Regulatory Phosphorylation of AMPA-Type Glutamate Receptors by CaM-KII During Long-Term Potentiation

See allHide authors and affiliations

Science  27 Jun 1997:
Vol. 276, Issue 5321, pp. 2042-2045
DOI: 10.1126/science.276.5321.2042


Long-term potentiation (LTP), a cellular model of learning and memory, requires calcium-dependent protein kinases. Induction of LTP increased the phosphorus-32 labeling of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)–type glutamate receptors (AMPA-Rs), which mediate rapid excitatory synaptic transmission. This AMPA-R phosphorylation appeared to be catalyzed by Ca2+- and calmodulin-dependent protein kinase II (CaM-KII): (i) it correlated with the activation and autophosphorylation of CaM-KII, (ii) it was blocked by the CaM-KII inhibitor KN-62, and (iii) its phosphorus-32 peptide map was the same as that of GluR1 coexpressed with activated CaM-KII in HEK-293 cells. This covalent modulation of AMPA-Rs in LTP provides a postsynaptic molecular mechanism for synaptic plasticity.

Long-term potentiation is a prolonged enhancement in synaptic efficacy that may underlie certain types of learning and memory, but its molecular mechanisms are unclear (1). Considerable evidence implicates changes in presynaptic transmitter release, postsynaptic responses, and synaptic structural changes. Postsynaptic elevations in Ca2+ and Ca2+-dependent protein kinases are required for establishment of LTP, and a likely target of these kinases is AMPA-Rs (2) because (i) their responsiveness is enhanced after elevations of postsynaptic Ca2+ (3) or by LTP induction (4), and (ii) these changes are blocked by KN-62 (3, 5), a CaM-K inhibitor (6). Activated CaM-KII enhances AMPA-R responsiveness in CA1 neurons in hippocampal slices and many other systems (7-9). In the slices, expression or infusion of activated CaM-KII also increases synaptic current and occludes subsequent induction of LTP. However, a key observation, direct phosphorylation of AMPA-Rs in response to LTP, has not been previously demonstrated.

Induction of LTP produces a small increase in the Ca2+-independent or constitutive activity of CaM-KII (10). CaM-KII can autophosphorylate on multiple sites (2, 11), so to confirm that the constitutive CaM-KII activity was due to autophosphorylation of Thr286, we used a phosphospecific antibody, AbP-Thr286. This antibody was specific for P-Thr286 in CaM-KII, reacting with autophosphorylated wild-type CaM-KII but not with the autophosphorylated Thr286-Ala mutant (Fig.1A). Protein immunoblot analyses with AbP-Thr286 of a hippocampal slice extract detected multiple immunoreactive bands, but only the one corresponding to the 50-kD α-CaM-KII was selectively blocked by preadsorption with the phosphopeptide antigen (Fig. 1B). Induction of LTP in the CA1 region of hippocampal slices with theta-burst stimulation (10) resulted in a small, stable increase in immunoreactivity to AbP-Th286 (Fig. 1C). The magnitude of the increase is in general agreement with the estimate that about 10% of synapses are potentiated by theta-burst stimulation (12). The enhanced phosphorylation of Thr286 was not due to an LTP-mediated increase in CaM-KII protein, because immunoreactivity with a general CaM-KII antibody (n = 4) did not increase at 5 min (93.6 ± 11.5% of control) or 15 min (98.7 ± 1.5%), and the small increase at 60 min was not significant (122 ± 16%,P > 0.1).

Figure 1

LTP increased phosphorylation of CaM-KII. (A) Protein immunoblot analysis of autophosphorylated CaM-KII using AbP-Thr286 (27). Wild-type CaM-KII (lanes 1, 2, and 5) or mutant Thr286-Ala (lanes 3, 4, and 6) were subjected to Ca2+- and CaM-dependent autophosphorylation (11) at 5°C for 5 min (lanes 1 and 3) or 10 min (lanes 2 and 4) and then at 30°C for 30 min (lanes 5 and 6). (B) Protein immunoblot of hippocampal slice homogenate with either a general CaM-KII antibody (lane 1) or AbP-Thr286 without (lane 2) or with preadsorption with the phosphopeptide antigen (lane 3) or non-phosphopeptide (lane 4). Arrow shows the position of α-CaM-KII. (C) AbP-Thr286 protein immunoblot of hippocampal slices at 5, 15, or 60 min after theta-burst stimulation to induce LTP (28) or 60 min after low-frequency stimulation (CON). Top panel, illustrative example; bar graph, composite values normalized per microgram of homogenate protein (mean ± SE, n = 4, *P < 0.05). (D) Homogenates from32P-labeled hippocampal slices (LTP or CON) were immunoprecipitated with general CaM-KII antibody and subjected to SDS-PAGE and autoradiography (inset), and 32P–CaM-KII was quantitated per microgram of protein (mean ± SE, n = 5 to 8 homogenates, *P < 0.05). Slashed bars represent the α subunit; solid bars, the β subunit.

