Intracellular membrane trafficking of glutamate receptors at excitatory synapses is critical for synaptic function. However, little is known about the specialized trafficking events occurring at the postsynaptic membrane. We have found that two components of the exocyst complex, Sec8 and Exo70, separately control synaptic targeting and insertion of AMPA‐type glutamate receptors. Sec8 controls the directional movement of receptors towards synapses through PDZ‐dependent interactions. In contrast, Exo70 mediates receptor insertion at the postsynaptic membrane, but it does not participate in receptor targeting. Thus, interference with Exo70 function accumulates AMPA receptors inside the spine, forming a complex physically associated, but not yet fused with the postsynaptic membrane. Electron microscopic analysis of these complexes indicates that Exo70 mediates AMPA receptor insertion directly within the postsynaptic density, rather than at extrasynaptic membranes. Therefore, we propose a molecular and anatomical model that dissects AMPA receptor sorting and synaptic delivery within the spine, and uncovers new functions of the exocyst at the postsynaptic membrane.
The exocyst is a multiprotein complex originally described in yeast to be required for exocytosis (TerBush et al, 1996). It consists of eight different subunits (Sec3, Sec5, Sec6, Sec8, Sec10, Sec15, Exo70 and Exo84), whose mammalian counterparts have already been identified (Ting et al, 1995; Hsu et al, 1996; Kee et al, 1997). The exocyst is localized at the plasma membrane and intracellular compartments (Shin et al, 2000; Yeaman et al, 2001), and it associates with cytoskeletal elements involved in vesicular trafficking (Hsu et al, 1998; Shin et al, 2000; Vega and Hsu, 2001; Wang et al, 2004). The exocyst is broadly expressed in adult brain and is enriched in synaptic membranes (Hsu et al, 1996; Sans et al, 2003; Vik‐Mo et al, 2003). However, its role in neurons has always been associated with membrane trafficking events required for neuronal development, such as neurite outgrowth (Vega and Hsu, 2001), synaptogenesis (Hazuka et al, 1999) and dendritic trafficking of newly synthesized NMDA receptors (NMDARs) (Sans et al, 2003). In contrast, the functional role of the exocyst in established synapses remains unclear.
Synaptic function is critically dependent on the targeting and delivery of a large number of different proteins at the postsynaptic membrane, including neurotransmitter receptors, ion channels, anchoring and cell adhesion molecules, and a variety of signal transduction modulators (Garner et al, 2000; McGee and Bredt, 2003; Yamagata et al, 2003; Kim and Sheng, 2004). AMPA‐type glutamate receptors (AMPARs) mediate most excitatory synaptic transmission in the brain. Nevertheless, they are highly dynamic components of excitatory synapses. AMPARs can be added to or removed from synapses in an activity‐dependent manner, leading to long‐term potentiation (LTP) or depression (LTD) of synaptic transmission (for recent reviews, see Sheng and Lee, 2001; Barry and Ziff, 2002; Malinow and Malenka, 2002; Bredt and Nicoll, 2003; Esteban, 2003). In addition to this regulated trafficking, AMPARs continuously cycle in and out of synapses in an activity‐independent manner, which requires NSF (Nishimune et al, 1998; Song et al, 1998; Luscher et al, 1999), Hsp90 (Gerges et al, 2004b) and NEEP21 (Steiner et al, 2005) function. Multiple proteins have been described to interact with AMPAR subunits and affect their synaptic localization (for recent reviews, see Scannevin and Huganir, 2000; Henley, 2003; Kim and Sheng, 2004); however, we still know very little about the membrane trafficking mechanism mediating AMPAR targeting and delivery into synapses.
Here we have explored a potential role for the exocyst in the dynamic behavior of AMPARs at the postsynaptic membrane. Using a combination of molecular biology, electrophysiology, fluorescence imaging and electron microscopy, we have found that different subunits of the exocyst mediate AMPAR synaptic trafficking at several distinct levels. Intracellular receptor targeting within the spine is mediated by the Sec8 subunit of the exocyst, possibly through PDZ‐dependent interactions with synaptic scaffolding molecules. Independently from vesicle targeting, receptor insertion into the postsynaptic membrane requires the Exo70 subunit. By separately manipulating vesicle targeting and receptor insertion, we now show that AMPAR synaptic delivery occurs from intracellular compartments within the spine directly into the postsynaptic density. Therefore, in addition to characterizing the exocytic machinery that operates at the postsynaptic membrane, these results delineate the subcellular pathways by which AMPARs reach their synaptic targets.
