Increasing the Receptor Tyrosine Kinase EphB2 Prevents Amyloid-β-induced Depletion of Cell Surface Glutamate Receptors by a Mechanism That Requires the PDZ-binding Motif of EphB2 and Neuronal Activity

Diverse lines of evidence suggest that amyloid-β (Aβ) peptides causally contribute to the pathogenesis of Alzheimer disease (AD), the most frequent neurodegenerative disorder. However, the mechanisms by which Aβ impairs neuronal functions remain to be fully elucidated. Previous studies showed that soluble Aβ oligomers interfere with synaptic functions by depleting NMDA-type glutamate receptors (NMDARs) from the neuronal surface and that overexpression of the receptor tyrosine kinase EphB2 can counteract this process. Through pharmacological treatments and biochemical analyses of primary neuronal cultures expressing wild-type or mutant forms of EphB2, we demonstrate that this protective effect of EphB2 depends on its PDZ-binding motif and the presence of neuronal activity but not on its kinase activity. We further present evidence that the protective effect of EphB2 may be mediated by the AMPA-type glutamate receptor subunit GluA2, which can become associated with the PDZ-binding motif of EphB2 through PDZ domain-containing proteins and can promote the retention of NMDARs in the membrane. In addition, we show that the Aβ-induced depletion of surface NMDARs does not depend on several factors that have been implicated in the pathogenesis of Aβ-induced neuronal dysfunction, including aberrant neuronal activity, tau, prion protein (PrPC), and EphB2 itself. Thus, although EphB2 does not appear to be directly involved in the Aβ-induced depletion of NMDARs, increasing its expression may counteract this pathogenic process through a neuronal activity- and PDZ-dependent regulation of AMPA-type glutamate receptors.

cause its degradation (32). Neuronal overexpression of hAPP/A␤ in transgenic mice or neuronal knockdown of EphB2 in wild-type mice each impaired long term potentiation and NMDAR-mediated currents (32). Furthermore, normalizing EphB2 levels in the dentate gyrus of hAPP mice reversed some of their synaptic and cognitive impairments (32). Taken together, these findings motivated us to investigate the relationship among A␤ oligomers, EphB2, and glutamate receptors in greater detail and to determine the extent to which their interactions depend on tau, PrP C , and neuronal activity.
Recombinant A␤ Oligomers-Unless indicated otherwise, the A␤ oligomers we used to treat primary neuronal cultures were prepared from recombinant A␤ peptides, and statements made about A␤ oligomers refer to this type of A␤ assembly. In brief, hydroxyfluroisopropanol-treated recombinant A␤ peptides (␤-amyloid , ultra pure, hydroxyfluroisopropanol from rPeptide, catalogue number A1163, 0.5 mg, primary lot number 9131142H) were first dissolved in 22 l of DMSO (at ϳ2.5 mM A␤  peptides, monomer equivalent) and then further diluted with 978 l of ice-cold Neurobasal A medium to generate a ϳ50 M A␤  solution. The A␤  solution was incubated at 4°C for 24 h to oligomerize A␤  peptides. Vehicle solution was prepared by following the same protocol except for omitting addition of A␤  peptides. On the day of the experiment, the concentration of A␤  peptides was determined by bicinchoninic acid (BCA) assay (Thermo Scientific, 23225), and vehicle control or oligomerized A␤  peptides (final concentration of 2 M monomer equivalent) were applied to primary neuronal cultures.
To characterize A␤ oligomers by atomic force microscopy (AFM), A␤  peptides dissolved in 22 l of DMSO were diluted with 978 l of ice-cold Dulbecco's phosphate-buffered saline without calcium or magnesium (DPBS-no Ca 2ϩ /Mg 2ϩ ; Life Technologies, 14190-144) because Neurobasal medium contains factors that interfere with AFM analysis. When added to the medium of neuronal cultures, A␤ oligomers prepared in DPBS-no Ca 2ϩ /Mg 2ϩ depleted cell surface GluN1 and EphB2 after 2 and 48 h, respectively (data not shown), confirming their bioactivity.
Size Exclusion Chromatography (SEC) and A␤ ELISA-We collected culture medium containing A␤ oligomers 2 and 48 h after treating cultured neurons and injected it onto a Superdex 75 (10/300GL) column (GE Healthcare) calibrated using a gel filtration standard kit (Bio-Rad, . Samples were eluted with 1 column volume of phosphate-buffered saline (PBS) at a flow rate of 0.8 ml/min into 1-ml SEC fractions.
