SAP97 directs NMDA receptor spine targeting and synaptic plasticity

Abstract

Abstract

SAP97 is a multidomain scaffold protein implicated in the forward trafficking and synaptic localization of NMDA- and AMPA-type glutamate receptors. Alternative splicing of SAP97 transcripts gives rise to palmitoylated αSAP97 and L27-domain containing βSAP97 isoforms that differentially regulate the subsynaptic localization of GluR1 subunits of AMPA receptors. Here, we examined whether SAP97 isoforms regulate the mechanisms underlying long-term potentiation (LTP) and depression (LTD) and find that both α- and β-forms of SAP97 impair LTP but enhance LTD via independent isoform-specific mechanisms. Live imaging of α- and βSAP97 revealed that the altered synaptic plasticity was not due to activity-dependent changes in SAP97 localization or exchange kinetics. However, by recording from pairs of synaptically coupled hippocampal neurons, we show that αSAP97 occludes LTP by enhancing the levels of postsynaptic AMPA receptors, while βSAP97 blocks LTP by reducing the synaptic localization of NMDA receptors. Examination of the surface pools of AMPA and NMDA receptors indicates that αSAP97 selectively regulates the synaptic pool of AMPA receptors, whereas βSAP97 regulates the extrasynaptic pools of both AMPA and NMDA receptors. Knockdown of βSAP97 increases the synaptic localization of both AMPA and NMDA receptors, showing that endogenous βSAP97 restricts glutamate receptor expression at excitatory synapses. This isoform-dependent differential regulation of synaptic versus extrasynaptic pools of glutamate receptors will determine how many receptors are available for the induction and the expression of synaptic plasticity. Our data support a model wherein SAP97 isoforms can regulate the ability of synapses to undergo plasticity by controlling the surface distribution of AMPA and NMDA receptors.

Introduction

Long-term potentiation (LTP) and depression (LTD) are two major forms of synaptic plasticity expressed in the hippocampus (Bliss & Collingridge, 1993; Malenka & Nicoll, 1999). The N-methyl-d-aspartate-type glutamate receptor (NMDAR) plays a key role in synaptic plasticity by acting as a coincidence detector to trigger the recruitment of α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptors (AMPARs) into the synaptic membrane (Malenka & Nicoll, 1999; Malinow & Malenka, 2002; Citri & Malenka, 2008). Thus NMDARs are critical for the induction of synaptic plasticity while AMPARs are responsible for its expression.

The recruitment of receptors to synapses depends on their interaction with scaffold proteins at the postsynaptic density (PSD) (Montgomery et al. 2004; Specht & Triller, 2008). PSD-95 and SAP97 are central organizers of the PSD, simultaneously binding to receptors and to other scaffold or regulator proteins through PDZ, SH3 and GK interaction domains (Montgomery et al. 2004). Both PSD-95 and SAP97 bind the AMPAR subunit GluR1, although PSD-95 also requires the auxiliary protein stargazin (Leonard et al. 1998; Cai et al. 2002; Fukata et al. 2005). PSD-95 and SAP97 undergo alternative N-terminal splicing that gives rise to the α variants which contain a palmitoylation sequence, or β variants with an L27 protein interaction domain (Muller et al. 1995; Mori et al. 1998; Chetkovich et al. 2002; McLaughlin et al. 2002; Schluter et al. 2006; Waites et al. 2009). The N-terminal sequences of SAP97 and PSD-95 control their dynamic properties (Chetkovich et al. 2002; Nakagawa et al. 2004; Waites et al. 2009). For example, the synaptic exchange rates of α- and βSAP97 differ widely, consistent with their association with structures localized within or outside the PSD, respectively (Waites et al. 2009). On many levels, αSAP97 mimics the properties of αPSD-95 (Nakagawa et al. 2004; Schluter et al. 2006; Waites et al. 2009). Nonetheless, given that αPSD-95 is the main expressed PSD-95 variant and βSAP97 the main form of SAP97, it appears that functional specialization has occurred between the two proteins (Chetkovich et al. 2002; Schluter et al. 2006; Waites et al. 2009). In contrast to αPSD-95, βSAP97 has multiple roles in the trafficking of receptor subunits, participating in the forward trafficking of AMPARs (Sans et al. 2001; Jeyifous et al. 2009), forming part of a multiprotein trafficking complex for Kir2 potassium channels (Leonoudakis et al. 2004), and trafficking NMDARs to synapses via Golgi outposts (Jeyifous et al. 2009).

Our previous work revealed that αSAP97 and βSAP97 differentially regulate the levels, dynamics and subsynaptic localization of AMPARs (Waites et al. 2009). Given the importance of AMPARs in the expression of synaptic plasticity, we examined how SAP97 isoforms could regulate synaptic plasticity mechanisms. By recording from synaptic connections between individual hippocampal neurons, we demonstrate that α- and βSAP97 independently regulate LTP at hippocampal synapses by altering the synaptic and extrasynaptic localization of AMPARs and NMDARs respectively. Our data reveal that these isoform specific SAP97-induced changes in glutamate receptor localization have the potential to control the plasticity of hippocampal synapses.

Methods

Expression constructs

A restriction fragment (NheI/XbaI) containing the sequence of rat βSAP97 (splice variant I1b, I3 and I5), fused at its N-terminus with enhanced green fluorescent protein (EGFP; Clontech, containing F64L/S65T) via the linker sequence SGLRSRAQASNS, was subcloned into the XbaI/XbaI backbone of the lentiviral vector pFUGW (Lois et al. 2002). This resulted in plasmid pFU-EG-rSAP97I3 (EMBL accession no. AM710296; http://www.ebi.ac.uk/embl), for the expression of the fusion protein EGFP-βSAP97. Furthermore, we used expression constructs for αSAP97-EGFP (Schluter et al. 2006) and βSAP97-EGFP (Waites et al. 2009), also based on pFUGW and tagged at their C-terminus with EGFP. The shRNA construct used to specifically knock down the expression of βSAP97 was targetted to the unique L27 domain of βSAP97. Both the βSAP97 shRNA and the scrambled control constructs have been described previously (Jeyifous et al. 2009). Sequence analysis reveals that αSAP97 has no unique domains: αSAP97 is identical to βSAP97 with the exception of the palmitoylation sequence, and αSAP97 shares this palmitoylation sequence with other MAGUKs (specifically αPSD93). Therefore it was not possible to create an shRNA that specifically targets αSAP97. Rescue of shRNA mediated knockdown was achieved by coexpressing an shRNA-resistant βSAP97*-EGFP and shRNA-SAP97 (S12) in the pFU-rSAP97I3-EG vector. Silent mutations in codons encoding N-terminal amino acid residues 6, 7, 8 and 9 of βSAP97-EGFP was achieved by PCR with oligonucleotides containing the following sequence ATGCCGGTCCGGAAGCAgGAcACaCAaAGA (bold: start codon; lower case: nucleotide changes). Note that this sequence is unique to βSAP97 and present near its L27 domain. The shRNA-SAP97 (S12) under the H1 polymerase III promoters was subcloned into the BsiW1-Pac1 sites upstream of the ubiquitin promoter/βSAP97*-EGFP expression cassette. Translocation studies (see Fig. 3) were conducted with yellow-fluorescent protein (YFP)-tagged rat CaMKIIα (a kind gift from J. Tsui) (Tsui et al. 2005).

Figure 3. cLTP and cLTD treatments cause YFP-CaMKIIα translocation into synaptic spines.

A, time-lapse images of YFP-CaMKIIα translocation into spines. Sample images of dendritic segments (DIV15) acquired during 3 min baseline conditions (left image; 1 min time point shown, see B), after application of 25 μm NMDA/10 μm glycine (centre image; 6 min), and after addition of 50 μm APV/10 μm CNQX (right image; 11 min time point). B, quantification of data shown in A. Mean spine intensity levels were reduced during NMDA application (n = 6 puncta). This quenching occurred in all cellular compartments and was not related to a redistribution of the YFP-CaMKIIα. After addition of the glutamate receptor blockers APV and CNQX, the fluorescence recovered over a period of 5 min. Note that the spine intensities after recovery exceed the levels prior to NMDA application as a result of CaMKIIα translocation (compare right and left image in A). C, neurons transfected with YFP-CaMKIIα were incubated with NMDA/glycine (cLTD) or bicuculline/glycine (cLTP) for 5 min before fixation and image acquistion (left panel) (DIV12). The enrichment of YFP-CaMKIIα in synapses was quantified as spine versus shaft fluorescence intensity (right panel) and showed that both treatments produced a translocation of CaMKIIα into dendritic spines (mean ± SEM; n = 3 cells with ≥17 synaptic puncta per cell; control: shaft/spine ratio 0.51 ± 0.06; cLTD: 0.93 ± 0.07; cLTP: 0.78 ± 0.12).

