Abstract
NMDA receptors (NMDARs) are key mediators of glutamatergic transmission and synaptic plasticity, and their dysregulation has been linked to diverse neuropsychiatric and neurodegenerative disorders. While normal NMDAR function requires regulated expression and trafficking of its different subunits, the molecular mechanisms underlying these processes are not fully understood. Here we report that the amyloid precursor protein intracellular domain associated-1 protein (AIDA-1), which associates with NMDARs and is encoded by ANKS1B, a gene recently linked to schizophrenia, regulates synaptic NMDAR subunit composition. Forebrain-specific AIDA-1 conditional knock-out (cKO) mice exhibit reduced GluN2B-mediated and increased GluN2A-mediated synaptic transmission, and biochemical analyses show AIDA-1 cKO mice have low GluN2B and high GluN2A protein levels at isolated hippocampal synaptic junctions compared with controls. These results are corroborated by immunocytochemical and electrophysiological analyses in primary neuronal cultures following acute lentiviral shRNA-mediated knockdown of AIDA-1. Moreover, hippocampal NMDAR-dependent but not metabotropic glutamate receptor-dependent plasticity is impaired in AIDA-1 cKO mice, further supporting a role for AIDA-1 in synaptic NMDAR function. We also demonstrate that AIDA-1 preferentially associates with GluN2B and with the adaptor protein Ca2+/calmodulin-dependent serine protein kinase and kinesin KIF17, which regulate the transport of GluN2B-containing NMDARs from the endoplasmic reticulum (ER) to synapses. Consistent with this function, GluN2B accumulates in ER-enriched fractions in AIDA-1 cKO mice. These findings suggest that AIDA-1 regulates NMDAR subunit composition at synapses by facilitating transport of GluN2B from the ER to synapses, which is critical for NMDAR plasticity. Our work provides an explanation for how AIDA-1 dysfunction might contribute to neuropsychiatric conditions, such as schizophrenia.
Introduction
NMDARs are ligand and voltage-gated heterotetrameric ion channels composed of two obligatory GluN1 subunits and a combination of GluN2 and/or GluN3 subunits (Traynelis et al., 2010; Paoletti et al., 2013). Their expression, trafficking, and composition at synapses are critical for normal synaptic function and behavior. In the forebrain, NMDAR transmission is mediated primarily through GluN2A-subunit-containing and GluN2B-subunit-containing NMDARs (Paoletti et al., 2013; Sanz-Clemente et al., 2013), and the specific subunit GluN2B has been implicated in diverse neuropsychiatric disorders (Loftis and Janowsky, 2003; Poot et al., 2011; Moghaddam and Javitt, 2012; Paoletti et al., 2013; Snyder and Gao, 2013; Mota et al., 2014). The precise composition of NMDAR subunits at synapses is spatiotemporally controlled, giving rise to receptor assemblies with distinct functional properties (Cull-Candy and Leszkiewicz, 2004; Paoletti et al., 2013). However, the molecular mechanisms governing the trafficking and subunit composition of NMDARs at synapses are not fully understood.
NMDAR subunits are synthesized in the endoplasmic reticulum (ER; Wenthold et al., 2003a). Through unknown mechanisms, assembled receptors are exported from the ER and routed to dendritic Golgi outposts via microtubule-dependent transport (Jeyifous et al., 2009). GluN2B-containing NMDARs are transported to synaptodendritic compartments by the molecular motor KIF17 via an evolutionarily conserved tripartite protein complex composed of Ca2+/calmodulin-dependent serine protein kinase (CASK), mammalian lin-7 protein 1 (MALS1), and munc-18-interacting 1 (Mint1; Setou et al., 2000). The membrane associated guanylate kinase (MAGUK) family member SAP97 also associates with CASK to facilitate ER export and delivery of NMDARs to synapses (Jeyifous et al., 2009).
Amyloid precursor protein intracellular domain associated-1 protein (AIDA-1) is an adaptor protein previously implicated in Cajal body (Xu and Hebert, 2005) and nucleolar function (Jordan et al., 2007). It is expressed throughout the brain, including the hippocampus, and is among the most abundant proteins at postsynaptic densities (PSDs; Jordan et al., 2004, 2007; Dosemeci et al., 2007; Jacob et al., 2010), where it binds to the postsynaptic scaffolding protein PSD95 and associates with NMDARs (Jordan et al., 2007). As with NMDARs and their associated proteins (Lau and Zukin, 2007; Paoletti et al., 2013), recent studies have linked the AIDA-1 gene, ANKS1B, to neuropsychiatric disorders, including schizophrenia (McClay et al., 2011a,b; Fromer et al., 2014; Purcell et al., 2014) and autism spectrum disorders (ASDs; Pinto et al., 2010; Hu et al., 2011; Meehan et al., 2011; Uddin et al., 2014). However, how or whether AIDA-1 affects synaptic or NMDAR function is unknown.
