Virtually all CNS synapses display the potential for activity-dependent long-term potentiation (LTP) and/or long-term depression (LTD). Intriguingly, the potential to exhibit LTP or LTD at many central synapses itself is powerfully modulated by previous synaptic activity. This higher-order form of plasticity has been termed metaplasticity. Here, we show that inhibitory autophosphorylation of Ca2+/calmodulin-dependent kinase II (CaMKII) is required for hippocampal metaplasticity at the lateral perforant path-dentate granule cell synapse. Brief 10 Hz priming, which does not affect basal synaptic transmission, caused a dramatic, pathway-specific and long-lasting (up to 18 h) reduction in subsequently evoked LTP at lateral perforant path synapses. In contrast, LTD was unaffected by priming. The induction of lateral perforant path metaplasticity required the activation of NMDA receptors during priming. In addition, metaplasticity was absent in knock-in mice expressing αCaMKII that cannot undergo inhibitory phosphorylation, indicating that inhibitory autophosphorylation of αCaMKII at threonines 305/306 is required for metaplasticity. Metaplasticity was not observed in the medial perforant pathway, consistent with the observation that CaMKII activity was not required for the induction of LTP at this synapse. Thus, modulation of αCaMKII activity via autophosphorylation at Thr305/Thr306 is a key mechanism for metaplasticity that may be of importance in the integration of temporally separated episodes of activity.
The ability to modify synaptic strength in an activity-dependent manner, either as long-term depression (LTD) or long-term potentiation (LTP), is a fundamental feature of most CNS synapses. A less well studied but particularly intriguing finding is that the capacity of many synapses for plastic changes itself is subject to considerable activity-dependent variation, or plasticity. This higher-order plasticity has been termed metaplasticity (Deisseroth et al., 1995; Abraham and Bear, 1996). Importantly, induction of metaplasticity does not necessarily lead to changes in synaptic strength per se but rather modifies the ability of synapses to exhibit synaptic plasticity in response to subsequent episodes of synaptic activity. For instance, a number of studies have demonstrated an inhibition of LTP after priming stimulation with frequencies ranging from 5 to 30 Hz (Huang et al., 1992; O'Dell and Kandel, 1994; Christie et al., 1995; Wang and Wagner, 1999). Stimulation within the same frequency range was also shown to enhance subsequent LTD (Christie and Abraham, 1992). Conversely, a facilitation of LTP after metabotropic glutamate receptor (mGluR) activation has been described (Bortolotto et al., 1994; Manahan-Vaughan and Reymann, 1996; Cohen et al., 1998). It was proposed on theoretical grounds that the expression of metaplasticity may consist of an activity-dependent shift of the transition threshold θm between LTD and LTP toward higher or lower stimulation intensities (Bienenstock et al., 1982), an idea for which experimental support was subsequently provided (Bear et al., 1987). Metaplasticity has been suggested to play an important role in normal neuronal function to keep synapses within a dynamic functional range, thus preventing them from entering states of saturated LTP or LTD (Abraham and Tate, 1997). It is particularly interesting to note that facilitation of LTD can also be induced by an alteration of behavioral state (Xu et al., 1998), indicating that metaplastic processes are invoked and may be important during normal behavior.
Despite the potential importance of metaplasticity, not much is known about the underlying cellular and molecular mechanisms. Several lines of evidence indicate that α Ca2+/calmodulin-dependent kinase II (αCaMKII) may be a promising candidate to mediate metaplasticity. After activation of αCaMKII by Ca2+ influx, activation is maintained by autophosphorylation at threonine 286, even after intracellular Ca2+ has returned to baseline levels (for review, see Lisman et al., 2002). This autophosphorylation has been shown to be required for LTP induction and spatial learning (Giese et al., 1998). Autophosphorylation at Thr305/306, conversely, interferes with the binding of calcium/calmodulin, thereby blocking the activation of CaMKII activity in vitro (Colbran and Soderling, 1990; Patton et al., 1990; Mukherji and Soderling, 1994). In addition, autophosphorylation at Thr305/306 was shown to regulate αCaMKII association with the postsynaptic density (PSD), which is essential for controlling hippocampal LTP and learning (Elgersma et al., 2002). Thus, CaMKII is bidirectionally regulated by activity and, in turn, profoundly affects synaptic plasticity. Thus, it appears likely that CaMKII could be a pivotal component in the bidirectional modulation of synaptic efficacy. In the present study, we characterize metaplasticity at the perforant path-dentate granule cell synapse and demonstrate a role for αCaMKII inhibitory autophosphorylation in this phenomenon.
Materials and Methods
αCaMKII mutant mice. The αCaMKII TT305/306VA point mutant has been described previously (Elgersma et al., 2002). Mutant mice were backcrossed 11 times with C57BL/6J. These mice were then crossed once with 129/Sv (Charles River, Maastricht, The Netherlands). Heterozygous offspring of these mice were crossed with each other to obtain homozygous mutants and wild-type littermates in the F2 hybrid C57BL/6/129Sv background. Adult mice (10-14 weeks) were used for slice electrophysiology as described below.
