β Ca²⁺/CaM-dependent kinase type II triggers upregulation of GluA1 to coordinate adaptation to synaptic inactivity in hippocampal neurons

Abstract

Prolonged AMPA-receptor blockade in hippocampal neuron cultures leads to both an increased expression of GluA1 postsynaptically and an increase in vesicle pool size and turnover rate presynaptically, adaptive changes that extend beyond simple synaptic scaling. As a molecular correlate, expression of the β Ca²⁺/CaM-dependent kinase type II (βCaMKII) is increased in response to synaptic inactivity. Here we set out to clarify the role of βCaMKII in the various manifestations of adaptation. Knockdown of βCaMKII by lentiviral-mediated expression of shRNA prevented the synaptic inactivity-induced increase in GluA1, as did treatment with the CaM kinase inhibitor KN-93, but not the inactive analog KN-92. These results demonstrate that, spurred by AMPA-receptor blockade, up-regulation of βCaMKII promotes increased GluA1 expression. Indeed, transfection of βCaMKII, but not a kinase-dead mutant, increased GluA1 expression on dendrites and elevated vesicle turnover (Syt-Ab uptake), mimicking the effect of synaptic inactivity on both sides of the synapse. In cells with elevated βCaMKII, relief of synaptic-activity blockade uncovered an increase in the frequency of miniature excitatory postsynaptic currents that could be rapidly and fully suppressed by PhTx blockade of GluA1 receptors. This increased mini frequency involved a genuine presynaptic enhancement, not merely an increased abundance of synapses. This finding suggests that Ca²⁺ flux through GluA1 receptors may trigger the acute release of a retrograde messenger. Taken together, our results indicate that synaptic inactivity-induced increases in βCaMKII expression set in motion a series of events that culminate in coordinated pre- and postsynaptic adaptations in synaptic transmission.

Synaptic properties can be modulated by chronic changes in overall cell activity (1) as well as by brief stimuli that evoke long-term synaptic potentiation (LTP) or depression (2). In either case, activity manipulation induces alterations in the number, properties, and composition of AMPA-type glutamate receptors (GluA subunits), which dominate excitatory transmission at central synapses (3). This finding raises interesting questions about the mechanisms that link the very different modes of induction to their ultimate functional effects. Although α Ca²⁺/CaM-dependent kinase type II (αCaMKII) is generally accepted as a critical player in inducing LTP, a varied group of signaling molecules have been implicated in the response to chronic activity deprivation, including BDNF (4–7), Arc (8), TNF-α (9, 10), retinoic acid (11, 12), and β3 integrin (13, 14). The interrelationship between these putative signaling molecules and the overall organization of the signaling in response to prolonged activity block remain obscure.

Characterizing the mechanisms underlying inactivity adaptation has been complicated by both the form of inactivity studied (e.g., blocking spikes, blocking synaptic transmission, blocking both concurrently) and the diversity of changes in synaptic properties that inactivity can produce. For example, in cortical neurons, tetrodotoxin (TTX) resulted in a scaling of AMPA receptor (AMPAR)-mediated miniature excitatory postsynaptic potentials (mEPSCs) (15). A similar effect was observed in spinal neurons and attributed to an increased accumulation of GluA1 subunits from slowed metabolic breakdown (16). In contrast, chronic block of AMPARs in cultured hippocampal neurons gave rise to clear increases in the rates of presynaptic vesicular turnover (17) and enlargement of presynaptic terminals and vesicle pools (18). In our experiments, chronic AMPAR blockade produced a multifaceted response that included an increase in mini amplitude and frequency, and a faster rate of mini decay (19). Thus, several fundamental questions remain about the nature of the responses to reduced AMPAR activity. Are the disparate changes in synaptic properties connected? What are the signaling mechanisms that link AMPAR blockade to a lasting modification of unitary synaptic properties?

We approach these questions with biochemical, immunocytochemical, and electrophysiological analyses. Previously, βCaMKII expression was shown to increase in response to inactivity and βCaMKII overexpression caused an elevation in mini frequency and speeding of mini decay, alterations similar to those seen after prolonged AMPAR blockade (19). Here we show that, in response to synaptic inactivity, increased βCaMKII activity is necessary to bring about increases in postsynaptic strength. In turn, alteration of the composition of postsynaptic AMPARs enables retrograde signaling from the postsynaptic cell to its immediately presynaptic terminals once transmission block is removed (6).