In this study, we generated AIDA-1 conditional knock-out (cKO) mice to determine whether AIDA-1 regulates synaptic and NMDAR function. We found that AIDA-1 regulates NMDAR subunit composition and NMDAR-mediated transmission at hippocampal synapses by affecting the delivery of GluN2B-containing NMDARs to synapses. Our findings uncover a new role for AIDA-1 in NMDAR synaptic physiology and introduce a novel component of the molecular machinery that regulates the subunit-specific trafficking of NMDARs to synapses.
Materials and Methods
All experiments were approved by and were in compliance with the Albert Einstein College of Medicine Institutional Animal Care and Use Committee. All rats and mice were housed and handled at the Albert Einstein College of Medicine.
Generation of AIDA-1 cKO mice.
AIDA-1fl/fl mice were generated by homologous recombination using 129SVEV/C57BL6 hybrid embryonic stem (ES) cells. A targeting vector (Fig. 1A) was constructed by flanking the first exon encoding the phosphotyrosine-binding (PTB) domain with loxP sequences. For homologous recombination, the long arm and short arm (6.6 kb upstream and 3.6 kb downstream DNA sequences from the PTB exon, respectively) were included. The linearized vector was injected into 129SVEV/C57BL6 hybrid ES cells and F1 progeny were generated at InGenious Targeting Laboratories (State University of New York, Stonybrook). Mice were genotyped at 3 weeks [postnatal day (P) 21] by PCR screening of DNA samples from tail snips using specific primers: WT (forward, 5′-TCT CCC CAG AAA TCC CAC AGT G; reverse, 5′-TAC TTC AGA ATG ACA GCA GCC CGC); floxed AIDA-1 (forward, 5′-GGG GAC TGG GGA GAA CCT TCC AT; reverse, 5′-CCG GAA CAC GGC ATC AG). To generate forebrain-specific AIDA-1 cKO mice, we crossed AIDA-1fl/fl mice with previously characterized CaMKIIα-Cre transgenic mice (Jackson Labs, stock #005359). We screened AIDA-1 cKOs by PCR using the following: Cre (forward, 5′-TAA AGA TAT CTC ACG TAC TGA CGG TG; reverse, 5′-TCT CTG ACC AGA GTC ATC CTT AGC); recombined AIDA-1fl/fl (forward: 5′-GGC TCT GGG GAT GCT TTT TAT G; reverse: 5′-ATG TCC CCA GAA TCA TAC GCC). We confirmed loss of AIDA-1 protein by Western blotting at P35 or later, unless otherwise stated (Fig. 1C,D).
Antibodies.
AIDA-1 mouse monoclonal antibodies were generated by Abmart to the following epitopes based on the rat AIDA-1 amino acid sequence: KRILASLGDR (clone: 1A11), PIGHDGYHPT (clone: 2K9), and KSVQIDPSEQ (clone: 2B22). Hybridoma cell lines were grown in advanced hybridoma medium (Invitrogen), antibodies purified from media using protein G columns and each used at 0.5 μg/ml final concentration. Additional antibodies used were as follows: rabbit anti-GluN2A and anti-GluN2B N-terminal antibodies (Alomone Labs; 1:1000); rabbit anti-GluN2A (H-54, 1:200, for immunocytochemistry; Santa Cruz Biotechnology); mouse anti-GluN1 N-terminal (Millipore Bioscience Research Reagents; 1:1000); rabbit anti-GluA1 (Millipore; 1:1000); rabbit anti-GluA1 N-terminal (Calbiochem; 1:200); mouse anti-GluA2 N-terminal [University of California (UC) Davis, NeuroMab Facility; 1:1000]; mouse anti-PSD95 (clone K28/43, UC Davis, NeuroMab Facility; 1:5000); mouse anti-CaMKIIα (Abcam; 1:5000); mouse anti-CASK and mouse anti-SAP97 (UC Davis, NeuroMab Facility; 1:2000); goat anti-KIF17 (Santa Cruz Biotechnology; 1:200); rabbit anti-calnexin (Genescript; 1:5000); anti-Golgi 58K protein (Abcam; 1:500).
Cell culture, transfection, transduction, and immunocytochemistry.