Different components of the exocyst play distinct roles in synaptic transmission
The C‐terminus of the exocyst subunit Sec8 contains a PDZ motif that interacts with the synaptic scaffolding proteins PSD95 and SAP102 (Riefler et al, 2003; Sans et al, 2003) and is required for NMDAR trafficking (Sans et al, 2003). Therefore, as a first step to evaluate the role of the exocyst in synaptic transmission, we expressed the C‐terminus (last 16 amino acids) of Sec8 as a GFP‐fusion protein in CA1 hippocampal neurons from organotypic slice cultures using the Sindbis virus system (see Supplementary data). This construct may act as a dominant negative by competing with the endogenous exocyst for its PDZ synaptic partners. The effect of Sec8 C‐terminus (Sec8‐Ct) on synaptic transmission was evaluated by simultaneous double whole‐cell recordings from neurons expressing the recombinant protein and nearby uninfected neurons (Methods). As shown in Figure 1A, Sec8‐Ct significantly reduced AMPAR and NMDAR responses in hippocampal CA1 neurons, while GABAA receptor‐mediated transmission remained unaffected. The depression of NMDA responses and lack of effect on GABAA responses are consistent with previously published observations using another Sec8 dominant negative construct (Sans et al, 2003). As control, overexpression of the full‐length Sec8 protein (Sec8‐wt) fused to GFP did not affect AMPA or NMDA transmission (Figure 1B). Importantly, expression of a truncated form of Sec8‐Ct lacking its last four amino acids (Sec8‐Ct‐Δ4), which are required for its PDZ interactions (Sans et al, 2003), did not alter AMPA or NMDA responses either (Figure 1C). These results suggest that the dominant negative effect of Sec8‐Ct is indeed due to competition for PDZ‐binding domains, which would be required for the synaptic delivery or stability of both AMPARs and NMDARs.
Although both AMPARs (Malinow and Malenka, 2002; Bredt and Nicoll, 2003) and NMDARs (Nong et al, 2004; Perez‐Otano and Ehlers, 2005; Prybylowski et al, 2005) undergo regulated synaptic trafficking, NMDARs are considered to be fairly stable components of the synapse, as compared with AMPARs (Sheng, 2001). Therefore, interference with a potential role of the exocyst specifically in receptor delivery is more likely to affect AMPARs to a larger extent than NMDARs. A truncated form of the exocyst subunit Exo70, lacking its C‐terminal half (Exo70‐N), has been recently shown to block the exocytic delivery of the glucose transporter Glut4 into the adipocyte cell surface (Inoue et al, 2003). Importantly, this truncated Exo70 is still able to interact with endogenous exocyst subunits (see Supplementary Figure 1 and Inoue et al, 2003). We therefore evaluated the effect on synaptic transmission of Exo70‐N as a GFP fusion protein in CA1 hippocampal neurons. As shown in Figure 1D, AMPAR‐mediated responses were significantly reduced in Exo70‐N‐expressing neurons. Importantly, NMDAR responses were not affected, in contrast to the result obtained with Sec8‐Ct. This result is also confirmed by the decrease in the ratio of AMPA to NMDA responses in Exo70‐N‐expressing neurons (Figure 1D; to note, AMPA/NMDA ratios are intrinsically very variable across slices; therefore, comparisons between infected and uninfected cells were always drawn from recordings within the same slice, and in most cases, as simultaneous double recordings). As control, overexpression of full‐length Exo70 (Exo70‐wt) fused to GFP did not affect AMPA or NMDA synaptic responses (Figure 1E). The differing effects of Exo70‐N on AMPA and NMDA synaptic transmission suggest that the Exo70 subunit of the exocyst mediates receptor insertion at synapses, which is more likely to affect the continuous synaptic delivery of AMPARs than the stable presence of NMDARs at synapses. (To note, none of these wild‐type and dominant negative constructs affected passive membrane properties of the infected cell, such as input resistance and holding current; see Supplementary Figure 2).
