Search tips
Search criteria 


Logo of hhmipaabout author manuscriptssubmit a manuscriptHHMI Howard Hughes Medical Institute; Author Manuscript; Accepted for publication in peer reviewed journal
Cell. Author manuscript; available in PMC 2012 October 27.
Published in final edited form as:
PMCID: PMC3442366

A role for neuronal piRNAs in the epigenetic control of memory-related synaptic plasticity



Small RNA-mediated gene regulation during development causes long-lasting changes in cellular phenotypes. To determine whether small RNAs of the adult brain can regulate memory storage, a process that requires stable and long-lasting changes in the functional state of neurons, we generated small RNA libraries from the Aplysia CNS. In these libraries we discovered an unexpectedly abundant expression of a 28 nucleotide-sized class of piRNAs in brain, which had been thought to be germ-line specific. These piRNAs have unique biogenesis patterns, predominant nuclear localization, and robust sensitivity to serotonin, a modulatory transmitter important for memory. We find that the piwi/piRNA complex facilitates serotonin-dependent methylation of a conserved CpG island in the promoter of CREB2, the major inhibitory constraint of memory in Aplysia, leading to enhanced long-term synaptic facilitation. These findings provide a small-RNA mediated gene regulatory mechanism for establishing stable long-term changes in neurons for the persistence of memory.


The lifetime of many human memories run on the order of years, whereas the RNA or protein molecules that may subserve these memory traces are thought to turnover on the order of days (Price et al. 2010). Several hypotheses have been proposed to explain how memories can remain stable in the face of constant molecular turnover: 1. Prion-like proteins at synapses can adopt active, stable, and self-perpetuating conformations that preclude turnover of the protein (Si et al. 2003, Bailey et al. 2004, Si et al. 2010), 2. Auto-regulatory and positive feedback loops within protein networks can allow persistent enzymatic activity of proteins or newly synthesized protein to take the place of existing protein machinery without loss in state and function (Lisman et al. 1985, Hayer et al. 2005, Song et al. 2007, Serrano et al. 2008), 3. And perhaps most tantalizing, the possibility that epigenetic mechanisms such as DNA methylation can alter gene-expression and thus the intrinsic properties of neurons in a long-term fashion, perhaps on the order of years (Crick 1984, David & Squire 1984, Weaver et al. 2004, Miller et al. 2010, Feng et al. 2010).

Small regulatory non-coding RNAs can cause long-lasting changes in cellular phenotypes during development, through both their involvement in auto-regulatory feedback loops (Hobert 2008, Rybak et al. 2008, Krol et al. 2010), as well as in the transcriptional and epigenetic regulation of gene-expression (Wassenegger 2005, Saito & Siomi 2010). To better understand the regulatory roles of miRNAs during long-term memory, we previously generated a miRNA library from the Aplysia central nervous system and demonstrated the role of a brain-specific miRNA, aca-miR-124, in constraining long-term synaptic facilitation through repression of the transcriptional activator CREB1 (Rajasethupathy et al. 2009). In the process of mining and characterizing miRNAs from Aplysia CNS, we made the unexpected discovery of neuronally expressed piRNAs (piwi-interacting RNAs).

piRNAs are a class of piwi-associated, 26–32 nucleotide small non-coding RNAs (Aravin et al. 2006, Girard et al. 2006, Grivna et al. 2006, Watanabe et al. 2006) that, unlike other small RNAs, are generated from long genomic clusters (Betel et al. 2007) and are thought to have germline-restricted expression. The function of piwi/piRNA RNP complexes is not fully understood as of yet, but some studies point to a possible role in the epigenetic regulation of transposable elements in the germline through de-novo DNA methylation (Aravin et al. 2007, Brennecke et al. 2008, Kuramochi-Miyagawa et al. 2008). Specifically, mice lacking one or more of its piwi homologs were shown to have substantial de-methylation and de-repression of transposable elements targeted by germline piRNAs.

In a recent study, Lee et al. (2009) reported the identification of piRNAs in neurons. However, it appears that the reported piRNA sequences are fragment of snoRNAs. This mis-annotation is perhaps due to snoRNA impurities that were present in the original piwi immunoprecipitates and therefore mis-classified as piRNAs (Girard et al. 2006). Here we provide a library of bonafide neuronally expressed piRNAs that have been validated in multiple ways: 1. Repeated deep sequencing of brain tissue with verification of discovered piRNAs by northerns, 2. Confirmation of known properties of piRNAs such as 3′end modification and stable association with a neuronally expressed piwi protein, and 3. Extensive bioinformatic analysis to show their distinct patterns of clustering within the genome.

We find that Aplysia piRNAs are broadly expressed outside of the ovotestes, and are amply present in neurons. These piRNAs are abundant, have unique biogenesis patterns, associate with a neuronal piwi protein, and are distinctly regulated by neuromodulators important for learning and memory. By analogy to their role in germline, we find that the piwi/piRNA complex in neurons can methylate target genes, but in this case it targets a critical plasticity-related gene and transcriptional repressor of memory, CREB2 (Bartsch et al. 1995), and methylates its promoter by first accessing its nascent transcript. The DNA methylation of CREB2 by the piwi/piRNA complex provides a mechanism by which transient external stimuli can cause long lasting changes in the gene-expression of neurons involved in long-term memory storage.