Primary cortical and hippocampal cultures were prepared from Sprague Dawley rat embryos at embryonic day 18 as previously described (Osten et al., 2000). Transfection of primary neuronal cultures [at 7–11 d in vitro (DIV)] or HEK cells was performed with vectors for the following: mKATE2-AIDA-1 (57 kDa isoform), eGFP-tagged GluN2A (Addgene plasmid 45445), GluN2B (Addgene plasmid 45447; Barria and Malinow, 2002), CASK (gift from Brett Abrahams, Albert Einstein College of Medicine), or YFP-KIF17 (gift from William Green, University of Chicago) using Lipofectamine 2000 (Invitrogen) following the manufacturer's instructions. Lipofectamine-transfected neurons (see Fig. 6E) were imaged at DIV 10–14. AIDA-1-specific shRNAs were cloned into pTRIP lentiviral vectors. The shRNA sequences were as follows: shAIDA#1 (5′-GCAATGTTATGGAAGATCA-3′), shAIDA#2 (5′-GCTCAACTGCCCCGTATTT-3′), and scrambled shNT (5′-CGTAAGCTGAGATAACTGA-3′). Lentiviruses were generated by triple transfection of pTRIP, psPAX2, and pMD.G into 293FT cells (Invitrogen) and infectious particles purified from the media. Neurons were transduced at 7–8 DIV using a multiplicity of infection of 0.5 and incubated for 5–14 d. Transduced neurons were processed for immunocytochemistry using standard conditions at DIV 16–21. Images were analyzed using ImageJ software.
Hippocampal slice and primary neuronal culture electrophysiology.
Acute transverse hippocampal slices (400 μm thick) were prepared from 5–6-week-old WT and AIDA-1 cKO mice of either gender. Briefly, hippocampi were isolated in a cutting solution containing the following (in mm): 215 sucrose, 2.5 KCl, 26 NaHCO3, 1.6 NaH2PO4, 1 CaCl2, 4 MgCl2, 4 MgSO4, and 20 glucose. Thirty minutes after sectioning, slices were incubated in an artificial CSF (ACSF) recording solution containing the following (in mm): 124 NaCl, 2.5 KCl, 26 NaHCO3, 1 NaH2PO4, 2.5 CaCl2, 1.3 MgSO4, and 10 glucose equilibrated with 95% O2/5% CO2, pH 7.4. The slices were incubated in this solution for ≥30 min before recordings.
All experiments, except where indicated, were performed at 28 ± 1°C in a submersion-type recording chamber perfused at ∼1–2 ml/min rate with ACSF supplemented with the GABAA receptor antagonist picrotoxin (PTX; 100 μm). Extracellular field potentials (fEPSPs) were recorded with a patch pipette filled with 1 mm NaCl and placed in the CA1 stratum radiatum (∼100 μm deep). EPSCs were recorded from CA1 pyramidal neurons voltage-clamped at −60 and +40 mV using patch-type pipette electrodes (∼3–4 MΩ) containing the following (in mm): 131 Cs-gluconate, 8 NaCl, 1 CaCl2, 10 EGTA, 10 glucose, 10 HEPES, pH 7.2, 292 mmol/kg osmolality. fEPSPs and EPSCs were evoked by stimulating Schaffer collateral inputs with a monopolar electrode filled with ACSF and positioned ∼100–150 μm away from the recording pipette. Miniature excitatory currents (mEPSCs) were recorded at 30°C from CA1 pyramidal neurons voltage clamped at −60 mV in ACSF containing 100 μm PTX and 0.5 μm TTX. mEPSCs were identified using a minimal threshold amplitude (≥5 pA) and analyzed using 10 min continuous recordings. Typically, series resistance (range, 8–15 MΩ) was monitored throughout the experiment with a −5 mV, 80 ms voltage step, and cells that exhibited significant change in series resistance (>20%) were excluded from analysis.
For recording of primary hippocampal cultures (see Fig. 5D,E), the extracellular and intracellular solutions were as described above. Whole-cell recordings were performed in shNT-expressing and shAIDA#2-expressing neurons voltage clamped at +40 mV in ACSF supplemented with NBQX (5 μm) and PTX (50 μm), to block AMPA/kainate and GABA receptors, respectively. Synaptic currents were obtained by electrical stimulation focally applied through a patch-type pipette filled with ACSF and placed close to the cell body of the recorded neuron (<50 μm). Each experiment has been repeated in ≥6 independent batches of cultures transfected on DIV 7 and recorded at DIV 12–21.
Pharmacological agents (Zn2+ and Ro 25-6981) were bath applied after establishment of a stable baseline (∼10 min), and their effects were measured after responses reached a new steady state (typically >15 min). Drugs were obtained from Sigma-Aldrich (Zn2+, PTX) and Tocris Bioscience (Ro 25-6981), except NBQX, which was obtained from the National Institute of Mental Health's Chemical Synthesis and Drug Supply Program. Stock solutions were prepared in water or DMSO (PTX, NBQX) and added to the ACSF as needed. Total DMSO in the ACSF was maintained at <0.1%. For Zn2+ experiments (see Fig. 4B), 10 mm tricine (N-[tris(hydroxymethyl)methyl]glycine) was added to the ACSF to buffer endogenous Zn2+. In the tricine-ACSF solution, NaHCO3 was increased up to 27 mm and the pH was adjusted with NaOH. A total extracellular concentration of free Zn2+ of 200 nm was estimated by adding 40 μm ZnCl2 to the tricine-ACSF.