Sec8 PDZ‐dependent interactions are required for the constitutive synaptic cycling of AMPA receptors
The activity‐independent, continuous delivery of AMPARs into synapses was originally inferred from intracellular peptide infusion experiments, in which AMPAR‐mediated synaptic transmission rapidly ran down when a specific peptide competing GluR2–NSF interactions was loaded in the recorded cell (Nishimune et al, 1998; Luscher et al, 1999). We have used the same approach to determine whether the PDZ motif of Sec8 participates in the constitutive cycling of AMPARs. To this end, we recorded AMPAR‐mediated synaptic responses from CA1 hippocampal neurons while infusing them with one of three peptides: Sec8 C‐t (last eight amino acids of Sec8), NR2B C‐t (last eight amino acids of the NMDAR subunit NR2B, which also contains a PDZ ligand motif (Kornau et al, 1995; Niethammer et al, 1996)) and Sec8‐mut C‐t (last eight amino acids of Sec8 with the PDZ ligand sequence –TTV mutated to –ATA) (see Supplementary data and Supplementary Figure 3). These experiments were carried out in a blind fashion, that is, the person carrying out the electrophysiological recordings and the analysis did not know which peptide was used in the experiment. As shown in Figure 2, the peptide corresponding to the Sec8 C‐terminus produced a fast run‐down of synaptic transmission, similar to the one observed for the GluR2–NSF peptide (Nishimune et al, 1998; Luscher et al, 1999; Gerges et al, 2004b). Importantly, the peptide corresponding to the PDZ ligand sequence of NR2B did not affect AMPAR‐mediated responses, suggesting that specific PDZ‐dependent interactions of Sec8 are required for the continuous synaptic cycling of AMPARs. As control, the mutated Sec8 C‐t peptide lacking the PDZ motif did not affect synaptic transmission either. Figure 2B indicates that the differences in run‐down of synaptic transmission obtained with the three peptides are not due to systematic changes in access resistance during the recordings.
The exocyst associates with AMPA receptors in hippocampal synaptosomes
Our data shown above suggest that the exocyst may be directly involved in AMPAR synaptic trafficking. Therefore, we tested whether an exocyst/AMPAR complex is present in the brain. To this end, we carried out immunoprecipitation experiments from hippocampal synaptosomes after Triton solubilization (see Supplementary data; all immunoprecipitations were replicated in at least three independent experiments). As shown in Figure 3A, immunoprecipitation of the exocyst subunits Sec8 or Exo70 co‐precipitated AMPARs but not metabotropic (mGluR) receptors. As previously published (Riefler et al, 2003; Sans et al, 2003), the exocyst also co‐precipitated the PDZ protein PSD95 (Figure 3A).
The exocyst complex has been recently shown to associate with NMDARs through its interaction with synaptic anchoring proteins, such as PSD95 and SAP102 (Sans et al, 2003). However, AMPA and NMDA receptors do not appear to associate in hippocampal extracts (Wyszynski et al, 1999; Sans et al, 2000) (see also Figure 3B). Therefore, we investigated whether the exocyst forms separate complexes with AMPARs and NMDARs at synaptic terminals. To this end, Triton‐solubilized synaptosomal extracts from hippocampus were immunoprecipitated with anti‐GluR2/3 (AMPARs) or anti‐NR1 (NMDARs) antibodies (see Supplementary data). As shown in Figure 3B, the exocyst subunits Exo70, Sec8 and Sec6 co‐precipitated predominantly with AMPARs, but only to low levels with NMDARs. This result supports the notion that the exocyst forms separate complexes with AMPARs and NMDARs.
A potential role for Sec8 but not for Exo70 in the dendritic trafficking of AMPA receptors
It has been recently described that the exocyst mediates the export into dendrites of newly synthesized NMDARs in primary neuronal cultures (Sans et al, 2003). In order to test whether the exocyst plays a similar role in the trafficking of AMPARs in hippocampal slices, we coexpressed a GFP‐tagged AMPAR subunit (GluR2‐GFP), together with Exo70‐N, Sec8 C‐t or NR2B C‐t (as control) tagged with a red fluorescence protein (Exo70‐N‐RFP, Sec8‐Ct‐RFP or NR2B‐Ct‐RFP; see Supplementary data). The efficiency of AMPAR transport along dendrites was quantified from the intensity of GFP fluorescence along the main apical dendrite, after background subtraction (see Supplementary Figure 4A). As shown in Figure 4, the trafficking of GluR2‐GFP into distal dendrites was virtually identical with or without coexpressed Exo70‐N, indicating that the transport of AMPARs along dendrites does not require a fully functional Exo70 subunit. In contrast, the dendritic transport of GluR2‐GFP was significantly impaired when co‐expressing Sec8‐Ct or NR2B‐Ct (Figure 4; see Supplementary Figure 4B for a complete quantification of AMPAR dendritic distribution in the presence of Exo70‐N, Sec8‐Ct and NR2B‐Ct). These results reinforce the interpretation that Exo70 and Sec8 play different roles in AMPAR trafficking: Exo70 probably mediates a local step in AMPAR synaptic delivery, whereas PDZ‐dependent interactions of Sec8 or other proteins are required for efficient receptor transport and targeting (since both Sec8‐Ct and NR2B‐Ct impaired dendritic trafficking, the identity of these PDZ interactions cannot be determined from these experiments).
