 ± 0.019*0.882 ± 0.07*1.906 ± 0.303*2.58 ± 0.1*1.34 ± 0.11.042 ± 0.1*1.178 ± 0.1*301 ± 42*
Moderate AD50.16 ± 0.029*0.223 ± 0.012*0.246 ± 0.02*0.236 ± 0.027*0.317 ± 0.036*1.020 ± 0.12*2.907 ± 0.3902.48 ± 0.187*1.25 ± 0.1*1.19 ± 0.15*1.21 ± 0.15*298 ± 34*
Advanced AD50.111 ± 0.01*0.237 ± 0.011*0.205 ± 0.014*0.184 ± 0.013*0.288 ± 0.018*0.940 ± 0.08*2.56 ± 0.255*1.78 ± 0.21*1.04 ± 0.1*1.01 ± 0.15*0.6 ± 0.23*229 ± 41*

Immunohistochemical analysis with the 82E1 antibody showed, in both patients with AD and in APPtg mice, immunostaining of abundant diffuse and dense core plaques (Fig. 3A–F). In addition, with this antibody, there were subtle linear Aβ immunoreactive deposits distributed along the neurons. Double-labeling studies utilizing antibodies against PSD95 and 82E1 showed that the linear and punctate deposits along the neurons co-localized with PSD95 in the dendritic processes (Fig. 3G–L). Linear regression analysis was performed to investigate the relationship between the levels of oligomers, synaptic proteins and cognitive impairment. The levels of dimers and pentamers were correlated with the severity of cognitive impairment [blessed information-memory-concentration (BIMC) score and mini-mental state examination (MMSE)] (Fig. S1A, see Supporting information, Table 2) and with the Braak stage (Fig. S1B, Table 2). Moreover, levels of dimers and pentamers correlated with the loss of synaptic proteins such as VAMP2 and PSD95 (Fig. S1C,D, Table 2). Consistent with this observation, the levels of the synaptic proteins VAMP2 and PSD95 were significantly correlated with the severity of cognitive impairment (Fig. S1E, Table 2). Levels of PSD95 were also correlated with the Braak stage (Fig. S1F, Table 2). A total of six bands, which represent the multimeric forms of Aβ, were significantly correlated with the BIMC and MMSE scores and the Braak stage (Table 2).

Figure 3.

 Immunohistochemical analysis of the patterns of Aβ immunoreactivity (82E1) in cases with AD and in APPtg samples. (A–F) Vibratome sections were immunolabeled with an Aβ antibody (82E1) and reacted with 3,3′-diaminobenzidine. (G–L) Vibratome sections were double-immunolabeled with an Aβ antibody (82E1, green channel) and PSD95 (red channel). (A–C) Compared with nondemented controls (A), in the frontal cortex of cases with MCI, Aβ was detected as discrete granular structures (arrows, B). In cases with advanced AD, the antibody detected abundant plaques (C). (D–F) Compared with nontg controls (D), in the frontal cortex of APPtg mice, the Aβ antibody detected discrete diffuse structures (arrows, E) in the neuropil, as well as fibrillar mature plaques (F). (G–I) In cases with mild AD, the discrete Aβ-positive granular structures co-localized with PSD95 (arrows). (J–L) In APPtg mice, the diffuse Aβ-positive structures co-localized with PSD95 (arrows). Scale bars: (A–F) 50 μm; (G–L) 20 μm.

Table 2.   Summary of correlation coefficients between levels of Aβ oligomers, synaptic proteins, dementia score and neuropathology. *P < 0.05 compared with nondemented control by one-way ANOVA with post hoc Dunnett’s test.
VariableMonomerDimerTrimerTetramerPentamerOligomers sumSynapsin IVAMP2SyntaxinSNAP25PSD95
BIMC0.2610.598*0.4210.2970.567*0.508*−0.240−0.627*−0.521*−0.507*−0.543*
MMSE−0.234−0.594*−0.402−0.206−0.532*−0.501*0.1090.542*0.585*0.4380.435
Braak score0.2570.755*0.4500.2060.593*0.525*−0.152−0.699*−0.586*−0.471−0.683*
Synapsin I−0.445−0.343−0.090−0.336−0.313−0.348
VAMP2−0.157−0.727*−0.565*−0.162−0.427*−0.396
Syntaxin−0.183−0.845*−0.706*−0.352−0.589*−0.561*
SNAP25−0.278−0.698*−0.030−0.574*−0.644*−0.659*
PSD95−0.223−0.665*−0.223−0.239−0.491*−0.444*