Slice preparation. Adult male Wistar rats (30-40 d old; 80-150 g) or mice (see above) were anesthetized with diethylether and decapitated. The brain was removed and immediately immersed in ice-cold carbogenated (95%O2/5%CO2) dissection saline that contained the following (in mm): 125.0 NaCl, 3.0 KCl, 0.2 CaCl2, 5.0 MgSO4, 1.25 NaH2PO4, 26.0 NaHCO3, and 13.0 d-glucose. Transverse slices (400 μm) were prepared with a vibratome (VT1000S; Leica, Wetzlar, Germany) and transferred to an interface chamber in which they were continuously superfused (1.8 ml/min) with artificial CSF (ACSF) containing the following (in mm): 125.0 NaCl, 3.0 KCl, 2.5 CaCl2, 1.3 MgCl2, 1.25 NaH2PO4, 26.0 NaHCO3, and 13.0 d-glucose, pH 7.35 (osmolality 300-308 mOsm/kg in all solutions). The temperature of the recording chamber was maintained at 35°C. The slices were allowed to recover for at least 1 h before starting recordings.
Electrophysiology. Extracellular borosilicate glass recording microelectrodes (∼2 MΩ) were filled with extracellular ACSF. The recording electrodes were placed within the outer and middle molecular layer for recording of lateral and medial perforant path field EPSPs (fEPSPs), respectively. Stimulation of the lateral and medial aspects of the perforant path was performed by placing bipolar stimulating electrodes within the outer and the middle molecular layer, respectively. The electrode positions were adjusted so that stimulation of the lateral perforant path caused a current sink in the outer molecular layer and a source in the middle molecular layer and vice versa for stimulation of the medial perforant path (see Fig. 2 A). As an additional criterion, lateral perforant path responses showed paired-pulse facilitation, whereas medial perforant path responses displayed paired-pulse inhibition (40 ms interpulse interval). This configuration was usually readily obtained. In some experiments, the identity of lateral perforant path recordings was confirmed pharmacologically by applying the selective group III mGluR agonist (l)-2-amino-2-methyl-4-phosphonobutyrate (l-AP-4) (10 μm). Stimulation was performed with square impulses of 50-200 μA lasting 0.1 ms via a stimulus isolator (World Precision Instruments, Sarasota, FL). Data were digitized at 10 kHz and analyzed with pClamp 8 software (Molecular Devices, Foster City, CA). Stimulus response curves were performed at the beginning of each experiment. The stimulus strength was then adjusted to yield fEPSPs that were 40-60% of the maximum response, and stimuli were applied at a frequency of 0.025 Hz for the remainder of the experiment. In some experiments, antidromic stimulation was performed by placing a stimulation electrode within the dentate hilar region.
Priming and conditioning stimuli. Priming stimulation in vitro was performed at 10 Hz (10 episodes of 20 pulses, spaced 1 s apart) in all homosynaptic, heterosynaptic, and antidromic priming experiments. Subsequent conditioning stimulation was performed with the same number of stimuli and spacing of episodes at frequencies ranging from 1 to 400 Hz. Both priming and conditioning stimulation were performed with a stimulation strength increased twofold compared with the baseline stimulation strength. In those cases in which priming stimulation caused a significant change of fEPSP amplitude, the stimulation strength was adjusted to prepriming baseline levels during the last 20 min before applying a conditioning tetanus.
Data analysis. To analyze synaptic efficacy, the maximal slope was determined in the rising phase of fEPSPs. Values from three successive fEPSPs were averaged to yield one data point. To determine the amount of LTP or LTD in the different experimental groups, we first determined as the average value of the fEPSP slope of the three last fEPSPs during the follow-up period, i.e., 60 min after conditioning stimulation. All values were normalized to the average fEPSP slope of the first 10 min of the baseline recording period, such that values >1 indicate potentiation and values <1 indicate depression. This value was then related to the average fEPSP slope during baseline recording. All data are expressed as mean ± SEM. Statistical comparisons were performed using Mann-Whitney U test with p < 0.05 as the threshold for statistical significance.
Drugs and reagents. All salts were obtained from Sigma (Deisenhofen, Germany). d(-)-2-amino-5-phosphonovaleric acid (APV) and l-AP-4 were purchased from Tocris Cookson (Bristol, UK), and R-(-)-2-amino-3-hydroxypropionic acid was obtained from Sigma. The latter drugs were dissolved in double-distilled water as a stock solution and then diluted in ACSF to a final concentration before use. KN93 (2-[N-(2-hydroxyethyl)-N-(4-methoxybenzenesulfonyl)]amino-N-(4-chloro-cinnamyl)-N-methylbenzylamine) was obtained from Merck (Darmstadt, Germany) and directly dissolved in ACSF.