Results

βCaMKII Knockdown Prevents Up-Regulation of GluA1 Induced by AMPAR Blockade.

Prolonged blockade of AMPARs in cultured hippocampal neurons [16–18 days in vitro (DIV)] leads to synaptic adaptation that is mediated by an increase in the GluA1 AMPAR subunit and is accompanied by an increase in βCaMKII levels (19) (Fig. S1). Here we used this model of adaptation to synaptic inactivity to characterize the role of βCaMKII in the homeostatic response. First, to determine whether βCaMKII is required for the up-regulation of GluA1 induced by synaptic inactivity, we generated lentiviruses that express either a shRNA against βCaMKII or a nonsilencing control shRNA. Infection with βCaMKII shRNA resulted in a ∼90% decrease in βCaMKII mRNA (P < 0.001, ANOVA followed by Newman-Keuls posthoc test, n = 4), but infection with the nonsilencing shRNA had no effect (Fig. 1A). NBQX (2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo[f]quinoxaline-2,3-dione) treatment for 24 h significantly increased βCaMKII protein levels (1.28 ± 0.13-fold relative to untreated cells, P < 0.05) (Fig. 1B), consistent with our previous data (19). In contrast, infection with the βCaMKII shRNA abrogated the inactivity-induced increase in βCaMKII levels (P > 0.8) (Fig. 1B). Treatment with NBQX also increased GluA1 protein (1.24 ± 0.13-fold relative to untreated cells, P < 0.05) (Fig. 1C) (20), which was also prevented (P > 0.5) by βCaMKII knockdown, just as predicted. These data support the idea that an inactivity-induced increase in βCaMKII is a prerequisite for the elevation in GluA1.

Fig. 1.

Knockdown of βCaMKII prevents inactivity-induced increase in GluA1. (A) βCaMKII shRNA decreased βCaMKII mRNA by ∼90%, as assessed by real-time PCR and normalized separately with three housekeeping genes, β-actin (white), α-tubulin (gray), and GAPDH (black). (B) βCaMKII shRNA also decreased βCaMKII protein levels and prevented the increase in βCaMKII protein induced by lack of neurotransmission. Shown are representative immunoblots (Upper) with the actin levels from the same experiment as loading controls shown below. (C) Similar examination of GluA1 levels shows that knockdown of βCaMKII expression also prevented the NBQX-mediated augmentation of GluA1 expression found in uninfected cells. Error bars denote SEM; *P < 0.05. Student's t test here, and for all statistics unless indicated otherwise.

If βCaMKII were necessary for the up-regulation of GluA1 and subsequent increases in mini frequency and amplitude induced by blocking AMPARs, knockdown of βCaMKII should prevent these adaptations in transmission. Consistent with this hypothesis, treatment with NBQX for 24 h produced a 1.7-fold increase in mini frequency in noninfected cells (1.78 ± 0.81 Hz, n = 5 in control, 3.96 ± 0.90 Hz, n = 8 in NBQX-treated cells, P < 0.05), but not in βCaMKII shRNA-infected neurons (Fig. 2 A and D). The average frequency was no different in basal (2.26 ± 0.44 Hz, n = 11) and 24-h NBQX-treated conditions (2.01 ± 0.32 Hz, n = 10 in NBQX). Similarly, the increase in mEPSC amplitude seen in noninfected cells (from 27.0 ± 2.0 pA in control to 31.9 ± 1.8 pA in NBQX-treated cells, P ≤ 0.01) was also abolished by βCaMKII knockdown (from 25.0 ± 0.5 pA in control to 20.1 ± 0.7 pA in activity-deprived cells) (Fig. 2 A–C). In fact, βCaMKII knockdown actually decreased mini amplitude in NBQX-treated cells (P ≤ 0.05). Taken together, these results demonstrate that βCaMKII is necessary for the increase in GluA1 that underlies the modifications in neurotransmission induced by prolonged AMPAR blockade.

Fig. 2.

