One conceptual mechanism for the induction of associative long-term memory is that a synaptic tag, set by a weak event, can capture plasticity-related proteins from a nearby strong input, thus enabling associativity between the 2 (synaptic tagging and capture, STC). So far, STC has been observed for only a limited time of 60 min. Nevertheless, association of weak memory forms can occur beyond this period and its mechanism is not well understood. Here we report that metaplasticity induced by ryanodine receptor activation or synaptic activation of metabotropic glutamate receptors prolongs the durability of the synaptic tag, thus extending the time window for associative interactions mediating storage of long-term memory. We provide evidence that such metaplasticity alters the mechanisms of STC from a CaMKII-mediated (in non-primed STC) to a protein kinase Mzeta (PKMζ)-mediated process (in primed STC). Thus the association of weak synapses with strong synapses in the “late” stage of associative memory formation occurs only through metaplasticity. The results also reveal that the short-lived, CaMKII-mediated tag may contribute to a mechanism for a fragile form of memory while metaplasticity enables a PKMζ-mediated synaptic tag capable of prolonged interactions that induce a more stable form of memory that is resistant to reversal.
How activated networks of neurons retain new information is a central question in understanding the processes of learning and memory. A wealth of data provides evidence that mechanisms of long-term potentiation (LTP) and long-term depression (LTD) contribute substantially to long-term information storage at the synaptic level (Malenka and Bear 2004). Associative forms of long-term memory may be formed or maintained by the processes of synaptic tagging and capture (STC), a prominent hypothesis that provides a conceptual basis for how short-term forms of plasticity are transformed to long-term plasticity in an associative and time-dependent manner (Frey and Morris 1997; Redondo and Morris 2011). STC has been validated from the single spine level to the behavioral level (Ballarini et al. 2009; Wang et al. 2010; Govindarajan et al. 2011; Moncada et al. 2011). It has also been reported recently that foot shock-induced spatial-memory impairment in rats can be rescued by novelty through the processes of STC (Almaguer-Melian et al. 2012). In addition, key molecules such as CaMKII, PKA, and TrkB are established as candidate synaptic tag mechanisms for LTP, while ERK1/2 is a candidate for LTD (Young et al. 2006; Nie et al. 2007; Sajikumar et al. 2007; Redondo et al. 2010; Ishikawa et al. 2011; Lu et al. 2011; Moncada et al. 2011). Activation of protein kinase Mzeta (PKMζ) by strong activation of a synaptic input is critical for the expression of STC and LTP in a weakly activated input (Sajikumar et al. 2005b). Thus PKMζ acts as an LTP-specific plasticity-related protein (PRP). Recently, CaMKIV and BDNF have been suggested as PRPs for late-LTP and late-LTD, respectively (Redondo and Morris 2011; Sajikumar and Korte 2011b).
Synaptic tag–PRP interactions within a specific time interval are a critical part of the associative interactions in STC, a process also called as late-associativity. Surprisingly, a synaptic tag duration of only about 60 min was proposed for LTP and LTD by the weak-before-strong paradigm (Frey and Morris 1998; Sajikumar and Frey 2004a), providing a temporal limit on the extent of STC-mediated associativity. However, in recent years, metaplasticity, in which the rules of plasticity are changed by the history of neuronal activation, has emerged as a prominent mechanism for more prolonged regulation of network activity in synaptic populations (Abraham 2008). For instance, activation of metabotropic glutamate receptors (mGluRs) or ryanodine receptors (RYRs) exert persistent effects on different forms of plasticity such as LTP and LTD (Manahan-Vaughan and Reymann 1996, 1997; Raymond et al. 2000; Mellentin et al. 2007). Synaptic activation of mGluRs before a weak tetanic stimulus facilitates the persistence of LTP in the area CA1 of rat hippocampal slices by activating protein synthesis machinery from pre-existing mRNA (Raymond et al. 2000). In addition, we have recently reported that metaplasticity processes like RYR activation lowers the threshold for subsequent STC, while mGluR activation alters the plasticity thresholds in apical dendritic compartments of CA1 pyramidal neurons, thus enabling effective interactions between different PRPs within the compartmentalization process of STC (Sajikumar et al. 2009; Sajikumar and Korte 2011b).
Because associativity between weak and strong synapses has so far been observed only over a period of 60 min, we set out to address the question whether late-associativity spanning hours can exist. We hypothesized that such late-associativity can be achieved through metaplasticity. In this study we provide evidence that RYR or mGluR activation substantially increases the synaptic tag duration and enables associativity in the later phase of LTP. Moreover, both RYR and mGluR activation alters the tag-setting process from a CaMKII-mediated one to a PKMζ-mediated one, thus enabling the synapses to express associative memory interactions over a prolonged time period.
