CaMKII Autophosphorylation is Necessary for Optimal Integration of Ca²⁺ Signals During LTP Induction but Not Maintenance

SUMMARY

CaMKII plays a critical role in decoding calcium (Ca²⁺) signals to initiate long-lasting synaptic plasticity. However, the properties of CaMKII that mediate Ca²⁺ signals in spines remain elusive. Here, we measured CaMKII activity in spines using fast-framing two-photon fluorescence lifetime imaging. Following each pulse during repetitive Ca²⁺ elevations, CaMKII activity increased in a stepwise manner. Thr286 phosphorylation slows the decay of CaMKII and thus, lowers the frequency required to induce spine plasticity by several fold. In the absence of Thr286 phosphorylation, increasing the stimulation frequency results in high peak mutant CaMKII^T286A activity that is sufficient for inducing plasticity. Our findings demonstrate that Thr286 phosphorylation plays an important role in induction of LTP by integrating Ca²⁺ signals, and it greatly promotes, but is dispensable for the activation of CaMKII and LTP.

INTRODUCTION

Activation of CaMKII in dendritic spines is considered to be one of the first biochemical steps translating the information encoded in rapid Ca²⁺ elevations (~100 ms) into a long-lasting increase of synaptic strength during LTP and spine enlargement associated with LTP (structural LTP or sLTP) (De Koninck and Schulman, 1998; Lisman et al., 2012; Matsuzaki et al., 2004). CaMKII links Ca²⁺ elevations to downstream cellular targets responsible for LTP and sLTP by phosphorylating diverse substrates (Barria et al., 1997; Murakoshi et al., 2011; Opazo et al., 2010; Walkup et al., 2015; Yang et al., 2013). At its basal state, CaMKII is locked in an inactive state through the binding of a regulatory segment to its substrate binding site. Upon the influx of Ca²⁺, Ca²⁺/calmodulin (Ca²⁺/CaM) binds to CaMKII and induces its activation. The subsequent autophosphorylation at Thr286 in CaMKIIα (Thr287 in CaMKIIβ), which is located in the mouth of the regulatory segment, prevents the rebinding of the regulatory segment to the kinase domain. This permits an autonomous activity of CaMKII that is independent of Ca²⁺/CaM binding (Chao et al., 2011; Hanson et al., 1989; Hanson et al., 1994; Lisman et al., 2012). It has been hypothesized that this autonomous activity is the biochemical memory storage for LTP, learning and memory (Lisman et al., 2012). Consistent with the critical role Thr286 phosphorylation in LTP, knock-in mice in which Thr286 of CaMKIIα is mutated to Alanine (Camk2a^T286A) were impaired in LTP, learning and memory (Giese et al., 1998). However, it was found that the memory deficits in Camk2a^T286A mice can be overcome by intense training, suggesting that the phosphorylation may be not required for the maintenance of memory once it is established (Irvine et al., 2005). Furthermore, studies using pharmacological or optogenetic inhibitors of CaMKII demonstrated that blocking CaMKII activity before LTP induction abolishes LTP, but inhibiting CaMKII after LTP induction does not reverse LTP, inconsistent with the autonomous CaMKII hypothesis of biochemical memory storage (Buard et al., 2010; Chen et al., 2001; Malinow et al., 1989; Murakoshi et al., 2017; Otmakhov et al., 1997; Sanhueza et al., 2011), but also see (Sanhueza et al., 2007).

More recently, CaMKII activation was directly measured during sLTP of spines using two-photon fluorescence lifetime imaging microscopy (2pFLIM) or ratiometric FRET imaging microscopy in combination with CaMKII activity sensors, Camuiα or its variants (Fujii et al., 2013; Lee et al., 2009; Takao et al., 2005; Yagishita et al., 2014). These sensors measured the conformation change of CaMKII by fluorescence resonance energy transfer (FRET) between fluorophores attached to the two ends of CaMKIIα subunit, and their FRET efficiency was correlated with CaMKII activity (Lee et al., 2009; Takao et al., 2005). The imaging results showed that CaMKII activation rapidly occurs in ~10 s and returns to basal levels within ~1 min during sLTP, suggesting that CaMKII activation is transient (Lee et al., 2009; Yagishita et al., 2014). However, the exact role of Thr286 phosphorylation in CaMKII activation during spine plasticity still remains elusive.

Here, in order to elucidate the biochemical step of CaMKII activation during sLTP of spines, we improved the temporal resolution of fluorescence lifetime imaging to the millisecond range. We found that CaMKII functions as a leaky biochemical integrator of Ca²⁺ signals, and Thr286 phosphorylation critically shapes Ca²⁺ integration. In the absence of Thr286 phosphorylation, CaMKIIα^T286A activity decays ~3 fold faster and thus impairs the integration capability of CaMKII. This integration deficit, however, could be overcome by increasing the stimulation frequency which allows the mutant to reach a level sufficient for inducing structural and electrophysiological LTP. Our results indicate that Thr286 autophosphorylation is required for the induction of spine plasticity by facilitating Ca²⁺ integration mediated by CaMKII, but phosphorylation is dispensable for plasticity maintenance.

RESULTS

To investigate the millisecond temporal profile of CaMKII activation in accord with Ca²⁺ transients induced by two-photon glutamate uncaging, we optimized our setup to improve the temporal resolution of 2pFLIM up to 128 ms/frame (32 × 32 pixels) (Figure 1A–1B, S1).

Figure 1. CaMKII Activation Measured with Millisecond Temporal Resolution.

(A) Representative fluorescence lifetime images of Camuiα during glutamate uncaging at 0.49 Hz. Warmer colors indicate higher fluorescence lifetime of Camuiα, corresponding to the active, open conformation of Camuiα.

(B) Time course of fluorescence lifetime of Camuiα in (A) of the stimulated spine (black) and dendritic region (blue). Inset is expanded view of the rising phase of Camuiα activation. Black dots represent uncaging pulses.

