Abstract
Synaptic dysfunction during early stages of Alzheimer’s disease (AD) is triggered by soluble amyloid-β (Aβ) oligomers that interact with NMDA receptors (NMDARs). We previously showed that Aβ induces synaptic protein loss through NMDARs, albeit through undefined mechanisms. Accordingly, we here examined the contribution of individual NMDAR subunits to synaptotoxicity and demonstrate that Aβ exerts differential effects on the levels and distribution of GluN2A and GluN2B subunits of NMDAR in dendrites. Treatment of cultured hippocampal neurons with Aβ1-40 (10 μM, 1 h) induced a significant increase of dendritic and synaptic GluN2B puncta densities with parallel decreases in the puncta densities of denritic and synaptic pTyr1472-GluN2B. Conversely, Aβ significantly decreased dendritic and synaptic GluN2A and dendritic pTyr1325-GluN2A puncta densities and increased synaptic pTyr1325-GluN2A puncta densities. Unexpectedly, Aβ treatment resulted in a significant reduction of GluN2B and pTyr1472-GluN2B protein levels but did not influence GluN2A and pTyr1325-GluN2A levels. These results show that Aβ exerts complex and distinct regulatory effects on the trafficking and phosphorylation of GluN2A and GluN2B, as well as on their localization within synaptic and non-synaptic sites. Increased understanding of these early events in Aβ-induced synaptic dysfunction is likely to be important for the development of timely preventive and therapeutic interventions.
Keywords
INTRODUCTION
Alzheimer’s disease (AD) is an age-related neurodegenerative disorder, characterized by a gradual loss of memory and other cognitive functions. Increasingly, evidence suggests that the cognitive decline observed in AD subjects reflects subtle synapticdysfunction rather than neuronal loss per se [1, 2]. Early synaptic dysfunction in AD has been attributed to soluble Aβ oligomers [3, 4] acting in a glutamate-dependent manner [5 –8]. The main species of Aβ found in amyloid deposits are Aβ1-40 and Aβ1-42 [9]. In human brain, the most abundant peptide is Aβ1-40 [10 –12], which has been implicated in synaptic loss in AD patients [13 –15]. Soluble Aβ oligomers are thought to exert synaptotoxic effects by binding near, or to, ionotropic N-methyl-D-aspartate receptors (NMDARs) [16], which play an essential role in mediating excitatory neural transmission, synaptic plasticity and glutamate-induced excitotoxicity by allowing the passage of Ca2 + into neurons.
NMDARs exist as heteromeric complexes, and include two obligatory GluN1 and two modulatory GluN2 (GluN2A and GluN2B) subunits; subunit composition determines NMDAR sensitivity and electrophysiological properties [17, 18], as well as Aβ-induced neurotoxicity [19]. The composition of synaptic NMDAR subunits changes during development, switching from predominantly GluN2B to GluN2A receptors [20]. The GluN2A subunit is primarily found at synaptic sites and is characterized by rapid offset kinetics, while the GluN2B subunit is preferentially located at extrasynaptic sites and displays slower kinetics [21 –23]. Many recent studies have demonstrated that NMDARs can exchange and diffuse laterally between synaptic and non-synaptic sites [24] under the modulatory influence of various factors, including scaffold proteins (e.g., PSD-95, SAP102) [25]. Dysregulated NMDAR subunit trafficking has been suggested as a mechanism that contributes to synaptic dysfunction in AD [26]. Importantly, post-translational modifications of NMDARs, such as tyrosine phosphorylation, are known to regulate NMDAR expression [26, 27]and function [26, 28]. To date, only GluN2 subunits have been demonstrated to be subject to tyrosine phosphorylation [29]; although phosphorylation of the Tyr 1325 and Tyr 1472 sites within the GluN2A and GluN2B subunits, respectively, are considered to be crucial events in NMDAR activation [30, 31], there is still little clarity regarding the precise role of each Tyr-phosphorylated GluN2 subunit in the execution of Aβ-induced synaptotoxicity [26 , 30–32].
Here, we demonstrate the differential regulation of GluN2A and GluN2B, the phosphorylation of that GluN2ATyr1325 and GluN2BTyr1472 by Aβ1-40 at synaptic sites of hippocampal neurons; Aβ also downregulated the total levels of GluN2B/ GluN2BTyr1472/Fyn proteins. Importantly, these changes were observed after just 1 h exposure to Aβ.