Materials and Methods
A total of 177 hippocampal slices prepared from 89 male Wistar rats (6–7 weeks old) were used for electrophysiological recordings. All procedures were approved by guidelines from the Animal Committee on Ethics in the Care and Use of Laboratory Animals of TU-Braunschweig. Briefly, after anesthetization using CO2, the rats were decapitated and the brains were quickly removed and cooled in 4 °C artificial cerebrospinal fluid (ACSF). Transverse hippocampal slices (400 μm) were prepared from the right hippocampus by using a manual tissue chopper, and the slices were incubated at 32 °C in an interface chamber (Scientific System Design; for details, see Sajikumar et al. 2005a). The ACSF contained the following (in mM): 124 NaCl, 4.9 KCl, 1.2 KH2PO4, 2.0 MgSO4, 2.0 CaCl2, 24.6 NaHCO3, 10 d-glucose, equilibrated with 95% O2–5% CO2 (32 L/h). In all experiments, 2 monopolar lacquer-coated, stainless-steel electrodes (5 MΩ; AM Systems, United States of America) were positioned at an adequate distance within the stratum radiatum of the CA1 region for stimulating 2 separate distal synaptic inputs S1 and S2 of 1 neuronal population (Fig. 1A). Pathway independence was tested by the paired-pulse facilitation (PPF) protocol with an interpulse interval of 30 ms as described previously (Sajikumar and Korte 2011a). Facilitation of field excitatory synaptic potentials (fEPSPs) was observed only when 2 consecutive pulses with an interval of 30 ms were applied to the same pathway. When 2 consecutive pulses were applied to different pathways within 30 ms, no cross-facilitation was observed (Sajikumar and Korte 2011a). For recording the fEPSP (measured as its initial slope function), 1 electrode (5 MΩ; AM Systems) was placed in the CA1 apical dendritic layer, and signals were amplified by a differential amplifier (Model 1700, AM Systems). The signals were digitized using a CED 1401 analog-to-digital converter (Cambridge Electronic Design). Slices were preincubated for 3 h, which is critical for late-plasticity studies (Sajikumar et al. 2005a). After the preincubation period, an input–output curve (afferent stimulation vs. fEPSP slope) was plotted prior to experiments. For setting the test stimulus intensity (biphasic constant-current pulses), fEPSP of 40% of its maximal amplitude was determined for both synaptic inputs, S1 and S2. Late-LTP (L-LTP) was induced using 3 stimulus trains of 100 pulses (“strong” tetanus [STET], 100 Hz; duration, 0.2 ms/polarity; intertrain interval, 10 min). Early-LTP (E-LTP) was induced using a weak tetanization protocol consisting of one 100-Hz train (21 biphasic constant-current pulses; pulse duration per half-wave, 0.2 ms) (Sajikumar et al. 2005a, 2007). For inducing depotentiation (DP), low-frequency stimulation was applied 5 min after the induction of primed E-LTP in the same synaptic input using 250 impulses at a frequency of 1 Hz (Sajikumar and Frey 2004b; Sajikumar et al. 2009). The 2 theta-burst stimulation (2xTBS) protocols [separated by 30 s consisted of 2 trains of 10 × 100 Hz bursts (5 biphasic pulses per burst)] with a 200 ms interburst interval, at the test pulse intensity (Raymond et al. 2000). The slopes of the fEPSPs were monitored online. The baseline was recorded for 60 or 90 min for control or priming experiments, respectively. Four 0.2-Hz biphasic constant-current pulses (0.1 ms/polarity) were used for baseline recording and testing at each time point (Sajikumar et al. 2009).
Ryanodine (RYA, 10 μM, dissolved in DMSO; Tocris) was prepared immediately before bath application (Sajikumar et al. 2009). The CaMKII inhibitors KN-62 (1-[NO-bis-1,5-isoquinolinesulfonyl]-N-methyl-l-tyrosyl-4-phenylpiperazine; Calbiochem) and KN-93 (2-[N-(2-hydroxyethyl)]-N-(4-methoxybenzenesulfonyl)amino-N-(4-chlorocinnamyl)-N-methylbenzylamine: Tocris) were dissolved in DMSO as stock solution (1 mM) (Sajikumar et al. 2007, 2009). The stock solution was dissolved in ACSF to reach the concentrations of 5 or 1 µM, respectively, immediately before bath application. Anisomycin and emetine (Sigma) were prepared as concentrated stock solutions in DMSO and diluted in ACSF to obtain a final concentration of 25 or 20 µM, respectively. The final concentration of DMSO was always 0.1%, a concentration that has no effect on basal synaptic transmission (Navakkode et al. 2004). d-2-Amino-phosphonopentanoic acid (AP-5) (Tocris) was used at a concentration of 50 µM (dissolved in ACSF and 0.1% DMSO). AIDA, (R,S)-1-aminoindan-1,5,dicarboxylic acid (Tocris), was used at a concentration of 500 µM dissolved in 0.1% DMSO. The myristoylated pseudosubstrate peptide myr-ZIP (ZIP) (AnaSpec) was prepared in distilled water as stock solution (5 mM). The required volume containing a final concentration of 1 μM was dissolved in ACSF immediately before bath application (Sajikumar et al. 2005b).