Primary Hippocampal Mouse Neurons-Unless indicated otherwise, experiments were carried out on primary hippocampal mouse neurons. Hippocampi of newborn mouse pups (P0-P1) were dissected in ice-cold Earle's balanced salt solution without CaCl 2 , MgSO 4 , and phenol red (Life Technologies, 14155). Dissected hippocampi were digested with papain (Worthington, LK003176; ϳ1 unit per hippocampus) in Earle's balanced salt solution at 37°C for 15 min and then triturated in a disposable plastic tube in low ovomucoid solution containing 1.5 mg/ml BSA (Sigma-Aldrich, A7030-10G), 1.5 mg/ml trypsin inhibitor (Sigma-Aldrich, T9253-5G), and 66.7 units/ml DNase I (Sigma-Aldrich, D5025) in DPBS (Life Technologies, 14040-182). After removing debris with a 70-m nylon strainer (BD Biosciences, 352350), neurons were spun at 1000 rpm for 5 min. Cell pellets were gently dissociated in Neurobasal A medium supplemented with 1ϫ B27 (Life Technologies, 17504-044), 2.4 mM L-glutamine (Life Technologies, 25030-081), and 100 units/ml penicillin/streptomycin (Life Technologies, 15410-122). They were then plated on poly-D-lysinecoated 12-well plates at a density of 500,000 neurons/well for Western blot analyses and 1,000,000 neurons/coverslip for proximity ligation assay. Half of the medium was replaced with new medium every week, and neurons were used for experiments at DIV 10 -14. Biotinylation and Isolation of Cell Surface Proteins-Cell surface proteins were biotinylated and isolated using a cell surface protein isolation kit (Thermo Scientific, 89881). After various treatments, primary hippocampal mouse neurons (DIV 10 -14) were washed once with ice-cold PBS and then incubated with sulfo-NHS-SS-biotin (sulfosuccinimidyl-2-(biotinamido)-ethyl-1,3-dithiopropionate; 0.25 mg/ml in ice-cold PBS) for 30 min at 4°C. After quenching the biotinylation reaction, neurons were washed twice with ice-cold Tris-buffered saline (TBS) and lysed in Pierce IP Lysis Buffer (Thermo Scientific, 87788) with Halt protease and phosphatase inhibitor mixture (Thermo Scientific, 78440). Lysates were then sonicated on ice using five 1-s bursts and centrifuged at 1000 rpm for 5 min at 4°C followed by determination of protein concentration by BCA protein assay. To isolate biotinylated surface proteins, 30 g of biotinylated total protein was incubated with NeutrAvidin gel slurry (25 l) at room temperature for 1 h followed by two washes with TBS and two additional washes with Pierce IP Lysis Buffer. Isolated biotinylated proteins were then solubilized in loading buffer for Western blot analyses.
Western For Western blot analysis of A␤ oligomers (Fig. 1F), PBS or Neurobasal medium containing 0.5 g of A␤ in 1ϫ NuPAGE lithium dodecyl sulfate sample buffer was loaded per gel lane. A␤ samples were electrophoresed on a 10 -20% Criterion Tris-Tricine gel (Bio-Rad, 3450068) in 2ϫ XT MES Running Buffer (Bio-Rad, 1610789) at 150 V for 3 h at 4°C. Gels were transferred to nitrocellulose membranes in 1ϫ NuPAGE transfer buffer (Life Technologies, NP0006-1) containing 10% methanol at 0.3 A for 2 h at 4°C. Membranes were then microwaved for 5 min in PBS for antigen retrieval, blocked in 5% BSA (Sigma-Aldrich, A3803-100G) in TBS overnight at 4°C, and incubated with a combination of two anti-A␤ antibodies (82E1 from IBL America (10323) at 1:1000 dilution and 6E10 from Covance (SIG-39320) at 1:2000 dilution) in TBS containing 5% BSA for 2 h at room temperature. Membranes were then washed with TBST four times for 5 min at room temperature and incubated with goat anti-mouse secondary antibodies conjugated to IRDye (0.1 g/ml) for 1 h at room temperature followed by washes in TBST (4 ϫ 5 min). Protein bands were visualized using an Odyssey CLx Infrared Imaging System and quantified with Image Studio software.
Generation of EphB2 Deletion Mutants-Deletion mutants of EphB2 with two FLAG tags inserted at the N-terminal side of the ligand-binding domain were generated using a QuikChange II XL site-directed mutagenesis kit (Agilent Technologies, 200522). FLAG-tagged EphB2 in pFUW plasmid (32) was used as a template in combination with the set of primers listed in Table 2. After mutagenesis, full sequences of deletion mutants were verified. Plasmids carrying the desired mutations were used to transform the Stbl3 Escherichia coli strain (Life Technologies, C7373-03) for maintenance.
Production and Purification of Lentiviral Particles-Lentiviral particles were generated by co-transfecting the transfer vector (pFUW with wild-type or mutant EphB2 cDNA insertion), the HIV-1 packaging vector (Delta8.9), and the VSVG envelope glycoprotein expression vector (pVSVG) into HEK293T cells.
Confluent HEK293T cells were transfected with three vectors (22.5 g of pFUW, 16.9 g of Delta8.9, and 11.25 g of pVSVG per 15-cm Petri dish) using CalPhos transfection reagent (Clontech, 631312) according to the manufacturer's instruction. Medium containing lentiviral particles was collected 48 h after transfection and filtered through a 0.22-m cellulose acetate filter (Corning Inc., 431154). Lentiviral particles in the medium were then concentrated by serial ultracentrifugation: 21,000 rpm for 2 h at 4°C in a Beckman SW28 and then 25,000 rpm for 2 h at 4°C in a Beckman SW55 with a sucrose cushion (2 ml of 20% sucrose in Hanks' balanced salt solution (Life Technologies, 14170) at the bottom of the SW55 tubes). Final pellets were dissolved in Hanks' balanced salt solution, aliquoted, and stored at Ϫ80°C until use. Lentiviral titers were determined with a p24 ELISA by Dr. David Chung (University of California, San Francisco). Primary cultured hippocampal neurons were transduced with lentiviral particles encoding EphB2 at 0.02 pg of p24/neuron. Lentiviral vectors encoding shRNA against GluN1 (sh-GluN1) or EphA2 were purchased from Sigma-Aldrich or GeneCopoeia, Inc., respectively. Primary cultured hippocampal neurons were transduced with these lentiviral particles at a multiplicity of infection of 2. We first compared the efficacies of five sh-GluN1 lentiviruses (Mission lentiviral transduction particles; clone IDs TRCN0000233326, TRCN0000233327, TRCN0000233328, TRCN0000233329, and TRCN0000257394) and selected two equally effective sh-GluN1 constructs (TRCN0000233326 and TRCN0000257394) for subsequent experiments. Because the results obtained with these two lentiviruses were similar (data not shown), they were combined for statistical analysis and data presentation.