Electrode implantation and in vivo priming. In some experiments, priming stimulation was applied in vivo to electrodes implanted into the lateral perforant path (see below, electrode implantation), with a subsequent analysis of synaptic plasticity in vitro. Seven to 8-week-old male Wistar rats underwent electrode implantation into the dentate gyrus as described previously (Kulla et al., 1999). Briefly, under sodium pentobarbitone anesthesia (Nembutal, 40 mg/kg, i.p., Serva Feinbiochemica, Heidelberg, Germany), animals underwent implantation of a monopolar recording electrode into the dentate gyrus granule cell layer and of a bipolar stimulating electrode into the lateral perforant path. Electrodes were made from 0.1-mm-diameter Teflon-coated stainless steel wire. A drill hole was made (1.0 mm diameter) for the recording electrode (3.1 mm posterior to bregma, 1.9 mm lateral to the midline), and a second drill hole (1 mm diameter, 7.9 mm posterior to bregma, and 5 mm lateral to the midline) was made for the stimulating electrode. The dura was pierced through both holes, and the recording and stimulating electrodes were lowered into the dentate gyrus granule cell layer and the lateral perforant path, respectively. Recordings of evoked field potentials via the implanted electrodes were taken throughout surgery. As a criterion for the lateral perforant path origin of the fEPSPs, we verified that double-pulse stimulation (interpulse interval of 40 ms) caused paired-pulse facilitation. Only animals in which clear paired-pulse facilitation was observed were used for the present experiments. Once verification of the location of the electrodes was complete, the entire assembly was sealed and fixed to the skull with dental acrylic (Paladur; Heraeus Kulzer, Mainz, Germany). The animals were allowed between 7 and 10 d to recover from surgery before experiments commenced.
For application of priming stimuli, electrodes were connected to an amplifier and stimulus isolator via a flexible cable in all animals, and the baseline stimulation strength was adjusted as described for the in vitro experiments. In one group of animals, priming stimulation with a protocol identical to the one described above (see above, Priming and conditioning stimuli) at a simulation strength increased twofold compared with baseline was applied (primed group). A control group of animals was handled identically, but no priming stimulus was applied. In all in vivo experiments, we verified immediately before application of priming that clear paired-pulse facilitation was observed in the stimulated pathway. On average, the ratio of the second to the first EPSP amplitude was 1.29 ± 0.01. Experiments were consistently conducted at the same time of day (commencing at 8:00 A.M.). Animals were maintained on a 12 h light/dark cycle. When slices were prepared from these animals, care was taken to visualize electrode position during preparation. Ipsilaterally to the side of electrode placement, the slices immediately adjacent to the one containing the electrode track were chosen for recording. Cutting slices comprising both hemispheres allowed us to select exactly corresponding slices on the contralateral side. In an additional group of animals, the effects of priming in vivo on lateral and medial perforant path EPSPs measured subsequently in vivo for up to 24 h was tested to exclude that priming stimulation in vivo induced long-lasting changes in synaptic strength.
Frequency dependence of synaptic plasticity at the lateral perforant path
It is well established that the amount of synaptic potentiation or depression varies as a nonmonotonic function with the strength of synaptic activation at many synapses. We have therefore initially examined the amount of synaptic plasticity elicited with conditioning stimulation of varying frequencies at the lateral perforant path (10 trains of 20 pulses spaced 1 s apart, frequencies of pulses within trains ranging from 1 to 400 Hz). Similar to previous studies, we found that low frequencies of 1 and 3 Hz induced LTD [3 Hz (square symbols in Fig. 1A), 0.79 ± 0.09 normalized fEPSP slope; n = 8], whereas higher frequencies resulted in expression of LTP [100 Hz (Fig. 1A, triangles), normalized fEPSP slope 1.55 ± 0.17; n = 8]. Conditioning stimulation with 10 Hz caused only low levels of potentiation (normalized fEPSP slope, 1.19 ± 0.16; n = 11), indicating that the crossover point between LTD and LTP is close to this stimulation frequency. The results obtained with the whole range of different conditioning frequencies are shown in Figure 1C (open symbols, numbers of experiments shown for each data point, significant differences to baseline EPSP slope indicated by asterisks).
Homosynaptic priming inhibits subsequent LTP at the lateral perforant path
We next addressed the question whether priming stimulation alters synaptic plasticity at the lateral perforant path. To this end, we used a stimulation protocol identical to those used in Figure 1A, at a frequency of 10 Hz. We chose this protocol because it did not cause significant long-term changes in synaptic transmission. After establishing a stable baseline, a 10 Hz priming stimulation was applied (Fig. 1B, open arrow). Forty minutes after priming, we tested the potential of the lateral perforant path to exhibit synaptic plasticity using the same range of conditioning stimulations as in Figure 1A. Clearly, priming stimulation caused a dramatic loss in the ability of lateral perforant path synapses to undergo potentiation. As shown in Figure 1B, conditioning stimulations at 100 Hz did not induce long-term potentiation (normalized fEPSP slope 1 h after tetanization, 0.91 ± 0.08; n = 8). The stimulation frequencies of 3 and 10 Hz caused depression (normalized fEPSP slope, 0.84 ± 0.03, n = 7 and 0.8 ± 0.08, n = 8, respectively). So far, these results indicate a dramatic loss of LTP, with LTD at low stimulation frequencies remaining intact (summarized in Fig. 1C, significant differences to baseline fEPSP slope indicated by asterisks). Significant differences between primed and control groups were found at conditioning frequencies of 50 and 100 Hz. We also tested whether inhibition of LTP by previous priming lasts for longer than 40 min, as depicted in Figure 1A-C. To this end, we prolonged the interval between homosynaptic 10 Hz priming and conditioning stimulation to the lateral perforant path to 2 h. After this longer postpriming interval, we still observed a significant priming-induced reduction in LTP (unprimed slices, normalized fEPSP slope 1 h after tetanization, 1.33 ± 0.04, n = 7; primed slices, normalized fEPSP slope, 1.13 ± 0.03, n = 8; p < 0.05).
