Extrasynaptic Membrane Trafficking Regulated by GluR1 Serine 845 Phosphorylation Primes AMPA Receptors for Long-term Potentiation*

Enhancement of synaptic transmission, as occurs in long-term potentiation (LTP), can result from several mechanisms that are regulated by phosphorylation of the AMPA-type glutamate receptor (AMPAR). Using a quantitative assay of net serine 845 (Ser-845) phosphorylation in the GluR1 subunit of AMPARs, we investigated the relationship between phospho-Ser-845, GluR1 surface expression, and synaptic strength in hippocampal neurons. About 15% of surface AMPARs in cultured neurons were phosphorylated at Ser-845 basally, whereas chemical potentiation (forskolin/rolipram treatment) persistently increased this to 60% and chemical depression (N-methyl-d-aspartate treatment) decreased it to 10%. These changes in Ser-845 phosphorylation were paralleled by corresponding changes in the surface expression of AMPARs in both cultured neurons and hippocampal slices. For every 1% increase in net phospho-Ser-845, there was 0.75% increase in the surface fraction of GluR1. Phosphorylation of Ser-845 correlated with a selective delivery of AMPARs to extrasynaptic sites, and their synaptic localization required coincident synaptic activity. Furthermore, increasing the extrasynaptic pool of AMPA receptors resulted in stronger theta burst LTP. Our results support a two-step model for delivery of GluR1-containing AMPARs to synapses during activity-dependent LTP, where Ser-845 phosphorylation can traffic AMPARs to extrasynaptic sites for subsequent delivery to synapses during LTP.

At central glutamatergic synapses, AMPAR 2 function is dynamically regulated, resulting in bidirectional synaptic plasticity, such as longterm potentiation (LTP) and long-term depression (LTD). Phosphorylation can regulate AMPARs by two distinct mechanisms that affect synaptic transmission (1): modulation of channel properties (2)(3)(4) and trafficking of AMPARs to the surface membrane (5)(6)(7)(8)(9). Phosphorylation of Ser-831 in GluR1 C terminus by calcium/calmodulin-dependent protein kinase II (CaM-KII) (10,11) potentiates single-channel conductance (2,4), and phosphorylation of Ser-845 by cAMP-dependent protein kinase (PKA) (12) increases channel open probability (3). Importantly, Ser-845 phosphorylation can also regulate the surface expression of AMPARs by increasing the pool of GluR1 recycled back to the surface after their endocytosis (13). CaM-KII activity is also critical for synaptic incorporation of AMPARs but requires a serine at position 845, which is not phosphorylated by CaM-KII, in GluR1 (14), suggesting an important role of Ser-845 phosphorylation for trafficking of AMPARs in an activity-dependent manner.
Although these studies indicate that AMPAR phosphorylation and trafficking are coupled (13,14), exactly how Ser-845 phosphorylation contributes to trafficking is unclear. Current understanding of the functional role of Ser-845 phosphorylation in GluR1 during synaptic plasticity is limited because previous studies have analyzed only relative changes in phospho-Ser-845. In fact, the functional significance of these changes is determined by the net level of phospho-Ser-845. In this report, we quantified the net Ser-845 phosphorylation and surface expression of GluR1 in hippocampal neurons during bidirectional synaptic plasticity. We observed that receptors phosphorylated at the Ser-845 trafficked specifically to extrasynaptic sites but not to synapses. However, extrasynaptic AMPARs can be incorporated into synapses by synaptic activity. Thus, Ser-845 phosphorylation serves as a "priming" step in enhancing synaptic strength during LTP by enhancing the extrasynaptic delivery of AMPARs.

MATERIALS AND METHODS
Cultured Hippocampal Neurons and Biochemistry-Hippocampal neurons were cultured from postnatal day 1-2 Sprague-Dawley rats as previously described (15,16) and used on 13-15 days in vitro. For chemical stimulations, cultures were first incubated in ACSF for 30 min at room temperature (in mM): 125 NaCl, 2.5 KCl, 1 MgCl 2 , 2 CaCl 2 , 33 D-glucose, and 25 HEPES (pH 7.3; 320 mosM final), followed by stimulation with 50 M forskolin (Sigma) and 0.1 M rolipram (Calbiochem) in ACSF (no MgCl 2 ) or 50 M NMDA. After 10 min of stimulation, neurons were replaced in regular ACSF and then subjected to surface biotinylation at indicated time points.