Statistical Analysis-Experimenters were blinded with respect to the genotype and treatment of cell cultures. Biological units were randomized during assays, sampling, and analyses. Statistical analyses were performed with Prism (version 6, GraphPad) and R (R Development Core Team). Individual culture wells (Western blot data) or individual transfected neurons (PLA data) were treated as independent biological units (n). Differences between genotypes and treatments were assessed, as appropriate, by unpaired Student's t test with Welch's correction or by one-way or two-way ANOVA with Bonferroni multiple comparison post hoc test. The null hypothesis was rejected at p Ͻ 0.05. In all figures, quantitative data are presented as means Ϯ S.E.

A␤ Oligomers Deplete NMDARs and EphB2 from the Neuronal Surface through Independent Mechanisms-Two-hour
treatment of primary hippocampal neuronal cultures with recombinant A␤ oligomers (2 M; monomer equivalent) reduced levels of surface GluN1 (sGluN1) (Fig. 1A), an obligatory subunit of NMDARs, consistent with previous reports (7,10). At this time point, levels of surface EphB2 (sEphB2) were still unchanged (Fig. 1A). Reductions of sEphB2 levels were observed after 48 h of A␤ treatment (Fig. 1B), consistent with previous findings (12,32). Notably, 2-h treatment with A␤ oligomers reduced sGluN1 levels also in primary hippocampal neurons from mice lacking EphB2 (Fig. 1C). Thus, at least in these mouse cultures, A␤-induced depletion of sGluN1 does not depend on EphB2 or alterations in its surface levels.
We confirmed the oligomeric nature of the recombinant A␤ preparations used in these experiments by atomic force microscopy (Fig. 1D). We also removed medium from neuronal cultures 2 and 48 h after addition of A␤ oligomers and characterized the A␤ species they contained by Western blot analysis as well as by SEC and ELISA (Fig. 1, E and F). The medium contained higher levels of A␤ oligomers at 2 h than at 48 h after addition of A␤ oligomers (Fig. 1, E and F), possibly because, with time, more and more A␤ oligomers bind to neuronal surface membranes (40) and are sequestered by the formation of A␤ fibrils. Western blot analysis revealed putative A␤ oligomers even in preparations of freshly solubilized A␤ peptides (Fig. 1F), likely due in part to the fact that SDS promotes the oligomerization of A␤  in gels (41).
In light of the known interactions between NMDARs and EphB2 (33), we next examined whether depletion of NMDARs may underlie the subsequent depletion of sEphB2. We first examined whether known inhibitors of A␤-induced sGluN1 depletion block the subsequent depletion of sEphB2. ␣-Bungarotoxin (BTX), a blocker of ␣-7 nicotinic acetylcholine receptors (␣-7), has been reported to partially block A␤-induced
To evaluate the dependence of sEphB2 levels on sGluN1 levels more directly, we transduced neurons with lentiviral vectors expressing an anti-GluN1 short hairpin RNA (sh-GluN1) or a scrambled shRNA (sh-SCR). Transducing neurons with lentivirus encoding sh-GluN1 depleted both total and surface levels of GluN1 more markedly and for much longer (DIV 2-14) than the 2-h A␤ exposure, but it had no effect on levels of total or surface EphB2 (Fig. 3A and data not shown). These results suggest that depletion of sGluN1 is not sufficient to lower sEphB2 levels and that A␤ oligomers may deplete sGluN1 and sEphB2 through parallel mechanisms.
Because NMDAR activity is required for A␤ oligomers to impair neuronal functions (42), we also examined whether blocking NMDAR activity attenuates A␤-induced sEphB2 depletion. Treatment of neuronal cultures with the NMDAR antagonist AP5 (100 M) had no effect on A␤-induced sEphB2 depletion (Fig. 3B), suggesting that the A␤-induced depletion of sEphB2 does not require NMDAR activity.
A␤-induced Depletion of sGluN1 and sEphB2 Does Not Require Tau, PrP C , or Neuronal Activity-We next investigated whether potential mediators of A␤-induced neuronal impairments, specifically tau, PrP C , and aberrant excitatory neuronal activity (15-17, 19 -21, 23-27), are required for A␤-induced depletion of sGluN1 and sEphB2. Treatment with A␤ oligomers depleted sGluN1 (Fig. 4, A-F) and sEphB2 (Fig. 5, A-F) in primary hippocampal neurons from mice lacking tau or PrP C and in wild-type neurons treated with TTX, which blocks action potentials by inhibiting voltage-gated sodium channels. The extent of A␤-induced sGluN1 and sEphB2 depletion in PrP C -deficient (Figs. 4, A and B, and 5, A and B), tau-deficient (Figs. 4, A and B, and 5, A and B), and TTX-treated wild-type (Figs. 4, D and E, and 5, D and E) cultures was comparable with that in wild-type control cultures (Fig. 1, A and B). To confirm that TTX was effective in these experiments, we measured levels of phosphorylated (i.e. activated) ERK, which TTX reduced (Figs. 4G and 5G). Thus, tau, PrP C , and aberrant excitatory neu-ronal activity are unlikely mediators of the A␤-induced depletion of sGluN1 and sEphB2, at least in these culture models.
