AMPA receptors are exocytosed in stimulated spines and adjacent dendrites in a Ras-ERK–dependent manner during long-term potentiation

Abstract

The exocytosis of AMPA receptors is a key step in long-term potentiation (LTP), yet the timing and location of exocytosis and the signaling pathways involved in exocytosis during synaptic plasticity are not fully understood. Here we combine two-photon uncaging with two-photon imaging of a fluorescent label of surface AMPA receptors to monitor individual AMPA receptor exocytosis events near spines undergoing LTP. AMPA receptors that reached the stimulated spine came from a combination of preexisting surface receptors (70–90%) and newly exocytosed receptors (10–30%). We observed exocytosis in both the dendrite and spine under basal conditions. The rate of AMPA receptor exocytosis increased ∼5-fold during LTP induction and decayed to the basal level within ∼1 min, both in the stimulated spine and in the dendrite within ∼3 μm of the stimulated spine. AMPA receptors inserted in the spine were trapped in the spine in an activity-dependent manner. The activity-dependent exocytosis required the Ras-ERK pathway, but not CaMKII. Thus, diffusive Ras-ERK signaling presumably serves as an important means for signaling from synapses to dendritic shafts to recruit AMPA receptors into synapses during LTP.

Long-term potentiation (LTP) at CA1 synapses in the hippocampus, a cellular model for learning and memory, is initiated by the influx of Ca²⁺ through NMDA-type glutamate receptors (NMDAR) into dendritic spines. The insertion of AMPA-type glutamate receptors (AMPAR) into the postsynaptic site and the associated enlargement of dendritic spines are believed to be critical for LTP induction (1–3). The insertion of AMPARs is likely a multistep process including the exocytosis of AMPARs from endosomes to extrasynaptic membranes, lateral diffusion of receptors into the synapse, and anchoring there (1, 4–11). The location of exocytosis during LTP is under some debate, with reports of exocytosis exclusively in the dendrite (9, 10, 12), as well as reports of exocytosis both in the spine and dendrite (13–16). The relative timing of AMPAR exocytosis during LTP is still ambiguous (11).

The signaling cascades linking spine Ca²⁺ elevation and AMPAR trafficking have been extensively studied, and a myriad of signaling molecules have been identified (1, 17–19). However, which signaling pathways underlie specific processes (e.g., exocytosis or anchoring) of AMPAR trafficking remains elusive. Recent studies using two-photon fluorescence lifetime imaging have show that different signaling pathways have different spatiotemporal dynamics during LTP: activity of Ca²⁺/calmodulin-dependent kinase II (CaMKII) is restricted to the stimulated synapse, whereas Ras signaling diffuses out of the stimulated spine and spreads along the dendrite over ∼5 μm (20, 21). Each step of AMPAR trafficking should have a specific spatiotemporal pattern dependent on its upstream signaling.

Here we image individual exocytosis events near spines in organotypic slices with subsecond temporal resolution using high-sensitivity two-photon imaging in combination with single-spine stimulation by two-photon glutamate uncaging. We provide evidence that AMPARs are exocytosed in stimulated spines as well as in dendrites near stimulated spines and are trapped in the stimulated spines. Furthermore, we show that the exocytosis is Ras-ERK dependent.

Results

To investigate the spatiotemporal dynamics of AMPAR trafficking during LTP, we imaged CA1 pyramidal neurons in organotypic hippocampal slices transfected with the GluA1 subunit of AMPAR fused with N-terminal EGFP or superecliptic pHluorin (SEP) (mCherry was used as a cell fill). SEP is a pH-sensitive GFP variant and is quenched in endosomes (22). Thus, SEP-GluA1 signal reports cell-surface receptors selectively, whereas EGFP-GluA1 reports both internal and surface receptors (11, 12, 22–24). To induce LTP and associated spine enlargement in single spines, we used two-photon glutamate uncaging (15–30 stimuli at 0.5 Hz) (25). After uncaging, the spine volume, as measured by mCherry fluorescence, increased transiently by 260% ± 75% (at ∼2 min) before relaxing to a plateau of 120% ± 30% increase (at 30 min) (Fig. 1 A–D) (3). Concomitant with the volume increase, EGFP-GluA1 fluorescence increased to a peak of 200% ± 60% and a sustained phase of 110% ± 12%, showing that the GluA1 was recruited to the spine (Fig. 1B). SEP-GluA1 fluorescence followed a similar time course (Fig. 1C), showing that surface GluA1 also increased in the spine (11, 24, 26). The increase in SEP-GluA1 fluorescence (peak 140% ± 12%, sustained 60% ± 7%) was smaller than the increase in the volume but consistent with that of the surface area as measured by YFP-CD8 (peak 155% ± 28%, sustained 57% ± 13%; Fig. 1D), which roughly scaled with the 2/3 power of the volume (Fig. 1 C and D). Imaging SEP-GluA1 and mCherry with higher temporal resolution (8 s) showed that the spine volume and the surface AMPARs in the spine increase at the same time (Fig. 1E).