Hippocampal neuron culture

Primary neuron cultures were prepared according to a modified Banker culture protocol (Banker & Goslin, 1998). Briefly, hippocampi from Sprague–Dawley rat embryos (E18–19) were dissociated in 0.05% trypsin (Gibco, no. 25300) and plated on poly-l-lysine coated coverslips (Carolina Biological Supply Co., Burlington, NC, USA) at a density of 165 cells mm⁻². Coverslips were inverted after 1 h, placed over a glial feeder layer and maintained at 37°C/5% CO₂ in Neurobasal medium (Gibco, no. 21103) containing 2 mm GlutaMAX (Gibco, no. 35050) and B27 (Gibco, no. 17504-044). Neuron cultures were infected with lentivirus expression constructs at DIV0–2 and used for experiments on DIV14–17. Alternatively, cultures were transfected by Ca₃(PO₄)₂ precipitation (on DIV7–9) or lipofectamine 2000 (Invitrogen) and used for experiments on DIV12–15. Most imaging experiments (localization studies, time-lapse imaging and fluorescence recovery after photobleaching (FRAP)) were done on lentivirus-infected neuron cultures, while the electrophysiological experiments were performed exclusively on transfected neurons. Control neurons were either untransfected neurons plated on the same coverslips as neurons transfected with α- or βSAP97, or neurons transfected with GFP alone. No significant difference in AMPAR or NMDAR EPSC amplitude was observed between untransfected and GFP-transfected neurons (Waites et al. 2009).

Hippocampal organotypic slice culture

Hippocampal slice cultures were prepared from P7 male rat pups as previously described (Montgomery et al. 2001). Briefly, 400 mm hippocampal sections were grown in MEM + 40% horse serum at 37°C on nitrocellulose membranes. At 3 DIV slices were moved to 34°C and maintained for 1–2 weeks. Slices were induced to express αSAP97-EGFP or EGFP-βSAP97 by lentiviral infection at 1 DIV. Paired whole cell recordings from CA3 pyramidal neurons were then performed at 7–11 DIV with the postsynaptic cell within the CA3 pair being GFP-positive indicative of expression of αSAP97-EGFP or EGFP-βSAP97.

Neuron transfection

A volume of 60 μl containing 2 μg DNA and 7.5 μl 2 m CaCl₂ was added dropwise to 60 μl of 2 × HBS buffer (274 mm NaCl, 10 mm KCl, 1.4 mm Na₂HPO₄, 15 mm glucose, 42 mm Hepes, pH 7.1). After 20 min incubation in the dark, the precipitate was pipetted onto a coverslip of cultured neurons in 1 ml conditioned medium containing 50 μm dl-2-amino-5-phosphonopentanoic acid (APV) and 10 μm 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX). Neurons were incubated for 30 min at 37°C/5% CO₂, rinsed three times with 2 ml pre-warmed HBSS and transferred back into culture dishes.

Lentivirus production

HEK cells were grown to confluence and cotransfected with pFUGW plasmids (10 μg) together with the helper plasmids VSVg (5 μg) and Δ8.9 (7.5 μg) in 1.5 ml Opti-MEM (Gibco, no. 51985), using 60 μl lipofectamine 2000 (Invitrogen, no. 11668) according to the supplier's protocol. Cells were transfected and maintained in neurobasal medium containing GlutaMAX and B27 at 32°C/5% CO₂. The medium was exchanged once after 24 h and the virus was harvested in culture medium after ∼55 h and stored at −80°C. Generally, lentivirus titres were in the range of 10⁸ ml⁻¹, as estimated by the number of infected HEK cells expressing EGFP fluorescence.

Imaging buffers

Buffers used for live imaging were based on Tyrode solution (120 mm NaCl, 2.5 mm KCl, 2 mm CaCl₂, 25 mm Hepes pH 7.4, 30 mm glucose, sterile filtered). For control experiments, neurons were incubated in buffer solution containing 2 mm MgCl₂, 0.5 μm tetrodotoxin (TTX), 1 μm strychnine, 50 μm APV and 10 μm CNQX, to fully block neuronal activity. To induce neuronal activity, we used conditions similar to those described by Ivanov et al. (2006). For the direct activation of NMDARs, we applied 25 μm NMDA/10 μm glycine in the presence of 0.5 μm TTX and 1 μm strychnine (= cLTD treatment). Enhanced synaptic activity was achieved through the application of 20 μm bicuculline/200 μm glycine in the presence of 1 μm strychnine (= cLTP). For this application, prior incubations in control buffer were done in the absence of TTX. Similar treatments with NMDA or high concentrations of glycine have been shown to induce LTD or LTP in cultured hippocampal neurons, respectively (Lu et al. 2001).

Synaptic localization assay

Coverslips with hippocampal cultures expressing αSAP97-EGFP or EGFP-βSAP97 were transferred into Tyrode control solution containing Mg²⁺, TTX (not for subsequent bicuculline/glycine treatment), strychine, APV and CNQX at 37°C. After 5 min, the solution was exchanged with NMDA/glycine (cLTD) or with bicuculline/glycine containing solution (cLTP), or with the identical control solution (see imaging buffer compositions). After 10 min the neurons were fixed for 10 min at 37°C in 0.1 m sodium phosphate buffer pH 7.4, containing 4% paraformaldehyde and 1% sucrose. To correct for the expression levels of individual neurons, we quantified the synaptic accumulation of SAP97, i.e. the mean fluorescence intensity of synaptic spines relative to the fluorescence in dendritic shafts, rather than the absolute levels of synaptic SAP97.

Live imaging

Time-lapse imaging was done on a scanning confocal microscope (custom-built by N. Ziv and S. Smith on a Zeiss Axiovert 100TV) with 488 nm and 514 nm lasers (Coherent; Sapphire 488–20CDRH and Compass 215M-20) and a 40× objective (1.3 NA; Zeiss Plan Neofluar), using OpenView software (by N. Ziv) for image acquisition. Coverslips with primary neurons were mounted in a perfusion chamber, maintained at 37°C and perfused with Tyrode-based solutions.

Fluorescence recovery after photobleaching (FRAP)

Synaptic puncta of EGFP-tagged SAP97 were bleached to ∼20% of initial fluorescence intensity by multiple scanning passes (15–20×) of a high intensity 488 nm laser beam. The fluorescence recovery was imaged for up to 30 min, initially every 30 s (14 frames) and then every 5 min (5 frames). Intensity values of bleached puncta at each time point were normalized to their fluorescence intensity prior to bleaching (I_t/I_pre) and to non-bleached control puncta (I_nb,t/I_nb,pre) in the same field of view. To calculate mean recovery traces the dynamic range of individual experiments was adjusted by setting the first value after photobleaching to zero. Thus, FRAP data were normalized according to the equations:

and

where F_norm is the normalized fluorescence intensity, F_t the intensity at time t, F₀ at time t = 0 and F_pre = 1 the intensity prior to photobleaching. The mean experimental data were fitted with an exponential equation with two components, as described previously (Tsuriel et al. 2006):

where τ₁ and τ₂ are the time constants and a and (1 –a) are the relative fractions of fluorescence in the two pools. The theoretical parameters were extracted by minimizing the sum of squared residuals, using a macro written in Excel (by N. Ziv). For time-lapse experiments, images were acquired every 3 min for up to 33 min (12 frames). Intensity values of synaptic SAP97 puncta were normalized to their fluorescence intensity prior to bleaching (I_t/I_pre), and expressed as mean intensity ± SEM.

Immunocytochemistry

Immunocytochemistry to detect α-actinin (Sigma, 1:2000) was performed as previously described (Cheyne & Montgomery, 2008; Waites et al. 2009). Primary antibody binding was performed overnight at 4°C. Secondary antibody was performed for 1 h at room temperature. Coverslips were washed and mounted onto slides (Vectashield) for imaging. Pyramidal neurons were identified by morphology and Z-stack (0.5 μm) images of dendrites were obtained on a Zeiss Axioskop with a CCD digital camera using a 63× oil objective. Image analysis was performed using ImageJ. Z-stacks were converted to 8-bit, merged into maximum projections and background subtracted to remove the diffuse protein expression within dendrites. Images in which staining was dim or the background was high were excluded from the analysis. Images were manually thresholded to select only puncta that were greater than 2-fold above image background. Puncta were analysed using the Analyze Particles function in ImageJ so that the average puncta intensity and number of puncta could be determined. All data were normalized to parallel controls and presented as means ± SEM were n is the number of fields analysed. Statistical analysis was performed using Student's t tests (one-tailed distribution, unequal variance).