Ability of EphB2 to Prevent A␤-induced Depletion of sGluN1 Depends on Its PDZ-binding Motif-We previously showed that normalizing EphB2 levels in the dentate gyrus of hAPP transgenic mice reversed deficits in NMDAR function (32), and others recently showed that EphB2 overexpression also counteracts A␤-induced NMDAR dysfunctions in neuronal cultures (43). Overexpression of FLAG-tagged wild-type EphB2 (EphB2 WT ) also prevented A␤-induced depletion of sGluN1 in neuronal cultures in the current study (Fig. 6). In one experiment, increasing the expression of EphB2 WT in primary cultures of rat hippocampal neurons also partially counteracted the depletion of sGluN1 caused by synthetic A␤ oligomers,

EphB2 PDZ Interactions Block NMDA Receptor Depletion by A␤
which, by Western blot analysis, contained more putative higher order assemblies than recombinant A␤ oligomers (data not shown).
To begin to explore the molecular mechanism by which EphB2 exerts this protective effect, we generated multiple constructs encoding mutant forms of FLAG-tagged EphB2 (Fig. 6, A-C), including EphB2 lacking the ligand-binding domain (EphB2 ⌬LB ), the fibronectin repeats (EphB2 ⌬FN ), the SAM domain (EphB2 ⌬SAM ), the PDZ-binding motif (EphB2 ⌬PDZ ), or kinase activity (EphB2 K661R ). Western blot analyses of wildtype neuronal cultures transduced with the different EphB2 constructs confirmed the expected changes in molecular weight and interaction with the anti-FLAG antibody (Fig. 6B). Although the EphB2 antibody used in this study is polyclonal and recognized all EphB2 mutants we generated, it showed relatively less reactivity with EphB2 ⌬FN , which was recognized readily by the anti-FLAG antibody (Fig. 6B). Kinase-deficient EphB2 K661R in which the lysine 661 residue critical for tyrosine kinase activity was mutated to arginine failed to mediate eph-rinB2-induced phosphorylation of the tyrosine 1472 residue (Tyr-1472) on GluN2B (Fig. 6C), confirming the desired impact of the mutation.
Except for EphB2 ⌬PDZ , all EphB2 mutants and EphB2 WT prevented A␤-induced sGluN1 depletion when overexpressed in primary cultures of hippocampal neurons (Fig. 6D). As documented in Figs. 1C and 6D (Empty), the endogenous EphB2, which was coexpressed in these cultures, would not be expected to modulate the A␤-induced depletion of sGluN1. It is also worth noting that deletion of the PDZ-binding motif did not alter cell surface expression, localization in spines, and ephrindependent clustering of FLAG-tagged EphB2 (44), suggesting that the lack of protective effect by EphB2 ⌬PDZ was not caused by alterations in its subcellular localization.
Over the time frame of this experiment, levels of total GluN1 were not significantly altered by treatment with A␤ oligomers or expression of EphB2 (Fig. 6E). The degree of EphB2 overexpression we achieved by lentiviral transduction of neurons was variable across constructs as determined by Western blot analysis of neuronal lysates (Fig. 6, F and G), possibly reflecting differences in the reactivity of EphB2 mutants with the anti-EphB2 antibody or in their stability. For most constructs, levels of sEphB2 correlated well with those of total EphB2 (tEphB2); however, sEphB2 ⌬LB levels were low despite high levels of tEphB2 ⌬LB expression (Fig. 6, F and G). Surface levels of EphB2 ⌬LB and likely EphB2 ⌬FN may have appeared lower than they actually were because these constructs lack large portions of the extracellular part of the molecule (Fig. 6A) and thus might have been less efficiently biotinylated and/or pulled down. Regardless, these two EphB2 mutants were able to counteract A␤-induced sGluN1 depletion.
Notably, overexpressing EphB2 did not prevent A␤-induced GluN1 depletion when neuronal cultures were treated with TTX (Fig. 7, A and B), suggesting that neuronal activity is required for EphB2 overexpression to counteract this A␤ effect. Treatment of cultures with A␤ oligomers or TTX and overexpression of EphB2 WT did not significantly alter total GluN1 levels (Fig. 7, A and C). Independently of whether cultures were treated with A␤ oligomers, TTX, or both, overexpression of EphB2 WT increased levels of total EphB2 and sEphB2 roughly 4 -6-fold over endogenous EphB2 levels found in cultures transduced with control virus (Fig. 7, A, D, and E). We again confirmed the efficacy of TTX in this experiment by monitoring phospho-ERK levels (Fig. 7, A and F).