MATERIALS AND METHODS
Chemicals and reagents
Aβ1–40 (Cat.#A1075), bovine serum albumin (BSA, Cat.# A7030), sodium borate, gelatin, poly-D-lysine (PDL), and trypsin inhibitor were obtained from Sigma. Antibodies to the following proteins were purchased from: Sigma (rabbit anti-phospho-NMDA NR2B(pTyr1472), Cat.#M2442; rabbit anti-SAP102, Cat.#S4687; mouse anti-β-actin, Cat.#A1978), Millipore (rabbit anti-NR1, Cat.#AB9864; rabbit anti-NR2A, Cat.#AB1555P; rabbit anti-NR2B, Cat.#AB1557P), BD Biosciences Pharmingen (mouse anti-synapsinI, Cat.#611392), Synaptic Systems (rabbit anti-SAP102, Cat.#124213). Rabbit anti-NMDAR2A (phospho Y1325) (Cat.# ab16646 and Cat.#ab106590) were from Abcam. Rabbit anti-Fyn (Cat.#4023) and mouse anti-Src (L4A1) (Cat.#2110) were from Cell Signaling. Anti-mouse and anti-rabbit sera, conjugated to horseradish peroxidase (HRP) were obtained from Bio-Rad. Alexa Fluor® 594-conjugated goat anti-mouse secondary antibody and Alexa Fluor® 488-conjugated goat anti-rabbit secondary antibody were purchased from Invitrogen.
Primary hippocampal neurons culture
Primary cultures of rat hippocampal cells were prepared from postnatal day 4 (P4) Wistar rats as described previously [15 , 34]. Trypsin dissociated cells were plated at a density of 450,000 cells/mm2 on gelatin/PDL coated glass coverslips. Cultures were maintained in Neurobasal medium supplemented with B27, containing Glutamax I, kanamycin and basic fibroblast growth factor (bFGF) (all materials from Invitrogen). Cells maintained in a humidified incubator (95% air, 5% CO2) at 37°C. Half of the medium was replenished every 3 days during the culture period. Cells were treated with Aβ and monitored after 7 days in vitro (DIV 7).
Soluble Aβ1-40 oligomers preparation and treatment
Soluble Aβ1-40 oligomers were prepared as described previously [15 , 36]. Briefly, Aβ1-40 peptide was dissolved in 2 mM DMSO, diluted in PBS, vortexed for 30 min, and centrifuged (1 h, 15,000 g, 4°C). The predominant aggregates in such preparations are reported to be low N-oligomers (mainly monomeric to tetrameric) [9 , 37–39]. It should be noted thatfibrillogenesis requires longer incubation times and higher concentrations (>10 μM) of the peptide [40, 41]. In our present experiment, the oligomer nature of Aβ1-40 in such preparation was also confirmed with atomic force microscopy (AFM) (Fig. 1).
The chosen dose (10 μm) and duration of exposure (1 h) to Aβ in this study was based on our previous observation that these parameters all detection of early synaptotoxicity [33].
Immunofluorescence
Cells were fixed in ice cold 4% paraformaldehyde, washed, permeabilized (0.3% Triton X-100 in PBS) and blocked. Specimens were incubated (overnight, 4°C) with primary antibodies against GluN1 (1:1000), GluN2A (1:1000), GluN2B (1:1000), Synapsin I (1:4000), SAP102 (1:1000), GluN2A/Tyr1325 (1:200), and GluN2B/pTyr1472 (1:500). Immunoreactive antigens were visualized after staining with Alexa Fluor® 594- conjugated goat anti-mouse IgG (1:500) or Alexa Fluor® 488-conjugated goat anti-rabbit IgG (1:500). Nuclear staining was achieved using Hoechst 33342 (1:1000). Optical sections and stacks of images from fluorescence labeled cells were captured with a Leica confocal laser scanning microscope (TCS SP5, with a plan apochromat 63×/1.4 numerical aperture oil lens) or a Leica fluorescence microscope (DM5000).
