Reinstatement of long-term memory following erasure of its behavioral and synaptic expression in Aplysia

Long-term memory (LTM) is believed to be stored in the brain as changes in synaptic connections. Here, we show that LTM storage and synaptic change can be dissociated. Cocultures of Aplysia sensory and motor neurons were trained with spaced pulses of serotonin, which induces long-term facilitation. Serotonin (5HT) triggered growth of new presynaptic varicosities, a synaptic mechanism of long-term sensitization. Following 5HT training, two antimnemonic treatments—reconsolidation blockade and inhibition of PKM—caused the number of presynaptic varicosities to revert to the original, pretraining value. Surprisingly, the final synaptic structure was not achieved by targeted retraction of the 5HT-induced varicosities but, rather, by an apparently arbitrary retraction of both 5HT-induced and original synapses. In addition, we find evidence that the LTM for sensitization persists covertly after its apparent elimination by the same antimnemonic treatments that erase learning-related synaptic growth. These results challenge the idea that stable synapses store long-term memories.

There is significant empirical support for the idea, proposed by Ramón Y Cajal more than a century ago (Cajal, 1894), that long-term memories are expressed in the brain, in part, by changes in synaptic connectivity. A corollary of this idea, accepted by many, if not most, modern neuroscientists, is that memories are maintained by persistent molecular and cellular alterations in synaptic structures themselves (Bailey and Kandel, 2008; Kandel et al., 2014). Here, we have tested the idea that long-term memory (LTM) is stored at synapses using the marine mollusk Aplysia californica. The relatively simple behavior and nervous system of this model invertebrate organism offer several major advantages for understanding memory at the level of modifications of individual synapses (Kandel, 2001). A form of learning in Aplysia whose cellular and molecular substrates are particularly well understood is sensitization of the gill- and siphon-withdrawal reflex (Carew et al., 1971; Brunelli et al., 1976; Antonov et al., 1999; Kandel, 2001; Glanzman, 2010). Sensitization of the withdrawal reflex exhibits a long-term (≥24 hr) form (Pinsker et al., 1973) due, in part, to long-term facilitation (LTF) of the monosynaptic connection between the sensory and motor neurons that mediate the reflex (Frost et al., 1985). Importantly, the monosynaptic sensorimotor connection can be reconstituted in dissociated cell culture, and LTF of the in vitro synapse can be induced by training with pulses of serotonin (5HT), the monoaminergic neurotransmitter that mediates sensitization in Aplysia (Brunelli et al., 1976; Glanzman et al., 1989; Marinesco and Carew, 2002).

Cellular and molecular analyses of this form of long-term synaptic plasticity have provided major mechanistic insights into long-term memory in Aplysia (Goelet et al., 1986; Dash et al., 1990; Bartsch et al., 1995; Martin et al., 1997), insights that have generalized to learning and memory in other organisms, including mammals (Yin et al., 1994, 1995; Frey and Morris, 1997; Kogan et al., 1997; Josselyn et al., 2001). Accordingly, we used the in vitro sensorimotor synapse in initial experiments to determine whether LTM is stored at synapses.

Current evidence supports the idea that LTM can be modified or even eliminated under certain circumstances. One of these goes under the rubric of reconsolidation blockade. Here, a stimulus is delivered to an animal that serves to reactivate the LTM for a previous learning experience. If, immediately after delivery of this reminder, the animal is treated with an inhibitor of protein synthesis (Nader et al., 2000), or subjected to electroconvulsive shock (Misanin et al., 1968), the LTM will be apparently eliminated. Based on this evidence, it has been proposed that the reminder stimulus returns the LTM to a labile state in which new protein synthesis is required to reconsolidate the memory; and that inhibition of protein synthesis during this period of reconsolidation can erase the original memory (Nader and Hardt, 2009) (but see Lattal and Abel, 2004). Another manipulation that can apparently erase LTM permanently is inhibition of the constitutively active catalytic fragment of the atypical protein kinase Cζ; the ongoing activity of this catalytic fragment, named PKMζ, appears to be required for the maintenance of several forms of LTM in mammals (Sacktor, 2011); inhibition of PKMζ, in the absence of a reminder, has been reported to abolish consolidated memory (but see Lee et al., 2013; Volk et al., 2013).

We, and others, have shown that the synaptic memory for LTF of in vitro sensorimotor connections can be apparently eliminated when sensorimotor cocultures are treated with a protein synthesis inhibitor following reminder training (Cai et al., 2012; Lee et al., 2012; Hu and Schacher, 2014). In addition, inhibiting the activity of PKM Apl III, the constitutively active fragment of the Aplysia atypical protein kinase C (PKC Apl III) (Villareal et al., 2009; Cai et al., 2011; Bougie et al., 2012), also erases consolidated LTF (Cai et al., 2011). LTF is mediated partly by the growth of new sensory neuron varicosities (Glanzman et al., 1990; Bailey and Kandel, 2008). Here we investigated whether blocking the reconsolidation of the memory for LTF, or inhibiting PKM Apl III, altered this long-term change in presynaptic structure. We found that the synaptic growth induced by LTF training was reversed by these two memory-disrupting manipulations; however, although the overall number of presynaptic varicosities reverted to the original, pretraining level, the resultant morphological pattern of sensorimotor synapses differed significantly from the original one. These results imply that the persistence of memory does not require the stability of particular synaptic connections.