Identification of neuronal piRNAs in Aplysia that stably associate with Piwi in nuclear compartments

Our previous generation of a small RNA library from Aplysia CNS resulted in the majority of sequence reads being mapped as miRNAs, with a minority of reads (~ 20 %) that mapped to the Aplysia genome but could not be annotated (Fig. 1A). Further examination of these non-annotated small RNA sequences revealed the unexpected presence in brain of another distinct class of small RNAs characterized by a predominant length of 28 nucleotide and a strong preference for a 5′U (at least 60% of cloned sequence reads for a given piRNA have uridine as the first nucleotide) (Fig. 1A, B). When these sequences were mapped to the unassembled genome trace files followed by the assembly of larger contigs comprising these regions, we identified clusters containing additional sequences with the same features, revealing a pattern characteristic of mammalian piRNAs (see methods for further annotation details). To more comprehensively survey piRNA expression in the Aplysia CNS, as well as other tissues, both in the juvenile animal as well as in the adult, we generated anew 10 different small RNA cDNA libraries using barcoded adapters and subjected the libraries to deep sequencing using the Illumina platform. Of the sequences that were annotated, the piRNA content per library averaged 15 %, compared with the miRNA content which averaged 60 % (Supplementary Tables S1 and S2). We identified 372 distinct piRNA clusters (Scaffold coordinates provided in Supplementary Table S2), of which a region of one representative cluster is shown (Fig. 1B). Aplysia piRNAs exhibit unusual biogenesis patterns in that within a cluster of piRNA reads, one or a few individual piRNAs were cloned hundreds of times more frequently than surrounding piRNAs in the same cluster (Fig 1B). This piRNA biogenesis pattern leads to an accumulation of specific piRNAs similar in read frequencies to miRNAs (Supplementary Table S3).

Figure 1
A. A size histogram of cloned small RNAs from Aplysia CNS revealed two populations, and further characterization confirmed the new class of sequences (shown in black) to be piRNAs.

Because piRNAs are preferentially expressed in germline cells in both vertebrates and invertebrates, we anticipated gonad-specific expression in Aplysia. Although we find the overall piRNA content (and piRNA-to-miRNA ratio) to be highest in the ovotestes (Supplementary Table S1), there are several abundant piRNAs that are selectively enriched in the CNS (Fig. 1C). To confirm the sequencing data, abundant piRNAs originating from 2 distinct clusters were analyzed by quantitative northern blots and detected in brain, as well as in ovotestis and heart, but to a lesser extent in other organs such as muscle or hepatopancreas (Fig. 1D). Since piRNAs in other species are known to be 2′O-methylated at their 3′ends, we asked whether neuronal piRNAs in Aplysia were also 2′-O-methylated at their 3′ends. We subjected total RNA from Aplysia CNS to periodate treatment and beta elimination, followed by northern blot for a piRNA (piR-1) and a miRNA (miR-22). Whereas the miRNA, which was expected to be unmodified at its 3′end, was sensitive to the treatment and showed an approximate 2-nucleotide shift (the expected change in mobility upon treatment and elimination of the last nucleotide with a remaining 3′ phosphate), the piRNA did not shift (Fig. 2E), suggestive of 2′-O-methylation previously documented for piRNAs in other species (Kirino & Mourelatos 2007).

Figure 2
A. The full length 964 kDa Aplysia piwi protein cloned from Aplysia CNS with conserved PAZ and PIWI domains and whose transcript is well expressed in both the ovotestes and CNS. In a homology tree, Aplysia piwi clusters more closely with the piwi genes ...

Consistent with piRNA expression in the CNS, we were also able to clone the full-length cDNA for the 964 kDa Piwi protein from the CNS. The sequences of these clones are homologous to vertebrate piwi proteins and have conserved PAZ and Piwi domains. The Aplysia piwi protein is much more closely related to piwi proteins, by homology, than to argonaute proteins of other species, and within the piwi family, more closely related to vertebrate than invertebrate piwi members (Fig. 2A), as is often the case with Aplysia proteins. We generated a polyclonal antibody for the Aplysia piwi protein that detects the induced recombinant protein as well as the protein in Aplysia neural extracts as a single band (Fig. 2B). To determine whether that the piwi protein stably interacts with piRNAs, we immunoprecipitated (IP) piwi from neural extracts (Fig. 2C) and extracted the RNAs from the piwi complex. When blots of the RNAs from the Piwi IP and Argonaute (Ago) IP were probed for a piRNA (aca-piR-1) and miRNA (aca-miR-22), the piRNA was detected only in piwi IP whereas the miRNA was detected only in Ago IP (Fig. 2D). We further find that RNA from neural extracts after piwi knockdown with 2′-O-methyl antisense oligoribonucleotides (piwi knockdown confirmed in Fig. 4A) are depleted in piRNAs when compared to control extracts, with no detectable change in the levels of other non-coding RNAs, such as miRNAs, or tRNA (Fig. 2E). These experiments demonstrate that there are indeed two distinct classes of small RNAs in Aplysia CNS, miRNAs and piRNAs, each of which associates with its respective Ago and Piwi protein.