Long-term potentiation (LTP) was induced after 15 min of stable baseline by high-frequency stimulation (HFS; see Fig. 6A) or by theta-burst stimulation (TBS; see Fig. 6B). HFS consisted of four 100 Hz, 100-stimuli bursts delivered 20 s apart. TBS consisted of a series of 10 bursts of five stimuli (100 Hz within the burst, 200 ms interburst interval), which was repeated four times (5 s apart). Long-term depression (LTD) was induced by low-frequency stimulation (900 pulses at 1 Hz; NMDAR-LTD; see Fig. 6C) or by transient (5 min) bath application of 50 μm (S)-3,5-dihydroxyphenylglycine, a group I mGluR agonist (mGluR-LTD; see Fig. 6D). The magnitude of LTP or LTD was compared between WT and cKO genotypes 35–40 min after induction. All electrophysiological recordings and analyses were done blind to the genotype of the mice.
All recordings were performed using a Multiclamp 700A amplifier (Molecular Devices), elicited at 20 s intervals, filtered at 2.2 kHz, and acquired at 5 kHz using a custom-made software written in Igor Pro 4.09A (Wavemetrics). NMDAR-EPSC decay kinetics was fitted with a single exponential function using OriginPro 8.6 software (OriginLab).
Biochemical fractionation—PSDs, ER, and Golgi.
PSD fractions were prepared as previously described using “the short method” (Jordan et al., 2004). Golgi and ER membranes were prepared following the Alternate Protocol 2 outlined by Graham (2001). Pellets from isolated fractions were resuspended in 2% SDS, 50 mm Tris, pH 7.4, buffer containing a protease inhibitor mixture, and resolved by SDS-PAGE and Western blotting. Golgi and ER fractions were identified using anti-58K Golgi protein antibody and anti-calnexin antibody, respectively.
Immunoprecipitation.
P42–P45 mice were killed using CO2 and their hippocampi were quickly dissected and Dounce homogenized 20 times in lysis buffer containing the following: 25 mm Tris, pH 7.4, 150 mm NaCl, 1% Tx-100, and 0.1% SDS containing protease inhibitor mixture (Thermo Scientific) and 2 mm EDTA. Lysates were briefly sonicated and incubated at 4°C with rocking for 1 h and then centrifuged at 20,000 × g for 30 min. Five hundred micrograms of lysate were incubated with 2 μg of primary antibody overnight at 4°C and immune complexes recovered with protein G agarose preblocked with 5% BSA/PBS. Immunoprecipitations were then processed by standard methods and Western blots analyzed by densitometry using ImageJ. For comparative quantification (see Fig. 6A,B), signal from the immunoprecipitation lane was normalized to the input signal.
Statistical analysis.
An unpaired, two-tailed t test was used for comparison of two groups. p < 0.05 was considered statistically significant. For >2 groups, a one-way ANOVA was used, followed by a post hoc unpaired, two-tailed t test with Tukey's correction for multiple comparisons. For electrophysiological experiments, statistical analysis was performed using one-way ANOVA at the p < 0.05 significance level in OriginPro 8.6 software (OriginLab). Unless otherwise stated, data are presented as mean ± SEM and illustrated traces are averages of 20–31 responses. For all figures, p values from comparisons where a statistically significant difference was found are indicated as follows: *p < 0.05, **p < 0.01, and ***p < 0.001.
Results
Generation of AIDA-1 cKO mice
To examine AIDA-1 function, we generated forebrain and neuron-specific AIDA-1 knock-out mice (AIDA-1 cKO). AIDA-1 is encoded by the exceptionally large ANKS1B gene (1.2 megabase). While multiple AIDA-1 splice isoforms have been reported (Ghersi et al., 2004a,b; Xu and Hebert, 2005; Jordan et al., 2007), the PTB domain is conserved across all known isoforms. Therefore, we flanked the exon encoding a portion of the PTB domain with loxP sequences and mated floxed AIDA-1 mice (AIDA-1fl/fl) with CaMKIIα-Cre mice for postnatal forebrain and neuron-specific ablation (Fig. 1A). CaMKIIα-driven Cre expression is observed as early as 2 weeks postnatally and reaches mature levels by P28 (Tsien et al., 1996), which enables the study of AIDA-1 function in neurons at later developmental stages, thereby mitigating concerns regarding compensatory developmental effects. Mice were genotyped at 3 weeks of age by PCR using specific primers for WT and floxed AIDA-1 alleles, and Cre recombinase (Fig. 1B). Western blot analysis using newly generated AIDA-1 monoclonal antibodies (see Materials and Methods) confirmed a progressive reduction in AIDA-1 levels in the forebrain of cKO mice from P7 to P35 (Fig. 1C), which is well after Cre has reached peak levels (Tsien et al., 1996). Interestingly, AIDA-1 is barely detectable at P7 but increases steadily through P35, showing that its expression is developmentally regulated. We confirmed that homozygous, AIDAfl/flCaMKIIαCre (cKO) mice had a 80–90% reduction, while heterozygous, AIDAfl/wtCaMKIIαCre (HET) mice had a 30–50% reduction in AIDA-1 protein expression in the hippocampus at P35 (Fig. 1D). AIDA-1 cKO mice appeared grossly normal and were indistinguishable from littermate controls.