We provide additional support for this idea with data from behavioral experiments in which we show that LTM can be reinstated in intact Aplysia following its apparent disappearance due to reconsolidation blockade or PKM inhibition. Because these two antimnemonic manipulations not only disrupt the behavioral expression of LTM, but also eliminate the synaptic changes—both electrophysiological (Cai et al., 2011, 2012; Lee et al., 2012) and morphological (present data)—closely associated with LTM in Aplysia (Bailey and Chen, 1983, 1988; Frost et al., 1985; Glanzman et al., 1990), our results challenge the idea that the synapse is a cellular site for long-term memory storage in Aplysia. In other behavioral experiments we show that both the disruption of LTM through inhibition of PKM Apl III and LTM induction require epigenetic changes. These results point to the nucleus of neurons as the potential locus of the engram in Aplysia.

In electrophysiological experiments involving 5HT-induced LTF of sensorimotor synapses in dissociated cell culture, we previously showed that a ‘reminder’ stimulus—a single, 5-min pulse of 5HT—could trigger the apparent reconsolidation of synaptic memory, as indicated by the vulnerability of consolidated LTF to disruption by administration of a protein synthesis inhibitor (anisomycin) following the reminder. Specifically, treatment with a single pulse of 5HT and anisomycin 24 hr or more after the induction of LTF reversed the synaptic facilitation (Cai et al., 2012; see also; Lee et al., 2012; Hu and Schacher, 2014). Here we asked whether the blockade of the reconsolidation of LTF also reverses the synaptic growth that underlies LTF (Glanzman et al., 1990).

Sensory neurons (SNs) of established sensorimotor cocultures were labeled with dextran fluorescein, and motor neurons (MNs) with dextran rhodamine, via pressure injection; the neurons were then imaged using laser scanning confocal fluorescence microscopy, and the presynaptic varicosities in contact with a postsynaptic structure (either a motor neurite or the postsynaptic soma) were quantified (Glanzman et al., 1990) (Figure 1A,B). After initial imaging some cocultures received five spaced 5-min pulses of serotonin (5X5HT training, 100 µM). 24 hr later the cocultures were reimaged and the varicosities requantified. As previously reported (Glanzman et al., 1990), the 5X5HT training produced a significant increase in the number of presynaptic varicosities contacting the MN (Figure 1C). After reimaging, two groups of the trained cocultures received a single 5-min pulse of 5HT (1X5HT, 100 µM) to reactivate the synaptic memory induced by 5X5HT training (Cai et al., 2012). Immediately following the reminder pulse of 5HT, one group (5X5HT-1X5HT-Aniso group) received anisomycin (10 µM) treatment for 2 hr. Two other groups of 5X5HT-trained cocultures did not get the reminder pulse of 5HT; one of these groups (5X5HT-Aniso) received the anisomycin treatment, whereas the other group (5X5HT) received vehicle solution instead. A final group of cocultures (Controls) did not receive either 5HT or anisomycin; instead, the Controls were treated with vehicle solution at the experimental times that the 5X5HT training and anisomycin treatment were administered to other groups. At 48 hr after the original imaging session all of the cocultures were imaged and the varicosities once more quantified.10.7554/eLife.03896.003Figure 1.Blockade of memory reconsolidation reverses 5HT-induced synaptic growth.(A) Experimental protocol.

The vertical blue bars represent pulses of 5HT, and the horizontal red bar represents anisomycin/vehicle treatment. A reminder pulse of 5HT (single blue bar) was delivered to the 5X5HT-1X5HT-Aniso cocultures prior to the anisomycin. (B) Sample confocal micrographs of a Control coculture. Blue arrowheads, presynaptic varicosities present at 0 hr; red arrowheads, new varicosities that appeared during 0–24 hr; and yellow arrowheads, varicosities formed during the 24–48 hr period. Scale bar, 20 μm. (C) Mean normalized varicosity number at 24 hr and 48 hr in the Control (n = 21), 5X5HT (n = 21), 5X5HT-Aniso (n = 19), and 5X5HT-1X5HT-Aniso (n = 26) groups. The number of varicosities measured at 24 hr and 48 hr was normalized to the number present at 0 hr (Note that differences in the data for the 5X5HT and 5X5HT-1X5HT groups were not significant, and two groups have been grouped together in the graph. The separate data for the 5X5HT-1X5HT group can be found in Figure 1—figure supplement 1). A repeated-measures ANOVA indicated that there was a significant group × time interaction (F[3,83] = 9.5, p < 0.0001). Planned comparisons using one-way ANOVAs indicated that the group differences at 24 and 48 hr were highly significant (F[3,83] = 8.6 for 24 hr and 12.5 for 48 hr; p < 0.0001 for the results of each ANOVA). SNK posthoc tests on the 24 hr data showed that the mean normalized varicosity number in the 5X5HT group (192.8 ± 12.7%), 5X5HT-Aniso group (180.9 ± 11.6%), and 5X5HT-1X5HT-Aniso group (179.5 ± 14.6%) was each significantly greater than that in the Control group (110.3 ± 8.5%; p < 0.001 for all comparisons). The increase in varicosity number persisted to 48 hr in the 5X5HT group (197.9 ± 12.2%, p < 0.001) and 5X5HT-Aniso group (191.6 ± 13.2%, p < 0.001), but not in the 5X5HT-1X5HT-Aniso group (124.4 ± 10.9%), when compared to the Control group (108.9 ± 15.2%).