Surface Biotinylation-After F/R or NMDA stimulations, cultured neurons or slices were transferred to ice-cold ASCF for 2 min, followed by biotinylation in 1 mg/ml of biotin (EZ-Link Sulfo-NHS-SS-Biotin; Pierce) for 20 or 45 min with slow agitation for cultured neurons or slices, respectively. This results in complete biotinylation throughout 400-m hippocampal slices (17). After three rinses in cold Tris-based ACSF (HEPES replaced with Tris) to quench free biotin, slices were snap frozen in liquid nitrogen. Cultures were immediately scraped without freezing in cold 1% Triton X-100 homogenization buffer (150 l/35-mm well) (in mM: 50 NaCl, 10 EDTA, 10 EGTA, 1 Na 3 VO 4 , 50 NaF, 25 NaPPi, 1 ␤-glycerophosphate, 1 phenylmethylsulfonyl fluoride, 0.001 microcystine, 1ϫ protease inhibitor mixture tablet (Roche Applied Science), 1ϫ phosphatase inhibitor mixture set I (Calbiochem), and 50 HEPES (pH 7.5). Solubilization was performed in 1% Triton X-100 because this protocol does not solubilize postsynaptic densities (36) and is therefore selective for extrasynaptic AMPARs. Two frozen slices were pooled for each condition and homogenized in 150 l of cold 1% Triton X-100 homogenization buffer. An additional 150 l of homogenization buffer was added to obtain 300 l of total slice homogenate. Homogenates from cultures (2 wells per condition) and slices were centrifuged at 10,000 ϫ g for 20 min to pellet the insoluble fraction. For the total fraction of GluR1 (surface plus internal), 75 l of the supernatant was mixed and heated with 25 l of 6ϫ SDS sample buffer. Biotinylated surface proteins in the remaining supernatant (ϳ225 l) were immunoprecipitated with 40 l of 50% avidin-agarose beads (ImmunoPure Immobilized Avidin; Pierce) for Ͼ2 h at 4°C. The beads were pelleted, and 75 l of the supernatant (internal fraction) was mixed and heated with 25 l of 6ϫ SDS sample buffer. The beads were then rinsed three times with 1% Triton X-100 homogenization buffer and heated in 100 l of 2ϫ SDS sample buffer (surface fraction). Equal volumes of the total and internal fractions were subjected to 10% SDS-PAGE, probed for total GluR1, and normalized to tubulin. The surface fraction was calculated as the following: surface ϭ 1 Ϫ (internal/total).
Quantitative Western Blotting-Primary antibodies were: anti-phospho-Ser-831-GluR1 (UBI), anti-phospho-Ser-845-GluR1 (UBI and Chemicon), anti-GluR1 (UBI), and anti-tubulin (Developmental Studies Hybridoma Bank). All surface GluR1 obtained from cultured neurons and slices were subjected to 10% SDS-PAGE with a set of calibration samples (at least 5/gel) on the same and every blot for net quantitation of phospho-Ser-845 (Fig. 2a). Calibration samples were prepared by phosphorylating a GST-fused consensus phosphorylation sequence (CPS) mutant for PKA (P842R) (10 M) with PKA (100 nM) in vitro for 30 min at 30°C in phosphorylation buffer (Fig. 1d). This fully phosphorylated Ser-845 GluR1 was mixed with specific amounts of non-phosphorylated proteins (10 M) to obtain 4,8,12,16,20,40,60,80, and 100% phosphorylated calibration samples. Gels were transferred to nitrocellulose membrane and immunoblotted using infrared dyecoupled secondary antibodies (anti-rabbit IgG Alexa Fluor 680, antimouse IgG IRdye800; Rockland). Image acquisition and data quantitation were performed on the Odyssey Infrared Imaging system (Li-Cor). Preliminary blots were run to normalize for the differences in GluR1 amount between biotinylated GluR1 and calibration samples. Calibration curves were calculated by determining the fraction of phospho-Ser-845 signals over total GluR1 signals and then plotting this fraction against percentage of phospho-Ser-845 of calibration standards. Only linear curves with R 2 Ͼ0.90 were used for analysis (Fig. 2a).