Loss of AIDA-1 selectively alters synaptic NMDAR subunit composition
AIDA-1 is highly expressed in the hippocampus (Jacob et al., 2010) and is enriched at PSDs in dendritic spines (Jordan et al., 2004, 2007; Dosemeci et al., 2007; Jacob et al., 2010). To characterize the effect of AIDA-1 deletion on excitatory synaptic transmission, we monitored Schaffer collateral to CA1 pyramidal cell synapses (Sch–CA1) in acute hippocampal slices from AIDA-1 cKO mice and littermate controls. We found that input–output function (Fig. 2A) and paired-pulse ratio (Fig. 2B) were unaltered in AIDA-1 cKO mice. Furthermore, we observed no differences in the amplitude or frequency of AMPAR-mediated mEPSCs (Fig. 2C). These results strongly suggest that loss of AIDA-1 does not affect AMPAR-mediated synaptic transmission or glutamate release probability.
Since AIDA-1 associates with NMDARs (Jordan et al., 2007), we characterized NMDAR-mediated synaptic transmission. We analyzed NMDAR/AMPAR ratios by recording evoked AMPAR-mediated postsynaptic currents (EPSCs) at −60 mV and NMDAR-mediated EPSCs at +40 mV in the presence of 10 μm NBQX (Fig. 3A), and also analyzed NMDAR-EPSC current–voltage (I–V) plots (Fig. 3B). Both NMDAR/AMPAR ratios and I–V plots were indistinguishable between genotypes. However, NMDAR-EPSCs showed significantly faster decay kinetics in AIDA-1 cKO mice compared with WT controls (Fig. 3C), which could reflect a change in synaptic NMDAR subunit composition (Cull-Candy and Leszkiewicz, 2004; Paoletti et al., 2013). To test this possibility, we used Ro25-6981 (500 nm) and low doses of Zn2+ (200 nm; see Materials and Methods) to selectively block GluN2B-containing and GluN2A-containing NMDARs, respectively. We found that NMDAR-EPSCs in AIDA-1 cKO mice were less sensitive to Ro25-6981 (Fig. 4A) but more sensitive to Zn2+ (Fig. 4B) than littermate controls.
This change in sensitivity to Ro25-6981 and low doses of Zn2+ suggests a decrease in the abundance of GluN2B-containing NMDARs and a corresponding increase in GluN2A-containing NMDARs at synapses in AIDA-1 cKO mice. To confirm this we isolated hippocampi from cKO mice and WT littermates and measured protein abundance using SDS-PAGE and Western blotting. However, we found no changes in the overall abundance of NMDAR or AMPAR subunits or other synaptic components in whole hippocampal lysates (Fig. 4C), suggesting that transcription, translation, or degradation of these proteins were unaltered in AIDA-1 cKO mice. In contrast, we observed an increase in the abundance of GluN2A subunits and a decrease in GluN2B subunits in biochemically isolated PSDs from hippocampi of AIDA-1 cKO mice compared with controls (Fig. 4D). There was also a small but significant reduction of PSD-95 in AIDA-1 cKO mice. Levels of synaptic GluN1 or AMPAR subunits GluA1 and GluA2 were unchanged (Fig. 4D).
To validate whether loss of AIDA-1 alters NMDAR subunit composition, we used lentiviral-delivered shRNAs to knockdown AIDA-1 expression in rat primary hippocampal neuronal cultures. We characterized two AIDA-1 targeting shRNAs (shAIDA#1 and shAIDA#2), either of which robustly downregulated AIDA-1 expression compared with a nontargeting shRNA control (shNT; Fig. 5A). We then performed immunocytochemistry to quantify the abundance of GluN2 subunits at synaptic junctions. As expected, knockdown of AIDA-1 decreased GluN2B and increased GluN2A distribution at synapses as assessed by colocalization with the synaptic marker Shank (Fig. 5B,C). Both AIDA-1-targeting shRNAs led to similar results compared with shNT, suggesting that off-target effects were unlikely (Fig. 5B).
We also recorded from neurons transduced with lentiviral shRNAs as identified by GFP expression and found that the decay time constant of NMDAR-mediated EPSCs (Fig. 5D) as well as the sensitivity to Ro25-6981 (Fig. 5E) were decreased in neurons expressing shAIDA#2 compared with shNT neurons. Phenotypes observed in primary neuronal cultures suggest that impaired NMDAR transmission in AIDA-1 cKO mice is not due to compensatory mechanisms. Altogether, both cKO and shRNA approaches strongly support an essential role for AIDA-1 in the regulation of GluN2A/GluN2B subunit composition at hippocampal synapses.