Evidence for involvement of βCaMKII as requisite player in adaptation of mEPSC properties. (A) Representative traces from control cells show an increase in mini frequency after 24 h of NBQX (Upper pair of traces); no change in frequency in shRNA-infected cell cultures (Lower pair of traces). (B) Averaged minis from the same experiment. Mini amplitude increased in transmission-deprived control cells, but not in transmission-deprived cell cultures infected with shRNA. (C and D) Average mini amplitude and frequency increased in neurotransmission-deprived control cells, but not in cells infected with shRNA. Error bars denote SEM; *P < 0.05.

βCaMKII Overexpression Mimics Inactivity-Dependent Enhancement of Synaptic Function.

Having examined whether βCaMKII is required for synaptic strengthening in response to AMPAR block, we next asked if βCaMKII overexpression altered GluA1 expression at synaptic sites. CA3-CA1 cultures (10–11 DIV) were transfected with EGFP-tagged βCaMKII, or EGFP alone as a control, and stained 24 to 36 h later with antibodies against surface GluA1 (red) and the synaptic marker Synapsin I (blue) (Fig. 3A). Analysis was performed in four independent cultures (βCaMKII-transfected cells, n = 14; EGFP-transfected cells, n = 16; 20–50 synapses per cell). In βCaMKII-transfected cells, the mean intensity of synaptic GluA1 was 1.24 ± 0.07, significantly higher (P < 0.03) than it was in EGFP-transfected cells, 1.00 ± 0.05 (Fig. 3B). Thus, the effects of βCaMKII overexpression are opposite to those of its suppression, compatible with the notion that βCaMKII regulates GluA1 levels at synapses.

Fig. 3.

βCaMKII increases surface expression of GluA1 at synapses. (A) Cell cultures transfected with EGFP-βCaMKII, fixed and stained for surface GluA1 (red) and synapsin (blue). (B) Synaptic GluA1 is significantly higher in cells transfected with EGFP-βCaMKII vs. EGFP alone. Shown as average level from about 200 synapses in four experiments normalized to average level from nontransfected cells in same experiment. Error bars denote SEM; *P < 0.05.

To test whether introducing βCaMKII could generate modifications of synaptic function, we recorded mEPSCs from CA3-CA1 neurons transfected at 10 to 11 DIV with constructs containing EGFP-tagged βCaMKII or a kinase-dead version of the enzyme (K43R) (21). Mini recordings from transfected cells displaying EGFP epifluorescence, and untransfected controls on the same coverslips, were directly compared (Fig. 4A). βCaMKII-transfected cells showed significantly increased mini frequency (2.7 ± 0.4 Hz, n = 9) relative to untransfected controls (0.7 ± 0.3 Hz, n = 12) (P < 0.005) (Fig. 4B), in line with previous findings (19). In contrast, the K43R-expressing cells displayed a mini frequency (0.9 ± 0.3 Hz, n = 8) no different from controls (0.7 ± 0.3 Hz, n = 12) (P > 0.2). In addition, βCaMKII-transfected cells displayed faster mEPSC kinetics (t_decay = 3.2 ± 0.4 ms) relative to controls (4.5 ± 0.1 ms) (P < 0.05), whereas the t_decay in K43R-βCaMKII transfected cells (4.4 ± 0.5 ms) was no different from in controls (P > 0.2) (Fig. 4C). The changes in mEPSC decay kinetics likely reflect the signature of GluA1 homomeric AMPARs, which have more rapid deactivation kinetics than GluA2-containing receptors (22). Thus, the catalytically active form of βCaMKII was able to reproduce some key features of the response to NBQX treatment. The mimicry was not perfect, inasmuch as no change in mini amplitude was seen in βCaMKII-transfected cells (19). This finding might be explained by the less-mature state of the transfected neurons used for recording (11–12 DIV for ease of transfection, not 16–18 DIV, as in experiments studying transmission blockade). At 10 DIV, suppression of activity with the potassium channel Kir2.1 also produced a highly significant increase in mini frequency but no difference in mini amplitude (23).

Fig. 4.