Western Blot Analysis of PKMζ
For western blot analysis, we used 5 groups of slices: 1) control slices which were incubated in the interface chamber for 3 h; 2) E-LTP group; 3) RYR-primed E-LTP group; 4) myr-ZIP together with RYR-primed E-LTP group; and 5) Anisomycin together with RYR-primed E-LTP group. In each group, 18 slices were used for the studies, thus a total of 90 slices prepared from thirty 6- to 7-week-old male Wistar rats were used. Tissues around the recording electrodes were cut carefully 2 h after the induction of E-LTP and used for further analyses. PKMζ expression was analyzed according to the methods as described previously (Hernandez et al. 2003; Sajikumar and Korte 2011b). Hippocampal tissue sections were collected in STKM buffer (250 mM saccharose, 50 mM Tris–HCl, pH 7.5, 25 mM KCl, 5 mM MgCl2, 1 μM leupeptin, 1 μM pepstatin A, 0.4 mM 4-(2-aminoethyl)-benzolsulfonylfluoride (AEBSF), 1 μM aprotinin) and lysated by 3 freeze/thaw cycles. After centrifugation for 10 min at 13 000 × g, the protein concentration of the supernatant was determined using Bradford assay. Four micrograms of total protein were subjected to SDS–PAGE and subsequent immunoblotting with antibodies against PKMζ (38–1400; Invitrogen) or tubulin (DM1A; Sigma-Aldrich), respectively. The amount of PKMζ was quantified by densitometric measurement of western blots using EasyWin (Herolab, Germany). The densitrometric values of each blot were normalized to the amounts of tubulin which served as a loading control and were calculated in relation to the control group. The values of each data points were represented as mean of 5 independent experiments.
The average values of the slope function of the fEPSP (mV/ms) per time point were analyzed using the Wilcoxon signed-rank test (Wilcoxon test) when compared within the group, or the Mann–Whitney U test (U test) when data were compared between the groups (Sajikumar and Korte 2011a, 2011b). One-way ANOVA with Dunnett's post hoc tests at the P < 0.05 significance level was used for the analysis of western blot results.
RYR Activation or Synaptic Priming by 2xTBS in E-LTP Results in an Intermediate Form of LTP
We first checked whether bath application of the RYR agonist ryanodine (RYA, 10 µM) has any long-term non-specific effects on the stability of the potentials from 2 independent synaptic inputs S1 and S2. RYA was applied 30 min after a stable baseline in S1 and S2 for the next 30 min. Both synaptic inputs showed stable potentials for up to 6 h (Supplementary Fig. 1). We have shown previously that the induction and persistence of short-term potentiation (STP) can be primed by the previous delivery of RYA (Sajikumar et al. 2009). In the present study, we investigated the priming effect of RYA on classical E-LTP. It is well documented that while longer-lasting than STP, E-LTP still lasts only 2–3 h and is protein synthesis independent (Sajikumar et al. 2005a). In the present control experiment, after recording a stable baseline in both synaptic input S1 and S2, the E-LTP induced by a weak tetanization protocol (WTET) in S1 lasted nearly 210 min, while S2 responses remained stable (Fig. 1B). Statistically significant potentiation was observed in S1 up to 210 min after the induction of E-LTP (U test, P = 0.02, Wilcoxon test, P = 0.01). In the next set of experiments, we probed the priming effect of RYA (10 µM) on E-LTP. A stable baseline of 30 min was recorded, followed by RYA bath application for 30 min. Thirty minutes after the washout of RYA (thus a total of 90 min baseline), E-LTP was induced in S1. As shown in Figure 1C, the induction and persistence of RYA primed E-LTP was significantly increased compared with the control E-LTP, without affecting the baseline in S2 (up to 285 min U-test, P= 0.02, up to 300 min, Wilcoxon test, P= 0.04). Interestingly, primed E-LTP showed statistically significant potentiation during induction and maintenance up to 105 min compared with normal E-LTP (see Fig. 1G for comparison [U test, 60 min (P= 0.003), 90 min (P= 0.004), 105 min (P= 0.02)].