The difference between the 5X5HT-Aniso and 5X5HT-1X5HT-Aniso groups was highly significant (p < 0.001). Asterisks indicate a significant difference for comparisons with the Control group; plus signs indicate a significant difference for comparisons with the 5X5HT-1X5HT-Aniso group. Here and in subsequent figures one symbol indicates p < 0.05; two symbols, p < 0.01; three symbols, p < 0.001. Error bars represent ±SEM.DOI: http://dx.doi.org/10.7554/eLife.03896.00310.7554/eLife.03896.004Figure 1—figure supplement 1.Graphs presenting the normalized mean varicosity numbers for the three 5HT-trained groups not subjected to reconsolidation blockade.The mean normalized varicosity number at 24 hr was 192.7 ± 15.0% in the 5X5HT group (n = 18), 193.3 ± 17.6% in the 5X5HT-1X5HT group (n = 3), and 180.9 ± 11.6% in the 5X5HT-Aniso group (n = 19). At 48 hr the mean normalized varicosity number was 192.9 ± 13.7% in the 5X5HT group, 228.3 ± 17.4% in the 5X5HT-1X5HT group, and 191.6 ± 13.2% in the 5X5HT-Aniso group. A repeated-measures ANOVA indicated that neither the interaction (F[2,37] = 0.8, p = 0.45), nor the overall effects of group (F[2,37] = 0.3, p = 0.73), was significant.DOI: http://dx.doi.org/10.7554/eLife.03896.004

The overall increase in the number of presynaptic varicosities induced by 5X5HT training on Day 1 persisted through the third imaging session in the 5X5HT and 5X5HT-Aniso groups of cocultures (Figure 1C). (Notice that because there were no significant differences between the 5X5HT and 5X5HT-1X5HT groups, these two groups have been combined into a single group [5X5HT] in Figure 1C. However, the data for the 5X5HT-1X5HT group are presented separately in the graph in Figure 1—figure supplement 1). The reminder pulse of 5HT coupled with inhibition of protein synthesis caused the number of varicosities to revert to the pretraining (0 hr) value in the 5X5HT-1X5HT-Aniso group. This structural result parallels the electrophysiological results previously reported for Aplysia sensorimotor cocultures (Cai et al., 2012; Lee et al., 2012; Hu and Schacher, 2014), and provides additional support for the notion that 1X5HT reactivated the synaptic memory induced by the 5X5HT training.

Besides quantifying changes in overall varicosity number, we tracked the fate of each SN varicosity in every coculture over the course of the experiments. The varicosities were put into one of three categories: ‘original’ varicosities, that is, the varicosities present at 0 hr; ‘5HT-induced’ varicosities, the varicosities that appeared during the 24 hr after 5HT treatment (this category pertained only to the groups given 5X5HT training); and ‘new’ varicosities, varicosities formed during the 24–48 hr period. For the analysis of varicosity fate the three groups of cocultures that received the 5X5HT training without reconsolidation blockade—that is, the 5X5HT, 5X5HT-1X5HT and 5X5HT-Aniso groups—were consolidated into a single group labeled ‘5HT-No reconsolidation/No blockade’ in Figure 2. Inspection of the fates of individual varicosities in this group yielded surprising results. First, many of the 5HT-induced SN varicosities in the 5HT-No reconsolidation/No blockade group did not persist until the final imaging session (Figures 2A and 3A); instead, there was significant retraction of the 5HT-induced varicosities between 24 and 48 hr in this group. Second, there was also retraction of the original varicosities during this period (Figures 2B and 3A). Varicosities in the 5X5HT-1X5HT-Aniso group—the group of trained cocultures subjected to reconsolidation blockade—exhibited a similar pattern of retraction of 5HT-induced and original varicosities, but the amount of retraction was significantly greater than that observed in the 5HT-No reconsolidation/No blockade cocultures (Figures 2A,B and 3B). A third unexpected finding was the substantial growth of new varicosities in all trained groups between 24 and 48 hr; the amount of the growth was significantly greater, however, in the 5-HT-No reconsolidation/No blockade cocultures than in the 5X5HT-1X5HT-Aniso group (Figure 2C).