Morphological dissection of AMPA receptor delivery into the spine surface
In order to directly test a local role of the exocyst in AMPAR trafficking, we monitored the transport of AMPARs from the dendritic shaft into adjacent spines and their delivery into the spine plasma membrane. To this end, we coexpressed GFP‐tagged GluR1 with a constitutively active, untagged form of αCaMKII (truncated αCaMKII, or tCaMKII) to trigger its synaptic delivery (see Supplementary data and Hayashi et al, 2000), with or without Exo70‐N. Protein expression was allowed for 36 h, then slices were fixed and the surface distribution of the recombinant receptor in CA1 hippocampal neurons was monitored by immunostaining using an anti‐GFP antibody under nonpermeabilized conditions (the GFP tag is placed at the extracellular N‐terminus of the receptor). The immunostaining was visualized with an infrared fluorophore (Cy5). This experimental approach allows us to independently monitor the total amount of receptor (GFP channel), the fraction exposed to the cell surface (Cy5 channel) and the presence of the exocyst dominant negative construct (RFP channel). The distribution of surface and total receptor in spines and dendrites was quantified by line plots of fluorescence intensity across spine heads and adjacent dendritic shafts (see Figure 5A for representative examples of the quantification of GFP and surface immunostaining from dendrite–spine pairs with or without coexpressed Exo70‐N). Surface ratios were then calculated as the ratio between the intensity of the Cy5 signal (surface receptor) and GFP signal (total receptor) after background subtraction (Gerges et al, 2004a; Brown et al, 2005). Importantly, this assay is internally normalized for immunostaining variability, because surface labeling is always acquired in pairs of spines and adjacent dendritic shafts. This also means that absolute changes in surface receptors at spines or dendrites cannot be evaluated, but only relative changes between the two compartments. Finally, spines are selected solely from their GFP image, so the analysis is done blind with respect to their surface immunostaining.
As shown in Figure 5B (first panel), surface expression of GluR1‐GFP was slightly higher in spines as compared to their adjacent dendritic shafts when the exocyst function is intact. In contrast, Exo70‐N led to a marked reduction in the amount of receptors that reach the spine surface, to the point that surface expression at spines became significantly lower than at dendrites (Figure 5B, second panel). This effect can also be evaluated from the leftward shift in the cumulative probability distribution of spine‐to‐dendrite ratios in the presence of Exo70‐N (Figure 5B, third panel).
A reduction of surface expression at spines may be due to an impaired translocation of the receptor into the spine. This step can be evaluated from the total receptor distribution (GFP channel) at spines and adjacent dendrites. As shown in Figure 5C (first panel), GluR1‐GFP partitions equally between dendrites and spines in the absence of exocyst dominant negative constructs. In marked contrast, and to our surprise, Exo70‐N led to a significant increase (almost doubling) in the amount of total receptor in spines as compared to dendrites (Figure 5C, second panel) (this effect can also be appreciated from the rightward shift in the cumulative probability distributions in Figure 5C, third panel). Therefore, the decrease in receptor surface expression produced by Exo70‐N is accompanied by an enrichment of receptors at spines.
The next trafficking step after translocation into spines is receptor insertion at the spine plasma membrane. This process can be evaluated independently by calculating ratios of surface to total receptor (Cy5/GFP) at spines and dendrites. This approach normalizes receptor surface expression to the total number of receptors in each compartment, and therefore, it reflects the intrinsic efficiency of receptor delivery at the plasma membrane. As shown in Figure 5D (first panel), surface ratios were slightly higher at spines than at dendrites when GluR1‐GFP was coexpressed only with tCaMKII. On the other hand, blockade of Exo70 function with Exo70‐N strongly impaired receptor insertion at the spine surface, as judged from the drastic reduction in the surface ratio at spines versus dendrites (Figure 5D, second and third panel).