Fig. 1.

SEP-GluA1 is recruited to spines after stimulation. (A) Images of a dendritic segment of a neuron transfected with mCherry (Left) and SEP-GluA1 (Right) before (Top), immediately after (Middle), and 30 min after (Bottom) single-spine stimulation. (Scale bar, 1 μm.) (B) Time course of mCherry (red) and GFP-GluA1 (green) fluorescence increase after single-spine stimulation. Stimulated spines (filled circles) increase in size transiently before plateauing. Adjacent spines do not grow (open circles). Fluorescence is normalized to three reference images before uncaging. n = 8 for stimulated spines, 17 adjacent spines, 3 cells. (C) Time course of mCherry and SEP-GluA1 fluorescence after single-spine stimulation. (Red fluorescence)^2/3 shown as blue dotted line. n = 34 for stimulated spines, 16 adjacent spines, 21 cells. (D) Time course of mCherry and membrane-tagged YFP-CD8 fluorescence after single-spine stimulation. (Red fluorescence)^2/3 shown as blue dotted line. n = 13 spines, 4 cells. (E) Time course of fluorescence increase during uncaging, normalized to peak increase. (F) Time course of mCherry fluorescence under control, MEK inhibitor U0126 (open circles, n = 18 spines, 7 cells), and CaMKII inhibitor KN62 (open triangles, n = 17 spines, 7 cells) conditions. (G) Time course of SEP-GluA1 fluorescence under drug conditions.

The CaMKII and Ras-ERK pathways have been reported to be required for the induction of LTP and associated spine enlargements (3, 20, 21, 27, 28), likely acting in parallel pathways (20). To determine whether these signaling pathways play a selective role in AMPAR recruitment, we performed uncaging experiments in the presence of blockers of ERK phosphorylation (U0126) and CaMKII activation (KN62) (Fig. 1 F and G). We found that both of these blockers partially but significantly (P < 0.05; paired t test) blocked structural plasticity as well as long-term AMPAR increases (3, 20, 21).

Because SEP-GluA1 fluorescence intensity follows the surface area increase during LTP (Fig. 1 C and D), the increase in SEP-GluA1 fluorescence may be due to the passive diffusion of preexisting surface AMPARs into spines from the dendritic shaft (11). To determine the contribution of newly exocytosed receptors to the increase in spine GluA1 content, we prebleached ∼85% of surface SEP-GluA1 fluorescence within an ∼8-μm radius of a select spine before inducing LTP, so that any increase in fluorescence comes from newly exocytosed receptors (Fig. 2A). Both before and after bleaching, applying NH₄Cl to deacidify internal stores caused fluorescence increases by ∼15% of the original fluorescence (Fig. S1), suggesting that our two-photon bleaching protocol does not bleach internal stores. mCherry was bleached little by this protocol (<10%) and thus was used to monitor spine structural plasticity (Fig. 2 A and B).

Fig. 2.