Electrophysiology

Dual whole cell recordings were performed at DIV9–14 on primary dissociated hippocampal cultures and at DIV7–11 on hippocampal organotypic slice cultures transfected with αSAP97-EGFP or βSAP97-EGFP as previously described (Montgomery et al. 2001; Waites et al. 2009). Briefly, hippocampal cultures were transferred to a recording chamber on a Zeiss Axioskop 2FS and visualized under infrared differential interference contrast microscopy. Cultures were perfused at room temperature at a rate of 2 ml min⁻¹ with artificial cerebrospinal fluid (ACSF, in mm: 119 NaCl, 2.5 KCl, 1.3 MgSO₄, 2.5 CaCl₂, 1 Na₂HPO₄, 26.2 NaHCO₃, 11 glucose). Presynaptic untransfected neurons were held in current clamp and induced to fire action potentials at 0.1 or 0.2 Hz by brief current injection (typically 20–50 pA for 20 ms). Postsynaptic neurons (GFP-positive) were held in voltage clamp at −65 mV. Series resistance (R_s) was continuously monitored and recordings in which R_s varied by more than 20% were excluded from analysis. Internal solution consisted of (in mm): 120 potassium gluconate (presynaptic cell) or 120 caesium gluconate (postsynaptic cell), 40 Hepes, 5 MgCl₂, 0.3 NaGTP, 2 NaATP, 5 QX314 (postsynaptic cell only), pH 7.2 with KOH or CsOH. Evoked AMPAR- and NMDAR-mediated EPSCs were measured as previously described (Montgomery et al. 2001, 2005; Montgomery & Madison, 2002; Waites et al. 2009). Isolated NMDAR EPSCs were measured at +40 mV in ACSF containing 10 μm 1,2,3,4-tetrahydro-6-nitro-2,3-dioxo-benzo[f]quinoxaline-7-sulfonamide (NBQX). AMPAR and NMDAR EPSCs in response to presynaptic action potentials were collected at 0.1 or 0.2 Hz. LTP and LTD were induced 5 min after the initiation of the postsynaptic whole cell recording using either pharmacological stimulation of NMDARs as described above for live imaging or by pairing presynaptic action potentials with postsynaptic depolarization (Montgomery et al. 2001, 2005; Montgomery & Madison, 2002). To measure surface AMPAR and NMDAR-mediated currents, exogenous AMPA (1 mm) or NMDA (1 mm) was applied with a picospritzer (pressure 2 bar, pulse 200 ms). The micropipette was placed at a standard distance of 200 μm from the dendrites of α- or βSAP97-EGFP transfected neurons. The peak amplitudes of AMPAR- and NMDAR-mediated surface currents were measured by whole cell patch clamp as detailed above. Online data acquisition and offline analysis for all electrophysiology experiments was performed with pCLAMP (Clampex v9.2). Statistical significance of changes in AMPAR and NMDAR EPSC amplitudes was tested using Student's t test with a level of significance set at P < 0.05.

Results

SAP97 isoforms regulate synaptic plasticity in hippocampal neurons

SAP97 is known to be critical for the trafficking of AMPA- and NMDA-type glutamate receptors to synapses (Sans et al. 2001; Jeyifous et al. 2009) and in directing the subsynaptic localization of AMPARs at excitatory synapses (Schluter et al. 2006; Jeyifous et al. 2009; Waites et al. 2009). Our previous data show that αSAP97 drives AMPARs into the PSD, whereas βSAP97 drives AMPARs to extrasynaptic sites (Waites et al. 2009). We therefore hypothesized that SAP97 isoforms are prime candidates to regulate changes in AMPAR expression that are known to occur with LTP and LTD. In order to determine the potential role of SAP97 isoforms in synaptic plasticity, we employed a combination of electrophysiology and live cell imaging in the dissociated hippocampal cell culture preparation. Dissociated cultures and pharmacological induction of plasticity were employed to enable us to visualize synaptic protein dynamics and to compare these data to the electrophysiogical changes seen with the same induction protocols.

LTP and LTD were induced in dissociated hippocampal cultures by pharmacological stimulation of NMDARs. Using paired whole-cell recordings from pyramidal neurons, we measured the amplitude of evoked excitatory postsynaptic currents (EPSCs) at synaptic connections between individual hippocampal neurons before, during and after the induction of chemical LTP (cLTP; 20 μm bicuculline + 200 μm glycine, in the presence of 1 μm strychnine to avoid activation of glycine receptors) or chemical LTD (cLTD; 25 μm NMDA + 10 μm glycine + 1 μm strychnine, see Methods).

In control hippocampal neurons, application of bicuculline and glycine induced an increase in the amplitude of the AMPAR EPSCs (Fig. 1A). Potentiation of the AMPAR-mediated currents developed gradually over the first 5 min after the pharmacological induction of plasticity and was sustained for the length of all paired recordings (n = 13 pairs). Twenty minutes after cLTP induction, the amplitude of the AMPAR EPSCs was increased to an average of 144.3 ± 16.4% of the baseline current amplitude (P < 0.01; n = 13 pairs). This cLTP was NMDAR dependent, as application of the NMDAR antagonist APV (50 μm) during cLTP induction prevented the increase in AMPAR EPSC amplitude (average AMPAR EPSC amplitude was 94.7 ± 2.4% of baseline 20 min after cLTP + APV, n = 5 pairs).

Figure 1. N-terminal SAP97 splice variants alter synaptic plasticity in hippocampal neurons.

A, AMPAR EPSC amplitudes measured from paired recordings between control hippocampal neurons or between neurons expressing α- or βSAP97-EGFP in the postsynaptic cell. cLTP was induced 5 min after attaining each paired recording by a 5 min application of 20 μm bicuculline and 200 μm glycine + 1 μm strychnine. EPSC amplitudes are normalized to baseline, pre-cLTP AMPAR EPSC amplitudes. Top panels, representative postsynaptic traces, overlaid, in control, α- or βSAP97-EGFP-expressing neurons before and after cLTP. One example of a presynaptic action potential trace is shown for each group. B, AMPAR EPSC amplitudes measured from paired recordings between hippocampal neurons as described in A. cLTD was induced by a 5 min application of 25 μm NMDA and 10 μm glycine + 1 μm strychnine and paired recordings were held for up to 20 min post-cLTD induction. C, surface AMPAR current amplitudes were measured by focal application of 1 mm AMPA in control, α- and βSAP97-expressing neurons immediately prior to and 20 min after the induction of cLTP or cLTD. Top, sample traces of surface AMPAR-mediated currents measured in control, α- and βSAP97-expressing neurons. Bottom, data were normalized to pre-cLTP or cLTD current amplitudes and are expressed as a percentage of the baseline surface AMPAR current.

We next examined whether the expression of two SAP97 isoforms, α- and βSAP97, altered this plasticity. These two N-terminal SAP97 isoforms were selected based on previous data showing that the palmitoylated αSAP97 recruits GluR1 containing AMPARs into the PSD and consequently increases basal synaptic transmission, while the L27 containing βSAP97 redirects GluR1 containing AMPARs into a peri-PSD compartment within dendritic spines, thus reducing basal synaptic transmission (Waites et al. 2009). Indeed, as expected the baseline levels of evoked AMPAR EPSC amplitudes were higher in neurons over-expressing αSAP97 and reduced in βSAP97-expressing neurons compared to control neurons, as reported previously (Waites et al. 2009). Average baseline AMPAR EPSC amplitude was 419.9 ± 54 pA in αSAP97-expressing neurons (n = 22 pairs) and 62.9 ± 16.1 pA in βSAP97-expressing neurons (n = 21 pairs). To assess the impact of α- and βSAP97 on the ability of excitatory synapses to undergo LTP, we measured EPSC amplitudes between pairs of hippocampal neurons in which the postsynaptic cells expressed EGFP-tagged αSAP97 or βSAP97. In contrast to untransfected neurons, we found that cLTP could not be induced in hippocampal neurons expressing αSAP97-EGFP or βSAP97-EGFP (Fig. 1A). In neurons expressing αSAP97, cLTP induction resulted in a decrease in the amplitude of AMPAR EPSCs. Twenty minutes after the induction of cLTP, the AMPAR EPSC amplitudes were 78.0 ± 13.9% of baseline control amplitudes (n = 10 pairs; Fig. 1A). Similarly, in neurons expressing βSAP97, cLTP induction resulted in a decrease in the amplitude of AMPAR EPSCs. Twenty minutes after the induction of cLTP, AMPAR EPSC amplitudes were decreased to 58.0 ± 13.7% of baseline current amplitudes (P < 0.01; n = 13 pairs). These data reveal that the overexpression of either SAP97 isoform impairs the induction and/or the expression of cLTP.

We next examined the ability of αSAP97 and βSAP97 to modulate long-term depression (LTD). In control neurons, the induction of cLTD produced an immediate and long-lasting depression of AMPAR-mediated EPSCs (Fig. 1B). AMPAR EPSC amplitudes were significantly decreased to 78.4 ± 14.8% of baseline current amplitudes (measured 20 min after cLTD induction; Fig. 1B; P < 0.01; n = 7 pairs). This cLTD was blocked by the NMDAR antagonist APV, demonstrating its NMDAR dependence (average AMPAR EPSC amplitude was 106.3 ± 1.2% of baseline currents 20 min after cLTD + APV; n = 5 pairs).

cLTD could also be induced in neurons expressing either αSAP97 or βSAP97 (Fig. 1B). As in control neurons, the amplitude of AMPAR EPSCs decreased immediately and this decrease was sustained in both αSAP97 and βSAP97-expressing hippocampal neurons for the length of the paired recordings. AMPAR EPSC amplitudes were 45.7 ± 11.1% of baseline current amplitudes in αSAP97-expressing neurons (n = 12 pairs; 20 min after cLTD induction). Similarly in βSAP97-expressing neurons, cLTD induction resulted in a significant decrease in AMPAR EPSC amplitudes to 59.6 ± 8.3% (n = 8 pairs) of baseline current amplitudes. The level of depression in αSAP97- and βSAP97-expressing hippocampal neurons was significantly greater than the LTD expressed in control neurons (P < 0.01). These data show that neither SAP97 isoform prevents the pharmacological induction of LTD. However, as reported above, we observed a dramatic difference in the baseline levels of evoked AMPAR EPSC amplitudes in neurons over-expressing αSAP97 (higher) and βSAP97 (reduced) compared to control neurons (Fig. 1A), as reported previously (Waites et al. 2009). These differences support the general conclusion that N-terminal SAP97 isoforms differentially affect the synaptic levels of AMPARs, and also imply that the changes in establishing LTP and/or LTD could have different underlying mechanisms.