These results support our interpretation that Exo70 plays a late step in AMPAR trafficking, even within the local environment of the synapse. Specifically, Exo70 seems to be required for the fusion of AMPAR‐containing vesicles with the spine plasma membrane.
Interestingly, the accumulation of GluR1‐GFP at intracellular compartments within the spine when Exo70 function is impaired also suggests that this pool of AMPARs reaches the synaptic membrane from intracellular compartments contained within the spine. We then tested whether a similar intra‐spine pathway is followed for the constitutive synaptic delivery of the cycling population of AMPARs. This pathway can be monitored by expressing GluR2‐GFP, which has been shown to constitutively reach spine and synaptic compartments (Shi et al, 2001). Therefore, we expressed GluR2‐GFP in CA1 hippocampal neurons with or without Exo70‐N‐RFP, Sec8‐Ct‐RFP or NR2B‐Ct‐RFP, as control. Total receptor (GFP channel) and surface (Cy5 channel) distribution of GluR2‐GFP was evaluated as described above (see Figure 5A). As shown in Figure 6A–C (first and second panels), and similar to the results obtained with GluR1‐GFP, interference with Exo70 function drastically impaired the surface delivery of GluR2 at the spine, and led to a concomitant enrichment of the receptor in the intracellular space of the spine. Interestingly, Sec8‐Ct also strongly impaired surface receptor delivery (Figure 6A, third panel), but this was not accompanied by receptor accumulation within the spine. In fact, Sec8‐Ct produced a significant decrease in the fraction of receptors present at spines (Figure 6B, third panel). As control, NR2B‐Ct did not affect receptor presence at spines, and only modestly decreased surface delivery at spines (Figure 6, fourth panels; see also Supplementary Figure 5 for a comparative analysis of the effects of Sec8‐Ct and NR2B‐Ct at the spine–dendrite pairs).
Taken together, these results indicate that the exocyst plays specialized roles in AMPAR synaptic trafficking, with Sec8 assisting both receptor translocation into the spine and delivery into the spine surface, while Exo70 mediates exclusively a final membrane fusion step within the spine.
In addition, these results strongly suggest that the exocyst mediates both constitutive and regulated delivery of AMPARs at synapses. As a functional test for the role of the exocyst in these two pathways, we carried out electrophysiological experiments in which the presence of recombinant receptors at synapses is detected from their inward rectification properties (Hayashi et al, 2000; Shi et al, 2001). These electrophysiological experiments reinforced our fluorescence imaging data, and confirmed that the exocyst is required for the functional insertion of all AMPAR populations at synapses (see Supplementary data and Supplementary Figure 6).
Distinct roles of Sec8 and Exo70 in targeting and insertion of AMPA receptors at the postsynaptic membrane
In yeast, different subsets of exocyst subunits have been proposed to associate with exocytic vesicles versus the target plasma membrane (Boyd et al, 2004). However, it is unknown what exocyst subassemblies reside in each compartment in the case of synaptic trafficking. Furthermore, it is still uncertain whether the target membrane for AMPAR exocytic delivery is directly at the synaptic membrane contained within the postsynaptic density (PSD), or whether AMPARs are inserted in extrasynaptic membranes and then reach their synaptic targets by lateral diffusion (Passafaro et al, 2001; Tardin et al, 2003; Adesnik et al, 2005; Triller and Choquet, 2005). Since Exo70 is required for local surface delivery of AMPARs (Figures 5 and 6), we then tested whether interference with this exocyst subunit would alter the association of AMPARs with the PSD. To this end, we isolated PSD fractions (protocol based on Peng et al (2004); see Supplementary data) from hippocampal extracts of control slices or slices overexpressing Exo70‐N. To our surprise, and despite the fact that Exo70‐N impairs receptor delivery into synapses (Figure 1D) and spine plasma membrane (Figures 5 and 6), Exo70‐N overexpression led to a marked accumulation of Sec8, Rab8 and AMPARs at the PSD fraction (Figure 7A and B; to note, none of these proteins is particularly enriched at the PSD fraction under basal conditions). Importantly, endogenous Exo70 and stable components of the PSD, such as PSD95 and αCaMKII, were not affected by expression of Exo70‐N (Figure 7A and B). Another exocytic Rab protein, Rab11, and the small GTPase Rho were undetectable at the PSD fraction under either condition. Therefore, these results together with those monitoring receptor delivery at the spine surface (Figures 5 and 6), strongly suggest that AMPARs are directly inserted at the postsynaptic membrane (or at membranes biochemically associated to the PSD) in an Exo70‐depending manner. Truncation of the Exo70 C‐terminus would lead to formation of an abortive complex at the PSD, in which incoming vesicles containing AMPARs, Sec8 and Rab8 remained associated but not yet fused with the postsynaptic membrane.
