|Home | About | Journals | Submit | Contact Us | Français|
Although the maintenance mechanism of late-LTP is critical for the storage of long-term memory, the expression mechanism of synaptic enhancement during late-LTP is unknown. The autonomously active PKC isoform, protein kinase Mzeta (PKMζ), is a core molecule maintaining late-LTP. Here we show that PKMζ maintains late-LTP through persistent N-ethylmaleimide-sensitive factor (NSF)/GluR2 subunit-dependent trafficking of AMPARs to the synapse. Intracellular perfusion of PKMζ into CA1 pyramidal cells causes potentiation of postsynaptic AMPAR responses; this PKMζ-mediated enhancement is mediated through NSF/GluR2 interactions, but not vesicle-associated membrane protein (VAMP)-dependent exocytosis. PKMζ may act through NSF to release GluR2-containing receptors from a reserve pool held at extrasynaptic sites by protein interacting with C-kinase 1 (PICK1), because disrupting GluR2/PICK1 interactions mimics and occludes PKMζ-mediated AMPAR potentiation. During LTP maintenance, PKMζ directs AMPAR trafficking, as measured by NSF/GluR2-dependent increases of GluR2/3-containing receptors in synaptosomal fractions from tetanized slices. Blocking this trafficking mechanism reverses established late-LTP and persistent potentiation at synapses that have undergone synaptic tagging and capture. Thus PKMζ maintains late-LTP by persistently modifying GluR2/NSF-dependent AMPAR trafficking to favor receptor insertion into postsynaptic sites.
The maintenance mechanism of the late, protein synthesis-dependent phase of LTP is critical for the storage of long-term memory (Pastalkova et al., 2006; Shema et al., 2007). Although the expression of the early induction phase of LTP has been studied extensively (Bliss and Collingridge, 1993), the mechanism for synaptic enhancement during the late phase of LTP is unknown. One approach to identify the expression mechanism of synaptic enhancement during late-LTP is to examine how the molecular mechanism that maintains late-LTP enhances synaptic transmission. A prime candidate for a core molecule maintaining late-LTP is PKMζ, an autonomously active isozyme of PKC (Bliss et al., 2006).
PKMζ maintains synaptic enhancement during late-LTP through its second messenger-independent and thus persistent kinase activity. PKMζ consists of an independent PKCζ catalytic domain produced from a brain-specific PKMζ mRNA, which, lacking an autoinhibitory PKCζ regulatory domain, is constitutively active (Sacktor et al., 1993; Hernandez et al., 2003). During LTP induction, afferent tetanic stimulation increases the synthesis of PKMζ from its mRNA (Hernandez et al., 2003; Muslimov et al., 2004; Kelly et al., 2007), and the resulting persistent increase in the autonomously active kinase is both necessary and sufficient for maintaining LTP (Ling et al., 2002). Postsynaptic perfusion of PKMζ increases the efficacy of AMPAR-mediated synaptic transmission (Ling et al., 2002; Ling et al., 2006), whereas inhibitors of the kinase activity of PKMζ reverse LTP even when applied hours after the tetanization, without affecting baseline synaptic transmission (Ling et al., 2002; Sajikumar et al., 2005; Serrano et al., 2005; Pastalkova et al., 2006).
PKMζ maintains a stable enhancement of synaptic transmission exclusively by increasing the number of functional postsynaptic AMPARs (Ling et al., 2006); therefore, the kinase may act to upregulate one of the AMPAR trafficking pathways that maintains a constant number of postsynaptic receptors under basal conditions. We therefore began by examining two core pathways that regulate basal AMPAR trafficking: one dependent upon the interaction between the chaperone-like protein NSF and the intracellular C-terminus of the GluR2 AMPAR subunit (Nishimune et al., 1998; Osten et al., 1998; Song et al., 1998); the other an exocytotic pathway requiring a membrane fusion event mediated by VAMP/synaptobrevin (Luscher et al., 1999).
For experiments in Figs. 1--4,4, and Supplementary Figs. 1-6, hippocampi of male, ~3-4 week-old Sprague Dawley rats were removed following decapitation under halothane-induced anesthesia, according to State University of New York Downstate Medical Center Animal Use and Care Committee standards. For tagging experiments in Supplementary Fig. 7, male, 7 week-old Wistar rats were used in accordance with Leibniz Institute for Neurobiology animal care standards.