Puncta analysis
For quantification, 5–10 neurons from 5 different batches of cultures and experiments were randomly chosen for evaluating puncta density (L) (puncta number/10 μm length dendrite close to the soma). Images (PDM) were captured for co-localization analysis using Image J software (NIH), and SynPAnal software [42] was used to analyze puncta density (L). A suitable threshold was set and maintained for images across the whole data set; the threshold was chosen to maximize the number of puncta while excluding noise, followed by selection of the appropriate color channel. Areas of interest (10 μm length of dendrite proximal to the soma) were selected and SynPAnal software was used to automatically compute puncta densities (L), based on the specified threshold settings. The software was able to distinguish whether individual pixels was part if a given puncta, and could distinguish between adjacent puncta.
In this study, the puncta densities of NMDARs on 10 μm length of dendrite close to the soma were evaluated. NMDARs that colocalized with synapsin I in dendrites were classified as synaptic NMDARs; all NMDARs found on dendrites were considered to be dendritic NMDARs.
Western blot
Primary hippocampal cells were briefly sonified in Tris HCl (50 mM, pH 7.4), containing 5 mM EDTA, 1% Triton X-100, 0.1% sodium dodecyl sulphate (SDS), protease inhibitors, and 150 mM NaCl, before clearing by centrifugation. Protein concentrations were determined using the DC protein assay (BioRad). After boiling, samples were electrophoresed (8% SDS polyacrylamide) and transferred onto nitrocellulose membranes. After blocking, membranes were incubated overnight at 4°C with the following primary antibodies GluN1 (1:1000), GluN2A (1:1000), GluN2B (1:1000), SAP102 (1:500), GluN2A/Tyr1325 (1:1000), GluN2B/pTyr1472 (1:1000), Fyn (1:1000), Src (1:1000), β-actin (1:4000), before incubation with HRP-conjugated anti-mouse or anti-rabbit antibodies (1:3000). Protein bands were visualized with ECL Plus (Amersham Biosciences), and signals were quantified using ImageJ software. Band density values were normalized to β-actin after subtracting background in each region of interest, and then to values from the corresponding controls.
Statistical analysis
Data are depicted as mean ± S.E.M. (5 indepen-dent experiments). Statistical analysis involved Independent-Sample t-test. A level of p < 0.05 was considered to be statistically significant.
RESULTS
Aβ acutely downregulates synaptic GluN2A
Aβ1-40 is one of the two most abundant species of amyloid deposits in the brain of Alzheimer’s patients. Aβ1-40 can stably assemble into soluble low molecular oligomers [43], and the soluble prefibrillar oligomers are recognized as the main species responsible for synaptic dysfunction [44, 45].
In the present experiment, Aβ1-40 was diluted in culture medium and incubated at 37°C for 1 h. The oligomeric nature of Aβ1-40 used to treat neurons was confirmed by AFM (Fig. 1). Spheroid or globular oligomers of similar size were detected. With section analysis, the average Z height of the probe was 3.1 nm, corresponding to the predicted size of soluble oligomers [46].
To quantify synaptic NMDAR puncta, double-labeling with antibodies against NMDAR and synapsin I were used. Synaptic and non-synaptic NMDARs puncta were distinguishable with the aid of Image J and SynPAnal softwares (Fig. 2). NMDARs were detected in the cytoplasm, dendritic shafts and spines. Those NMDAR clusters in dendrites that were not colocalized with the presynaptic marker synapsin I were considered to represent non-synaptic NMDARs.
Treatment of neurons with Aβ1-40 (10 μM, 1 h) resulted in differential distributions of NMDARs at synaptic and non-synaptic sites; the treatment also altered total protein levels of the different NMDARs (Table 1).
When immunocytochemical methods were used to monitor GluN2A puncta density, our results show that Aβ1-40 (10 μM, 1 h) treatment induced a significant reduction puncta densities of dendritic GluN2A (Table 1, Fig. 3A, D; Aβ, 160.3 ± 3.4 versus control, 209.6 ± 3.7 puncta/10 μm dendritic length, p < 0.01) and synaptic GluN2A in 10 μm dendritic length close to the soma (Table 1, Fig. 3C, D; Aβ, 30.5 ± 1.5 versus control, 71.7 ± 2.5 puncta/10 μm dendritic length, p < 0.01). SynapsinI (Fig. 3B, D; Aβ: 176.1 ± 16.4 versus control: 190.2 ± 16.4 puncta/10 μm dendritic length, p > 0.05) puncta density was not affected by exposure to Aβ. At the same time, the treatment does not alter total GluN2A protein levels in immunoblots of cell lysates (Table 1, Fig. 3E; Aβ: 110.5 ± 30.3% of control, p > 0.05).
