Synaptic memory requires CaMKII

  1. Wucheng Tao
  2. Joel Lee
  3. Xiumin Chen
  4. Javier Díaz-Alonso
  5. Jing Zhou
  6. Samuel Pleasure
  7. Roger A Nicoll  Is a corresponding author
  1. Key Laboratory of Brain Aging and Neurodegenerative Diseases, Fujian Medical University, China
  2. Department of Cellular and Molecular Pharmacology, University of California, San Francisco, United States
  3. Department of Neurology, University of California, San Francisco, United States
  4. Physiology, University of California, San Francisco, United States

Abstract

Long-term potentiation (LTP) is arguably the most compelling cellular model for learning and memory. While the mechanisms underlying the induction of LTP (‘learning’) are well understood, the maintenance of LTP (‘memory’) has remained contentious over the last 20 years. Here, we find that Ca2+-calmodulin-dependent kinase II (CaMKII) contributes to synaptic transmission and is required LTP maintenance. Acute inhibition of CaMKII erases LTP and transient inhibition of CaMKII enhances subsequent LTP. These findings strongly support the role of CaMKII as a molecular storage device.

Editor's evaluation

The article addresses the ‘decades-old’ unresolved question as to whether CaMKII is required for the maintenance of synaptic long-term potentiation, and shows, based on a set of elegant experiments, that it indeed is.

https://doi.org/10.7554/eLife.60360.sa0

eLife digest

How the brain stores information is a question that has fascinated neuroscientists for well over a century. Two general ideas have emerged. The first is that groups of neurons hold information by staying active. The second is that they hold information by strengthening their connections to one another, making it easier for them to work together in the future. Scientists call this second idea 'long-term potentiation'.

One of the molecules involved in long-term potentiation is a protein called calcium-calmodulin-dependent kinase II, or CaMKII for short. Blocking CaMKII, or deleting its gene, stops the connections between neurons from becoming stronger. This suggests neurons need CaMKII to learn, but it remains unclear whether neurons also use CaMKII to maintain neuronal memories after they have been created. If CaMKII does play a role in maintaining memories, blocking it after learning should reverse the learning process, but so far, experiments have not been able to show this.

Tao et al. revisited these experiments to find out more. They examined slices of brain tissue from mice that had been treated with fast-acting CaMKII inhibitors. It took tens of minutes, but the inhibitors were able to reverse long-term potentiation, both for newly acquired neuronal memories and for older memories that had formed when the mice were alive. The choice of CaMKII inhibitor and the time lag could explain why scientists have not observed the effect before.

Understanding long-term potentiation is a fundamental part of understanding learning and memory. It could also reveal more about the opposite phenomenon: long-term depression. This is a type of learning where the connections between neurons become weaker. Long-term depression also takes tens of minutes to occur, suggesting that future research into CaMKII might shed light on how it works.

Introduction

The mechanism by which the brain stores information has fascinated neuroscientists for well over a century. Two general ideas have emerged. The first proposes that information is held by ongoing activity in neuron ensembles. This, indeed, underlies short-term working memory (D’Esposito and Postle, 2015; Inagaki et al., 2019; Wang, 2001). However, it is clear that such a mechanism cannot account for long lasting, enduring memories. For instance, it is well established that patients who recover from prolonged periods of isoelectric EEG and brain stem silence following barbiturate overdose have no detectable cognitive deficits, for example, Bird and Plum, 1968. The second and generally accepted view, first enunciated by Cajal over a century ago (Cajal, 1911), is the modification of neuronal connections. How these changes might be stored has remained a mystery.

Over the past few decades, Ca2+-calmodulin-dependent kinase II (CaMKII) has emerged as a very attractive molecular candidate for information storage (Bayer and Schulman, 2019; Bhattacharyya et al., 2020; Hell, 2014; Kennedy, 2013; Lisman et al., 2002; Lisman et al., 2012; Lisman and Goldring, 1988). This dodecameric kinase is activated by Ca2+/CaM resulting in autophosphorylation and the activity of the kinase remains after the removal of Ca2+ (Miller and Kennedy, 1986), a state referred to as autonomy. Furthermore, evidence suggests that the enzyme can undergo activation-triggered subunit exchange such that subunits that did not experience the original Ca2+ transient are phosphorylated, a mechanism postulated to allow for continued phosphorylation in the face of protein turnover (Bhattacharyya et al., 2020; Bhattacharyya et al., 2016; Stratton et al., 2014) Thus, CaMKII has many attractive biochemical features expected for a memory molecule. What is the physiological evidence for such a role?

Long-term potentiation (LTP), in which brief high-frequency synaptic stimulation results in a lasting increase in synaptic strength, is an attractive cellular model for learning and memory (Choquet, 2018; Collingridge et al., 2004; Huganir and Nicoll, 2013; Malinow and Malenka, 2002; Nicoll, 2017). A role for CaMKII in LTP is compelling. Pharmacologically blocking CaMKII (Malenka et al., 1989; Malinow et al., 1989; Otmakhov et al., 1997) or genetically deleting CaMKII (Giese et al., 1998; Incontro et al., 2018; Silva et al., 1992) blocks LTP and constitutively active CaMKII both mimics and occludes LTP (Lledo et al., 1995; Pettit et al., 1994; Pi et al., 2010; Poncer et al., 2002). However, despite the attractive biochemical properties of CaMKII, whether CaMKII is involved in the maintenance of LTP and, by extension, memory storage remains problematic.

There are two predictions if CaMKII is responsible for LTP maintenance and synaptic memory (Lisman, 2017; Sanhueza and Lisman, 2013). First, transiently blocking CaMKII after the induction of LTP should cause a long-lasting erasure. Attempts at reversing LTP have failed on numerous occasions (Buard et al., 2010; Chen et al., 2001; Malinow et al., 1989; Murakoshi et al., 2017; Otmakhov et al., 1997), but see Feng, 1995. Second, if CaMKII is involved in synaptic memory, one would expect it to leave a lasting trace at synapses (Lisman, 2017; Sanhueza et al., 2011; Sanhueza and Lisman, 2013). Thus, silencing CaMKII should reduce synaptic transmission. Interestingly, the CaMKII/NMDAR complex does exist under basal conditions (Leonard et al., 1999). However, the genetic deletion of CaMKII (Achterberg et al., 2014; Giese et al., 1998; Silva et al., 1992), but see Hinds et al., 1998, or pharmacological inhibition of CaMKII does not affect baseline synaptic responses (Buard et al., 2010; Chen et al., 2001; Feng, 1995; Malinow et al., 1989; Murakoshi et al., 2017; Otmakhov et al., 1997; Wang and Kelly, 1996). Finally, using a FRET-based CaMKII sensor, CaMKII activation in spines after LTP induction persists for only ~1 min (Lee et al., 2009). Thus, physiological support for a role of CaMKII in maintaining LTP or synaptic transmission is lacking.

Recent findings have prompted us to reevaluate CaMKII’s role in the maintenance of LTP. Deleting CaMKII with CRISPR (Incontro et al., 2018) or expressing a peptide inhibitor of CaMKII (Goold and Nicoll, 2010; Sanhueza et al., 2011) cause a substantial reduction in synaptic transmission. Furthermore, studies using a membrane permeable peptide inhibitor of CaMKII (tatCN21 or antCN27) have reported a lasting depression in synaptic transmission (Barcomb et al., 2016; Gouet et al., 2012; Sanhueza et al., 2011; Sanhueza et al., 2007) and evidence for a reduction in LTP maintenance (Sanhueza et al., 2011; Sanhueza et al., 2007).

To reevaluate the role of CaMKII in synaptic memory we have used two different rapidly acting and reversible CaMKII inhibitors. As would be expected if LTP contributes to synaptic memory, inhibition of CaMKII depresses synaptic transmission and the properties of this persistent action of CaMKII are remarkably similar to those of LTP. Furthermore, we demonstrate that applying these inhibitors after inducing LTP fully reverses LTP, indicating that CaMKII is required for the persistence of LTP. Thus, our findings strongly support CaMKII as a molecular storage device. Reasons for why the present results differ from most previous studies are discussed.

Results

Blocking CaMKII inhibits synaptic transmission

Previous studies established, both with the expression of inhibitory peptides (Goold and Nicoll, 2010; Sanhueza et al., 2011) or CRISPR deletion of CaMKII (Incontro et al., 2018) that CaMKII contributes ~50% to AMPAR responses. In order to study the basis for this activity as well as the role of CaMKII in LTP maintenance, rapid and reversible block of CaMKII is essential. Since peptide inhibitors have proved most effective, this requires making these peptides membrane permeable either by cell-penetrating peptides (CPPs) (e.g., tat) or by protein lipidation (e.g., myristoylation). Delivery of peptides into cells has profound therapeutic potential and therefore has received a great deal of study (Allen et al., 2018; LeCher et al., 2017; Nelson et al., 2007; Patel et al., 2019). The effectiveness of CPPs is hotly debated, because, although these peptides clearly enter the cell by endocytosis, it is not certain to what degree these peptides actually have access to the cytoplasm. This appears not to be an issue with lipid modification.

As an initial assay we bathed acute slices in a peptide inhibitor and measured the AMPAR/NMDAR ratio. We would expect the ratio to be reduced by ~50%, since expressing these peptides had no effect on NMDAR currents. We first tested the effectiveness of tatCN21 (5 μM) (Barcomb et al., 2016; Buard et al., 2010; Sanhueza et al., 2011) in reducing the AMPAR/NMDAR ratio (Figure 1—figure supplement 1). In agreement with previous reports, we failed to see an effect on AMPAR responses. Since higher concentrations of tatCN21 have significant nonspecific effects (Barcomb et al., 2016), we turned to other peptides. Previous work (Goold and Nicoll, 2010) showed that prolonged (1–2 days) incubation with myr-CN27 (10 µM) caused a ~50% reduction in the AMPAR/NMDAR ratio. We, therefore, repeated this experiment with shorter incubations (1–2 hr) and a lower concentration (1 µM) and found that myr-CN27 reduced the ratio by ~50% (Figure 1—figure supplement 1). What might account for the lack of effect of tatCN21? To address this, we repeated the experiment using myr-CN21 (5 μM) and found that it reduced the ratio by ~50% (Figure 1—figure supplement 1), suggesting that lipidation is a more effective mode of delivery than CPPs. We, therefore, selected myr-CN27 for our studies.

Application of myr-CN27 to acute slices (1 μM) caused a highly reproducible slowly developing depression of synaptic AMPAR mediated EPSCs, which stabilized after approximately 40 min (Figure 1A, B). A number of experiments were carried out to ensure that the effect we observe with myr-CN27 is due to the specific inhibition of CaMKII. First, in many experiments we measured the size of the NMDAR EPSC, by depolarizing the cell to +40 mV before and after the application of the peptide. We found that myr-CN27 had no effect on the NMDAR EPSC (Figure 1A, red circles, Figure 1B). Second, it is important to note that the magnitude of the decrease in transmission seen with myr-CN27 is similar to that reported with the genetic deletion of CaMKIIα (Incontro et al., 2018) suggesting that myr-CN27 fully silences CaMKII. Finally, we examined the effects of myr-CN27 in cells lacking CaMKIIα (Figure 1C, D). CaMKIIα was deleted using CRISPR (Incontro et al., 2018). We simultaneously recorded from control cells and cells lacking CaMKIIα in slice culture. In the absence of CaMKIIα (green circle) synaptic responses were ~50% of control (black circles), as expected. Application of myr-CN27 depressed controls cells to the level recorded in cells lacking CaMKIIα (Figure 1C, D, black circles), but it had no additional effect on cells lacking CaMKIIα (green circles) (n = 6). These results demonstrate that the effect of CN27 is fully explained by its selective and complete silencing of CaMKIIα.