Importantly, this interpretation also implies that AMPAR targeting towards the postsynaptic membrane does not require a fully functional Exo70. Since Sec8 is the exocyst subunit that interacts with synaptic scaffolding molecules (Riefler et al, 2003; Sans et al, 2003) and it contributes to the transport of AMPARs into spines (Figure 6), we then tested whether Sec8 may be involved in the targeting aspects of AMPAR delivery into the postsynaptic membrane. To this end, we carried out similar PSD fractionation experiments with hippocampal extracts from uninfected slices, and from slices overexpressing Sec8‐Ct. As shown in Figure 7C and D, Sec8‐Ct led to a drastic reduction in the amount of AMPARs and Sec8 at the PSD fraction. Interestingly, the association of NMDARs and PSD95 with the PSD was also decreased, in agreement with the depression of NMDA responses produced by Sec8‐Ct (Figure 1A). The same result was obtained for the actin‐dependent motor protein myosin Va, which is enriched in dendritic spines (Naisbitt et al, 2000; Walikonis et al, 2000). However, Sec8‐Ct did not produce a generalized disassembly of the postsynaptic complex since Exo70 and αCaMKII were not affected by Sec8‐Ct overexpression. This result with Sec8‐Ct is virtually opposite to the one observed with Exo70‐N (compare quantifications in Figure 7B and D), and reinforces the interpretation that Sec8 mediates the targeting of AMPAR‐containing vesicles towards the synapse, whereas Exo70 is required for the fusion of these vesicles with the postsynaptic membrane.
Ultrastructural analysis of the role of the exocyst in AMPA receptor insertion at the postsynaptic membrane
The results presented above strongly suggest that AMPARs are able to reach the PSD and accumulate within it when Exo70 function is blocked. In order to test anatomically this interpretation and map with high‐spatial resolution the role of Sec8 and Exo70 in AMPAR synaptic insertion, we have carried out immunogold electron microscopy. Organotypic hippocampal slices overexpressing Exo70‐N or Sec8‐Ct, or control (uninfected) slices were processed for postembedding immunogold staining with anti‐GluR2/3 antibodies (see Supplementary data). Electron micrographs were taken randomly from excitatory (asymmetric) synapses in the stratum radiatum (CA1) of these slices (see representative examples in Figure 8A). As shown in Figure 8A and B, AMPAR immunogold particles were significantly enriched (2.6‐fold) within the PSD in Exo70‐N‐expressing slices. In contrast, Sec8‐Ct produced a reduction in the fraction of AMPAR labeling at the PSD, although the difference was not significant with respect to uninfected slices.
This result confirms our biochemical observations and strengthens the interpretation that AMPARs approach the postsynaptic membrane but fail to fuse with it in the absence of a functional Exo70. In order to precisely map the trafficking step mediated by Exo70, we measured the distance between each gold particle and the synaptic cleft from slices expressing Exo70‐N, Sec8‐Ct or uninfected slices. The results are presented as cumulative distributions in Figure 8C (only gold particles within 60 nm of the synaptic cleft—approximately three times the average PSD thickness—were considered for this quantification). As shown in Figure 8C, AMPARs were found to accumulate in close proximity to the synaptic cleft in slices expressing Exo70‐N (red line; leftward shift of the cumulative distribution), even from the shortest measurable distances within the PSD. Sec8‐Ct produced the opposite effect, that is, AMPARs were less abundant in the vicinity of the PSD (blue line; rightward shift in the cumulative distribution). These results indicate that, in the absence of functional Exo70, AMPARs approach the postsynaptic membrane as much as physically possible, within the resolution of the immunogold labeling technique.