Differential regulation of GluN2B and SAP102 by Aβ
The GluN2B subunit of the NMDAR is highly expressed at birth but its levels gradually diminish with age, as the expression of the GluN2A subunit increases [47 –49]. Like GluN2A, this subunit binds glutamate and it is widely considered that the ratio of GluN2A:GluN2B subunits shifts during development and aging [50, 51]. The expression of GluN2B and GluN2A show similar levels of decline in AD [52 –54]. During synaptogenesis, GluN2B trafficking is facilitated by SAP102 [55].
In our present study, Aβ induced a significant increase of puncta densities of dendritic GluN2B (Table 1, Fig. 4A, D; Aβ: 194.2 ± 9.7 versus control: 169.6 ± 7.7 puncta/10 μm dendritic length, p < 0.05) and synaptic GluN2B (Table 1, Fig. 4C, D; Aβ: 119.2 ± 7.9 versus control: 70.5 ± 5.6 puncta/10 μm dendritic length, p < 0.05). In addition, Aβ-induced decreases in GluN2B protein levels in cell lysate were found by immunoblotting (Table 1, Fig. 4F; 83.8 ± 4.7% of control, p < 0.05).
Aβ led to a significant reduction in SAP102 puncta density (Fig. 4E; Aβ, 123.3 ± 2.5 versus control, 163.2 ± 2.5 puncta/10 μm dendritic length, p < 0.01) when dendrities of Aβ-treated neurons were examined. At the same time, Aβ did not influence SAP102 protein expression levels (Fig. 4G; 99.9 ± 12.8% of control, p > 0.05), as measured by immunoblotting. In contrast to the increased puncta densities of dendritic or synaptic GluN2B, the presently observed reduction of SAP102 puncta density in dendrites suggests that SAP102 does not directly affect anchoring of NR2B at synaptic sites.
Aβ leads to differential phosphorylation of GluN2A (pTyr1325) and GluN2B (pTyr1472)
Tyrosine phosphorylation is thought to influence the trafficking of GluN2 subunits [56, 57]. Immunocytochemical analysis revealed that Aβ treatment leads to a significant downregulation of dendritic pTyr1325-GluN2A puncta density (Table 1, Fig. 5A, D; Aβ: 170.8 ± 1.9 versus control: 185.5 ± 2.8/10 μm dendritic length, p < 0.05) and a significant increase in synaptic pTyr1325-GluN2A puncta density (Table 1, Fig. 5C, D; Aβ: 136.2 ± 2.8 versus control: 109.8 ± 2.5/10 μm dendritic length, p < 0.01).
Comparing to pTyr1325-GluN2A, Aβ downregulates puncta densities of both dendritic (Table 1, Fig. 6A, D; Aβ: 179.0 ± 3.1 versus control: 199.0 ± 3.4/10 μm dendritic length, p < 0.05) and synaptic (Table 1, Fig. 6C, D; Aβ: 126.2 ± 2.3 versus control: 141.5 ± 2.5, p < 0.01) pTyr1472-GluN2B.
At the same time, immunoblotting revealed that exposure of hippocampal neurons to 10 μM Aβ1-40 for 1 h leads to a significant decrease in the total protein levels of phosphorylated GluN2B (Tyr residue 1472; 84.6 ± 2.6% of control, p < 0.05) (Table 1, Fig. 7B), but not of phosphorylated GluN2A (Tyr residue 1325; 94.7 ± 3.0% of control, p > 0.05) (Table 1, Fig. 7A).
In light of previous studies that showed tyrosine phosphorylation of GluN2A and GluN2B to be mediated by Src and Fyn, respectively [58], we monitored the protein levels of these kinases by immunoblotting to corroborate the changes seen in Fig. 7A and B. Consistent with the data shown in Fig. 7A and B, Aβ treatment did not alter Src expression (Fig. 7C; 101.9 ± 0.4% of control, p > 0.05) and resulted in a significant decrease in Fyn levels (Fig. 7D; 89.2 ± 0.9% of control, p < 0.05).