Figure 1 with 1 supplement see all
Ca2+-calmodulin-dependent kinase II (CaMKII) inhibitor myr-CN27 inhibits AMPAR synaptic transmission through CaMKIIα.

(A) Time course of myr-CN27 (1 µM) inhibition on AMPAR EPSCs and NMDAR EPSCs in acute slices. (B) Summary graphs showing myr-CN27 inhibited AMPAR EPSCs (before drug: 63.9 ± 12 pA; after drug: 27.6 ± 3.9 pA; n = 10, p < 0.01, two-tailed Wilcoxon Signed Rank Test), but not NMDAR EPSCs (before drug: 43.2 ± 13.7 pA; after drug: 42 ± 13.4 pA; n = 5, p > 0.05, two-tailed Wilcoxon Signed Rank Test). (C) Time course of the effect of myr-CN27 on AMPAR EPSCs in wt cells and simultaneously recorded CRISPR-CaMKIIα transfected cells, normalized to wt baseline (from culture slices). While myr-CN27 inhibited AMPAR EPSCs in wt cells, it had no effect on CRISPR-CaMKIIα transfected cells (n = 6, p > 0.05, two-tailed Wilcoxon Signed Rank Test). (D) Sample traces showing that myr-CN27 inhibited AMPAR EPSCs in control cells, but had no effect in CRISPR-CaMKIIα transfected cells. Black traces are control cell, green traces are transfected cell. Mean ± standard error of the mean (SEM). **p < 0.01. (E) A comparison of the reduction of synaptic transmission in field and whole cell recording. For the field recording (black, n = 8) after a 60-min stable baseline, 1 μM myr-CN27 was applied to the slice for 100 min. For the whole cell recording (blue, n = 7), the drug was applied after a 10-min stable baseline. In both, the response was reduced to 50% of baseline. Example traces for both are shown taken at the timepoints indicated by 1 and 2. The field EPSPs shown on the bottom right have scale bars vertical 0.1 mV and 5 ms horizontal. The whole cell recording shown bottom left have scale bars 50 pA vertical and 20 ms horizontal. Both recordings are done in WT slices with the whole cell recordings done in slice culture DIV3–14 while the fields were done in acute slices from P15–28 mice.

Figure 1—source data 1

Ca2+-calmodulin-dependent kinase II (CaMKII) inhibitor myr-CN27 inhibits AMPAR synaptic transmission through CaMKIIα.

In this dataset, the results of the effects of myr-CN27 on AMPAR EPSCs and NMDAR EPSCs are included.

https://cdn.elifesciences.org/articles/60360/elife-60360-fig1-data1-v2.xlsx

Previous studies have relied on field potential recording and thus we repeated our experiments with this approach (Figure 1E). We still recorded a depression, but in this case the depression took tens of minutes to develop (black symbols). We interleaved these experiments with whole cell recording and confirmed that the depression is considerably faster (blue symbols). We interpret the difference in time course to the fact that whole cell recordings are made from superficial cells, while field potential responses are generated by populations of neurons throughout the depth of the slice. Penetration of peptides into the slice is expected to be slow.

We next examined the effects of AIP, a peptide inhibitor designed from the autoinhibitory domain. We first expressed AIP in neurons in slice culture and compared AMPAR and NMDAR responses to those in simultaneously recorded neighboring control cells (Figure 2A). Similar to the effect of expressing CN27 (Goold and Nicoll, 2010) or applying myr-CN27 (see above), there was a selective ~50% reduction in the AMPAR EPSC, but no change in the NMDAR response. Given these positive results, we bathed acute slices in myr-AIP (20 μM) and measured the AMPAR/NMDAR ratio (Figure 2B) (n = 12). The ratio was significantly reduced. Acute application of myr-AIP in slice culture caused a slowly developing depression in synaptic activity (Figure 2C, black circles) (n = 6) and this effect is due to the selective inhibition of CaMKII, because interleaved recording from cells in which CaMKIIα had been deleted (Figure 2C, green circles), myr-AIP no longer had any effect (n = 5). These findings show that myr-CN27 and myr-AIP selectively and completely inhibit CaMKII, both in acute slices and in slice culture. It is interesting that, while CaMKII is expressed presynaptically and reported to effect transmitter release (Benfenati et al., 1992; Hinds et al., 2003), there was no presynaptic effect of the peptide inhibitors.

Ca2+-calmodulin-dependent kinase II (CaMKII) inhibitor AIP inhibits AMPAR synaptic transmission through CaMKIIα.

(A) In culture slices, transfection of AIP reduced AMPAR EPSCs (left, n = 10, p < 0.01, two-tailed Wilcoxon Signed Rank Test) but not NMDAR EPSCs (right, n = 10, p > 0.05, two-tailed Wilcoxon Signed Rank Test). Sample traces show the effects of AIP on AMPAR EPSCs and NMDAR EPSCs. (B) AMPA/NMDA ratios compared to wild-type (WT) (n = 34) are reduced after myr-AIP (20 µM) treatment (n = 12). Scale bar = 50 pA vertical and 30 ms horizontal. (C) Left, summary of the effect of myr-AIP (20 µM) on AMPAR EPSCs in wt cells (n = 6) and interleaved CRISPR-CaMKIIα transfected cells (n = 5) from culture slices, normalized to each cell’s baseline. The difference between control and CRISPR-CaMKIIα transfected cells at 30 min: p < 0.01, Mann–Whitney U test. Right, sample traces showing myr-AIP inhibition of AMPAR EPSC in control cells, but not in CRISPR-CaMKIIα transfected cells. Black traces are control cell, green are transfected cell. Mean ± standard error of the mean (SEM).

Figure 2—source data 1

Ca2+-calmodulin-dependent kinase II (CaMKII) inhibitor AIP inhibits AMPAR synaptic transmission through CaMKIIα.

In this dataset, the results of the effects of AIP on AMPAR EPSCs and NMDAR EPSCs are included.

https://cdn.elifesciences.org/articles/60360/elife-60360-fig2-data1-v2.xlsx

The time course of inhibition for both myr-CN27 and myr-AIP is slow, requiring 10’s of minutes. Is this due to the slow access of these peptides to the cell interior, or is it due to the slow reversal of the action of CaMKII? To address this, we turned to a novel light inducible inhibitor of CaMKII, in which the inhibitory peptide AIP2 (Ishida et al., 1998) is linked to the LOV-Jα helix domain of Phototropin (referred to as paAIP2) (Murakoshi et al., 2017). Blue light rapidly exposes the AIP2 peptide, which then returns to its closed inactive conformation in approximately 40 s following light exposure. We expressed paAIP2 in individual neurons with biolistics, which, unlike the myristoylated peptides, has the advantage of limiting the manipulation to individual postsynaptic cells. The synaptic properties of neurons expressing paAIP2 were not different from the properties of neighboring control cells (Figure 3—figure supplement 1A, B) (n = 16). Exposure of these cells to blue light resulted in a slowly developing inhibition of AMPAR responses (Figure 3A) (n = 12), similar to the kinetics of myr-CN27 and myr-AIP. The blue light had no effect on simultaneously recorded control cells (Figure 3A, black circles) or in cells coexpressing paAIP2 with CRISPR to delete CaMKIIα (Figure 3—figure supplement 2) (n = 5), establishing the specificity of paAIP2. These results with paAIP2 suggest that the slow time course is not due to slow access, but rather due to the slow reversal of the action of CaMKII.

Figure 3 with 2 supplements see all
Light-activated Ca2+-calmodulin-dependent kinase II (CaMKII) inhibitor paAIP2 inhibits AMPAR synaptic transmission through CaMKIIα.

(A) Time course of paAIP2 effect on AMPAR and NMDAR synaptic transmission in culture slices. Blue light exposure inhibited AMPAR synaptic transmission in paAIP2 expressing cells (green circles), but not in simultaneously recorded control cells (black circles) or NMDAR synaptic transmission (red circles). (B) Summary data showing that paAIP2 inhibited AMPAR synaptic transmission (before light: 51 ± 13.8 pA; after light: 26.5 ± 7.2 pA; n = 12, p < 0.01, two-tailed Wilcoxon Signed Rank Test), but had no effect on NMDAR synaptic transmission (before light: 31.2 ± 7.6 pA; after light: 29.8 ± 7.7 pA; n = 5, p > 0.05, two-tailed Wilcoxon Signed Rank Test). Mean ± standard error of the mean (SEM). **p < 0.01.

Figure 3—source data 1

Light-activated Ca2+-calmodulin-dependent kinase II (CaMKII) inhibitor paAIP2 inhibits AMPAR synaptic transmission through CaMKIIα.

In this dataset, the results of the effects of paAIP2 on AMPAR EPSCs and NMDAR EPSCs are included.

https://cdn.elifesciences.org/articles/60360/elife-60360-fig3-data1-v2.xlsx

The origin of the constitutive action of CaMKII

The persistent CaMKII activity has two possible origins. First, it could represent ongoing constitutive Ca2+ activation of CaMKII from various sources. It seems unlikely that resting Ca2+ levels (Maravall et al., 2000) stimulate CaMKII, since the sensitivity of CaMKII to Ca2+ requires considerably higher levels of Ca2+ (Schulman, 1984). To directly test whether constitutive CaMKII activity is Ca2+ dependent, we loaded cells with the Ca2+ chelator BAPTA (15 mM). However, myr-CN27 retained its depressive effect (Figure 4—figure supplement 1A). This finding also rules out spontaneous background NMDAR activity as a source for the constitutive CaMKII activity. In addition, acute application of the NMDAR antagonist APV has little effect on AMPAR EPSCs (Incontro et al., 2018).

The constitutive activity might reflect an LTP process induced by slice preparation, which is known to depolarize cells and release glutamate. To address this possibility, we pretreated animals with a high dose of MK-801, which is known to block in vivo NMDAR responses (Davies et al., 1988). To verify that NMDARs were blocked by this procedure, we recorded NMDAR EPSCs from slices from these animals and found that the NMDAR EPSC was blocked (Figure 4—figure supplement 1B). Having confirmed the absence of NMDAR responses the slices were perfused with MK-801. Under such conditions myr-CN27 still exerted its inhibitory effect (Figure 4—figure supplement 1C).

