Selective erasure of distinct forms of long-term synaptic plasticity underlying different forms of memory in the same postsynaptic neuron

SUMMARY

Generalization of fear responses to non-threatening stimuli is a feature of anxiety disorders. It has been challenging to target maladaptive generalized memories without affecting adaptive memories. Synapse-specific long-term plasticity underlying memory involves the targeting of plasticity-related proteins (PRPs) to activated synapses. If distinct tags and PRPs are used for different forms of plasticity, one could selectively remove distinct forms of memory. Using a stimulation paradigm in which associative long-term facilitation (LTF) occurs at one input and non-associative LTF at another input to the same postsynaptic neuron in an Aplysia sensorimotor preparation, we found that each form of LTF is reversed by inhibiting distinct isoforms of protein kinase M (PKM), putative PRPs, in the postsynaptic neuron. A dominant negative atypical PKM selectively reversed associative LTF, while a dominant negative classical PKM selectively reversed non-associative LTF. While both PKMs are formed from calpain-mediated cleavage of PKCs, each form of LTF is sensitive to a distinct dominant negative calpain expressed in the postsynaptic neuron. Associative LTF is blocked by dominant negative classical calpain, while non-associative LTF is blocked by dominant negative small optic lobe (SOL) calpain. Interfering with a putative synaptic tag, the adaptor protein KIBRA, which protects the atypical PKM from degradation, selectively erases associative LTF. Thus, the activity of distinct PRPs and tags in a postsynaptic neuron contribute to the maintenance of different forms of synaptic plasticity at separate inputs allowing for selective reversal of synaptic plasticity and providing a cellular basis for developing therapeutic strategies for selectively reversing maladaptive memories.

INTRODUCTION

Multiple memories can become persistent when neutral stimuli are presented close in time with a strong behaviorally relevant stimulation that induces protein synthesis-dependent long-term memory [1]. In some circumstances, the long-term memories generated by the neutral stimuli – referred to as behavioral generalization or emotional tagging - are maladaptive resulting in anxiety disorders [2]. One cellular analog for these phenomena is the synaptic tagging and capture hypothesis, which posits that stimulation insufficient on its own to produce long-term plasticity can nevertheless produce long-term changes in synaptic strength when paired with stronger stimulation at a different synapse [3–6].

The expression of plasticity-related proteins (PRPs) is induced by strong stimulation that produces persistent plasticity. The synaptic tagging and capture hypothesis describes the ‘capture’ of PRPs at additional sites ‘tagged’ by the timely presentation of weak stimulation [7–14]. Neither the identity of the molecular tags at activated synapses nor the PRPs that are captured by these tags in either presynaptic [12–14] or postsynaptic neurons [7–11] are definitively known. Potential candidates for PRPs include a number of kinases such as atypical PKCs or persistently active forms of PKC, named PKMs ([8–11]; but see also Volk et al [15] that question the role of the atypical PKMζ in long-term memory and plasticity). Although it is postulated that the weakly stimulated synapses ‘capture’ the same PRPs as the synapse receiving the stronger stimuli, it is unknown whether the expression of long-term potentiation (LTP) or long-term facilitation (LTF) at all synaptic inputs share the same molecular tags and PRPs for consolidating and maintaining the plasticity. If they do not share the same molecules, it may be possible to selectively reverse one form of plasticity without affecting the other.

Aplysia sensorimotor synapses express two different forms of persistent LTF – non-associative LTF (a cellular analog of long-term sensitization) and associative LTF (a cellular analog of long-term classical conditioning) [16–19]. The maintenance of each form of LTF requires stimulus-dependent activation of specific calpain isoforms, which cleave the different PKC isoforms to generate multiple cell-specific constitutively active PKMs [20–22]. Do different constitutively active PKMs generated by different calpains in a postsynaptic neuron maintain the different forms of persistent plasticity at separate inputs? Can manipulating the activity of specific molecules in the postsynaptic neuron selectively reverse each form of persistent plasticity expressed at the separate synaptic inputs?

We found that distinct molecules in the postsynaptic motor neuron maintain two different forms of LTF expressed at separate synaptic inputs. Activities of different calpain isoforms, which cleave PKC isoforms into their respective constitutively active PKMs, are required for the consolidation of the different forms of LTF. Different PKMs maintain the two forms of LTF, and the adaptor protein KIBRA, critical for long-term plasticity and cognition [23–25], stabilizes specific PKMs needed to sustain LTF. These results provide a potential strategy for selectively reversing long-term plasticity at specific synapses to reverse maladaptive long-term memories that form as a result of behavioral or emotional tagging.

RESULTS

A cellular analog of long-term sensitization expressed at Aplysia sensorimotor synapses produced by 5 bath applications of 5-HT (5 μM, each lasting 5 min at 20 min intervals) lasts more than a week after two days of 5-HT applications [18]. A cellular analog of long-term classical conditioning produced by the pairing of action potential activity in the sensory neurons (2 sec tetanus at 20 Hz) with a 5 min bath application of 5-HT (5 μM) lasts more than a week after two pairings of activity plus 5-HT [19]. The two different forms of persistent LTF have different properties, are maintained by different PKMs, and can be reversed by different manipulations [18–22, 26]. Can separate sensory neuron inputs onto an L7 express the two different forms of persistent LTF?

Two pairings of stimuli at one input induces persistent LTF at both inputs

Two sensory neurons were plated with one motor neuron L7 (Figure 1A and 1B) and allowed four days to form stable synapses [18, 19, 22]. We examined the consequences of two pairings of activity and 5-HT to one input on day 0 (4 days after plating) and day 1, while the other input was exposed only to the two bath applications of 5-HT. We compared the changes in synaptic strength to those detected after treatment with L15 + seawater (Cont), 5-HT application alone on day 0 and one pairing of activity plus 5-HT to one input on day 1, which produces a transient form of LTF lasting less than 48 h [19, 27], or two unpaired stimuli (tetanus and 5-HT application separated by 5 min) on day 0 and day 1. Synaptic strength (EPSP amplitude) was examined on day 0 before stimuli and re-examined after stimuli on day 3, 5 and 7 to determine the duration of the plasticity at each sensory neuron input (Figure 1A).