Aβ treatment does not influence GluN1 subunit expression
The GluN1 is an obligatory component of NMDAR; each NMDAR receptor complex contains two GluN1 subunits which bind the glutamate co-agonist,gylcine.
Exposure of primary hippocampal neurons to 10 μM Aβ1-40 oligomers for 1 h did not alter dendritic and synaptic GluN1 puncta densities (Table 1, Fig. 8A, D; dendritic GluN1, Aβ: 174.9 ± 3.1 versus control:187.3 ± 6.18 puncta/10 μm dendritic length, p > 0.05; Fig. 8C, D; synaptic GluN1, Aβ: 112.7 ± 3.1 versus control: 115.1 ± 8.3 puncta/10 μm dendritic length, p > 0.05) or GluN1 protein levels of expression (Table 1, Fig. 8E; 92.4 ± 18.6% of control, p > 0.05), as evaluated by the complementary methods of immunocytochemistry and immunoblotting.
DISCUSSION
The important role of Aβ oligomers in synaptic dysfunction and neuronal loss in AD is indisputable. In addition, there is strong evidence that the neurotoxic actions of Aβ depend on activated NMDAR [19, 59], whose expression is altered in AD patient brains [60 –64]. We and others have shown that synaptic dysfunction is an early event induced by Aβ oligomers [15, 33]. However, the exact contribution of specific NMDAR subunits or subpopulations to AD pathology, especially during early stages of the disease, is unclear. This is an especially important issue in light of an increasing recognition that the pathological processes underpinning AD may precede clinical symptoms by many years [65, 66]. The latter suggests the existence of an approximately 50-year window of opportunity during which preventive or delay strategies can be initiated, well before the appearance of clinically-detectable symptoms [66]. The present in vitro experiments used cultured hippocampal cells that were obtained from early postnatal rats (aged 4 days), which after DIV7 display mature networks and functional synapses. Mature neurons in culture are characterized by extensive dendritic arborization and expression of “mature” GluN1/GluN2A receptors puncta in the vicinity of dendrite spines, mostly colocalized with synaptic proteins [22 , 67–69]. Here, we found that NMDA receptors are mainly clustered in the soma and proximal dendritic shaft, and that GluN2A subunits are abundantly located at non-synaptic sites (i.e. not colocalized with the synaptic protein, synapsin I).
GluN2A and GluN2B show dynamic changes, and both are critical regulators of NMDAR localization and function, with significant roles in learning and memory [70, 71]. Excitatory synaptic transmission is tightly regulated by the number of synaptic NMDAR, and Aβ is known to modulate synaptic NMDAR activity [72 –74]. Our finding supports the view that Aβ induces a shunting of GluN2A and GluN2B between synaptic and non-synaptic membrane sites and/or the cytoplasmic compartment of neurons [35] from early stages of Aβ pathology. Accordingly, the altered localization of GluN2A and GluN2B at dendrites in close proximity to the soma may be a key event in the early stages of AD.
The present results show that puncta densities of dendritic and synaptic GluN2A are significantly reduced after Aβ treatment. Since GluN2A are normally stabilized at the synapse by binding to the scaffold protein PSD-95 [23], the observed reduction in GluN2A after Aβ may be explained by our previous finding that Aβ decreases PSD-95 puncta density [33]. The loss of synaptic GluN2A would result in lower spine concentrations of Ca2 + and thus, decreased synaptic strength and efficacy [23]. Previous work has shown that tyrosine phosphorylation is an important mechanism for the potentiation of NMDAR activity [75]. The present results, showing decreased puncta density of dendritic pTyr1325-GluN2A is consistent with reduced dendritic GluN2A puncta density under control conditions. The finding that Aβ1-40 treatment increases synaptic pTyr1325-GluN2A puncta density is therefore likely to represent an early response. On the other hand, the lack of change in total protein levels (western blotting) may indicate a redistribution of GluN2A and pTyr1325-GluN2A away from synaptic sites and toward to non-synaptic site or other subcellular compartments (e.g., distal dendrites, cytoplasm, membrane of the cell body, axons). Further, a contribution of glial GluN2A/pTyr1325-GluN2A [76] to the total levels of this protein measured cannot be excluded.