Accumulation of Aβ oligomers and loss of synaptic proteins in APPtg mice

To analyze the levels of Aβ in APPtg mice, immunoblot assays were performed with cytosolic and membrane fractions homogenized with Buffer A (Fig. S2A,B, see Supporting information) or Buffer B (Fig. S2C,D), and probed with antibodies against Aβ clones 82E1 (Fig. S2A,C), 6E10 (Fig. S2B,D) and 4G8 (not shown). Similar to the studies in AD brains (Fig. 1), in APPtg mice we found a clear banding pattern consistent with the estimated molecular weight of the Aβ monomers and multimers when using the membrane fraction of samples prepared with Buffer A and probed with the 82E1 antibody (Fig. S2A). Compared with nontransgenic (nontg) mice, in 6-month-old APPtg mice, we observed abundant levels of monomers, dimers, trimers and, to a lesser extent, other Aβ multimers (Fig. 4A,B). In agreement with the studies in patients with AD (Fig. 2), levels of the synaptic proteins VAMP2, syntaxin and PSD95 were reduced significantly in homogenates from APPtg mice compared with nontg controls (Fig. 4C).

Figure 4.

 Analysis of the Aβ and synaptic protein bands detected by immunoblot in APPtg mice. Samples were homogenized using Buffer A and probed with antibodies against Aβ (82E1) and synaptic proteins (VAMP2, syntaxin, PSD95). All panels are from the brains of 6-month-old mice. (A) Representative western blot of the membrane fractions from 6-month-old nontg control and APPtg mice displaying bands corresponding to Aβ monomer (4 kDa) and multimers (8–28 kDa) and PSD95 (95 kDa). (B) Semi-quantitative analysis of the bands representing Aβ monomer (4 kDa), dimer (8 kDa) and higher order oligomers (12–28 kDa). (C) Semi-quantitative analysis of the immunoblots for VAMP2, syntaxin, PSD95. n = 8 mice per group. *P < 0.01 compared with nontg control by unpaired, two-tailed Student’s t-test.

Interactions between Aβ and PSD95 in the brains of patients with AD and in APPtg mice

To further investigate the interactions between Aβ and synaptic proteins in AD, co-immunoprecipitation and double-labeling experiments were performed. For this purpose, control and AD brain homogenates extracted with Buffer A were immunoprecipitated with antibodies against Aβ (82E1 clone) or synaptic proteins (PSD95, VAMP2, syntaxin and SNAP25). When samples from human brains were immunoprecipitated with the antibody against Aβ (82E1 clone) and then analyzed by western blot with an antibody against PSD95, the strongest interaction was observed in cases with AD when compared with nondemented controls (Fig. 5A,B). No interacting bands were detected in control experiments with samples immunoprecipitated with a nonimmune IgG (Fig. 5A) or when the tissue sample was excluded. No differences were detected between cases with AD and nondemented controls in co-immunoprecipitation experiments with antibodies against Aβ (82E1 clone) and the synaptic proteins syntaxin, SNAP25 or VAMP2 (not shown).

Figure 5.

 Co-immunoprecipitation studies for Aβ and PSD95 in cases with AD and in APP tg mice. Samples from the frontal cortex of human nondemented controls and cases with AD, or from the brains of nontg and APP tg mice, were homogenized with Buffer A and membrane fractions were processed for immunoprecipitation. (A) Samples from the brains of control and cases with AD were immunoprecipitated (IP) with an Aβ antibody (82E1), and analyzed by western blot (WB) with an antibody against PSD95. The reactive band was more intense in cases with AD (arrow); no reactive bands at 95 kDa were observed under control conditions. (B) Semi-quantitative analysis of the co-immunoprecipitated band showed higher levels in cases with AD. (C) Mouse brain cortex samples from 6-month-old animals were immunoprecipitated with Aβ antibody (82E1), and analyzed by western blot with an antibody against PSD95. The reactive band was more intense in APP tg mice (arrow). (D) Semi-quantitative analysis of the co-immunoprecipitated band showed higher levels in APP tg mice. (E) Mouse brain cortex samples from 6-month-old animals were immunoprecipitated with PSD95 antibody, and analyzed by western blot with an antibody against Aβ (82E1). The reactive band was more intense in APP tg mice (arrow). n = 3 cases or mice per group, *P < 0.05 compared with control by one-way ANOVA with post hoc Dunnett’s test. (F) Mouse brain cortex samples from 6-month-old APP tg mice were immunoprecipitated with an antibody against Aβ (82E1), and the resulting co-precipitates were analyzed by mass spectroscopy. Both Aβ tryptic peptides were identified together with three PSD95 peptides. The charge and XCorr score of the identified peptides are indicated.