Figure 4
A. Knockdown of piwi causes a robust up-regulation of CREB2, which is specific as there is no significant change in expression levels of C/EBP or CPEB (effects quantified as a mean of 4 independent trials ± S.D).

To better understand the sub-cellular localization of piwi and piRNAs in Aplysia neurons, we first separated neural protein and RNA extracts into nuclear and cytoplasmic fractions, and probed for the piwi protein on western blots and piRNAs on northern blots. Effective fractionation was confirmed by the presence of GAPDH only in cytoplasmic compartments, and Histone H3 in nuclear compartments. We detected the Piwi protein primarily in the nuclear compartment (Fig. 2G). Consistent with this finding, over-expression of GFP-tagged piwi in Aplysia sensory neurons shows a predominant nuclear localization of the piwi protein (Fig. 2F). A northern blot comparing small RNA content in the nuclear and cytoplasmic fractions with total unfractioned RNA also revealed that the piRNAs were primarily nuclear, whereas the miRNAs were primarily cytoplasmic (Fig. 2H). Taken together, both piwi and piRNAs in Aplysia neurons have predominant nuclear localization suggesting a nuclear function for the piwi/piRNA complex.

Piwi/piRNA complexes enhance memory-related synaptic plasticity by regulating the transcriptional repressor, CREB2

To determine whether piRNAs have a regulatory role in memory-related synaptic plasticity, we screened some of the abundant neuronal piRNAs for changes in expression levels upon exposure to serotonin (5HT), a neuromodulator important for learning and memory. A subset of the selected piRNAs was significantly up-regulated (Fig. 3A). aca-piR-4 and aca-piR-15 are examples of piRNAs that were robustly induced by 5HT. The former was transiently induced while the latter had a more delayed but enduring activation. The increase in piRNA expression in response to 5HT was particularly interesting in comparison with the activity of several Aplysia miRNAs, which by contrast, were rapidly down-regulated in neurons in response to neuromodulators and to neuronal activity (Rajasethupathy et al. 2009). These observations suggest that the two classes of small RNAs in the Aplysia CNS could exercise coordinated bi-directional activity of their targets during memory-related synaptic plasticity.

Figure 3
A. Aplysia CNS were treated either with vehicle (−) or with 5 × 5HT and RNA was extracted 1 h and 4 h later and northern blotted. tRNA bands are shown to control for equal loading of samples. Changes in piRNA levels are quantified in ( ...

To better understand the functional relevance of these 5HT-induced piRNAs, we explored their role in memory-related synaptic plasticity in cultured neurons in response to 5HT. The co-cultures used in these experiments consisted of two sensory neurons that each synapse on a single target motor neuron. We first depleted piwi (and consequently its associated piRNA population) from sensory neurons that form synapses with motor neurons in culture, and assayed for changes in the strength of the sensory-motor synapse. We injected an antisense 2′-O-methyl oligoribonucleotide to piwi in one sensory neuron of the co-culture while the other sensory neuron was left unmodified as an internal control. In each case electrical activity was recorded in the motor neuron after exposure to 5HT to determine the change in baseline synaptic transmission and in memory-related long-term facilitation (LTF) at these synapses. We found that knockdown of piwi significantly impaired long-term facilitation (LTF) as measured at 24 and 48 h after exposure to five pulses of 5HT (n=34), when compared with uninjected controls in the same co-culture (n=37; F(3,95)=13.63; p<0.001 repeated measures ANOVA; p<0.02 and p<0.04 at 24 and 48h, respectively, Newman-Keuls post hoc test; Fig. 3C). The observed differences between the two groups were not due to differences in the basal strength of the synaptic connections. We confirmed the efficacy of piwi knockdown by western blotting, as the antibody was not able to detect piwi by immunostain (Fig. 4A), and also confirmed that the piwi knockdown specifically prevented the accumulation of mature piRNAs (Fig. 2E). Control experiments with the injection of scrambled antisense 2′-O-methyl oligoribonucleotides did not show changes in LTF (n=23, scrambled AS vs n=9, 5×5-HT; p>0.6 at both 24 and 48h, Newman-Keuls post hoc test; Fig. 3D). We next determined whether over-expression of piwi had the opposite effect. Over-expression of Piwi-GFP (n=22) caused a significant enhancement of 5HT-dependent long-term synaptic facilitation with respect to untreated controls (n=40) as measured at 24 and 48 h (F(2,78)=44.04; p<0.001 repeated measures ANOVA; p<0.001 Newman-Keuls post hoc test at both 24 and 48h; Fig. 3E). Taken together, we conclude that 5HT induces the activity of piwi-associated piRNAs, which in turn act to enhance LTF.

To identify genes through which piwi might act to enhance 5HT-dependent long-term facilitation, we screened many plasticity-related genes for changes in expression levels after knockdown of piwi. Desheathed pleural ganglia were incubated in antisense 2′-O-methyl oligoribonucleotides conjugated with penetratin to inhibit piwi (confirmed by Western blot Fig. 4A), and total protein was extracted and Western blots prepared and probed with specific antibodies. We found that inhibition of piwi led to a reproducible two-fold up-regulation of the transcriptional repressor, and major inhibitory constraint on LTF, CREB2, when compared to neurons treated with scrambled control 2′O-methyl oligoribonucleotides. This effect was specific to CREB2, as piwi inhibition had no effect on several other plasticity-related genes such as C/EBP and CPEB (Fig. 4A). The observed increase in CREB2 protein levels was supported by an even greater increase in CREB2 mRNA levels (Fig. 4B).