As differences in glutamate receptor trafficking have been suggested to occur in dissociated versus slice hippocampal preparations (Shi et al. 1999; Waites et al. 2009), we sought to determine whether αSAP97 and βSAP97 also altered the ability of synapses to express plasticity in hippocampal slices. We performed paired whole cell recordings from CA3 pyramidal cell pairs in which the postsynaptic neuron was expressing either EGFP-tagged α- or βSAP97. On average, baseline AMPAR-mediated EPSCs in control pyramidal cell pairs were 16.0 ± 2.3 pA, compared with 27.3 ± 8.4 pA and 14.0 ± 2.4 pA in pairs in which the postsynaptic neuron expressed αSAP97 or βSAP97 respectively. The increase in AMPAR EPSC amplitude in αSAP97-expressing neurons was significant (P < 0.01), demonstrating that αSAP97 exerts similar effects in dissociated and hippocampal slice in vitro preparations. The AMPAR EPSC amplitude in βSAP97-expressing neurons was not significantly different from controls. The smaller magnitude of the change in average AMPAR EPSC amplitude in the slice cultures may reflect the lower expression levels of α- and βSAP97 induced by lentiviral infection. However, we also noted a decrease in the probability of βSAP97-expressing CA3 pyramidal cell pairs having an evident AMPAR EPSC indicative of a synaptically connected pair (27.3% in βSAP97-expressing neurons, compared with 38.5% in control slices and 53.8% in αSAP97-expressing slices), suggesting that AMPAR-mediated currents may be decreased to undetectable levels or that CA3 pyramidal neuron connectivity is decreased in βSAP97-expressing neurons. We then induced LTP or LTD in each connected CA3 pyramidal cell pair by our electrical synaptic stimulation protocols described previously, resulting in robust LTP and LTD (Supplemental Fig. S1; Montgomery et al. 2001, 2005, Montgomery & Madison, 2002). Twenty minutes after the LTP pairing protocol, average AMPAR EPSC amplitude was 210.2 ± 71.0% of baseline amplitude in control CA3 pyramidal cell pairs (n = 5 pairs, Fig. S1A). The induction of LTP between pyramidal cell pairs in hippocampal organotypic slices was also altered by the expression of αSAP97 or βSAP97 but to a differing degree from that observed with cLTP in dissociated hippocampal cultures. Twenty minutes after the induction of LTP, average AMPAR EPSC amplitude was 126.4 ± 7.3% and 86.9 ± 5.6% of the baseline EPSC amplitude in pyramidal cell pairs in which the postsynaptic neuron expressed α- (n = 6 pairs) or βSAP97 (n = 5 pairs), respectively. Thus αSAP97 impaired and βSAP97 prevented the induction of LTP in hippocampal slices. Induction of LTD with 1 Hz presynaptic stimulation for 5 min induced robust LTD in control, αSAP97 and βSAP97-expressing neurons (Fig. S1B). AMPAR EPSC amplitudes were 47.1 ± 10.6%, 24.8 ± 6.3% and 29.0 ± 4.6% of baseline current amplitudes in control (n = 5 pairs), αSAP97- (n = 6 pairs) and βSAP97 (n = 6 pairs)-expressing neurons, respectively. The magnitude of the LTD induced by 1 Hz stimulation was stronger compared to that induced by chemical stimulation in dissociated hippocampal cultures (Fig. 1B). Interestingly, the relative amounts of LTD expressed in control, α- and βSAP97-expressing neurons showed the same relationship in both forms of LTD. That is, the level of electrically induced LTD in αSAP97- and βSAP97-expressing hippocampal neurons was significantly greater than that expressed in control neurons (P < 0.05 in both cases), as was observed in response to the induction of cLTD (Fig. 1B).

SAP97 isoforms differentially alter synaptic and extrasynaptic AMPAR pools

Our above analysis of αSAP97- and βSAP97-induced changes in LTP and LTD were measured as changes in synaptic AMPAR-mediated currents. Since we have previously shown that a significant proportion of AMPARs in βSAP97-expressing neurons are located in the extrasynaptic membrane, we examined whether the size of the surface pool of receptors was altered by the induction of synaptic plasticity in neurons expressing αSAP97 or βSAP97. Any changes observed in the size of the surface AMPAR currents could then be compared to the changes previously observed in synaptic AMPAR currents (Fig. 1A and B), providing insights into the redistribution of synaptic and extrasynaptic AMPAR pools during synaptic plasticity. In these experiments, we measured the amplitude of the surface AMPAR-mediated current in response to focal application of AMPA prior to and 20 min after cLTP or cLTD treatment (Fig. 1C). In control neurons, we found that the amplitude of the surface AMPAR-mediated current was not significantly altered by cLTP (average current amplitude was 90.6 ± 11.7% of baseline current amplitude 20 min after cLTP induction; n = 6; P = 0.08; Fig. 1C), despite the LTP-induced increase in the amplitude of the synaptic AMPAR-mediated currents (Fig. 1A). These data are consistent with the hypothesis that under control conditions, the increase in synaptic AMPAR current measured during LTP results from the recruitment of extrasynaptic AMPARs to synapses (Makino & Malinow, 2009).

We next examined how the surface pool of AMPARs was altered in neurons expressing αSAP97 or βSAP97. Consistent with βSAP97 inducing a higher level of total surface AMPARs localized to extrasynaptic sites (Waites et al. 2009), baseline surface current amplitudes were significantly higher in βSAP97-expressing neurons (Fig. 1C) (average current amplitude was 198.6 ± 45.3 pA in βSAP97 neurons versus 107.7 ± 25.5 pA in αSAP97 neurons; P < 0.001; n = 6), as previously reported (Waites et al. 2009). Following the pharmacological induction of LTP, we found that the amplitude of the surface AMPAR current was decreased in αSAP97-expressing neurons. Twenty minutes after the induction of LTP average current amplitude was 76.2 ± 13.5% of baseline currents (P < 0.05; n = 6; Fig. 1C), a decrease similar to that occuring in the synaptic pool (Fig. 1A). In contrast, the surface AMPAR-mediated current amplitude did not change with cLTP in βSAP97 neurons (95.7 ± 4.4% of baseline AMPAR current amplitude, P > 0.1, n = 6; Fig. 1C), despite the significant decrease in synaptic AMPARs with cLTP (Fig. 1A), suggesting that βSAP97 facilitates an increase in the size of the extrasynaptic pool of AMPARs with cLTP.

We also examined the amplitude of the surface AMPAR mediated current in response to the induction of cLTD. Here we found that cLTD resulted in a decrease in the amplitude of the surface AMPAR-mediated currents in control, αSAP97- and βSAP97-expressing neurons (control: 80.1 ± 11.3% of baseline AMPAR current amplitude; αSAP97: 64.6 ± 5.5% of baseline AMPAR current amplitude; βSAP97: 65.14 ± 13.4% of baseline AMPAR current amplitude; Fig. 1C, P < 0.01 in all cases). This decrease in the size of the total surface AMPAR-mediated currents is similar to the changes seen in AMPAR-mediated synaptic currents in control, αSAP97- and βSAP97-expressing neurons (Fig. 1B), suggesting that with LTD expression both synaptic and extrasynaptic pools of AMPARs are changing in parallel and to a similar degree.

Synaptic plasticity does not alter the synaptic localization of SAP97 isoforms

Activity-dependent changes in synaptic strength have been directly linked to changes in the number of receptors localized in the PSD. If α- and βSAP97 are involved in directing AMPAR movement, then changes in α- and βSAP97 localization in response to synaptic activity may be responsible for the re-distribution of AMPARs during cLTP or cLTD. This concept is supported by several studies indicating that both the synaptic levels (Mauceri et al. 2004) and the protein dynamics (Nakagawa et al. 2004) of SAP97 may be regulated by synaptic activity. We designed a set of experiments to assess whether our cLTP or cLTD protocols caused a redistribution of SAP97 isoforms that could explain the observed changes of synaptic plasticity in α- and βSAP97-expressing neurons (Fig. 2). Initially, we examined the synaptic levels of EGFP-tagged α- and βSAP97 using time-lapse imaging. The intensity of synaptic SAP97 puncta were measured prior to, during and after the induction of cLTP or cLTD (Fig. 2A–C). No significant changes in the levels of α- or βSAP97 were detected either during or after the cLTP treatment (αSAP97: 99 ± 1.5% after 4 min of cLTP treatment and 99 ± 6.5% at 20 min after the treatment; βSAP97: 97 ± 2.0% after 4 min, 101 ± 5.4% after 20 min; Fig. 2B).

Figure 2. SAP97 localization is unchanged by cLTP or cLTD.