Since Exo70‐N seems to trap AMPARs at their final fusion step with the postsynaptic membrane, we then decided to map where this insertion takes place laterally with respect to the postsynaptic density. To this end, we measured the lateral distance to the edge of the PSD for all immungold particles present within the PSD or on the lateral extrasynaptic membrane. As shown in Figure 8D, in the presence of Exo70‐N, AMPARs predominantly accumulate within the PSD, but in close proximity to its edge (almost 40% of the labeling is found within the distal 10% of the PSD) (Sec8‐Ct was not included in this analysis due to the low number of gold particles present at the PSD under those conditions). Once again, this analysis supports the interpretation that AMPARs are inserted into the synaptic membrane directly from the PSD or close to its edge, in a process mediated by the Exo70 subunit of the exocyst.
Delivery of AMPARs into synapses requires a precise and continuous coordination between sorting of AMPAR‐containing vesicles towards the synapse, and fusion of these incoming vesicles with the postsynaptic membrane. Here, we show that the exocyst complex links these two distinct functions through the concerted actions of Sec8 and Exo70. This conclusion is based on four main experimental observations. First, both Sec8 and Exo70 are required for AMPAR function at synapses, as assayed electrophysiologically. Second, both Sec8 and Exo70 dominant negative mutants prevent receptor delivery to the spine surface, but in addition, only blockade of Exo70 function leads to accumulation of AMPARs at intracellular compartments within the spine. Third, exocytic vesicle proteins and AMPARs remain associated with the PSD upon interference with Exo70 function, whereas a Sec8 dominant negative produces the opposite effect. And fourth, AMPARs accumulate within the distal part of the PSD in very close proximity to the synaptic cleft when Exo70 function is blocked. These observations lead us to propose a model for the anatomical and molecular events that lead to the delivery of AMPARs into synapses (Figure 9). First, AMPARs are transported through intracellular membranes, probably belonging to the trans‐Golgi network and recycling endosomes (Horton and Ehlers, 2003, 2004; Gerges et al, 2004a; Park et al, 2004), to reach the spine compartment. This process is constitutive for GluR2/GluR3 receptors, but requires CaMKII activation for GluR1/GluR2 receptors (Piccini and Malinow, 2002). The translocation of AMPARs into the spine is assisted by PDZ‐dependent interactions of Sec8, but is independent from the Exo70 subunit of the exocyst. Exocytic vesicles containing AMPARs bud from intra‐spine compartments and are targeted towards the postsynaptic density, in a process probably directed by interactions between the Sec8 subunit of the exocyst and synaptic scaffolding molecules (Figure 9, left panel). These vesicles dock with the postsynaptic membrane within the PSD, leading to a transient association between AMPARs, vesicular proteins, such as Sec8 and Rab8, and the postsynaptic density (Figure 9, upper right panel). This step is blocked by Sec8‐Ct, presumably by competing with the Sec8‐PDZ‐dependent interactions that direct the vesicles towards the PSD. Finally, this complex is resolved by exocytic membrane fusion mediated by the Exo70 subunit of the exocyst (Figure 9, lower right panel). The C‐terminus of Exo70 is required for this step, since Exo70‐N prevents AMPAR delivery into the synaptic surface.
Our results reinforce the current view that different exocyst subunits perform individual functions during the exocytic process (Boyd et al, 2004), with Sec8 being specialized in vesicle transport and targeting, and Exo70 in membrane fusion. More importantly, this model provides new information about the distinct organization of this exocytic machinery in neurons for the delivery of AMPARs at synapses. The exocyst was originally proposed to act as a spatial landmark for polarized exocytosis in yeast (Finger et al, 1998). We now know that the exocyst mediates surface delivery of a variety of membrane proteins in different cell types. Therefore, it is expected that cell‐ and structure‐specific proteins will recruit the exocyst to localize delivery of different cargo to their specific targets. In the case of the synaptic delivery of AMPA receptors, the PDZ‐dependent interaction between Sec8 and synaptic scaffolding molecules, such as PSD95, seems to be the determining factor for membrane targeting. This interpretation is consistent with the observation that up‐ or downregulation of PSD95 synaptic levels bidirectionally controls AMPAR delivery at synapses in hippocampus (El‐Husseini et al, 2000; El‐Husseini Ael et al, 2002; Colledge et al, 2003; Ehrlich and Malinow, 2004) and cerebral cortex (Beique and Andrade, 2003).