Similarly, when brains from APPtg mice were immunoprecipitated with an antibody against Aβ and analyzed by western blot with an antibody against PSD95, the strongest interaction was observed in APPtg samples relative to nontg controls (Fig. 5C,D). Similar results were obtained when reverse co-immunoprecipitation was performed using the antibody against PSD95 followed by western blot analysis with the Aβ antibody (82E1 clone) (Fig. 5E). No significant interactions were detected by co-immunoprecipitation between Aβ (82E1 clone) and the synaptic proteins syntaxin, SNAP25 or VAMP2 (not shown). To further investigate the potential interactions between Aβ and other synaptic proteins in an unbiased manner, we immunoprecipitated Aβ with the 82E1 antibody from the brains of APPtg mice and analyzed the copurifying proteins by multidimensional protein identification technology mass spectrometry. Analysis of Aβ immunoprecipitates revealed that, indeed, PSD95 interacts with Aβin vivo (Fig. 5F). In contrast, no significant recovery of PSD proteins was obtained in samples precipitated with IgG alone or in control samples.

Alterations in proteins involved in the dendritic spine motility apparatus in patients with AD and in APPtg mice

Spine motility plays an important role in learning and memory [28], and previous studies have shown that Aβ oligomers might interfere with this function [29]. PSD95 has been shown to play an important role in spine motility by providing a scaffold for other dendritic proteins, such as Shank, Homer and actin [30–32]. Given that Aβ oligomers have been shown to interfere with PSD95, it is possible that abnormalities in spine motility in AD might be associated with alterations in the downstream effectors. To investigate this possibility, immunoblot analysis for postsynaptic proteins was performed in fractionated homogenates from human and mouse brain samples. In the membrane fraction of samples extracted with Buffer A, Shank1, Shank3 and pan-Shank were detected as triple or quadruple bands ranging in size from 160 to 240 kDa; these multiple bands are consistent with the known alternative splicing of Shank. The other PSD proteins, such as Homer 1, were detected at ∼ 40 kDa, whereas SAP90/PSD95-associated protein 1 (SAPAP1) was detected as a single band at around 110 kDa (Fig. 6A,B). When compared with controls, in cases with MCI and AD, there was a significant reduction in the levels of the bands corresponding to Shank1, Shank3 and total pan-Shank (Fig. 6C). In contrast, levels of Homer and SAPAP1 were not altered significantly (Fig. 6E). Similarly, compared with nontg controls, in APPtg mice, levels of pan-Shank, Shank1 and Shank3 were reduced (Fig. 6D), whereas levels of Homer and SAPAP1 were not significantly different (Fig. 6F).

Figure 6.

 Analysis of the dendritic scaffold proteins by immunoblot in AD brains and APP tg mice. (A) Frontal cortex of controls and cases with MCI and AD prepared with membrane fractions in Buffer A. (B) Nontg and APP tg cortex prepared with membrane fractions in Buffer A. (A) Representative western blot analysis of human brain samples probed with antibodies against pan-Shank, Shank1, Shank3, Homer and SAPAP1. (B) Representative western blot analysis of mouse brain samples probed with antibodies against pan-Shank, Shank1, Shank3, Homer and SAPAP1. (C, D) Semi-quantitative analysis showing a reduction in Shank proteins in MCI and AD compared with nondemented control (C), and a similar reduction in APP tg mice compared with nontg controls (D). (E, F) Semi-quantitative analysis showing no changes in Homer or SAPAP1 levels in diseased human (E) or tg mouse brains (F). n = 5 cases per group for control, MCI and AD samples. *P < 0.05 compared with nondemented controls by one-way ANOVA with post hoc Dunnett’s test. n = 8 mice per group for nontg and APP tg mice. *P < 0.01 compared with nontg controls by unpaired, two-tailed Student’s t-test.

To further investigate the effects of Aβ on the dendritic scaffold proteins, primary neuronal cultures were exposed to conditioned medium from Chinese hamster ovary (CHO) cells that produce Aβ oligomers. Compared with vehicle-treated controls, after 6 or 24 h of exposure to Aβ oligomers, primary neuronal cells displayed a reduction in the numbers of pan-Shank-positive punctae along the dendritic arbor (Fig. 7). This effect was detected after 6 h of treatment (Fig. 7C) and, after 24 h, these effects became more evident (Fig. 7A–E). In addition, a similar reduction in the numbers of PSD95-positive punctae was detected after 24 h of treatment with Aβ (Fig. S3, see Supporting information). By lactate dehydrogenase assay, levels of neuronal survival were comparable between cells treated with control conditioned medium and Aβ-containing medium (Fig. S3). Together, these results support the possibility that Aβ reduces Shank and PSD95 content in the dendrites and that the effects are not the result of cell death.