For whole-cell recordings, hippocampal slices (400 μm) were prepared using a Vibratome tissue sectioner. The slices were transferred to an incubation chamber at 32°C in oxygenated (95% O2- 5% CO2) physiological saline containing (in mM): 124 NaCl, 5 KCl, 26 NaHCO3, 1.6 MgCl2, 4 CaCl2, 10 glucose, for a minimum of 1.5 hr prior to recording. Following incubation, single slices were transferred to a 1.5 ml recording chamber placed on the stage of an upright microscope (Zeiss Axioskop 2, Carl Zeiss, Thornwood, NY) and perfused with warm (31-33°C) saline at ~5 ml/min. Visualized CA1 pyramidal cells were held at –75 mV, the empirically determined reversal potential of the GABAA receptor (Ling et al., 2002). Synaptic events were evoked every 15 sec (or every 2.5 sec in Supplementary Fig. 1) by extracellular stimulation with bipolar electrodes placed in stratum radiatum. The recording pipettes had tip resistance of 2-4 MΩ and contained (in mM): 130 Cs-MeSO4, 10 NaCl, 2 EGTA, 10 HEPES, 1 CaCl2, 2 Na-ATP, 0.5 Na-GTP, ± PKMζ (7-20 nM, 0.5-0.9 pmol/min/μl phosphotransferase activity (Ling et al., 2002)), ± peptides (100 μM, unless otherwise specified): pep2m (KRMKVAKNAQ, Tocris), scr-pep2m (VRKKNMAKQA, Pepsyn), pepNSF3 (TGKTLIARKIGTMLNAREPK, Pepsyn), pep-ΔA849-Q853 (KRMKLNINPS, Quality Control Biochemicals), pep2-EVKI (YNVYGIEEVKI, Tocris), pep2-EVKE (YNVYGIEEVKE, Quality Control Biochemicals), or Botox B (0.5 μM, List Biological Laboratories). (A myristoyl group was added to the N-terminal during synthesis of myr-pep2m and scr-myr-pep2m.) The pH value of the pipette solution was adjusted to 7.3 with CsOH. Cs was used to block potassium currents, including GABAB responses. Evoked AMPAR-mediated EPSCs were recorded under voltage-clamp mode with a Warner PC-501A amplifier (Warner Instruments, Hamden CT) and filtered at 2 kHz (-3dB, 4 pole Bessel). Brief voltage steps (-5 mV, 5 or 10 ms) from holding potential were applied during the course of recording to monitor cell access resistance, input resistance, and capacitance. Only cells with an initial input resistance of > 100 MΩ, and an initial access resistance of < 20 MΩ with insignificant change (< 20%) during the course of recording were accepted for study. Signals were digitized with Digidata 1322A, acquired and analyzed with pClamp software (Axon Instruments, Foster City, CA) running on Pentium microcomputer. The peak amplitude of EPSCs was further analyzed with Excel (Microsoft). The mean ± SEM of 1 min bins of the responses were plotted in the figures.
Field EPSPs (fEPSPs) were recorded using glass microelectrodes with a resistance of 3-8 MΩ, filled with the recording saline, and positioned 200 μm from the stimulating electrodes in stratum radiatum. Current intensity of test stimuli (20-40 μA, 0.1 ms duration) was set to produce one-third maximal fEPSPs (1-2 mV). The frequency of test stimulation was every 15 s, and the mean ± SEM of 1 min bins of responses plotted in the figures. The slope of the fEPSP was measured between 10-50% of the initial phase of the fEPSP response. The tetanization protocol for biochemical experiments (Fig. 3) was optimized to produce a relatively rapid onset of a protein synthesis-dependent phase of potentiation (Osten et al., 1996; Tsokas et al., 2005) (Supplementary Fig. 4), and consisted of four 100 Hz-1 sec trains, set at 75% of the maximal fEPSP response (15-30 μA above the test stimuli current intensity), delivered at 20 sec intertrain intervals. For late-LTP in Fig. 4, tetanization was performed as previously described (Serrano et al., 2005), and consisted of four 100 Hz-1 sec trains, set at 75% of the maximal fEPSP response, delivered at 5 min intervals. Control synaptic pathways were shown to be independent by the absence of paired-pulse facilitation between the pathways, as previously described (Serrano et al., 2005). Tagging experiments in Supplementary Fig. 7 were performed with different stimulation and recording techniques as detailed previously (Sajikumar et al., 2005). Briefly, strong tetanization producing late-LTP consisted of three 100 Hz-1 sec trains, duration 0.2 ms per polarity, delivered at 10 min intervals, using monopolar stainless steel electrodes. Weak tetanization producing early LTP consisted of one 100 Hz train of 21 biphasic constant-current pulses, 0.2 ms pulse duration per half-wave at the stimulus intensity of the population spike threshold. Test stimulation for tagging experiments consisted of four 0.2 Hz biphasic, constant-current pulses (0.1 ms per polarity), delivered 1, 3, 5, 11, 15, 21, 25, and 30 min after tetanization and thereafter once every 15 min. Data were analyzed with custom-made software (PWIn; HN Magdeburg).