Loss of AIDA-1 impairs NMDAR-dependent synaptic plasticity at hippocampal Sch–CA1 synapses
An increase in GluN2A/GluN2B ratio has been associated with a higher threshold for the induction of NMDAR-dependent LTP (NMDAR-LTP; Yashiro and Philpot, 2008; Paoletti et al., 2013). Consistent with this notion, NMDAR-LTP induced either by HFS (Fig. 6A) or TBS (Fig. 6B) was reduced at hippocampal Sch–CA1 synapses in AIDA-1 cKO mice compared with WT littermates. Moreover, we found that NMDAR-dependent LTD induced by low-frequency stimulation (Fig. 6C), but not mGluR-dependent LTD (Fig. 6D), was also reduced in AIDA-1 cKO mice. Together, these results strongly suggest that AIDA-1 plays an important role in NMDAR-dependent synaptic plasticity in the hippocampus by regulating NMDAR subunit composition.
AIDA-1 preferentially associates with GluN2B-containing NMDARs
The altered synaptic GluN2A/GluN2B ratio without a change in the total expression of GluN2A or GluN2B raises the possibility that AIDA-1 differentially regulates NMDAR subunit trafficking. To determine whether AIDA-1 interacts with NMDARs in a subunit-specific manner, we performed coimmunoprecipitation experiments from mouse hippocampal lysates. Consistent with previous work (Jordan et al., 2007), AIDA-1 bound to NMDARs and PSD95 but not AMPARs (Fig. 7A). By measuring the amount of GluN2 subunits coimmunoprecipitating with AIDA-1 and normalizing to total GluN2 subunit levels in the input, we found that AIDA-1 associates preferentially with GluN2B over GluN2A (Fig. 7A). Similarly, immunocytochemical analysis showed that endogenous AIDA-1 colocalizes more with GluN2B than with GluN2A subunits in rat primary hippocampal neurons (Fig. 7B,C). Furthermore, overexpressed GFP-GluN2B and mKATE2-AIDA-1 colocalized extensively at synapses as well as in somatodendritic clusters in primary hippocampal cultures (Fig. 7D). Overexpressed GluN2A and AIDA-1 also colocalized in dendrites but to a lesser extent (Fig. 7D). To determine whether AIDA-1 could interact with GluN2B, we coexpressed AIDA-1, GluN1, and GluN2B in HEK293FT cells. However, we could not detect an interaction between AIDA-1 and GluN2B using coimmunoprecipitations (Fig. 7E). This indicates that either AIDA-1 does not bind to GluN2B directly or specific conditions or factors exist in neurons that enable the formation of GluN2B-AIDA complexes. Together, these results suggest that AIDA-1 primarily associates with and controls the localization of GluN2B-containing NMDARs.
AIDA-1 interacts with the GluN2B transport complex and regulates GluN2B ER export
The extensive colocalization of AIDA-1 and GluN2B in somatodendritic compartments (Fig. 7D) implies that AIDA-1 regulates trafficking of GluN2B throughout the secretory pathway. We therefore asked whether AIDA-1 interacts with MAGUK family members SAP97 and CASK, and/or the kinesin-2 molecular motor KIF17, which regulate the ER export and transport of GluN2B-containing NMDARs to synapses (Setou et al., 2000; Guillaud et al., 2003; Lau and Zukin, 2007; Jeyifous et al., 2009). We found that AIDA-1 formed complexes with CASK and SAP97, as well as with the GluN2B molecular motor KIF17, but not with KIF1b (Fig. 8A). Moreover, we found that AIDA-1 bound to CASK and KIF17 when coexpressed in HEKs, suggesting a direct interaction (Fig. 8B). As acute knockdown of CASK or KIF17 have also been shown to reduce synaptic GluN2B levels (Guillaud et al., 2003; Jeyifous et al., 2009), we asked whether loss of AIDA-1 altered the levels of these proteins, but found no significant changes (Fig. 8C).
Because CASK and KIF17 transport GluN2B-containing NMDARs in ER-derived vesicles (Jeyifous et al., 2009), we reasoned that loss of AIDA-1 might lead to the accumulation of GluN2B in the ER. We therefore performed subcellular fractionation of AIDA-1 cKO and WT hippocampal tissues using iodixanol gradient centrifugation to obtain enriched Golgi and ER fractions. As expected, the ER marker calnexin was enriched in heavier fractions while the Golgi marker 58K protein was enriched in lighter fractions (Fig. 8D). Densitometry-based quantification indicated that GluN2B levels were significantly increased in calnexin-positive ER fractions from AIDA-1 cKO mice compared with controls (Fig. 8D). Together these results show that GluN2B accumulates in the ER in the absence of AIDA-1 and are consistent with a role for AIDA-1 in the ER export of GluN2B-containing NMDARs.