Figure 7.

 Double immunolabeling analysis for microtubule-associated protein 2 (MAP2) and pan-Shank in primary neuronal cultures treated with conditioned medium containing Aβ oligomers. Hippocampal neuronal cells from P1 mice were treated for 6 or 24 h with conditioned medium from APP-expressing CHO cells (80 pm, a sublethal dose). Fixed cells on coverslips were immunolabeled with antibodies against MAP2 (green channel) and pan-Shank (red channel) and analyzed with a laser scanning confocal microscope. All images are from cells treated for 24 h; the graph represents data from both 6 and 24 h time points. (A, B) Confocal images showing neurons after 24 h of treatment with vehicle (A) or Aβ (B). Compared with vehicle-treated cells, Aβ treatment resulted in a reduction in pan-Shank-positive punctae along the dendrites. (C) Analysis of levels of pan-Shank and MAP2-immunoreactive structures after 6 and 24 h of treatment with vehicle or Aβ. (D, E) Confocal images at higher power showing the detail of MAP2-labeled dendritic branches and pan-Shank-immunoreactive punctae (arrows) along the dendrites. Scale bars: (A, B) 20 μm; (D, E) 10 μm. n = 3 samples per condition, *P < 0.05 compared with vehicle-treated controls by unpaired, two-tailed Student’s t-test.

Discussion

The present study shows that the levels of Aβ oligomers, assessed with the 82E1 antibody in fractionated brain homogenates from patients with AD at different stages, correlate with the severity of cognitive impairment (BIMC score and MMSE) and with the loss of synaptic markers, such as the synaptic vesicle protein VAMP2 and the postsynaptic protein PSD95. The levels of Aβ dimers and pentamers correlate more closely with the severity of cognitive impairment and alterations in synaptic proteins.

This is consistent with recent studies in cellular and rodent models, showing that small soluble oligomers are toxic because they damage the synapses [20,33,34]. Aβ dimers, trimers and pentamers secreted by cultured neurons inhibit long-term potentiation and damage spines [18]. In hippocampal slices and in animal models, Aβ oligomers, ranging in size from two to 12 subunits, impair synapse function [20]. A recent study has shown that intra-axonal injection of oligomeric Aβ42 acutely inhibits synaptic transmission at the squid giant synapse by disrupting synaptic vesicles [35]. Interestingly, Aβ dimers recovered from the cerebrospinal fluid of patients with AD cause memory deficits and synaptic dysfunction when infused in vivo [19]. Moreover, soluble Aβ dimers purified from AD brains have been shown to inhibit long-term potentiation, enhance long term depression and reduce dendritic spine density in rodent hippocampus [36]. The co-administration of antibodies specific for the N-terminus of Aβ prevented these synaptic and functional deficits, whereas antibodies against the C-terminus were less effective [37].