Two CA1 regions per group were homogenized in 100 μl of homogenization buffer containing (in mM): 50 Tris-HCl [pH 7.4], 1 EDTA, 1 EGTA, 0.1 phenylmethylsulfonyl fluoride, 5 benzamidine, aprotinin (17 kallikrein units/ml), 0.1 leupeptin, and 1% phosphatase inhibitor cocktails 1 and 2. All reagents were from Sigma-Aldrich and all biochemical experiments were at 4°C, unless otherwise stated. Total protein was measured by bicinchoninic acid assay (Pierce Biotechnology). Control slices were from the same hippocampus; in parallel experiments, no changes in AMPAR or PKMζ levels were detected between slices maintained with and without test stimulation (data not shown); therefore unstimulated slices were used for comparisons with LTP. Homogenization was performed in number 19 Kontes glass homogenizing tubes using a Talboy 102 homogenizer (20 strokes at 60 rpm). To measure total AMPARs, 10 μl of this homogenate was spun at 100,000 X g for 30 min, resulting in a membrane pellet that was resuspended in 20 μl homogenization buffer and heated for 10 min at 90°C with sample buffer. To obtain the synaptosomal fraction (Chin et al., 1989), 90 μl of the homogenate was added to 110 μl of 2 M sucrose (resulting in 200 μl of a 1.1 M sucrose homogenate); 160 μl of 0.8 M sucrose was then layered on the 1.1 M sucrose homogenate, and the preparation spun at 1625 X g for 10 min. The 0.8 M sucrose gradient was collected, added to 3 X its volume of homogenization buffer, and spun at 15,555 X g for 8 min. The pellet was resuspended in 20 μl homogenization buffer with sample buffer and heated to 90°C for 10 min. Co-enrichment of AMPAR subunits and the postsynaptic marker, postsynaptic density-95 protein (PSD-95; antiserum 1:500, Chemicon), is shown in Supplementary Fig. 6.
The samples were subjected to SDS-PAGE (20 μl/1 cm lane). After electrophoresis the proteins were transferred to nitrocellulose membrane, and the membrane blocked for 90 min in 1% bovine serum albumin and 1% hemoglobin in Tris-buffered saline with 1% NP-40 (TBSN) (Sacktor et al., 1993). Multiple proteins from a single band were measured by cutting the lanes into up to 8 separate strips and placing strips from different lanes together for probing with specific primary antisera (Sacktor et al., 1993). Primary antisera were GluR1 (1:100, Oncogene), GluR2 (1:500, Chemicon), GluR3 (1:500, Chemicon), incubated overnight at 4°C. Immunostaining for glyceraldehyde-3-phosphate dehydrogenase (GAPDH, 1:10,000, Sigma) was used to normalize AMPARs with respect to protein loading. The membrane was then washed in TBSN followed by a 90 min incubation with secondary antibody coupled to alkaline-phosphatase (1:2000). The blots were developed with 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium and optical density analyzed using NIH Image. Representative images of the bands from these strips are presented in Fig. 3C.
Surface GluR2 labeling was adapted from methods previously described (Grosshans et al., 2002). CA1 regions were dissected from the slices and incubated for 45 min in 1 mg/ml biotin in phosphate-buffered saline at 4°C. After sonication for 10 sec, the surface protein was separated from total protein using streptavidin beads overnight at 4°C, and the surface and total protein subjected to SDS-PAGE and immunoblotting, as described above except development was with chemiluminescence (Supersignal West Femto, Pierce Biotechnology). The relative level of surface GluR2 is expressed as surface/total immunostaining X 100. PKMζ was measured by C-terminal ζ antiserum (1:100) (Sacktor et al., 1993).
Immunoprecipitation (IP) of PKMζ from hippocampal homogenates was performed as previously described (Kelly et al., 2007). For IP of PICK1, the hippocampal homogenate was incubated overnight at 4°C with anti-PICK1 antiserum (N-18, 5 μg/200 μl homogenate, Santa Cruz Biotech). Immunoblots of the immunoprecipitate were performed with the following antisera: PKCα (H-7, 1:100, Santa Cruz Biotech), PKCε (E-5, 1:100, Santa Cruz Biotech), PKMζ (C-terminal, 1:500) (Kelly et al., 2007), and PICK1 (2 μg/ml, Affinity BioReagents).