Postsynaptic βCaMKII activity triggers enhanced presynaptic functioning and is required for up-regulation of GluA1 in response to activity deprivation. (A) Recordings in TTX+bicuculline, from cells transfected with βCaMKII (Upper), an inactive form of βCaMKII (K43R mutant) (Lower), and an untransfected cell from the same culture (Upper trace). (B) Cumulative distribution of mini frequency (one per interval) shows an increase in frequency for βCaMKII-transfected cells (thin black line) compared with control cells (thick black line, P < 0.005, K-S test), but not in cells transfected with the inactive K43R (gray line). (C) Average mini from one cell, showing the shorter decay time in βCaMKII-transfected cells compared with control, but not to K43R. (D) Average GluA1 or βCaMKII levels, quantified from immunoblots normalized to actin (representative blots, Upper), in cultures treated with NBQX for 24 h in the presence of the CaM kinase inhibitor KN-93 (Right) or its inactive analog KN-92 (Left). Error bars denote SEM; *P < 0.05.

βCaMKII Kinase Activity Drives GluA1 Increase in the Response to AMPAR Blockade.

Next, we turned to biochemical studies to verify that it is the up-regulation of βCaMKII kinase activity that causes GluA1 elevation during AMPAR blockade. This test is important because CaMKII can also act as a structural, noncatalytic component (24, 25). To this end, we used KN-93 (2 μM), a blocker of protein kinase activity, and for comparison, its inactive analog KN-92 (2 μM). NBQX treatment for 24 h increased GluA1 levels in the KN-92 controls (1.89 ± 0.19 times control) (n = 6, P < 0.05) (Fig. 4D, Left), but not in the presence of KN-93 (n = 5), which left GluA1 levels somewhat lower than in nonactivity-deprived neuronal cultures (0.75 ± 0.05 times control) (Fig. 4D, Right). The differential effects on GluA1 levels were echoed in mini amplitude, frequency, and time course, which were prevented by KN-93 but not by KN-92 (19). In addition, we monitored βCaMKII levels to ensure that they are unaltered by KN-93 application. Indeed, KN-93 treatment did not prevent the increase in βCaMKII generated by activity deprivation: βCaMKII was increased 1.42 ± 0.07-fold with NBQX+KN-93 compared with KN-93 treatment alone (n = 7, P < 0.05) (Fig. 4D). This finding was similar to the rise in βCaMKII in control experiments with KN-92 (1.86 ± 0.15-fold increase with NBQX+KN-92 compared with KN-92 alone, n = 8, P < 0.05). Thus, up-regulation of βCaMKII kinase activity, supported by an overall increase in βCaMKII level, acts upstream of the GluA1 elevation to mediate the biochemical and physiological effects of AMAPR blockade.

Does βCaMKII Elevation Trigger Structural Changes?

An increase in the synapse number could contribute to the increase in mini frequency seen after prolonged AMPAR blockade or βCaMKII overexpression. Indeed, βCaMKII appears to regulate dendritic branching and synapse number in immature hippocampal cultures (<11 DIV) (21). To look for possible changes in synapse density, we counted individually-resolved puncta of synapsin I along GFP⁺ dendrites of transfected cells. The average synapse number of βCaMKII-transfected cells appeared 1.41 ± 0.24 times greater than in EGFP-transfected controls (Fig. 5A), but the increase was not significant (P = 0.13), in part because of high variability, likely arising from random cell death during transfection. We also counted the average density of synapses per unit length of dendrite but found no significant difference (P > 0.4) between cells transfected with βCaMKII (0.31 ± 0.02 synapses/μm, 14 cells, 2,717 synapses) and those transfected with EGFP alone (0.36 ± 0.02 synapses/μm, 16 cells, 3,602 synapses) (Fig. 5B). A spatial rearrangement of synapses, moving closer to the soma, might improve mini detection; however, the ratio of synapse densities in regions proximal (10–30 μm) and distal (110–130 μm) to the soma was no different in βCaMKII- (1.1 ± 0.1) than in EGFP-transfected cells (1.2 ± 0.3) (P > 0.7) (Fig. 5C). To look for a recruitment of silent synapses (synapses lacking AMPARs), we counted the fraction of synapsin-positive puncta that also stained positively for GluA1 (Fig. 5D), but found no difference in βCaMKII- (0.47 ± 0.1) compared with EGFP-transfected controls (0.36 ± 0.1) (P > 0.3). Apparently, the density of GluA1-containing synapses remained unchanged.

Fig. 5.