Taken together, these results suggest that Aβ dimers and other small oligomers might initiate the cascade of events leading to synaptic damage and cognitive impairment in patients with AD. However, correlational studies can only suggest that such interactions between Aβ and synapses might be at play in AD, and additional studies are needed to confirm this possibility. The mechanisms through which the accumulation of Aβ monomers and oligomers might damage the synapses are not completely clear. One possibility is that this effect might be mediated in part by interactions between Aβ and pre- and postsynaptic proteins, such as VAMP2 and PSD95, respectively. Another possible interpretation is that, once Aβ toxic arrays cross from the presynaptic to the postsynaptic site, the damage in the postsynaptic site results in denervation and secondary changes in the presynaptic site. However, it is possible that Aβ has toxic effects in both the pre and postsynaptic sites. This study also showed that Aβ co-immunoprecipitated with PSD95 in the brains of patients with AD and in APPtg mice, and that discrete Aβ aggregates co-localized with PSD95 along the dendrites in the brains of AD patients and in APPtg mice. In addition, levels of PSD95, Shank1 and Shank3 were reduced. This is consistent with previous studies showing alterations in the levels of PSD95 in AD [21,22,24] and in APPtg models [23,25]. Moreover, and in agreement with our publication, a recent study has shown a reduction in PSD95 levels in the hippocampus of subjects with MCI, which was accompanied by decreased levels of two proteins associated with PSD95, namely the N-methyl-d-aspartic acid receptor (NMDA-R) subunit A and the low-density lipoprotein receptor-1 [24]. In support of a role for Aβ in the mechanisms of postsynaptic damage in AD, previous studies have shown that soluble Aβ oligomers induce the degradation of PSD95 [38] and promote alterations in PSD architecture by depleting the synaptic pool of Homer1b and Shank1 clusters [39]. It is thought that signaling pathways, such as phosphatidylinositol 3-kinase and extracellular signal-regulated kinase, might mediate the actions of soluble Aβ on Homer1b and Shank levels [39]. In support of this possibility, we found that Aβ co-immunoprecipitated with PSD95 in the brains of patients with AD and in APPtg mice, and that discrete Aβ aggregates co-localized with PSD95 along the dendrites in AD and in APPtg mice brains. Further supporting the co-immunoprecipitation results, mass spectrometry studies showed that PSD95 fragments can be recovered from the brains of APPtg mice precipitated with an antibody against Aβ but not with an IgG control. However, these results should be interpreted with caution, given the known tendency of Aβ to bind other proteins. Moreover, interactions between Aβ and PSD95 may be indirectly mediated by other, as yet unidentified, proteins. The distributions of the discrete Aβ aggregates detected with 82E1 are similar to those previously described using the A11 antibody against oligomers [40]. This is consistent with previous studies showing that Aβ-derived diffusible ligands and other naturally occurring Aβ oligomers bind to the postsynaptic density and interfere with dendritic spine function [33,41]. Studies utilizing real-time two-photon microscopy have demonstrated that spine motility plays an important role in learning and memory, and Aβ oligomers are capable of impairing this process [29,36]. Oligomers might disturb spine motility by disrupting the spine scaffold supported by PSD95, Shank1 and Shank3; however, the mechanisms through which Aβ might disturb PSD95 are not completely understood. One possibility might be that soluble monomers and oligomers secreted at the presynaptic site might be taken up by the postsynaptic site. This is supported by recent studies in which fluorescently labeled Aβ oligomers were shown to be taken up via the endocytic pathway in neuronal cell cultures [42]. The second is that oligomers could leak from lysosomal compartments, as has been suggested in previous studies [43,44].

PSD95 can join NMDA-Rs to the postsynaptic membrane by the interaction of the first and second PDZ domain with the NR2 subunit of the heteromeric NMDA-R complex [45–48]. In addition to its ability to cluster NMDA-Rs, SAP90/PSD95 is most probably involved in targeting and NMDA-R signaling, as indicated by the analysis of various tg mouse models [49,50]. In addition, PSD95 is believed to play a central role in the process of spine motility by serving as a scaffold to bind and organize other integral postsynaptic membrane proteins and PSD components, such as the Shank proteins. The Shank family proteins are major components of the postsynaptic density [31,51,52] and interact with the PDZ ternary complex composed of PSD95, discs large, and zona occludens-1 (PDZ). This allows Shank to bind indirectly to the multiprotein NMDA-R and α-amino-3-hydroxyl-5-methyl-4-isoxazole-propionate receptor complexes via the guanylate kinase-associated protein [53] and to the C-terminus of group I metabotropic glutamate receptors [54]. The Shank proteins contain a ligand motif for Homer [54], which, in turn, binds to group I metabotropic glutamate receptors, inositol 1,4,5-trisphosphate receptors and ryanodine receptors [55]. Shank can also bind to cytoskeletal proteins and regulate spine motility via interactions with cortactin. Shank1 and Shank3 bind to spectrin [56] and F-actin-associated proteins [30–32].

Consistent with the possibility that alterations in PSD95 might lead to downstream changes in dendritic proteins involved in spine function, we found that the levels of the postsynaptic proteins Shank1 and Shank3 were reduced in patients with AD and in the brains of APPtg mice when compared with controls. This is consistent with a recent study showing alterations in glutamate receptors and Shank proteins in AD [57]. Interestingly, haploinsufficiency of Shank3 in humans causes a syndrome with dendritic spine dysgenesis [31] which results in a learning disability known as the 22q13 deletion syndrome [58,59].