Homosynaptic priming does not affect subsequent LTP at the medial perforant path
The medial perforant path shows a termination area onto dentate granule cell dendrites that is distinct from the lateral perforant path. Specifically, medial perforant path synapses are formed on the proximal dentate granule cell dendrites within the middle molecular layer, whereas the lateral perforant path terminates within the outer portion of the molecular layer. This segregation is thought to have functional implications for information processing and plasticity in the dentate gyrus. Consistent with this possibility, significant differences in short- and long-term synaptic plasticity have been observed in these two pathways (McNaughton, 1980; Dahl and Sarvey, 1990). We therefore examined whether homosynaptic priming also affects the potential to express LTP in the medial perforant path. To this end, we used a conditioning frequency at which pronounced priming effects were observed in the lateral perforant path [50 Hz (Fig. 1C)]. As shown in Figure 1D, 10 Hz priming performed 40 min before conditioning stimulation caused only small effects on the amount of LTP (unprimed, 1.42 ± 0.05 normalized fEPSP slope, n = 8; 10 Hz primed, 1.27 ± 0.08 normalized fEPSP slope, n = 9). No statistical difference was found between the two groups. Although we did not examine conditioning stimulations over the whole range of frequencies tested for the lateral perforant path, these results suggest that the potential to exhibit priming-induced metaplasticity of LTP is much lower at medial perforant path synapses (see comparison of lateral and medial perforant path for conditioning stimulations of 50 Hz in Fig. 1E).
Heterosynaptic priming of the medial perforant path inhibits subsequent LTP at the lateral perforant path
Having established that the lateral perforant path displays significant homosynaptic priming effects, we further examined the conditions under which such effects can be elicited. The dentate gyrus afforded us with the opportunity to test whether priming can be induced heterosynaptically because the two anatomically distinct aspects of the perforant path can be functionally discriminated in vitro. As described previously (Dietrich et al., 1997), we positioned a pair of stimulating and recording electrodes within the outer and middle molecular layer of the dentate gyrus for lateral and medial perforant path recordings, respectively. Electrode positions were adjusted to yield clear current sinks in the outer and sources within the middle molecular layer for lateral perforant path stimulation and vice versa for medial perforant path stimulation (Fig. 2A, left). Furthermore, fEPSPs in the lateral perforant path showed paired-pulse facilitation, whereas those in the medial perforant path displayed paired-pulse inhibition (interpulse interval of 40 ms). As an additional criterion, we applied the group III metabotropic glutamate receptor agonist l-AP-4 (10 μm) at the end of each experiment. l-AP-4 selectively inhibited only lateral but not medial perforant path responses [fEPSP reduction by 26.4 ± 4.8%, n = 13 in the lateral perforant path and -0.5 ± 2.5%, n = 13 in the medial perforant path (Fig. 2A, right)]. Any recordings that did not fulfill these criteria were excluded from additional analysis.
Once baseline recording was established in this configuration, a 10 Hz priming stimulation was applied to the medial perforant path (Fig. 2B, open arrows), which did not affect lateral perforant path responses (Fig. 2B1). Forty minutes after priming to the medial perforant path, a 50 Hz conditioning stimulation was applied to the lateral perforant path (Fig. 2B1, filled arrow). The amount of LTP induced in the lateral perforant path by this conditioning stimulation was significantly reduced after heterosynaptic priming (normalized fEPSP slope, 1.01 ± 0.04; n = 6; p < 0.05) compared with unprimed control slices in the same recording configuration [normalized fEPSP slope, 1.39 ± 0.06; n = 7 (Fig. 2 B1,C)]. These data are in good agreement with published data, in which priming applied to the medial perforant path in vivo raised the threshold for induction of lateral perforant path LTP (Abraham et al., 2001). These findings indicate that heterosynaptic metaplasticity at these pathways is a robust phenomenon present both in vivo and in vitro. It should be noted that, unlike the study published by Abraham et al. (2001), the priming protocols used in the present study did not affect basal synaptic transmission (Fig. 2 B2). Likewise, application of the conditioning 50 Hz stimulation to the lateral perforant path failed to cause large changes in the medial perforant path fEPSPs (Fig. 2 B2).