Second, if CaMKII is involved in synaptic memory, one would expect that LTP acquired when the animal was alive, would contribute to synaptic transmission, as previously proposed (Lisman, 2017; Sanhueza and Lisman, 2013). If this is the case, there are a number of predictions. One would expect that the transient inhibition of CaMKII should result in long-lasting inhibition, akin to the resetting of a molecular switch. In a number of cells, we monitored the depression over time following the removal of myr-CN27 and found no recovery (Figure 4A). One might argue that myr-CN27 fails to washout of the cells. To circumvent this possibility, we repeated these experiments with paAIP2 whose action rapidly reverses upon terminating the blue light (~40 s). Again, we failed to observe any recovery (Figure 4B). We confirmed that this was not due to the continued action of the inhibitor, by demonstrating that LTP could be induced following light termination (see below, Figure 5).

Figure 4 with 1 supplement see all
The depression caused by transient inhibition of Ca2+-calmodulin-dependent kinase II (CaMKII) is long lasting and is absent in cells lacking NMDARs.

(A) In acute slices, time course showing that following myr-CN27 application AMPAR synaptic transmission remains depressed (the difference between before and after myr-CN27: n = 4, p < 0.01, two-tailed Wilcoxon Signed Rank Test). (B) In culture slices, time course showing that following blue light exposure on paAIP2 expressing cells AMPAR synaptic transmission remains depressed (the difference between before and after blue light exposure: n = 5, p < 0.01, two-tailed Wilcoxon Signed Rank Test). (C) Left panel, paired recordings from acute slices of control cells and those expressing Cre in GluN1 floxed mice (n = 5, p < 0.01, two-tailed Wilcoxon Signed Rank Test). Middle panel, time course of the myr-AIP effect on AMPAR synaptic transmission in control cells (n = 5) and interleaved cells expressing Cre in GluN1 floxed mice (n = 5). p < 0.01, Mann–Whitney U test. Right panel, sample traces showing myr-AIP inhibited AMPAR synaptic transmission in control cell (black traces), but not in a GluN1KO cell (green traces). Mean ± standard error of the mean (SEM). (D) Left panel, paired recordings from acute slices of control cells and those expressing Cre and N812W (pore dead, PD) in GluN1 floxed mice (n = 6, p < 0.05, two-tailed Wilcoxon Signed Rank Test). Middle panel, the action of myr-CN27 is not rescued by expressing a GluN1 pore dead mutant (n = 8). Right panel, sample traces showing myr-CN27 inhibited AMPAR synaptic transmission in control cell (black traces), but not in a GluN1KO + PD cell (green traces).

Figure 4—source data 1

The depression caused by transient inhibition of Ca2+-calmodulin-dependent kinase II (CaMKII) is long lasting and is absent in cells lacking NMDARs.

In this dataset, the results of the effects of myr-CN27 and paAIP2 on AMPAR EPSCs over time are included.

https://cdn.elifesciences.org/articles/60360/elife-60360-fig4-data1-v2.xlsx
Transient inhibition of Ca2+-calmodulin-dependent kinase II (CaMKII) enhances subsequent long-term potentiation (LTP).

(A) Timeline of experimental procedure. Following in utero electroporation acute slices were prepared at P15–P25. Slices were then exposed to blue light for 40 min. Paired recordings were then made from a control cell (black circles) and a paAIP2 expressing cell (green circles). (B) One representative experiment showing that prior paAIP2 activation reduced baseline AMPAR EPSCs, but LTP was larger than that recorded in the neighboring control cell. (C) Normalized representative experiment showing that prior paAIP2 activation increased LTP. (D) Summary data showing that prior paAIP2 activation reduced baseline AMPAR EPSCs, but enhanced LTP (n = 5). (E) Normalized summary data showing that prior paAIP2 activation doubles the size of LTP. n = 5, p < 0.01, two-tailed Wilcoxon Signed Rank Test. Mean ± standard error of the mean (SEM).

Figure 5—source data 1

Transient inhibition of Ca2+-calmodulin-dependent kinase II (CaMKII) enhances subsequent long-term potentiation (LTP).

In this dataset, the results of the effects of LTP induction on control cells and preinhibition cells are included.

https://cdn.elifesciences.org/articles/60360/elife-60360-fig5-data1-v2.xlsx

If the constitutive activity resulted from prior NMDAR-dependent LTP, one would expect that this inhibitory action should be absent in cells in which NMDARs have been deleted embryonically. We carried out in utero electroporation to express cre recombinase in neurons in GluN1 floxed mice. Transfected cells were recorded at P16–20 in acute slices. We first carried out paired recordings in the AMPAR inhibitor CNQX to ensure that cre expressing neurons lacked NMDAR EPSCs (Figure 4C, left panel). We then compared the action of myr-AIP in transfected cells with that in interleaved control cells and normalized the baseline to 100%. Myr-CN27 had no effect in these cells. Application of myr-AIP had no effect in these cells (Figure 4C, middle and right panels, green circles and traces), whereas it had its normal inhibitory effect in interleaved control cells (Figure 4C, black circles and traces), further supporting the conclusion that the persistent CaMKII activity represents prior LTP. A caveat to these experiments is that the synaptic action of CaMKII requires its binding to the GluN2B C-terminus of the NMDAR. Thus, the loss of GluN2B in the GluN1 lacking neurons could equally explain the loss of effect of myr-CN27. To address this concern we repeated these experiments, replacing GluN1 with a pore dead GluN1 mutant (N812W) (Amin et al., 2017). While this mutant assembles normally with GluN2 subunits and traffics normally to the membrane, it generates little current (Amin et al., 2017). We carried out in utero electroporation in GluN1 floxed mice to transfect Cre and GluN1(N812W). Thus, this genetic experiment is the equivalent of having APV present in utero and until acute slices were made at P14–18. The NMDAR EPSC in transfected neurons was severely reduced, when compared to neighboring control cells recorded in CNQX (Figure 4D, left panel). The depressant effect of myr-CN27 recorded in these transfected cells was absent (Figure 4D, middle and right panels). These results suggest that in behaving animals, cells that lack NMDAR function, but with GluN2B intact, Ca2+ signaling independent of the NMDAR appears incapable of activating CaMKII, at least in regard to its action on synaptic function.

Since LTP is saturable, one might expect that when constitutive activity is transiently silenced, the magnitude of LTP should be larger than that in control conditions. To test this prediction, we transfected paAIP2 in utero and prepared acute slices at days P15–25. The slices were exposed to blue light for 40 min and then simultaneous recordings were made from a transfected cell and a control cell (Figure 5A). In a typical experiment (Figure 5B), as well as in the summary of all experiments (Figure 5D), the size of baseline EPSCs in paAIP2 expressing cells following blue light exposure was, as expected, approximately half of that of the control neighboring cells. LTP was then induced in both cells. To compare the magnitude of LTP in the two cells we normalized the baseline EPSCs to 100%. As is clear, both in the single experiment (Figure 5C) and in the summary (Figure 5E), the magnitude of LTP in cells expressing paAIP2 is approximately twice that of control cells. This interaction between the constitutive potentiation and LTP is consistent with their sharing the same underlying mechanism. Taken together these results strongly support the conclusion that while the animal is alive, LTP is acquired and leaves a synaptic memory trace.

Blocking CaMKII erases LTP

The experiments presented thus far are all consistent with the notion that synapses acquire LTP that contributes to synaptic transmission. If this is correct, then inhibiting CaMKII after LTP induction must reverse the potentiation. Yet, as discussed in the introduction, this experiment has failed on numerous occasions (Buard et al., 2010; Chen et al., 2001; Malinow et al., 1989; Murakoshi et al., 2017; Otmakhov et al., 1997), but see Feng, 1995. Thus, it was mandatory that we revisit this crucial experiment. We therefore carried out a ‘two pathway experiment’, in which the responses from two independent pathways were recorded from a single cell (Figure 6A). Following the recording of baseline responses, LTP was induced in one of the pathways (Figure 6B, red circles) by pairing depolarization of the cell to 0 mV with continued synaptic stimulation, while in the other control pathway, stimulation was stopped during the depolarization (Figure 6B, black circles). Once LTP had been established, myr-CN27 was applied and the effect on both pathways recorded. Myr-CN27 caused a complete reversal of established LTP and as expected, a ~50% reduction in the control pathway (Figure 6B). A summary of all the experiments (Figure 6B, n = 11) demonstrates that in the presence of myr-CN27 the LTP pathway and control pathway converge to approximately 50% of the control. This experiment establishes the essential role of CaMKII in maintaining LTP. A previous field potential study using tatCN21 (5 μM) failed to observe a reversal of LTP (Buard et al., 2010). We, therefore, repeated our two pathway experiment using field potential recording. Again, we observed that myr-CN27 (1 μM) caused a slow reduction the EPSPs in the control pathway and a reversal of LTP (Figure 6—figure supplement 1). Give that tatCN21 (5 μM) had no effect on synaptic transmission (Buard et al., 2010; Figure 1—figure supplement 1), it is not surprising that it failed to reverse LTP in the previous study (Buard et al., 2010).

Figure 6 with 1 supplement see all
Myr-CN27 erases long-term potentiation (LTP) in acute slices.

(A) Cartoon diagram of two pathway experiment. To record the response from two independent pathways, two bipolar stimulating electrodes were positioned to either side of the recorded cell with a distance of around 100 μm. Stimuli were applied alternately every 20 s. (B) A sample experiment showing that myr-CN27 inhibited the LTP pathway more strongly than the control pathway. (C) Left, summary data showing that myr-CN27 reduced the control pathway (black circles) 50%, while completely reversing LTP (red circles) (the difference between control and LTP pathway at 60 min: n = 11, p > 0.05, two-tailed Wilcoxon Signed Rank Test). Responses are mean ± standard error of the mean (SEM). Right, sample traces showing the effect of myr-CN27 on AMPA EPSCs in control and LTP pathway. LTP is induced by 2 Hz stimulation for 90 s, while holding the cell at 0 mV.

Figure 6—source data 1

Myr-CN27 erases long-term potentiation (LTP) in acute slices.

In this dataset, the results of the effects of myr-CN27 on LTP and control pathways are included.

https://cdn.elifesciences.org/articles/60360/elife-60360-fig6-data1-v2.xlsx

Discussion

The role of CaMKII in the maintenance of LTP and information storage has been one of the most vexing issues in the field of synaptic plasticity. On the one hand, the biochemical properties of CaMKII have for decades made this molecule an extremely attractive candidate for molecular storage (Bhattacharyya et al., 2020; Coultrap and Bayer, 2012; Hell, 2014; Lisman et al., 2002; Lisman et al., 2012). On the other hand, numerous physiological experiments over the years have failed to support the role of CaMKII in the maintenance of LTP and by extension memory. To reevaluate the physiological role of CaMKII in synaptic memory, we have used two classes of CaMKII inhibitory peptides. With these inhibitors we find that CaMKII is required for the maintenance of LTP and provide evidence that LTP acquired while the animal was alive leaves a lasting synaptic memory trace.