In earlier studies, it was observed that knockdown of CREB2 in sensory neurons could prime the sensory-motor synapse such that a weak stimulus (1 pulse instead of 5 pulses of 5HT) was sufficient to cause LTF that lasts days (Bartsch et al. 1995). We therefore asked whether over-expression of piwi, through its repression of CREB2, could prime neurons in a similar way. Indeed, we found that cells over-expressing piwi (n=15) gave rise to LTF when exposed to just a single pulse of 5HT, whereas control cells (n=25) required 5 pulses of 5HT to elicit LTF that could last days (F(4,100)=5.05; p<0.001 repeated measures ANOVA; p<0.001 Newman-Keuls post hoc test at 24 h and 48 h; Fig. 4C). The facilitation produced by one pulse of 5HT was robust as it was observed in 13 out of 15 piwi over-expressing cells. Moreover, the facilitation seen at 24 h with one pulse was comparable in magnitude to that seen at 24 h in control cells treated with 5 pulses of 5HT. In earlier studies (Bartsch et al. 1995), while knockdown of CREB2 allowed for priming by 1 pulse of 5HT, there was no enhancement of LTF seen from 5 pulses of 5HT. Our observation, therefore, that piwi-overexpressing cells enhance LTF from 5 pulses of 5HT (Fig. 3E) suggests that piwi likely has other targets in addition to CREB2. In summary, the sensitivity of piwi-over-expressing cells to one pulse of 5HT suggests that it primes neuronal excitability through regulation of CREB2, while the enhancement of LTF caused by 5 pulses in piwi over-expressing cells suggests that piwi likely regulates other genes as well, in addition to CREB2.

CREB2 is methylated at its promoter in response to 5HT induced synaptic plasticity

To gain insight into the mechanism of CREB2 regulation by piwi, we asked whether 5HT acts on CREB2 at the level of transcription, as suggested by the nuclear localization of piwi and its effect on both CREB2 protein and RNA levels. Earlier studies followed CREB2 expression levels up to 3–4 h after exposure to 5HT, and in this time-frame, no change in CREB2 protein or RNA was noted (Bartsch et al. 1995). We therefore monitored the levels of CREB2 for days after the initial exposure to 5HT and noticed that CREB2 protein levels begin to drop at 12 h and continue to remain low for up to 48 h without returning to the initial base line level of expression (Fig. 4D). This long-lasting drop in CREB2 levels is consistent with, and may be responsible for, the observation that 5HT induces long-lasting elevation in the transcriptional activator, CREB1 (Liu et al. 2008). At the protein level, the reduction in CREB2 levels was modest, but the effect was more pronounced at the mRNA level (Fig. 4E). The long-lasting effect on both the CREB2 protein and RNA levels suggests that a stable 5HT-dependent repressive state is established. Since piwi and piRNAs have known roles in epigenetic regulation in the germline through DNA methylation, we asked whether CREB2 also is being regulated by piwi through methylation at its promoter.

The Aplysia DNA methyltransferase (DNMT) is well expressed in neurons (Moroz et al. 2006) and its enzymatically active domain is highly conserved among the vertebrate homologs of DNMTs. We therefore inhibited ApDNMT enzymatic activity chronically in neurons with the DNMT inhibitor RG108 and observed a strong increase in CREB2 levels. To determine if DNMT activity on CREB2 was dependent on 5HT, we applied RG108 to neurons in the presence of 5HT and found that 12 h later, the 5HT-dependent long-lasting down-regulation of CREB2 was abolished (Fig. 4F). These effects of RG108 appear to be specific to CREB2 since there was no significant up-regulation of CREB1 levels. In fact, a modest down-regulation was apparent. To determine whether the effects of DNMT inhibition on CREB2 levels were functionally important during memory-related plasticity, we again performed electrophysiological experiments on sensory-motor co-cultures, in the absence and presence of RG108. Remarkably, bath application of the inhibitor RG108 (n=38) almost fully abolished 5HT-dependent long-term facilitation with respect to controls (n=37), as measured at both 24 and 48 h (F(3,100)=12.86; p<0.001 repeated measures ANOVA; p<0.03 and p<0.02 at 24 and 48h, respectively, Newman-Keuls post hoc test; Fig. 4G). The effect of RG108 was entirely dependent on 5HT, as the application of RG108 alone, in the absence of 5HT had no effect on the baseline activity of the cells (Fig. 4G).