Taken together, these studies suggest that, in AD, Aβ oligomers might lead to synaptic dysfunction by the sequestration of PSD95, which, in turn, might result in alterations to the dendritic spine scaffold and Shank proteins. This could then result in cytoskeletal changes and reduced spine motility. Alternatively, alterations in PSD95–Shank complexes could also result in the dysregulation of glutamate receptors. This latter possibility is consistent with recent studies [60] showing that Aβ triggers alterations in the endocytosis of α-amino-3-hydroxyl-5-methyl-4-isoxazole-propionate receptor, which compromises synaptic plasticity. Similarly, other studies have shown that cell-derived oligomers decrease dendritic spine density in the hippocampus by an NMDA-dependent signaling pathway [36], suggesting a model in which exposure to Aβ oligomers mimics a state of NMDA-R blockade, either by reducing NMDA-R activation, reducing NMDA-R-dependent calcium influx or enhancing NMDA-R-dependent activation of calcineurin.

In conclusion, this study has shown that the presence of certain species of small Aβ oligomers in the brains of patients with AD correlates with alterations in selected synaptic proteins and cognitive impairment. These results suggest that Aβ could interact directly with postsynaptic proteins, leading to alterations in the spine scaffold system.

Experimental procedures

Please note that additional methodological details are included in Doc. S1 (see Supporting information).

Subjects

A total of 20 human cases was included in this study. These were divided into several groups: control (neurologically unimpaired), MCI, moderate AD and advanced AD. A summary of the demographic and clinicopathological characteristics of these cases is presented in Table 3. The autopsy cases in this study came from patients evaluated at a number of sites associated with the Alzheimer’s Disease Research Center at the University of California, San Diego (UCSD), CA, USA. Written informed consent for neurobehavioral evaluation, autopsy and for the collection of samples and subsequent analysis was obtained from the patient and caregiver (usually the next of kin) before neuropsychological testing and after the procedures of the study had been fully explained. The study methodologies conformed to the standards set by the Declaration of Helsinki and Federal guidelines for the protection of human subjects. All procedures were reviewed and approved by the UCSD Institutional Review Board.

Table 3.   Summary of clinicopathological characteristics of the controls and cases with MCI and AD. CDR, Clinical Dementia Rating.
GroupnAge (years)Gender (male/female)Duration (years)Education (years)CDR scoreBIMC score (range)MMSE score (mean)Brain weight (g)Braak stage
Control584.4 ± 152/3013.3 ± 5.801–228.5 ± 3.41181 ± 73.30–I
MCI589.4 ± 4.83/27.0 ± 1.113.8 ± 3.90.51–1025.2 ± 5.71234 ± 90.1I–II
Moderate AD589.8 ± 7.92/37.5 ± 1.711.8 ± 6.8111–2416.6 ± 6.51097 ± 48.0III–IV
Advanced AD580.4 ± 6.82/38.6 ± 3.915.6 ± 2.9225–3311.2 ± 8.4986 ± 146.3V–VI

Neurobehavioral and neuropathological examination

Please refer to Doc. S1.

APPtg mouse samples

For studies in animal models, brain samples from 16 6-month-old mice (n = 8 nontg; n = 8 APPtg) were included for immunoblot analysis. The characteristics of thy1-APPmut tg (line 41) [61] have been described previously. APPtg mice express mutated (London V717I and Swedish K670M/N671L) human APP751 under the control of the murine Thy1 promoter [61]. This tg model was selected because mice produce high levels of AΒ1–42 and exhibit performance deficits in the water maze, synaptic damage and early plaque formation, beginning around 3 months of age [61,62]. tg lines were main-tained by crossing heterozygous tg mice with nontg C57BL/6 × DBA/2 F1 breeders. All mice were heterozygous with respect to the transgene. All experiments were performed in accordance with National Institutes of Health legislation, all animals were handled in strict accordance with good animal practice and all procedures were completed under the specifications set forth by the UCSD Institutional Animal Care and Use Committee.

Tissue fractionation

For the analysis of Aβ and synaptic proteins, tissues from humans and APPtg mice were processed utilizing two different methods. The first method utilized a sucrose-containing buffer that allows for the separation of Aβ oligomers [‘Buffer A’ containing NaCl/Pi (pH 7.4), 0.32 m sucrose, 50 mm Hepes, 25 mm MgCl2, 0.5 mm dithiothreitol, 200 μg·mL−1 phenylmethanesulfonyl fluoride, 2 μg·mL−1 pepstatin A, 4 μg·mL−1 leupeptin, 30 μg·mL−1 benzamidine hydrochloride]. The second method utilized a buffer that facilitates the separation of the membrane and cytosolic fractions (‘Buffer B’ containing 1.0 mm Hepes, 5.0 mm benzamidine, 2.0 mm 2-mercaptoethanol, 3.0 mm EDTA, 0.5 mm magnesium sulfate, 0.05% sodium azide; final pH 8.8).