Electrophysiology and Biochemistry of Hippocampal Slices-Acute hippocampal slices were prepared and recorded from 4 -6-week-old male Sprague-Dawley rats as previously described (18,19). Slice recording buffer was saturated with 95% O 2 /5% CO 2 (in mM): 125 NaCl, 2.5 KCl, 22.6 NaHCO 3 , 1.25 NaH 3 PO 4 , 2 CaCl 2 , 1 MgCl 2 , and 11.1 D-glucose (pH 7.4). Recordings were digitized at 100 kHz, and initial linear slopes of fEPSPs were analyzed. Basal stimulation was induced every min, and recordings were averaged to 3-min bins for analysis. Control and test conditions were interleaved within each animal to control for differences between animals. After 20 min of stable baseline, 25 M forskolin and 0.1 M rolipram (no MgCl 2 ) were added in slice recording buffer for 10 min with or without basal stimulation. Theta burst stimulation protocols were used as described (18). After recordings, slices were snap frozen, microdissected for CA1 region, homogenized in 150 l of ice-cold 1% Triton X-100 homogenization buffer, and subjected to 10% SDS-PAGE with calibration samples for quantitative phospho-Ser-845 Western blotting. For Chem-LTD experiments (Fig. 2, b and c), F/R stimulations were immediately followed by 3 min of 25 M NMDA. Slices were replaced in regular slice recording buffer after stimulation and then subjected to surface biotinylation without microdissection at indicated time points.
Statistical Analysis-All statistics were performed by Student's twotailed, unpaired t-test compared with controls or basal (0-min time point) within each group. In all the figures, *, p Ͻ0.05; **, p Ͻ0.01; ***, p Ͻ0.001 All data are presented as mean Ϯ S.E.

Ser-845 Phosphorylation Increases Surface GluR1 in Cultured
Hippocampal Neurons-To quantify net Ser-845 phosphorylation in hippocampal neurons, we generated standards of GluR1 with known phosphorylation stoichiometries. This was accomplished by in vitro phosphorylation of purified GST fusions of the C-terminal 75 residues of GluR1 (GST-GluR1-CT) with purified protein kinases (11). Using [␥-32 P]ATP, CaM-KII (10 nM) catalyzed phosphorylation of Ser-831 to a molar stoichiometry of ϳ1 (Fig. 1a). Surprisingly, PKA (10 -200 nM) gave a stoichiometry of Ͻ0.1 for Ser-845 (Fig. 1b). Examination of the sequence around both Ser-831 and Ser-845 revealed that neither is a CPS for CaM-KII or PKA, respectively, since neither contains an arginine in the p-3 position (Fig. 1e). Therefore, the CPS GST-GluR1-CT fusion mutants P828R and P842R were expressed, purified, and examined for phosphorylation. CaM-KII phosphorylated the P828R mutant the same as wild-type ( Fig. 1c), but now PKA catalyzed stoichiometric phosphorylation of the CPS P842R mutant (Fig. 1d). The results obtained by 32 P incorporation were verified using phospho-specific antibodies for Ser-831 and Ser-845 (Fig. 1f). The CPS mutant P842R was stoichiometrically phosphorylated by PKA on Ser-845 and mixed with known amounts of non-phosphorylated GST-GluR1-CT to make calibrations standards. These in vitro phospho-Ser-845 calibration standards were run on every SDS-PAGE ( Fig. 2a) with extracts of cultured neurons or hippocampal slices and used for Western blotting with phospho-specific antibody to quantify the net Ser-845 phosphorylation levels of endogenous GluR1 in cultured hippocampal neurons.