A–C, lentivirus-infected neuronal cultures expressing αSAP97-EGFP or EGFP-βSAP97 were imaged for 5 min under baseline conditions, during 5 min application of bicuculline/glycine (cLTP treatment) or NMDA/glycine (cLTD) and after drug washout. A, sample images of EGFP-βSAP97 in spines (DIV15), imaged before (3 min time point shown; left image) and during NMDA/glycine application (9 min; centre image) and after washout (18 min timepoint; right image). B, quantification of time-lapse imaging data of αSAP97- (grey trace, n = 13 cells) and βSAP97-expressing neurons (black, n = 24) before, during and after cLTP induction, expressed as means ± SEM. No significant changes in mean spine levels of SAP97 were detected at 3 versus 33 min time points; Mann–Whitney U test. C, application of NMDA and glycine was used to induce cLTD in neurons expressing αSAP97-EGFP (grey, n = 6) or EGFP-βSAP97 (black, n = 11). The treatment caused a temporary quenching of EGFP fluorescence, particularly in βSAP97-expressing neurons (see also Fig. 2A), but had no long-term effect on synaptic SAP97 levels (means ± SEM, Mann-Whitney U test). D and E, cLTP or cLTD treatments were applied for 10 min in neurons expressing αSAP97-EGFP or EGFP-βSAP97 and the cells were immediately fixed. After fixation, the fluorescence intensity in spines relative to the shaft was quantified. For the cLTP application, TTX was omitted from the pre-incubation buffer (see methods). D, quantification of the spine/shaft ratio of αSAP97-EGFP fluorescence (DIV15; mean ± SEM; n≥ 11 cells with 11 synaptic puncta per cell; right panel). αSAP97 enrichment did not change by either treatment (P = 0.8; Mann–Whitney U test, two-tailed). E, similarly, the cLTP and cLTD treatments produced no significant changes in the levels of βSAP97 in synaptic spines (n≥ 14 random fields of view with ≥ 27 synaptic puncta per cell, analysis blind to condition; right panel).

In contrast, there was a significant decrease in the fluorescence intensity of EGFP-βSAP97, and to a lesser extent of αSAP97-EGFP, during the induction of cLTD (αSAP97: 88 ± 6.1% after 4 min of cLTD treatment, P = 0.2; βSAP97: 63 ± 2.3%; P < 0.001, Mann–Whitnery U test; Fig. 2A and C). These data imply that conditions that stimulate the reduction of synaptic AMPARs (i.e. cLTD) are perhaps causing the simultaneous removal of α- and βSAP97 from the spine. However, we also noted that this loss in fluorescence was reversible and was not associated with a long-term change in synaptic levels of αSAP97-EGFP or EGFP-βSAP97 (αSAP97: 103 ± 4.5% at 20 min after cLTD treatment; βSAP97: 101 ± 1.3%; Fig. 2C). Upon closer inspection, we noticed that the loss of fluorescence during the addition of NMDA/glycine occurred throughout the entire cell including dendrites (Fig. 2A) and soma (not shown), putting into doubt that an actual redistribution of the protein was taking place as had been previously proposed (Nakagawa et al. 2004). Instead, we hypothesized that the loss of EGFP fluorescence was due to the acidification of the postsynaptic neuron during depolarization (Chesler, 2003) and thus due to quenching of the fluorophore (EGFP). In the presence of 25 μm NMDA, the intracellular pH of primary neurons decreases to about 6.5, recovering over a time course of minutes (Irwin et al. 1994). Furthermore, the fluorescence of EGFP is strongly pH dependent, with a pK_a≈ 6 and a decrease by ΔpH_i≈−0.5 leading to a reduction of fluorescence intensity of 20% (Kneen et al. 1998). These observations suggest that pH-dependent quenching of EGFP-based fluorophores is a confounding factor when using protocols that induce a prolonged depolarisation of neurons.

To test this quenching hypothesis, we repeated the time-lapse experiment with neurons expressing YFP-CaMKIIα (Fig. 3), which is known to be recruited to synapses by the activation of NMDARs (Shen & Meyer, 1999). Following the addition of NMDA/glycine to the imaging buffer, the YFP fluorescence was quickly lost in all cellular compartments (Fig. 3A). Furthermore, the loss/quenching of YFP-CaMKIIα was more pronounced than that observed with EGFP-βSAP97, as expected for a fluorophore with a substantially higher pH sensitivity at physiological levels (Llopis et al. 1998). Moreover, after blocking excitatory neurotransmission with APV and CNQX, the fluorescence slowly recovered over a period of about 5 min (Fig. 3B). In a second experiment, neurons expressing YFP-CaMKIIα were stimulated with bicuculline/glycine or NMDA/glycine for 5 min and then immediately fixed (Fig. 3C). Here, we observed the translocation of YFP-CaMKIIα fluorescence into spines with both treatments, as reported previously (Shen & Meyer, 1999). These data clearly show that the apparent loss of fluorescence of YFP-CaMKIIα following the addition of NMDA/glycine is largely due to quenching and not to a redistribution of the protein. This quenching effect is likely to have caused the reduction of YFP-βSAP97 fluorescence during NMDA application observed in an earlier study (Nakagawa et al. 2004). In our experiments the quenching by NMDA was more pronounced in neurons expressing EGFP-βSAP97 compared to αSAP97-EGFP (Fig. 2). This difference may result from α and βSAP97 occupying different subsynaptic compartments that may be differentially affected by pH changes (Degiorgis et al. 2008; Waites et al. 2009).

To evaluate the effect of cLTP and cLTD on the synaptic localization of EGFP-tagged α- and βSAP97 independently of quenching, we immediately fixed SAP97-expressing neurons after incubation in Tyrode buffer containing either bicuculline/glycine or NMDA/glycine (Fig. 2D and E). We found that neither cLTP nor cLTD induced significant changes in the localization of αSAP97 or βSAP97 (αSAP97: control spine/shaft levels 2.8 ± 0.2, LTD 2.9 ± 0.4, LTP 2.8 ± 0.4; βSAP97: control 2.5 ± 0.2, LTD 2.4 ± 0.1, LTP 2.5 ± 0.2; n≥ 11 fields of view), indicating that the induction of synaptic plasticity does not change the steady-state synaptic levels of either of the two SAP97 isoforms.

Synaptic plasticity does not alter the exchange rate of SAP97

Our previous imaging studies revealed that the steady-state levels of synaptic scaffold proteins such as SAP97 are composed of soluble, membrane- and cytoskeleton-associated pools (Waites et al. 2009). With regard to changes in synaptic strength, the most important is the size of the synaptic fraction of the scaffold proteins as this pool largely dictates the number and distribution of binding sites for neurotransmitter receptors within the PSD (Bredt & Nicoll, 2003; Bats et al. 2007; Waites et al. 2009). As such, shifts in the ratio of PSD-associated versus soluble pools of scaffold proteins could have a significant impact on receptor docking sites and synaptic strength. We therefore explored whether the induction of synaptic plasticity altered the exchange kinetics of EGFP-tagged α- or βSAP97 by performing fluorescence recovery after photobleaching (FRAP) experiments (Fig. 4).

Figure 4. SAP97 dynamics are unchanged during induction and expression of cLTD or cLTP.

A and B, the exchange of synaptic αSAP97-EGFP and EGFP-βSAP97 was measured by FRAP during and after the application of NMDA/glycine (cLTD) or bicuculline/glycine (cLTP) or under control conditions. Sample images of αSAP97-EGFP in lentivirus-infected hippocampal neurons (control condition) show the limited recovery of a bleached αSAP97 punctum (A, arrowhead) compared to the faster exchange of βSAP97 (B). C and D, cLTD or cLTP were induced in lentivirus-infected neurons expressing αSAP97-EGFP or EGFP-βSAP97. Fluorescence recovery of bleached synaptic SAP97 was recorded with a delay of 20 min after drug washout. C, fluorescence recovery of αSAP97-EGFP after cLTD (dark grey trace, n = 22) or cLTP treatment (light grey, n = 13) did not vary from the exchange kinetics observed during blocked synaptic activity (control, black, n = 27). D, quantification of FRAP showed that the recovery of synaptic EGFP-βSAP97 was unchanged between control conditions (black trace, n = 20) and the neurons that had undergone cLTP or cLTD treatments followed by a 20 min delay (light and dark grey traces, n = 20 for both conditions).

Since the expression of long-term changes in synaptic strength are acquired over time, we evaluated whether the exchange kinetics of αSAP97-EGFP and EGFP-βSAP97 were changed 20 min after the pharmacological induction of synaptic plasticity (Fig. 4C and D). However, we could not detect any changes in the exchange kinetics of α- or βSAP97 after the induction of LTP or cLTD compared to control conditions. Therefore, the dynamics of α- and βSAP97 seem to be independent not only of the acute level of synaptic activity but were also unchanged by the chemical induction of LTP or LTD in these cultures, putting into doubt to what extent synaptic plasticity is actually involved in the regulation of SAP97 dynamics and vice versa. Taken together our data demonstrate that conditions that alter the synaptic strength in our system do not change the synaptic levels or protein dynamics of αSAP97 or βSAP97. This implies that the observed failure to induce LTP in neurons overexpressing α- or βSAP97 is not linked to changes in the dynamic properties of the scaffold proteins.

SAP97 regulates the synaptic localization of NMDARs

The activation of NMDARs is critical for the induction of many forms of synaptic plasticity. As LTP was absent in both αSAP97- and βSAP97-expressing neurons, we investigated whether SAP97 isoforms altered the synaptic expression of NMDARs. Using paired whole cell recordings, we first examined whether evoked NMDAR-mediated EPSCs between individual pairs of pyramidal neurons were altered by the overexpression of αSAP97 or βSAP97 in the postsynaptic neuron. Isolated synaptic NMDAR-mediated currents were measured between individual cultured hippocampal pyramidal neurons in response to presynaptic action potentials.