Priming stimulation either in vitro or in vivo induces a long-lasting inhibition of LTP at the lateral perforant path
An examination of time intervals >8 h between priming and conditioning stimulation is difficult in vitro because of possible changes in recording conditions. We therefore opted for another approach to determine whether priming effects persist for >12 h. We stereotactically implanted stimulation electrodes unilaterally into the lateral perforant path of intact rats. After appropriate recovery periods (see Materials and Methods), 10 Hz priming stimulation identical to that used in the in vitro experiments was delivered in vivo to the lateral perforant path (in vivo priming) (Fig. 3A,C). A control group of animals was connected to the stimulus isolator and handled identically to primed animals but without application of the priming stimulation (sham priming) (Fig. 3A,C). Sixteen to 18 h after in vivo priming or sham priming, animals were killed, and hippocampal slices were prepared with a vibratome as described for the above experiments. After equilibration in the interface chamber for ∼1 h and baseline recording, conditioning stimulation was applied to the lateral perforant path using the same protocol as for the in vitro experiments (50 Hz, average time from application of priming/sham priming to in vitro induction of LTP, 17.8 ± 0.9 h, n = 10 and 16.3 ± 0.6 h, n = 12, respectively; NS).
Because the in vivo priming stimulation was performed unilaterally, we separately analyzed the hippocampi ipsilateral and contralateral to the stimulation electrode. The hippocampus implanted with the stimulation electrode showed a remarkable decrease of LTP after priming stimulation (normalized fEPSP slope, 1.12 ± 0.04; n = 4; p < 0.05), whereas robust LTP was evoked in sham-primed animals (1.5 ± 0.05; n = 6) (Fig. 3A). Interestingly, a significant but less pronounced effect of priming was observed on the contralateral side (primed group, 1.19 ± 0.04 normalized fEPSP slope, n = 6; sham-primed animals, 1.39 ± 0.05, n = 6; p < 0.05) (Fig. 3B). These experiments are summarized in Figure 3C. Because unilateral application priming stimulation in vivo caused reduction of LTP in both hippocampi, these findings also imply the existence of heterosynaptic priming effects, possibly via commissural fibers. These results presented in Figure 3 suggest that the priming-induced profound loss of LTP is strikingly stable, lasting at least 2 h in vitro and up to 16-18 h in vivo. Metaplasticity lasting for several hours in vitro has been seen previously (Holland and Wagner, 1998). Moreover, in vivo analyses have shown that some forms of metaplasticity induced in the dentate gyrus may persist for several days after priming stimulation (Abraham et al., 2001).
To exclude that 10 Hz priming stimulation itself caused LTP in vivo that might occlude later LTP in vitro, we monitored the long-term effects of 10 Hz priming stimulation in vivo. Application of the 10 Hz priming stimulation in vivo did not cause significant long-term changes in lateral perforant path synaptic efficacy compared with sham-primed animals either after 1 h (primed, 0.97 ± 0.02, n = 7; sham primed, 1.03 ± 0.02, n = 9) or after 24 h (primed, 1.12 ± 0.06, n = 7; sham primed, 1.00 ± 0.03, n = 9; NS). The medial perforant path behaved differently. Medial perforant path priming caused a significant transient reduction in synaptic efficacy at medial perforant pathway immediately after conclusion of the priming stimulation (primed, 0.50 ± 0.065, n = 4; sham primed, 0.97 ± 0.02, n = 10; p < 0.001), which returned to baseline values 2 h after priming and remained unaltered thereafter (24 h after priming, primed, 1.08 ± 0.05, n = 4; sham primed, 0.99 ± 0.03, n = 10; NS)
NMDA receptors and postsynaptic cell firing in the induction of metaplasticity
It has been proposed that Ca2+ influx into postsynaptic neurons is the key trigger for the induction of metaplasticity. Entry of Ca2+ may arise via either synaptic Ca2+-permeable receptors, most probably NMDA receptors, or voltage-gated Ca2+-permeable channels during action potential firing. We first examined whether NMDA receptors are required for the induction of this form of metaplasticity. We blocked NMDA receptors with d-APV during priming stimulation [25 μm for 30 min before priming (application indicated by horizontal bar in Fig. 4A)] and subsequently applied a 10 Hz priming stimulation. After 40 min of perfusion with normal ACSF, a 50 Hz conditioning stimulation was applied to elicit LTP. As shown in Figure 4A, LTP was successfully induced in both primed (normalized fEPSP slope, 1.45 ± 0.08; n = 8) and unprimed (normalized fEPSP slope, 1.49 ± 0.07; n = 7) groups, with no significant difference between the two groups (Fig. 4B). Thus, as seen previously for other forms of priming stimulation (Huang et al., 1992), induction of metaplasticity with 10 Hz stimulation relies on the activation of NMDA receptors.
Although NMDA receptors appear necessary for synaptic induction of metaplasticity, it could be that an important effect of NMDA receptor activation is facilitation of synaptically driven postsynaptic firing (Abraham et al., 2001). In this case, eliciting postsynaptic action potentials might also be an effective stimulus for induction of metaplasticity. We therefore induced granule cells to fire action potentials by antidromic stimulation with a 10 Hz priming paradigm identical to that used for synaptic priming. To exclude that the antidromic stimulation might recruit progressively less granule neurons during the course of 10 Hz stimulation, we measured the amplitude of the antidromic population spike during priming stimulation. The amplitudes of the antidromic population spikes elicited by the first and 20th stimulus during priming, however, did not reveal a reduction in amplitude (20th population spike 97.42% of first population spike; n = 9; data not shown). Application of antidromic priming stimulation caused some depression of lateral perforant path EPSPs, similar to although less pronounced than LTD induced by antidromic stimulation in other preparations (Christofi et al., 1993). Antidromic priming also resulted in a potent inhibition of subsequent LTP (1.16 ± 0.03; n = 9; p < 0.05) compared with unprimed control recordings (normalized fEPSP slope, 1.59 ± 0.14; n = 8), similar to homosynaptic and heterosynaptic priming (Fig. 1E, 2C). It should be noted that, although most of the granule cell activity will arise from the anti-dromically evoked action potentials, we cannot completely exclude some degree of synaptic activation during the antidromic priming stimulation by activation of hilar axons. Nevertheless, this experiment suggests that the average rate of postsynaptic cell firing can be a critical integration mechanism for adjusting subsequent synaptic plasticity, as proposed previously (Bienenstock et al., 1982; Abraham et al., 2001).