To determine if DNMT acted indirectly on CREB2 or whether it directly methylated the promoter of CREB2, we examined its promoter region for possible CpG islands. We found two predicted CpG islands, one that is proximal to the translational start site (~200 bp upstream of the first ATG) and which encompasses a CRE-binding element and TATA-binding site, and the other that is distal (~700 bp upstream of the first ATG) (Fig. 5A). We also noticed that the promoter of ATF4, the human homolog of CREB2, contains a conserved CpG island ( To test whether either of the predicted CpG islands were functional, we extracted genomic DNA and treated it with bisulfite. This procedure allows recognition of methylated bases in DNA (Callinan & Feinberg 2006), because bisulfite converts all genomic cytosine residues to uridine excepting the methylated cytosines, which are inert to bisulfite treatment. By scoring the C to T conversion rates of the CpG sites in genomic DNA after bisulfite treatment, one can determine the fraction of DNA at every CpG site that exists in the methylated versus un-methylated state. We first asked whether methylation specific primers (MSP -designed to detect only the methylated copies of genomic CREB2) have a differential ability to amplify genomic DNA from cells that either have, or have not, been treated with 5HT. We found that exposure to 5HT dramatically increases the methylated fraction of the proximal CpG island but not the distal CpG island (Fig. 5A). We next designed both USPs (un-methylated specific primers, designed to detect only the unmethylated copies of genomic DNA) and MSPs for the promoter regions of CREB2 and CREB1 to compare the fractional representation of the methylated and unmethylated states of the CpG islands at baseline and after exposure to 5HT. We found that in the basal state the CREB2 promoter exists in both methylated and un-methylated forms, but 12 h after exposure to 5HT, the promoter is almost entirely in the methylated form, and in the presence of DNMT inhibitors, the promoter is almost entirely in the un-methylated form (Fig. 5B). This pattern of methylation of the CREB2 promoter is in direct contrast to the CREB1 promoter, which exists almost entirely in the un-methylated form at baseline, remains un-methylated after exposure to 5HT, and again remains un-methylated in the presence of DNMT inhibitors (Fig. 5B).

Figure 5
A. The genomic locus for Aplysia CREB2 is shown, and areas in green are predicted CpG islands. Methylation specific primers (MSP) designed to detect methylation at the distal CpG island shows no change in methylation patterns within 12 h after exposure ...

To more quantitatively measure the methylated and un-methylated fraction of the CREB2 promoter, we designed primers that lie outside the proximal CpG island and amplified the region in between by pyro-sequencing, to score the C to T conversions at every CpG site. Since the C to T conversion rate at each site reflects the un-methylated fraction, to display % methylation we plot 1 – (C to T conversion rate) at each CpG site. We found that, at baseline, almost every CpG site in the CREB2 promoter exists in approximately 50% methylated form, which is striking particularly when compared with the promoters of CREB1 and PKA-R, which display little to no methyation (Fig. 5C, D, E). This finding suggests that the CREB2 promoter is dynamically regulated by methylation and that its methylation state at baseline may reflect experience. After exposure to 5HT, every CpG site within the CpG island of CREB2 has increased methylation, with those at the beginning and end of the CpG island showing the most significant increase (Fig. 5C). Extraction and bisulfite treatment of genomic DNA after exposure to DNMT inhibitors prevents, as expected, the 5HT-induced increase in methylation and drops methylation levels to below baseline (Fig. 5C). Taken together, these data reveal that 5HT causes direct methylation of the proximal CpG islands in the CREB2 promoter, and that this methylation leads to a long-term down-regulation of CREB2 RNA and protein levels, which may be responsible for the resulting persistence of memory-related synaptic plasticity.

Piwi/piRNA complexes control the methylation state of the CREB2 promoter

Given that that piwi is regulating CREB2 at the transcriptional level (Fig. 4A, B), we asked whether piwi was required for the observed serotonin-dependent methylation of CREB2 in neurons. We inhibited piwi in sensory neurons and extracted the genomic DNA after exposure to 5HT. Following bisulfite treatment, we scored the % methylation by pyrosequenceing and found that inhibition of piwi completely abolished the serotonin-dependent increase in methylation at the promoter (Fig. 6A). The reversal in methylation patterns was most significant at the beginning and ends of the CpG island consistent with the observation that the same sequence areas were most sensitive to serotonin (Fig. 6A). To determine which piRNA mediates this effect, we searched the CREB2 locus for potential piRNA bindings sites and identified four well-expressed candidate piRNAs that had good complimentary to the promoter, 5′UTR, and initial coding segment of the CREB2 mRNA (Fig. 6B). Through a series of knockdown experiments using 2′-O-methyl oligoribonucleotides specific to each of the four piRNAs, we observed that one piRNA, aca-piR-F had the strongest effect on CREB2 expression. Knockdown of aca-piR-F, but not aca-piR-A, C, or D, increased the baseline levels of CREB2, both at the protein and RNA level, demonstrating that aca-piR-F is a transcriptional regulator of CREB2 (Fig. 6C). The effects of piR-F are specific since use of 4-nucleotide mismatch oligos antisense to piR-F had no effect on CREB2 levels (Fig. 6C). If aca-piR-F were indeed mediating the observed 5HT-dependent methylation effects of CREB2, then aca-piR-F should be regulated by 5HT on a similar time course. We followed aca-piR-F levels with exposure to 5HT as a function of time, and noticed a slightly delayed, but more enduring up-regulation of aca-piR-F that peaked at 3–4 hours before dropping back to baseline at 12 hours (Fig. 6D). This time course is consistent with the observed drop in CREB2 RNA levels, which begin between at 6 hours after exposure to 5HT. Since the putative binding site for aca-piR-F lies near the translational start site of CREB2, we propose a model in which the piwi/piRNA complex, through aca-piR-F, binds the nascent CREB2 transcript, thereby bringing it within close proximity for regulation of the CREB2 promoter during 5HT dependent long-term memory (Fig. 6E). The observed stable silencing of CREB2 by the piwi/piRNA complex (Fig 6), when placed in the context of Figs. 4A and B where transient knockdown of piwi reverses CREB2 silencing, is suggestive of active de-methylation at the CREB2 promoter. While this is consistent with a wider literature that demonstrates active and ongoing de-methylation of promoters in adult neurons, we in this study have not explicitly demonstrated active de-methylation of the CREB2 promoter. Transcriptional control of gene-expression through complementary base pairing of a small RNA with a nascent mRNA transcript has been previously discovered, first in the exciting work from S. pombe and more recently from a study in c. elegans (Verdel et al. 2004, Guang et al. 2010).