Thus, the 5X5HT trained cocultures subjected to reconsolidation blockade exhibited significantly more retraction and significantly less growth of varicosities during the 24–48 hr period than did the other trained cocultures. Interestingly, whereas there was greater retraction of 5HT-induced varicosities than of original varicosities between 24–48 hr in the 5HT-No reconsolidation/No blockade group, the retraction of 5HT-induced varicosities and original varicosities did not differ in the 5X5HT-1X5HT-Aniso group (Figure 2—figure supplement 1).10.7554/eLife.03896.005Figure 2.Effect of reconsolidation blockade on the fates of individual varicosities.For varicosities in the 5HT-induced and original categories, the varicosity count in a given category for each coculture at 48 hr was normalized to the number of varicosities in the same category in that coculture at 24 hr. For the varicosities formed >24 hr, the varicosity count for each coculture at 48 hr was normalized to the total number of varicosities in that coculture at 24 hr. (A) The mean normalized decrease in the number of 5HT-induced varicosities (varicosities that appeared within 24 hr after 5X5HT training) at 48 hr was 67.9 ± 4.2% in the 5X5HT-1X5HT-Aniso group, which was greater than that in the 5HT-No reconsolidation/No blockade group (47.7 ± 3.7%, p < 0.001, two-tailed t test). (B) Retraction of original varicosities in the 5X5HT-1X5HT-Aniso group (56.4 ± 6.1%) was significantly greater than in the 5HT-No reconsolidation/No blockade group (32.3 ± 5.1%, p < 0.01, two-tailed t test). (C) Reversal of the morphological changes induced by 5X5HT training also involved inhibition of synaptic growth from 24–48 hr.

There was significantly less growth of new varicosities during this period in the 5X5HT-1X5HT-Aniso group (36.8 ± 5.2%) than in the 5HT-No reconsolidation/No blockade group (52.9 ± 4.2%, p < 0.05, two-tailed t test). Blue bars, 5HT-No reconsolidation/No blockade; red bars, 5X5HT-1X5HT-Aniso group. Asterisks indicate significant differences between the 5HT-No reconsolidation/No blockade and 5X5HT-1X5HT-Aniso groups. Error bars represent ±SEM.DOI: http://dx.doi.org/10.7554/eLife.03896.00510.7554/eLife.03896.006Figure 2—figure supplement 1.Comparison of retraction of original and 5HT-induced varicosities between 24–48 hr in the reconsolidation blockade experiments.(A) The percent retraction of 5HT-induced varicosities (47.7 ± 3.7%) was significantly greater than that of the original varicosities (32.3 ± 5.1%, p < 0.02, two-tailed t test) in the 5HT-No reconsolidation/No blockade group. (B) The percent retraction of the 5HT-induced varicosities (67.9 ± 4.2%) did not differ significantly from that of original varicosities (56.4 ± 6.1%) in the 5X5HT-1X5HT-Aniso group.DOI: http://dx.doi.org/10.7554/eLife.03896.00610.7554/eLife.03896.007Figure 3.Confocal fluorescence micrographs illustrating the structural effects of 5X5HT training, reconsolidation blockade, and chelerythrine treatment.(A) Coculture that received 5X5HT training alone. (B) Coculture that received the 5X5HT training plus 1X5HT and anisomycin treatment immediately after the 24 hr imaging session. (C) Coculture that received the 5X5HT training plus chelerythrine treatment immediately after the 24 hr imaging session. Blue arrowheads, varicosities present at 0 hr; red arrowheads, varicosities formed during 0–24 hr; and yellow arrowheads, varicosities formed during 24–48 hr. Scale bars, 20 μm.DOI: http://dx.doi.org/10.7554/eLife.03896.007

The substantial morphological changes in the 5HT-No reconsolidation/No blockade cocultures between 24 and 48 hr were surprising, because the overall number of varicosities remained stable in these cocultures during this period (see the data for the 5X5HT and 5X5HT-Aniso groups in Figure 1C). Moreover, as demonstrated in our earlier electrophysiological investigations of LTF, the amplitude of the sensorimotor EPSP was also stable between 24 and 48 hr after 5X5HT training (Cai et al., 2011). Our morphological data reveal the operation of a heretofore unrecognized homeostatic mechanism that adjusts the number of SN varicosities—and, presumably, active synaptic sites (Schacher et al., 1990; Kim et al., 2003)—for sensorimotor connections according to their learning-related experience, but that seems unconcerned with the identities of the individual varicosities/synaptic sites. In the 5HT-No reconsolidation/No blockade cocultures the significant retraction of SN varicosities during 24–48 hr after 5X5HT training was compensated for by significant growth of new varicosities, so that the overall number remained constant (Figures 1C and 2). By contrast, the number of varicosities was reset to the original (0 hr) value in the 5X5HT-1X5T-Aniso group; this resetting involved increased retraction of both 5HT-induced and original varicosities, and decreased growth of new varicosities between 24 and 48 hr. Consequently, the morphological state of the cocultures in the 5X5HT-1X5T-Aniso group at 48 hr differed significantly from that at 0 hr with respect to the identities of individual SN varicosities.