Besides, our results that Aβ1-40 rapidly downregulates NR2B protein level is consistent with that of some reports on the effects of soluble Aβ oligomers in in vitro and in vivo experiments [77, 78]. The Aβ-induced loss of GluN2B is likely to affect synaptic plasticity [61, 79] and may explain why chronic treatment with the GluN2B antagonist piperidine18 does not protect against synapse loss in an Aβ-overexpressing mouse model of AD [80]. At the same time, we unexpectedly found increased puncta densities of dendritic and synaptic GluN2B; these may reflect a compensatory response to the reductions of total GluN2B levels following exposure to Aβ, adding support for the idea that dendrites that are close to the soma are particularly functionally sensitive to Aβ.
In addition to reducing GluN2B protein levels, Aβ treatment of hippocampal neurons also reduced the protein levels of pTyr1472-GluN2B, as well as dendritic and synaptic puncta densities. Phosphorylation of the Tyr1472 epitope is reported to be associated with enrichment of synaptic GluN2B [81]. The present inconsistent changes of increased GluN2B and decreased pTyr1472-GluN2B puncta densities, may suggest that phosphorylation of Tyr1472-GluN2B is not required for translocation of GluN2B to the synapse [81], and the related regulation mechanism between GluN2B and phosphorylation of Tyr1472 may be more complex at synaptic sites. In addition, it should be noted that the postsynaptic GluN2B is a substrate for Fyn [82], and that phosphorylation of GluN2B at Tyr1472 strengthens the interaction between GluN2B and PSD-95 [83], a mechanism that links NMDAR signaling with Aβ toxicity. Further, tau protein, which plays an important role in the pathology of AD, was recently shown to be involved in the postsynaptic targeting of Fyn in dendrites [84, 85], while the accumulation of hyperphosphorylated tau within dendritic spines disrupts NMDAR anchoring and trafficking [86]. Whether the reductions of Fyn and pTyr1472-GluN2B found here are related to alterations in tau hyperphosphorylation and relocation from axons to dendrites is a question for future investigations. Meanwhile, it would appear from the present results that reductions of GluN2B-pTyr1472 -GluN2B-Fyn protein levels precede the trafficking of hyperphosphorylated tau to dendritic spines.
This study also examined the potential role of SAP102, an important postsynaptic scaffolding protein, in mediating the actions of Aβ. SAP102 is highly expressed in adult human brain where it plays an important role in synaptic function, plasticity and regulation of NMDAR trafficking [2, 87]. Like glutamate receptors, postsynaptic scaffold proteins (e.g., PSD-95, SAP102) are also dynamically altered [88]. We here found that Aβ does not lead to significant alterations in the levels of SAP102 protein; however, Aβ did result in significantly smaller puncta density in dendrites. The non-tandem alterations in GluN2B and SAP102 likely suggests that SAP102 does not play a direct role in anchoring GluN2B at synaptic sites, although previous studies indicated preferential interactions between GluN2B and SAP102 [55, 89]. Notably, a recent study revealed a non-scaffolding role for SAP102 in clearing GluN2B from synapses [90]; thus, the decreased SAP102 puncta density in dendrites may lead to the reduced clearance of GluN2B from synapse sites. On the other hand, the finding that total SAP102 protein levels are unchanged could indicate that SAP102 relocalizes to the cytoplasm where it participates in GluN2B trafficking.
In conclusion, our findings demonstrate that Aβ exerts complex and distinct regulatory effects on the trafficking and phosphorylation of GluN2A and GluN2B, as well as on their localization within synaptic and non-synaptic sites. Increased understanding of these early events in Aβ-induced synaptic dysfunction is likely to be important for the development of timely preventive and therapeutic interventions.
Footnotes
ACKNOWLEDGMENTS
This work was supported by the Program for New Century Excellent Talents in University (NCET-10-0015), the Importation and Development of High-Caliber Talents Project of Beijing Municipal Institutions (CIT&TCD201304175), Beijing Natural Science Foundation (5152004), National Natural Science Foundation of China (81301100, 81200968), and the Science and Technology Development Project of Beijing Education Committee (SQKM201210025003), as well as the EU-funded project SwitchBox (Contract 259772).