Figure 1. Two pairings at one input produced persistent plasticity at both sensory neuron inputs.

(A) Experimental protocol; see Results and Methods.

(B) Two sensory neurons plated with a single motor neuron L7. The scale bar equals 100 μm.

(C) Summary of the changes in synaptic strengths produced by treatment. A two-way ANOVA indicated a significant effect of group × repeated measures (F_{9, 72} = 100.086, p < 0.0001 at S1; F_{9, 72} = 144.302, p < 0.0001 at S2). Pairwise comparisons at each time point (Bonferroni) indicated that two pairings of activity plus 5-HT (stimulated alone) evoked significant facilitation at each input compared to Cont (**p < 0.01) and the other stimulated groups (^#p < 0.05; ^##p < 0.01) on day 3, 5 and 7.

See also Figures S1 and S2.

Compared to control (N = 7) or one paired stimuli (N = 7), two paired stimuli (N = 8) produced LTF lasting more than 6 days at both the sensory neuron input stimulated with 2 pairings of activity plus 5-HT (S1) and the sensory neuron input exposed to two 5-HT applications (S2; Figure 1C and Figure S1). A 5-HT application on day 0 plus a paired stimulation on day 1 produced only a transient form of LTF at the input receiving the paired stimuli [see 19, 27] that is no longer significantly different than control by day 3, and the sensory neuron input exposed to two 5-HT applications (S2) showed no significant change in synaptic strength at any time point. Two unpaired stimuli (N = 6) failed to evoke LTF at either the sensory neuron input stimulated with the unpaired stimuli or the sensory neuron input exposed to two 5-HT applications (Figure 1C and Figure S1). Thus, two paired stimuli that evoked persistent LTF at one sensory neuron input recruit persistent LTF at a second input onto the same postsynaptic neuron that received two 5-HT applications, which failed to produce LTF at that sensory neuron input if persistent LTF was not expressed at the other input. The expression of persistent LTF at both inputs under these conditions matches the persistent plasticity at multiple inputs explained by the synaptic tagging and capture hypothesis.

Each sensory neuron input expresses a different form of persistent LTF

Persistent associative LTF at sensorimotor synapses is accompanied by the attenuation in the kinetics of homosynaptic depression (HSD), a form of plasticity evoked by low frequency stimulation of the sensory neuron (homosynaptic stimulation – HS, a single action potential at 20 sec intervals) [19], while synapses expressing persistent or transient non-associative LTF show HSD kinetics indistinguishable from controls [19]. Following the different stimuli to each input on day 0 and day 1, the sensory neuron input receiving the two pairings of activity plus 5-HT (S1) expressed HSD with attenuated kinetics, while the sensory neuron input exposed to two applications of 5-HT (S2) had the same HSD kinetics as controls (see Figure S2). Thus, the sensory neuron inputs stimulated with two pairings express features characteristic of persistent associative LTF, while the inputs exposed to two 5-HT applications do not.

The ‘reminders’ that induce a labile state, which can lead to the reversal of LTF when synaptic reconsolidation is blocked, also distinguish the two forms of persistent LTF. For associative LTF, a reminder is low frequency stimulation in the sensory neuron, while for non-associative LTF a reminder is a brief application of 5-HT [18, 19]. Importantly these reminders are specific: low frequency stimulation in the sensory neuron do not make synapses expressing non-associative LTF labile, and 5-HT does not make synapses expressing associative LTF labile. In both cases, synaptic reconsolidation is blocked by a 2-hour incubation with rapamycin immediately following the reminder. If the single postsynaptic motor neuron can maintain two forms of persistent plasticity, one would predict that blocking reconsolidation after distinct reminders could selectively reverse each form of plasticity. We examined whether reversal occurred by 24 hours (day 4) after the different types of reminder plus rapamycin applied on day 3, 48 hours after the last stimulation, and on day 6 to determine if any reversal is permanent.

After stimulation led to persistent facilitation at both S1 and S2, when 5-HT is used as the reminder followed by rapamycin [2X (Tet + 5-HT) + (5-HT + Rapa); N = 8] on day 3, the persistent LTF expressed at the sensory neuron input S1 is unaffected, while the persistent LTF expressed at the sensory neuron input S2 (Figure 2B) was reversed. Treatment with 5-HT alone (N = 7) or rapamycin alone (N = 7) failed to alter persistent LTF at either input. Basal synaptic strength at both inputs (S1 and S2) was unaffected by 5-HT + rapamycin (N= 6), 5-HT alone (N = 6) or rapamycin alone (N = 6) (Figure 2B). Thus, the persistent LTF evoked at the sensory neuron input exposed to two applications of 5-HT (S2), but not the persistent LTF produced at the sensory neuron input stimulated with two pairings of activity plus 5-HT (S1), is reversed with rapamycin after a brief application of 5-HT – a feature characteristic of persistent non-associative plasticity [18].

Figure 2. Reactivation-dependent synaptic reconsolidation blockade at each input after rapamycin.

(A) Experimental protocol; see Results and Methods.

(B) Summary of the changes in synaptic strengths +/− 5-HT application on day 3 followed by +/− rapamycin. A two-way ANOVA indicated a significant effect of group × repeated measures for both S1 (F_{18, 120} = 237.19; p < 0.0001) and S2 (F_{18, 120} = 119.814; p < 0.0001). Pairwise comparisons at each time point indicated significant facilitation in S1 for all stimulated groups compared to Cont + 5-HT on day 3, 4 and 6 (**p < 0.01). In S2, pairwise comparisons indicated significant facilitation for all stimulated groups compared to Cont + 5-HT on day 3, 4 and 4 (**p < 0.01), except for 5-HT reactivation plus rapamycin, which was significantly different than stimulated alone on day 4 and 6 (^##p < 0.01).