It has been shown previously that LTP can be primed by synaptically released glutamate through Group 1 mGluR-mediated mechanisms (Raymond et al. 2000). To test this, we primed E-LTP by synaptically activating mGluRs through the delivery of 2 TBS (2xTBS) but in the presence of the NMDA receptor antagonist AP-5 (50 μM) to prevent LTP induction and permit selective activation of the mGluRs (Raymond et al. 2000). The AP-5 was bath-applied 30 min after a stable baseline in S1 and S2 for 10 min and then washed out for 20 min before the delivery of WTET for inducing E-LTP. Interestingly, similar to RYA primed E-LTP, we observed an enhanced E-LTP (2xTBS primed E-LTP) lasting 300 min (Fig. 1D, U test, P= 0.01, Wilcoxon test, P= 0.02).
Next we were interested to investigate by which mechanism RYA or 2xTBS priming exerts its effect. Thus we applied PKMζ inhibitor ZIP (1 µM) initially for 30 min and then co-applied either with RYA for 30 min in Figure 1E or with AP-5 for 10 min in Figure 1F. The persistence of RYA or 2xTBS-primed E-LTP in the presence of ZIP was similar to the control E-LTP in non-primed slices, without affecting the baseline in S2. Statistically significant potentiation was observed in S1 up to 210 min in Figure 1E (U test, P= 0.02, Wilcoxon test, P= 0.01) or up to 180 min in Figure 1F (Wilcoxon test, P= 0.03) or 135 min (U test, P= 0.04). A comparison of E-LTP with RYA primed E-LTP and 2xTBS-primed E-LTP is presented in Figure 1G. Thus, RYR activation or 2xTBS priming before the induction of E-LTP results in an intermediate form of LTP, lasting nearly 5 h.
Associative Properties of Primed E-LTP
Weak tetanization in S1 prior to strong tetanization in S2 (the so-called weak-before-strong paradigm) reveals the synaptic tag duration in STC (Frey and Morris 1998; Sajikumar and Frey 2004a) and an effective time window for the tag-PRP interaction of about 60 min. Since RYR priming of E-LTP results in an intermediate form of LTP, it was of interest to test how long the synaptic tag can persist in RYA or 2xTBS-primed E-LTP. To address these questions, we first probed whether associativity can be achieved between weak and strong synapses over a longer interval of 240 min. Consistent with earlier findings, we noticed that L-LTP was not expressed in S1 (weakly activated synapses) when subsequent induction of L-LTP occurred 240 min later in S2 (Supplementary Fig. 2A). We also conducted PPF experiments to prove the used pathways (S1 and S2) were independent (Supplementary Fig. 2B). We next asked whether RYR or 2xTBS-primed E-LTP can take part in the processes of STC with an interval of 240 min. It was interesting to investigate this question because earlier we reported that RYR priming creates a new synaptic tag, and by this means enables associativity (Sajikumar et al. 2009). To address this question, a stable baseline of 30 min was recorded, after which RYA was bath-applied for 30 min and then washed out for a further 30 min (Fig. 2A). E-LTP was induced in the synaptic input S1 30 min after washout of the drug (thus, a total of 90 min baseline). Primed E-LTP decayed to baseline within 240 min (Fig. 2A). Statistically significant potentiation was observed up to 210 min (Wilcoxon test, P= 0.02). Subsequently, 240 min after the induction of E-LTP in S1, L-LTP was induced in S2 by STET. Surprisingly, the potentials in S1 slowly recovered to its initial potentiation level from 275 min onwards, thus expressing STC. Potentials in S1 again showed statistically significant potentiation from 275 min onwards (Wilcoxon test, P= 0.04) to 480 min (Wilcoxon test, P= 0.02), indicating that priming prolongs the STC duration up to at least 4h. It was interesting to test whether RYA priming of STC requires protein synthesis. Thus, we repeated the experiments presented in Figure 2A with protein synthesis inhibitors anisomycin (ANI, 25 µM) or emetine (20 µM) (Supplementary Fig. 2C and D). Consistent with our earlier findings of mGluR-primed LTP (Sajikumar and Korte 2011b), we observed that priming stimulation by RYA also activates protein synthesis to prolong STC.