Values are presented as mean ± SEM. Within group differences were determined by paired t-test and between group comparisons by unpaired t-test or one-way ANOVA followed by Sheffe post-hoc test. One-way ANOVA repeated measurements followed by least significant difference and Tukey post-hoc tests were used for comparisons of subunit levels and multiple time points, respectively.
Blocking the interaction between NSF and GluR2 by postsynaptic perfusion of pep2m, a peptide that mimics the NSF-binding site in GluR2, causes a rundown of basal AMPAR responses (Nishimune et al., 1998; Song et al., 1998). This rundown is dependent on afferent synaptic activity (Luscher et al., 1999) and is reversed by periods of synaptic inactivity (Duprat et al., 2003). This contrasts with the enhancement of synaptic transmission by PKMζ, which is independent of afferent stimulation (Ling et al., 2006). Therefore, in order to isolate an effect of pep2m on PKMζ-mediated potentiation, we minimized the basal rundown induced by the peptide by stimulating at a relatively low rate. Thus, whereas postsynaptic perfusion of pep2m (100 μM) produced baseline rundown with afferent stimulation at a test frequency of every 2.5 sec (Supplementary Fig. 1), it caused only minimal rundown with test stimulation every 15 sec (Fig. 1A, p = 0.30 between responses at 1 and 13 min of recording; p = 1.0 between responses at 13 min with and without pep2m). Using this test stimulation frequency, we examined the effect of pep2m on PKMζ-mediated AMPAR potentiation.
Perfusion of PKMζ alone into CA1 pyramidal cells doubled the amplitude of AMPAR responses within 5 min of recording, as previously described (Ling et al., 2002; Serrano et al., 2005; Ling et al., 2006) (Fig. 1A, p < 0.005 between responses at 1 and 13 min). In interleaved experiments, perfusion of pep2m together with the kinase strongly attenuated the potentiation (Fig. 1A, p < 0.01 between responses at 13 min with PKMζ alone and responses with PKMζ and pep2m, p = 0.20 between 1 and 13 min of recording with PKMζ and pep2m, and p = 0.24 at 13 min between PKMζ with pep2m and pep2m alone. A scrambled, inactive version of the peptide (scr-pep2m) (Liu and Cull-Candy, 2005) had no effect on PKMζ-mediated potentiation (Fig. 1B, p = 1.0 at 13 min between PKMζ alone and PKMζ with scr-pep2m). PKMζ-mediated AMPAR potentiation required the activity of NSF because pep-NSF3, a peptide that inhibits the ATPase activity of NSF (Lee et al., 2002), also blocked the synaptic enhancement (Fig. 1C, p < 0.01 between responses at 13 min with PKMζ alone and with PKMζ and pep-NSF3, p = 0.30 between responses at 1 and 13 min with PKMζ and pep-NSF3, p = 0.16 at 13 min between PKMζ with pep-NSF3 and pep-NSF3 alone). In contrast, a peptide that blocks the interaction between the clathrin adaptor AP2 and GluR2 at a binding site that partially overlaps with that of NSF (Lee et al., 2002) did not affect AMPAR potentiation by PKMζ (Supplementary Fig. 2, responses at 13 min with PKMζ alone and with PKMζ and pep-ΔA849-Q853 are not significantly different, p = 0.83).
We then examined the constitutive, exocytotic AMPAR trafficking pathway to the synapse, which is blocked by botulinum toxin B light chain (Botox B) that cleaves VAMP/synaptobrevin (Luscher et al., 1999). As expected, postsynaptic perfusion of Botox B (0.5 μM) depressed baseline AMPAR-mediated transmission (Fig. 1D, p < 0.05 between responses at 1 and 13 min in the presence of Botox). In contrast to pep2m, however, the toxin did not prevent PKMζ-mediated AMPAR potentiation (p = 0.36, comparing responses at 13 min with PKMζ alone and PKMζ plus Botox with baselines subtracted, Fig. 1D, right insert).