To induce Ser-845 phosphorylation (12), we used a recently described chemical stimulation protocol of forskolin plus rolipram (F/R) that results in prolonged NMDA receptor (NMDAR)-dependent LTP (cLTP) (20) and recruits CaM-KII to dendritic spines (21). Treatment of hippocampal cultures with F/R (50 M/0.1 M) for 10 min persistently increased basal Ser-845 phosphorylation (10 -20%) of surface GluR1 to ϳ60% at 30 -40 min (Fig. 2b), which correlated with a significant increase in surface expression of GluR1 (Fig. 2c). Importantly, there was no net increase in Ser-831 phosphorylation during this same time period (Fig. 2e), suggesting that Ser-845 phosphorylation may play a specific role in GluR1 trafficking (13,14). The Ser-845 phosphorylation stoichiometry did not differ between the surface and total GluR1 fractions (data not shown). Neither Ser-845 phosphorylation nor surface GluR1 delivery was blocked by the NMDAR antagonist D-2 amino-5-phosphonopentanoic acid (D-APV) (Fig. 2, b and c) or tetrodotoxin (data not shown). The plot of net phospho-Ser-845 versus normalized surface GluR1 yielded a linear relationship with an R 2 value of 0.997 and slope of 0.75 (Fig. 2d). This tight relationship strongly implicates GluR1 Ser-845 phosphorylation in regulating surface expression of AMPARs.
Dephosphorylation of Ser-845 Is Associated with a Decrease in Surface GluR1-Bath application of NMDA induces chemical LTD (Chem-LTD) and Ser-845 dephosphorylation in hippocampal slices (22,23). We hypothesized that blocking Ser-845 phosphorylation during F/R stimulation may also block the surface delivery of new AMPARs. Bath application of NMDA (50 M) to cultured hippocampal neurons resulted in a significant Ser-845 dephosphorylation of surface GluR1 and also prevented subsequent F/R-mediated Ser-845 phosphorylation (Fig. 3a). Importantly, the decrease in phospho-Ser-845 was associated with a decrease in the surface pool of GluR1 (Fig. 3b). These results suggest that the fraction of GluR1 delivered to the surface was closely correlated with net phospho-Ser-845 (13).
Ser-845-phosphorylated AMPARs Are Trafficked to Extrasynaptic Sites in Hippocampal Slices-The functional role of GluR1 Ser-845 phosphorylation was further investigated in acute hippocampal slices by testing whether F/R increases phospho-Ser-845, surface GluR1, and synaptic strength. Surface GluR1 was measured by biotinylation using conditions previously shown to penetrate the entire slice (17). Similar percentages of AMPARs (ϳ30 -35%) were found on the surface in both slices and cultured neurons, indicating that biotinylation was equally effective in these preparations (data not shown). A 10-min F/R (25 M/0.1 M) stimulation resulted in a robust Ser-845 phosphorylation from a basal value of 15 to ϳ50% that persisted for at least 1 h after the washout of drugs (Fig. 4a) that was associated with a significant increase in surface GluR1 fraction (Fig. 4c). Bath treatment with 25 M NMDA for 3 min induced Chem-LTD (65 min after NMDA treatment, normalized fEPSP slope: Control ϭ 0.94 Ϯ 0.4, n ϭ 6; NMDA ϭ 0.75 Ϯ 0.5, n ϭ 12; p Ͻ0.05) (19) and completely blocked F/R-mediated Ser-845 phosphorylation (Fig. 4b) and surface delivery of GluR1 (Fig. 4c). Thus, pharmacological stimulations that either increased (F/R) or decreased (NMDA) net phospho-Ser-845 significantly increased or decreased the surface pool of GluR1, respectively, supporting the hypothesis that Ser-845 phosphorylation is involved in regulating the size of surface pool of GluR1 in hippocampal slices. CaM-KII (10 nM) phosphorylated wild-type GluR1 (10 M) to a molar stoichiometry of ϳ1, and this was completely blocked by S831A mutation. b, PKA (10 nM) did not phosphorylate wild-type GluR1, but mutating Ser-845 into consensus phosphorylation site for PKA (P842R) allowed robust phosphorylation (d), whereas a similar mutation at the CaM-KII site (P828R) had no effect (c). Panels a-d, n ϭ 4. e, amino acid sequence around the two phosphorylation sites revealed that neither Ser-831 nor Ser-845 sites are consensus phosphorylation sequences for CaM-KII or PKA, respectively, with a proline rather than an arginine at the p-3 position. f, Western blot analysis using phospho-specific and total GluR1 antibodies of in vitro CaM-KII and PKA phosphorylated GST-GluR1-CTs gave similar results as 32 P incorporation assays (a-d). We evaluated