(C) Averaged change in fluorescence lifetime of Camuiα (n = 36 spines/14 neurons). Left panel is expanded view of the rising phase of right panel. The orange curve indicates the decay kinetics of fluorescence lifetime signal obtained by curve fitting of a double-exponential function: F(t) = F₀ ▪ [P_fast ▪ e^−t/τ_fast + P_slow ▪ e^−t/τ_slow], where F₀ is the initial fluorescence lifetime, τ_fast and τ_slow are the fast and slow decay time constants and P_fast and P_slow are the respective populations. The time constants are obtained as τ_fast = 6.4 ± 0.7 s (P_fast = 74%) and τ_slow = 92.6 ± 50.7 s (P_slow = 26%).

(D) Representative fluorescence lifetime images of Camuiα in response to a single glutamate uncaging pulse.

(E) Time course of fluorescence lifetime of Camuiα in (D).

(F) Averaged change in fluorescence lifetime of Camuiα (n = 35 spines/8 neurons) in response to a single glutamate uncaging pulse. The orange curve indicates the kinetics of fluorescence lifetime signal obtained by curve fitting of a function: F(t) = [a ▪ e^−t/τ_fast + c] ▪ [1- e^−t/τ_rise], where c is a constant which represents the slow decay component as described in (C). The time constants are obtained as: τ_rise = 0.3 ± 0.1 s and τ_fast = 8.2 ± 1.7 s. All data are shown in mean ± s.e.m., and s.e.m. of time constants is obtained by bootstrapping. Scale bar, 1 µm.

In response to repetitive pulses of two-photon glutamate uncaging in the absence of extracellular Mg²⁺, Ca²⁺ was elevated to the micromolar level and decayed over ~100 ms (Figure S2). During a typical structural LTP (sLTP) induction protocol consisting of 30 pulses of glutamate uncaging at 0.49 Hz (Lee et al., 2009; Matsuzaki et al., 2004), we observed a rapid and robust activation of CaMKII in stimulated spines. CaMKII activity decayed over tens of seconds (Figure 1A, 1B). The averaged time course of CaMKII activation during sLTP (Figure 1C) showed stepwise increases following each glutamate uncaging pulse until plateauing within ~10 s. After the cessation of glutamate uncaging, CaMKII activity decayed with time constants of τ_fast = 6.4 ± 0.7 s (74%) and τ_slow = 92.6 ± 50.7 s (26%).

To clarify how CaMKII integrates Ca²⁺ signals, we measured CaMKII activation in response to a single glutamate uncaging pulse (Figure 1D-1F). The time course showed a rapid activation that peaked within 1 s and then decayed over tens of seconds. On average, the time constant of CaMKII activation was 0.3 ± 0.1 s, and the fast decay component was τ_fast = 8.2 ± 1.7 s (Figure 1F). Comparing this curve with CaMKII activation during sLTP induction (Figure 1C, Figure S4A), CaMKII appears to linearly summate its activity in response to repetitive glutamate uncaging pulses until reaching its plateau, suggesting that CaMKII activity is modulated by Ca²⁺ elevations and acts as a leaky integrator with time constants of ~6 s and ~1 min (Figure S4A).

The above experiments were performed with relatively long glutamate uncaging pluse (4 – 6 ms) at room temperature (25 ± 0.5°C). We found that a shorter duration of laser pulses (0.5 ms) activated CaMKII in a similar way (Figure S4A). At a near physiological temperature (34–35°C), CaMKII activity decayed faster (τ_fast = 1.8 ± 1.6 s (45%) and τ_slow = 11.0 ± 19.0 s (55%); Figure S4B, S4C). From these results, the temperature dependency of the decay kinetics in spines was determined to be Q₁₀ = 3.6 (τ_fast ) and 8.4 (τ_slow).

CaMKII activity is known to be regulated by autophosphorylation at Thr286 (Lisman et al., 2012). However, the role of Thr286 phosphorylation in the integration of Ca²⁺ signals remains unknown. To address this, we first used the Camuiα mutant in which the phosphorylation site is mutated to alanine (Camuiα^T286A), which occludes Thr286 phosphorylation. We compared the activation of the T286A mutant to that of Camuiα^WT in response to a single glutamate uncaging pulse (Figure 2A). The activity of CaMKIIα^T286A rapidly increased to a level similar to CaMKIIα^WT, but decayed with a single component that was ~3 fold faster than CaMKIIα^WT (τ_decay = 1.9 ± 0.3 s). These results indicate that Thr286 phosphorylation substantially slows the decay of CaMKII activity.

Figure 2. Activation of Camuiα^T286A and Camuiα^T286A During sLTP Induction.

(A) Activation of Camuiα^T286A (green) and Camuiα^WT (black) in response to a single glutamate uncaging pulse (black dot). The data and fitted curve for Camuiα^WT are from Figure 1F for comparison. The orange curve on Camuiα^T286A indicates the decay kinetics obtained by curve fitting of a function: F(t) = C ▪ [1 − e^−t/τ_rise] ▪ e^{−t/τ_decay}, where τ_rise is adapted from Figure 1F and is fixed during curve fitting (τ_rise = 0.3 s). The decay time constant for Camuiα^T286A is 1.9 ± 0.3 s (n = 30 spines/9 neurons). Inset is expanded view of the rising phase of Camuiα^T286A activation.

(B) Averaged fluorescence lifetime of Camuiα^WT , Camuiα^T286A , Camuiα^T286D , and Camuiα^{T286D/T305A/T306A} during sLTP induction. Right panel is fluorescence lifetime averaged over −4–0 s: Camuiα^WT: 1.75 ± 0.02 ns; Camuiα^T286A: 1.69 ± 0.03 ns; Camuiα^T286D: 2.06 ± 0.01 ns; Camuiα^{T286/T305A/T306A}: 2.10 ± 0.01 ns. Asterisks denote the statistical significance (p < 0.05; ANOVA followed by post hoc Bonferroni test).

(C–E) Averaged change in fluorescence lifetime of Camuiα^T286A (C; green), Camuiα^T286D (D; navy), Camuiα^{T286D/T305A/T306A} (E; light blue) and Camuiα^WT (black) in the stimulated spine during glutamate uncaging. The data and fitted curve for Camuiα^WT are from Figure 1C for comparison. Left panel is expanded view of the right panel. The orange curve on Camuiα^T286A is obtained by curve fitting of a function: F(t) = C ▪ e^{−t/τ_decay}. The decay time constant is obtained as 1.9 ± 1.2 s (26 spines/12 neurons). The orange curve on Camuiα^T286D and Camuiα^{T286D/T305A/T306A} are obtained by curve fitting of a function: F(t) = C ▪ e^{−t/τ_decay} + d₀, where d₀ is a constant. The decay time constant is obtained as 3.0 ± 0.6 s for Camuiα^T286D (33 spines/10 neurons) and 5.7 ± 0.5 s for Camuiα^{T286D/T305A/T306A} (32 spines/7 neurons). All data are shown in mean ± s.e.m., and s.e.m. of time constants is obtained by bootstrapping.