Autophosphorylation of αCaMKII at Thr305/306 is required for metaplasticity at the lateral perforant path
It has been shown recently that autophosphorylation of αCaMKII at Thr305/306 is important for regulating the CaMKII association with the postsynaptic density and for setting the threshold for LTP induction in naive slices (Elgersma et al., 2002). Therefore, this could be an attractive mechanism to control metaplasticity at the lateral perforant path. To test the involvement of CaMKII Thr305/306 phosphorylation in metaplasticity at the lateral perforant path, we used mice in which the Thr305 and Thr306 amino acids of endogenous αCaMKII are substituted by nonphoshorylatable residues (valine and alanine, respectively, TT305/6VA mutation) (Elgersma et al., 2002) (see Materials and Methods). In these mice, αCaMKII cannot undergo inhibitory autophosphorylation. We predicted that the priming-induced inhibition of LTP at the lateral perforant path should be significantly reduced in these animals if priming requires inhibitory Thr305/306 autophosphorylation. Because LTP levels with tetani at 50 Hz were not fully saturated in mice (data not shown), as opposed to rats (Fig. 1), we used 100 Hz conditioning stimuli for these experiments. In wild-type mice, application of this tetanus led to stable LTP after 1 h (1.35 ± 0.06; n = 9) (Fig. 5A, open symbols, C). The effects of previous application of a priming stimulus was very similar to that observed in rats: LTP was considerably diminished to 1.10 ± 0.04 (n = 9; p < 0.05) (Fig. 5A, filled symbols, C). In marked contrast, Thr305/306VA mice exhibited completely unaltered LTP levels after priming (unprimed, 1.33 ± 0.04, n = 7 vs primed, 1.33 ± 0.04, n = 7, respectively) (Fig. 5B,C). It is of note that the levels of LTP in wild-type and Thr305/306VA mice were comparable, indicating that the capacity for expression of LTP per se is intact in mutant mice. It is also unlikely that the thresholds for the induction of LTP are altered in mutant mice. We addressed this question by analyzing synaptic transmission after 10 Hz stimulation (see Materials and Methods). This stimulation paradigm corresponds to the threshold at which LTD transitions to LTP (θm). A 10 Hz stimulation did not cause significant changes in basal synaptic transmission in either wild-type (0.98 ± 0.07 compared with pre-10 Hz baseline, 40 min after stimulation; n = 12) or mutant (1.04 ± 0.09; n = 12) mice. Collectively, these results provide strong evidence that 10 Hz priming dramatically reduces the potential for LTP induction in the lateral perforant path via inhibitory autophosphorylation of αCaMKII at Thr305/306.
CaMKII activity is required for LTP in the lateral but not the medial perforant path
These experiments raise the question why metaplasticity is observed in the lateral but not the medial perforant path. One possibility that might account for this pathway-specific phenomenon is that CaMKII activity is required for LTP only in the lateral but not the medial perforant path. In this case, a reduction of CaMKII activity by inhibitory autophosphorylation during priming would not be expected to reduce subsequent LTP in the medial perforant path. We tested this idea by examining the effects of inhibiting CaMKII with KN93 (10 μm) on LTP in the medial versus lateral perforant path. Indeed, inhibition of CaMKII caused a marked reduction of LTP in the lateral perforant path (1.35 ± 0.06, n = 9 under control conditions vs 1.06 ± 0.04, n = 8 in the presence of KN93; p < 0.05) (Fig. 6A,B). In marked contrast, inhibiting CaMKII did not significantly affect LTP elicited in the medial perforant path (1.37 ± 0.07 under control conditions vs 1.22 ± 0.05 in the presence of KN93; n = 5 in both groups) (Fig. 6C,D).
Many CNS synapses appear to encode previous activity not only as changes in synaptic strength but also as a change in the potential to subsequently exhibit LTP or LTD. This phenomenon has been termed metaplasticity. In our experiments, we describe metaplasticity in the lateral perforant path invoked by stimulation at 10 Hz (close to θm), expressed as a dramatic and long-lasting loss in subsequent LTP in the lateral perforant path. A similar decrease in the potential to express LTP after priming stimulations that do not change basal synaptic transmission has been observed previously in the CA1 region (Huang et al., 1992; Larkman et al., 1992; O'Dell and Kandel, 1994). Alternatively, priming stimulation with high-frequency trains was performed in the presence of NMDA receptor antagonists to prevent the induction of LTP, with similar results (Wang and Wagner, 1999). Evidence of metaplasticity induced without changes in basal synaptic transmission in other hippocampal synapses has so far been limited (Christie and Abraham, 1992). Remarkably, in the present study, priming affected only subsequent lateral perforant path LTP but not LTD. Changes in LTD elicited with associative but not nonassociative stimulation protocols have been shown previously to occur only within a narrow frequency range, being present with priming frequencies of 5 Hz but not 1 or 15 Hz (Christie and Abraham, 1992). This raises the intriguing possibility that specific forms of metaplasticity may be invoked depending on the priming frequency band.