Next we examined the effect of inhibiting PKM on SN varicosity number. PKM Apl III, a homolog of mammalian PKMζ (Sacktor, 2011), is formed from the atypical Aplysia protein kinase C (PKC Apl III) (Bougie et al., 2009) by calpain-dependent cleavage; furthermore, this cleavage is induced by prolonged treatment with 5HT (Bougie et al., 2012). In previous behavioral and electrophysiological investigations we found that inhibition of PKM Apl III, either with the pseudosubstrate sequence of the regulatory domain of atypical PKC (ZIP) or with chelerythrine, a PKC inhibitor selective for PKM at low concentrations (Ling et al., 2002; Villareal et al., 2009), abolished both consolidated long-term sensitization and LTF (Cai et al., 2011). To determine whether inhibition of PKM Apl III reverses the structural growth that mediates long-term sensitization (Bailey and Chen, 1983, 1988) and LTF (Glanzman et al., 1990), we gave some cocultures (5X5HT-Chel group) 5X5HT training followed 24 hr later by treatment with chelerythrine (Figure 4A). Another group of cocultures (5X5HT group) received the 5X5HT training alone. Finally, a third group (Controls) received the vehicle solution instead of either 5HT or chelerythrine. The SNs and MNs of all of the cocultures were labeled with fluorescent dye as before and then imaged on Day 1 (0 hr) of the experiment. Immediately afterwards cocultures in the 5X5HT and 5X5HT-Chel groups were given 5X5HT training. All cocultures were imaged for a second time at 24 hr, after which cocultures in the 5X5HT-Chel group received chelerythrine (10 µM, 1 hr), while the other two groups received the vehicle solution.

(The 5X5HT-Chel group did not receive 1X5HT prior to chelerythrine treatment.) The cocultures were imaged for a final time at 48 hr.10.7554/eLife.03896.009Figure 4.Inhibition of PKM also reverses 5HT-induced synaptic growth.(A) Experimental protocol. The vertical blue bars represent pulses of 5HT; the horizontal red bar represents the period of chelerythrine/vehicle treatment. The 5X5HT and Control groups received vehicle treatment during this time. (B) Mean SN varicosity number measured at 24 hr and 48 hr for the Control (n = 14), 5X5HT (n = 21) and 5X5HT-Chel (n = 14). The number of varicosities measured at 24 hr and 48 hr was normalized to the number of varicosities at 0 hr. A repeated-measures ANOVA indicated that there was a significant group × time interaction (F[2,46] = 7.5, p < 0.001). Planned comparisons showed that the group differences were significant for both the 24 and 48 hr measurements (24 hr, F[2,46] = 3.7, p < 0.04; 48 hr, F[2,46] = 12.8, p < 0.0001). The mean normalized varicosity number at 24 hr was 171.2 ± 14.0% in the 5X5HT group, 166.9 ± 22.8% in the 5X5HT-Chel group, and 114.0 ± 8.9% in the Control group. Posthoc comparisons indicated that the 5X5HT training produced a significant increase in varicosity number in the 5X5HT and 5X5HT-Chel groups at 24 hr (p < 0.05 for both comparisons with the Control group). The mean normalized varicosity number at 48 hr in the 5X5HT group (197.9 ± 18.7%) was greater than that in the Control group (104.4 ± 7.2%, p < 0.001), but the 5X5HT-Chel group mean (116.1 ± 9.5%) was not significantly different from the Control mean. The difference between the mean varicosity numbers for the 5X5HT and 5X5HT-Chel groups was highly significant (p < 0.001). Asterisks, comparison between 5X5HT and Control groups; plus signs, comparison between 5X5HT and 5X5HT-Chel groups.

Error bars represent ±SEM.DOI: http://dx.doi.org/10.7554/eLife.03896.009

As before, the 5X5HT training produced a significant net increase in the number of presynaptic varicosities in the two trained groups at 24 hr (Figure 4B). This net increase persisted in the 5X5HT group, but was reversed in 5X5HT-Chel group at 48 hr. Monitoring of the fate of individual varicosities in the 5X5HT group revealed the same pattern of structural growth and retraction observed previously in the 5X5HT-trained cocultures not subjected to reconsolidation blockade. Specifically, some varicosities induced by the 5X5HT training were lost between 24 and 48 hr, as were some of the original varicosities; this loss, however, was compensated for by the growth of new varicosities; consequently, the overall varicosity number remained stably elevated during the 24–48 hr period (Figures 3C and 5). The varicosities in the 5X5HT-Chel group exhibited significantly greater retraction and significantly less growth between 24 and 48 hr than those in the 5X5HT group, resulting in an overall loss of varicosities. Thus, the number of varicosities in the 5X5HT-Chel group at 48 hr was returned to the value at 0 hr. This result indicates that inhibiting PKM Apl III, like reconsolidation blockade (Figure 1C), engages a homeostatic mechanism that resets the presynaptic varicosities in sensorimotor cocultures to the number present prior to training.