Ability of EphB2 to Counteract the Effect of A␤ Oligomers May Depend on Its Interaction with GluA2-EphB2 can phosphorylate NMDARs through its tyrosine kinase activity and can directly bind to them via its extracellular domains (33). However, overexpression of EphB2 mutants lacking kinase activity or specific extracellular domains still prevented A␤-induced depletion of surface GluN1, which makes it unlikely that the protective ability of EphB2 is mediated by direct effects on NMDARs. Therefore, we focused on GluA2, which can become indirectly associated with the PDZ-binding motif of EphB2 through direct interactions with PDZ domain-containing proteins (44,45). Additionally, endocytosis of GluA2 is required for A␤ to depress NMDAR currents and synaptic function (9).   (data not shown). D--G, cultures of primary hippocampal neurons (DIV 10 -14) from wild-type mice were treated with A␤ oligomers or vehicle for 2 h followed by Western blot analysis of sGluN1, tGluN1, phosphorylated (p) ERK, and total (t) ERK levels. Some cultures were treated with TTX (1 M) 30 min before and throughout exposure to A␤ or vehicle. Phospho-ERK levels were normalized to total ERK levels. D, representative Western blot. E-G, quantitations of sGluN1 (E), tGluN1 (F), and phospho-ERK (G) levels. For ERK levels, the sum of two bands (ERK1 and -2) was quantitated. n ϭ 16 wells per condition from four independent experiments. Two-way ANOVA revealed an interaction between the effects of A␤ and TTX on phospho-ERK levels (p Ͻ 0.05) but not on sGluN1 levels (p ϭ 0.98). One-way ANOVA revealed a significant (p Ͻ 0.001) TTX effect on phospho-ERK levels. *, p Ͻ 0.05; **, p Ͻ 0.01 versus vehicle by unpaired t test with Welch's correction (B) or Bonferroni test (E). Bars and error bars represent means and S.E., respectively.

EphB2 PDZ Interactions Block NMDA Receptor Depletion by A␤
Treatment of neuronal cultures with A␤ oligomers decreased surface levels of the AMPAR subunit GluA2 (Fig.  8A), consistent with previous reports (9,13,14). sGluA2 was depleted by A␤ within 2 h, which is similar in time frame to the depletion of sGluN1 and much faster than the depletion of sEphB2. Interestingly, EphB2 overexpression counteracted the A␤-induced depletion of sGluA2 through a mechanism that depended on the presence of the PDZ-binding motif of EphB2 (Fig. 8A). To monitor interactions between EphB2 and GluA2 in primary neuronal cultures, we used a PLA that allows for the in situ detection of two antigens only when they are in close proximity (Ͻ40 nm) (38).
At comparable levels of overexpression, EphB2 WT showed more colocalization with GluA2 than EphB2 ⌬PDZ indepen- JANUARY 22, 2016 • VOLUME 291 • NUMBER 4 JOURNAL OF BIOLOGICAL CHEMISTRY 1727 dently of whether cultures were treated with vehicle or A␤ (Fig.  8, B-F). Treatment with A␤ oligomers reduced the colocalization of EphB2 and GluA2 in neurons overexpressing EphB2 WT or EphB2 ⌬PDZ (p Ͻ 0.01 by two-way ANOVA). TTX treatment had no effect on colocalization between GluA2 and endogenous EphB2 or overexpressed EphB2 WT but increased colocalization between GluA2 and overexpressed EphB2 ⌬PDZ (Fig. 8F), possibly due to increased surface GluA2 resulting from synaptic scaling (46). For unclear reasons, TTX treatment reduced the levels of overexpressed EphB2 WT but not EphB2 ⌬PDZ (Fig. 8D). This difference may have obscured some effects of the mutation, for example, resulting in the detection of comparable levels of interaction between GluA2 and EphB2 WT versus EphB2 ⌬PDZ in the presence of TTX (Fig. 8F). GluA2 levels were not altered by overexpression of EphB2 WT or EphB2 ⌬PDZ under control conditions (Fig. 8E). GluA2 levels were reduced by A␤ in neurons that were transfected with empty vector or vector encoding EphB2 ⌬PDZ but not in those overexpressing EphB2 WT (Fig. 8E). Taken together, these results suggest that EphB2 overexpression may counteract A␤-induced depletion of sGluN1 by increasing sGluA2 levels.

EphB2 PDZ Interactions Block NMDA Receptor Depletion by A␤
Lastly, we tested and refuted the hypothesis that overexpression of any protein bearing a PDZ-binding motif can counteract the effect of A␤ oligomers on cell surface glutamate receptors. For this purpose, we focused on EphA2, another receptor tyrosine kinase with a PDZ-binding motif (47). To our knowledge, EphA2 has not been demonstrated to interact with or regulate glutamate receptors. Overexpression of EphA2 in cultured neurons did not prevent A␤-induced depletion of sGluN1 (Fig. 9).

Discussion
This study demonstrates that the ability of EphB2 to counteract A␤-induced depletions of AMPARs and NMDARs depends on its PDZ-binding motif and the presence of neuronal activity. We also obtained evidence suggesting that this effect may involve interactions between the PDZ-binding motif of EphB2 and GluA2, which could promote the retention of GluA2 at the surface membrane and prevent A␤-induced depletion of surface NMDARs (9,14). From a therapeutic per-spective, it is important to note that these protective EphB2 effects were observed only when EphB2 was expressed at supraphysiological levels and that they did not depend on its kinase activity.
We also found that A␤-induced depletions of NMDARs do not depend on EphB2 depletion, tau, PrP C , or aberrant neuronal activity, all of which have been implicated in A␤-induced neuronal dysfunction (12, 15-17, 20, 21, 23-27, 31, 32, 43). One interpretation of these findings is that NMDAR depletion occurs upstream of the other factors within the pathogenic cascade that A␤ oligomers trigger. However, we found that A␤ oligomers depleted EphB2 even in the presence of cyclosporin or AP5, which prevented A␤-induced depletion of NMDARs. An alternative possibility is that A␤ oligomers activate parallel pathways that affect neuronal functions through distinct mechanisms.