Next, we measured the activity of CaMKIIα^T286A during sLTP induction (glutamate uncaging at 0.49 Hz). In contrast to the activation of Camuiα^WT, Camuiα^T286A failed to accumulate its activity due to rapid decay kinetics (Figure 2B, 2C). CaMKII^T286A activity remained attenuated, activating and decaying in accord with each uncaging pulse (τ_decay = 1.9 ± 1.2 s; Figure 2C, Figure S4B). To determine whether activation of Camuiα^T286A might be affected by its incorporation into endogenous CaMKII holoenzyme with wild-type CaMKII subunits, we measured Camuiα^T286A activation in hippocampal slices from Camk2a^T286A knock-in mice. In this scenario, Thr286 of all subunits in the holoenzyme are mutated. We observed similar results when using hippocampal slices from Camk2a^T286A knock-in mice (τ_decay = 1.9 ± 0.2 s; Figure S4D, S4E), suggesting that the assembly with wild-type CaMKII subunits does not alter the kinetics of Camuiα^T286A.

To further investigate the role of Thr286 phosphorylation, we also imaged the activity of Camuiα^T286D phosphomimetic mutant, and Camuiα^{T286D/T305A/T306A}, which additionally prevents inhibitory autophosphorylation at Thr305 and Thr306 (Pi et al., 2010b; Shen et al., 2000). Consistent with its constitutively active nature, the basal fluorescence lifetime of Camuiα^T286D and Camuiα^{T286D/T305A/T306A} were higher than the maximum activation of Camuiα^WT (Figure 2B). This observation suggests pseudo-phosphorylation at Thr286 is sufficient to force CaMKII to adopt an open conformation in the absence of Ca²⁺ elevation. Furthermore, Camuiα^T286D and Camuiα^{T286D/T305A/T306A} showed a small, but significant activation by glutamate uncaging, presumably due to the binding of Ca²⁺/CaM to the subunit (Figure 2D, 2E). While the first uncaging pulse activated all Camuiα variants to a similar degree, the phosphorylation mutants showed markedly less integration of activity compared to WT. The decay time constants, which likely reflects Ca²⁺/CaM dissociation from pseudo-phosphorylated forms of CaMKII, were 3.0 ± 0.6 s for Camuiα^T286D (Figure 2D) and 5.7 ± 0.5 s for Camuiα^{T286D/T305A/T306A} (Figure 2E). The prolonged decay time of Camuiα^{T286D/T305A/T306A} is likely due to the enhanced Ca²⁺/CaM binding affinity, since Camuiα^T305A/T306A similarly exhibits prolonged decay (τ_fast = 11.8 ± 0.5 s (72%) and τ_slow = 255 ± 153 s (28%); Figure S4F). Together, these results suggest that the incapability of Camuiα^T286D to integrate Ca²⁺ signals is not due to the inhibitory phosphorylation (Figure S3). Instead, our results suggest that active Thr286 phosphorylation is a prerequisite for CaMKII to integrate Ca²⁺ transients during sLTP induction.

The above results suggest that the impaired CaMKIIα^T286A activation during sLTP induction results from the fast decay of Camuiα^T286A activity (τ ~1.9 s), which is comparable to the interval of glutamate uncaging (~2 s). If this is the case, uncaging intervals shorter than the decay time should result in a higher degree of accumulation of CaMKII activity. As expected, when we applied higher frequency glutamate uncaging stimulations (1.9 Hz and 7.8 Hz), we found that CaMKIIα^T286A activity increased and plateaued at a significantly higher level in the stimulated spines (Figure 3A-3C) as well as in adjacent dendrite shafts (Figure 3D-3F). The decay time of Camuiα^T286A was comparable between different stimulation frequencies: τ_decay = 1.9 ± 0.7 s (1.9 Hz for 30 pulses, Figure 3A), τ_decay = 1.5 ± 1.0 s (1.9 Hz for 120 pulses, Figure 3B), and τ_decay = 3.1 ± 1.0 s (7.8 Hz for 120 pulses, Figure 3C). In comparison, τ_decay in wild-type CaMKIIα showed longer decay at all frequencies: τ_fast = 6.9 ± 0.5 s (81%) and τ_slow = 127 ± 84 s (19%) (1.9 Hz for 30 pulses, Figure 3A), τ_fast = 7.8 ± 0.9 s (69%) and τ_slow = 192 ± 132 s (31%) (1.9 Hz for 120 pulses, Figure 3B), and τ_fast = 7.2 ± 1.0 s (59%) and τ_slow = 103 ± 31 s (41%) (7.8 Hz for 120 pulses, Figure 3C). Our results indicate that even in the absence of Thr286 phosphorylation, CaMKIIα^T286A activity can be boosted by increasing the frequency of Ca²⁺ transients to rates that allow for their integration by CaMKIIα^T286A.

Figure 3. CaMKII Activation in Dendritic Spines in Response to High-frequency Glutamate Uncaging.

(A–C) Averaged change in fluorescence lifetime of Camuiα^WT (black) and Camuiα^T286A (green) in response to high frequency glutamate uncaging at 1.9 Hz (A, B) or 7.8 Hz (C) for 30 pulses (A) or 120 pulses (B, C) in the dendritic spines. Insets are expanded view. The orange curves indicate the decay kinetics obtained by a double-exponential fitting for Camuiα^WT (see details in Figure 1C) and mono-exponential fitting for Camuiα^T286A (see details in Figure 2C). The obtained decay time constants are as follows (number of samples: spines/neurons): (A) Camuiα^WT: τ_fast = 6.9 ± 0.5 s (81%) and τ_slow = 127 ± 84 s (19%) (21/14). Camuiα^T286A: τ_decay = 1.9 ± 0.7 s (22/16). (B) Camuiα^WT: τ_fast = 7.8 ± 0.9 s (69%) and τ_slow = 192 ± 132 s (31%) (18/11). Camuiα^T286A: τ_decay = 1.5 ± 1.0 s (17/10). (C) Camuiα^WT: τ_fast = 7.2 ± 1.0 s (59%) and τ_slow = 103 ± 31 s (41%) (25/13). Camuiα^T286A: τ_decay = 3.1 ± 1.0 s (15/8).