Another similarity between the morphological effects of reconsolidation blockade and inhibition of PKM was that there was equal retraction of 5HT-induced and original varicosities in the 5X5HT-Chel cocultures during 24–48 hr, whereas, as before, the retraction of 5HT-induced varicosities was greater in the 5X5HT cocultures during this period (Figure 5—figure supplement 1).10.7554/eLife.03896.010Figure 5.Effect of inhibition of PKM on varicosity fate.For the 5HT-induced and original varicosities, the number of varicosities in each category for each coculture at 48 hr was normalized to the number in the same category in that coculture at 24 hr. For the varicosities formed >24 hr, the number of varicosities for each coculture at 48 hr was normalized to the total number of varicosities in the coculture at 24 hr. (A) The mean normalized decrease in the number of 5HT-induced varicosities at 48 hr was 53.4 ± 5.8% in the 5X5HT group and 68.0 ± 5.8% in the 5X5HT-Chel group (p < 0.05, one tailed t test). (B) The mean normalized decrease in the number of original varicosities at 48 hr was 23.5 ± 4.7% in the 5X5HT group and 63.4 ± 9.3% in the 5X5HT-Chel group (p < 0.001, two-tailed t test). (C) More presynaptic varicosities were formed during 24–48 hr in the 5X5HT group (63.9 ± 9.0%) than in the 5X5HT-Chel group (43.2 ± 5.8%, p < 0.05, one-tailed t test). Blue bars, 5X5HT group; red bars, 5X5HT-Chel group. Asterisks, comparison between 5X5HT and 5X5HT-Chel groups.

Error bars represent ±SEM.DOI: http://dx.doi.org/10.7554/eLife.03896.01010.7554/eLife.03896.011Figure 5—figure supplement 1.Comparison of retraction of original and 5HT-induced varicosities between 24–48 hr in the chelerythrine treatment experiments.(A) The difference in the percent retraction between 5X5HT-induced (53.4 ± 5.8%) and original (23.5 ± 4.7%) varicosities was highly significant in the 5X5HT group (p < 0.001, two-tailed t test). (B) There was no significant difference in the percent retraction between 5HT-induced (68.0 ± 5.8%) and original (63.4 ± 9.3%) varicosities in the 5X5HT-Chel group.DOI: http://dx.doi.org/10.7554/eLife.03896.011

Inspection of the synaptic structure of the Control cocultures revealed a complex underlying pattern of structural change similar to that observed in the reconsolidation experiment. There was significant growth and retraction of SN varicosities, both original and new, over the 48 hr of the experiments (Figures 6 and 1B). This structural dynamism contrasted with the constancy of overall varicosity number in these cocultures (Figures 1C and 4B), as well as with the stability of the sensorimotor EPSP amplitude observed in Control cocultures over similar time periods in our earlier studies (Cai et al., 2011, 2012).10.7554/eLife.03896.008Figure 6.Changes in presynaptic varicosities in Control cocultures.(A) The normalized increase in the number of varicosities was 54.9 ± 6.5% during the 0–24 hr period and 41.5 ± 7.2% during the 24–48 hr period (green bars). (B) The mean normalized number of original varicosities (blue bar) that retracted during the 0–24 hr period was 43.2 ± 4.4%. During the 24–48 hr period 15.5 ± 2.0% of the original varicosities (those present at 0 hr, blue segment), and 30.2 ± 5.1% of the varicosities that formed during 0–24 hr period (red segment), retracted; thus, the mean total normalized retraction of varicosities from 24–48 hr was 45.7%. Note that the numbers of newly formed varicosities during the 0–24 and 24–48 hr periods in each coculture were normalized to the total number of varicosities in that coculture measured at 0 hr. The percentage of retracted varicosities was obtained by normalizing the number of varicosities that disappeared during the 0–24 and 24–48 hr periods for a given coculture to the total number of varicosities in that coculture at 0 hr.DOI: http://dx.doi.org/10.7554/eLife.03896.008

The present morphological results challenge the notion that the persistence of sensitization memory in Aplysia depends on the persistence of particular facilitated synapses. To further investigate this idea, we tested whether the LTM for behavioral sensitization can be reinstated in Aplysia following reconsolidation blockade and inhibition of PKM, two treatments previously shown to eliminate LTF, the synaptic basis of long-term sensitization (Cai et al., 2011, 2012; Lee et al., 2012; Hu and Schacher, 2014).