In control neurons, NMDAR-mediated EPSC amplitudes remained constant throughout the length of the paired recordings (minimum of 20 min; average NMDAR EPSC amplitude was 30.30 ± 1.28 pA, n = 10 pairs; Fig. 5A). Interestingly, we found that αSAP97 and βSAP97 had differential effects on NMDAR-mediated EPSCs. In neurons expressing αSAP97, NMDAR EPSC amplitudes were not significantly different from control EPSCs (29.77 ± 0.78 pA, n = 12 pairs; P > 0.1). However, NMDAR EPSC amplitude was remarkably decreased in neurons expressing βSAP97 (9.41 ± 0.61 pA; n = 13 pairs; Fig. 5A), which was significantly decreased from both control and αSAP97-expressing neurons (P < 0.001 in both cases). This reduction in NMDAR-mediated EPSC amplitude in βSAP97-expressing neurons was not a result of a decrease in synapse number as we found no effect of either αSAP97 or βSAP97 on spine density in hippocampal neurons: in βSAP97-expressing neurons α-actinin density was 1.18 ± 0.13 μm⁻¹, which was not significantly different from control neurons (1.01 ± 0.09 μm⁻¹; P > 0.05) or αSAP97-expressing neurons (0.98 ± 0.16 μm⁻¹; P > 0.05).

Figure 5. βSAP97 is a negative regulator of synaptic NMDARs.

A, NMDAR EPSC amplitudes measured from paired recordings between pyramidal neurons in hippocampal cultures. NMDAR EPSCs were measured at +40 mV and paired recordings were maintained for a minimum of 20 min. Inset: representative traces of NMDAR EPSCs in control, α- and βSAP97-expressing neurons. Five postsynaptic traces are shown overlaid for each group, with 1 example of a presynaptic action potential. B, surface NMDAR-mediated currents evoked by focal application of NMDA to α- and βSAP97-expressing neurons. Right: example traces of surface NMDAR-mediated currents shown overlaid for control, α- and βSAP97-expressing neurons.

In order to determine whether the surface expression of functional NMDARs, comprising both synaptic and extrasynaptic pools, is altered in αSAP97- versusβSAP97-expressing neurons, we measured the surface NMDAR-mediated currents in hippocampal neurons by focal application of NMDA. Interestingly, despite the fact that βSAP97 decreases synaptic NMDARs, neurons expressing βSAP97 were more responsive to exogenous NMDA than those expressing αSAP97. NMDA-evoked current amplitudes were 112.47 ± 17.90 pA in βSAP97-expressing neurons, significantly higher than both αSAP97-expressing neurons and control neurons (αSAP97: 52.29 ± 5.38 pA, control: 60.00 ± 5.84 pA; both significantly different from βSAP97-expressing neurons, P < 0.001; Fig. 5C). Therefore, while βSAP97-expressing neurons have fewer functional NMDARs at the synapse for the detection of synaptically released glutamate, βSAP97 supports a significantly higher total surface expression of functional NMDARs.

Specific knockdown of βSAP97 increases synaptic expression of AMPA and NMDA receptors

To determine whether βSAP97 negatively controls the size of the synaptic pool of NMDARs in the endogenous situation, we used a short hairpin RNA (shRNA) to specifically downregulate the expression of endogenous βSAP97 in hippocampal neurons. Our previous studies have shown that this shRNA suppresses the expression of βSAP97 by more than 90% (Jeyifous et al. 2009). In contrast, previous studies have designed shRNA to the common GUK domain of SAP97 or utilized Dlgh1 mice lacking any functional SAP97 protein and therefore result in knockdown of all SAP97 isoforms (Howard et al. 2010). We found that specific knockdown of βSAP97 reverses the inhibitory effect previously observed on synaptic NMDAR-mediated EPSCs and results in a significant increase in synaptic NMDARs (Fig. 6A). Paired recordings revealed that the amplitude of NMDAR EPSCs in shRNA-expressing neurons were significantly higher than control neurons and in neurons expressing a scrambled shRNA (average NMDAR EPSC amplitude in shRNA-expressing neurons was 115.44 ± 47.03 pA (n = 6 pairs), compared with 33.87 ± 8.15 pA in control neurons (n = 5 pairs) and 33.50 ± 5.95 pA in neurons expressing a scrambled shRNA (n = 5 pairs); P < 0.01). Similarly, knockdown of βSAP97 reversed the previously observed decrease in synaptic AMPARs and also resulted in an increase in synaptic AMPARs (average AMPAR EPSC amplitude was 554.86 ± 287.42 pA in shRNA-expressing neurons (n = 6 pairs), compared with 141.18 ± 40.10 pA in control neurons (n = 5 pairs) and 139.38 ± 21.35 pA in neurons expressing the scrambled shRNA (n = 5 pairs); P < 0.01; Fig. 6B). The original effect of βSAP97, that is a decrease in both AMPAR and NMDAR EPSC amplitude compared with controls, could be rescued by expression of a modified βSAP97* that is resistant to knockdown (see Methods; Fig. 6A and B). Together these data reveal that endogenous βSAP97 decreases the synaptic pools of NMDARs and AMPARs.

Figure 6. Knockdown of endogenous βSAP97 releases extrasynaptic receptors to the synaptic pool.

A, knockdown of endogenous βSAP97 results in a significant increase in the amplitude of NMDAR-mediated EPSCs that was not observed in the shRNA scrambled controls. Expression of the altered form βSAP97* that is resistant to knockdown rescued the phenotype observed in βSAP97-expressing neurons. Example traces of NMDAR-mediated EPSCs are shown for control (untransfected), shRNA βSAP97 and βSAP97* with a representative action potential. Scale bars: 50 pA/50 ms and 20 mV/20 ms. B, knockdown of endogenous βSAP97 also significantly increases the amplitude of AMPAR mediated EPSCs. Expression of the altered form βSAP97* that is resistant to knockdown rescued the phenotype observed in βSAP97-expressing neurons. In both A and B, average NMDAR and AMPAR EPSCs were measured from paired whole cell recordings in which the postsynaptic cells were either untransfected (controls), or expressing shRNA βSAP97, scrambled shRNA or βSAP97* (see Methods). Example traces of AMPAR-mediated EPSCs are shown for control (untransfected), shRNA βSAP97 and βSAP97* with a representative action potential. Scale bars: 100 pA/100 ms and 20 mV/20 ms.

We also measured surface (i.e. synaptic + extrasynaptic) NMDAR- and AMPAR-mediated currents in βSAP97 shRNA-expressing neurons with exogenous application of NMDA or AMPA, respectively. In response to exogenous application of NMDA, average surface NMDAR current amplitude was 136.98 ± 13.49 pA (n = 21 neurons) in βSAP97 shRNA-expressing neurons. In response to exogenous application of AMPA, average surface AMPAR current amplitude was 435.12 ± 59.57 pA (n = 24 neurons). Neither the surface NMDAR- nor AMPAR-mediated current amplitudes in βSAP97 shRNA-expressing neurons were significantly different from the synaptic NMDAR or AMPAR EPSC amplitudes in βSAP97 shRNA-expressing neurons (NMDAR EPSC 115.44 ± 47.03 pA Fig. 6A, AMPAR EPSC 554.86 ± 287.42 pA Fig. 6B; P > 0.05). Together these data are consistent with the redistribution of NMDA and AMPA receptors from extrasynaptic to synaptic sites in the absence of βSAP97.

SAP97 isoforms alter the distribution of NMDARs during synaptic plasticity

The differential localization of NMDARs at synaptic and extrasynaptic sites in αSAP97 and βSAP97-expressing neurons, respectively, suggests that N-terminal isoforms of SAP97 may regulate the size of the NMDAR pools and contribute to a shift in NMDAR distribution between synaptic and extrasynaptic sites during synaptic plasticity. We therefore measured the amplitude of the surface NMDAR-mediated current in response to exogenous NMDA before and 20 min after the pharmacological induction of LTP or LTD (Fig. 7). We found that the surface pool of NMDARs was not altered by cLTP in control neurons or in neurons expressing αSAP97 (control: 123.6 ± 14.6%, n = 13; αSAP97: 116.5 ± 30.2% of baseline total surface NMDAR-mediated currents, n = 9; Fig. 7A). However, we observed that βSAP97 induced a significant increase in the surface pool of NMDARs after the pharmacological induction of LTP (202.6 ± 10.9% of baseline surface NMDAR current amplitude; n = 9, P < 0.001; Fig. 7A). To determine whether the βSAP97-induced increase in surface NMDARs was induced by changes in synaptic or extrasynaptic NMDARs, we performed paired whole cell recordings to measure isolated NMDAR-mediated EPSCs. The amplitudes of the NMDAR EPSCs were monitored before and after the pharmacological induction of LTP or LTD in control, α- and βSAP97-expressing neurons (Fig. 7). We found that cLTP induction caused a significant increase in the amplitude of the synaptic NMDAR EPSCs in control and αSAP97-expressing neurons (baseline current EPSC amplitude increased from 37.85 ± 6.4 pA to 58.5 ± 25.9 pA in control neurons 30 min post-cLTP, n = 6 pairs, and from 34.18 ± 8.27 pA to 55.57 ± 9.4 pA in αSAP97-expressing neurons 30 min post-cLTP, n = 8 pairs; Fig. 7B). In contrast, no significant change occurred in the amplitude of the NMDAR EPSCs in βSAP97-expressing neurons (EPSC amplitudes were 16.7 ± 2.1 pA and 13.4 ± 1.97 pA, prior to and 30 min after LTP, respectively, n = 10 pairs; Fig. 7B). Therefore, the LTP-induced increase in total surface NMDAR mediated currents in βSAP97-expressing neurons must result from an increase in the size of the extrasynaptic pool of NMDARs.