(D–F) Averaged change in fluorescence lifetime in dendritic region as shown in (A–C), respectively. All data are shown in mean ± s.e.m., and s.e.m. of time constants is obtained by bootstrapping.

We next examined the role of Thr286 phosphorylation in sLTP. We measured structural plasticity in mEGFP-expressing CA1 pyramidal neurons of organotypic hippocampal slices prepared from Camk2a^T286A knock-in mice (Figure 4A-4G). We found that the sustained, but not the transient, phase of spine enlargement was significantly attenuated in Camk2a^T286A homozygous mice with the standard induction protocol (glutamate uncaging at 0.5 Hz, 30 pulses; ΔV_sustained = 16 ± 6% in Camk2a^T286A and ΔV_sustained = 89 ± 14% in littermate controls) (Figure 4A, 4B, 4F, 4G). Heterozygous mice exhibited a trend towards partial impairments in sLTP, although not statistically different (ΔV_sustained = 57 ± 11%) (Figure 4B, 4G).

Figure 4. Structural and Electrophysiological Plasticity Induced by High Frequency Stimulation.

(A) Fluorescence intensity images (mEGFP) of spine structural plasticity during sLTP. The arrowhead indicates the spot of two-photon glutamate uncaging. Scale bar, 1 µm.

(B–E) Structural LTP of neurons from Camka^T286A wild-type (WT/WT), heterozygous (WT/T286A) or homozygous (T286A/T286A) mice induced by glutamate uncaging at 0.5 Hz for 30 pulses (B), 2 Hz for 30 pulses (C), 2 Hz for 120 pulses (D) or 7.8 Hz for 120 pulses (E). Number of samples (spines/neurons) are 22/22 for WT, 27/26 for Het, and 25/25 for Homo in (B); 17/16 for WT, 18/18 for Het, and 12/12 for Homo in (C); 15/15 for WT, 13/13 for Het, and 14/14 for Homo in (D), and 15/15 for Homo, 20/19 for Het, and 16/16 for Homo in (E).

(F–G) Quantifications of spine volume change during the transient phase (F; peak value recorded at 1 min) or the sustained phase (G; averaged over 25–30 min). Asterisks denote the statistical significance (p < 0.05; ANOVA followed by post hoc Bonferroni test).

(H) Whole-cell recording of LTP induced at the Schaffer collateral in CA1 neurons from Camk2a^T286A and litter-mate control mice. LTP is induced by electrical stimulations at 2 Hz for 60 s or 40 Hz for 15 s with depolarization to 0 mV. Number of neurons is 19 for WT, 20 for T286A (2 Hz) and 27 for T286A (40 Hz). Typical EPSC traces before (-5–0 min) and after (43–63 min) are presented on top of the panel.

(I) Quantification of EPSC potentiation averaged over 40 – 63 min. Asterisks denote the statistical significance (p < 0.05; ANOVA followed by post hoc Bonferroni test). All data are shown in mean ± s.e.m.

If this impairment in sLTP is caused by the inability of CaMKIIα^T286A to integrate Ca²⁺ signals stimulated at this frequency, sLTP should be rescued by increasing the frequency of glutamate uncaging. Indeed, when sLTP is induced by 2 Hz glutamate uncaging (30 or 120 pulses), the partial sLTP impairment observed in heterozygous Camk2a^T286A mice was rescued (ΔV_sustained = 74.8 ± 14.9% for 2 Hz, 30 pulses and ΔV_sustained = 101.8 ± 21.4% for 2 Hz, 120 pulses; Figure 4C, 4D, 4G). Furthermore, structural plasticity in homozygous Camk2a^T286 mice was fully rescued by 7.8 Hz glutamate uncaging (for 120 pulses) (ΔV_sustained = 85.2 ± 10.4%) (Figure 4E and 4G).

To confirm that sLTP induced by high frequency glutamate uncaging stimulation protocols is CaMKII-dependent, we transfected slices prepared from CaMKIIα conditional knock-out mice (Camk2a^fl/fl) (Hinds et al., 2003) with tdTomato fused Cre recombinase (tdTM-Cre) and mEGFP (or tdTM and mEGFP as a control). We found that tdTM-Cre transfected neurons displayed significant impairment in the sustained phase of spine enlargement (averaged over 25–30 min; p < 0.05) at all frequencies, while the transient phase (peak value recorded at 1 min) was partially blocked (p < 0.05; Figure S5A, S5B, S5C), consistent with the reported necessity of CaMKII activity to support the sustained phase of, but not the transient phase of sLTP (Lee et al., 2009; Matsuzaki et al., 2004). These results demonstrate that structural plasticity induced by high frequency glutamate uncaging is still CaMKIIα dependent.

To exclude the possibility that the rescue of sLTP is caused by CaMKIIβ activity, we knocked-down endogenous CaMKIIβ by shRNA and measured sLTP induced by 7.8 Hz glutamate uncaging (Figure S5D–H). CaMKIIβ knock-down partially inhibited sLTP when 0.5 Hz glutamate uncaging was used, suggesting that CaMKIIβ also regulates sLTP (Figure S5E, S5F) (Borgesius et al., 2011; Kim et al., 2015). However, normal sLTP was induced by 7.8 Hz glutamate uncaging in CA1 neurons of Camk2a^T286A mice expressing CaMKIIβ shRNA (ΔV_sustained = 94 ± 20 %) (Figure S5G, S5H). Thus, sLTP induced with high frequency in Camk2a^T286A is CaMKIIβ-independent.

In addition, to examine the possibility that the rescue of sLTP is caused by altered Ca²⁺ dynamics in the mutant mice, we measured Ca²⁺ elevations during sLTP induction (Figure S2). We found that Ca²⁺ dynamics in Camk2a^T286A were similar to that in WT, indicating that Ca²⁺ transients are not affected in Camk2a^T286A mice. Taken together, the rescue of sLTP induced by high frequency glutamate uncaging is due to augmented CaMKIIα^T286A activity, by Ca²⁺ elevations elicited at higher frequency than CaMKIIα decay, in the stimulated spines.

Since it has been reported that overexpression of CaMKII^T286D affects spine structure and plasticity (Mayford et al., 1995; Pi et al., 2010a; Rotenberg et al., 1996), we quantified sLTP and basal spine size in neurons expressing Camuiα^T286D_. While 0.49 Hz glutamate uncaging failed to induce sLTP in Camuiα^T286D-expressing neurons, 7.8 Hz glutamate uncaging fully rescued sLTP (ΔV_sustained = 71.6 ± 14.9 %) (Figure S5I, S5J). In addition, we observed a ~70% increase in the basal spine volume in neurons expressing Camuiα^T286D (Figure S5K, S5L), consistent with a previous study (Pi et al., 2010a).

Lastly, we investigated whether the impaired electrophysiological LTP in Camk2a^T286A mice could be rescued by high frequency electrical stimulation. To test this, we prepared acute hippocampal slices from Camk2a^T286A knock-in mice or littermate controls (4–5 weeks). We performed whole-cell patch clamp recordings from CA1 pyramidal neurons and measured EPSCs in response to stimulation of Schaffer collateral inputs. LTP was induced by pairing low frequency synaptic stimulation (2 Hz for 15 s) with postsynaptic depolarization to 0 mV. Assuming a ~20% release probability (Branco and Staras, 2009), 2 Hz presynaptic electrical stimulation corresponds to ~0.5 Hz glutamate uncaging, our standard sLTP protocol. While robust LTP was induced in wild-type neurons by this protocol (ΔEPSC = 96 ± 16 %), Camk2a^T286A neurons showed impaired LTP (ΔEPSC = 18 ± 10 %) (Figure 4H, 4I), in agreement with previous studies (Giese et al., 1998). However, when we applied high frequency synaptic stimulation (40 Hz, 15 s; corresponding to ~8 Hz glutamate uncaging), LTP in Camk2a^T286A mice was rescued (ΔEPSC = 76 ± 20%; Figure 4H, 4I). LTP induced by 40 Hz stimulation was abolished by the NMDAR antagonist APV, suggesting it is NMDAR-dependent (Figure S6A, S6B), unlike some other forms of LTP (Villers et al., 2014). To confirm that LTP induced by 40 Hz synaptic stimulation is CaMKII-dependent, we co-injected AAV encoding Cre and floxed-tdTM to the lateral ventricle of Camk2a^fl/fl embryos. We prepared acute hippocampal slices at P21-P30, which were sparsely labeled with CaMKIIα KO CA1 hippocampal cells (see Supplemental Information). We found that LTP induced by 40 Hz synaptic stimulation was abolished in Cre- positive CA1 cells lacking CaMKIIα (p < 0.05; Figure S6C, S6D). Our results demonstrate that the impaired structural and electrophysiological LTP in CA1 neurons of Camk2a^T286A mice can be rescued by high frequency synaptic stimulation (paired with postsynaptic depolarization), which elicits peak activation of CaMKII^T286A. These results suggest phosphorylation at Thr286 significantly broadens the frequency domain over which CaMKII can integrate Ca²⁺ signals, and permits CaMKII to integrate Ca²⁺ signals at physiological relevant frequencies to induce synaptic plasticity.

DISCUSSION

In this study, we found that the rapid activation (~0.3 s) and the relatively slow decay (~6 s and ~1 min) of CaMKII underlie the stepwise activation, revealing its function as a leaky biochemical integrator of Ca²⁺ signals. The integration window of CaMKII (6–8 s) also defines the frequency of stimulation required for inducing plasticity (Matsuzaki et al., 2004).

Our observations clarify the critical role of Thr286 phosphorylation in plasticity: it lowers the stimulation frequency required to induce synaptic plasticity, and permits CaMKII to integrate Ca²⁺ signals at physiologically relevant frequencies. The decay time of the non-phosphorylated mutant CaMKIIα^T286A and phospho-mimetic mutant CaMKIIα^T286D was ~3- and ~2-fold faster than that of wild-type CaMKII, respectively. The faster decay time of the mutants substantially compromised the integration capability of CaMKII for the same stimulation frequency. This result suggests that Thr286 phosphorylation is required for optimal integration of Ca²⁺ signals (Figure 2). The fast decay of CaMKIIα^T286D activity, perhaps due to the dissociation of Ca²⁺/ CaM, suggests that dephosphorylation of Thr286 plays an important role in determining the optimum decay time (6 – 8 s) of CaMKII (but see also (Otmakhov et al., 2015)).

We also found that spine plasticity can be induced in Camk2a^T286A knock-in mice by increasing the frequency of stimulation. Our results indicate that peak CaMKIIα^T286A activity, achieved even without Thr286 phosphorylation, is able to relay Ca²⁺ signals to downstream molecules, such as Rho GTPase proteins, whose sustained activities have been shown to last > 30 min to support spine plasticity (Hedrick et al., 2016; Murakoshi et al., 2011; Nishiyama and Yasuda, 2015). Our findings suggest that Thr286 phosphorylation is dispensable for the maintenance of structural and elecrophysiological LTP, although prolonged Thr286 phosphorylation may occur during LTP under some conditions (Lengyel et al., 2004). Our results are consistent with previous studies demonstrating that inhibition of autonomous CaMKII activity inhibits LTP induction but does not impact LTP maintenance (Buard et al., 2010; Chen et al., 2001; Murakoshi et al., 2017). In addition, our results provide a molecular basis for the previous observation that Thr286 phosphorylation is required for rapid learning, but not memory (Irvine et al., 2005). Interestingly, structural plasticity can be also rescued by higher frequency glutamate uncaging in neurons expressing Camuiα^T286D. This result suggests that the net increase in CaMKII activity during sLTP induction is essential for synaptic plasticity (Bach et al., 1995; Mayford et al., 1995). Together, we provide evidence that CaMKIIα phosphorylation at Thr286 is not the biochemical trace of memory as previously suggested (Lisman et al., 2012). Instead, our findings demonstrate that Thr286 phosphorylation is essential for the optimal integration of Ca²⁺ signals, which is only required during plasticity induction.

STAR methods

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Ryohei Yasuda (ryohei.yasuda@mpfi.org).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

All experimental protocols were approved by Max Planck Florida Institute IACUC, Duke University Medical Center or National Institutes of Natural Sciences and meet the guidelines of the National Institutes of Health guide for the Care and Use of Laboratory Animals.

Animals

Mice from BL6/C57 strain (purchased from Charles River Laboratories, both males and females) were used as wild-type group in 2pFLIM measurements. Camk2a^T286A knock-in mice were from Dr. KP. Giese (Giese et al., 1998). CaMKIIα conditional knock-out mice (Camk2a^fl/fl, strain name: C57BL/6-Camk2a^tm1Vyb; deposited by Dr. S. Tonegawa) that carry a floxed exon 2 allele were from the Jackson Laboratory (Hinds et al., 2003). All experimental animals were bred in-house under the animal care and guidelines of Duke University Medical Center and Max Planck Florida Institute for Neuroscience.

Hippocampal Slices

Organotypic cultured hippocampal slices were prepared from postnatal 4–7 day mice (both mail and females (see a detailed protocol at (Stoppini et al., 1991)). Briefly, tissue slices were plated on cell culture inserts (hydrophilic PTFE, 0.4 µm, Millipore) and maintained in tissue medium (minimum essential medium eagle (MEM) 8.4 mg/ml, horse serum 20%, L-glutamine 1 mM, CaCl₂ 1 mM, MgSO₄ 2 mM, D-glucose 12.9 mM, NaHCO₃ 5.2 mM, HEPES 30 mM, insulin 1 µg/ml, ascorbic acid 0.075%) at 37°C supplemented with 5% CO₂ until experiments (DIV 12–19). Hippocampal slices were biolistically transfected with plasmids at DIV 5–10 (12 mg gold particle, size: 1 µm, 30 µg plasmid). Acute slices were prepared from postnatal 30–40 day Camk2a^T286A mice and littermates. Preparation of slice cultures was in accordance with the animal care and guidelines of Duke University Medical Center and Max Planck Florida Institute for Neuroscience.

Acute Slices

Mice (both male and females) were sedated by isoflurane inhalation, and perfused intracardially with a chilled choline chloride solution. Brain was removed and placed in the same choline chloride solution composed of 124 mM Choline Chloride, 2.5 mM KCl, 26 mM NaHCO₃, 3.3 mM MgCl₂, 1.2 mM NaH₂PO₄, 10 mM glucose and 0.5 mM CaCl₂, pH 7.4 equilibrated with 95% O2/5% CO₂. Coronal slices (250 µm) were prepared from Camk2a^T286A mice and wild-type littermates age between P30-P40.

Primary Neuronal Culture

Neocortex dissected from 4 newborn mice (postnatal day 0; Both males and females) were triturated and plated into 35 mm dishes coated with 50 µg/ml PLL (Sigma) in culture medium composed of basal medium Eagle (BME) supplemented with 10% heat-inactivated fetal bovine serum (Invitrogen), 35 mM glucose (Sigma), 1 mM Glutamax (Sigma), 100 U/ml penicillin (Sigma), and 0.1 mg/ml streptomycin (Sigma). Proliferation of non-neuronal cells was inhibited by Cytosine arabinoside (2.5 µM) treatment on the second day after plating.

Method Details

Plasmids

Molecular cloning and mutations were carried out using QuikChange site-directed mutagenesis kit (Agilent Technologies) and InFusion cloning kit (Clontech). Expression of Camuiα was under pCAG promoter. Camuiα (dimVenus-CaMKII-mEGFP or Green-Camuiα) was described previously (Lee et al, 2009). We used dimVenus (Venus_{A208K, T145W}) instead of sREACh, which has slightly better characteristics as FRET donor (Murakoshi et al., 2008), since this version has been already extensively characterized (Lee et al., 2009). shRNA against CaMKIIβ (Mus musculus) were purchased from GE-Dharmacon. pGIPZ Camk2b shRNA (TCAGAGAGGAACTTCCCAG; V2LMM2804) were used to knock-down Camk2b. pGIPZ non-silencing shRNA (CTTACTCTCGCCCAAGCGAGAG) were used as a control. pCMV-CaMKIIβ (Rattus) is from Dr. Hayashi (Okamoto et al, 2007).

Microscope

Fluorescent lifetime of Camuiα was measured by a home-built two-photon fluorescence lifetime imaging microscopy (2pFLIM). Camuiα was excited with a Ti:Sapphire laser tuned at 920 nm (Coherent, Chameleon) with laser power measured under the water immersion objective (Olympus, NA = 0.9, 60x) in the range of 1–1.5 mW, see detailed information at (Murakoshi et al., 2008; Yasuda et al., 2006). A second Ti:Sapphire laser at 720 nm (laser power measured under the objective: 2.5–3 mW), pulse duration of 4–6 ms was used to photolysis MNI-caged L-glutamate.

CaMKII activity imaging

Hippocampal slices were bathed in artificial cerebrospinal fluid (ACSF) bubbled with carbogen (95% O₂/ 5% CO₂) during the image recordings. Final ion concentrations (in mM) in imaging solution: NaCl 127, NaHCO₃ 25, D-glucose 25, KCl 2.5, NaH₂PO₄ 1.25, supplemented with CaCl₂ 4, MNI-caged L-glutamate (Tocris) 4, TTX 0.001, Trolox (Sigma) 1. Between DIV 12–19, we imaged individual transfected CA1 pyramidal neurons. Dendritic spines on the secondary and tertiary apical dendrites were used for imaging. Images were acquired by a home-built 2pFLIM microscope controlled by custom software (MatLab). Experiments were performed at 25 ± 0.5°C or 34–35 °C as indicated. The temperature was controlled with a control syringe heater and an inline solution heater (TC344C, SW-10/6 and SH-27B, Warner Instruments). Recordings were performed with 32×32 pixels (pixel size: 12.3 ± 1.72 pixel/µm) at 128 ms/frame (7.8 Hz).

2pFLIM data analysis

Fluorescence lifetime of mEGFP in Camuiα is affected by the FRET efficiency. Fluorescence lifetime of mEGFP in Camuiα has at least 3 populations: closed conformation (basal), open conformation (active), and mEGFP (donor) with unfolded dimVenus (acceptor). The third population can be regarded as a constant component throughout the measurement. Since Camuiα is a monomeric FRET pair sensor (1:1 ratio of mEGFP: dimVenus), the change of mean fluorescence lifetime of Camuiα (τ_m) reflects the change of FRET efficiency and thus the conformation change of Camuiα. The mean fluorescence lifetime of Camuiα (τ_m) was derived from the mean photon arrival time 〈t〉 as follows:

$${\tau }_m=\langle t\rangle -t_0=\frac{\int \mathit{\text{dt}}\cdot \mathit{\text{tF}}(t)}{\int \mathit{\text{dt}}\cdot F(t)}-t_0,$$

where F(t) is the fluorescence lifetime decay curve and t₀ is the instrumental time offset. We estimated t₀ by fitting to the fluorescence decay curve summing all pixels in all frames over a whole image session (typically 1024 frames) with a double exponential function convolved with the Gaussian pulse response function:

$$F(t)=F_0{\sum }_{i=1,2}{P}_{i}H(t,t_0,{\tau }_i,{\tau }_G),$$

where P_i is the population of fluorophore (mEGFP) with the fluorescence lifetime of τ_i, respectively, and H(t) is the fluorescence lifetime curve with a single exponential function convolved with the Gaussian pulse response function:

$$H(t,t_0,{\tau }_i,{\tau }_G)=\frac{1}{2}\text{exp}(\frac{{\tau }_G^2}{2{\tau }_i^2}-\frac{t-t_0}{{\tau }_i})\text{erfc}\left(\frac{{\tau }_{\mathrm{G}}^2-{\tau }_i(t-t_0)}{\sqrt{2}{\tau }_i{\tau }_G}\right)-t_0^{\prime },$$

in which τ_G is the width of the Guassian pulse response function, F₀ is the peak fluorescence before convolution, $t_0^{\prime }$ is the instrumental time offset measured by the fitting, and erfc is the error function. Since the mean fluorescence lifetime can be obtained from the fitting as

$${\tau }_m=\frac{{\sum }_i{P}_i{\tau }_i^2}{{\sum }_i{P}_i{\tau }_i}$$

t₀ can be calculated as:

$$t_0=\langle t\rangle -\frac{{\sum }_i{P}_i{\tau }_i^2}{{\sum }_i{P}_i{\tau }_i}$$

In our typical setting, t0 is ~ 1–2 ns. For Green-Camuiα, the parameters are typically, τ1 ~ 1.9–2.1 ns, τ2 ~ 0.4–0.6 ns, τG ~ 0.15 ns and P1 ~ 0.4–0.5 (P2 = 1 − P1). A change in the population (P1 and P2) by ~ 10% causes a change in the mean fluorescence lifetime (τm) by ~ 0.1 ns.. A change in the population (P₁ and P₂) by ~ 10% causes a change in the mean fluorescence lifetime (τ_m) by ~ 0.1 ns. It should be noted that changes in the fluorescence lifetime are independent of the fitting.

Measurements of structural plasticity

Two-photon laser scanning microscopy (2pLSM) was used to quantify the spine volume change during glutamate uncaging of sLTP (Matsuzaki et al., 2004). When the experiments were performed in Camk2a ^T286A knock-in mice, hippocampal slices were cultured from Camk2a^T286A/T286A (homozygous), Camk2a^T286A/WT (heterozygous), and wild-type littermates. Hippocampal slices were biolistically transfected with mEGFP (12 mg gold particle, size: 1 µm, 30 µg plasmid) at DIV 7–10. Transfected CA1 neurons were imaged between DIV 12–17. For the experiments performed in Camk2a^fl/fl mice, the transfected plasmids are as follows: 1) conditional knock-out group: co-transfected with tdTomato labeled Cre recombinase (pCAG-tdTM-cre, 25 µg) and mEGFP (pCAG-mEGFP, 20 µg); 2) control group: co-transfected with tdTomato (pCAG-tdTM, 25 µg) and pCAG-mEGFP (20 µg). Transfected CA1 neurons with Cre positive cells were excited with Ti:Sapphire laser at 920 nm (Coherent, Chameleon) with laser power measured under the objective (Olympus, NA = 0.9, 60x) in the range of 1–1.5 mW. Green-channel fluorescence collected from PMT (photoelectron multiplier tubes) placed after the 510 nm/70 nm (bandwidth) wavelength filters were used for spine volume change analysis.

Calcium imaging

We performed whole-cell patch clamp to hippocampal cultured neurons from Camk2a^T286A mice and wild-type littermates with the pipette containing Fluo-4FF (Ca²⁺ indicator) and Alexa-594 in Potassium Gluconate internal solution (in mM: K gluconate 130, Na phosphocreatine 10, MgCl2 4, Na₂ATP 4, MgGTP 0.3, L- Ascorbic acid 3, HEPES 10, pH 7.4, 310 mosm). 2pLSM and Ti:Sapphire laser at 920 nm (Coherent, Chameleon) was used to simultaneously excite the two fluorescent dyes and thus the fluorescence from Fluo-4FF can be used to quantify the [Ca²⁺] transients during glutamate uncaging of sLTP. Images were acquired every 64 ms.

AAV injection

E14.5/15.5 timed-pregnant Camk2a^fl/fl mice were anesthetized with ~2% isoflurane and administered 0.1 mg buprenorphine SR (ZooPharm) for analgesia. The uterine horns were exposed through an abdominal incision and the right lateral ventricle of each embryo was injected with approximately 1–2 µl of AAV solution. Half of the embryos were injected with AAV1.CAG.EGFP, and half were injected with AAV1.CAG.Flex.tdTomato/ AAV1.hSyn.Cre (all from U Penn vector core).