Long-term sensitization (LTS) of the siphon-withdrawal reflex (SWR) was induced in intact Aplysia using five bouts of tails shocks (5XTrained). Brief reminder training (one bout of tail shocks, 1XTrained) was applied at 48 hr after the original training to trigger reconsolidation of the LTM for sensitization (Cai et al., 2012). Immediately following the reminder training an intrahemocoel injection of anisomycin was administered to some animals (Figure 7A). As previously reported (Cai et al., 2012; Lee et al., 2012), this treatment eliminated LTS, assessed here at 72 hr (Figure 7B). Afterwards, some of the animals received three additional bouts of tail shocks (3XTrained). Importantly, the three additional bouts of sensitization training did not induce LTM in naïve animals (Control-Veh-3XTrained; Figure 7B). However, the three bouts of training completely restored LTM following its disruption by reconsolidation blockade (Figure 7B).10.7554/eLife.03896.012Figure 7.Reinstatement of LTS after its elimination by reconsolidation blockade in Aplysia.(A) Experimental protocol. The timing of the pretests, training, posttests, and drug/vehicle injections is shown relative to the end of the last training session. The time of the intrahemocoel injection of either anisomycin or vehicle is indicated by the red arrow. Animals in the 5XTrained-1XTrained-Aniso and 5XTrained-1XTrained-Aniso-3XTrained groups received a reminder episode of sensitization training (one bout of tail shocks) immediately after the 48 hr posttest (black bar) and prior to the injection of anisomycin, whereas the animals in the 5XTrained-Aniso group did not. After the 72 hr posttest animals in the Control-Veh-3XTrained and 5XTrained-1X5HT-Aniso-3XTrained groups received truncated sensitization training (three bouts of tail shocks).

(B) The mean duration of the SWR measured at 48 hr, 72 hr and 96 hr for the Control-Veh-3XTrained (n = 12), 5XTrained-Aniso (n = 4), 5XTrained-1XTrained-Aniso (n = 7), and 5XTrained-1XTrained-Aniso-3XTrained (n = 11) groups. A repeated-measures ANOVA indicated that there was a significant group × time interaction (F[9,90] = 51.0, p < 0.0001). Subsequent planned comparisons indicated that the overall differences among the four groups were highly significant on all of the posttests (48 hr, F[3,30] = 137.2, p < 0.0001; 72 hr, F[3,30] = 50.5, p < 0.0001; and 96 hr, F[3,30] = 43.8, p < 0.0001). SNK posthoc tests on the 48 hr data indicated that the initial sensitization training produced significant LTS in the 5XTrained-Aniso (53.0 ± 7.0 s), 5XTrained-1XTrained-Aniso (59.9 ± 0.2 s), and 5XTrained-1XTrained-Aniso-3XTrained (53.7 ± 2.9 s) groups compared to the Control-Veh-3XTrained group (2.5 ± 0.9 s, p < 0.001 for each test). The responses of the trained groups did not differ significantly at 48 hr after sensitization training. However, the mean duration of the SWR in the 5XTrained-Aniso group (53.8 ± 6.3 s) remained prolonged at 72 hr, and was significantly longer than that in the 5XTrained-1XTrained-Aniso group (6.4 ± 1.9 s) as well as in the 5XTrained-1XTrained-Aniso-3XTrained group (4.8 ± 3.2 s, p < 0.001 for both comparisons). LTS was restored by the three additional tail shocks applied after the 72 hr posttest. The mean duration of the SWR in the 5XTrained-1XTrained-Aniso-3XTrained group at 96 hr was 56.7 ± 2.7 s, which was significantly longer than that for the 5XTrained-1XTrained-Aniso group (10.4 ± 3.9 s) at 96 hr (p < 0.001). The SWR of the 5XTrained-Aniso group at 96 hr (53.8 ± 3.8 s) was also significantly longer than that in the 5XTrained-1XTrained-Aniso group (p < 0.001).

Asterisks, comparisons of the Control-Veh-3XTrained group with the 5XTrained-Aniso group, the 5XTrained-1XTrained-Aniso group, and 5XTrained-1XTrained-Aniso-3XTrained group at 48 hr; comparison of the Control-Veh-3XTrained group with the 5XTrained-Aniso group at 72 hr; and comparisons of the Control-Veh-3XTrained group with the 5XTrained-Aniso group and the 5XTrained-1XTrained-Aniso-3XTrained group at 96 hr. Plus signs, comparisons of the 5XTrained-1XTrained-Aniso group with the 5XTrained-Aniso group and the 5XTrained-1XTrained-Aniso-3XTrained group at 96 hr. Error bars represent ±SEM.DOI: http://dx.doi.org/10.7554/eLife.03896.012

To test whether LTS could be reinstated after its apparent erasure by inhibition of PKM Apl III, animals were again initially trained using five bouts of tail shocks. 24 hr after the original sensitization training the animals received an intrahemocoel injection of chelerythrine (Figure 8A). Animals treated with chelerythrine exhibited no LTS at 48 hr (Figure 8B), confirming our previous finding (Cai et al., 2011). The animals that received the original sensitization training were then given three additional bouts of tail shocks (5XTrained-Chel-3XTrained group), as were control animals that had not received the original sensitization training (Control-Veh-3XTrained group). The modest additional training reinstated LTS following its apparent erasure by chelerythrine, but did not produce LTS in the control animals (Figure 8B).10.7554/eLife.03896.013Figure 8.Reinstatement of LTS following its erasure by PKM inhibition.(A) Experimental protocol. The timing of the pretests, training, posttests, and drug/vehicle injections is shown relative to the end of the last training session. The time of the intrahemocoel injection of either chelerythrine or vehicle is indicated by the red arrow. After the 48 hr posttest animals in both groups received three additional bouts of tail shocks. (B) Mean duration of the SWR measured at 24 hr, 48 hr and 72 hr for the Control-Veh-3XTrained (n = 9) and 5XTrained-Chel-3XTrained (n = 9) groups. A repeated-measures ANOVA indicated that there was a significant interaction effect between group and time (F[3,14] = 25.4. p < 0.0001). A planned comparison indicated that the mean duration of the SWR at 24 hr in the 5XTrained-Chel-3XTrained group (44.7 ± 5.6 s) was significantly longer than that in the Control-Veh-3XTrained group (1.1 ± 0.1 s, F[1,16] = 60.2, p < 0.001).