When total autophosphorylation of CaM-KII was measured in32P-labeled hippocampal slices by immunoprecipitation with a general CaM-KII antibody, LTP induction produced a significant increase in 32P–CaM-KII at 15 min that was further elevated at 60 min (Fig. 1D). The fact that phosphorylation of Thr286 was maximal at 5 min (Fig. 1C) whereas total autophosphorylation was slower (Fig. 1D) is consistent with the fact that autophosphorylation of Thr286 is much faster than autophosphorylation of other sites or of exogenous substrate proteins (11).

We next examined whether induction of LTP increased32P labeling of AMPA-Rs. Conditions were optimized to immunoprecipitate more than 90% of GluR1 from the homogenate (Fig.2A, lanes 1 and 2). The immunoprecipitate was also reactive to a GluR2/3 antibody (13) (Fig. 2A, lanes 3 and 4), indicating immunoprecipitation of the hetero-oligomeric AMPA-R complex. The amount of 32P–AMPA-R was significantly elevated 15 min (24.0 ± 11.6% over control, P < 0.05, n = 5, normalized by immunoreactivity) and 60 min (42.5 ± 8.6%, P < 0.01, n = 8) after induction of LTP compared with the low-frequency stimulated control (Fig. 2B). This enhanced phosphorylation of AMPA-Rs, as well as the phosphorylation of CaM-KII, was not due to an LTP-induced increase in total protein phosphorylation (Fig. 2C, n = 5 to 8), thereby ruling out an increase in adenosine triphosphate (ATP)–specific activity.

Figure 2

LTP enhanced phosphorylation of AMPA-R. (A) Immunoprecipitation of AMPA-R complex (28). Protein immunoblots with GluR1 antibody (lanes 1 through 3) or GluR2/3 antibody (lane 4) of either slice supernatants before (lane 1) and after (lane 2) immunoprecipitation with GluR1 antibody or of GluR1 immunoprecipitates (lanes 3 and 4). Representative autoradiographs from SDS-PAGE of immunoprecipitated AMPA-R (lanes 5 and 6) from 60-min control and LTP 32P-labeled slices. (B) AMPA-R immunoprecipitates from CON or LTP slices at the indicated times were quantitated for 32P incorporation (mean ± SE, n = 5 to 8 homogenates, *P < 0.05, **P < 0.01). Data were normalized to GluR1 protein immunoblots from the same gel (slashed bars) or 1 μg of total homogenate protein per sample (solid bars). No significant difference in AMPA-R immunoreactivity was detected between CON and LTP samples at any time points. (C) 32P incorporation into total cellular proteins was determined in 32P-labeled hippocampal slices at the indicated times (n = 5 to 8). Aliquots of slice homogenates were spotted on P81 paper and analyzed for total 32P-protein (28).

To examine whether AMPA-R phosphorylation was catalyzed by the activated CaM-KII, we used the CaM-kinase inhibitor KN-62, which blocks induction of LTP (5) as well as Ca2+-dependent potentiation (3) in hippocampal slices. The KN-62 did not affect basal synaptic transmission or suppress short-term potentiation (STP) (that is, the first 10 min of potentiation), but it blocked the expression of LTP (Fig.3A) and the enhanced phosphorylations of CaM-KII and AMPA-Rs (Fig. 3B). Because KN-62 inhibits CaM-KII competitively with Ca2+ and CaM and does not inhibit protein kinase A or C (PKA or PKC) in vitro or in vivo (6,14), the data of Fig. 3, A and B, strongly support the conclusion that CaM-KII was the catalyst of the AMPA-R phosphorylation. This conclusion is important, because PKA and PKC can also be activated during LTP (15, 16), and both can phosphorylate AMPA-Rs in cultured cells (17, 18). However, activation of PKA is transient (5 to 15 min) and requires stimuli much stronger than theta-burst stimulation (16), and effects of PKC on AMPA-R responsiveness are controversial (19). As expected, theN-methyl-d-aspartate receptor (NMDA-R) antagonist D-AP5 also blocked AMPA-R phosphorylation (Fig.3C).

Figure 3

KN-62 and D-AP5 block LTP and phosphorylation of AMPA-R and CaM-KII. (A) Changes in excitatory postsynaptic potential (EPSP) slope before or after induction of LTP (10) in slices treated without (open circles; n = 4) or with (solid diamonds;n = 5) 10 μM KN-62 (Seigakagu). Data are expressed as a percentage of baseline values. (B) 32P labeling of AMPA-R (n = 12 homogenates, normalized to AMPA-R immunoreactivity) and CaM-KII (n = 11 homogenates, normalized to protein) subunits were analyzed at 60 min in slices treated with 10 μM KN-62. Slices were incubated in phosphate-free medium plus KN-62 for 30 min and32P-labeled in the presence of KN-62 for another 30 min before theta-burst stimulation. Treatment with KN-62 had no effect on total 32P-protein. (C) Effect of AP5 on AMPA-R phosphorylation. Slices were treated as in (B) but with 50 μM D-AP5. Data are the average of two experiments.