In the classical view, the expression mechanism of metaplasticity underlying these phenomena was thought to be a shift in θm, with the maximal amount of depression and potentiation left unchanged (Bienenstock et al., 1982; Bear et al., 1987) (Fig. 7A). Our results, as well as changes observed after visual deprivation in rats (Kirkwood et al., 1996), suggest that metaplasticity may also be expressed as a change in the maximal amount of LTD or LTP synapses can exhibit. In the visual cortex, visual deprivation causes marked loss of LTD, whereas LTP is left intact (Kirkwood et al., 1996) (depicted schematically in Fig. 7B). In our experiments, priming caused loss of LTP, with LTD left intact (depicted in Fig. 7C). In contrast to the purely “horizontal shift” in θm, these forms of metaplasticity may be described as a “vertical shift” in the Bienenstock-Cooper-Munro (BCM) curve (Philpot et al., 2003), leading to loss of either LTP or LTD.
Mechanisms underlying metaplasticity in the lateral perforant path
As reported previously in the CA1 region (Huang et al., 1992), as well as in the dentate gyrus (Abraham et al., 2001), metaplasticity induced by synaptic 10 Hz priming proved to be dependent on the activation of NMDA receptors. Thus, NMDA receptor-mediated Ca2+ influx appears to contribute to intracellular Ca2+ transients required for both induction of metaplasticity and LTP. However, this does not appear to be an absolute requirement, because antidromic priming in the absence of synaptic stimulation also induced metaplasticity in our experiments. Indeed, metaplasticity can be evoked even in the presence of NMDA receptor blockers if the intensity of the priming stimulation is sufficiently high (Wang and Wagner, 1999). The most parsimonious explanation for these results is that the absence of NMDA-mediated synaptic transmission can be compensated by other Ca2+ influx pathways. Moreover, the induction of metaplasticity with antidromic priming suggests that postsynaptic action potentials are sufficient to induce metaplasticity even in the complete absence of synaptic stimulation (Abraham et al., 2001).
Several studies have implied that CaMKII may be an attractive candidate to control metaplasticity (Abraham and Tate, 1997; Tompa and Friedrich, 1998; Elgersma et al., 2004). Indeed, CaMKII appears to be well suited to allow bidirectional shifts in synaptic plasticity thresholds. This capability of CaMKII likely hinges on the complex phosphorylation-dependent regulation of CaMKII at two major regulatory sites, Thr286 and Thr305/306. The current view holds that strong synaptic activation, such as during a high-frequency tetanus, causes binding of Ca2+/calmodulin to CaMKII subunits and subsequent translocation from the actin cytoskeleton to the postsynaptic density. Sufficient levels of Ca2+/calmodulin cause autophosphorylation at Thr286, which dramatically increases the affinity for Ca2+/calmodulin and enables the holoenzyme to remain autonomously active even after Ca2+ levels have decayed (for review, see Lisman et al., 2002). As long as Ca2+/calmodulin is bound to CaMKII, phosphorylation at the Thr305/306 sites is inhibited because these sites are located within the Ca2+/calmodulin binding domain of CaMKII. After translocation to the postsynaptic density, CaMKII then binds to the NR2B subunit of the NMDA receptor (Bayer et al., 2001). This binding has several major consequences. First, it causes autonomous CaMKII activity that is not dependent on the phosphorylation state of the kinase, as well as trapping of calmodulin. Moreover, binding to the NR2B subunit blocks subsequent phosphorylation at Thr305/306. Thus, a large increase of postsynaptic Ca2+/calmodulin (e.g., during high-frequency tetanization) cannot immediately result in the subsequent inhibition of CaMKII activity by Thr305/306 autophosphorylation. Indeed, our experiments using mutant mice expressing αCaMKII that cannot undergo inhibitory autophosphorylation show that blockade of inhibitory autophosphorylation does not influence the maximal levels of LTP at lateral perforant path synapses obtained with strong tetanic stimulation.