The first class of inhibitory peptide we used is derived from an endogenous CaMKII inhibitory protein referred to as CaMKIINtide (CN27) (Chang et al., 1998; Goold and Nicoll, 2010; Pellicena and Schulman, 2014; Vest et al., 2007). The second class of inhibitory peptide is derived from the autoinhibitory domain of CaMKII (Bayer et al., 2001; Leonard et al., 1999). Although these inhibitory peptides were thought to bind to separate sites (S and T sites), recent structural studies of the binding of GluN2B and other interacting peptides to CaMKII indicate that these peptides use similar interactions to bind across the substrate binding pocket of the CaMKII active site (Özden et al., 2020). Thus, at present it is not possible to use these two classes of inhibitors to distinguish between the disruption of GluN2B binding and the blockade of kinase activity. These peptides included a myristoylated version of CN27 (myr-CN27), AIP (myr-AIP) and a recently developed photoactivatable peptide inhibitor (paAIP2) (Murakoshi et al., 2017). Both classes of peptides block kinase activity with high affinity (Chang et al., 2001; Ishida et al., 1998), but also interfere with the binding of CaMKII to GluN2B (Sanhueza et al., 2011; Vest et al., 2007).

Our experiments clearly establish that inhibitory peptides fully reverse LTP. How long is CaMKII responsible for maintaining LTP? Experiments addressing the role of constitutive CaMKII in the maintenance of LTP are constrained by the length of recording; approximately 1 hr for whole cell and a few hours for field potentials. However, if LTP is the substrate for memory, it should leave a lasting memory trace at synapses, acquired while the animal is alive (Lisman, 2017; Sanhueza and Lisman, 2013). A number of studies have reported a lasting depression following transient inhibition of CaMKII (Barcomb et al., 2016; Gouet et al., 2012; Sanhueza et al., 2011; Sanhueza et al., 2007). Furthermore, even under basal conditions, CaMKII is found in isolated PSDs (Petersen et al., 2003; Strack et al., 1997) and synaptic puncta (Bayer et al., 2006) and this is in the autophosphorylated state (Strack et al., 2002; Trinidad et al., 2005). As discussed above all three inhibitory peptides (myr-CN27, myr-AIP, and paAIP2) depress synaptic transmission and this depression requires the presence of CaMKII. A number of trivial explanations, such as ongoing constitutive Ca2+ activation of the enzyme, were excluded.

We then focused on the possibility that this constitutive action of CaMKII reflects prior LTP. First, the finding that the inhibition does not affect NMDAR responses is consistent with an LTP mechanism, since LTP preferentially enhances AMPAR responses (reviewed in Nicoll, 2017). Second, there should be no recovery from the transient inhibition of CaMKII, analogous to the resetting of a molecular switch. This is, indeed, the case. Third, cells in which the NMDAR has been deleted embryonically should be devoid of the constitutive activity. Again, this is the case. This is unlikely to be due to the loss of GluN2B subunit, because cells expressing a pore dead NMDAR mutant, in which assembles with GluN2B, had little CaMKII constitutive activity. This finding suggests that non-NMDAR sources of Ca2+ in the behaving animal are incapable of triggering CaMKII-dependent enhancement in synaptic transmission. This tight linking of NMDAR sources of Ca2+ to CaMKII is critical, because it prevents the degrading of the Hebbian nature of plasticity. Finally, it is well established that LTP is saturable. Thus, if constitutively action CaMKII represents LTP, removing this component should allow for larger LTP. We find that in cells in which CaMKII is transiently blocked LTP is approximately twice as large as in neighboring control cells. Taken together these results suggest that constitutive CaMKII represents LTP acquired at synapses while the animal was alive, thus supporting a role for CaMKII in synaptic memory.

Based on the present results the term ‘basal synaptic transmission’ needs to be reevaluated. It has generally been assumed that the synapses studied in a hippocampal slice, in which much of the afferent drive from multiple inputs have been removed in the slicing, are at a ground ‘basal’ state. The present results indicate that the synaptic currents we measure are actually maintained by a persistent enhancement acquired prior to slicing.

It should be pointed out that the contribution of CaMKII to synaptic transmission is unlikely to be a simple addition to the overall preexisting excitatory drive onto the cell. It has long been argued that such a scenario would be unstable and quickly saturate (Bienenstock et al., 1982; Cooper and Bear, 2012; Fusi et al., 2005; Morrison et al., 2008; Toyoizumi et al., 2014; Turrigiano, 2008). A number of well-established mechanisms exist to maintain stability, including heterosynaptic depression (Scanziani et al., 1996), NMDAR-dependent long-term depression (Malenka and Bear, 2004) and on a longer time scale cell wide homeostasis (Turrigiano, 2008). Thus, it is envisaged that the CaMKII memory trace is embedded in a network of synapses with no overall net change in the excitatory drive onto the cell or network.

It is striking to compare the action of paAIP2 on LTP induction (Murakoshi et al., 2017) and on synaptic transmission (present study). A brief light exposure (~1 min), delivered immediately prior to inducing LTP, is sufficient to prevent LTP (Murakoshi et al., 2017). This finding indicates that the blocking of CaMKII activation by paAIP2 is very rapid. In contrast, the effect on synaptic transmission and the reversal of LTP required 10’s of minutes. What might explain the dramatic difference in the kinetics of inhibition of LTP induction compared to LTP maintenance and constitutive CaMKII? There is a well-accepted sequence of events for the role of CaMKII in the induction of LTP (Coultrap and Bayer, 2012; Hell, 2014; Lisman et al., 2012; Nicoll, 2017). Following NMDAR activation Ca2+ binds to CaM, which then binds to CaMKII activating the enzyme. This causes the autophosphorylation of T286 and the translocation of cytosolic CaMKII to the PSD where it binds to the C-tail of GluN2B. It seems reasonable to expect that under baseline conditions much of CaMKII and paAIP2 are freely diffusible in the spine cytoplasm, as are phosphatases. The activation of paAIP2 would quickly inhibit CaMKII autophosphorylation and its recruitment to the PSD and phosphatases would quickly reverse its effect. During baseline transmission and during the maintenance of LTP a fraction of CaMKII is bound to GluN2B. Evidence suggests that the CaMKII/GluN2B complex is protected from phosphatases (Cheriyan et al., 2011; Lisman and Raghavachari, 2015; Mullasseril et al., 2007). Thus, reversing action of this sequestered CaMKII would be more difficult than preventing phosphorylation of cytosolic CaMKII.

The present results could shed light on the underlying mechanism of long-term depression (LTD). Although this topic remains contentious (Collingridge et al., 2010; Dore et al., 2016; Goodell et al., 2017; Lisman and Zhabotinsky, 2001; Malenka and Bear, 2004; Stein et al., 2021; Wong and Gray, 2018), the long-held view is that LTD is mediated by the activation of phosphatases, in particular protein phosphatase 1 (PP1) (Mulkey et al., 1994; Mulkey et al., 1993). One of the limitations to this model is the lack of evidence that synaptic transmission is maintained by kinase activity. Our demonstration that CaMKII maintains synaptic transmission provides a very simple model for LTD, that is, a depotentiation or reversal of this prior LTP. One of the features of LTD is that it takes many minutes for its induction. Our finding that it takes many minutes for CaMKII inhibitors to depress synaptic transmission provides an explanation for this property. A formal model for the interplay between CaMKII and PP1 was proposed some time ago (Lisman and Zhabotinsky, 2001). The absence of LTD in CaMKII knockout mice (Coultrap et al., 2014; Stevens et al., 1994) supports such a model, but does not rule out other models.

What might account for the failure of previous studies to detect a component of synaptic transmission driven by constitutive CaMKII (Achterberg et al., 2014; Buard et al., 2010; Chen et al., 2001; Feng, 1995; Giese et al., 1998; Malinow et al., 1989; Murakoshi et al., 2017; Otmakhov et al., 1997; Silva et al., 1992; Wang and Kelly, 1996), but see Hinds et al., 1998, or to detect a role of constitutive CaMKII in maintaining LTP (Buard et al., 2010; Chen et al., 2001; Malinow et al., 1989; Murakoshi et al., 2017; Otmakhov et al., 1997), but see Feng, 1995? We suggest three factors. The first concerns the duration of peptide application, which in our hands takes 10’s of minutes to act, especially with field potential recording. The second concerns the concentration and time. As discussed above it is more difficult to reverse the constitutive action of CaMKII, than it is to block its activation. Finally, the membrane permeabilizing agent is important. In our hands myristoylation was more effective and specific than the CPP, tat.

In summary our results overcome many of the obstacles that have prevented embracing CaMKII as a molecular storage device. Specifically, our results show that transient inhibition of CaMKII results in a lasting depression of synaptic transmission with properties consistent with the erasure of prior LTP acquired while the animal was alive. Furthermore, the finding that CaMKII inhibition after the induction of LTP reverses LTP establishes its role in LTP maintenance. Thus, these physiological results compliment the rich biochemical literature on CaMKII, making CaMKII a particularly attractive molecular storage device. It is important to note, that our results have focused on LTP and it remains open as to whether CaMKII actually stores memories. Recent experiments (Rossetti et al., 2017) using the viral expression of the catalytically dead CaMKII K42M mutant, presumed to act in a dominant negative manner, support its role in memory.

Materials and methods

Animals

All the experimental procedures on animals were approved by the UCSF Animal Care and Use Committee, BUA # BU002466-04C. For acute slice recordings, typically we use 5–10 animals to obtain complete dataset; for culture slice recordings, we usually used 5–7 animals to obtain complete dataset.

Experimental constructs and chemical agents

Request a detailed protocol

The paAIP2 plasmid was obtained from Dr. Ryohei Yasuda and had been characterized (Murakoshi et al., 2017). The CRISPR construct targeting at CaMKIIα was previously characterized (Incontro et al., 2018). For biolistic experiments, all the plasmids were expressed in pCAGGS vector, which contains an internal ribosome entry site (IRES) followed by the fluorophore GFP. myr-CaMKIINtide (myr-CN27) and myr-AIP were purchased from Calbiochem. Inc (catalog#, 208921; catalog#, 189482). MK-801 maleate was purchased from HelloBio. Inc (catalog#, HB30004). Myr-CN21 was custom ordered from Elim Biopharm, Inc by myristoylating the N-terminal of CN21:

KRPPKLGQIGRSKRVVIEDDR amino acid sequence.

Slice culture and biolistic transfection

Request a detailed protocol

Hippocampal cultured slices are obtained from 6- to 8-day-old rats (Stoppini et al., 1991). Biolistic transfection was done 1 day after sectioning, by using a Helio Gene Gun with 1 μm DNA-coated gold particles (BioRad). Slices were maintained at 34°C and the medium was changed every 2 days. Typically, slices are used for electrophysiological recording 6–8 days after transfection, except for CRISPR experiments, in which slices are maintained for another week before recording.

In utero electroporation

Request a detailed protocol

E15.5 pregnant mice were anesthetized with 2.5% isoflurane in O2 and injected with buprenorphine for analgesia. The lateral ventricles of embryos were injected with 1 µl mix plasmid DNA (1 µg/µl) with beveled micropipette. Each embryo was electoporated with 5 × 50 ms, 35–40 V pulse.

Acute slice preparation

Request a detailed protocol

Mice of 2–3 weeks of age were anesthetized with 4% isoflurane, decapitated, and the brain dissected free. The whole brain was sliced into 300 µm slices in cutting solution as described (Granger et al., 2013); recovery at 34°C for half an hour and then stored at room temperature. Solutions were continuously gassed with 95% O2/5% CO2.