the functional effect of GluR1 Ser-845 phosphorylation on synaptic strength by recording fEPSPs from the CA1 region of hippocampal slices. After 20 min of stable baseline, basal stimulation was stopped and F/R was applied to slices for 10 min, followed by basal stimulation 40 min later to assess changes in synaptic strength. The 40-min pause was chosen to allow surface GluR1 phospho-Ser-845 to increase to almost 50% (Fig. 4, a and b). When basal stimulation was resumed, synaptic strength in F/R treated slices was comparable with the control (Me 2 SO) (Fig. 5a, closed diamonds and open circles, respec-tively). This was surprising because the increase in surface GluR1 phospho-Ser-845 (Fig. 5b) should increase surface AMPARs (Fig. 4c), thereby enhancing synaptic strength. There was also no detectable presynaptic effect, as the pair-pulse facilitation (PPF) ratio remained at 1.5 during baseline and 40 min after F/R treatment (data not shown). These findings revealed that in the absence of basal stimulation, F/R treatment was sufficient to robustly increase surface AMPA receptors (Fig. 4c) but not synaptic strength (Fig. 5a), indicating that AMPA receptors phosphorylated at Ser-845 were probably trafficked to extrasynaptic sites.
Ser-845 Phosphorylation Primes GluR1 for Synaptic Incorporation -What additional signal is required to drive the extrasynaptic receptors into synapses? When F/R was applied with basal synaptic stimulation, the synaptic responses were dramatically increased, resulting in cLTP (20) that persisted for over 1 h (Fig. 5a, open squares). As shown previously, cLTP required concurrent synaptic activity, involving NMDARs (20). Based on our finding that synaptic activity is not required for surface expression of AMPARs in response to F/R, we hypothesized that Ser-845 phosphorylation primed AMPARs for synaptic potentiation by trafficking them to extrasynaptic sites, possibly followed by synaptic incorporation requiring synaptic activity.
Can extrasynaptic AMPARs regulate synaptic strength during activity-evoked plasticity? To test this hypothesis, we primed hippocampal slices with F/R treatment in the absence of basal stimulation to increase extrasynaptic AMPARs, followed by theta burst stimulation to induce electrical LTP. Indeed, three theta trains resulted in significantly larger LTP in slices pretreated with F/R (1.63-fold over baseline, p Ͻ0.05) compared with control slices (1.34-fold over baseline) (Fig. 5c). In this and subsequent experiments, pair-pulse facilitation ratios were not significantly different between F/R Ϯ NMDA and control groups (data not shown), suggesting a predominantly postsynaptic locus of expression under these conditions. Furthermore, CA1 regions microdissected after recordings had significantly elevated phospho-Ser-845 in slices pretreated with F/R compared with controls (Fig. 5d). These results support the hypothesis that GluR1 Ser-845 phosphorylation increases the number of AMPARs available for synaptic incorporation (14) by increasing the extrasynaptic surface pool.
Blocking Ser-845 Phosphorylation Prevents Priming of AMPARs for Synaptic Incorporation-If Ser-845 phosphorylation is important for regulating the surface pool of extrasynaptic AMPARs, then blocking Ser-845 phosphorylation should also prevent priming of AMPARs and the corresponding enhancement of LTP observed in Fig. 5c. We tested this hypothesis by using a Chem-LTD paradigm (3-min bath application of 25 M NMDA) that alone resulted in basal dephosphorylation of Ser-845 (Figs. 1e and 2b) and decreased the surface pool of GluR1 (Figs. 1f and 2c) (28). We anticipated that NMDA treatment may block F/Rmediated Ser-845 phosphorylation and thus the priming of LTP. Indeed, application of NMDA after F/R treatment significantly suppressed Ser-845 phosphorylation from 46% (Fig. 5d) in slices treated with F/R alone to 19% (Fig. 5f) when F/R was followed by NMDA (p Ͻ0.001). This NMDA treatment also completely blocked F/R-induced cLTP in the presence of basal stimulation (data not shown). More importantly, NMDA treatment completely blocked the priming effect of F/R on theta burst LTP (Fig. 5e). These results suggest that GluR1 Ser-845 phosphorylation plays an important role in modulating the dynamic range of LTP amplitude by regulating the amount of extrasynaptic AMPARs available for synaptic incorporation (Fig. 5g).