Conversely, our experiments do demonstrate a pivotal role of inhibitory Thr305/306 autophosphorylation for expression of priming-induced metaplasticity. How could subthreshold priming stimuli efficiently induce Thr305/306 autophosphorylation? A likely explanation is that subthreshold priming stimulation leads to activation of only few CaMKII subunits within the entire holoenzyme. This would be insufficient to activate CaMKII to a level that it can move to the PSD and bind to the NR2B subunit of the NMDA receptor. In fact, CaMKII may even fail to dissociate from the actin cytoskeleton because this process requires Ca2+/calmodulin binding (Shen et al., 1998; Shen and Meyer, 1999). In this scenario, subsequent dissociation of Ca2+/calmodulin from CaMKII subunits would permit phosphorylation of these subunits at Thr305/306. This phosphorylation has two important consequences. First, although this CaMKII may have some autonomous activity attributable to a small amount of Thr286 phosphorylation, Thr305/306 phosphorylation abolishes the ability to be fully reactivated, because T305/306 phosphorylation interferes with Ca2+/calmodulin binding (Hashimoto et al., 1987; Colbran and Soderling, 1990; Patton et al., 1990; Mukherji and Soderling, 1994). More importantly, in vitro as well as in vivo studies have demonstrated that Thr305/306 phosphorylation of αCaMKII strongly reduces the affinity of αCaMKII to the postsynaptic density (Strack et al., 1997; Shen et al., 2000; Elgersma et al., 2002). The association of αCaMKII with the postsynaptic density, in turn, is thought to be essential for the expression of LTP (Hu et al., 1999; Lisman and Zhabotinsky, 2001; Elgersma et al., 2002). Because phosphorylation at Thr305/306 has a strong dominant-negative effect (Elgersma et al., 2002), Thr305/306 phosphorylation of only a few subunits may be sufficient to prevent PSD association of the CaMKII holoenzyme. Thus, although a subsequent stimulus of 100 Hz may induce Thr286 phosphorylation at the remaining subunits that are not blocked by previous Thr305/306 phosphorylation, this is apparently not sufficient to overcome the PSD dissociating properties of the subunits within the holoenzyme that are phosphorylated at Thr305/306. This mechanism would readily explain why subthreshold activation causes phosphorylation at Thr305/306 and why this, in turn, powerfully inhibits subsequent LTP.
Our results are also in good agreement with experiments performed in other types of αCaMKII mutant mice. Mice expressing αCaMKII that mimics inhibitory phosphorylation display a dramatic reduction in LTP and learning (Elgersma et al., 2002). Similarly, because phosphorylation at Thr305/306 occurs rather efficiently (Elgersma et al., 2002), the observed shift toward LTD in the Thr286Asp mice expressing autonomously activated CaMKII (Mayford et al., 1995) could also be attributable to increased Thr305/306 phosphorylation in these mice. We did not observe changes in LTD after priming stimulation, suggesting that this CaMKII-dependent mechanism does not affect the mechanisms underlying long-term depression of synaptic responses. Indeed, either genetic or pharmacological inhibition of CaMKII (Kirkwood et al., 1997; Lee et al., 2000) or genetic manipulation of inhibitory autophosphorylation fails to alter the levels of LTD (Elgersma et al., 2002). Although our results suggest that priming-induced modulation of αCaMKII association with the PSD may be a key contributor to metaplasticity (Fig. 7D), the data do not exclude that additional enzymatic pathways may also play a role in metaplasticity at other synapses, as has been suggested for protein phosphatases (Zeng et al., 2001; Woo and Nguyen, 2002). It is an intriguing possibility that protein phosphatase pathways might affect αCaMKII and, by extension, CaMKII-dependent metaplasticity. In addition, we should note that, although inhibitory autophosphorylation of αCaMKII appears to be required for metaplasticity expressed during the first hour after priming, we cannot exclude that additional, downstream mechanisms may be involved in metaplasticity expressed at longer latencies after priming.
Pathway specificity of priming effects
Intriguingly, priming stimulations that readily induced metaplasticity in the lateral perforant path failed to do so in the medial perforant path, although the lateral and medial perforant paths terminate on the same postsynaptic target. In principle, such a pathway specificity may arise in two ways. First, inhibitory autophosphorylation of CaMKII could be induced only in the lateral but not the medial perforant path. Such a difference might arise via expression of different types of NMDA receptors or Ca2+ channels, giving rise to local differences in priming-induced Ca2+ increases. Downstream of Ca2+ influx pathways, it might be that inhibitory autophosphorylation occurs at both the medial and lateral perforant paths, but that CaMKII activity is not required for LTP induction at the medial perforant path. Indeed, we found that blocking CaMKII with KN93 completely inhibited LTP at the lateral perforant path while only weakly affecting LTP at the medial perforant path. This idea is in good agreement with recent work showing that medial perforant path LTP is only affected if either p42/44 mitogen-activated protein kinase or protein kinase A are blocked in addition to CaMKII (Wu et al., 2004). These results indicate that loss of CaMKII activity can be compensated by other kinases in the medial but not the lateral perforant pathway. More generally, the requirement for different sets of kinases at individual synaptic pathways may constitute an attractive mechanism to generate pathway-specific and stimulation-paradigm-specific modulation of synaptic plasticity.
This work was supported by Deutsche Forschungsgemeinschaft Sonderforschungsbereich Tr3/C4 (D.M.-V., H.B.) and by a grant of the National Epilepsy Fund, “The Power of the Small” (Y.E.). We are grateful to Jens Klausnitzer for technical support.
Correspondence should be addressed to Heinz Beck, Department of Epileptology, University of Bonn, Sigmund-Freud-Strasse 25, D-53105 Bonn, Germany. E-mail:.
Copyright © 2005 Society for Neuroscience 0270-6474/05/257697-11$15.00/0