Tissue extraction with Buffer A

Briefly, as described previously [8,9], frontal cortex from human and mouse brain samples (0.1 g) was homogenized in 0.4 mL of Buffer A containing phosphatase and protease inhibitor cocktails (Calbiochem, San Diego, CA, USA). The samples were centrifuged at 1000 g for 10 min at 4 °C. Supernatants were retained and placed into appropriate ultracentrifuge tubes and the pellets were rehomogenized in 0.3 mL of Buffer A and recentrifuged at 1000 g for 10 min at 4 °C. The second supernatant was collected and combined with the first supernatant and centrifuged at 100 000 g for 1 h at 4 °C. This final supernatant was collected to serve as the cytosolic fraction and the remaining pellet was resuspended in 0.2 mL of Buffer A and rehomogenized; this was the membrane fraction. The bicinchoninic acid protein assay was used to determine the protein concentration of the samples.

Tissue extraction with Buffer B

Briefly, as described previously [26,27], frontal cortex from human and mouse brain samples (0.1 g) was homogenized in 0.7 mL of Buffer B containing phosphatase and protease inhibitor cocktails (Calbiochem). Samples were centrifuged at 5000 g for 5 min at room temperature. Supernatants were retained and placed into appropriate ultracentrifuge tubes and centrifuged at 100 000 g for 1 h at 4 °C. This supernatant was collected to serve as the cytosolic fraction, and the pellets were resuspended in 0.2 mL of Buffer B and rehomogenized; this was the membrane fraction. The bicinchoninic acid protein assay was used to determine the protein concentration of the samples.

Antibodies

For immunoblot and immunohistochemical detection of Aβ, the mouse monoclonal antibodies 4G8 (Signet Laboratories, Dedham, MA, USA), 82E1 (IBL, Minneapolis, MN, USA) and 6E10 (Signet Laboratories) were used. For the analysis of synaptic proteins, mouse monoclonal antibodies against SNAP25 (Abcam, Cambridge, MA, USA), Syntaxin (Abcam) and PSD95 (UC Davis/NIH Neuro-Monoclonal Antibody Facility, Davis, CA, USA) were used. The synaptic protein VAMP1 was detected with a rabbit polyclonal IgG (Abcam), and actin levels were determined with the mouse monoclonal C4 antibody (Millipore, Temecula, CA, USA). For the detection of pan-Shank, Shank 1, Shank 3 and pan-SAPAP, mouse monoclonal antibodies from the UC Davis/NIH Neuro-Monoclonal Antibody facility were used. Homer protein was detected with a rat polyclonal antibody (Millipore). Table 4 presents a summary of the antibodies used for this study.

Table 4.   Antibodies for immunohistochemical and immunoblot analysis. aa, amino acid.
AntigenAntibodyCloneConcentrationSpecificitySource
Mouse monoclonal82E11 : 1000aa residues 1–16IBL
Mouse monoclonal6E101 : 1000aa residues 1–16Signet
Mouse monoclonal4G81 : 500aa residues 17–24Signet
Rabbit polyclonalA111 : 2000OligomersUC Irvine
Synapsin IMouse monoclonal1 : 1000C-terminusMillipore
VAMP2Mouse monoclonalVAMP1 : 1000C-terminusMillipore
Syntaxin 1AMouse monoclonalSP81 : 1000aa residues 4–190Abcam
SNAP25Mouse monoclonalSP121 : 1000Isoforms A and BAbcam
PSD95Mouse monoclonalK28/431 : 1000aa residues 77–299Antibodies Inc. (Davis, CA, USA)
Pan-ShankMouse monoclonalN23B/491 : 1000aa residues 84–309UC Davis
Shank1Mouse monoclonalN22/211 : 1000aa residues 469–691UC Davis
Shank3Mouse monoclonalN69/461 : 1000aa residues 840–857UC Davis
HomerRat monoclonal1 : 1000Homer 1a, 1b, 1c, 2a, 2b, 3Millipore
SAPAPMouse monoclonalN127/311 : 1000aa residues 772–992UC Davis
ActinMouse monoclonalC41 : 1000aa residues 50–70Millipore

Immunoblot analysis

Western blot analysis was performed essentially as described previously [26,63]. For additional details, please refer to Doc. S1.