The mean duration of the SWR at 48 hr (4.1 ± 1.4 s) in the 5XTrained-Chel-3XTrained group was not significantly longer than that in the Control-Veh-3XTrained group (2.6 ± 0.8 s), as indicated by a planned comparison (F[1,16] = 0.95, p = 0.34). The SWR in the 5XTrained-Chel-3XTrained group at 72 hr was 48.7 ± 5.0 s, which was significantly longer than that in the Control-Veh-3XTrained group (4.6 ± 1.2 s; planned comparison, F[1,16] = 73.1, p < 0.001). Asterisks, comparison between 5XTrained-Chel-3XTrained and Control-Veh-3XTrained groups. Error bars represent ±SEM.DOI: http://dx.doi.org/10.7554/eLife.03896.013

The above behavioral results indicate that LTM in Aplysia can persist despite elimination of its behavioral expression by reconsolidation blockade and inhibition of PKM Apl III. Moreover, because these two antimnemonic treatments also eliminate LTF (Cai et al., 2011, 2012; Lee et al., 2012; Hu and Schacher, 2014) (and Figures 1C and 4B), the results imply that some component of LTM, perhaps a priming process, may persist in the absence of synaptic alterations. Possibly, LTM, or the primer for LTM, resides in the nuclei of neurons within the SWR circuitry, encoded as epigenetic changes (Levenson and Sweatt, 2005; Rahn et al., 2013; Graff and Tsai, 2013a). To investigate this possibility, we tested whether chelerythrine's apparent disruption of LTM involves alterations of chromatin structure. For this test we used the histone deacetylase (HDAC) inhibitor trichostatin A (TSA) (Graff and Tsai, 2013b). Aplysia were given five bouts of tail shocks, and the strength of the SWR was tested 24 hr later (Figure 9A). Immediately after this test the sensitization-trained animals received an intrahemocoel injection of the vehicle solution alone or TSA; 10–15 min later some of the animals treated with TSA were given an injection of chelerythrine (5XTrained-TSA-Chel group). (Untrained control animals received an injection of the vehicle solution after the 24 hr test.) In addition, another group of trained animals were given the chelerythrine without a prior injection of TSA (5XTrained-Chel group). All animals were retested at 48 hr. TSA treatment blocked the disruption of LTS by chelerythrine (Figure 9B). The injection of TSA 24 hr after training by itself did not alter LTS (5XTrained-TSA group).

These results indicate that one of chelerythrine's antimnemonic actions is to reverse histone acetylation induced by LTS training.10.7554/eLife.03896.014Figure 9.Epigenetic regulation of LTM in Aplysia.(A) Experimental protocol. The times of the pretests, training, posttests, and drug/vehicle injections are shown relative to the end of the last training session. The times of the trichostatin A (TSA)/vehicle and chelerythrine/vehicle injections are indicated by blue arrow and red arrow respectively. (B) TSA treatment blocks erasure of LTS by chelerythrine. The graph presents the mean duration of the SWR measured at 24 hr and 48 hr in the Control-Veh (n = 9), 5XTrained-Veh (n = 5), 5XTrained-TSA (n = 6), 5XTrained-Chel (n = 5), and 5XTrained-TSA-Chel (n = 6) groups. A repeated-measures ANOVA showed a significant group × time interaction (F(8,52) = 53.2, p < 0.0001). Subsequent planned comparisons indicated that the group differences for the 24 and 48 hr posttests were highly significant (24 hr, F[4,26] = 45.5, p < 0.0001; 48 hr, F[4,26] = 94.4, p < 0.0001). For the 24 hr posttest, SNK posthoc tests indicated that the training produced significant sensitization in all four trained groups (5XTrained-Veh, 54.0 ± 6.0 s; 5XTrained-TSA, 49.0 ± 7.6 s; 5XTrained-Chel, 59.6 ± 0.4 s; and 5XTrained-TSA-Chel, 57.7 ± 2.3 s) compared to the Control-Veh group (1.6 ± 0.6 s, p < 0.001 for all tests). Comparisons of the four trained groups showed that their responses did not differ significantly on the 24 hr posttest. However, the responses of 5XTrained-Veh (49.0 ± 6.9 s), 5XTrained-TSA (56.0 ± 4.0 s), and 5XTrained-TSA-Chel (59.8 ± 1.7 s) on the 48 hr posttest were significantly more enhanced than those for both the 5XTrained-Chel (2.8 ± 1.8 s) and Control-Veh (1.7 ± 0.5 s) groups (p < 0.001 for all tests).