Newly exocytosed AMPARs are recruited to stimulated spines. (A) Images of mCherry (Left) and SEP-GluA1 (Right) fluorescence before bleaching (Top), after bleaching (Middle), and after uncaging (Bottom). (B) Fluorescence time course of mCherry (Left) and SEP-GluA1 (Right) for the spine and dendrite shown in A. (C) Population data of mCherry (Left) and SEP-GluA1 (Right) fluorescence during bleaching protocol. SEP normalized to prebleach fluorescence. n = 18 stimulated spines, 9 adjacent spines, and 16 neurons. (D) mCh (Left) and SEP fluorescence (Right) in the presence of TeTX (black crosses), KN62 (open blue squares), and U0126 (filled triangles). n = 10 spines, 6 cells for TeTX; 10 spines, 6 cells for U0126; and 15 spines, 5 cells for KN62. (E) mCh (Left) and SEP (Right) fluorescence after lamp bleaching and uncaging. Only the stimulated spine (filled square) recovered SEP fluorescence, whereas the dendrite (filled triangles) and adjacent spines (dotted line, cross) did not. TeTX blocked SEP-GluA1 fluorescence recovery in stimulated spines (open circles). n = 9 spines, 6 cells for lamp bleaching; 6 spines, 3 cells for TeTX under lamp bleach.

After uncaging at a single spine, the volume of the stimulated spine increased similarly to nonbleached spines (410% ± 60%; Figs. 2 A–C). At the same time, SEP-GluA1 fluorescence in the stimulated spine increased from 12% ± 2% to 60% ± 11% of the original fluorescence (Fig. 2 A–C). Without stimulation, the SEP-GluA1 fluorescence recovery in the spine was slow and small (from 16% ± 2% to 26% ± 3%; Fig. S2). The dendritic area immediately below the spine, or neighboring spines during stimulation, had similar recoveries (from 15% ± 2% to 24% ± 3% for dendrite; from 16% ± 2% to 22% ± 2% for adjacent spines; Fig. 2C). The contribution from stimulation-dependent exocytosis is the SEP fluorescence increase in the stimulated spine, ∼50%, minus the background recovery, ∼10%, or roughly 40%. These results suggest that AMPARs are exocytosed near stimulated spines and retained in the spine within a few minutes.

To confirm that the increase in spine SEP-GluA1 fluorescence was due to the exocytosis of AMPAR, we transfected neurons with tetanus toxin light chain (TeTX), which cleaves vesicle-associated membrane protein (VAMP) specifically (29). TeTX expression partially blocked structural plasticity (Fig. 2D, Left) and reduced SEP-GluA1 accumulation to the background level (Fig. 2D, Right). This result indicates that the recovery is mostly due to VAMP-dependent exocytosis. As an alternative way to exclude the possibility of the diffusion of AMPARs from the nonbleached area, we measured SEP fluorescence recovery after bleaching of a larger area of the secondary dendrite (>50-μm radius) using an arc lamp. Under this condition, nonstimulated spines and dendritic shafts did not show any recovery, indicating that basal exocytosis is not sufficient to drive recovery (Fig. 2E). In contrast, stimulated spines recovered ∼15% (Fig. 2E), which is smaller than but qualitatively consistent with the two-photon bleaching results (Fig. 2C). This spine recovery was inhibited by TeTX, confirming that the recovery was due to the exocytosis of AMPA receptors (Fig. 2E).

Dividing the SEP recovery (15–40%) by the increase of SEP fluorescence without bleaching (15–40% vs. 140%; Fig. 1C) suggests that approximately 10–30% of the total GluA1 increase during LTP induction (Fig. 1) is from newly exocytosed SEP-GluA1, and the rest (70–90%) is presumably due to the diffusion of receptors from the parent dendrites (11). For both two-photon bleaching (Fig. 2C) and lamp bleaching (Fig. 2E) the recovery in the stimulated spine was saturated within 1 min of the stimulation, suggesting that receptor exocytosis occurs within 1 min.

We also analyzed the signaling pathways underlying the fluorescence recovery during LTP. A CaMKII inhibitor (KN62) and ERK inhibitor (U0126) partially blocked the mCherry fluorescence increase, as well as the SEP-GluA1 fluorescence increase (Fig. 2D), suggesting that both CaMKII and ERK signaling are required for insertion of exocytosed AMPARs into the stimulated spine (P < 0.05; paired t test).

Although the previous experiments allowed us to make inferences on the importance of exocytosis, they did not yield information about the actual exocytosis events themselves. To determine the location and timing of individual AMPAR exocytosis events, we imaged while continuously photobleaching all surface receptors on a ∼10-μm stretch of dendrite to prevent fluorescence recovery (bleaching τ = 8.3 ± 0.7 s for dendrite, 7.3 ± 0.85 s for spine; Fig. S3). Under this condition, we observed fast fluorescence increases in spines and dendrites, reporting single exocytosis events (10, 12) (Fig. 3 and Movies S1, S2, and S3). We observed exocytosis events with a distribution of sizes with a subset of large quanta events (Fig. S4). The events with large quanta size occur primarily in dendrites, whereas the events with small quanta size occur both in spines and dendrites (Fig. S4).