Immunoprecipitation assays

Briefly, homogenates from human and mouse brains were prepared in Buffer A as for immunoblot analysis. Samples from the membrane fractions were diluted in immunoprecipitation buffer [20 mm Tris/HCl, pH 7.5, 150 mm NaCl, 1 mm EDTA, 1 mm NaVO4, 50 mm NaF, with protease inhibitors (Roche, Basle, Switzerland) containing 1% Triton X-100], and immunoprecipitation assays were carried out essentially as described previously [64]. The lysates were then centrifuged for 20 min at 13 000 g, and the protein concentrations were determined with a bicinchoninic acid protein assay kit. Three hundred micrograms of each of the supernatants were incubated with 1 μg of the antibody against synaptic proteins PSD95, VAMP, syntaxin or SNAP25 overnight at 4 °C. Then, the immunocomplexes were adsorbed to protein A–Sepharose 4B or protein G–Sepharose (Amersham, Piscataway, NJ, USA). After extensive washing with immunoprecipitation buffer, which contained 1% Trion X-100, samples were heated in NuPAGE SDS sample buffer (Invitrogen, Temecula, CA, USA) for 5 min and subjected to gel electrophoresis on Tris-Tricine gels, followed by immunoblot analysis with an antibody against either synaptic proteins or Aβ (6E10 or 82E1). Samples were also immunoprecipitated with an antibody against Aβ (6E10 or 82E1), separated on a Bis-Tris 4–20% gel, and then subjected to immunoblot analysis with mouse monoclonal antibodies against synaptic proteins.

Aβ ELISA

Levels of Aβ in the frontal cortex of patients diagnosed with MCI, moderate AD, advanced AD and of age-matched controls were assessed via ELISA, which was performed according to the manufacturer’s instructions (Life Technologies, Temecula, CA, USA).

Lactate dehydrogenase assay

Cell viability was evaluated by the lactate dehydrogenase assay. Cells were plated on 96-well plates in complete medium. After treatment, assays were performed following the manufacturer’s instructions (Promega, Madison, WI, USA). Results were expressed as the percentage of cell death.

Primary neuronal cultures

Please refer to Doc. S1.

Preparation and treatment with natural Aβ

Natural Aβ was prepared according to Walsh et al. [16] (kindly provided by Dr Eddie Koo) by incubating control CHO cells or CHO cells expressing the APP V717F mutation (also referred to as 7PA2 cells) with B27 conditioned medium for 16 h. Total Aβ concentration was determined as described previously [65]. Neurons were treated with 80 pm of natural Aβ for 6 and 24 h. Cells were then fixed with 4% paraformaldehyde–4% sucrose.

Double labeling and laser scanning confocal microscopy

To evaluate the co-localization between Aβ and synaptic markers, double immunohistochemical analysis was performed as described previously [26]. Vibratome sections were immunolabeled with a monoclonal antibody against PSD95 (1 : 10 000, UC Davis), detected with the Tyramide Signal Amplification™-Direct (Red) system (1 : 100; NEN Life Sciences, Boston, MA, USA), and the mouse monoclonal antibody against Aβ (clone 82E1, 1 : 500), detected with fluorescein isothiocyanate-conjugated secondary antibodies (1 : 75; Vector Laboratories, Burlingame, CA, USA) [26]. All sections were processed simultaneously under the same conditions, and the experiments were performed twice to assess reproducibility. Sections were imaged with a Zeiss 63 × (N.A. 1.4) objective on an Axiovert 35 microscope (Zeiss, Thornwood, NY, USA) with an attached MRC1024 LSCM system (BioRad, Hercules, CA, USA) [26]. To confirm the specificity of primary antibodies, control experiments were performed in which sections were incubated overnight in the absence of primary antibody (deleted) or preimmune serum and primary antibody alone.

Multidimensional protein identification technology, linear trap quadrupole mass spectrometry and analysis of tandem mass spectra

Please refer to Doc. S1.

Statistical analysis

Unless otherwise noted, all data were presented as the mean ± SEM. Mean values were compared using the Kruskal–Wallis test for Braak scores; nonparametric and one-way analysis of variance (ANOVA) tests were used for all other comparisons. If a significant global result was obtained (overall P value < 0.05), the Kruskal–Wallis test was followed by Dunn’s multiple comparison test, and ANOVA was followed by either Student–Newman–Keuls or Bonferroni’s multiple comparison tests. Pearson product moment correlations were used to determine the intragroup association of MMSE and BIMC to oligomers and synaptic proteins.

Acknowledgements

This work was supported by National Institutes of Health grants AG18440, AG5131, AG022074 and AG11385.

Ancillary