It should be noted in this context that all our findings were obtained in dissociated primary neuronal cultures. We cannot exclude the possibility that the relationships among the factors we studied are different in vivo. As is true for most data obtained in experimental models, the relevance of our findings to patients with AD also remains uncertain and deserves to be further explored in future studies. Notwithstanding these caveats, the novel mechanistic insights our study provides could guide the development of strategies to counteract A␤-induced neuronal dysfunction and help make the brain more resistant against pathogenic A␤ assemblies.
A␤ oligomers deplete and dysregulate glutamate receptors and related molecules, including EphB2 (7, 9 -12, 14, 32). However, it has remained uncertain whether these alterations are causally linked with each other and, if so, in which sequence or constellation. Our results suggest that A␤-induced endocytosis of GluA2 acts upstream of and may promote the depletion of GluN1, consistent with previous studies showing that GluA2 endocytosis is required for A␤-induced NMDAR dysfunction and synaptic depression (9,14). Interestingly, the A␤-induced depletions of GluA2 and GluN1 could be prevented by neuronal overexpression of EphB2 WT but not EphB2 ⌬PDZ . EphB2 . The same Western blot was probed with a polyclonal anti-EphB2 antibody (top) and a monoclonal anti-FLAG antibody (middle) and labeled with secondary antibodies conjugated to distinct fluorophores. Wild-type and mutant EphB2 bands are indicated by black and red arrowheads, respectively. The faint bands between 75 and 100 kDa in the EphB2 blot are probably nonspecific as they were also present in primary neurons from EphB2-deficient mice (data not shown). C, substitution of Lys-661 for arginine (K661R) specifically disrupts the kinase activity of EphB2. HEK293T cells were transiently transfected with plasmids encoding EphB2 WT or GluN2B or co-transfected with GluN2B plus EphB2 WT or EphB2 K661R . Two days after the transfection, cells were treated for 1 h at 37°C with preclustered human Fc-EphrinB2 (500 ng/ml) followed by Western blot analysis for GluN2B, GluN2B phosphorylated at tyrosine 1472 (GluN2B p1472 ), EphB2, and ␣-tubulin. D--G, cultures of primary hippocampal neurons (DIV 2) from wild-type mice were transduced with different EphB2 constructs as in B. At DIV 12-14, neurons were treated with A␤ oligomers (2 M) or vehicle (Veh; 0.08% DMSO) for 2 h followed by quantification of sGluN1 (D), tGluN1 (E), tEphB2 (F), and sEphB2 (G) levels by Western blot analysis. Representative Western blots are shown on the left, and quantitations of Western blot signals are shown on the right. The red bars represent the boundary between different membranes. Two representative Western blots with duplicate samples for some EphB2 constructs are shown. n ϭ 15-33 wells per condition from four to nine independent experiments. In D, overexpression of each EphB2 construct, except for EphB2 ⌬PDZ , prevented the A␤-induced depletion of sGluN1. In E and F, the ␤III-tubulin blots are the same. In F and G, bands of EphB2 WT , EphB2 K661R , EphB2 ⌬SAM , and EphB2 ⌬PDZ overlap with that of endogenous EphB2 at ϳ120 kDa, whereas bands of EphB2 ⌬LB and EphB2 ⌬FN reside below the faint nonspecific band around 80 kDa (see also B). Wild-type and mutant EphB2 bands are indicated by black and red arrowheads, respectively. Quantitations of EphB2 levels represent the sum of endogenous (ϳ120 kDa) and exogenous (WT or mutant) EphB2 signals. Two-way ANOVA revealed a significant (p Ͻ 0.01) interaction between the effects of A␤ treatment and EphB2 transduction on sGluN1 levels (D), significant effects of EphB2 constructs (p Ͻ 0.0001) but not of A␤, and no significant interaction between EphB2 constructs and A␤ (F and G). **, p Ͻ 0.01; ***, p Ͻ 0.001; ****, p Ͻ 0.0001 versus vehicle condition (D) or empty condition (F and G) by Bonferroni test. Bars and error bars represent means and S.E., respectively. regulates trafficking of GluA2 by a mechanism that depends on its PDZ-binding motif: the PDZ-binding motif of EphB2 binds to the PDZ domain-containing scaffold protein glutamate receptor-interacting protein 1, which binds to GluA2 and regulates the localization of GluA2-containing AMPARs (33,44,45). Using a PLA approach, we confirmed that this motif is required for the association of EphB2 with GluA2.
In contrast, the PDZ-binding motif does not appear to be involved in interactions between EphB2 and NMDARs. Instead, EphB2 can regulate NMDAR localization and function by directly binding to GluN1 through its extracellular domain and phosphorylating GluN2B through its kinase activity (33,39,48). However, although the kinase activity of EphB2 can influence NMDAR localization and function in vitro (39), NMDAR impairments in EphB2-deficient mice could be normalized by expression of kinase-deficient EphB2 (49), suggesting a more limited or different role for this activity in vivo.