These results indicate that PKMζ enhances AMPAR responses selectively through the NSF/GluR2-dependent AMPAR trafficking pathway. A primary action of NSF on GluR2 is to disrupt the interaction between GluR2 and PICK1 (Hanley et al., 2002). We therefore tested whether blocking this interaction mimics and occludes the effect of PKMζ. Perfusing pep2-EVKI, which selectively disrupts GluR2/PICK1 interactions, relative to GluR2 interactions with glutamate receptor AMPAR binding protein (ABP)-interacting protein (GRIP)/(ABP) (Chung et al., 2000), caused synaptic potentiation (Fig. 2A, 100 μM, p < 0.01 between 1 and 13 min of recording), consistent in kinetics but higher in amplitude than that previously described (Kim et al., 2001) (but see also (Daw et al., 2000), in which no potentiation was observed). A control peptide (pep2-EVKE) had no effect on basal synaptic transmission (Fig. 2A, p = 0.8 between 1 and 13 min of recording). Increasing the amount of pep2-EVKI in the pipette from 100 to 500 μM did not augment the synaptic enhancement, indicating that these concentrations of the peptide in the pipette produced a maximal effect (Fig. 2A, p = 0.9 at 13 min of recording). Adding pep2-EVKI (100 μM) to PKMζ did not increase postsynaptic AMPAR enhancement, as compared to PKMζ alone (Fig. 2B, the potentiation by PKMζ alone vs PKMζ and pep2-EVKI at 13 min of recording was not significantly different, p = 0.2). Although occlusion experiments must be interpreted cautiously, these results suggest that the action of PKMζ and the release of GluR2 from PICK1 enhance postsynaptic AMPAR responses by the same mechanism.
Because PICK1 interacts with PKCα but not other conventional or novel PKC isoforms (Staudinger et al., 1997), we asked if PICK1 also interacts with PKMζ, which had not previously been examined. PKMζ, as well as PKCα, but not the novel PKCε that is abundant in brain, co-immunoprecipitated from hippocampal extracts (Supplementary Fig. 3).
Does PKMζ regulate GluR2/NSF-dependent AMPAR trafficking during LTP? To disrupt NSF/GluR2 interactions during recordings of long-term synaptic plasticity we tested the effect of a cell-permeable, myristoylated version of pep2m (myr-pep2m) (Garry et al., 2003). We first confirmed the effectiveness of bath applications of myr-pep2m to hippocampal slices by testing the ability of the agent to block AMPAR potentiation by postsynaptic perfusion of PKMζ. As with the intracellular application of pep2m, bath applications of myr-pep2m (10 μM) blocked PKMζ-mediated AMPAR potentiation (Fig. 3A, p < 0.01 between responses at 13 min with PKMζ alone and responses with PKMζ and myr-pep2m; p = 0.58 between responses at 1 and 13 min with PKMζ and myr-pep2m).
We determined whether blocking NSF/GluR2 interactions affects LTP by applying myr-pep2m to the bath and stimulating with a tetanization protocol optimized to produce a relatively rapid onset of the protein synthesis-dependent phase of LTP (Osten et al., 1996; Tsokas et al., 2005), as shown by a rapid decline in potentiated responses in the presence of the protein synthesis inhibitor anisomycin (Supplementary Fig. 4). Myr-pep2m prevented the expression of LTP persistence at 1 hr posttetanization (Fig. 3B, p < 0.0001 between responses in the presence and absence of myr-pep2m 1 hr after tetanization; p = 0.35 between responses in the presence of myr-pep2m before and 1 hr after tetanization). The agent did not significantly affect potentiation within the first 15 min posttetanization, although a trend for a decrease was observed (for the 15 min after tetanization, repeated measures of ANOVA showed no significant effect with respect to drug (saline vs myr-pep2m) (F(1, 16) = 4.37, p = 0.053); time (F(14, 224) = 1.56, p = 0.093); and drug by time interaction (F(14, 224) = 0.71, p = 0.77)). Myr-pep2m also did not affect baseline responses (Fig. 3B, right insert, p = 1.0 between responses 1 min before and 60 min after myr-pep2m), and a scrambled version of the peptide did not affect LTP (Supplementary Fig. 5). Interestingly, the decrement in potentiation was somewhat faster with myr-pep2m than with anisomycin (Supplementary Fig. 4, insert), suggesting that NSF/GluR2 interactions might play a role in both protein synthesis-independent and protein synthesis-dependent phases of LTP.
To confirm that PKMζ mediates NSF/GluR2-dependent AMPAR trafficking during LTP, we examined the trafficking of AMPARs during LTP maintenance that can be observed by biochemical fractionation techniques (Heynen et al., 2000; Grosshans et al., 2002; Whitlock et al., 2006; Williams et al., 2007). One hr after tetanization we isolated synaptosomes from the CA1 regions (Chin et al., 1989) (Fig. 3C; Supplementary Fig. 6) and found that the AMPAR subunits GluR2 and 3 increased in the synaptosomal fraction (p < 0.02), whereas GluR1 subunits decreased (p < 0.01). Myr-pep2m prevented the increases in synaptic GluR2 and 3 (Fig. 3C), demonstrating that these changes in distribution of AMPARs during LTP were dependent on NSF/GluR2 interactions.