Fig. 3.

Kinetics of exocytosis events. (A) mCherry (Upper) and SEP-GluA1 images (Lower) of spines undergoing stimulation. Filtered spatially (0.75 μm) and temporally (1.25 s). Exocytosis is measured as a sharp increase in spine fluorescence. Stimulated spine shown by open arrowhead. Exocytosis shown by closed arrowhead. (Left) Example of transient spine exocytosis. (Right) Example of sustained spine exocytosis. Numbers indicate time after starting uncaging(s). (Scale bar 1 μm.) (B) mCherry and SEP-GluA1 fluorescence during dendritic exocytosis. (Left) Example in dendrite immediately beneath spine, showing movement of fluorescence into the stimulated spine (40–41 s). (Right) Example that is 2 μm away. (C) Fluorescence time course for region of interest (ROI) shown as yellow circle in A (filtered with 1.25-s window). (D) Fluorescence time course for ROI shown in B. (E) Average of all unstimulated spine exocytosis events (thick green line) and four example exocytosis events (thin lines). n = 25 events. Average was taken of unfiltered data, whereas individual traces have been filtered (1.25 s). (F) Average of all stimulated exocytosis events (thick green line), trace examples in A (black lines), and four other example exocytosis time courses. n = 46 events. (G) Average fluorescence time course for unstimulated dendritic exocytosis (thick green) and four individual example time courses. (H) Average time course of stimulated dendritic exocytosis events (thick green line), examples from B (black lines), and three other example time courses. n = 134 events.

When not stimulated, the fluorescence increase in spines quickly returns to baseline within a few seconds (Fig. 3E). Because the decay is faster than bleaching (7 to 8 s; Fig. S3) or diffusion of AMPARs out of spine (60 s; Fig. S5), and fluorescence did not increase in the adjacent dendritic shaft after AMPAR is exocytosed in the spine (Fig. S6), the decay is presumably due to reinternalization and reacidification of exocytosed AMPARs (15). During stimulation, the fluorescence is sustained more than ∼10 s in the stimulated spine (Fig. 3F), suggesting that during spine growth AMPARs exocytosed in the stimulated spine are trapped there. This sustained exocytosis no longer occurs after 1 min of the cessation of stimulation (Fig. S7). We further segregated stimulated spine exocytosis into transient and sustained types, using a threshold of 30% increase at 4 s after exocytosis, and found that exocytosis events with sustained GluA1 fluorescence are associated with increases in spine volume (Fig. S8 A and B). The sustained exocytosis was also observed in the presence of CaMKII inhibitor (KN62) or ERK inhibitor (U0126) during stimulation, suggesting that the trapping of AMPARs in the spine depends on neither CaMKII nor ERK signaling (Fig. S8C). In contrast to the spine exocytosis, the decay time of dendritic exocytosis was independent of stimulation (∼10 s) (Fig. 3 G and H). We occasionally observed that AMPARs exocytosed into dendrites move into the stimulated spine (Fig. 3B, Left, 40 to 41 s; Movie S3).

The exocytosis rate before stimulation was 0.11 ± 0.05 events/min in the spine (Fig. 4A) and 0.034 ± 0.01 events/min per μm in the dendrite (Fig. 4B). The rate increased by ∼5-fold within 15 s of the first stimulus and remained elevated for the duration of the stimulus protocol to ∼0.61 ± 0.1 events/min in the spine (Fig. 4A) and ∼0.18 ± 0.04 events/min/μm within 2.5 μm of the spine in the dendrite (Fig. 4B). After the stimulus ended, the exocytosis rate in both the spine and dendrite quickly returned to baseline within ∼1 min (Fig. 4 A and B), consistent with the fluorescence recovery due to exocytosis after photobleaching dendrite, which saturated within ∼1 min (Fig. 2). The increase in exocytosis rate in dendrites during stimulation was highest in the area beneath the stimulated spine and decayed with distance from the stimulated spine with a length constant of ∼3 μm (Fig. 4C). Cotransfection of SEP-GluA1 with TeTX abolished the activity-dependent increase of AMPAR exocytosis, suggesting that these observed events are VAMP-dependent exocytosis (Fig. 4D and Fig. S9 A and B).

Fig. 4.

Location and timing of activity-dependent exocytosis. (A) Time course of exocytosis in spine. Fifty-second movies were taken and split into 25-s epochs. (B) Time course of dendritic exocytosis within 2.5 μm of stimulated spine. (C) Distance from stimulated spine exocytosis occurred before (open circles), during (filled triangles), and after (gray squares, filled diamonds) stimulation. (D) (Left) Exocytosis rate in spine during stimulation (0–50 s). Same color scheme as in A. *Significant differences from control rate (ANOVA, P < 0.05). (Right) Exocytosis rate in the dendrite, near (0–2.5 μm) and away from (3.5–6.5 μm) the stimulated spine during stimulation (0–50 s). n = 112 spines/dendrites from 31 cells for control; 39 spines, 7 cells for U0126; 41 spines, 9 cells for KN62; 25 spines, 6 cells for tetanus toxin (TeTX); and 49 spines, 12 cells for dnRas.

To identify the signaling pathways involved in exocytosis, we recorded exocytosis events in the presence of CaMKII and ERK inhibitors KN62 and U0126, as well as in cells expressing dominant negative Ras mutant (S17N) (dnRas; Fig. 4). KN62 did not alter the increase in the exocytosis rate during LTP in spines and dendrites (Fig. 4 A and B). In contrast, both U0126 and dnRas partially blocked the increase in the exocytosis rate in spines and dendrites (Fig. 4 A, B, and D). The partial effect of U0126 was not due to partial block of ERK activity: we performed Western blots to confirm the block of ERK phosphorylation (Fig. S10). Because dnRas has a stronger effect than U0126, Ras may regulate exocytosis through other pathways, such as phosphoinositide 3 kinase (30). Transfection with a constitutively active Ras mutant (G12V) did not increase the dendritic exocytosis rate (Fig. S9C), indicating that increasing basal Ras activity is not sufficient for increasing AMPAR exocytosis. These results suggest that the Ras-ERK pathway, but not the CaMKII pathway, is required for the increase in exocytosis rate during uncaging.

Discussion

In this study, we used two-photon imaging and glutamate uncaging to simultaneously monitor structural plasticity and trafficking of AMPARs in single dendritic spines. We found that AMPARs are recruited to spines rapidly after LTP induction, at the same time as the volume increases (within ∼10 s), in contrast to reports that AMPAR recruitment is delayed (11, 24) (Fig. 1). By prebleaching surface AMPARs, we found that newly exocytosed AMPARs are rapidly trapped in the stimulated spine (Fig. 2). The increase after bleaching was a fraction of the increase without bleaching (140% in Fig. 1C vs. 15–40% in Fig. 2 C and E), suggesting that AMPAR spine recruitment is mainly supplied by the diffusion of preexisting surface receptors (70–90%) from the dendritic shaft, as well as exocytosis near spines (10–30%) (6–8, 11, 12, 23). There is no single explanation for the difference in spine recovery between the two-photon and lamp bleaching protocols. It may be a combination of diffusion of unbleached receptors (10% dendritic recovery vs. none) or photodamage in the lamp bleaching protocol from bleaching a wide area (∼50-μm radius). It should be noted that the relative contribution of surface diffusion may be overestimated, because overexpression of GluA1 may increase extrasynaptic AMPAR and the apparent diffusive fraction.

We also visualized individual exocytosis events while inducing LTP and structural plasticity in single synapses (Figs. 3 and 4). Without stimulation, the exocytosis rate in the spine and dendrite were both low (<0.1 exocytosis/min in the spine and ∼0.03 exocytosis/min per μm in the dendrite) (Fig. 4 and Fig. S9) (10, 12). Upon stimulation the exocytosis rate in both the spine and dendrite increased ∼5-fold and then returned to the baseline rate after cessation of stimulation. Exocytosis was restricted around the stimulated spine within ∼3 μm and was dependent on Ras-ERK signaling.

Previous studies using SEP-GluA1 imaging have reported GluA1 exocytosis exclusively in dendrites (10–12), whereas experiments using similar techniques have reported spine exocytosis (13–16). We found that exocytosis events could have small and large amplitude and that large-amplitude exocytosis events occurred preferentially in dendrites, whereas small-amplitude events occurred in both spine and dendrite (Fig. S4 and Movies S1, S2, and S3). These small events are indeed exocytosis and not due to the clustering of AMPA receptors: first, these events are inhibited by TeTX, a specific exocytosis inhibitor (12); second, we observed rapid fluorescence increases in the whole stimulated spine with the rise time <250 ms (Fig. 3 and Fig. S4), which is much faster than diffusion of AMPAR from dendrite into spines (∼60 s; Fig. S5). The apparent disagreement with previous studies reporting nonexistence of spine events is likely due to the difference in the detection threshold, but there may be other reasons, such as culture condition or stimulation paradigm.

Without stimulation, AMPAR exocytosis fluorescence has a short half-time, perhaps due to reinternalization (Fig. 3E and Fig. S7). During stimulation, the fluorescence increase was more persistent (Fig. 3F and Fig. S7). Others have similarly reported two types of exocytosis, transient and persistent, in both dendrites (10, 12) and spines (16). The persistence of spine exocytosis during stimulation may be due to the trafficking of AMPAR into synapses and trapping there (7). Although our fluorescence recovery after bleach analysis (Fig. S5) did not reveal a clear difference in spine–dendrite diffusion coupling between before and after LTP induction, previously activity-dependent regulation of AMPAR diffusion was observed using various techniques (6, 11, 15, 31). This persistent fluorescence depended neither on CaMKII nor Ras-ERK signaling (Fig. S8C). Exocytosis events with a sustained fluorescence change were associated with a simultaneous increase in spine volume (Fig. S8 A and B), supporting the hypothesis that the fusion of endosomes with the plasma membrane provides membrane to enable rapid spine growth (13, 16).

The length constant of dendritic exocytosis was ∼3 μm (Fig. 4C), which is similar to that of Ras activity during LTP induction in single spines (11, 20). Application of a Ras-ERK pathway inhibitor, U0126, as well as genetic blockade by dnRas, impaired the increase of exocytosis rate, implying that the Ras-ERK pathway is necessary for exocytosis. This is consistent with previous studies showing that the Ras-ERK pathway is required for LTP (28) and associated spine growth (20). Although it has been suggested that Ras-ERK signaling invades adjacent spines (20), we did not observe an increase in the exocytosis rate in adjacent spines. Thus, Ras-ERK signaling is required but not sufficient for increasing exocytosis in spines (20).

Synaptic cross-talk experiments indicate that the initiation of plasticity at a spine can lower the threshold for plasticity in nearby spines in an ERK-dependent manner (20). Our exocytosis results provide one possible mechanism for this cross-talk: AMPARs are exocytosed in the dendritic area around a stimulated spine and are available to adjacent spines. Thus, the spreading of Ras-ERK activity is likely important for signaling to recruit AMPARs to the stimulated spine during LTP, as well as for signaling on the micrometer-length scale, such as the facilitation of LTP (20).

Our data also indicate that CaMKII is not required for AMPAR exocytosis (Fig. 4). This is consistent with the previous observation showing that the activity of CaMKII is restricted to the stimulated spines (21), unlike exocytosis events (Fig. 4). Notably, CaMKII was not required for trapping AMPARs on the seconds time scale (Fig. S8C). Because the CaMKII inhibitor inhibited the accumulation of newly exocytosed AMPA receptors in the stimulated spine in the minute time scale (Fig. 3), CaMKII may play roles in stabilizing AMPA receptor in synapses by, for example, recruiting scaffolding protein to synapse (32, 33).

Materials and Methods

Preparations.

SEP-GluA1 consists of SEP (22) tagged to the extracellular N terminus of GluA1 after the predicted signal peptide cleavage site, where it does not interfere with trafficking or signaling (2). SEP-GluA1 was expressed throughout the cell body and dendrites of neurons (34). Overexpression of GluA1 did not effect structural plasticity (Fig. 1) (27, 35). The SEP-GluA1 plasmid was generously provided by Scott Soderling (Duke University, Durham, NC), the mCherry-IRES-TeTX plasmid by Matt Kennedy and Michael Ehlers (Duke University, Durham, NC), and the dnRas, H-Ras with N17S mutation, by Linda van Aelst (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY).

Hippocampal slice cultures were prepared from postnatal day 6 or 7 rats, in accordance with the animal care and use guidelines of the Duke University Medical Center (36). After 7–12 d in culture, cells were biolistically transfected with 1-μm gold beads at a 1:1 molar ratio of SEP-GluA1:mCherry (for GFP-GluA1, the ratio was 1:4 GFP-GluA1:mCherry; for YFP-CD8, the ratio was 1:1). Experiments were performed 3–4 d later to allow for full, bright expression of both SEP-GluA1 and mCherry.

All experiments were performed at room temperature (∼25 °C) in standard artificial cerebral spinal fluid (ACSF) containing 4 mM CaCl₂, 0 mM MgCl₂, 1 μM TTX, and 2.5 mM 4-Methoxy-7-nitroindolinyl (MNI)-caged-L-glutamate aerated with 95% O₂ and 5% CO₂. For pharmacology and genetic perturbations, all experiments were performed in pairwise fashion. For plasticity experiments (Figs. 1 and 2), we uncaged on two spines from an identified neuron, then incubated with drugs for ∼30 min, then uncaged two to three more times on the same neuron. For exocytosis experiments, we preincubated all slices with drugs for 1 h before imaging, while performing control experiments on other slices prepared on the same day. For experiments using TeTX and dnRas, we recorded exocytosis movies from slices prepared on the same day as controls.

Imaging and Glutamate Uncaging.

We used a custom-built two-photon microscope with two Ti:sapphire lasers (Spectra-Physics). One laser was tuned to 920 nm to excite both SEP-GluA1 for AMPAR trafficking and mCherry for morphology. The second laser was tuned to 720 nm for glutamate uncaging. Each lasers’ intensity was controlled independently using electrooptical modulators (Pockels cells; Conoptics). The beams were combined using a beam-splitting cube and passed through the same set of scan mirrors and objective (60×, 0.9 NA; Olympus). SEP and mCherry fluorescence were separated using a dichroic mirror (565 nm) and band-pass filters (510/70, 620/60; Chroma). Fluorescence signals from cooled GaAsP sPMTs (H7422-40P, Hamamatsu) were acquired by ScanImage using a data acquisition board (PCI-6110, National Instruments) (37).

Two-photon glutamate uncaging was performed in ACSF lacking Mg²⁺ in the presence of MNI-caged-L-glutamate (2.5 mM) and TTX (1 μM). For uncaging pulses, 6–8 mW laser power (720 nm) was delivered to the back focal aperture of the objective for 6 ms. The uncaging beam was parked at a manually selected location ∼0.5 μm from the tip of the spine head, away from the parent dendrite. Only spines well separated from the parent dendrite and nearby spines were selected for experiments. Typically, 15–30 pulses were applied, saturating structural plasticity.

All images are 128 × 128 pixels (10 μm × 10 μm). For long-term imaging, images were taken as a stack of five slices with 1-μm separation, averaging six frames each for each slice (Fig. 1). For imaging while uncaging (Figs. 1E and 2), images were acquired in a single plane every 8 s, averaging six frames. For the exocytosis imaging (Figs. 3 and 4), images were acquired in a single plane at 4 Hz for 50 s, yielding 200 frames.

Data Analysis.

To identify exocytosis events from movies of SEP-GluA1 fluorescence, we filtered movies using a Gaussian spatial filter of three pixels (0.75 μm) and a temporal filter of five frames (1.25s). Background was corrected by simple subtraction of surrounding fluorescence. Spines and dendrites were typically well bleached, and exocytosis events were identified in filtered time courses as increases above the noise level (Fig. 3C). In the dendrite, exocytosis events were semiautomatically identified by: filter movies; drawing a kymograph along the dendrite; identifying points with rapid increases (<1 s) in fluorescence (threshold ∼30%); then playing movies to verify that they were not artifacts due to endosomes moving along the dendrite (when all surface fluorescence is bleached, the small fluorescence from receptors in endosomes is higher than the background, and moving endosomes can appear as rapid fluorescence increase). The identified exocytosis events in spines and dendrites were further verified for a rapid (<0.5 s) fluorescence increase lasting more than ∼1 s by looking at the unfiltered fluorescence time course by eyes.