The experimental design for Figure 2B was similar to that of 2A but with the exception that instead of RYA, 2xTBS was applied in presence of AP-5 (50 µM). Interestingly, similar to Figure 2A, the intermediate form of E-LTP in S1 was transformed to L-LTP showing STC (Fig. 2B). To this end, we checked whether the synaptic stimulation specifically activates mGluRs in our experimental conditions, similar to that of an earlier report (Raymond et al. 2000). Thus, we repeated the experiments depicted in Figure 2B but with the exception that 2xTBS was delivered in the presence of the Group 1 mGluR antagonist AIDA (500 μM; dark bar), in combination with AP-5 (Supplementary Fig. 2E). In agreement with the earlier report (Raymond et al. 2000), we observed that mGluRs are required for priming by 2xTBS stimulation because priming stimulation in the presence of AIDA prevented the enhancement of E-LTP (Supplementary Fig. 2E) and its subsequent capture process. Co-application of ZIP (1 µM) either with RYA or with 2xTBS abolished the primed STC (Fig. 2C and D). In Figure 2C, S1 showed significant potentiation up to 270 min (Wilcoxon test, P= 0.01) or up to 225 min (U test, P= 0.01). In Figure 2D, S1 showed significant potentiation up to 195 min (Wilcoxon test, P= 0.01, U test, P= 0.04). Potentials in S2 showed significant increases from the time of tetanization until 480 min later (both Fig. 2C and D Wilcoxon test, P= 0.01)
In a critical next series of experiments, we increased the time interval of induction of L-LTP from 240 to 300 min or 360 min. At the 300 min interval, the potentials in S1 decayed to baseline within 240 min but regained to original potentiated levels from 325 min onwards (significant potentiation from 325 to 540 min; Wilcoxon test, P= 0.01, Fig. 2E) but at an interval of 360 min (Fig. 2F) the primed E-LTP failed to show STC. Statistically significant potentiation up to 270 min was observed in primed E-LTP (U test, P= 0.03, Wilcoxon test, P= 0.04). Post-tetanization potentials in S2 showed significant potentiation compared with their pre-potentiation values at all time points in all experiments (Fig. 2E and F, Wilcoxon test, P= 0.02). Taken together, these experiments reveal a very interesting aspect of STC: RYR or 2xTBS priming prolongs the duration of the synaptic tags from 1 to 5 h.
RYR Priming Alters the Mechanisms of Synaptic Tag-Setting Process
In normal STC, setting of a synaptic tag is mediated by CaMKII (Sajikumar et al. 2007; Redondo et al. 2010). This tag can last only 60 min, after which the tag degrades probably due to dephosphorylation. So, the question is, by what means can RYR-primed E-LTP maintain its synaptic tag for a prolonged duration without degradation? For example, does priming of RYR change the molecular identity of the synaptic tag that enables the tag to persist for a longer period? To address these questions, we co-applied an inhibitor of CaMKII, KN-62 (5 µM) along with RYA for 30 min and then KN-62 alone for next 60 min. Thirty minutes after RYA application, E-LTP was induced in S1 in the presence of KN-62. We have earlier reported that KN-62 at this specific concentration can effectively reset the synaptic tag (Sajikumar et al. 2007). The induction of primed E-LTP was significantly decreased in the presence of a CaMKII inhibitor, KN-62. Surprisingly, CaMKII inhibition had no effect on primed STC (Fig. 3A), in contrast to its effectiveness in blocking STC under non-primed conditions (Sajikumar et al. 2007; Redondo et al. 2010). Primed E-LTP showed statistically significant potentiation initially up to 220 min (Wilcoxon test, P= 0.04), and the potentials again regained to a statistically significant level after the induction of L-LTP in S2 (from 285 min onwards Wilcoxon test, P= 0.02). The next main question was to check whether CaMKII dominates in the tag-setting during the early phase of STC in the RYA primed condition, similar to that of normal STC. Thus, we used a similar experimental design that of Figure 3A but with the exception that STET was delivered to S2 at 60 min for inducing L-LTP. Interestingly, CaMKII inhibition did not prevent STC because the E-LTP was transformed to L-LTP expressing STC (Fig. 3B). The E-LTP was statistically significant for an initial 30 min (Wilcoxon test, P= 0.04) and then again from 90 min (Wilcoxon test, P= 0.04) until 360 min (Wilcoxon test, P= 0.01). It is noteworthy that the observed effect of KN-62 may not be specific to CaMKII because it can also inhibit CaMKIV and voltage-dependent K+ channels (Ledoux et al. 1999). To ensure specificity, we confirmed these results with another CaMKII inhibitor KN-93 (1 µM) (Supplementary Fig. 3A and B) which has been shown to prevent the setting of CaMKII-mediated synaptic tag at this concentration by an earlier report (Redondo et al. 2010). If CaMKII does not mediate the synaptic tag in primed E-LTP, what other molecular mechanism becomes involved? We started our mechanistic approach by studying the role of PKMζ in mediating the setting of the synaptic tag because recently we showed that mGluR-dependent priming induces PKMζ as a PRP for STC (Sajikumar and Korte 2011b). We used the same experimental design as that of Figure 3A but instead of KN-62, we used PKMζ inhibitor ZIP. PKMζ inhibition prevented primed E-LTP from expressing late-STC (Fig. 3C). Statistically significant potentiation was observed only up to 240 min in S1 (Wilcoxon test, P = 0.03), but no significant potentiation after the STET to S2. To check whether PKMζ acts as a PRP as well, we bath-applied ZIP from 270 min onwards until the end of the experiment. PKMζ inhibition not only prevented the capture processes in S1, but also deteriorated the potentiation of L-LTP in S2 (Fig. 3D).