Possible structural changes associated with inactivity-induced elevation of βCaMKII. (A) Total number of synapses (normalized to untransfected cells) was not significantly increased in βCaMKII- (P = 0.13) relative to EGFP-transfected cells. (B) The number of synapses, when counted along an identified dendrite, was not significantly different in βCaMKII- compared with EGFP-transfected cells. (C) The ratio of proximal synapses (10–30 μm from the soma) to distal synapses (110–130 μm from soma) was no different in EGFP- and βCaMKII-transfected cells. (D) The fraction of synapses containing GluA1 (synapses defined by synapsin staining) was not different in βCaMKII- and EGFP-transfected cells. (E) The total number of dendritic branch intersections with concentric circles at increasing distances from the soma (Scholl analysis, 20-μm intervals) was slightly higher in βCaMKII- (filled symbols) relative to EGFP-transfected cells (empty symbols). At any given distance however, there was no significant difference in the number of branches. (F) The sum of the number of branch intersections at all distances was slightly higher in βCaMKII- (filled bar) relative to EGFP-transfected cells (empty bar). Error bars denote SEM; *P < 0.05.

Could βCaMKII overexpression enlarge the dendritic tree, thus indirectly increasing the total number of synapses? In Scholl analysis, we assessed the extent of the dendritic tree by counting branch intersections with concentric circles at increasing distances from the center of the soma (Fig. 5E). βCaMKII-transfected neurons (n = 53) showed more intersections than EGFP-transfected controls (n = 21), but none of the differences was statistically significant. In pooled data for all circles (Fig. 5F), βCaMKII-transfected neurons averaged 40.5 ± 1.9 branch intersections per cell, 25% more (P < 0.05) than EGFP controls (32.5 ± 2.7 branch intersections per cell). Because synapse density held steady (Fig. 5B), the expansion of the dendritic tree would imply an increase in synapse number, as already hinted in the total synapse count (Fig. 5A). A greater abundance of synapses could contribute to the overall elevation in mini frequency.

Increased Mini Frequency Reflects a Genuine Enhancement of Presynaptic Function.

To pin down functional aspects of the mini frequency increase independent of synapse number, we directly measured vesicle turnover in the presynaptic terminals adjacent to a transfected cell, using uptake of an antibody against a luminal epitope of the vesicle protein synaptotagmin (Syt-ab) (20, 26, 27). Hippocampal cultures transfected with βCaMKII or EGFP at 10 to 11 DIV were allowed to take up Syt-ab for 4 h (with bicuculline and TTX present, as in mini recordings). After fixation, the cells were costained with synapsin to identify putative synapses. Syt-ab staining was measured only at synapses onto transfected cells (Fig. 6A) and its intensity was averaged and normalized to the mean intensity in EGFP-transfected cells (Fig. 6B). Syt-ab uptake was 1.17 ± 0.06-fold greater for βCaMKII-transfected targets than for EGFP controls, without any increase in synapsin staining (1.06 ± 0.02 of EGFP controls). Thus, we conclude that increased levels of postsynaptic βCaMKII, whether induced globally by activity-deprivation, or cell-by-cell via transfection, bring about an up-regulation of presynaptic activity in individual synapses contacting the postsynaptic cell.

Fig. 6.

GluA1 homomers mediate the βCaMKII-induced increase in presynaptic activity. (A) βCaMKII-transfected cells (green) incubated with Syt-ab (red) for 4 h in the presence of TTX and bicuculline and then stained for synapsin as a synaptic marker (blue). (B) Average Syt-ab uptake per synapse from each cell shows increased Syt-ab uptake in βCaMKII-transfected cells. (C) Mini recordings from βCaMKII-transfected cells in the absence (Left) and presence (Right) of PhTx to block GluA2-lacking (likely GluA1 homomeric) AMPA receptors. (D) Mini frequency per minute in βCaMKII- or EGFP-transfected cells, normalized to average frequency in untransfected cells. After 3 min of baseline recording, PhTx was added. (E) Average decay time in βCaMKII- and EGFP-transfected cells before (control) and after adding PhTx. Error bars denote SEM; *P < 0.05.

Presynaptic Enhancement Is Triggered by Postsynaptic Signaling.

Increases in mini frequency induced by chronic NBQX are eliminated within minutes by philanthotoxin (PhTx), a blocker of GluA1 homomers (20, see also refs. 28 and 29). This finding suggests that increased levels of GluA1 drove formation of GluA1 homomers that led in turn to enhanced presynaptic vesicle turnover. We tested whether PhTx would act similarly on changes in neurotransmission brought about by βCaMKII overexpression, which would be a crucial advance because sparse βCaMKII transfection targets individual neurons rather than the entire population, thus specifying the potential origin of the retrograde signaling. Hippocampal cultures were transfected with βCaMKII-EGFP (n = 9) or EGFP alone (n = 8) and mEPSCs were recorded from EGFP-positive cells 20 to 30 h later. Three minutes after initiation of recording, cells were superfused with 10 μM PhTx, and the mEPSC properties before and after drug treatment were compared (Fig. 6C). The βCaMKII-induced increase in mini frequency (2.8 ± 0.4-fold, P < 0.005, K-S test) was reverted by PhTx to levels not significantly different from controls (1.1 ± 0.2-fold, P > 0.2, K-S test) (Fig. 6D), a consistent finding in every cell examined. In contrast, in EGFP controls, PhTx had no significant effect on mini frequency (0.9 ± 0.1-fold) relative to baseline frequency before toxin application (1.0 ± 0.5-fold, P > 0.2, K-S test). All features of the PhTx effect on frequency were similar to those previously reported in NBQX-pretreated cells (20). Furthermore, in βCaMKII-transfected cells, minis displayed faster decay kinetics (3.2 ± 0.40 ms) than EGFP controls (4.6 ± 0.5 ms, P < 0.02, K-S test) (Fig. 6E). After application of PhTx, however, t_decay was significantly lengthened (5.7 ± 0.7 ms, P < 0.01, K-S test) and wound up no different from PhTx-treated controls (P > 0.4). In contrast, mini decay was not slowed in EGFP-transfected control cells by PhTx (4.6 ± 0.5 ms before PhTx, 5.8 ± 0.7 ms after PhTx, P > 0.1, K-S test). The differential effect of PhTx on mini decay reinforces the idea that the toxin specifically blocked βCaMKII-induced GluA1 homomers, and thus prevented their ability to trigger the increase in mini frequency through retrograde transmission (6, 7).

Discussion

We focused on a form of synaptic adaptation to reduced activity that features enhancements of synaptic function on both sides of the synapse. Our experiments uncovered a key role for βCaMKII in the signal-transduction pathway that couples prolonged inactivity to coordinated modifications of post- and presynaptic efficacy. AMPAR inhibition with NBQX causes an increase in βCaMKII levels and βCaMKII activity that drives expression of GluA1 subunits at synapses. Our data implicate postsynaptic βCaMKII as a crucial contributor to adaptation to inactivity.

βCaMKII Links Inactivity to Multiple Aspects of Adaptation.

Participation of βCaMKII was first suggested on the basis of inactivity-driven rises in βCaMKII levels and synaptic effects of increasing βCaMKII via transfection (19). Here we show that the various effects of AMPAR blockade on GluA1 levels and mini properties were all abolished by specific knockdown of βCaMKII with shRNA (Figs. 1 and 2). Furthermore, direct introduction of βCaMKII into individual neurons increased the surface expression of GluA1 to the same degree as NBQX pretreatment (Fig. 3), and also elevated mini frequency and presynaptic vesicle turnover (Fig. 4). Both the rise in GluA1 levels and the faster kinetics of AMPAR minis (an indicator of GluA1 function), depend on βCaMKII being a functional kinase, not merely a structural component (Fig. 4).

The increase in βCaMKII protein and subsequent elevation in GluA1 levels induced by synaptic inactivity develop over the course of hours. We recently found that these slow, adaptive changes enable retrograde communication within minutes once AMPAR blockade is removed. Indeed, relief of inactivity allows Ca²⁺ permeation through GluA1 homomers, triggering rapid BDNF signaling back to the presynaptic terminal to elevate release probability that can be interrupted by chelation of postsynaptic Ca²⁺ (6). Consistent with this scheme (Fig. 6), we show here that the increased mini frequency resulting from overexpression of βCaMKII is acutely mediated by increased GluA1 homomers, as PhTx rapidly reverses the effect. Collectively, these results indicate that increases in the abundance and catalytic activity of βCaMKII are necessary to drive downstream steps in the signaling cascade (Fig. 7). This scenario does not exclude participation of other signaling molecules, such as presynaptic Ca²⁺ channels or neurotrophin receptors as further requisites of the retrograde signaling (7, see also ref. 30).

Fig. 7.

Postsynaptic βCaMKII activity sets in motion adaptation to inactivity, culminating in post- and presynaptic enhancements. Scheme depicting key signaling events, some taking place during prolonged blockade of neurotransmission, and others specifically triggered by removal of the blockade. Blockade of excitatory transmission initiates signaling through βCaMKII and GluA1, which support signaling back to the presynaptic terminal once AMPARs are unblocked (see also refs. 6 and 7).

Our findings suggest that prolonged synaptic inactivity differs strikingly from brief, LTP-inducing activity in reliance on signaling by βCaMKII instead of αCaMKII, even though the ultimate outcome, enhanced synaptic strength through increased AMPAR surface expression, is largely similar. The association between changes in βCaMKII and GluA1 befits a homeostatic response to a reduction of AMPAR activity and a consequent decrease in cytosolic Ca²⁺. Up-regulation of βCaMKII, which is approximately eightfold more sensitive to Ca²⁺/CaM than αCaMKII (31), would tune the enzyme to a lower ambient Ca²⁺ level. By the same token, generation of GluA1 homomers would favor Ca²⁺ entry when AMPAR blockade is removed, possibly aiding in the rebound of neuronal activity levels. Both changes can be viewed as adaptive to a state of reduced neurotransmission; together, the changes help explain how retrograde transmission can be set in motion by spontaneous minis alone (6).

Because α- and βCaMKII play such different roles in response to changes in synaptic input, wide swings in their relative abundance (19) provide a molecular signature of previous network activity and a determinant of future metaplasticity. Molecular mechanisms controlling the level of βCaMKII are unclear, but could involve retinoic acid (11, 32). Interestingly, changes in βCaMKII above or below its basal level both diminish αCaMKII protein. Although overexpression of βCaMKII reduces αCaMKII protein (19), lowering βCaMKII by shRNA also decreased αCaMKII protein levels (Fig. S2B), possibly accounting for the reduction in mini amplitude (Fig. 2C). The αCaMKII mRNA was spared (Fig. S2A), consistent with a specific action of the βCaMKII shRNA; thus, the basis of this bell-shaped relationship remains unknown.

The signaling that links βCaMKII to glutamate receptors also remains largely unexplored. Elevation of βCaMKII might promote local protein synthesis of GluA1, or retard the endocytosis of surface GluA1. That βCaMKII overexpression alone is unable to increase mEPSC amplitude indicates that additional factors are required to prevent endocytosis of GluA2-containing AMPARs during inactivity. One possibility is that inactivity triggers an additional down-regulation of GluA2 endocytosis (16). Future work will be aimed at determining how synaptic strengthening arises from such diverse signaling mechanisms, triggered by very different changes in activity (blockade of AMPAR versus short, intense bursts of activity). It will be interesting to determine how long GluA1 surface expression remains enhanced following removal of activity blockade and how this relates to resetting of βCaMKII levels.

The heightened levels of homomeric GluA1 and βCaMKII following experimental blockade of neurotransmission are reminiscent of elevations during development itself. Whereas GluA2-lacking AMPARs are abundant in early development, GluA2-containing AMPARs dominate in mature synapses (33). Similarly, βCaMKII levels peak earlier in development relative to αCaMKII (34, 35). Perhaps AMPAR blockade and activity deprivation cause neurons to revert to an earlier developmental state in which neuronal signaling pathways are geared up for establishing new synaptic connections.

Materials and Methods

See SI Materials and Methods for transfections, lentiviral-mediated knockdown of βCaMKII, real-time PCR, Western blotting, Syt-ab uptake, immunocytochemistry, and electrophysiology.

Cell Culture.

CA3-CA1 hippocampal neurons were cultured as previously described (36). Cells were deprived of synaptic activity for 24 h by adding 10 μM NBQX (Ascent Scientific) to culture wells after 14 to 16 DIV.

Statistics.

Unless otherwise noted, data are shown as mean value ± SEM and data shown as cumulative distributions were tested by the nonparametric Kruskal-Wallis test and then, in case of statistically significantly different means, compared pairwise using the Kolmogorov-Smirnov test (K-S test). Mean values shown as bar graphs were compared using Student's t test (if only two groups) or ANOVA followed by Newman-Keuls posttest for multiple groups (if more than two groups were compared).