We then determined the role of PKMζ in mediating the NSF/GluR2-dependent redistribution of AMPARs in LTP maintenance. We first found that myr-pep2m did not block the increase of PKMζ in the hippocampal slices during LTP maintenance, indicating that blocking NSF/GluR2 interactions did not prevent the induction of PKMζ synthesis (Hernandez et al., 2003; Kelly et al., 2007) (Fig. 3D, p < 0.05, comparing PKMζ levels 1 hr posttetanization to levels in control nontetanized slices, both in the presence and absence of myr-pep2m; p = 0.31 between posttetanization PKMζ levels with and without the agent). Conversely, however, inhibiting PKMζ with the selective, cell-permeable, myristoylated ζ inhibitory peptide (ZIP) prevented both the late phase of LTP, as previously reported (Ling et al., 2002; Sajikumar et al., 2005; Serrano et al., 2005) (data not shown), and blocked the NSF/GluR2-dependent increase in synaptic GluR2/3 (Fig. 3C; a significant decrease in GluR2 was observed). Consistent with the lack of effect of PKMζ on the exocytotic pathway of AMPAR trafficking, the total surface expression of GluR2 measured by biotin-labeling did not increase 1 hr posttetanization (Fig. 3E, left). The total amount of GluR2 (surface and internal) also did not change (Fig. 3E, right). Thus PKMζ mediates the NSF/GluR2-dependent changes in AMPAR distribution during LTP.
The hallmark of PKMζ is its unique role in maintaining the persistence of LTP. Indeed, PKMζ inhibitors are the only agents known to reverse established late-LTP (Ling et al., 2002; Sajikumar et al., 2005; Serrano et al., 2005; Pastalkova et al., 2006). We therefore asked whether blocking the expression mechanism of PKMζ-mediated synaptic potentiation by myr-pep2m could also reverse established LTP. To test for a persistent effect on maintenance, we strongly tetanized one synaptic pathway and then applied myr-pep2m 3 hr posttetanization (Fig. 4A). Myr-pep2m reversed potentiation in the tetanized pathway (p < 0.01 between baseline and 3 hr posttetanization, and between 3 and 5 hr posttetanization; p > 0.05 between baseline and 5 hr posttetanization), but did not affect a second, independent nontetanized pathway simultaneously recorded within each of the slices. The inactive scrambled version of myr-pep2m had no effect (Fig. 4B, p < 0.001 between baseline and 5 hr posttetanization; p > 0.05 between 3 and 5 hr posttetanization; p < 0.0001 between myr-pep2m and scrambled myr-pep2m at 5 hr posttetanization).
PKMζ-mediated late-LTP can also be “captured” by weakly stimulated synapses through the process of synaptic tagging (Frey and Morris, 1997; Sajikumar et al., 2005). Blocking NSF/GluR2-mediated AMPAR trafficking with myr-pep2m reversed persistent potentiation at both the strongly stimulated synapses and the weakly stimulated synapses that underwent synaptic tagging and capture (Supplementary Fig. 7).
Although late-LTP is dependent on new protein synthesis for its induction, the mechanism of expression of synaptic enhancement during late-LTP has previously been unknown. Because PKMζ is synthesized in LTP induction (Hernandez et al., 2003; Muslimov et al., 2004; Kelly et al., 2007) and is then necessary and sufficient for maintaining synaptic potentiation during late-LTP (Ling et al., 2002; Serrano et al., 2005), our approach was first to determine the mechanism by which the kinase enhances AMPAR-mediated synaptic transmission. We found that PKMζ works selectively through upregulation of an NSF/GluR2-dependent AMPAR trafficking pathway to increase postsynaptic AMPARs. We then showed that PKMζ regulates NSF/GluR2-dependent AMPAR trafficking during LTP maintenance, and finally that synaptic potentiation during late-LTP maintenance is expressed through this pathway. Because NSF/GluR2-dependent trafficking serves to maintain a constant number of postsynaptic AMPARs under basal conditions, late-LTP maintenance can be viewed as a persistent resetting by an autonomously active kinase of this homeostatic trafficking mechanism to favor an increased number of receptors at postsynaptic sites (Fig. 5). Future work will be required to determine which substrates of PKMζ mediate this transformation.