In addition to the electrophysiological experiments, quantification of PKMζ in non-primed and primed groups provided further evidence that PKMζ mediates setting of synaptic tag in primed E-LTP. Western blot analysis revealed an increased expression of PKMζ 2 h following induction of RYA primed E-LTP (Fig. 3E, Group 2) compared with either control (Group C), E-LTP without RYA priming (Group 1) or RYA primed E-LTP but protein synthesis-inhibited group with anisomycin (Group 4). There was also a significantly increased amount of PKMζ in the RYA-primed Group 2 in comparison to Group 1 (P = 0.01), and Group 4 (P = 0.01) (Fig. 3F). Although the application of ZIP together with RYA during priming inhibits PKMζ function as shown in Figure 3C, it had no effect on the expression rate of PKMζ (group 3, Fig. 3E and F, no statistically significant increase in the expression of PKMζ compared with Group 2, P > 0.05) similar to our earlier report (Sajikumar and Korte 2011b). Thus, RYA priming alters the CaMKII-mediated tag-setting process to a PKMζ-mediated tag-setting process.
RYR Priming or 2xTBS Priming Creates Long-Lasting Synaptic Tags
We have reported earlier that DP 5 min after the induction of E-LTP can reset the synaptic tag mediated by CaMKII and prevent subsequent STC (Sajikumar and Frey 2004b; Sajikumar et al. 2009). Thus, we tested whether DP 5 min after primed E-LTP also affects primed E-LTP and subsequent STC. Five minutes after RYA or 2xTBS primed E-LTP (Fig. 4A and B), DP was applied to the same input. Surprisingly, and in contrast to our earlier findings in normal STC (Sajikumar and Frey 2004b), normally depotentiating 1 Hz stimulation resulted in only a transient depression for 10 min (Fig. 4A and B). The fEPSPs recovered to the initial potentiation level and slowly decayed to baseline potentials similar to that of primed E-LTP. After the induction of L-LTP in S2 at 240 min (Fig. 4A and B), the potentials in S1 regained to a stable L-LTP lasting 8 h. Next we checked whether DP stimulation during PKMζ inhibition could reset the synaptic tag and thus interfere with STC. To test this idea, we bath-applied ZIP for the next 1 h immediately after RYA or AP-5 washout (Fig. 4C and D). WTET and DP were delivered in the presence of ZIP (Fig. 4C and D, open circles). Interestingly DP completely prevented E-LTP in S1 and in addition no STC was observed after the delivery of STET in S2. The primed E-LTP was statistically significant for an initial 5 min (U test, Wilcoxon test, P= 0.01) and the L-LTP showed statistically significant potentiation from 240 min onwards until 480 min (Wilcoxon test, P= 0.01 in both Fig. 4C and D). RYA- and 2xTBS-primed E-LTP are resistant to DP and its effects on late associativity, while PKMζ inhibition during DP blocked both of the priming effects, thus restoring the sensitivity of the LTP and STC to DP.
Metaplasticity is an emerging mechanism that has been proposed to govern different aspects of functional plasticity (Abraham 2008; Sajikumar and Korte 2011b). Metaplasticity not only creates new synaptic tags but also regulates the function of different plasticity factors such as BDNF and PKMζ in distal dendritic clusters (Sajikumar et al. 2009; Sajikumar and Korte 2011b). Normally the associative interaction in STC is expressed and restricted to the early phase of LTP, i.e. up to a time period of 60 min (Redondo and Morris 2011), which limits the ability of synaptic populations to integrate information beyond this period. This could be due to the fragile nature of the synaptic tag marked during a weak synaptic potentiation event. Extending the time interval for the interaction of weak and strong events in a synaptic population can result in a more precise consolidation process, especially for tuning the synapses to promote or prevent long-term memory storage (Abraham 2008). Our earlier findings provided compelling evidence that activation of RYRs before the induction of STP is capable of creating new synaptic tags for functional plasticity (Sajikumar et al. 2009). Consistent with these findings, our present data provide evidence that RYR activation before the induction of E-LTP can extend the duration of synaptic tag from the normal 1 h to at least 5 h, so that an associative integration is possible throughout this time period. In addition to the prolongation of the functionality of the synaptic tag, RYR priming also increased the persistence of E-LTP, similar to that of the intermediate form of LTP dependent on Group 1 mGluR activation (Raymond et al., 2000). It is to be noted that the setting of synaptic tag is dependent on the type and area of stimulation. For instance, a very weak stimulation that induces STP is unable to mark synaptic tags in distal dendrites (Sajikumar et al. 2009) but does so in basal dendrites (Sajikumar et al. 2007). Interestingly, activation of RYR before the induction of STP is able to create synaptic tags, but these tags survive only for a shorter duration (Sajikumar et al. 2009) than the RYR primed E-LTP in the present study.