Figure 6
A. Real-time pyrosequencing of the CREB2 promoter region shows increased methylation in response to 5HT (maroon) that is fully reversed when 5HT is applied in the presence of a piwi inhibitor (blue). The results were quantified as a mean of 4 independent ...


The discovery that piRNAs exist outside the germline, in several major organs of Aplysia, but significantly in the nervous system, suggests much broader roles for piRNAs than has been previously appreciated. In addition to their presence and in certain cases enrichment in neurons, Aplysia piRNAs are unique from those previously described in that they derive from hotspots in the genome where they are abundantly expressed and, notably, several piRNAs are regulated by neuromodulators important for learning-related synaptic plasticity, suggesting functions in memory storage.

An understanding of the role of piRNAs in the epigenetic regulation of long-term memory is significant for several reasons. First, the role of epigenetic modifications in differentiated cells, especially in adult neurons, has been controversial. It is commonly thought that changes in gene expression during development are permanent, but that they are not permanent in adult neurons where the plastic nature of synaptic connections, by definition, requires bi-direction and reversible changes in gene expression. In recent years, the identification of DNA demethylase activity in adult neurons (Baretto et al. 2007, Rai et al. 2008, Ma et al. 2009) brought forth the possibility that epigenetic changes in the adult brain may not necessarily be permanent, but may simply be more long-lasting, and more permanent than the other known modifications so far described. Subsequent studies have identified individual gene loci that are methylated in response to neurotransmitter activity, though the time course of onset and persistence of methylation are unclear and require further study. Our study provides a piRNA-mediated mechanism for epigenetic regulation in neurons, and further, explores the electrophysiological properties of DNA Methyltransferase and of Piwi in synaptic plasticity.

The findings that the piwi/piRNA complex regulates the CREB2 promoter by DNA methylation in an activity-dependent manner provides an attractive explanation for how neurons translate transient stimuli into stable internal representations, and is consistent with several earlier studies that show a role for epigenetic regulation in memory (Miller and Sweatt 2007, Miller et al. 2010, Feng et al. 2010). Our data on the variability in baseline CREB2 methylation levels suggests further that each neuron may have a different basal level of CREB2 expression, which reflects its experience and immediate history. This would be consistent with earlier observations showing that variations in baseline levels of CREB1 across populations of neurons in the amygdala determine the sequence in which these neurons are recruited for memory and for recall (Han et al. 2007). Since CREB2 is antagonistic to CREB1, long-lasting changes in CREB2 levels could set up this CREB1 distribution in neuronal cells, based on experience, which in turn can dictate which neurons are already holding a memory trace and which neurons are readily drawn into new memory traces (Han et al. 2007, Won et al. 2008). The likelihood that CREB2 set-points within a neuron can dictate its functional capacity for both memory and flexibility is further supported by a ubiquitin ligase over-expressing mouse model that is phenotypically much smarter than wild-type mice (Pavlopoulos et al. 2011), and these mice show significantly reduced baseline CREB2 expression when compared with wildtype mice (Unpublished Observations, Pavlopoulos et a. 2011).

The discussion above, however, does not address the question as to how piwi mediated transcriptional, and therefore cell-wide changes in neurons (intrinsic plasticity), effectively mediate synapse-specific events (synaptic plasticity)? As studies previously have emphasized, it is likely that both forms of plasticity co-exist such that one can fine-tune the other, but it is also possible, that in certain contexts the two exist entirely independently. While synaptic plasticity affords orders of magnitude more computational power and is therefore ideal for storage of explicit memories requiring attention to detail, intrinsic plasticity such as those driven by piRNA mediated epigenetics has the advantage of priming memories and allowing for robust generalized learning where the same association rules are applicable to experiential learning in various contexts. Since human life is characterized by a great deal of habit-formation and repetition based associative learning, the use of intrinsic plasticity alone in some parts of the human brain may turn out to be an efficient method for this type of memory storage.