Photostimulation

Request a detailed protocol

Blue light pulses (0.1 Hz, 1 s, 20 mW/cm2) were generated by 473 nm blue DPSS laser (Shanghai Laser & Optics Century, BL473T8-300FC). The blue light from a laser was delivered through a optical patch cable connected to the optical fibers. Light pulses were controlled by a Master-8 (A. M. P. I). The blue light is applied for 1-s duration with interval of 10 s.

Injection of MK-801

Request a detailed protocol

MK-801 was dissolved in saline solution and administered as a single injection i.p. (10 mg/kg body weight) 1 hr before brain slicing.

Electrophysiological recording

Request a detailed protocol

Whole-cell voltage clamp recordings were obtained from either wild-type cells or fluorescent transfected pyramidal cell in CA1 region of hippocampus (Schnell et al., 2002). Where indicated dual recordings were made from control and transfected cells. Pyramidal neurons were identified by location and morphology. All recordings were made at 20–25°C. Internal solution (in mM): 135 CsMeSO4, 8 NaCl, 10 HEPES, 5 QX314-Cl, 4 Mg-ATP, 0.3 Na-GTP, 0.3 EGTA, 0.1 spermine. Osmolarity was adjusted to 290–295 mOm and pH was buffered at 7.3–7.4. External solution (mM): 119 NaCl, 2.5 KCl, 4 CaCl2, 4 MgCl2, 1 NaH2PO4, 26.2 NaHCO3, 11 glucose, bubbled continuously with 95% O2/5% CO2. For field recordings, the internal solution was 3 M NaCl with a large opening pipette tip. Synaptic currents were evoked every 10 s with bipolar stimulating electrodes placed in s. radiatum. To record EPSCs, picrotoxin (100 µM) was added to the external solution; for recording of AMPAR EPSCs, the cell membrane was held at −70 mV, while for NMDAR EPSCs, it was hold at +40 mV; LTP is induced by stimulating at 2 Hz for 90 s while clamping the cell at 0 mV. For two independent synaptic pathway experiments, two bipolar stimulating electrodes were positioned to each side of the recording electrode with a distance of around 100 µm and alternately stimulated every 20 s. Current responses were collected with a Multiclamp 700B amplifier (Axon Instruments), filtered at 2 kHz, and digitized at 10 kHz. Cells with series resistance larger than 20 MOhm were excluded from analysis.

Statistical analysis

Request a detailed protocol

Data analysis was carried out in Igor Pro (Wavemetrics), Excel (Microsoft), and GraphPad Prism (GraphPad Software). All paired recording data were analyzed statistically with a Wilcoxon Signed Rank Test for paired data. For unpaired data, a Mann–Whitney U test was used. Statistical parameters including the types of the statistical tests used, exact value of n, precision measures (mean ± standard error of the mean) and statistical significance are reported in the figure legends. All statistical tests performed were two sided, and with all tests a p value of <0.05 was considered statistically significant. All error bars represent standard error of the mean.

Data availability

All data generated or analysed during this study are included in the manuscript and supporting files.

References

Decision letter

  1. Nils Brose
    Reviewing Editor; Max Planck Institute of Experimental Medicine, Germany
  2. John R Huguenard
    Senior Editor; Stanford University School of Medicine, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Decision letter after peer review:

Thank you for submitting your article "Synaptic memory is maintained by persistent CaMKII activity" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by John Huguenard as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

As the editors have judged that your manuscript is of interest, but as described below that additional experiments are required before it is published, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

Summary

Tao et al. describe a study on an important issue in the field of nerve cell biology that has been unresolved for decades despite substantial research efforts: The role of CaMKII in the maintenance of LTP. By using various means to perturb CaMKII activity and activation, the authors provide evidence for the notion that CaMKII remains activated for extended periods of time after LTP induction and might thus be part of a synaptic memory mechanism.

Essential revisions

Revisions Requiring Additional Data/Experiments

1. The new data of the present study are in disagreement with a host of previously published studies, which indicated that CaMKII activation is generally rather short-lived and transient. Instead, the authors propose that a subfraction of CaMKII remains activated to store memory. The authors' main explanation for the data discrepancy (i.e. prolonged inhibition >30 min) is not fully convincing as some effects documented in the present study are seen earlier and within a time frame that others have also looked at. The reviewers are missing experimental approaches to resolve the striking data discrepancy, e.g. by comparing their assay conditions with those used by others.

2. No data to identify the persistently active CaMKII subfraction or its target proteins are presented, leaving the entire notion of memory-CaMKII still speculative. The manuscript would be strengthened substantially with direct evidence on the 'memory-CaMKII', e.g. biochemical proof of activated CaMKII. For instance, the authors could test their model of GluN2B/CaMKII interaction biochemically with CoIPs to identify the constitutively active CaMKII (e.g. through T286 phosphorylation, which should be absent with no LTP/memory). This experiment could be controlled with a GluN2B mutant that lacks the interaction.

3. The control experiment for no LTP is problematic. Knocking out NMDA receptors increases synaptic strength, as the authors showed previously. Furthermore, NR2B, which the authors propose in their model to be the scaffolding protein for active CaMKII, is not present. Thus, multiple changes occur in this experiment, of which the authors pick one as support for their claim that LTP is missing. The authors need to show how this control prevents the formation of the 'memory-CaMKII'. It seems that this is not possible with the documented experiment as synaptic GluN2B is removed, which is proposed to be the candidate interactor/scaffold of the 'memory-CaMKII'.

4. Figure 4 and S4: The run-up of the baseline has a similar slope as the increase after LTP induction, which is problematic. Peptide experiments should be performed with fEPSP recordings with 'solid' baseline rather than 3-point 'baseline'. In essence, it would be helpful to have different LTP protocols to show the same effect. Using cell-permeable peptides does not require whole cell recordings. With fEPSP, the issue of LTP wash-out as a rationale for the short baseline can be circumvented.

5. Figure 6: Compared to Figure 4, the LTP in control is rather reduced and not enhanced in paAIP2. It is not uncommon that effect sizes vary, but here it is used as the single line of evidence to indicate that the peptides create more 'room' for later LTP. The experiment needs to be better controlled, e.g. with interleaved experiments.

Revisions Requiring Text Redaction

6. The peptides used to perturb CaMKII function are probably among the best tools currently available. However, their specificity remains an issue. They have never been tested systematically but only on a handful of targets. CaMKIINtides have other 'structural' effects than the one on NR2B binding, i.e. interactions with other proteins, and AIP peptides are known to have effects on processes that are not CaMKII-related (Pellicena and Schulman – Front Pharmacol 2014). This issue should be included in the discussion as a general caveat.

7. There are no data to show that LTP accumulates during life. Such evidence would be important to support the model propose. Age and LTP 'experience' are not generally applicable criteria when switching back and forth between cultured and acute slices. In essence, one would not expect 'memories' in slice culture, but the effects of the peptides appear to be the same in the two culture systems. With regard to the NMDAR KO experiment, it is then argued that it prevents the formation of in vivo memories and so that no effect of the peptide inhibitors is seen. Thus, whether in vivo or in vitro, synapses appear to 'learn' the same amount. The line of logic in the discussion of these data is not entirely convincing. Given that showing that the corresponding synapses have been LTPed in vivo is highly non-trivial, it is suggested that alternative models and explanations are discussed.

8. Generally, and in the context of the point above, the reviewers suggest to substantially tone down the claim that a mechanism of memory is identified. CaMKII has both basic and potentiation-related synaptic function, which all might well overlap.

9. LTP is synapse specific. Erasing LTP by blocking the memory molecule would thus be expected to be synapse specific as well. However, both control and LTP pathway are affected by CaMKII perturbation. The authors argue that this is due to LTP accumulation during life. The support for this claim is limited. As the authors show, basal synaptic transmission is reduced by the CaMKII inhibitors. Whether this is LTPed synaptic strength is not clear. An alternative explanation would be that during early LTP expression, before consolidation, LTP can be reversed by inhibiting CaMKII. This should be discussed.

10. In CaMKII-α KO mice, LTP is impaired but not absent (Silva et al. – Science 1992). Furthermore, CaMKII autophosphorylation was proposed to facilitate LTP induction, but is not essential (Chang et al. – Neuron 2017). These results are consistent with an important role of CaMKII in LTP, but indicate that the magnitude of LTP blockade depends on the induction protocol and is more in agreement with induction and consolidation of LTP rather than long-term memory. This should be discussed.

11. Halt et al. (EMBO J 2012) proposed that memory consolidation requires the CaMKII/NR2B interaction. The effects of the inhibiting peptides after 10s of minutes, as seen in the present study, are consistent with this notion; whether long-term memory is affected, as proposed in the present study, is not shown. The major novelty of the present study would be that a fraction of CaMKII remains constitutively active to store memory. Reversal of LTP in 30-60 min is not necessarily related to long-term memory, and the notion that reduction of basal transmission reflects the same mechanism is not straight forward. This should be discussed. If more evidence in support of the memory-CaMKII idea can be provided (see above), additional experiments in the context of this item are not expected.

12. Figure 2: The effect of AIP expression appears less than 50%. A summary point of two different cell populations is incorrect (see e.g. Levy et al. – Neuron 2015; Kim et al. – Neuron 2015). The data need to be plotted differently and an explanation as to why the effect is smaller needs to be attempted.

13. Figure S5: The effect of myr-CN27 with MK-801 appears smaller than in other conditions. Does activity in slices before recording have an effect?

14. Line 36: It is unclear what data would support the notion that LTP accumulates during life. The corresponding passage should be referenced or rephrased.

15. Lines 76-78: This claim is incorrect. Halt et al. proposed a function for CaMKII in memory consolidation.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Synaptic memory is maintained by persistent CaMKII activity" for further consideration by eLife. Your revised article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by John Huguenard as the Senior Editor.

The manuscript has been improved very substantially, but there are several remaining issues that we regard to be important:

Data Interpretation and Discussion

1. CaMKII Enzymatic Activity vs. Protein Interactions. The firm conclusion that it is continued CaMKII activity rather than CaMKII binding to relevant postsynaptic binding partners (e.g. the C-terminus of GluN2B) is still not justified by the data. Notably, CaMKII binding to GluN2B was shown to be important for LTP (Barria and Malinow, 2005; Halt et al., 2012). Earlier evidence indicates that the CN27-related CN21 peptide reverses LTP by displacing CaMKII from GluN2B and not by inhibiting CaMKII but (Sanchueza et al., 2011). The authors argue that the AIP peptide only inhibits CaMKII activity without displacing it from GluN2B because earlier work showed that addition of untagged AIP during immunoprecipitation of CaMKII does not disrupt co-precipitation of the NMDAR at 20 μm (Leonard et al., 1999). However, although AIP does not disrupt the preformed CaMKII-NMDAR complex (Leonard et al., 1999), expression of it in HeLa cells prevents binding of CaMKII to GluN2B (see Suppl Figure 2 in Murakoshi et al., 2017). In other words, when expressed in intact cells, AIP does inhibit CaMKII binding to GluN2B. This could also be the case when myrAIP is applied to intact cells,. Perhaps in intact cells CaMKII has a tendency to unbind from GluN2B (potentially due to dephosphorylation of T286) and AIP augments this unbinding or prevents re-binding (potentially by inhibiting re-phosphorylation), while in vitro in immunoprecipitates there is no unbinding and AIP cannot displace CaMKII from GluN2B. The issue of AIP acting by inhibiting CaMKII activity vs. CaMKII binding to GluN2B is exacerbated by recently published data indicating that S and T site binding of substrates vs. docking proteins for CaMKII might not be separable (Özden et al., 2020: CaMKII binds both substrates 1 and effectors at the active site. bioRxivhttps://doi.org/10.1101/2020.10.25.354241). This work indicates that T and S sites form a continuum and that most – if not all – binding peptides will affect both, kinase activity and binding to anchoring proteins. Furthermore, the authors argue that their myrCN27 acts via CaMKII inhibition and not CaMKII displacement from GluN2B. Their reasoning is that they had a full effect at 1 μm of myrCN27 whereas Sanchueza et al. (2011) reported that 20 but not 5 μm of the related tat-CN21 displaced CaMKII from GluN2B (and that this correlated with a decrease in transmission and reversal of LTP during maintenance phase), when 5 μm tat-CN21 was sufficient to block CaMKII activity and LTP induction presumably because 5 μm is sufficient to block kinase activity. However, myrCN27 is longer than CN21 and the myr tag could promote binding of this peptide to CaMKII, so it could be much more potent than tat-CN21 in displacing CaMKII from GluN2B and already accomplish displacement at 1 uM.