The synaptic activation of mGluRs by 2xTBS also revealed a similar result to that of the RYR priming experiments, and thus both forms of mGluR activation share similar subcellular mechanisms for the metaplasticity of E-LTP (see below). E-LTP induction following either RYR activation or mGluR activation produced PKMζ as a PRP, through either RYR activation or through mGluR activation. These results confirm and extend the earlier reports by Raymond et al. (2000) that Group 1 mGluR activation couples to nearby protein synthesis machinery from pre-existing mRNA (here PKMζ mRNA) (Raymond et al. 2000; Sajikumar and Korte 2011b). In general, these results give motivation for exploring whether priming can be used to expand the temporal interval of associations during behavior.
Metaplasticity Alters the Mechanisms of STC
It has been reported that different concentrations of RYR agonists such as ryanodine or caffeine have different effects on ventral and dorsal hippocampus and these differences are due to the different distribution of RYR's (Grigoryan et al. 2012). It is well established that RYR activation amplifies activity-dependent calcium influx via calcium-induced calcium release leading to the activation of different subcellular signaling pathways. These pathways activate CaMKII, PKA, PKC, mTOR and ERK1/2 (Abraham 2008). Large elevations of intracellular calcium, either via very high-frequency stimulation or indeed via RYR, are known to induce NMDAR-independent LTP (Wang et al. 1996; Malenka and Bear 2004; Raymond 2008). However, the RYA-primed E-LTP in our case was NMDAR dependent (Supplementary Fig. 4), an important prerequisite for STC (O'Carroll and Morris 2004). Indeed, it has been reported recently that RYR activation by intrahippocampal injection of ryanodine in rats improved spatial learning and enhanced memory consolidation by specifically increasing the RYR subtypes such as RYR2 and RYR3, plus BDNF and PKMζ (Adasme et al. 2011). These results are consistent with our results presented here and also our earlier findings that RYA primes LTP (Mellentin et al. 2007) and that PKMζ is a critical PRP generated during mGluR-mediated metaplasticity (Sajikumar and Korte 2011b). Interestingly, CaMKII mediates STC in RYR-primed synapses that express an equivalent to E-LTP (Sajikumar et al. 2009). CaMKII also mediates the setting of synaptic tags necessary for behavioral tagging (Moncada et al. 2011). A plasticity-related extracellular protease, neuropsin, which is important for STC at apical dendrites also acts via CaMKII signaling (Ishikawa et al. 2008, 2011). In contrast to the tag-setting process of normal STC, the present data indicate that synaptic tagging in RYR or synaptic activity primed E-LTP is altered from a CaMKII-mediated process to a PKMζ-mediated process through the processes of metaplasticity. This finding might explain how the “life-time” of a synaptic tag is extended in primed E-LTP, as PKMζ is a persistently active kinase that can maintain plasticity up to hours and weeks, much longer than the activation of CaMKII (Sacktor 2011).
But how does PKMζ extend the duration of synaptic tags in RYR primed E-LTP? It may act by preventing the degradation of the CaMKII-mediated synaptic tag by protecting it from dephosphorylation, or it may act by replacing CaMKII from the tag-setting process with PKMζ in tagged synapses. If PKMζ protects the CaMKII-mediated tag from degradation, tagging should still be blocked by a CaMKII inhibitor (Sajikumar et al. 2007). However, even in the CaMKII-inhibited situation, the primed E-LTP expressed STC both at an early tagging period of 1 h and a late tagging interval of 4 h, thus confirming the possibility of a switch from the CaMKII-mediated tag-setting process. On the other hand, our second assumption is supported by the finding that PKMζ inhibition during the priming process, or during the induction of E-LTP, prevents STC. Thus, consistent with our earlier findings on LTP in stratum oriens (Sajikumar et al. 2007), we confirmed here that in certain conditions synaptic tagging can be mediated by PKMζ.
Overall, we propose that the normal CaMKII-mediated tag setting is a “short-lived tag setting process” enabling the associative process for only a limited time period, while PKMζ-mediated tag setting is a “long-lived tag setting process” that extends the associativity for a longer period of time (Fig. 5). This proposal is consistent with the recent finding that CaMKII signals through TrkB for synaptic tagging, but that this is a fragile process lasting only about 60 min (Lu et al. 2011). It is also consistent with our earlier findings that DP 5 min after the induction of E-LTP resets the synaptic tag and interferes with the expression of STC (Sajikumar and Frey 2004b; Sajikumar et al. 2009). Moreover, we showed here that primed E-LTP was resistant to DP and its effects on STC in primed E-LTP. In addition, the tag-resetting experiments during the inhibition of PKMζ resulted in a complete DP of primed E-LTP, thus strengthening our observation that metaplasticity by RYR or synaptic activity priming promotes a stable tag-setting process through PKMζ-mediated mechanisms.