The action of PKMζ to enhance synaptic transmission may be through NSF's ability to disrupt the binding of GluR2-containing receptors with PICK1, a trafficking protein that is critical for both hippocampal LTD and LTP (Terashima et al., 2008). A pool of GluR2 appears to be excluded from synapses by binding with PICK1, because disrupting GluR2/PICK1 interactions by pep2-EVKI increases postsynaptic AMPAR responses. Although occlusion experiments are to be interpreted cautiously, PKMζ may act, through NSF, to decrease the size of this extrasynaptic pool of receptors, which are then free to traffic to the synapse. In addition, because PICK1 is critical for endocytosis of AMPARs (Xia et al., 2000; Perez et al., 2001), PKMζ may act to redirect AMPARs away from endocytosis, allowing more to be available for postsynaptic insertion. Interestingly, the AMPAR trafficking driven by PKMζ is opposite in direction to the pathway mediated by PKCα, which directs GRIP/ABP-bound, postsynaptic GluR2-containing receptors away from the synapse to PICK1 and thus to endocytosis during LTD (Xia et al., 2000; Perez et al., 2001). Indeed, we found both PKMζ and PKCα in complexes with PICK1 in hippocampal extracts. Variation in the balance between these PKC isoforms prior to experimentation may set the level of the PICK1-bound extrasynaptic AMPAR pool and thus contribute to the differences in the response to pep-EVKI seen in the literature (Daw et al., 2000; Kim et al., 2001). An alternative hypothesis for PKMζ-mediated potentiation is that the kinase acts to form and maintain a dynamic postsynaptic “slot” (Shi et al., 2001; McCormack et al., 2006), to which the trafficking of AMPARs requires NSF/GluR2 interactions. Directing AMPARs to the synapse and maintaining a structural slot to receive the receptors are not mutually exclusive mechanisms.
During LTP, PKMζ regulates the trafficking of GluR2/3 receptors through NSF/GluR2-dependent interactions, as measured by changes in the amount of the subunits in synaptosomal fractions. These results are generally consistent with the delayed postsynaptic increase in overexpressed GluR2/3 subunits observed during LTP (Shi et al., 2001; McCormack et al., 2006), and with the protein synthesis-dependent increase in GluR2/3 surface expression in synaptoneurosomes during LTP in vivo (Williams et al., 2007). In our study, the increase in synaptosomal GluR2/3 occurs without a change in the total amount of surface GluR2-containing AMPARs, consistent with previous observations (Grosshans et al., 2002), suggesting a lateral redistribution of the receptors from extrasynaptic to synaptic sites. Local exocytosis near the synapse by a non-VAMP/synaptobrevin-mediated process, however, below our level of detection or compensated by endocytosis elsewhere, cannot be excluded (Beretta et al., 2005). Although these crude fractionation techniques provide only limited information as to receptor location, they demonstrate that an NSF/GluR2-dependent redistribution of AMPARs occurs in LTP maintenance and is mediated by PKMζ activity.
This PKMζ-mediated, NSF/GluR2-dependent AMPAR trafficking is functionally critical for the expression of LTP maintenance. Blocking the trafficking reversed late-LTP at either strongly tetanized synapses or at weakly tetanized synapses that had undergone synaptic tagging and capture. These findings relate late-LTP expression to other forms of synaptic plasticity, particularly in the cerebellum, where a critical role for NSF/GluR2 interactions has been observed for both synaptic depression and potentiation (Steinberg et al., 2004; Gardner et al., 2005; Kakegawa and Yuzaki, 2005; Liu and Cull-Candy, 2005). These other forms of synaptic plasticity have been studied in their early phase, when second messenger-dependent protein kinases are typically thought to regulate receptor trafficking as transient mechanisms. Although it is currently unknown whether the persistent increase in PKMζ observed in LTP maintenance (Osten et al., 1996) extends throughout the late-phase of LTP (and long-term memory), the only agents known to reverse established late-LTP are inhibitors of PKMζ kinase activity and now an inhibitor of the AMPAR-trafficking expression mechanism of PKMζ-mediated synaptic potentiation. Thus late-LTP can be viewed as a sustained version of a short-term plasticity mechanism, in which the synthesis of an autonomously active kinase continually drives receptors to postsynaptic sites. This persistent, but dynamic mechanism may allow for stable, yet flexible storage of information at synapses to be the physiological substrate of long-term memory (Pastalkova et al., 2006; Shema et al., 2007).