Because of the considerations above, it is mandatory that the authors modify the respective discussion and do not dismiss the notion that 1 μm myrCN27 and AIP in general could have acted by displacing CaMKII.

2. In-Vivo-LTP. The authors argue that GluN1 KO prevents LTP expression in vivo and that this experimental approach could therefore test whether previous LTP in vivo is required for the dependence of synaptic strength on CaMKII. Indeed, the results are consistent with this line of argumentation. However, GluN1 KO has multiple additional effects: It increases synaptic strength and eliminates synaptic NR2B expression, which is thought to be the anchor for active CaMKII. Furthermore, multiple forms of synaptic plasticity (e.g. LTP, LTD, scaling, a.o.) occur during the lifetime of an animal, and notably, NMDARs, CaMKIIα, and CaMKIIαβ were reported to be involved in these forms of plasticity. Thus, the interpretation that in GluN1 KO neurons LTP is absent in vivo is reasonable, but given the other consequences of the KO, it is not conclusive. To further link the requirement of constitutively active CaMKII to in vivo LTP, the authors use a type of occlusion experiment. They express the photoactivatable AIP in neurons in utero. Then activate AIP for 40 min in slices after which they start recording. Like inhibition while stimulating, the inhibition without stimulation reduced AMPAR EPSC by 50% compared to neighboring control neurons. LTP in the previously inhibited neurons was about 2-times bigger than that in control neurons. The authors argue that the inhibition of CaMKII liberated "space" for more LTP. Given that LTP is saturable, the origin of this increased "space" is not clear. It could be due to a shift of the baseline further away from the ceiling or, as argued, to the reversal of in vivo LTP that is then newly triggered. Again, a suggestive interpretation, but not conclusive. In a final experiment, the authors report that 1 µM myr-CN27 reverts LTP during the expression phase. The experiment is performed in a two-pathway experiment, in which both the control non-induced pathway, as well as the induced pathway are reduced. Notably, the strength of the LTPed pathway settles at the same strength as the control pathway, indicating that both basal and the new LTPed synaptic strength is reduced. Myr-AIP is reported to have a similar effect at 20 µM, but not 1 µM. While the 1 µM serves as an important peptide control, the effect of the 20 µM is not as efficient as the 1 µM myr-CN27 as after around 60 min, some potentiation is left and EPSCs are not reduced to 50% of baseline. In a single experiment with field EPSP, myr-CN27 also reduces the fEPSPs of both the LTPed and control pathway, but in the time window observed, the responses do not converge. While these results indicate that CaMKII activity during expression is required, they do not necessarily translate into a mechanism selective for LTP, e.g. that the reduction in control pathway is based on in-vivo-acquired LTP. The different outcomes of the myr-CN27 peptide, the myr-AIP peptide in whole-cell recordings, and the myr-CN27 peptide in fEPSP are fully consistent with different mechanisms, e.g. with CaMKII activity being required for the expression of LTP, while structural assembly being required to stabilize basal synaptic transmission, as reported previously. To distinguish between these possibilities, information about on which CaMKII populations the different peptides operate would be important, e.g. whether myr-CN27 dissociates CaMKII from NR2B, or whether myr-AIP or myr-CN27 change the autoactivation of CaMKII.

In conclusion, the present study identifies an essential role of CaMKII activity for basal synaptic transmission and LTP expression. It is convincingly shown that this relates to differential effects of cell-penetrating peptides. However, the title and abstract appear too far-fetched as the results do not conclusively show that basal synaptic transmission is the product of acquired in vivo LTP. The reviewers urge the authors to address this issue by more conservative wording of title, abstract, and relevant text passages.

https://doi.org/10.7554/eLife.60360.sa1

Author response

Essential revisions

Revisions Requiring Additional Data/Experiments

1. The new data of the present study are in disagreement with a host of previously published studies, which indicated that CaMKII activation is generally rather short-lived and transient. Instead, the authors propose that a subfraction of CaMKII remains activated to store memory. The authors' main explanation for the data discrepancy (i.e. prolonged inhibition >30 min) is not fully convincing as some effects documented in the present study are seen earlier and within a time frame that others have also looked at. The reviewers are missing experimental approaches to resolve the striking data discrepancy, e.g. by comparing their assay conditions with those used by others.

This is a reasonable concern. We have rewritten the beginning of the Results section by discussing the background involved in our focusing on myr-CN27. We discuss the pros and cons of using cell permeabilizing peptides (CPPs) and protein lipidation, which we believe has an impact on our results. We also touch base with previous results with tatCN21. In addition, given that almost all of the previous data used field potential recordings, we repeated some of our key experiments with field potential recording. First, we tested myr-CN27 on baseline transmission. As shown in Author response image 1 (black symbols) the effect of myr-CN27 is considerably slower that seen with interleaved whole cell recording (blue symbols) and takes on average about 2 hours to stabilize. The most likely explanation for this is that with whole cell recording, cells near the surface are recorded, whereas with field potentials the peptide must diffuse throughout the slice.

Author response image 1

Second, we repeated the myr-CN27 “eraser” experiment with field potential recording (Figure 6—figure supplement 1).

2. No data to identify the persistently active CaMKII subfraction or its target proteins are presented, leaving the entire notion of memory-CaMKII still speculative. The manuscript would be strengthened substantially with direct evidence on the 'memory-CaMKII', e.g. biochemical proof of activated CaMKII. For instance, the authors could test their model of GluN2B/CaMKII interaction biochemically with CoIPs to identify the constitutively active CaMKII (e.g. through T286 phosphorylation, which should be absent with no LTP/memory). This experiment could be controlled with a GluN2B mutant that lacks the interaction.

We believe that this type of experiment has already been reported. In the classic paper of Seth Grant {Husi, 2000 #1497} he showed the CoIP of CaMKII with NMDARs. Furthermore {Leonard, 1999 #2681} showed that CaMKII/NMDAR complex exists under basal conditions. Finally, Halt et al. {Halt, 2012 #2216} showed that this interaction is reduced in their knockin mouse, in which the C-tail of GluN2B no longer binds to CaMKII. Since CaMKII must be in its active form to bind to GluN2B and that this interaction “locks” CaMKII in an active conformation {Bayer, 2001 #955}, we feel that this set of results does address the reviewer’s request. If the reviewer insists that we perform further biochemical experiments, we will be forced to withdraw the manuscript, since my lab is not currently capable of such experiments.

3. The control experiment for no LTP is problematic. Knocking out NMDA receptors increases synaptic strength, as the authors showed previously. Furthermore, NR2B, which the authors propose in their model to be the scaffolding protein for active CaMKII, is not present. Thus, multiple changes occur in this experiment, of which the authors pick one as support for their claim that LTP is missing. The authors need to show how this control prevents the formation of the 'memory-CaMKII'. It seems that this is not possible with the documented experiment as synaptic GluN2B is removed, which is proposed to be the candidate interactor/scaffold of the 'memory-CaMKII'.

The reviewer has a very good point. We had planned to address this issue by comparing the effect of deleting GluN1 to that of replacing GluN1 with a pore dead mutant form of GluN1 from Lonnie Wollmuth. If the pore dead mutant recues the defect our hypothesis would have to be discarded. Alternatively, the failure to rescue would argue for the need for a functional NMDA receptor. This negative result would be couple to immunocytochemistry to ensure that the GluN1 pore dead mutant colocalizes with PSD-95. Unfortunately, we have worked on this experiment for over six months. It has involved generating GluN1 forebrain KO mice and because of the pandemic and numerous technical problems we have been unsuccessful. We suggest two possible alternatives. First, we include the data and explain explicitly that an alternative explanation for the results is the loss of the GluN2B C-tail (current version). Or, if the reviewer insists, we can remove this data from the manuscript.

4. Figure 4 and S4: The run-up of the baseline has a similar slope as the increase after LTP induction, which is problematic. Peptide experiments should be performed with fEPSP recordings with 'solid' baseline rather than 3-point 'baseline'. In essence, it would be helpful to have different LTP protocols to show the same effect. Using cell-permeable peptides does not require whole cell recordings. With fEPSP, the issue of LTP wash-out as a rationale for the short baseline can be circumvented.

This is an important issue. We have studied LTP with whole cell recording for many years and there is definitely a trade off in terms of how much baseline one can get and the rate at which LTP washes out. In most of our experiments we carry out interleaved control experiments to assess the stability of the recordings. (Typically, there is a run up of ~30-50%) In other cases we use a control pathway to monitor stability. In the experiments in Figure 4 we have a control pathway. Both pathways show the same run up prior to LTP, but the control pathway remains stable during the recording until the peptide is applied. This, to me, is a tightly controlled experiment. In Figure S4 the experiments with the 20 μm Myr-AIP were obtained immediately after those with the lower dose, so they are obtained under the same conditions. As discussed above we have repeated some of our key experiments with field potential recording in which longer baselines are recorded.

5. Figure 6: Compared to Figure 4, the LTP in control is rather reduced and not enhanced in paAIP2. It is not uncommon that effect sizes vary, but here it is used as the single line of evidence to indicate that the peptides create more 'room' for later LTP. The experiment needs to be better controlled, e.g. with interleaved experiments.

The reviewer is correct in saying that the magnitude of LTP varies considerably among groups of experiments. Traditionally this variability is controlled for by carrying out interleaved experiments. In the present series of experiments, we have actually used simultaneous recordings from control and transfected cells in the same slice. I have studied LTP for 40 years and it is my belief that this is the most controlled experiment possible. I am not aware of other labs using simultaneous recordings to study LTP.

Revisions Requiring Text Redaction

6. The peptides used to perturb CaMKII function are probably among the best tools currently available. However, their specificity remains an issue. They have never been tested systematically but only on a handful of targets. CaMKIINtides have other 'structural' effects than the one on NR2B binding, i.e. interactions with other proteins, and AIP peptides are known to have effects on processes that are not CaMKII-related (Pellicena and Schulman – Front Pharmacol 2014). This issue should be included in the discussion as a general caveat.