Discussion
Here we report a previously unknown role for AIDA-1 in the regulation of synaptic NMDAR subunit composition and NMDAR-dependent plasticity in the hippocampus. Using a forebrain-specific AIDA-1 cKO mouse and a knockdown strategy, we demonstrate that loss of AIDA-1 decreases GluN2B-mediated and increases GluN2A-mediated synaptic transmission without affecting the total protein abundance of either subunit, consistent with a dysregulation of GluN2 subunit trafficking. We also showed that AIDA-1 preferentially associates with GluN2B-containing NMDARs and plays a role in their transport from the ER. To our knowledge, our findings provide the first evidence that AIDA-1 can control NMDAR subunit composition and synaptic function by regulating GluN2B distribution.
The basis for the preferential association between AIDA-1 and GluN2B over GluN2A is unclear. While AIDA-1 associates with GluN2B in hippocampal tissue, this does not appear to be a direct interaction (Fig. 7E). Among candidate intermediary proteins are MAGUKs that bind to both NMDARs and AIDA-1 (Kim and Sheng, 2004; Jordan et al., 2007; Lau and Zukin, 2007), and which may exhibit GluN2 subunit selectivity, although this latter view remains controversial (Paoletti et al., 2013). AIDA-1 could also interact with PDZ domain-containing adaptors, such as MALS-2, which binds to the GluN2B C-terminal domain and increases the surface expression of GluN2B but not GluN2A (Jo et al., 1999; Iwamoto et al., 2004). Another candidate is amyloid precursor protein, which binds to both AIDA-1 (Ghersi et al., 2004a,b) and NMDARs (Cousins et al., 2009) and regulates cell-surface GluN2B levels in neurons without affecting surface GluN2A expression (Hoe et al., 2009). Although we cannot rule out a direct role for AIDA-1 in controlling GluN2A levels at synapses, our imaging and quantitative coimmunoprecipitations support a direct role for AIDA-1 in regulating GluN2B. Previous findings show that GluN2A can compensate for the loss of GluN2B at synapses (Barria and Malinow, 2002), raising the possibility that upregulation of synaptic GluN2A represents a compensatory effect in AIDA-1 cKO mice. Still, ∼20% of GluN2A associates with AIDA-1 by immunoprecipitation and immunofluorescence analysis (Fig. 7A–D), suggesting AIDA-1 binding to GluN1/GluN2A/GluN2B triheterotrimeric NMDARs (Paoletti et al., 2013; Tovar et al., 2013).
We found that loss of AIDA-1 leads to the accumulation of GluN2B in the ER. Mechanisms governing NMDAR export from the ER include alternative splicing of GluN1 subunits (Scott et al., 2001) and masking of C-terminal ER-retention motifs by the binding of PDZ-domain-containing proteins (Wenthold et al., 2003a). We previously reported that AIDA-1 binds to the NMDAR complex and can accumulate in the nucleus following NMDAR activity (Jordan et al., 2007). While a role for altered gene expression or nucleolar function cannot be excluded, we found no changes in the overall abundance of factors involved in NMDAR expression or trafficking (Fig. 8C). Instead, we provide evidence that AIDA-1 associates with SAP97, CASK, and the molecular motor KIF17, which are components of a complex critically involved in GluN2B transport from the ER (Jeyifous et al., 2009) and trafficking along dendrites to synapses (Setou et al., 2000; Guillaud et al., 2003; Jeyifous et al., 2009). Remarkably, knockdown of SAP97 or CASK leads to GluN2B accumulation in the soma (Jeyifous et al., 2009). Moreover, acute loss of KIF17 reduces the dendritic transport of GluN2B, decreases synaptic GluN2B, and increases GluN2A levels (Guillaud et al., 2003), exactly phenocopying the synaptic alteration of NMDAR subunits in AIDA-1 cKO mice. Altogether these results strongly suggest that AIDA-1 is a functional component of the CASK/KIF17 complex that regulates the ER export and trafficking of GluN2B to synapses.
Precisely how AIDA-1 regulates the CASK/KIF17 complex and GluN2B export from the ER is unclear. AIDA-1 may act as an adaptor that sorts cargos from the ER to vesicles for transport. In support of this possibility, the PTB domain of AIDA-1 is ∼45% homologous to the PTB domain of Dab2, which regulates the endocytosis of lipoprotein receptor-related proteins and recognizes inositide-4,5-bisphosphates, which are enriched in membrane subdomains that undergo increased rates of endocytosis (Mishra et al., 2002). Alternatively, ubiquitination has emerged as an elaborate cellular mechanism for sorting proteins into different subcellular compartments (Hicke and Dunn, 2003). The sterile-α-motif domains of the AIDA-1 homolog Odin can bind to ubiquitinated EphA8 receptors and prevent their internalization (Kim et al., 2010). As GluN2B is ubiquitinated by the E3 ligase Mindbomb-2 (Jurd et al., 2008), AIDA-1 may bind to ubiquitinated GluN2B and regulate its localization, although we were unable to detect a direct interaction between AIDA-1 and GluN2B expressed in heterologous cells.