Supplementary Figure 1. Postsynaptic perfusion of pep2m causes rundown of baseline AMPAR responses recorded at 0.4 Hz test stimulation frequency. AMPAR responses in the presence of the peptide show rundown (red open circles, p < 0.01 between responses at 1 and 13 min in the presence of 100 μM pep2m; p < 0.005 between responses at 13 min with and without pep2m, n = 4). Baseline stimulation without peptide shows no difference between responses at 1 and 13 min (black open circles, p = 0.32, n = 4). Inserts above show representative traces recorded ~1 min (left) and ~13 min (right) after cell breakthrough.
Supplementary Figure 2. Pep-ΔA849-Q853 (100 μM), which blocks GluR2 interactions with AP2, does not affect PKMζ-mediated potentiation of AMPAR responses (n = 4).
Supplementary Figure 3. PICK1 and PKMζ co-immunoprecipitate from hippocampal homogenates. A, PICK1 antiserum immunoprecipitates PKMζ, PKCα, and PICK1, but not PKCε. The amount of protein loaded in the input lane is equivalent to 6% of the amount of protein used for immunoprecipitation and loading in the IP lanes. IPs with (+) and without (−) primary precipitating antisera are shown. B, PKMζ antiserum immunoprecipitates both PICK1 and PKMζ. As can be seen in the input lanes, the immunoblots for the PKMζ IP were developed longer than those for the PICK1 IP. Equivalent results were obtained in 3 separate experiments.
Supplementary Figure 4. The brief tetanization protocol (four 100 Hz, 1 sec trains, 20 sec intertrain interval) produces a relatively rapid onset of a protein synthesis-dependent phase of LTP. In the presence of bath applications of anisomycin (50 μM), significant potentiation was observed at 15 (p < 0.05) but not 60 min (p > 0.05) posttetanization (red filled circles), relative to an independent, unstimulated synaptic pathway recorded in the same slice (black open circles); n = 4. Insert, the return to 100% pretetanus baseline occurs earlier in the presence of myr-pep2m (red open circles, data from Fig. 3B) than with anisomycin (red filled circles) (linear regression of data points beginning 40 min posttetanization showed myr-pep2m returned to baseline at 52.4 ± 3.6 min; anisomycin at 81.3 ± 4.3 min, p < 0.005).
Supplementary Figure 5. Scrambled myr-pep2m had no effect on LTP. A 2-way ANOVA was used to analyze scrambled myr-pep2m (red filled circles) and saline fEPSP responses (black open circles, data from Fig. 3B) collected 1-60 min posttetanization. There was no significant difference between the drug- and saline-treated samples (F(1, 10) = 1.172, p = 0.30).
Supplementary Figure 6. Co-enrichment of postsynaptic proteins in hippocampal synaptosomal fraction. Representative immunoblot showing enrichment of GluR1-3 and PSD-95 in synaptosomal fraction, compared with crude hippocampal homogenate.
Supplementary Figure 7. Blocking NSF/GluR2 interactions reverses LTP maintenance after synaptic tagging. A, Two independent pathways within a hippocampal slice are stimulated: one by strong tetanization (filled circles, filled arrows) and the second 30 min later by weak tetanization (open circles, open arrow). Applications of myr-pep2m (10 μM) 30 min after the weak stimulation reversed synaptic potentiation at these latter tagged synapses, as well as at the strongly stimulated synapses (p < 0.05 for both pathways between baseline and 1 hr and between 1 and 6 hr posttetanization; p > 0.05 between baseline and 6 hr). B, Applications of the scrambled myr-pep2m (10 μM) 30 min after the weak stimulation had no effect on synaptic tagging and capture, with both the weakly and strongly stimulated pathways showing persistent potentiation (p < 0.05 between baseline and 6 hr posttetanization). C, Myr-pep2m reverses persistent potentiation at the strongly stimulated pathway (filled circles) without affecting an independent, nontetanized pathway (open circles) (p < 0.05 between baseline and 1 hr posttetanization, and between 1 and 6 hr posttetanization; p > 0.05 between baseline and 6 hr posttetanization, and for nontetanized pathway between 0 and 6 hr of recording). D, Scrambled myr-pep2m has no effect on either tetanized or nontetanized pathways (p < 0.001 between baseline and 6 hr posttetanization; p > 0.05 between 1 and 6 hr posttetanization; p < 0.0001 between myr-pep2m and scrambled myr-pep2m at 6 hr posttetanization). N = 5 for all experiments.
Supported by NIH RO1 MH53576 and MH57068 (TCS), the American Heart Association grant 051226T (PB), and the German BMBF FKZ-01GW0553 (JUF).