We have reported recently that metaplasticity elicited by mGluR activation before the induction of a local form of LTP can generate locally synthesized PKMζ that acts as a PRP for functional plasticity (Sajikumar and Korte 2011b). Our present findings extends these results by adding the notion that RYR priming can also generate PKMζ as a new plasticity factor that can mediate the setting of a synaptic tag in primed E-LTP. The biochemical evidence supports this interpretation. It has been suggested earlier that local translational processes can provide input-specific synaptic immunity against synaptic tag degradation via DP (Woo and Nguyen 2003; Sajikumar and Frey 2004b). This suggests that the generation of PKMζ by local protein synthesis caused by RYR or mGluR priming might prevent the degradation of the synaptic tag. By this means, the PKMζ-mediated synaptic tag stays intact up to 4–5 h. As we observed that a time window of 6 h was ineffective for STC, this may be due to degradation of the PKMζ-mediated tag by this late time point. It remains to be determined whether newly transcribed PKMζ can rescue synaptic tags from degradation and further prolong the STC window. Nevertheless, metaplasticity can generate stable “synaptic tags” that enable the coding of memory engrams for an extended period of time, allowing associativity in the “late” stage of LTP (Fig. 5).
PKMζ: a Mediator of Synaptic Tags and a PRP
How does PKMζ enable associativity for an extended period during STC? PKMζ promotes synaptic strengthening by releasing α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs) from an extrasynaptic pool and enhances N-ethylmaleimide-sensitive factor/glutamate receptor-2 (NSF/GluR2)-mediated trafficking (Yao et al. 2008; Migues et al. 2010). Recently, a model of “PKMζ-synaptic autotagging” has been proposed for the formation of long-term memory at the cellular level (Sacktor 2011). According to this model, PKMζ in tagged synapses phosphorylates a substrate, possibly the GluR2 C-terminal or its associated proteins, resulting in the release of the receptors from protein interacting with C kinase-1 (PICK1) by NSF. This enables the redistribution of the extrasynaptic AMPAR to postsynaptic sites. The increased amount of GluR2 at the potentiated synapse acts as a “tag” that captures the PKMζ–PICK1 complex. Our present findings are also in line with this prediction, since PKMζ generated due to the processes of metaplasticity alters the normal tag-setting process.
We have reported recently that STC is compartment restricted in distinct dendritic branches and that each of the compartments further contains “synaptic units” or “clusters” with different plasticity thresholds (Sajikumar and Korte 2011b). Metaplasticity can govern the compartmentalization processes of STC by altering the plasticity thresholds of “synaptic units” or “clusters” by providing new PRPs like PKMζ and regulating PRPs like BDNF in STC. Our present findings support the “synaptic unit” model for the following reasons. 1) Metaplasticity by RYR or mGluR's (through synaptic activity) activation alter the normal tag setting process and 2) the PKMζ-mediated tag-setting process resulted in altered plasticity thresholds for expressing STC for a longer period of time. In short, these findings suggest that the “PKMζ-mediated synaptic tag” can capture PKMζ–PRP generated due to the strong activity of the nearby synapses and by this means mediate associativity in the late-phase of LTP. Thus PKMζ acts as a molecule for mediating a synaptic tag as well as a PRP for an extended period of associativity.
Outlook and Conclusion
The most interesting aspect of the whole family of metaplasticity phenomena is that they clearly demonstrate that neural circuits are capable of associating events at one point in time with a much later strong stimulus/event. This greatly expands temporally a network's capacity for associating stimuli. Here we provide the first evidence for how activated neuronal networks can initiate and maintain memory for an extended period of time. Our findings suggest that the initial fingerprint of memory, the “synaptic tag,” is a dynamic molecular complex that can alter its mechanisms based on previous neuronal activity. We propose that metaplasticity can tune activated neuronal networks for coding stable memory engrams for an extended period of time by switching the synaptic tag from a “fragile” state to a more “stable” state. The present work demonstrates that, through metaplasticity, the capacity of a neural network for encoding memory can be extended even further than previously understood. The relation between these synaptic plasticity phenomena and memory phenomena remains an important experimental issue for further investigation.
Supplementary material can be found at: http://www.cercor.oxfordjournals.org/
This work was supported by DFG (SA 1853/1-1 to S.S. and M.K.). S.S. was in addition supported by an Alexander von Humboldt Fellowship. W.C.A. was supported by the Health Research Council of New Zealand and the New Zealand Marsden Fund. Z.C.X was supported by Monash Professorial Fellowship and the talent program of Yunnan Province. Q.L was supported by DAAD fellowship (A/09/98265). A part of this study was also supported by new investigator grant from the National Medical Research Council (NMRC/NIG/0021/2008) of Singapore to S.S.
We are grateful to Dr. Sheeja Navakkode for her helpful discussions and to Tania Meßerschmidt and Rheinhard Huwe for their excellent technical assistance. Conflict of Interest: None declared.