Figure 7. Modulation of NMDARs by cLTP and cLTD in α- and βSAP97-expressing neurons.

A, surface NMDAR-mediated current amplitudes evoked by focal application of NMDA before and 20 min after the induction of cLTP in control, α- and βSAP97-expressing neurons. Amplitudes are expressed as a percentage of the baseline NMDAR current amplitude. B, NMDAR EPSC amplitudes measured from paired recordings between hippocampal neurons. cLTP was induced 5 min after attaining the postsynaptic whole cell recording. NMDAR EPSCs were measured at +40 mV and paired recordings were held for 40 min. C, surface NMDAR-mediated currents evoked by focal application of NMDA (1 mm) were measured before and 20 min after the induction of cLTD. Amplitudes are expressed as a percentage of the baseline NMDAR current amplitude. D, NMDAR EPSC amplitudes measured from paired recordings between hippocampal neurons. cLTD was induced 5 min after the attainment of a postsynaptic whole cell recording and paired recordings were maintained for 30 min. All presynaptic neurons were untransfected, while the postsynaptic partner was either untransfected (control) or expressed α- or βSAP97-EGFP.

We also examined how the pharmacological induction of LTD altered the amplitude of surface NMDAR-mediated currents in control, α- and βSAP97-expressing neurons (Fig. 7C). Five minutes after cLTD induction, surface NMDAR-mediated currents were significantly decreased in control neurons (average total NMDAR-mediated surface current was 55.6 ± 13.1% of baseline, P < 0.01, n = 6; not shown), but recovered to near baseline levels 20 min after the pharmacological induction of LTD (Fig. 7C: 87.8 ± 13.4% of baseline current, P > 0.05, n = 6). Similarly, 5 min after cLTD induction, we found that surface NMDAR-mediated currents were significantly decreased in α- and βSAP97-expressing neurons (αSAP97: 56.2 ± 14.8% baseline current, P < 0.01, n = 6; βSAP97: 48.8 ± 6.6% baseline current, P < 0.01, n = 9; not shown). However in α- and βSAP97-expressing neurons, the depression of NMDAR-mediated currents did not recover, such that 20 min after cLTD induction total surface NMDAR-mediated currents were still significantly below baseline levels (αSAP97: 67.6 ± 14.6% and βSAP97: 44.8 ± 7.2% of baseline currents, n = 6 and 9, respectively, P < 0.01 in both cases; Fig. 7C).

To determine whether these changes in NMDAR-mediated currents occurred at synaptic receptors, we performed paired whole cell recordings to measure the amplitude of isolated NMDAR EPSCs in control, α- and βSAP97-expressing neurons. We found that NMDAR EPSC amplitudes decreased significantly in control neurons, as previously described for electrically evoked LTD (Montgomery & Madison, 2002; Montgomery et al. 2005). Similarly, in α- and βSAP97-expressing neurons, we found that NMDAR EPSC amplitudes decreased significantly and that this decrease persisted for the length of the paired recording (≥35 min; Fig. 7D; control: average baseline EPSC amplitude decreased from 31.51 ± 2.59 pA to 12.68 ± 1.66 pA at 25 min post-LTD, n = 7 pairs; αSAP97: from 31.42 ± 3.28 pA to 16.14 ± 3.94 pA at 25 min post-LTD, n = 6 pairs; βSAP97: from 14.93 ± 0.81 pA to 11.87 ± 0.95 pA at 25 min post-LTD, n = 7 pairs). Thus, in response to cLTD the size of the synaptic pool of NMDARs decreases in control, α- and βSAP97-expressing neurons, albeit to varying degrees. These decreases are likely to contribute to the observed reduction in surface NMDAR-mediated currents, but the differences we observed in the changes of these two pools of NMDARs again reflect the differential regulation of synaptic and extrasynaptic NMDARs by αSAP97 and βSAP97.

Discussion

In this study we have examined how N-terminal SAP97 isoforms regulate the induction and the expression of synaptic plasticity, and whether this occurs through activity-dependent regulation of the number of synaptic binding sites, or through the differential localization of NMDARs and AMPARs. The use of pharmacological protocols to induce plasticity enabled us to directly examine how the same protocols influence synapse function via electrophysiology and protein dynamics using live cell imaging. Our data suggest that the capacity to induce and express synaptic plasticity is at least in part determined by how many receptors are bound to either the PSD-dominant palmitoylated αSAP97, inducing increased clustering of AMPARs within the PSD, or to extrasynaptic L27 domain-containing βSAP97 that sequesters both AMPARs and NMDARs at extrasynaptic sites.

αSAP97 occludes synaptic potentiation by increasing synaptic AMPARs

α- and βSAP97 appear to prevent LTP via different mechanisms. As NMDAR-mediated synaptic currents were entirely normal in αSAP97-expressing neurons, these neurons possess the ability to induce LTP. We believe that the lack or the reduction of LTP is due to an increased steady-state localization of AMPARs in the PSD by αSAP97 (Fig. 8) (Waites et al. 2009). Therefore LTP and αSAP97 overexpression appear to engage the same mechanisms to increase synaptic strength and the expression of LTP is occluded by αSAP97. As LTP was not accompanied by a change in αSAP97 exchange kinetics, this reveals that the number of αSAP97 AMPAR binding sites does not change, but that during LTP αSAP97 promotes the selective localization of AMPARs to the synaptic pool.

Figure 8. Model of SAP97 isoform-dependent regulation of synaptic glutamate receptor localization.

Our physiological data reveal that α- and β-isoforms of SAP97 play strikingly different roles in controlling the localization of AMPA- and NMDA-type glutamate receptors at synaptic and extrasynaptic sites. A, βSAP97 creates docking sites for AMPA and NMDA receptors out of the PSD, sequestering AMPA and NMDA receptors at extrasynaptic sites where they are unable to contribute to the induction and expression of LTP. The net result is a greater shift in receptor localization to the extrasynaptic sites. Both receptor types are able to freely diffuse away from the synaptic site, however, and be internalised in response to LTD. B, αSAP97 creates docking sites for AMPA receptors in the PSD, shifting the localization of AMPA receptors into the synapse and increasing synaptic strength. No effect of αSAP97 on NMDAR localization or function was evident in our studies.

βSAP97 regulates plasticity by sequestering NMDARs outside of synapses

Our observation that βSAP97 prevents LTP, decreases synaptic NMDARs and increases surface NMDARs reveals that βSAP97 localizes NMDARs to extrasynaptic sites, as it does with AMPARs (Waites et al. 2009). This is supported by converse experiments where in the absence of βSAP97, synaptic AMPARs and NMDARs are increased whereas extrasynaptic receptors are decreased, indicating that a property of endogenous βSAP97 is to exclude these receptors from the synapse (Fig. 8). These data have led us to the conclusion that βSAP97 is a negative regulator of synaptic pools of NMDARs as well as AMPARs, consequently preventing the induction of LTP.

A previous model proposed that βSAP97 functions to either (a) facilitate the delivery of AMPARs to the extrasynaptic membrane from intracellular pools or (b) trigger the release of AMPARs that are retained in an extrasynaptic pool so they can be translocated to synapses (Rumbaugh et al. 2003; Schluter et al. 2006). While this model was in relation to AMPARs, our independent measurement of synaptic and surface NMDARs and AMPARs has enabled us to distinguish between these two possibilities. The observed high extrasynaptic levels of glutamate receptors are consistent with βSAP97 facilitating the delivery of AMPARs and NMDARs to the extrasynaptic membrane from intracellular pools. However, our data indicate that these receptors are sequestered outside of the synapse, preventing the induction of LTP and the movement of extrasynaptic AMPARs into synaptic sites.

Previous studies have produced conflicting data on the effect of βSAP97, and the deletion or knockdown of SAP97, on AMPAR and NMDAR EPSCs (Nakagawa et al. 2004; Schluter et al. 2006); (Howard et al. 2010). A major reason for these discrepancies lies in the differing strategies employed to knockdown the expression of SAP97. In previous studies, only the effects of βSAP97 overexpression were examined. However, in the loss-of-function experiments the expression of all SAP97 isoforms was decreased by knockdown/knockout targeted to the GUK domain or exon 4 of SAP97 (Nakagawa et al. 2004; Howard et al. 2010). Therefore the combined synaptic effects of all SAP97 isoforms were reflected in these data. In the current study, we designed an shRNA to specifically target βSAP97, enabling us to exclusively determine the role of this dominant isoform in regulating the extrasynaptic pools of AMPA and NMDA receptors. Moreover, the extracellular stimulation employed in the previous studies may have activated both synaptic and extrasynaptic receptors, preventing the specific effect on extrasynaptic receptors being identified. Furthermore, the data discrepancies could also be caused by I3/I4/I5 inserts which are present in SAP97 (SAP97 with I3 and I5 was used in the present study; not stated in other studies) and also the level and temporal expression profiles of βSAP97 having different effects on AMPAR and NMDAR currents (Howard et al. 2010). The fact that βSAP97 is also important in the forward trafficking of glutamate receptors (Sans et al. 2001; Jeyifous et al. 2009) adds yet another degree of complexity to the regulation of receptors by βSAP97. Our previously observed reduction in NR1 puncta intensity in βSAP97-deficient hippocampal neurons (Jeyifous et al. 2009) is likely to reflect a decrease in βSAP97-bound extrasynaptic surface receptors. However, by specifically decreasing βSAP97 expression in the present study, we further show that independent of its role in receptor trafficking, βSAP97 shifts the dynamic equilibrium of glutamate receptors from a synaptic to an extrasynaptic location.