An alternative hypothesis to ER export and sorting is that AIDA-1 stabilizes GluN2B-containing NMDARs at synaptic sites similar to MAGUKs (Kim and Sheng, 2004). This possibility is consistent with biochemical and electron microscopy evidence showing that AIDA-1 is enriched and localized within the core of the PSD (Jordan et al., 2004, 2007; Dosemeci et al., 2007; Jacob et al., 2010). Yet another possibility is that AIDA-1 stabilizes GluN2B at synapses indirectly through PSD-95, as PSD-95 can control surface expression of AMPARs and NMDARs (Elias and Nicoll, 2007; Xu, 2011) and loss of AIDA-1 reduces synaptic levels of PSD-95 (Fig. 4D). However, neither AMPAR mEPSCs, nor input/output of AMPAR-mediated transmission, nor AMPAR subunits at the PSD are altered in AIDA-1 cKO mice. Therefore the significance of reduced PSD95 expression at synapses will require additional experimentation. Studies on glutamatergic receptor trafficking have shown that it is difficult to assign distinct roles to receptor adaptor proteins (Paoletti et al., 2013). For example, MAGUKs, which stabilize and cluster AMPARs and NMDARs at synapses, have also been observed throughout the secretory pathway and promote ER export (Wenthold et al., 2003b). Additional experiments will therefore be required to determine the precise molecular mechanisms underlying the AIDA-1-dependent regulation of GluN2B ER export, transport, and localization at synapses.
Several human studies have linked the AIDA-1 gene ANKS1B to schizophrenia and ASDs. A recent large exome sequencing study identified small de novo mutations in ANKS1B in patients with schizophrenia (Fromer et al., 2014; Purcell et al., 2014), and a single nucleotide polymorphism (SNP) in ANKS1B has been associated with improved response to antipsychotics, such as olanzapine and respiridone (McClay et al., 2011a,b). In addition, copy number variations and SNPs in ANKS1B have also been identified in patients with autism (Pinto et al., 2010) and correlate positively with impaired play skills (Hu et al., 2011; Meehan et al., 2011). Furthermore, ANKS1B contains “brain-critical” exons that are enriched in autism (Uddin et al., 2014). Despite these important links to disease, how altered AIDA-1 function may contribute to neuropsychiatric disorders is unknown. One possibility is that impaired NMDAR-dependent LTP and LTD contributes to cognitive deficits present in diseases associated with AIDA-1 dysfunction. NMDAR dysfunction has been implicated in schizophrenia (Lau and Zukin, 2007; McClay et al., 2011b; Poot et al., 2011; Paoletti et al., 2013; Snyder and Gao, 2013; Fromer et al., 2014; Purcell et al., 2014), ASD (Poot et al., 2011; Tarabeux et al., 2011; Jiang and Ehlers, 2013; Palumbo et al., 2014), and Alzheimer's disease (AD; Lau and Zukin, 2007; Mota et al., 2014). GluN2B (GRIN2B), in particular, has been identified as a candidate gene in genome-wide association studies of patients with sporadic ASD (Poot et al., 2011) and deletions or copy number variations of GluN2B regulating factors, such as SHANK (Jiang and Ehlers, 2013) and TBR1 (Palumbo et al., 2014), have been reported in patients with ASD. In AD, synaptic alterations induced by amyloid-β are mediated by NMDARs, especially GluN2B-containing NMDARs (Mota et al., 2014). Although more evidence is required, our finding that AIDA-1 regulates NMDAR function by controlling synaptic GluN2B levels raises the possibility that mutations in AIDA-1 might contribute to the pathogenesis of these neuropsychiatric disorders by impairing GluN2B function.
Footnotes
This work was supported by National Institutes of Health Grants AG039521 (to B.A.J.), MH081935, and DA017392 (to P.E.C.). A.E.C. was supported by a National Alliance for Research on Schizophrenia and Depression Young Investigator Grant from the Brain & Behavior Research Foundation and partially supported by Núcleo milenio Nu-MIND (NC 130011). J.O.T. was partially supported by T32 GM007288.
The authors declare no competing financial interests.
- Correspondence should be addressed to Bryen A. Jordan, PhD, Dominick P. Purpura Department of Neuroscience, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Rose Kennedy Center, Room 825, Bronx, NY 10461. bryen.jordan{at}einstein.yu.edu