AMPARs to extrasynaptic sites on the surface membrane. Previous studies (3,14,24) examined Ser-845 phosphorylation in GluR1 upon various stimulation paradigms, but only relative changes were analyzed, making it difficult to interpret the likely physiological significance of these changes. This report is the first investigation to quantify the net phosphorylation state of Ser-845 in endogenous hippocampal GluR1 during chemically induced bidirectional synaptic plasticity. This allowed analyses of functional effects in changes of net phospho-Ser-845 and surface GluR1 delivery during bidirectional synaptic plasticity by combining quantitative immunoblotting, surface biotinylation, and slice field recordings. Under basal conditions ϳ30 -35% of GluR1 was on the surface membrane, and only 15% was phosphorylated at Ser-845. Induction of cLTP, in the absence of electrical stimulation, increased surface expression and Ser-845 phosphorylation without affecting Ser-831 phosphorylation, whereas Chem-LTD decreased surface expression and Ser-845 phosphorylation. Combined with a functional analysis of synaptic responses, we demonstrated that Ser-845 phosphorylationregulated surface GluR1 delivery was to extrasynaptic sites. Results from this study indicate that Ser-845 phosphorylation can modulate the dynamic range of synaptic plasticity by regulating the pool of AMPARs available for synaptic incorporation. Thus, Ser-845 phosphorylation might be one of several mechanisms that neurons may use to regulate  120 min (n ϭ 9 -12). b, after recordings, CA1 regions were microdissected from slices and subjected to quantitative Western blotting for net phospho-Ser-845. F/R stimulation resulted in significant increase in phospho-Ser-845, suggesting there was an increase in surface GluR1 with F/R stimulation regardless of basal stimulation (n ϭ 5-7). c, F/R pretreatment (black bar, solid diamonds) without basal stimulation resulted in larger potentiation in response to triple theta bursts (arrow) (at 120 min after F/R: Control, 1.34 Ϯ 0.04, n ϭ 7; F/R, 1.63 Ϯ 0.10, n ϭ 9; p Ͻ0.05. Step 1) stimulates trafficking of internal GluR1-containing AMPARs to extrasynaptic sites on the surface membrane, which primes AMPARs for synaptic incorporation (3 in middle panel). During strong synaptic activation (Step 2), synaptic NMDARs are activated, resulting in increased intracellular calcium (4 in right panel). Calcium triggers the activation of signaling cascades, likely involving CaM-KII (7,14), which drives GluR1-containing AMPARs to synapses from extrasynaptic sites by lateral diffusion (5 in right panel). Thus, the two-step model for synaptic delivery of AMPARs consists of delivery of GluR1-containing AMPARs to extrasynaptic sites in a phospho-Ser-845dependent manner (Step 1, the priming step), followed by synaptic incorporation of AMPARs, which requires synaptic NMDAR activation and calcium (Step 2). *, p Ͻ0.05; **, p Ͻ0.01; ***, p Ͻ0.001 for all the panels. the amplitude of synaptic potentiation. Although there are likely other mechanisms for GluR1 delivery to the surface membrane during synaptic plasticity, the critical role of Ser-845 phosphorylation is in the enrichment of extrasynaptic AMPARs by shifting the equilibrium between the surface and internal pools of AMPARs. It is interesting that F/R treatment gave equivalent increases in Ser-845 phosphorylation of surface and total GluR1. A possible model to explain this observation is that in the absence of Ser-845 phosphorylation the equilibrium between surface and internal AMPARs strongly favors the internal pool (i.e. endocytosis). Upon Ser-845 phosphorylation, this equilibrium between surface insertion and endocytosis changes (13), such that approximately equal numbers of AMPAR may cycle between the surface and internal pools. Thus, F/R treatment could increase the pool of total Ser-845phosphorylated AMPAR, thereby favoring surface expression by mass action.