When peptide mapping was performed on 32P–AMPA-Rs from control and LTP slices, both peptide maps appeared identical (Fig.4, A and B), suggesting that LTP increased phosphorylation of the same site. This result is similar to our previous observation on AMPA-R phosphorylation upon stimulation of NMDA-R in cultured hippocampal neurons (20). To further confirm that phosphorylation of AMPA-R in hippocampal slices was catalyzed by CaM-KII, GluR1 was expressed in HEK-293 cells with or without CaM-KII. Little basal 32P labeling of the GluR1 was observed, but coexpression of GluR1 with CaM-KII gave a 10-fold increase in 32P-GluR1 (Fig. 4C, top panel) that showed a major 32P-peptide (Fig. 4C, bottom panel), which must have been the same as that observed in the hippocampal slices because a mixture of the two did not reveal any differences (Fig. 4D). CaM-KII enhanced the AMPA-R responsiveness of recombinant and native receptors (7-9) and the response in HEK-293 cells (Fig. 4, E and F). These results indicate that the site phosphorylated by CaM-KII in HEK-293 cells and in hippocampal slices after induction of LTP enhanced AMPA-R responsiveness. This phosphorylation site is as yet unidentified because several previously identified regulatory sites appear to be extracellular (21). A recent study has identified PKA and PKC phosphorylation sites in the intracellular COOH-terminus of GluR1 (18).

Figure 4

32P-peptide mapping of AMPA-Rs. (A and B) Peptide maps of AMPA-Rs from hippocampal slices. Immunoprecipitated 32P–AMPA-Rs from 60-min CON or LTP slices were excised from the SDS-PAGE, subjected to complete tryptic digestion, and resolved by two-dimensional (2D) peptide mapping (29). The main phosphopeptide is circled. (C and D) Phosphorylation of GluR1 (flip) expressed in HEK-293 cells (30) with or without coexpression of the CaM-KII mutant H282R (26), which has 20% constitutive activity. Two days after transfection, cells were32P-labeled for 1.5 hours, GluR1 was immunoprecipitated and analyzed by SDS-PAGE and autoradiography [(C), duplicates shown], and the 32P-GluR1 from (C) was subjected to 2D mapping as in (A) and (B). The main phosphopeptide is circled. Trypsin-digested hippocampal slice (LTP-induced) AMPA-Rs were mixed with equal amounts (in counts per minute) of HEK-293 cell GluR1 (plus CaM-KII H282R) and subjected to 2D peptide mapping (D). (E and F) Effect of CaM-KII on whole cell currents recorded from HEK-293 cells expressing GluR1 (flip) (30). (E) Responses elicited by application of 10 mM glutamate (100-ms pulses, −80 mV) immediately after formation of the patch (0 min) and 20 min after infusion of activated CaM-KII (30). (F) Peak currents, normalized to 0 time, for cells infused with activated (solid circles,n = 9) or heat-inactivated (open squares, n = 10) CaM-KII.

Our results, combined with reports of Tyr phosphorylation of NMDA-Rs after LTP (22), provide evidence for prolonged biochemical postsynaptic changes during LTP consistent with a current model (23). The Ca2+ influx through the NMDA-R in the dendritic spine triggers the rapid Ca2+-dependent autophosphorylation of Thr286 (Fig. 1C) and activation of CaM-KII (10). This activated CaM-KII then catalyzes slow Ca2+-independent autophosphorylation (Fig. 1D) and phosphorylation of AMPA-Rs (Fig. 2B) on a site (Fig. 4, B through D) that enhances AMPA-R responsiveness (Fig. 4F). This slow phosphorylation of AMPA-R correlated temporally with the previously reported increase in AMPA-R responsiveness after LTP induction (4). The delay of 10 to 15 min in AMPA-R phosphorylation by CaM-KII suggests that this mechanism is not involved in STP, consistent with the observation that KN-62 also does not block STP but is only inhibitory after about 10 to 15 min (Fig. 3A). Although ours and other data (7-9) implicate a critical role for AMPA-R phosphorylation by CaM-KII in LTP, it is likely that CaM-KII modulates additional aspects of synaptic plasticity. Mutant mice lacking α-CaM-KII show deficits in LTP induction and spatial learning, but mice overexpressing a partially active CaM-KII mutant exhibit more complex alterations in long-term potentiation and depression (24). Thus, it is likely that additional mechanisms, regulated by CaM-KII and other enzymes, are also essential for the multiple facets (for example, neurotransmitter release and increased gene expression) of this complex phenomenon.


Stay Connected to Science

Navigate This Article