(C) Summary of the changes in synaptic strength +/− HS followed by +/− rapamycin. A two-way ANOVA indicated a significant effect of group × repeated measures for both S1 (F_{18, 120} = 72.987; p < 0.0001) and S2 (F_{18, 120} = 74.701; p < 0.0001). Pairwise comparisons indicated significant facilitation in S1 for all stimulated groups compared to Cont + HS_S1 on day 3, 4 and 6 (**p < 0.01), except for HS_S1 reactivation plus rapamycin, which was significantly different than stimulated alone on day 4 and 6 (^##p < 0.01). In S2, pairwise comparisons indicated significant facilitation for all stimulated groups compared to Cont + HS_S1 (**p < 0.01) on day 3, 4 and 6.

In contrast, the persistent LTF at S1 is reversed when low frequency stimulation of the sensory neuron (HS) at S1 is used as a reminder followed by rapamycin [2X (Tet + 5-HT) + (HS_S1 + Rapa); N = 10] on day 3, while the persistent LTF expressed at input S2 is unaffected (Figure 2C). Unlike HS at input S1, HS at input S2 followed by rapamycin [2X (Tet + 5-HT) + (HS_S2 + Rapa); N = 7] on day 3 failed to affect the persistent LTF expressed at either input. HS alone at either input failed to alter either form of persistent LTF (N = 6 for each input). Synaptic baseline was unaffected by HS alone, or HS followed by rapamycin (N = 6 for each treatment; data for S2 not shown). The persistent LTF produced at input S1 (two pairings of activity plus 5-HT), but not the persistent LTF produced at input S2 (two applications of 5-HT), is reversed with rapamycin only after low frequency stimulation of the sensory neuron (HS) – a property characteristic of persistent associative LTF [19]. Thus, specific reminders followed by a blockade of reconsolidation selectively reverse each form of persistent plasticity expressed at each input onto L7. This reversal was long lasting, remaining three days after the blockade.

Different PKM isoforms in L7 sustain LTF at the separate inputs

Associative and non-associative LTF require different isoforms of PKM in the postsynaptic neuron for the maintenance of LTF [22]. In L7, activity of PKM Apl I, but not PKM Apl III, is required to sustain persistent non-associative LTF. In contrast, PKM Apl III, but not PKM Apl I, in L7 is required to sustain persistent associative LTF. We examined whether interfering with the activities of each PKM isoform in L7 by over expressing specific dominant negative (dn) constructs on day 3, 48 hours after presenting the last pairing of activity plus 5-HT, would reverse persistent LTF at specific sensory neuron inputs by day 4, 24 hours after construct injection (Figure 3A and Figure S3) and whether any reversal is retained on day 6, 3 days after construct injection.

Figure 3. Over expression of specific dn-PKM constructs in L7 reversed distinct forms of persistent LTF at each input.

(A) Experimental protocol; see Results and Methods.

(B) Phase contrast views of sensorimotor cultures and epiflourescent views of construct expression in the same view area 24 hours after injection on day 3. Scale bar equals 100 m.

(C) Summary of the changes in synaptic strengths after over expression of control or dominant negative (dn) PKM constructs. A two-way ANOVA indicated a significant effect of group × repeated measures for both S1 (F_{21, 132} = 74.8; p < 0.0001) and S2 (F_{21, 132} = 35.802; p < 0.0001). Pairwise comparisons at each time point indicated significant facilitation in S1 for all stimulated groups compared to Cont on day 3, 4 and 6 (**p < 0.01), except after over expression of dn-PKM Apl III, which was significantly different than stimulated alone on day 4 and 6 (^#p < 0.05; ^##p < 0.01). In S2, pairwise comparisons indicated significant facilitation for all stimulated groups compared to Cont on day 3, 4 and 6 (**p < 0.01), except after over expression of dn-PKM Apl I, which was significantly different than stimulated alone on day 4 and 6 (^##p < 0.01).

See also Figure S3.

Over expression of dn-PKM Apl III in L7 reversed persistent LTF expressed at input S1 (two pairings of activity plus 5-HT), while the increase in synaptic strength is sustained at input S2 (two applications of 5-HT; Figure 3C and Figure S3). In contrast, over expression of dn-PKM Apl I in L7 reversed persistent LTF at input S2, while the increase in synaptic strength is sustained at input S1 (Figure 3C and Figure S3). Over expression of either dominant negative construct in L7 did not affect basal synaptic strength. Over expression of the control construct mRFP in L7 did not affect persistent LTF (N = 7) or basal synaptic strength (N = 7) expressed at either input (Figure 3C and Figure S3). Thus, different PKM isoforms in L7 are required to sustain different forms of LTF at the separate sensory neuron inputs.

Adaptor protein KIBRA is required for maintaining PKM Apl III-dependent LTF

Persistent calpain activities do not appear to be required for the continued presence of the PKMs responsible for maintaining LTF after day 3 [22]. What other molecules stabilize the PKMs for sustaining persistent plasticity? If multiple PKMs are present in L7, how are different PKMs targeted to distinct synaptic inputs expressing different forms of LTF? KIBRA is an adaptor protein that interacts with atypical PKMs to prevent degradation and is required to maintain LTP in hippocampus [23–25]. KIBRA also represents an interesting candidate to link PKMs to AMPA receptors [23]. We cloned Aplysia KIBRA and it retains all the conserved domains of vertebrate KIBRA including the atypical PKM binding site (Figure 4A). We examined the ability of KIBRA to stabilize the different isoforms of PKM in Aplysia and the dependence of stabilization on this binding site, by mutating the three critical residues identified to be required for binding vertebrate PKMζ [24] to alanine (KIBRA_AAA).

Figure 4. KIBRA stabilizes PKMs in isolated Aplysia sensory neurons.

(A) Description of domains and conservation in Aplysia KIBRA. The homology in the atypical PKC binding domain, and the residues mutated to form KIBRA_AAA are shown.

(B) Examples of expression of eGFP and mRFP PKM Apl I (top), mRFP PKM Apl II (middle) and mRFP PKM Apl III (bottom) either with vector (pNEX3), KIBRA, or KIBR_AAA in neurites of isolated Aplysia sensory neurons.