Glutamate receptors have previously been proposed to be differentially regulated in hippocampal organotypic slices in comparison to dissociated hippocampal cultures, with GluR1-containing AMPARs being inserted upon stimulation in slices but under basal conditions in dissociated cultures (Shi et al. 1999). Our data support this conclusion in that we were able to induce LTP in αSAP97-expressing neurons in hippocampal slices but not in dissociated cultures, suggesting that a proportion of GluR1-containing receptors remained intracellular in the slice preparation and were inserted in response to LTP stimuli. As a result LTP was partially occluded by the expression of αSAP97, but was fully occluded in the dissociated culture system where GluR1 readily moves to the surface under basal conditions. Intriguingly, LTP was also blocked by the expression of βSAP97 in hippocampal organotypic slices, consistent with our conclusion from dissociated hippocampal cultures that βSAP97 prevents the movement of receptors into the synaptic space that is required for the expression of LTP.

The observed differences in synaptic AMPAR- and NMDAR-mediated synaptic currents induced by the expression of α- or β-SAP97 do not appear to result from differential effects on synapse number as we observed no effect of either αSAP97 or βSAP97 on spine density in hippocampal neurons or in surface GluR1 puncta (Waites et al. 2009). In addition, the significant decrease in the amplitude of the AMPAR- and NMDAR-mediated synaptic currents in βSAP97-expressing neurons, and the selective increase in the amplitude of the AMPAR- but not NMDAR-mediated synaptic currents in αSAP97-expressing neurons is not consistent with SAP97 isoforms increasing presynaptic glutamate release. Effects of SAP97 on presynaptic function have previously been observed in response to chronic overexpression of βSAP97 in vivo from E16–P8, but presynaptic changes were not observed in immature neurons acutely expressing βSAP97 (Howard et al. 2010). Thus, our results likely reflect that N-terminal SAP97 isoforms regulate receptor targeting but not presynaptic function in early development.

α- and β-isoforms also exist for the MAGUK proteins PSD-95 and PSD-93. As the dominant SAP97 variant expressed at synapses is the β-isoform, and αPSD-95 the main form of PSD-95 (Chetkovich et al. 2002; Schluter et al. 2006), functional specialization may occur at the synapse as αPSD-95 stabilizes AMPARs in the PSD, while βSAP97 stabilizes receptors in the extrasynaptic pool. Similar to αSAP97, the α-isoform of PSD-95 has been shown to promote the synaptic clustering of AMPARs (Schluter et al. 2006), suggesting that there is some functional redundancy between α-isoforms with regards to LTP. The strong effects produced by βSAP97 expression or knockdown on AMPARs and NMDARs demonstrate that the L27 domain of βSAP97 bestows a unique and functionally important set of properties to this isoform. Our data also imply that the L27 domains in PSD-93 and PSD-95 may similarly perform functions that modulate the ratios of synaptic versus extrasynaptic receptor complexes and thus may regulate synaptic plasticity in unexpected ways.

Following cLTP treatment, βSAP97 induced an increase in the extrasynaptic NMDAR pool, revealing that the size of extrasynaptic pools of receptors can also be regulated by plasticity. As there was no parallel increase in the synaptic NMDARs, these extrasynaptic receptors do not appear to merge with synaptic receptor pools as occurs during LTP in control neurons. Currently the role of the large extrasynaptic NMDAR pool in βSAP97-expressing neurons is unknown, but extrasynaptic signalling through NMDARs has been shown to be detrimental, activating cell death pathways and contributing to pathological conditions (Groc et al. 2009; Milnerwood et al. 2010). Therefore βSAP97 may play a role in the altered NMDAR signalling and trafficking observed in conditions such as Alzheimer's and Huntington's disease, schizophrenia and addiction (Marcello et al. 2007; Groc et al. 2009; Milnerwood et al. 2010).

As LTD could still be induced in α- and βSAP97-expressing neurons, N-terminal SAP97 isoforms allow receptor movement away from synaptic sites. In control, α- and βSAP97-expressing neurons, depression of the amplitude of AMPAR- and NMDAR-mediated currents was observed in the synaptic and the surface pools, reflecting a reduction in the size of both the synaptic and extrasynaptic receptor pools during LTD. This is consistent with the movement of AMPARs and NMDARs to membrane sites lateral to the PSD and their subsequent internalization at extrasynaptic endocytic zones to drive synapses to a depressed state (Blanpied et al. 2002; Montgomery & Madison, 2002; Ashby et al. 2004; Racz et al. 2004; Montgomery et al. 2005; Petrini et al. 2009). The internalization of NMDARs was most pronounced in neurons expressing βSAP97, where we observed a strong reduction of the surface receptor levels after LTD. This may result from the high baseline levels of extrasynaptic surface NMDARs in these neurons that facilitate NMDAR-dependent cLTD induction and receptor movement to extrasynaptic endocytic sites.

Synaptic plasticity does not affect the dynamics or distribution of N-terminal SAP97 isoforms

Previous studies have shown contradictory results with respect to the activity dependence of SAP97 localization and dynamics. Treatment of hippocampal neurons with NMDA has been suggested to recruit βSAP97 to the synapse by CaMKII phosphorylation at Ser39 (Mauceri et al. 2004), or conversely to lead to the dissipation of βSAP97 from synapses (Nakagawa et al. 2004). In contrast, we have shown that neither α- nor βSAP97 localization or dynamics was altered by the chemical LTP or LTD protocols. These differences could be due to fluoresence quenching described earlier, the βSAP97 isoform used in these studies (βSAP97 containing the I3/I5 inserts in our experiments; not stated in the other reports), or the level of SAP97 overexpression in comparison to the low levels expressed in the current study by lentiviral infection. Another possibility is the differences in concentration and length of treatment with NMDA utilized in the previous studies (50 μm for 15 min or 3 μm for 20 min). It is not known what, if any, changes in synaptic strength were induced by these protocols. In contrast, the protocols used in the current study were utilized in both imaging and electrophysiology experiments, and were shown in control neurons to induce LTP or LTD and to cause the translocation of CaMKII to synapses.

Ultimately, we aimed to determine whether there is a relationship between SAP97 dynamics and synaptic plasticity. We found little evidence that either α or βSAP97 alter their distribution or exchange kinetics in response to changes in synaptic activity. This implies that they exert their effects on synaptic function by creating surface glutamate receptor docking sites within the PSD and extrasynaptic space. At present, we find no evidence to support a model wherein either α- or βSAP97 performs a chaperone function for these receptors. This is not to say that receptor binding to these scaffold proteins is not regulated. Our cLTD data clearly demonstrate that surface AMPA and NMDA receptors can uncouple from the cytoskeletal matrix and be internalized. Such a model is consistent with studies showing that both AMPA and NMDA receptor binding to scaffold proteins is regulated by post-translational modifications such as phosphorylation and that SAP97 isoforms use their multidomain structure to create tertiary complexes with both kinase and phospatases (AKAP/calcineurin) capable of responding to changes in synaptic signalling (Colledge et al. 2000; Tavalin et al. 2002; Montgomery et al. 2004; Dell'Acqua et al. 2006; Shepherd & Huganir, 2007). Taken together, our data show that α- and βSAP97 do not regulate synaptic plasticity via activity-dependent changes in their synaptic localization or dynamics, but rather are a platform for regulating the tethering of receptors in synaptic or extrasynaptic compartments.

Conclusions

We have performed parallel functional and imaging studies that revealed that N-terminal α and β splice variants of SAP97 differentially regulate the synaptic localization of AMPARs and NMDARs, resulting in major changes in the ability of neurons to express synaptic plasticity. Our data show that αSAP97 is important for localizing AMPARs in the PSD and increases the synaptic strength, but that these receptors are readily removed from the synapse in response to LTD. In contrast, βSAP97 appears to be a negative regulator of synaptic potentiation, favouring a shift towards synaptic depression by sequestering both AMPARs and NMDARs at extrasynaptic sites. As a result βSAP97 can block and/or modulate both the induction and the expression of synaptic plasticity. Together these data reveal that synaptic SAP97 isoforms can play functionally distinct roles in regulating glutamate receptor levels at synapses and as such influence synaptic plasticity mechanisms.

Glossary

Abbreviations

AMPAR

AMPA receptor

CaMKII

calcium–calmodulin-dependent kinase II

cLTD

chemical LTD

cLTP

chemical LTP

LTD

long-term depression

LTP

long-term potentation

NMDAR

NMDA receptor

PSD

postsynaptic density

SAP97

synapse associated protein of 97 kDa

shRNA

short hairpin RNA

Author contributions

Electrophysiology experiments were performed and analysed in the Montgomery lab by D.L., C.B.M. and J.M.M. Supporting electrophysiology experiments were performed by D.G. All imaging experiments and analysis were performed in the Garner lab by C.G.S. and C.L.W. Minor immunocytochemistry experiments were performed by J.W.F. and D.L. SAP97 constructs were built by C.G.S., S.O. and C.C.G. J.M.M., C.C.G., D.L., C.G.S. and C.L.W. conceived and designed the experiments. The manuscript was written by J.M.M., C.C.G., D.L. and C.G.S. All authors have approved the final version of the manuscript.

Supplementary material

Figure S1

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors