Driving AMPA Receptors into Synapses by LTP and CaMKII: Requirement for GluR1 and PDZ Domain Interaction

See allHide authors and affiliations

Science  24 Mar 2000:
Vol. 287, Issue 5461, pp. 2262-2267
DOI: 10.1126/science.287.5461.2262


To elucidate mechanisms that control and execute activity-dependent synaptic plasticity, α-amino-3-hydroxy-5-methyl-4-isoxazole propionate receptors (AMPA-Rs) with an electrophysiological tag were expressed in rat hippocampal neurons. Long-term potentiation (LTP) or increased activity of the calcium/calmodulin-dependent protein kinase II (CaMKII) induced delivery of tagged AMPA-Rs into synapses. This effect was not diminished by mutating the CaMKII phosphorylation site on the GluR1 AMPA-R subunit, but was blocked by mutating a predicted PDZ domain interaction site. These results show that LTP and CaMKII activity drive AMPA-Rs to synapses by a mechanism that requires the association between GluR1 and a PDZ domain protein.

Long-term potentiation (LTP) of synaptic transmission is a well-characterized form of activity-dependent plasticity likely to play important roles in learning and memory (1). A key mediator of this plasticity is CaMKII, an enzyme that is strongly expressed at excitatory synapses (2). Although the cell biological processes underlying this form of plasticity are poorly understood, the trafficking of synaptic receptors appears to play a crucial role (3).

Genes of interest were delivered to neurons in organotypically cultured hippocampal slices, using the Sindbis virus expression system (3–5). Neurons carrying foreign genes were identified by green fluorescent protein (GFP) expression and whole-cell recordings were obtained (Fig. 1A) (6). To examine the effect of elevated CaMKII activity on neuronal function, we generated a construct encoding the catalytic domain of this enzyme fused with GFP (tCaMKII-GFP). The expression of this construct increased constitutive CaMKII activity in baby hamster kidney (BHK) cells (Fig. 1B) (7). In hippocampal slice neurons expressing this construct, fluorescence was detected in dendritic arbors and spines (8). To determine the effect on synaptic transmission, we measured synaptic responses in two nearby neurons, one infected with tCaMKII-GFP (indicated by GFP fluorescence) and the other uninfected (Fig. 1, A and D). Such pairwise comparisons of synaptic responses to stimuli delivered at the same site showed that tCaMKII-GFP enhanced transmission (Fig. 1D) (9–11). Cells infected with GFP alone did not show any change in synaptic response (Fig. 1C).

Figure 1

Enhanced synaptic transmission in neurons expressing tCaMKII-GFP. (A) Hippocampal CA1 pyramidal neuron infected with Sindbis virus and expressing GFP. Fluorescent (top) and differential interference contrast (bottom) images of the same field during electrophysiological recording. Bar: 30 μm. Schematic diagram on right: Whole-cell recordings were obtained from a fluorescent (infected) and an adjacent nonfluorescent (uninfected) neuron with identical stimulation position and intensity. (B) Calcium/calmodulin–independent kinase activity of tCaMKII constructs. BHK cells were infected with respective Sindbis viruses, and Ca2+/calmodulin–independent kinase activity was determined. Controls were uninfected cells (top) or lacZ-infected cells (bottom). (C) Synaptic responses from neurons expressing GFP or from nearby nonexpressing cells. (Left panel) For each pair of cells, the amplitude of response from infected cell is plotted against amplitude of response in uninfected cell (n = 27). Mean of all values is shown in filled circle (uninfected: 35.1 ± 6.3; infected: 33.4 ± 5.0). (Right panel) Summary results of measured rectification for uninfected (2.4 ± 0.1, n = 15) and infected (2.4 ± 0.1, n = 17) cells. Sample responses from nearby uninfected and infected cells are overlaid and shown on right side of each panel. For rectification, responses were obtained at –60 and +40 mV; responses from infected and uninfected cells are not necessarily from nearby cells. Scaled: responses from uninfected cell were scaled so that the current at +40 mV matched that of the infected cell. Bars: 20 pA, 25 ms. Same symbols, trace display conventions, and bar values are used in subsequent figures. (D) tCaMKII-GFP produces enhancement of synaptic transmission in expressing neurons (left panel: uninfected: 15.4 ± 2.4; infected: 25.3 ± 2.5, n = 35) with no effect on rectification (right panel: 2.2 ± 0.1,n = 14 for uninfected and 2.2 ± 0.3,n = 12 for infected cells).

To examine if this increase in AMPA-R–mediated transmission was due to a delivery of receptors to synapses, we developed an electrophysiological assay. The current-voltage (I-V) relationship of AMPA-Rs is determined by the GluR2 subunit: AMPA-Rs with GluR2 show linearI-V relations; AMPA-Rs lacking GluR2 show little outward current at +40 mV (12). Most AMPA-Rs in hippocampal pyramidal cells contain the GluR2 subunit (13), consistent with the linear I-V relationship of synaptic transmission (14). We overexpressed the AMPA-R GluR1 subunit in hippocampal slice neurons (Fig. 2A). This protein was tagged with GFP to facilitate identification of expressing neurons and immunoprecipitation (3). As demonstrated by coimmunoprecipitation experiments, most of the resulting recombinant AMPA-Rs lacked GluR2 (Fig. 2B) (15). Such receptors were functional and showed complete inward rectification when expressed in HEK293 cells (3). Thus, incorporation of these recombinant receptors into synapses would be expected to increase rectification of synaptic responses (16).

Figure 2

Increased CaMKII activity delivers GluR1-GFP into synapses. (A) Fluorescence image of a hippocampal slice expressing GluR1-GFP. Sindbis virus particles expressing GluR1-GFP were injected into several sites of CA1 pyramidal cell layer. DG, dentate gyrus. Bar: 300 μm. (B) Immunoprecipitation study indicates that GluR1-GFP forms largely homomers. GluR1-GFP was immunoprecipitated from hippocampal slices with anti-GFP (two left lanes) or anti-GluR1 (two right lanes) and blotted with anti-GFP (top), anti-GluR1 (middle), and anti-GluR2 (bottom). GluR2 immunoreactivity did not coprecipitate with GFP immunoreactivity. In contrast, GluR2 coprecipitated with endogenous GluR1, indicating coimmunoprecipitation itself was successful. The lower bands observed in the GluR2 blot derive from antibody used for immunoprecipitation. (C) Synaptic responses from neurons expressing GluR1-GFP and nearby nonexpressing cells. Expression of GluR1-GFP did not affect the amplitude (uninfected: 32.7 ± 4.1; infected: 31.0 ± 3.3,n = 13) or rectification (2.2 ± 0.1,n = 42 for uninfected; 2.4 ± 0.1,n = 41 for infected cells). (D) Coexpression of GluR1-GFP with tCaMKII increased amplitude (uninfected: 25.7 ± 5.3; infected: 41.3 ± 5.1, n = 13) and rectification (2.2 ± 0.4, n = 75 for uninfected; 4.3 ± 0.7, n = 29 for infected cells).

GluR1-GFP is widely distributed throughout the dendritic arbor, but little is incorporated into synapses in the absence of activity (3). In agreement with this, expression of GluR1-GFP had no effect on either the amplitude or rectification of synaptic transmission (Fig. 2C). To determine if CaMKII activity could drive the recombinant GluR1-GFP into synapses, we coexpressed GluR1-GFP and tCaMKII by using an internal ribosomal entry site (IRES) construct (5, 17). BHK cells expressing this construct showed increased constitutive CaMKII activity (Fig. 1B) (7), and slices expressing this construct showed fluorescence (indicating GluR1-GFP expression) (8). Pairwise recordings from infected and noninfected cells showed that transmission was enhanced (Fig. 2D), consistent with an increase of CaMKII activity (10, 18). Notably, transmission showed increased rectification, indicating a contribution of the homomeric GluR1-GFP to transmission (Fig. 2D). This effect on rectification was due to coexpression of the two proteins, because transmission onto cells expressing either tCaMKII or GluR1-GFP alone had rectification comparable to that in uninfected cells (Figs. 1D and 2C). These results show that CaMKII activity induces the insertion of homomeric GluR1-GFP into the synapse.

GluR1 is phosphorylated by CaMKII at Ser831 during LTP (19). To examine if direct phosphorylation of the receptor at this site is required for delivery, we substituted Ser831 with Ala, thus creating GluR1(S831A)-GFP (Fig. 3, A and B). This mutation, however, did not block delivery. Expression of this construct alone changed neither amplitude nor rectification (Fig. 3C), and coexpression with tCaMKII produced potentiated transmission that showed the same increase in rectification as that seen with GluR1-GFP-IRES-tCaMKII (Fig. 3D).

Figure 3

CaMKII-driven delivery of GluR1 to synapses is not dependent on phosphorylation at Ser831 but on interaction with putative PDZ domain protein. (A) COOH-terminus of GluR1 wild-type (WT) and two point mutants used in this study. Asterisks indicate stop codon. (B) Current-voltage relationship of current evoked by kainate (1 mM) onto HEK cells expressing GluR1(S831A)-GFP (filled circles, n = 3) and GluR1(T887A)-GFP (open circles, n = 3). Responses were recorded at –60 to +40 mV (in 20-mV steps) and are normalized by the response at –60 mV. Sample traces are shown on right. Bar: 50 pA, 200 ms. (C) GluR1(S831A)-GFP alone did not have any effect on either amplitude (uninfected: 35.8 ± 4.5; infected: 36.0 ± 3.7, n = 9) or rectification (2.4 ± 0.3, n = 14 for uninfected cells; 2.3 ± 0.3, n = 16 for infected cells, P = 0.55; see Web figure 2). (D) Coexpression of tCaMKII and GluR1(S831A)-GFP resulted in increased amplitude (uninfected: 14.4 ± 2.6; infected: 38.7 ± 2.7, n = 9) and rectification (2.1 ± 0.1, n = 36 for uninfected cells; 3.6 ± 0.2, n = 15 for infected cells,P < 10−6; see Web figure 2) indicating that receptor delivery is not dependent on the phosphorylation of the receptor at Ser831. (E) Expression of GluR1(T887A)-GFP had no effect on synaptic amplitude (uninfected: 25.5 ± 4.9; infected: 27.7 ± 5.1,n = 10) or rectification (2.2 ± 0.2,n = 14 for uninfected cells; 2.4 ± 0.2,n = 11 for infected cells, P = 0.51; see Web figure 2). (F) Coexpression of GluR1(T887A)-GFP with tCaMKII blocked potentiation by tCaMKII. Coexpression resulted in depressed transmission (infected: 28.9 ± 4.3; uninfected: 38.9 ± 4.2, n = 13) with no change in rectification (2.2 ± 0.1, n = 36 for uninfected cells; 2.4 ± 0.2, n = 15 for infected cells,P = 0.73; see Web figure 2).

The subcellular localization of many membrane proteins is controlled by associations with a class of proteins containing PDZ domains (20). (The name PDZ derives from the proteins PSD-95, Dlg, and ZO1, which contain the domain.) In particular, the very COOH-terminus of the cytosolic tail of such surface proteins has a consensus (S/T)X(V/L) (20, 21). Serine or threonine at −2 position appears to be crucial because a mutation at this site can prevent associations (20). The last three amino acids of the COOH-terminus of GluR1 are TGL, which conforms to the consensus sequence. We converted the GluR1 COOH-terminus from TGL to AGL, thus creating GluR1(T887A)-GFP (Fig. 3A). This protein, when expressed in HEK293 cells, formed functional AMPA-Rs that showed the normal rectification (Fig. 3B). When expressed in hippocampal neurons, this protein was detected in dendrites (8). This construct showed no effect on transmission when expressed alone (Fig. 3E). However, when GluR1(T887A)-GFP and tCaMKII were coexpressed in hippocampal slice neurons, the effects of tCaMKII on synaptic response amplitude and rectification were completely blocked. Indeed, transmission onto these neurons was depressed (Fig. 3F).

To determine if LTP delivers AMPA-Rs to synapses through a similar mechanism, we examined LTP in cells expressing GluR1-GFP. Whole-cell recordings were obtained from cells expressing (Fig. 4A, top) or not expressing (Fig. 4A, bottom) GluR1-GFP. LTP was induced with a pairing protocol (6). We addeddl-2-amino-5-phosphonovaleric acid (APV) to the bath 30 min after potentiated transmission was measured, in order to isolate pure AMPA-R–mediated responses. The holding membrane potential was then switched to measure rectification of the AMPA-R–mediated responses. Similar to the effect of coexpressed CaMKII and GluR1-GFP, rectification was increased after LTP in cells expressing GluR1-GFP (4.6 ± 0.7; n = 14) compared to cells not expressing GluR1-GFP (2.7 ± 0.2; n = 15) (Fig. 4B). In 5 of the 14 experiments on cells expressing GluR1-GFP, a control (nonpaired) pathway was monitored, and rectification in this pathway was not increased (Fig. 4B).

Figure 4

LTP induction delivers GluR1 into synapses. Delivery requires interaction between GluR1 and a PDZ-domain protein. (A and B) Increased rectification after LTP in cells expressing GluR1-GFP, but not in uninfected cells. Recordings from CA1 pyramidal neurons infected with GluR1-GFP [(A), top] or uninfected neurons [(A), bottom]. LTP was induced as described (3, 6). After a stable potentiation period of 30 min, APV was added to isolate the AMPA component of transmission, and rectification was then determined (6). Synaptic current at 0 mV was not significantly different in infected and noninfected cells (5.3 ± 0.5% and 3.7 ± 0.8%, respectively, expressed as percent of current at –60 mV, P > 0.1) (B) Rectification after LTP induction was statistically different between infected and uninfected cells (P < 0.04, Mann-Whitney test), and between potentiated and control pathways of GluR1-GFP–expressing neurons (in five cells which two pathways were recorded, paired t test, P < 0.04). Sample traces [recorded at periods indicated in (A)] are shown in bottom. (C and D) Expression of GluR1(T887A)-GFP blocks LTP and leads to depression after pairing. Whole-cell recording from neurons expressing either GluR1-GFP or GluR1(T887A)-GFP. (C) Plot of excitatory postsynaptic current amplitude versus time for individual experiments from a cell expressing GluR1-GFP (top) or GluR1(T887A)-GFP (bottom). (D) Ensemble averages from 14 cells expressing GluR1-GFP (open circles) and 21 cells expressing GluR1(T887A)-GFP (closed squares). Average of four control pathways recorded from GluR1(T887A)-GFP cells are also shown (open squares). Responses from cells expressing GluR1(T887A)-GFP showed a short-term potentiation followed by a persistent depression below baseline levels (P < 0.04 Wilcoxon test), whereas cells expressing GluR1-GFP showed potentiation (P < 0.01 Wilcoxon test). Sample traces [recorded at periods indicated in (C)] are shown at bottom.

We also examined the effects of GluR1(T887A) on LTP. As shown above, this receptor, which has the PDZ-interaction site mutated, can completely block the potentiation produced by tCaMKII. In a series of blind experiments, we recorded synaptic responses from cells expressing either GluR1-GFP or GluR1(T887A)-GFP. After pairing, cells expressing the control construct displayed stable potentiation lasting at least 50 min, at which time the recording was terminated (Fig. 4C, top, and Fig. 4D, top). Cells expressing GluR1(T887A)-GFP displayed a very different response. After a pairing protocol, these cells showed a short-lasting potentiation that decayed over the next 20 min; 45 min after pairing, the responses were significantly depressed from baseline levels (Fig. 4C, bottom, and Fig. 4D, top). In 4 of the 21 experiments with GluR1(T887A), a control (nonpaired) pathway was monitored, which did not show depression (Fig. 4D, top).

The cell-biological mechanisms underlying synaptic plasticity have been difficult to delineate. In part, this is due to the lack of techniques in intact preparations allowing molecular perturbations with spatial and temporal control, as well as the absence of assays for specific molecular events linked to synaptic plasticity. Here, we generated electrophysiologically tagged receptors to monitor their synaptic delivery during LTP and increased CaMKII. In the absence of plasticity-inducing stimuli, we saw no evidence for their contribution to transmission. This is consistent with previous results indicating that in the absence of evoked activity, GluR1 is retained within the dendrite (3). Upon coexpression with constitutively active CaMKII or following LTP induction, we see that tagged receptors contribute to transmission, indicating their delivery to synapses.

Previous studies indicate that LTP induction increases the CaMKII-dependent phosphorylation of GluR1 at Ser831 (19). Although such phosphorylaton may enhance the function of synaptic receptors (22), this phosphorylation does not seem to be required for receptor delivery: tCaMKII can deliver GluR1(S831A)-GFP to the synapse (23). Our results indicate that some protein(s) other than GluR1 must be substrate(s) of CaMKII and participate in the regulated synaptic delivery of AMPA-Rs.

The most surprising of our results relate to the effects of GluR1(T887A). This protein forms functional receptors and has no detectable effects on basal synaptic transmission. However, this mutant receptor can block the effects of tCaMKII and LTP (24). This has several implications: (i) It reinforces the view that CaMKII and LTP act through similar mechanisms. (ii) It indicates that both CaMKII-potentiation and LTP exert their effects through GluR1. (iii) It indicates that an interaction between GluR1 and a protein with a PDZ domain plays a key intermediate in these forms of plasticity. (iv) GluR1(T887A) depresses transmission, but only after increased CaMKII or LTP. This last finding suggests that activity enables the mutant protein to interrupt a constitutive delivery of endogenous AMPA-Rs (25).

These results demonstrate that incorporation of GluR1-containing AMPA-Rs into synapses is a major mechanism underlying the plasticity produced by activation of CaMKII and LTP. This process requires phosphorylation of protein(s) other than GluR1. Furthermore, this delivery requires interactions between the COOH-terminus of GluR1 and PDZ domain proteins.

  • * These authors contributed equally to this work.

  • Present address: INSERM U261, Institut Pasteur, 25 rue du Dr. Roux, 75015 Paris, France.

  • To whom correspondence should be addressed. E-mail: malinow{at}


Stay Connected to Science

Navigate This Article