(C) Quantification of stabilization. All results are normalized in each experiment to the average mRFP/EGFP ratio of vector alone (pNEX3) (solid horizontal line). ANOVAs were performed separately for PKM Apl I (pNEX, N = 29; KIBRA, N = 43; KIBR_AAA, N = 26; F_{2, 97} = 6.7, p < 001), PKM Apl II (pNEX, N = 18; KIBRA, N = 15; KIBR_AAA, N = 22; F_{2, 54} = 4.04, p < 0.05) and PKM Apl III (pNEX, N = 16; KIBRA, N = 31; KIBR_AAA, N = 31; F_{2, 77} = 12.1, p < 0.0001). *p < 0.05 for a post-hoc comparison between this group and the control pNEX (Bonferroni post hoc test). All experiments were repeated in at least three separate preparations of sensory neurons, N = number of neurons. Error bars are SEM.

(D) After live imaging of mRFP and eGFP, a subset of neurons were fixed and stained with anti-KIBRA antibody. Levels of KIBRA increased in KIBRA expressing cells (4.2 ± 1.3 fold, N = 15) and KIBRA_AAA expressing cells (4.8 ± 2.1 fold, N = 18) to a similar extent, p > 0.1 two-tailed Students t-test between the fold increase in KIBRA and KIBRA_AAA expressing cells.

We expressed each mRFP-PKM together with eGFP as a control to monitor the level of expression in isolated sensory neurons. We then compared the mRFP/eGFP ratio in cells expressing vector alone, wild type KIBRA or mutant KIBRA_AAA with the atypical PKM binding site removed. As expected, KIBRA increased the mRFP/eGFP ratio for PKM Apl III. No change in the ratio of mRFP-PKM Apl III/eGFP is observed for KIBRA_AAA (Figure 4B; quantified in Figure 4C), suggesting that this site is required for stabilization of PKM Apl III. In contrast, KIBRA did not increase the mRFP/eGFP ratio for PKM Apl I, although surprisingly, KIBRA_AAA did (Figure 4B; quantified in Figure 4C). This suggests that not only do these mutations disrupt a site important for stabilization of PKM Apl III, but that they cause a conformational change in KIBRA that reveals a cryptic site for stabilizing PKM Apl I. This suggests that alternative splicing of KIBRA or post-translational modifications of KIBRA could alter the specificity of KIBRA for PKMs. Both KIBRA and KIBRA_AAA increased the ratio for PKM Apl II (Figure 4B; quantified in Figure 4C), suggesting that PKM Apl II could be stabilized by both the site used by PKM Apl III and the cryptic site that could stabilize PKM Apl I. Using an antibody to KIBRA, we confirmed that KIBRA and KIBRA_AAA are over-expressed at relatively equivalent levels (Figure 4D). These results suggest multiple isoform-specific stabilizing sites in KIBRA for the PKMs, with the site mutated being required specifically for PKM Apl III stabilization. If KIBRA is important for localizing and stabilizing PKMs at the appropriate synapse, then one would expect KIBRA_AAA to act as a specific dominant negative for PKM Apl III, binding to the appropriate synaptic complex, but not recruiting PKM Apl III to that complex. In contrast, KIBRA_AAA would not act as a dominant negative for PKM Apl I or PKM Apl II, since this variant still stabilizes these isoforms.

We examined whether KIBRA interactions in L7 are required to maintain persistent plasticity at a specific sensory neuron input after the two inputs receive stimulation that resulted in persistent plasticity at both inputs. We injected dominant negative construct KIBRA_AAA (dn-KIBRA) on day 3, 48 hours after control treatment or stimulation. We determined whether KIBRA_AAA would reverse persistent LTF at each sensory neuron input by day 4, 24 hours after construct injection (Figure 5A and Figure S4) and whether any reversal would be retained on day 6, 3 days after construct injection.

Figure 5. Blocking KIBRA interactions in L7 selectively reversed persistent associative plasticity.

(A) Experimental protocol; see Results and Methods.

(B) Summary of changes in synaptic strength following over expression of dn-KIBRA in L7 on day 3. A two-way ANOVA indicated a significant effect of group × repeated measures for both S1 (F9, 69 = 154.881; p < 0.0001) and S2 (F_{6, 69} = 254.243; p < 0.0001). Pairwise comparisons at each time point indicated significant facilitation in S1 and S2 for all stimulated groups compared to Cont on day 3, 4 and 6 (**p < 0.01), except in S1 after over expression of dn-KIBRA, which was significantly different than stimulated alone on day 4 and 6 (^##p < 0.01).

See also Figures S4–S6.

Over expression of KIBRA_AAA reversed persistent associative LTF at the sensory neuron input S1, but did not affect persistent non-associative LTF expressed at the sensory neuron input S2 (N = 8; Figure 5B). The increase in synaptic strength at input S2 is comparable to the increase at input S2 after 2X (Tet + 5-HT) alone (N = 7). Over expression of KIBRA_AAA (N = 6) did not affect basal synaptic strength at either input (Figure 5B). Thus, the adaptor protein KIBRA in L7 acts by stabilizing PKM Apl III, whose constitutive activity is required for sustaining associative LTF selectively at synaptic sites of S1 receiving two pairings of activity plus 5-HT. The effects of KIBRA_AAA also match the requirements for PKM Apl III when a single sensory neuron synapses with L7 [22]. When expressed in L7, KIBRA_AAA reverses associative LTF, but not non-associative LTF, which requires PKM Apl I (Figure S5) [22]. When expressed in the sensory neuron, KIBRA_AAA reverses non-associative LTF, but not associative LTF, which requires PKM Apl II (Figure S6) [22].

Specific calpain isoforms are required for persistent LTF at the separate inputs

Aplysia neurons express several calpains, and some are known to cleave the PKC isoforms into their respective PKMs [21] Although the calpains are not required to maintain plasticity (after day 3), the classical calpain activity is required in the sensory neurons and L7 for the consolidation of persistent associative LTF, while SOL calpain activity is required in the sensory neurons and L7 for the consolidation of persistent non-associative LTF [22]. We examined whether interfering with these calpain activities, by over expressing specific dominant negative constructs in L7 on day 1 soon after the second pairing, affected the consolidation of persistent plasticity at the separate sensory neuron inputs on L7. We re-examined synaptic strength starting 24 hours after construct injection to determine the time course of the changes in synaptic strength (Figure 6A) and on day 5 to determine if the block of persistent facilitation lasted.

Figure 6. Expression of persistent LTF required calpain activity.

(A) Experimental protocol; see Results and Methods.

(B) Summary of the changes in synaptic strength following over expression of either dn-SOL or dn-Classical calpain in L7. A two-way ANOVA indicated significant effect of group × repeated measures for both S1 (F_{15, 99} = 71.482; p < 0.0001) and S2 (F_{15, 99} = 47.761; p < 0.0001). Pairwise comparisons at each time point indicated significant facilitation in S1 for all stimulated groups compared to Cont on day 2, 3 and 5 (**p < 0.01), except after over expression of dn-Classical calpain, which was significantly different than stimulated alone on day 2, 3 and 5 (^#p < 0.05; ^##p < 0.01). In S2, pairwise comparisons indicated that only stimulation alone produced a significant facilitation compared to Cont on day 2, 3 and 5 (**p < 0.01). After over expression of either dn-Classical or dn-SOL calpain, these stimulated groups were significantly different than stimulated alone on day 2, 3 and 5 (^#p < 0.05; ^##p < 0.01).

(C) Summary of the changes in synaptic strength following over expression of dn-Classical calpain in the sensory neuron (input S1) receiving two pairings of activity plus 5-HT. A two-way ANOVA indicated a significant effect of group × repeated measures for S1 (F_{9, 66} = 80.395; p < 0.0001) and S2 (F_{9, 66} = 76.666; p < 0.0001). Pairwise comparisons at each time point indicated significant facilitation in both S1 and S2 after stimulation alone compared to Cont on day 2, 3 and 5 (**p < 0.01). After over expression of dn-Classical calpain, there were significant differences in both S1 and S2 compared to stimulated alone on day 2, 3 and 5 (^#p < 0.05; ^##p < 0.01).

Over expressing dn-SOL calpain in L7 beginning on day 1 (about 4 hours after the second pairing of stimuli; N = 7) selectively interfered with non-associative plasticity expressed at the sensory neuron input S2 (Figure 6B) such that it was not significantly different than controls (N = 6). The other sensory neuron input (S1) was unaffected by dn-SOL calpain and expressed persistent associative plasticity (Figure 6B). Over expression of the dn-SOL calpain in L7 (N = 6) did not affect basal synaptic strength. Thus Sol calpain activity in L7 is required for the consolidation of persistent non-associative LTF at sensory neuron input S2.

Over expressing dn-Classical calpain in L7 beginning on day 1 (about 4 hours after the second pairing of stimuli; N = 7) interfered with LTF expressed at both sensory neuron inputs (Figure 6B) compared to the persistent LTF expressed at both inputs (N = 7) after stimulation when the constructs were not injected. Over expressing dn-Classical calpain (N = 6) did not affect basal synaptic strength (S1 and S2; Figure 6B). Thus, Classical calpain activity appears to be required for the consolidation of persistent plasticity at both inputs.

The expression of persistent associative plasticity at input S1 that received the strong stimuli is required for persistent plasticity at input S2 that received the weaker stimuli (see Figure 1C and Figure S1). Thus, blocking Classical calpain activity may have a selective role in the expression of persistent associative LTF at the sensory neuron input receiving two pairings of activity plus 5-HT (S1), which in turn could interfere with the expression of the persistent LTF at input S2 (see Figure 1C and Figure S1). Since the expression of persistent associative plasticity also required the activation of Classical calpain in the sensory neuron [22], we examined whether the expression of persistent plasticity at the S2 input would be affected when dn-Classical calpain construct is over expressed only in the sensory neuron (S1) receiving two pairings of activity plus 5-HT (Figure 6C). Over expressing dn-Classical calpain on day 1 in the sensory neuron receiving the two pairings of stimuli (S1) reversed persistent plasticity at both sensory neuron inputs (S1 and S2; N = 8; Figure 6C) compared to the persistent plasticity at both inputs after stimulation when the construct is not injected (N = 6). Basal synaptic strength at each input is not affected by over expressing dn-Classical calpain in the sensory neuron (N = 6) compared to vehicle treatment (N = 6). Thus, the expression of persistent associative plasticity requires Classical calpain activity in both the sensory neuron and L7, and Classical calpain-dependent expression of persistent plasticity at input S1 receiving two pairings of activity plus 5-HT is required for the expression of persistent plasticity at input S2 receiving two applications of 5-HT.

DISCUSSION

Our results indicate that separate synaptic inputs on a postsynaptic neuron can simultaneously express two different forms of persistent plasticity that mediate different forms of memory, which can be reversed independently by different manipulations. The postsynaptic neuron maintains these different forms of persistent plasticity with different constitutively active PKM isoforms: PKM Apl III sustains associative LTF (cellular analog of classical conditioning) and PKM Apl I sustains non-associative LTF (cellular analog of sensitization). The expression of persistent plasticity at the two synapses shares a property characteristic of synaptic tagging and capture - persistent associative plasticity at one input is required for the expression of persistent non-associative plasticity at the other input. Once consolidated, however, the two forms of plasticity appear to be independent of each other. Each can be reversed selectively by synaptic reconsolidation blockade, or by interfering in L7 with the activity of a specific PKM or the adaptor protein KIBRA.

How are distinct PRPs and tags made at each postsynaptic site to sustain associative LTF or non-associative LTF? A major difference is the downstream consequences of the local stimuli at the L7 sites contacted by S1 and S2 during induction – activation of NMDA, AMPA and metabotropic glutamate receptors [30–33] paired with multiple serotonergic receptors at S1, while only serotonergic receptors are activated at S2 sites. The distinct signals induced in both the presynaptic sensory neuron terminal and the postsynaptic sites in L7 contacted by the sensory neuron - kinases [34, 35], calpains [21], local protein synthesis [36, 37], and the synthesis and secretion of neurotrophin-like peptides [38–40] are all candidates for generating distinct tags and PRPs in both neurons.

What are the tags or PRPs, and how does capture work? A clear difference between the complexes important for associative and non-associative LTF is their dependence on distinct PKMs. Since they are generated by calpain cleavage of PKCs, one possibility is that the PKMs are generated globally downstream of gene expression- and protein synthesis-dependent activation of the calpains. In this case the PKMs are PRPs that would be captured by specific tags (e.g. KIBRA) at the distinct synapses. Alternatively, distinct calpain signaling complexes [27, 38, 40, 41] in L7 may locally produce PKMs at the different synaptic sites. However, since the dominant negative calpains interfere with the induction of persistent plasticity even when expressed hours after stimulation, local activation of the calpains would be a downstream effector of the combination of gene expression and tag, as opposed to the calpains themselves acting as synaptic tags.

KIBRA, which stabilizes PKM Apl III, but not PKM Apl I, may serve as a distinct tag in L7 for maintaining associative LTF. We postulate that another adaptor protein maintains PKM Apl I in L7 for sustaining non-associative LTF at the other input. The synapses presumably have additional tags in both sensory neurons and L7 [42], including those that contribute to the different properties at each input, such as the different HSD kinetics and the different reminders in synaptic reconsolidation.

Our results parallel the cross-tagging and capture phenomenon where the weakly stimulated input that normally produces short-term E-LTD, now expresses long-term L-LTD if other inputs are stimulated strongly to produce L-LTP [11, 43]. Two different forms of persistent plasticity - L-LTP and L-LTD - are co-expressed at synapses contacting the same postsynaptic neurons. Moreover, an atypical PKM may be a PRP sustaining L-LTP but not L-LTD [11]. However our results greatly expand the repertoire of cross-tagging suggesting that two different forms of facilitation initiated by distinct cascades lead to distinct maintenance complexes. From a therapeutic perspective, our results suggest: a) early interference with the expression of the associative plasticity/memory can reverse both associative and non-associative memories, while at later times these become independent and more recalcitrant to reversal [44, 45] and b) identifying the reactivation that selectively activates the inputs recruited to express persistent plasticity or memory followed by specific pharmacological perturbation may provide one strategy for selectively reversing some maladaptive memories produced by emotional tagging during traumatic episodes [46].

In summary, multiple PKMs in a postsynaptic neuron sustain different forms of long-term synaptic plasticity. These PKMs are generated from their respective precursors by specific calpains. The maintenance of the PKMs is aided by adaptor proteins such as KIBRA, which via interaction with specific PKMs, protect constitutively active kinases from degradation. Other constitutively active forms of PKC compensate for removal of the atypical PKMζ in the mammalian CNS [47, 48]. Our results suggest that other PKMs may also contribute to persistent memories when distinct stimuli produce different forms of persistent plasticity.

STAR METHODS

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit polyclonal anti-Kibra	this paper	N/A
Alexa Fluor 647-conjugated goat anti-rabbit antibody	ThermoFisher	A-21245; RRID: AB_2535813
Chemicals, Peptides, and Recombinant Proteins
Potassium chloride (KCl)	Sigma	P5405-500G; CAS: 7447-40-7
Potassium acetate (K-acetate)	Sigma	P1190-500G; CAS: 127-08-2
Potassium HEPES (K-HEPES)	Sigma	H0527-100G; CAS: 82207-62-3
Serotonin creatinine sulfate monohydrate (5-HT)	Sigma	H7752-250MG; CAS: 61-47-2
Fast Green FCF	Sigma	F7252-5G; CAS: 2353-45-9
L-15 Medium (Leibovitz)	Sigma	L5520-500ML
Rapamycin	Calbiochem	Cat# 553211
Hemolymph	From adult Aplysia californica	N/A
Artificial sea water from Instant Ocean sea salt	Instant Ocean	Product No. SS15-10
Critical Commercial Assays
MidiPrep for generating DNA for injections	Invitrogen	K210004
Experimental Models: Organisms/Strains
Aplysia californica	NIH/University of Miami national Resource for Aplysia	N/A
Oligonucleotides
KIbra start inside for cloning GGGTCTAGAGAAATATGCCAGAGAGGGGCAG	this paper	N/A
KIbra start outside for cloning GTTTAGATGAAATATGCCAGA	this paper	N/A
KIbra end outside for cloning CAACAGAAACTCCACCAAGCT	this paper	N/A
Kibra end inside for cloning GGGGGATCCGCCTGCTTACACTTCTTCACC	this paper	N/A
Kibra outside reverse for mutation CACTCAGCCAACTTGGCAACT	this paper	N/A
Kibra reverse for mutation AGCTAGAGCTCGAGCGCTGACAGTATTTCTCTGA	this paper	N/A
KIbra forward for mutation	this paper	N/A
GCTCGAGCTCTAGCTTGGAAGCGGGCTGATGGCA	this paper	N/A
KIbra outside forward for mutation ACACGCTGAGGGAAGTGAAG	this paper	N/A
Recombinant DNA
pNEX mRFP-dn PKM Apl I	[21]	N/A
pNEX mRFP-dn PKM Apl III	[20]	N/A
pNEX mRFP-PKM Apl I	[21]	N/A
pNEX mRFP PKM Apl II	[22]	N/A
pNEX mRFP PKM Apl III	[20]	N/A
pNEX Kibra	this paper	N/A
pNEX dn Kibra	this paper	N/A
pNEX eGFP	[49]	N/A
pNEX mRFP	[50]	N/A
pNEX dn ApCCAL1 calpain	[21]	N/A
pNEX dn ApSOL calpain	[21]	N/A

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, Jiangyuan Hu (jh2004@cumc.columbia.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Aplysia cell culture

Sensory neurons were isolated from adult animals (60–80 g; Aplysia californica, which are hermaphrodites) and each motor neuron L7 was isolated from the abdominal ganglia of juvenile animals (2 g). The co-culture contained one or two sensory neurons and one L7, and maintained up to 11 days [18, 19]. Each culture represented one sample, since each motor neuron L7 was derived from a separate animal. The sensory neurons added to each L7 for cultures prepared on the same day (between 9–12 cultures) were derived from the pleural ganglia of 2 adult animals (4 hemi-ganglia in total). Cultures were maintained at 18 °C and fed every other day with medium containing 50% filtered hemolymph and 50% L15 that had salts added to make it isotonic with seawater. The cultures prepared on the same day were divided into different experimental and control groups for each experiment.

METHOD DETAILS

Electrophysiology

Standard intracellular recording technique was used to record the amplitude of excitatory postsynaptic potential (EPSP) evoked in motor neuron L7 [22]. The soma of L7 was impaled with a recording microelectrode (10–15 MΩ) filled with a solution containing 2.0 m K-acetate, 0.5 m KCl, and 10 mm K-HEPES, pH 7.4. An AxoClamp-2A amplifier and pCLAMP software (Axon Instruments, Foster City, CA) were used for data acquisition. The membrane potential of L7 was hyperpolarized to −80 mV (~20 mV below resting potential), and EPSP for each input was evoked in L7 by stimulating each sensory neuron with a brief (0.3 – 0.5 msec) depolarizing pulse to evoke an action potential using an extracellular electrode placed near the cell body of the sensory neuron. EPSP amplitudes were monitored before and after various treatments at the indicated time points in the different experiments. The resting membrane potential or input resistance measured in the soma of L7 was not significantly altered by time in culture, stimulation and/or over expression of control or experimental constructs. These measures of cell viability changed by less than 10% over the course of the experiments (day 4 – 11 in culture). Cultures with initial EPSP amplitudes on day 4 in culture in the range of 10 – 30 mV were selected (85% of total cultures) for control and experimental conditions. The synaptic strength on day 4 in culture represented the basal synaptic strength (or synaptic baseline), since there was little change in synaptic strength over the next week in the absence of stimulation [18, 19, 41]. Assignment of any culture to control and experimental groups at the start of each experiment on day 4 in culture (defined as day 0 for each experiment) was to insure that there were no significant differences in the mean and variance of the initial strength of the synapses between the groups prior to control or experimental stimulation or treatments.

Persistent LTF and homosynaptic depression

For cultures containing one sensory neuron and one L7 (see Figure S5A and S6A), persistent LTF lasting more than a week was evoked in one of two ways [18–19, 22, 41]: 1) cultures were exposed to 2 consecutive days (day 0 and day 1) of five applications of 5-HT (5 μM; Sigma) each lasting 5 min at 20 min intervals to evoke persistent non-associative LTF. Each application of 5-HT was washed out with solution containing 50% seawater and 50% L15 (vehicle). Control cultures received mock treatments with vehicle. 2) Persistent associative LTF was evoked by two pairings of a presynaptic tetanus with an application of 5-HT on 2 consecutive days. The sensory neuron was stimulated with a tetanus (20 Hz for 2 seconds) while L7 was maintained at its resting membrane potential that was followed immediately (about 1 second) by a bath application of 5-HT (5 μM) for 5 min. Cultures were rinsed with L15-sea water, and then with culture medium (50% filtered hemolymph and 50% L15). On day 1, cultures received the other pairing of tetanus in the sensory neuron and application of 5-HT.

For cultures containing two sensory neurons (inputs S1 and S2) and one L7 (see Figure 1A and 1B) persistent LTF was evoked as follows: One input, S1, was designated to receive the pairing of tetanus in the sensory neuron and application of 5-HT (once on day 1, or twice on day 0 and day 1) as described above (regardless of position or synaptic strength), and at the same time the other input S2 received the 5 min application of 5-HT on 2 consecutive days (day 0 and day 1; see also Figures 2A, 3A, 5A and 6A). In some cultures (Figure 1A), S1 received two unpaired stimuli and S2 received the 5 min application of 5-HT on 2 consecutive days (day 0 and day 1). For each experiment, the mean of the EPSP amplitudes monitored on day 0 for S1 and S2 was not significantly different from each other.

After recording EPSP strength on day 3 (2 days after the last stimulus or controls; see Figure S2), the kinetics of homosynaptic depression (HSD) was monitored in some cultures by homosynaptic stimulation (HS) - stimulating each sensory neuron with low frequency stimulation (one action potential every 20 seconds; 8 stimuli in total). The motor neuron was maintained at −80 mV during the stimulation to accurately record the amplitude of each EPSP [18, 19].

Different reactivations and synaptic reconsolidation blockade

After recording EPSP strength on day 3 (2 days after the stimuli or control treatment), some cultures were exposed to 5-HT (5 μM) for 5 minutes to reactivate the synapses (see Figure 3A). Immediately after the application of 5-HT, some cultures were incubated with rapamycin (100 nM) for 2 hours [18, 19]. EPSP amplitudes were re-examined 24 hours later on day 4 and again on day 6 to test synaptic reconsolidation blockade.

In other cultures one of the sensory neurons was reactivated by homosynaptic stimulation (HS; see Figure 3A). After recording EPSP amplitude on day 3, sensory neuron S1 or S2 was stimulated by firing an action potential at 20 sec intervals (8 stimuli). Immediately after HS, some cultures were incubated with rapamycin for 2 hours [18, 19]. EPSP amplitudes were reexamined 24 hours later on day 4 and again on day 6 to test synaptic reconsolidation blockade.

Plasmid constructs and microinjection

DNA constructs: all constructs use the pNEX3 plasmid background and promoter [49]. Dominant negative (dn) constructs PKMs [dn monomeric red fluorescent protein (mRFP)-PKM Apl I and mRFP-PKM Apl III] and dn calpains (dn-SOL calpain and dn-Classical calpain) have been previously described [20, 21]. Dominant negative mRFP-PKM Apl I and mRFP-PKM Apl III have a conserved aspartic acid in the catalytic site converted to alanine (Apl III D392-A; Apl I D444-A). This mutation removes > 95% of kinase activity, but allows for stabilizing phosphorylation sites. Both dn-PKM Apl I and dn-PKM Apl III have been shown to block intermediate forms of facilitation [20, 21]. The Classical and SOL calpain were the most prominent calpains expressed in sensory and motor neurons, although other calpains also exist in Aplysia databases [21]. The calpains were cloned by PCR based on transcriptosome data (www.aplysiagenetools.org), and dominant negative calpains were generated by converting the catalytic cysteine to serine [21]. KIBRA was cloned using primers (primers in key resources table) derived from National Center for Biotechnology Information (NCBI) XP_012936697.1 with XBA and BAMHI sites at their ends and cloned into the pNEX3 vector [48] cut with XBA and BAMHI. The insert was confirmed by sequencing. The dn-KIBRA (KIBRA_AAA) was generated using overlap PCR to insert the mutations into a carboxy-terminal fragment of KIBRA cloned in the pJET vector (Thermo Fisher Scientific) using unique StuI and PaeI sites surrounding the mutated region. This region was then replaced in the full length KIBRA with Drd1 and XbaI.

Microinjection was performed as previously described [50]. Various dominant negative constructs were microinjected into L7s or sensory neurons by a pneumatic picopump (PV820, World Precision Instruments). A solution containing a construct (0.3 μg/μl DNA in distilled water) plus 0.2% fast green was microinjected into the nuclei of L7 or sensory neurons on day 1 (dn-calpains; see Figure 6A) or on day 3 (dn-KIBRA; see Figure 5A, Figure S5A and Figure S6A), or into the nuclei of L7 on day 3 (dn-PKMs; see Figure 3A). The tip of the micropipette was inserted into the cell nucleus. A short pressure pulse (20–50 ms duration; 20–40 psi) was delivered until the nucleus of the sensory neurons became uniformly green. Because the volume of the nucleus in L7 is about 30X that of the nucleus of the sensory neuron, a pressure pulse (20– 40 psi) of 600–1200 ms was delivered until the nucleus of L7 became uniformly green. The fluorescent images of construct expression were captured with a Nikon Diaphot microscope 24 hours after injections. For stabilization experiments molar equivalent levels of plasmids encoding Kibra, dn-KIBRA and the mRFP-PKMs were used. For images of eGFP and mRFP-PKMs (Figure 4D), sensory neurons were imaged live using a LSM 710 (Zeiss) laser confocal scanning microscope equipped with an Axiovert 100 inverted microscope (Zeiss) and a 63×, NA 1.4 objective. The eGFP and mRFP images were acquired sequentially.

Antibodies

KIBRA C-terminal peptide (N-terminal LESFFHDDRIGEEV C-terminal) was synthesized for both antigen production and antibody purification. Peptide was coupled to a bovine serum albumin-Maleimide and Sulfo-link (Pierce, Rockford, IL, USA) according to the manufacturer’s instruction. After conjugation to BSA-Maleimide, rabbits were injected and final serum (after three boosts) was affinity purified on the Sulfo-link column. KIBRA C-term antibody (~1.32 μg/μL) was used for immunocytochemistry [23, 24] at a 1:1000 concentration.

Immunocytochemistry

Cells were fixed 24 hours following microinjection with 4% paraformaldehyde with 30% sucrose in PBS for 30 min and washed with PBS. Fixed cells were permeabilized with 0.1% Triton X-100 with 30% sucrose in PBS for 10 min, and washed briefly with PBS. Free aldehydes were quenched with 50 mM ammonium chloride for 10–15 min followed by brief PBS wash. To block nonspecific antibody binding, cells were incubated with 10% normal goat serum (Sigma) plus 0.5% Triton X-1000 in PBS for 30 min. Cells were incubated with KIBRA C-term antibody in blocking solution (1:1000) for 1 hour, followed by 4 PBS washes of 5–10 min. Cells were then incubated in the dark with Alexa Fluor 647-conjugated goat anti-rabbit antibody in blocking solution (1:200, Invitrogen) for 1 hour and washed as above. Images were captured by a LSM 710 (Zeiss) laser confocal scanning microscope equipped with an Axiovert 100 inverted microscope (Zeiss) and a 63×, NA 1.4 objective.

QUANTIFICATION AND STATISTICAL ANALYSIS

All data are expressed as the mean ± SEM (Standard Error of the Mean) produced by the indicated treatments. The EPSP amplitude on day 0 was normalized as 100%. The changes in EPSP amplitudes were measured by dividing the EPSP amplitudes on subsequent days by the EPSP amplitude on day 0 multiplied by 100%. To measure the kinetics of HSD on day 3 (Figure S2), the EPSP amplitude evoked by first stimulation on day 3 is normalized as 100% (total 8 stimuli). Some cultures expressing persistent LTF (less than 5%) had an increase in synaptic strength that was expressed as an action potential evoked in L7 were excluded from all analyses. Since previous experiments on cultures containing one sensory neuron and one L7 [18, 19, 22] indicated that 2 days of the different stimulations evoked different forms of persistent LTF that are reversed with different treatments, the number of cultures per control or experimental groups for each experiment ranged from 6–10 samples. Based on these earlier studies, an N of 6 per group (with 6 groups) gave statistical power of > 96.8% probability of rejecting the null hypothesis. The IBM SPSS 24.0 statistics software package was used to test for significance. A two-way ANOVA (group × repeated measures) was first used to assess overall significant difference. If significant, we then performed pairwise post hoc tests (Bonferroni) at each time point to determine the consequences of different types of treatments on persistent plasticity evoked in stimulated groups or on control synaptic baseline. The comparisons included: A control group versus all other control or stimulated groups (designated with *P < 0.05 or **P < 0.01), and stimulated alone [2X (Tet + 5-HT) or 2X (5X5-HT)] versus other stimulated groups (designated with ^#P < 0.05 or ^##P < 0.01). In all experiments we found no significant differences between the control group versus all other control groups.

Images for the expression of constructs in co-cultures were viewed with a Nikon Diaphot microscope attached to a silicon-intensified target (SIT) (Dage 68; Dage-MTI) video camera. Fluorescent intensity (arbitrary units) was measured by the Microcomputer-Controlled Imaging Device (MCID) software package (Imaging Research). All images of construct expression were quantified blindly.

For analysis of PKM stabilization (Figure 4), single processes for each neuron were outlined using NIH Image J and values for red and green fluorescence measured. Background fluorescence was subtracted from each image. Each experiment contained three injected plasmids encoding an mRFP-PKM, eGFP, and either vector, KIBRA or KIBRA_AAA. The red/green ratios for all three groups were normalized to the average ratio seen in vector-injected neurons in that experiment. For each PKM, a one-way ANOVA of the normalized ratio (n = neuron) was performed with Bonferroni post hoc tests. All quantification was done blindly.