Future work on the role of small RNAs in learning and memory should provide further insight into the varying roles of miRNAs versus piRNAs. While notable exceptions exist (Wayman et al. 2008, Fiore et al. 2009), we and others have previously found a rapid turn-over of several neuronal miRNAs in response to neuromodulators and neuronal activity (Rajasethupathy et al. 2009, Krol et al. 2010), which contrasts to the slow but more enduring up-regulation of the few neuronal piRNAs observed in this study. In addition, while aca-miR-124 (Rajasethupathy et al. 2009) and aca-miR-22 (In Preparation) constrain serotonin dependent long-term facilitation, piwi-dependent piR-F enhances it. We currently have very few cases from which to draw generalizable conclusions, but future large-scale studies of small RNA function in neurons may highlight the possible existence of two distinct classes of small RNAs that are bi-directionally regulated by neuromodulators, and that act on a functionally segregated population of targets, to effect either facilitation or constraint on memory related synaptic plasticity. Further studies would also benefit from genome wide analysis of piRNA/piwi occupied promoter regions during serotonin-mediated synaptic plasticity to obtain a more complete picture of the epigenetic landscape during memory. One attractive possibility is that piRNAs are directed only toward inhibitors of plasticity, and that with each repeated training trial (either behavioral training or pulses of serotonin) the promoters of more inhibitory genes are silenced, such that eventually the cell is maximally primed and excitable, allowing for the strongest associative memories. Finally, future experiments with chromatin IP of RNA Polymerase and/or of piwi at the CREB2 locus would greatly increase our understanding of the mechanisms governing piwi dependent methylation. It would substantiate the idea that piwi is recruited to CREB2 in an activity dependent manner, and further that methylation of the promoter is directly responsible for the observed reduction in transcription of the gene. It is also possible that other small RNAs play a role in epigenetic regulation during plasticity. Irrespective of their biogenesis properties, small RNAs confer versatile sequence specificity to mechanisms of gene regulation, and therefore, any small RNA that evolves functionality for its guide protein to recruit methylation elements to the target promoter could prove equally effective. It is possible therefore, that one of the many rapidly multiplying classes of nuclear small RNAs take over the same task in other species.

In summary, we find that piwi/piRNAs control the activity dependent epigenetic regulation of the transcription factor CREB2, which may prove to be an important and general mechanism of small RNA mediated long-lasting regulation of gene-expression in neurons that contributes to long-term memory storage. This initial study compels the further exploration of a genome-wide approach toward understanding the extent of small-RNA-mediated epigenetic regulation in neurons during learning and memory.

Experimental Procedures

Small RNA Cloning, Sequencing, and annotation

Tissue Preparation and RNA isolation was as described in Rajasethupathy et al. 2009. Starting amount of total RNA was 5 μg per library. Small RNA cloning was performed as described in Hafner et al. 2008, with the exception of using barcoded libraries and the Illumina platform for sequencing. A total of about 15,000 reads were obtained for each solexa library. miRNAs were annotated as described in Rajasethupathy et al. 2009. piRNA candidates were chosen based on their length (between 26 and 33 nt), a 5′ terminal U, and their property of clustering together in a genomic contig (Genomic contigs that contained at least 10,000 total piRNA reads, with at least a 60% U start bias, and no more than 1000 bp interval between individual piRNAs were defined as piRNA clusters.) 372 such piRNA clusters were generated. All small RNA sequences were mapped to these clusters and any clone that mapped perfectly or within one mismatch to a piRNA cluster was annotated as a piRNA.

Pharmacological treatment, Northern and Western blot and Q-PCR analysis

Inhibition of Aplysia DNA Methyltransferase (DNMT) were performed by incubating ganglia or cultured cells in RG108 (Sigma) at a final concentration of 200 μM. Inhibition of piwi and piRNAs were carried out using penetratin conjugated 2′-O-methyl antisense oligoribonucleotides as described in Rajasethupathy et al. 2009. 150 μl of 200 nM penetratin conjugated oligonucleotides were applied to desheathed pleural ganglia in Eppendorf tubes, for a minimum of 8 hrs (for piRNA knockdown) and a minimum of 1 d (for piwi knockdown) before washout, after which RNA or protein was harvested. Northern blot (Landgraf et al. 2007), Western blot and QPCR (Rajasethupathy et al. 2009), were performed as previously described. Probe sequences and antibody information are provided in supplementary information.

Periodate Treatment & Beta Elimination

Total RNA from each Aplysia CNS was extracted, ethanol precipitated, and re-dissolved in 13.5 μl of H20. 4 μl of Borate buffer 1 (25 ml 0.8 M boric acid + 8.75 ml 0.2 M sodiumborate in 100 ml H2O, pH8.6) and 2.5 μl of 0.2 M sodiumperiodate were added to the RNA solution and incubated in the dark at 24 °C for 10 min. 5 μl of 50% (v/v) glycerol were added to quench the reaction and the resulting solution was concentrated in a speedvac to 5 μl. 50 μl of Borate buffer 2 (6.25 ml 0.8 M boric acid + 16.875 ml 0.2 M sodiumborate in 400 ml H2O, pH9.5 with NaOH) were added and incubated for 90 min at 45 °C. RNA was then precipitated in 3 volumes of ethanol, washed once in 70 % ethanol, and run on a 20 % acrylamide gel along with untreated total RNA from Aplysia CNS, as well as with synthetic piRNA that was either 2′-O-methyl modified at its 3′end or unmodified, and each treated or untreated, as positive and negative controls, loaded at 30 fmol each with carrier tRNA. Blots were probed for piR-1 and miR-22.