Electrophysiology

Slices were maintained in a submerged chamber at 32 °C for 1h and then at room temperature in oxygenated ACSF. Whole cell recordings and LTP protocol: CA1 pyramidal neurons were visualized using oblique illumination. Whole cell recordings were performed using a patch-clamp amplifier (Multiclamp 700B, Molecular Devices). Patch pipettes (3–6 ΩM) were filled with a K Gluconate solution (130 mM K gluconate, 10 mM Na phosphocreatine, 4 mM MgCl₂, 4 mM Na₂ATP, 0.3 mM MgGTP, 3 mM L-Ascorbic acid, 10 mM HEPES, pH 7.4, 310 mosm). Series resistances (10 to 40 MΩ) and input resistances (100 to 300 MΩ) were monitored throughout the experiment using negative voltage steps (-5 mV, 50 ms). The membrane potential was held at −70 mV. Experiments were performed at room temperature and slices were perfused with oxygenated ACSF (127 mM NaCl, 2.5 mM KCl, 10 mM glucose, 10 mM NaHCO₃, 1.25 mM NaH₂PO₄, 2 mM MgCl₂, 2 mM CaCl₂, 0.1 mM Picrotoxin). Excitatory postsynaptic currents (EPSCs) were evoked by extracellular stimulation of the Schaffer collateral fibers using a concentric bipolar stimulating electrode (World Precision Instruments) at 0.03 Hz. LTP was induced by pairing synaptic stimulation (2 Hz for 120 pulses, or 40 Hz for 600 pulses at the end of the depolarization) with a postsynaptic depolarization to 0 mV (60 s). EPSP potentiation was assessed for 50–60 min after LTP condition stimulation. For the APV experiments, APV (50 µM) was applied 10 min before recording. All data was analyzed with an in-house program written in MatLab.

Validation of shRNA

Dissociated cortical neuron cultures were infected with GIPZ lentiviral CAMKIIβ shRNA (1 µl/ml) or GIPZ lentiviral scramble shRNA (titer: 1.5×1013 genome copies/µl; Dharmacon) at DIV 6. On DIV 14–16, neurons were washed with PBS and immediately extracted with SDS buffer. Samples were separated on 4–20% acrylamide gel (Mini-PROTEAN TGX precast gels, Bio-Rad), then transferred onto 0.45 µm pore size PVDF membranes (Millipore) using semidry immunoblotting (transfer buffer containing 25 mM Tris, 200 mM glycine and 20% methanol). Membranes were blocked with 5% nonfat milk in TBS-T (Tris Buffered Saline with 0.1% Tween-20) for 1 hour at room temperature, then incubated overnight at 4°C with primary antibodies diluted in 5% BSA in TBS-T. We used the following commercially available antibodies: mouse anti-CAMKIIβ (1:200; LS-B5767; LSBio), and mouse anti-β-actin (1:2000; Sigma). Membranes were washed 3 times for 15 minutes in TBS-T, followed by incubation for 1 hour at room temperature with HRP-conjugated rabbit anti-mouse secondary antibody (Bio-Rad), diluted 1:2000 in 5% nonfat milk in TBS-T. Membranes were washed 3 times for 15 minutes in TBS-T, then incubated with Pierce ECL Plus western blotting substrate (for CAMKIIβ) or Pierce ECL western blotting substrate (for β-actin) to detect western blotted proteins. We used the Image Quant LAS4000 Imaging System (GE Healthcare) to visualize protein bands. Image J software was used to quantify endogenous CAMKIIβ levels normalized to β-actin to ensure equal loading.

The Quantification and Statistical Analysis

Error bars shown in the figures represent standard error of the mean (s.e.m). One-way ANOVA analysis with post hoc Bonferroni test are used to compare different conditions (α = 0.05). n indicates the number of spines/neurons for spine imaging results and the number of cells for electrophysiology results. Asterisks denote the statistical significance (*p < 0.05). s.e.m of time constants is obtained by bootstrapping.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-CaMKIIβ	LSBio	Cat# LS-B5767-50, RRID: AB_10915143
Anti-Actin	Sigma	Cat# A5316 RRID: AB_476743
Bacterial and Virus Strains
AAV1.CAG.Flex.tdTomato	University of Pennsylvania Vector Core	N/A
AAV1.hSyn.Cre	University of Pennsylvania Vector Core	N/A
pGIPZ Camk2b shRNA	GE-Dharmacon	Cat# VGM5524 Source Clone ID: V2LMM_2804
pGIPZ non-silencing shRNA	GE-Dharmacon	Cat# RHS4348
Biological Samples
Chemicals, Peptides, and Recombinant Proteins
MNI-caged-L-glutamate	Tocris	Cat# 1490
Tetrodotoxin	Tocris	Cat# 1078
MEM	Sigma-Aldrich	Cat# M4642
Critical Commercial Assays
Pierce BCA Protein Assay Kit	Thermo Scientific	Cat# 23225
Deposited Data
Experimental Models: Cell Lines
Experimental Models: Organisms/Strains
Mouse- C57BL/6NCrl	Charles River	Strain Code: 027
Mouse- Camk2aT286A/T286A	Giese et al. 1998	Dr. K. Peter-Giese
Mouse- Camk2afl/fl C57BL/6-Camk2atm1Vyb	Hinds et al., 2003	Jackson lab, Stock # 006575
Oligonucleotides
Recombinant DNA
Camuiα (Green-Camuiα or VenusA208K,T145W-CaMKIIα- mEGFP)	Lee et al., 2009	N/A
pCMV-CaMKIIβ (Rattus)	Okamoto et al., 2007	Dr. Y. Hayashi
Software and Algorithms
Prism 6	GraphPad Software	https://www.graphpad.com/scientific-software/prism/
Matlab	MathWorks	https://www.mathworks.com/
Other
Millicell Cell Culture Insert, 30 mm, hydrophilic PTFE, 0.4 µm	EMD Millipore	Cat# PICM0RG50

In brief.

Chan et al. showed that CaMKII activation acts as a leaky integrator of Ca2+ pulses in dendritic spines, and demonstrated that phosphorylation of T286 is necessary for optimal integration of Ca²⁺ pulses during the induction of LTP but dispensable for the maintenance of LTP.

Highlights.

• CaMKII acts as a leaky integrator of Ca²⁺ pulses in dendritic spines

• Phosphorylation at T286 is necessary for optimal integration of Ca²⁺ pulses

• Phosphorylation at T286 is dispensable for the maintenance of LTP