In the current study, the prevention of A␤-induced GluN1 depletion by overexpression of EphB2 was independent of the kinase activity of EphB2 as well as of its ligand-binding domain and fibronectin repeats, which together make up most of the extracellular domain. In our view, the most parsimonious interpretation of these findings is that overexpression of EphB2 prevents A␤-induced NMDAR depletion by PDZ-binding motifmediated retention of GluA2 at the surface membrane, which would be expected to counteract A␤-induced enhancement of GluA2 endocytosis and consequent GluN1 depletion (9).
Because this protective effect of EphB2 was observed only when it was overexpressed and A␤ oligomers depleted sGluN1 to similar degrees in wild-type and EphB2-deficient neurons, it is likely that the association between endogenous EphB2 and GluA2 is not strong enough to retain GluA2 at the surface, at least in the presence of pathologically elevated levels of A␤ oligomers. However, because genetic ablation of EphB2 during early stages of develop- Treatments are shown below the bar graphs; whether A␤ oligomers or TTX (ϩ) or vehicle (Ϫ) was applied is indicated. n ϭ 9 -10 wells per condition from three independent experiments. A, representative Western blot. tERK, total ERK. B-F, quantitation of Western blot signals for sGluN1 (B), tGluN1 (C), sEphB2 (D), tEphB2 (E), and phosphorylated (p) ERK (F). In B, two-way ANOVA revealed no significant interaction between the effects of A␤ and TTX on sGluN1 levels in cultures transduced with empty (p ϭ 0.49) or EphB2 WT (p ϭ 0.09)-expressing lentivirus. For the empty condition in F, two-way ANOVA revealed a significant effect of TTX (p Ͻ 0.001) and of A␤ (p Ͻ 0.01) and significant (p Ͻ 0.01) interaction. n.s., not significantly different; **, p Ͻ 0.01; ***, p Ͻ 0.001 versus vehicle condition or as indicated by brackets (Bonferroni test). Bars and error bars represent means and S.E., respectively. ment may engage compensatory mechanisms (32,50), it remains possible that endogenous EphB2 counteracts the A␤-induced depletion of sGluN1 in wild-type neurons of the adult brain.
It is interesting that neuronal activity was required for overexpression of EphB2 to prevent A␤-induced GluN1 depletion even though neuronal activity did not influence the association Neurons were transduced with different EphB2 constructs and treated with A␤ as described in Fig. 6, D-G, followed by quantitation of sGluA2 levels by Western blot analysis. Two representative Western blots are shown on the left, and quantitations of Western blot signals are shown on the right. n ϭ 15-18 wells per condition from three to four independent experiments. Two-way ANOVA revealed a significant (p Ͻ 0.05) interaction between the effects of A␤ treatment and EphB2 transduction on sGluA2 levels. B-F, putative interactions between EphB2 and GluA2 in individual transfected neurons were monitored with a PLA in which fluorescence signal above background indicates close proximity (Ͻ40 nm) between EphB2 and GluA2. Cultures of primary hippocampal neurons (DIV 7) from wild-type mice were transfected with empty pFUW plasmid (Empty) or with pFUW plasmid encoding EphB2 WT or EphB2 ⌬PDZ . GFP-encoding plasmid was co-transfected to visualize transfected neurons. Some cultures were treated with A␤ oligomers (2 M) or vehicle (Veh; 0.08% DMSO) for 24 h, and others were treated with TTX (1 M) or vehicle for 6 h prior to fixation. Fixed cultures were analyzed by PLA and immunostained for EphB2 or GluA2. B, for each treatment and transfection condition, PLA signals are shown on the left, EphB2 immunoreactivity is shown in the middle, and GFP fluorescence is shown on the right. GFP fluorescence was observed only in transfected neurons. EphB2 images were also thresholded to identify primarily EphB2-transfected neurons, which displayed stronger EphB2 immunoreactivity than untransfected neurons expressing only endogenous EphB2. Scale bar, 100 m. C, GluA2 immunostaining for each treatment and transfection condition. Scale bar, 100 m. D and E, relative intensity of EphB2 (D) and GluA2 (E) immunofluorescence in cell bodies of GFP-positive neurons. n ϭ 16 -59 neurons per condition from two independent experiments. F, relative PLA signal indicating EphB2-GluA2 colocalization. EphB2-GluA2 interactions were 1) increased when EphB2 was overexpressed, 2) reduced by A␤, and 3) partly dependent on the PDZ-binding motif of EphB2. n ϭ 51-85 neurons per condition from four independent experiments. Two-way ANOVA revealed significant effects of EphB2 constructs (p Ͻ 0.0001), A␤ (p Ͻ 0.05), and TTX (p Ͻ 0.05) and no significant interactions between EphB2 constructs and A␤ or TTX. n.s., not significantly different; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; ****, p Ͻ 0.0001 versus vehicle (A) or versus control (Ctrl) condition or as indicated by brackets (D-F) by Bonferroni test. Bars and error bars represent means and S.E., respectively.

EphB2 PDZ Interactions Block NMDA Receptor Depletion by A␤
between EphB2 and GluA2. These results suggest that neuronal activity plays a role downstream of GluA2 engagement, possibly through homeostatic synaptic scaling (46,51).