Finally, these results have offered unexpected anatomical information about the subcellular pathways followed by AMPARs to reach their synaptic targets. It has been previously proposed that the constitutive delivery of GluR2 receptors occurs by direct insertion at the postsynaptic membrane, whereas GluR1‐containg receptors would reach the synaptic membrane by lateral translocation, after being inserted at extrasynaptic membranes (Passafaro et al, 2001). This model is compatible with the reported lateral diffusion of AMPARs on the dendritic plasma membrane (Tardin et al, 2003; Groc et al, 2004). However, our results do not support this model for synaptic delivery. We have found that impairment of AMPAR surface delivery with the Exo70 dominant negative leads to the accumulation of both recombinant GluR1 and GluR2 receptors within the spine, and not in the surrounding dendritic space. Using immunogold electron microscopy, we have shown that endogenous AMPARs accumulate within the PSD, in very close proximity to the synaptic membrane, when Exo70 function is blocked. And finally, AMPARs biochemically associate with PSD fractions when surface delivery is impaired. Taken together, these results strongly suggest that AMPARs are targeted directly into the PSD, but become locked in an abortive complex at its intracellular side when membrane fusion is prevented. We do not know the reason for this discrepancy in the spatial pattern of AMPAR delivery, although a potentially important experimental difference may be the use of organotypic slice cultures in our experiments versus dissociated neuronal cultures for the reports on the lateral translocation of AMPARs (see also two recent reports supporting intra‐spine delivery in slices (Kopec et al, 2006) and lateral diffusion in dissociated neurons (Adesnik et al, 2005)). Future experiments will be needed to resolve the physiological role of these different delivery pathways.
In conclusion, our results shed light into the complex array of events that coordinate the transport, targeting and insertion of AMPARs at the postsynaptic membrane, and provide a molecular and anatomical map of the intracellular machinery that catalyzes this process along dendrites and within the dendritic spine.
Materials and methods
Simultaneous whole‐cell recordings were carried out from nearby pairs of infected/transfected and uninfected/untransfected CA1 pyramidal neurons using fluorescence and transmitted light illumination. Synaptic responses were evoked with two bipolar electrodes (stimulation pulses of 200 μs, up to 20 V) placed on the Schaffer collateral fibers between 300 and 500 μm of the recorded cell. The responses obtained from the two stimulating electrodes were averaged for each cell and counted as an ‘n’ of 1. This experimental configuration ensures that the potential effects of the expressed recombinant proteins are postsynaptic, since expression is always in CA1 neurons and stimulation is applied on Schaffer collateral fibers from CA3 neurons. Synaptic responses mediated by AMPA or NMDA were acquired at −60 or +40 mV, respectively, in the presence of 100 μM picrotoxin to block GABAA receptors. GABAA responses were acquired at 0 mV in the absence of receptor antagonists; therefore, they should be considered as a combination of mono‐ and disynaptic IPSCs. All synaptic responses were averaged over 50–100 trials. Voltage‐clamp whole‐cell recordings were acquired with a Multiclamp 700 A amplifier (Axon Instruments). Composition of perfusion and internal solutions are described in the Supplementary data.
Confocal and electron microscopy experiments are described in the Supplementary data.
Biochemical techniques for membrane fractionation and immunoprecipitations from synaptosomal preparations are described in the Supplementary data.
Supplementary data are available at The EMBO Journal Online.
We thank Alan Saltiel for the plasmids containing the Exo70 and Sec8 coding sequences, Shu‐Chan Hsu for the monoclonal antibody against Exo70, Roger Tsien for the red fluorescence protein variant tdimer2(12), Dorothy Sorenson and Chris Edwards for their assistance with the immunogold electron microscopy, members of the Esteban lab and María Soengas for critical reading of the manuscript. This work was supported by grants from the National Institute of Mental Health (grant MH070417) and the National Alliance for Research on Schizophrenia and Depression (NARSAD) to JAE, and grant GM63032 from the National Institute of General Medical Sciences to MRS. We also thank Dr Thomas Coles for sponsoring JAE as an Investigator of the National Alliance for the Mentally Ill (NAMI).
- Copyright © 2006 European Molecular Biology Organization