Another essential finding of our study was that the surface delivery of AMPARs can be dissociated from synaptic targeting that required concomitant synaptic activity (Fig. 5a), suggesting that synaptic NMDAR activation is required (20). Ser-845 phosphorylation has previously been proposed to be important for regulating the amount of GluR1-containing AMPARs available for synaptic incorporation, whereas CaM-KII activity appears to be critical for synaptic incorporation (14). Our results provide a specific mechanism for this previous finding; Ser-845 phosphorylation-dependent delivery is selective to extrasynaptic sites. The increase in extrasynaptic pool of GluR1 resulted in larger LTP following triple theta burst stimulation (Fig. 5c), suggesting that synaptic NMDAR activity can drive GluR1-containing AMPARs into synapses from the extrasynaptic pools. This priming effect was suppressed by Chem-LTD, which blocked the F/R-mediated increase in phospho-Ser-845 and surface GluR1 (Fig. 3) (25). Therefore, our results support a two-step model for synaptic delivery of AMPARs (Fig. 5g), where GluR1-containing AMPARs are first delivered to the extrasynaptic sites in a Ser-845 phosphorylation-regulated manner. Activated synaptic NMDARs then cause an influx of calcium that likely triggers the signal transduction cascades necessary for anchoring AMPARs in synapses. Thus, Ser-845 phosphorylation primes the GluR1-containing AMPARs for synaptic incorporation.
The two-step model for synaptic delivery of AMPARs (Fig. 5g) is consistent with reports from other laboratories. For example, studies show that extrasynaptic AMPARs in hippocampal dendrites are increased nearly 2-fold during LTP (26) and extrasynaptic AMPARs are significantly depleted in GluR1 knock-out mice (27). Furthermore, recent data show that removal of GluR2-containing AMPARs also occurs first at extrasynaptic sites, immediately followed by a decrease in synaptic AMPARs (28). AMPARs have been shown to rapidly move in and out of synapses via lateral movement along the surface membrane, which may account for the increase in postsynaptic AMPAR responses during LTP (29 -32). Elevating local intracellular calcium causes accumulation of AMPARs on the surface membrane, suggesting that calcium triggers the mechanism for synaptic anchoring of laterally diffusing AMPARs (30,31). Whether calcium first increases the number of anchors at synapses or blocks the lateral diffusion of AMPARs, which then recruit the anchors, remains to be determined. Because AMPARs diffuse freely in extrasynaptic sites (31), it seems more likely that synaptic calcium recruits more anchors to postsynaptic densities, which then sequester laterally diffusing AMPARs. One potential component of this anchor might be postsynaptic density-95, which has been shown to be important for regulating the number of synaptic AMPARs (33)(34)(35). Our data do not exclude that AMPARs can be incorporated into synapses directly from the cytoplasm but do indicate that extrasynaptic AMPARs are important contributors to synaptic plasticity.
An important regulator of AMPAR trafficking is stargazin, which has been shown to be critical for the delivery of AMPARs to the surface membrane (36). Importantly, overexpression of stargazin significantly increases AMPA-mediated currents in CA1 pyramidal neurons but not their synaptic currents (36), mirroring present findings with F/R treatment and leading to similar conclusions on stargazin-mediated delivery of AMPARs selectively to extrasynaptic sites. Alternatively, direct regulation of AMPA receptor function by stargazin (37,38) is also consistent with its potentiating effects. However, our data showed unambiguously, using surface biotinylation, that AMPARs can be trafficked to extrasynaptic sites. Interestingly, postsynaptic density-95 overexpression results in synaptic potentiation of AMPARs (33,39). What might be the Ca 2ϩ -sensitive transducer that promotes anchoring of AMPARs at synapses? CaM-KII has been implicated in inserting AMPARs into synapses (7,14), and F/R stimulation is known to recruit CaM-KII to dendritic spines (21). Thus, calcium influx through synaptic NMDARs can activate CaM-KII (40), which may then drive extrasynaptic AMPARs into synapses. The exact target of CaM-KII important for this delivery is unknown, as it does not appear to be Ser-831 in GluR1 (7). Intriguingly, stargazin has multiple phosphorylation sites, including CaM-KII and PKC sites that appear to be involved in regulating AMPAR trafficking (37). Clearly, additional studies are needed to elucidate the mechanisms of synaptic capture of AMPARs.