In another potential pathway, A␤ oligomers were proposed to serially activate ␣-7 and striatal enriched protein tyrosine phosphatase (STEP 61 ), leading to dephosphorylation of GluN2B by STEP 61 and depletion of GluN1 (7). In contrast to results obtained by Snyder et al. (7), the ␣-7 antagonist BTX did not prevent A␤-induced GluN1 depletion in the current study. In both studies, high concentrations (10 M) of BTX were used as compared with its subnanomolar IC 50 (52). Therefore, variable degrees of ␣-7 blockade are unlikely to account for the discrepancy. In a similar vein, although the calcineurin inhibitors FK506 and cyclosporin A have been reported to prevent APP/A␤-induced NMDAR dysfunction (7,9), FK506 was ineffective in our study, and cyclosporin A prevented A␤-induced GluN1 depletion during the first 2 h as reported previously (7) but not after 48 h of A␤ exposure. Differential effects of these calcineurin inhibitors (53) and differences in experimental systems (organotypic slice culture versus primary neuronal culture and APP overexpression versus exposure to recombinant A␤(1-42) oligomers) may explain the different results. Notably, both genetic ablation (10) and pharmacological blockade (7) of ␣-7 blocked A␤-induced GluN1 depletion only partially. Based on our results and those obtained by others (9), we suspect that enhanced GluA2 endocytosis contributes to the A␤-induced depletion of NMDARs more strongly than activation of ␣-7 and STEP 61 .
Although exposure to A␤ oligomers depleted both sEphB2 and sGluN1 and the A␤-induced depletion of GluN1 could be prevented by overexpression of EphB2, the depletion of GluN1 did not depend on the depletion of EphB2, suggesting that A␤ oligomers deplete EphB2 and GluN1 through independent mechanisms. Notably, the A␤ oligomers generated from recombinant A␤  in the current study depleted surface, but not total, EphB2 even when we used them at high concentration (up to 5 M) or exposed cultures to them for longer periods (up to 6 days) (data not shown). These results are consistent with those obtained in a previous study using similar A␤ oligomers (12) but differ from those we obtained with naturally secreted A␤ oligomers from 7PA2-CHO cells that depleted both surface and total EphB2 (32). The discrepancy could be due to differences in the A␤ preparations used. Although increasing the expression of EphB2 counteracted the depletion of sGluN1 by recombinant or synthetic A␤ oligomers that dif-fered in their content of putative higher order assemblies, it remains possible that overexpression of EphB2 protects against the detrimental effects of some types of A␤ oligomers but not others.
Our findings that A␤ oligomers depleted both NMDARs and AMPARs in neuronal cultures are consistent with those of previous studies (7,9,10,12,14,32,54,55,57). We also demonstrated that overexpression of EphB2 rescued both types of glutamate receptors in these culture models. In contrast, electrophysiological recordings in acute hippocampal slices from hAPP-J20 mice have so far revealed primarily deficits in the function of NMDARs but not of AMPARs (32,57), and normalization of neuronal EphB2 levels in the dentate gyrus of these mice appeared to specifically rescue NMDAR function (32). These discrepancies may be due to differences in the (a) duration of exposure to elevated A␤ levels (several months in hAPP-J20 mice versus 2 or 48 h in the current study), (b) APP metabolites (mixture of A␤ peptides and other APP metabolites produced by neurons in brain versus recombinant A␤  added to cell culture medium), (c) cell types exposed to A␤ (diverse populations of mature neurons, glia, and endothelial cells in hAPP-J20 mice versus primary cultures enriched for young hippocampal neurons), and (d) duration and extent of EphB2 expression (several months of normalized levels in hAPP-J20 mice versus 10 -12 days of overexpression in cultured hippocampal neurons). Thus, the mechanisms by which EphB2 normalization reversed functional deficits in hAPP-J20 mice may well have been at least partly different from those by which EphB2 overexpression prevented depletion of surface glutamate receptors in the current study.
Although many A␤-induced effects on neuronal integrity and function depend on the presence of tau (15-17, 58 -66), others do not. In the current study, tau was not required for A␤ oligomers to deplete GluN1 and EphB2. Tau also does not appear to be required for the A␤-induced loss of dendritic spines (60,67). Given the roles of GluN1 and EphB2 in the formation and maintenance of dendritic spines (33,68), it is tempting to speculate that the depletion of these molecules promotes spine loss.
Another factor that has been implicated in A␤-induced neuronal dysfunction is PrP C . However, different groups have obtained perplexingly disparate results in regard to this potential mediator. Indeed, PrP C has been shown to be required for hAPP/A␤ to impair neuronal functions in vitro (24, 25, 69 -71) and in vivo (27,71). In contrast, other groups have identified various hAPP/A␤-induced neuronal impairments that do not depend on the presence of PrP C (28 -30, 71, 72). In the current study, PrP C ablation did not prevent A␤-induced sGluN1 depletion, contrary to results obtained by Um et al. (25). Although differences between the specific A␤ assemblies used in these studies may explain the discrepancy (56), our findings clearly demonstrate that some A␤ oligomers can deplete surface NMDARs independently of any mediation by PrP C . Conceivably, different types of A␤ oligomers cause neuronal dysfunctions by engaging different mediators. Further adding to this complexity, our findings suggest that even the same type of A␤ oligomers can impair neuronal functions through distinct pathways that, at a minimum, involve GluA2/GluN1 and EphB2.
Author Contributions-T. M. designed, conducted, and analyzed all the experiments and wrote the manuscript. D. K. characterized A␤ oligomers by AFM. J. A. K. contributed to the proximity ligation assay shown in Fig. 8, B-F. E. J. contributed to the characterization of A␤ oligomers. L. M. supervised the study and wrote the manuscript. All authors reviewed the results and approved the final version of the manuscript.