Point well taken. We now mention that caution should be used in interpreting our results with these peptides (end of the first paragraph of the discussion).

7. There are no data to show that LTP accumulates during life. Such evidence would be important to support the model propose. Age and LTP 'experience' are not generally applicable criteria when switching back and forth between cultured and acute slices. In essence, one would not expect 'memories' in slice culture, but the effects of the peptides appear to be the same in the two culture systems. With regard to the NMDAR KO experiment, it is then argued that it prevents the formation of in vivo memories and so that no effect of the peptide inhibitors is seen. Thus, whether in vivo or in vitro, synapses appear to 'learn' the same amount. The line of logic in the discussion of these data is not entirely convincing. Given that showing that the corresponding synapses have been LTPed in vivo is highly non-trivial, it is suggested that alternative models and explanations are discussed.

This is an important issue and we agree with the reviewer that further discussion is warranted. I should point out that Lisman {Sanhueza, 2013 #2828;Lisman, 2017 #2794} has asserted that one of the predictions for LTP to be a memory molecular is that there is constitutive CaMKII activity due to prior LTP while the animal was alive. We agree with the reviewer and suggest that “acquired” is a better word than “accumulates”. What we would propose is that NMDAR-dependent Hebbian processes occur early on as neuronal networks are being established. We suggest that this process is occurring by the time that slice cultures are made and are maintained during the time in culture. The issue of whether the basal CaMKII represents prior LTP has been raised throughout. We have done the best that we can to try to link the basal activity to prior LTP. We would welcome any further suggestions to strengthen (or disprove) this link. However, since the link isn’t perfect, we have decided to reorganize the paper so that we first present all of our data characterizing the role of CaMKII in maintaining basal synaptic transmission and then present the data on LTP.

8. Generally, and in the context of the point above, the reviewers suggest to substantially tone down the claim that a mechanism of memory is identified. CaMKII has both basic and potentiation-related synaptic function, which all might well overlap.

We are very careful to refer to our results as “synaptic memory”, with the implication that it could underlie behavior memory. We now make this point more explicit in the discussion. In addition, we have reordered the paper, so that the linkage of the basal action of CaMKII to LTP is played down.

9. LTP is synapse specific. Erasing LTP by blocking the memory molecule would thus be expected to be synapse specific as well. However, both control and LTP pathway are affected by CaMKII perturbation. The authors argue that this is due to LTP accumulation during life. The support for this claim is limited. As the authors show, basal synaptic transmission is reduced by the CaMKII inhibitors. Whether this is LTPed synaptic strength is not clear. An alternative explanation would be that during early LTP expression, before consolidation, LTP can be reversed by inhibiting CaMKII. This should be discussed.

We believe that we have addressed this concern in our previous comments.

10. In CaMKII-α KO mice, LTP is impaired but not absent (Silva et al. – Science 1992). Furthermore, CaMKII autophosphorylation was proposed to facilitate LTP induction, but is not essential (Chang et al. – Neuron 2017). These results are consistent with an important role of CaMKII in LTP, but indicate that the magnitude of LTP blockade depends on the induction protocol and is more in agreement with induction and consolidation of LTP rather than long-term memory. This should be discussed.

I am not entirely clear as to how to address this. There is much variability in the field as to how complete the block of LTP is in the absence of CaMKII or its activity. Certainly no one would disagree that CaMKII is essential for most of NMDAR-dependent LTP. However, it is not entirely clear how this issue relates to constitutively active CaMKII.

11. Halt et al. (EMBO J 2012) proposed that memory consolidation requires the CaMKII/NR2B interaction. The effects of the inhibiting peptides after 10s of minutes, as seen in the present study, are consistent with this notion; whether long-term memory is affected, as proposed in the present study, is not shown. The major novelty of the present study would be that a fraction of CaMKII remains constitutively active to store memory. Reversal of LTP in 30-60 min is not necessarily related to long-term memory, and the notion that reduction of basal transmission reflects the same mechanism is not straight forward. This should be discussed. If more evidence in support of the memory-CaMKII idea can be provided (see above), additional experiments in the context of this item are not expected.

I am hesitant to focus on behavioral memory and terms like “memory consolidation” when discussing synaptic physiology. How the field defines “long-term” is open to discussion. In most studies of LTP, measurements are taken out to 45 min. – 1hour.

12. Figure 2: The effect of AIP expression appears less than 50%. A summary point of two different cell populations is incorrect (see e.g. Levy et al. – Neuron 2015; Kim et al. – Neuron 2015). The data need to be plotted differently and an explanation as to why the effect is smaller needs to be attempted.

Point well taken. We have removed the summary points from the scatter plot and have added a bar graph.

13. Figure S5: The effect of myr-CN27 with MK-801 appears smaller than in other conditions. Does activity in slices before recording have an effect?

We are not certain why the depression is somewhat less. We do not think that it is due to activity in the slice prior to the recording, because incubating slices from the outset in APV does not result in a smaller depression.

14. Line 36: It is unclear what data would support the notion that LTP accumulates during life. The corresponding passage should be referenced or rephrased.

This is a criteria put froward by John Lisman in many of his reviews e.g., {Sanhueza, 2013 #2828}. Throughout we have changed “accumulates” to “acquired”.

15. Lines 76-78: This claim is incorrect. Halt et al. proposed a function for CaMKII in memory consolidation.

I am unable to locate this in the text. What page and paragraph in the reviewer referring to?

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

The manuscript has been improved very substantially, but there are several remaining issues that we regard to be important:

Data Interpretation and Discussion

1. CaMKII Enzymatic Activity vs. Protein Interactions. The firm conclusion that it is continued CaMKII activity rather than CaMKII binding to relevant postsynaptic binding partners (e.g. the C-terminus of GluN2B) is still not justified by the data.

We accept that the crux of the issue is CaMKII Enzymatic Activity vs. Protein Interactions and we also accept that it is impossible to definitively distinguish between these two alternatives, using CN27 and AIP. Therefore, we remove that part of the discussion that argues for “activity” over “protein interactions”. We also removed the paragraph in the Discussion critiquing the Lee et al. results. Instead, we cite the recent Ozden et al. paper (pg. 12, paragraph 2,line 8 and immediately after this citation we state “Thus, at present it is not possible to distinguish between the disruption of GluN2B binding and the blockade of kinase activity.” We reiterate this on pg 14, paragraph 2, last 3 lines. We have removed “activity” from the title and throughout by replacing “CaMKII activity” with “CaMKII” or the “action of CaMKII”). We also abandon the T site/S site model that has featured prominently in the CaMKII literature, but now appears to have lost its value.

Notably, CaMKII binding to GluN2B was shown to be important for LTP (Barria and Malinow, 2005; Halt et al., 2012). Earlier evidence indicates that the CN27-related CN21 peptide reverses LTP by displacing CaMKII from GluN2B and not by inhibiting CaMKII but (Sanchueza et al., 2011). The authors argue that the AIP peptide only inhibits CaMKII activity without displacing it from GluN2B because earlier work showed that addition of untagged AIP during immunoprecipitation of CaMKII does not disrupt co-precipitation of the NMDAR at 20 μm (Leonard et al., 1999). However, although AIP does not disrupt the preformed CaMKII-NMDAR complex (Leonard et al., 1999), expression of it in HeLa cells prevents binding of CaMKII to GluN2B (see Suppl Figure 2 in Murakoshi et al., 2017). In other words, when expressed in intact cells, AIP does inhibit CaMKII binding to GluN2B. This could also be the case when myrAIP is applied to intact cells,. Perhaps in intact cells CaMKII has a tendency to unbind from GluN2B (potentially due to dephosphorylation of T286) and AIP augments this unbinding or prevents re-binding (potentially by inhibiting re-phosphorylation), while in vitro in immunoprecipitates there is no unbinding and AIP cannot displace CaMKII from GluN2B.

Since we are no longer making any special claims that AIP is acting differently from CN27, we don’t think that the issues raised by the reviewer are any longer relevant.

The issue of AIP acting by inhibiting CaMKII activity vs. CaMKII binding to GluN2B is exacerbated by recently published data indicating that S and T site binding of substrates vs. docking proteins for CaMKII might not be separable (Özden et al., 2020: CaMKII binds both substrates 1 and effectors at the active site. bioRxivhttps://doi.org/10.1101/2020.10.25.354241). This work indicates that T and S sites form a continuum and that most – if not all – binding peptides will affect both, kinase activity and binding to anchoring proteins. Furthermore, the authors argue that their myrCN27 acts via CaMKII inhibition and not CaMKII displacement from GluN2B. Their reasoning is that they had a full effect at 1 μm of myrCN27 whereas Sanchueza et al. (2011) reported that 20 but not 5 μm of the related tat-CN21 displaced CaMKII from GluN2B (and that this correlated with a decrease in transmission and reversal of LTP during maintenance phase), when 5 μm tat-CN21 was sufficient to block CaMKII activity and LTP induction presumably because 5 μm is sufficient to block kinase activity. However, myrCN27 is longer than CN21 and the myr tag could promote binding of this peptide to CaMKII, so it could be much more potent than tat-CN21 in displacing CaMKII from GluN2B and already accomplish displacement at 1 uM.

We agree that with myr-CN27 it is not possible to dissociate protein interactions from enzymatic activity, although it is worth pointing out that this is the argument Sanhueza et al. (2011) used to explain why 5 μm CN21 blocked LTP induction (activity), but did not reverse LTP (protein interaction).

Because of the considerations above, it is mandatory that the authors modify the respective discussion and do not dismiss the notion that 1 μm myrCN27 and AIP in general could have acted by displacing CaMKII.

We have removed this from the discussion.

2. In-Vivo-LTP. The authors argue that GluN1 KO prevents LTP expression in vivo and that this experimental approach could therefore test whether previous LTP in vivo is required for the dependence of synaptic strength on CaMKII. Indeed, the results are consistent with this line of argumentation. However, GluN1 KO has multiple additional effects: It increases synaptic strength and eliminates synaptic NR2B expression, which is thought to be the anchor for active CaMKII. Furthermore, multiple forms of synaptic plasticity (e.g. LTP, LTD, scaling, a.o.) occur during the lifetime of an animal, and notably, NMDARs, CaMKIIα, and CaMKIIαβ were reported to be involved in these forms of plasticity. Thus, the interpretation that in GluN1 KO neurons LTP is absent in vivo is reasonable, but given the other consequences of the KO, it is not conclusive.

While the loss of constitutive CaMKII activity following GluN1 deletion is consistent with prior in vivo LTP, the reviewer pointed out an equally valid conclusion. Since CaMKII binding to the NMDAR is essential for its synaptic action, this disruption could explain our result. To distinguish between these alternatives we have replaced GluN1 with a pore dead mutant. This subunit associates normally with GluN2 subunits and traffics normally to the membrane (Amin, et al., 2017). Thus, the only difference between wt synaptic GluN1/GluN2B synaptic receptors and the mutant is that it is pore dead. Thus, if any non-NMDAR sources of Ca2+ can activate CaMKII, it would bind the GluN2B and exert its synaptic action. The fact that the pore dead mutant failed to rescue constitutive CaMKII, suggests that in the freely moving animal non-NMDAR sources of Ca2+ are incapable of activating CaMKII, at least in terms of synaptic function. If there were to occur, it would degrade the Hebbian mechanism of plasticity. Aside from the NMDA KO experiment, the reviewer continues to express resistance to the proposal that the constitutive basal CaMKII action represents prior LTP when the animal was alive. I believe there is little doubt that CaMKII contributes to synaptic transmission (Sanhueza, 2007;Sanhueza, 2011;Barcomb, 2016;Gouet, 2012:Goold, 2010;Barcomb,2014;Incontro, 2018). We have done everything we can to test whether this constitutive activity is related to prior LTP. If this is incorrect, what do the reviewers propose is the basis for this activity? We welcome any suggestions the reviewer might have to either strengthen the connection or come up with an experiment that will permanently put this idea to rest. We love doing these kinds of Popper experiments. However, we emphasize that throughout the manuscript we use terms such as “suggests” and “is consistent with this proposal”. We make no claims that it is “conclusive”.

To further link the requirement of constitutively active CaMKII to in vivo LTP, the authors use a type of occlusion experiment. They express the photoactivatable AIP in neurons in utero. Then activate AIP for 40 min in slices after which they start recording. Like inhibition while stimulating, the inhibition without stimulation reduced AMPAR EPSC by 50% compared to neighboring control neurons. LTP in the previously inhibited neurons was about 2-times bigger than that in control neurons. The authors argue that the inhibition of CaMKII liberated "space" for more LTP. Given that LTP is saturable, the origin of this increased "space" is not clear. It could be due to a shift of the baseline further away from the ceiling or, as argued, to the reversal of in vivo LTP that is then newly triggered. Again, a suggestive interpretation, but not conclusive.

We are perplexed by the reviewers’ reticence to accept the conclusion of these experiments. The use of occlusion is not a new idea. It has been used by many investigators, and is generally accepted, in a variety of experiments studying LTP. Just one example is the linkage of the enhancement of CaMKII to LTP (e.g., Lledo et al., 1995), which is part of a figure in one of Lisman’s reviews (Lisman et al., Nat. Rev. Neurosci., 2002). It is also the same logic that Sanhueza et al. (2011) provids to conclude that tat21 reverses saturated LTP (end of the Results section).

In a final experiment, the authors report that 1 µM myr-CN27 reverts LTP during the expression phase. The experiment is performed in a two-pathway experiment, in which both the control non-induced pathway, as well as the induced pathway are reduced. Notably, the strength of the LTPed pathway settles at the same strength as the control pathway, indicating that both basal and the new LTPed synaptic strength is reduced. Myr-AIP is reported to have a similar effect at 20 µM, but not 1 µM. While the 1 µM serves as an important peptide control, the effect of the 20 µM is not as efficient as the 1 µM myr-CN27 as after around 60 min, some potentiation is left and EPSCs are not reduced to 50% of baseline. In a single experiment with field EPSP, myr-CN27 also reduces the fEPSPs of both the LTPed and control pathway, but in the time window observed, the responses do not converge. While these results indicate that CaMKII activity during expression is required, they do not necessarily translate into a mechanism selective for LTP, e.g. that the reduction in control pathway is based on in-vivo-acquired LTP. The different outcomes of the myr-CN27 peptide, the myr-AIP peptide in whole-cell recordings, and the myr-CN27 peptide in fEPSP are fully consistent with different mechanisms, e.g. with CaMKII activity being required for the expression of LTP, while structural assembly being required to stabilize basal synaptic transmission, as reported previously. To distinguish between these possibilities, information about on which CaMKII populations the different peptides operate would be important, e.g. whether myr-CN27 dissociates CaMKII from NR2B, or whether myr-AIP or myr-CN27 change the autoactivation of CaMKII.

We agree with the reviewers’ summary of the data and this is exactly the way we present it in the paper. However, we feel that the reviewer is over interpreting the results. First, we do not agree that this data can be interpreted as distinguishing the mechanism underlying CaMKII’s effects on basal synaptic transmission and LTP. We would argue that any difference in effects on baseline transmission and LTP reflect the effectiveness of the peptide, which varies with the depth of the recorded cell as well as other variables. We also explain that the use of peptides with field potential recording is limited because of the poor access of the peptides to the center of the slice. Thus, the fact that the LTPed pathway does not fall below baseline after one hour of peptide application in Figure S5, although it is still declining, cannot, in our opinion, be used to argue for different mechanisms.

In conclusion, the present study identifies an essential role of CaMKII activity for basal synaptic transmission and LTP expression. It is convincingly shown that this relates to differential effects of cell-penetrating peptides. However, the title and abstract appear too far-fetched as the results do not conclusively show that basal synaptic transmission is the product of acquired in vivo LTP. The reviewers urge the authors to address this issue by more conservative wording of title, abstract, and relevant text passages.

The conclusion that “the present study identifies an essential role of CaMKII activity for basal synaptic transmission and LTP expression” is most surprising. Most of the criticism of our manuscript is that it is impossible to distinguish between “activity” and “protein binding”. We agree and have removed “CaMKII activity” throughout the manuscript.

https://doi.org/10.7554/eLife.60360.sa2

Article and author information

Author details

  1. Wucheng Tao

    1. Key Laboratory of Brain Aging and Neurodegenerative Diseases, Fujian Medical University, Fuzhou, China
    2. Department of Cellular and Molecular Pharmacology, University of California, San Francisco, San Francisco, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2577-8161
  2. Joel Lee

    Department of Cellular and Molecular Pharmacology, University of California, San Francisco, San Francisco, United States
    Contribution
    Investigation
    Contributed equally with
    Xiumin Chen
    Competing interests
    No competing interests declared
  3. Xiumin Chen

    Department of Cellular and Molecular Pharmacology, University of California, San Francisco, San Francisco, United States
    Contribution
    Data curation, Investigation, Validation
    Contributed equally with
    Joel Lee
    Competing interests
    No competing interests declared
  4. Javier Díaz-Alonso

    Department of Cellular and Molecular Pharmacology, University of California, San Francisco, San Francisco, United States
    Present address
    Department of Neuroscience, University of California, Irvine, United States
    Contribution
    Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4980-7441
  5. Jing Zhou

    Department of Neurology, University of California, San Francisco, San Francisco, United States
    Contribution
    Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2809-7097
  6. Samuel Pleasure

    Department of Neurology, University of California, San Francisco, San Francisco, United States
    Contribution
    Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8599-1613
  7. Roger A Nicoll

    1. Department of Cellular and Molecular Pharmacology, University of California, San Francisco, San Francisco, United States
    2. Physiology, University of California, San Francisco, San Francisco, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review and editing
    For correspondence
    roger.nicoll@ucsf.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6977-4632

Funding

National Institutes of Health (MH070957)

  • Roger A Nicoll

The funders had no role in study design, data collection, and interpretation, or the decision to submit the work for publication.

Acknowledgements

We wish to thank the late John Lisman and Johannes Hell for their unfailing input during the course of this study. We thank Dr. R Yasuda for the paAIP2 plasmids, Dr. Wei Lu for contributing the data for Figure 3A and Dr. Lonnie Wullmuth for GluN1 pore dead plasmid (GluN1(N812W)). We thank members of the Nicoll lab for comments on the manuscript. We appreciate the technical assistance from Eric Dang. This research was supported by grant MH070957.

Ethics

All the experimental procedures on animals were approved by the UCSF Animal Care and Use Committee, BUA # BU002466-04C.

Senior Editor

  1. John R Huguenard, Stanford University School of Medicine, United States

Reviewing Editor

  1. Nils Brose, Max Planck Institute of Experimental Medicine, Germany

Publication history

  1. Received: June 24, 2020
  2. Accepted: December 14, 2021
  3. Accepted Manuscript published: December 15, 2021 (version 1)
  4. Version of Record published: January 28, 2022 (version 2)

Copyright

© 2021, Tao et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 3,249
    Page views
  • 685
    Downloads
  • 6
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Wucheng Tao
  2. Joel Lee
  3. Xiumin Chen
  4. Javier Díaz-Alonso
  5. Jing Zhou
  6. Samuel Pleasure
  7. Roger A Nicoll
(2021)
Synaptic memory requires CaMKII
eLife 10:e60360.
https://doi.org/10.7554/eLife.60360
  1. Further reading

Further reading

    1. Biochemistry and Chemical Biology
    2. Neuroscience
    Jinli Geng, Yingjun Tang ... Xiaodong Liu
    Research Article Updated

    Dynamic Ca2+ signals reflect acute changes in membrane excitability, and also mediate signaling cascades in chronic processes. In both cases, chronic Ca2+ imaging is often desired, but challenged by the cytotoxicity intrinsic to calmodulin (CaM)-based GCaMP, a series of genetically-encoded Ca2+ indicators that have been widely applied. Here, we demonstrate the performance of GCaMP-X in chronic Ca2+ imaging of cortical neurons, where GCaMP-X by design is to eliminate the unwanted interactions between the conventional GCaMP and endogenous (apo)CaM-binding proteins. By expressing in adult mice at high levels over an extended time frame, GCaMP-X showed less damage and improved performance in two-photon imaging of sensory (whisker-deflection) responses or spontaneous Ca2+ fluctuations, in comparison with GCaMP. Chronic Ca2+ imaging of one month or longer was conducted for cultured cortical neurons expressing GCaMP-X, unveiling that spontaneous/local Ca2+ transients progressively developed into autonomous/global Ca2+ oscillations. Along with the morphological indices of neurite length and soma size, the major metrics of oscillatory Ca2+, including rate, amplitude and synchrony were also examined. Dysregulations of both neuritogenesis and Ca2+ oscillations became discernible around 2–3 weeks after virus injection or drug induction to express GCaMP in newborn or mature neurons, which were exacerbated by stronger or prolonged expression of GCaMP. In contrast, neurons expressing GCaMP-X were significantly less damaged or perturbed, altogether highlighting the unique importance of oscillatory Ca2+ to neural development and neuronal health. In summary, GCaMP-X provides a viable solution for Ca2+ imaging applications involving long-time and/or high-level expression of Ca2+ probes.

    1. Neuroscience
    Maria Cecilia Martinez, Camila Lidia Zold ... Mariano Andrés Belluscio
    Research Article

    The automatic initiation of actions can be highly functional. But occasionally these actions cannot be withheld and are released at inappropriate times, impulsively. Striatal activity has been shown to participate in the timing of action sequence initiation and it has been linked to impulsivity. Using a self-initiated task, we trained adult male rats to withhold a rewarded action sequence until a waiting time interval has elapsed. By analyzing neuronal activity we show that the striatal response preceding the initiation of the learned sequence is strongly modulated by the time subjects wait before eliciting the sequence. Interestingly, the modulation is steeper in adolescent rats, which show a strong prevalence of impulsive responses compared to adults. We hypothesize this anticipatory striatal activity reflects the animals’ subjective reward expectation, based on the elapsed waiting time, while the steeper waiting modulation in adolescence reflects age-related differences in temporal discounting, internal urgency states, or explore–exploit balance.