Nuclear/Cytoplasmic Fractionation

Aplysia CNS and heart tissue were subject to nuclear and cytoplasmic extraction using the Pierce NE-PER (Thermo Scientific 78833) Kit. All procedures were performed as described in the instruction manual, though we quadrupled all starting volumes so that we may split the extraction in half, twice, to end up with the following 4 samples for any given tissue (1. Total RNA, 2. Fractioned RNA, 3. Total protein, 4. Fractioned protein). And for step 6, we re-dissolved the nuclear pellet in a final volume that would match the cytoplasmic extract, rather than the amount stated in the manual, to allow comparison across fractions and with the total input. The effectiveness of fractionation was confirmed by western blotting the protein fractions and probing for nuclear (Histone H3) and cytoplasmic (GAPDH) specific markers. This kit was already optimized for extraction of Aplysia CNS and Heart tissue, but requires further optimization of reagent volumes and incubation times to successfully separate muscle or ovotestes fractions.

DNA Methylation Assays

DNA purification (DNA mini kit; Qiagen) was performed on Aplysia sensory neuron clusters. Purified DNA was then processed for bisulfite modification (Epitect Bisulfite Kit; Qiagen). Quantitative PCR was used to determine the DNA methylation status of the CREB2 and CREB1 genes. Methylation-specific PCR primers were designed using the Methprimer software (available at For quantitative methylation analysis through real-time PCR and pyrosequencing of the CREB2, CREB1, and PKA-R promoters, we used the Sequenom massArray facility at Cornell ( and all primers were designed using the epidesigner software (available at

All primer sequences are provided in supplementary information.

Cell Culture, Injections, Treatments and Electrophysiology

Co-cultures of synaptically-paired sensory and motor neurons were prepared as previously reported (Schacher and Proshansky 1983; Montarolo et al. 1986). For intracellular injections, either a piwi antisense oligoribonucleotide (5 uM) was used for knockdown of piwi, or a piwi-GFP expression vector (1 μg/ul) was used for over-expression of piwi. To generate the pNEX-apPiwi-eGFP, the apPiwi ORF was PCR-amplified from cDNA and subcloned into a pNEX3-eGFP vector (Kaang, 1996) modified by the insertion of a Gateway Destination cassette (Invitrogen) within the polylinker (pNEX3-eGFP-DEST vector).

Supplementary Material






We would like to thank Tom Maniatis and Praveen Sethupathy for comments on the manuscript and for helpful discussions throughout. We also thank Ferdinando Fiumara, Doron Betel, Pavankumar Puvulla and Pierre Trifilieff for experimental advice and discussions, and Frank Grasso for gifting invertebrate CNS samples. We thank Vivian Zhu and Edward Konstantinov for technical assistance with cell cultures. This work is supported by an HHMI grant P50 HG002806, NIH grant P01 GM073047 and NRSA training grant 5F30MH086267.


Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.


  • Aravin A, Gaidatzis D, Pfeffer S, Lagos-Quintana M, Landgraf P, Iovino N, Morris P, Brownstein MJ, Kuramochi-Miyagawa S, Nakano T, Chien M, Russo JJ, Ju J, Sheridan R, Sander C, Zavolan M, Tuschl T. A novel class of small RNAs bind to MILI protein in mouse testes. Nature. 2006;442:203–7. [PubMed]
  • Aravin AA, Sachidanandam R, Girard A, Fejes-Toth K, Hannon GJ. Developmentally regulated piRNA clusters implicate MILI in transposon control. Science. 2007 May 4;316(5825):744–7. [PubMed]
  • Bailey CH, Kandel ER, Si K. The persistence of long-term memory: a molecular approach to self-sustaining changes in learning-induced synaptic growth. Neuron. 2004 Sep 30;44(1):49–57. [PubMed]
  • Barreto G, Schafer A, Marhold J, Stach D, Swaminathan SK, Handa V, Doderlein G, Maltry N, Wu W, Lyko F, Niehrs C. Gadd45a promotes epigenetic gene activation by repair-mediated DNA demethylation. Nature. 2007;445:671–675. [PubMed]
  • Bartsch D, Ghirardi M, Skehel PA, Karl KA, Herder SP, Chen M, Bailey CH, Kandel ER. Aplysia CREB2 represses long-term facilitation: relief of repression converts transient facilitation into long-term functional and structural change. Cell. 1995 Dec 15;83(6):979–92. [PubMed]
  • Betel D, Sheridan R, Marks D, Sander C. Computational analysis of mouse piRNA sequence and biogenesis. PLoS Comp Biol. 2007;3 [PMC free article] [PubMed]
  • Brennecke J, Malone CD, Aravin AA, Sachidanandam R, Stark A, Hannon GJ. An epigenetic role for maternally inherited piRNAs in transposon silencing. Science. 2008 Nov 28;322(5906):1387–92. [PMC free article] [PubMed]
  • Callinan PA, Feinberg AP. The emerging science of epigenomics. Hum Mol Genet. 2006 Apr 15;15(Spec No 1):R95–101. [PubMed]
  • Crick, Memory and molecular turnover. Nature. 1984;312:101. [PubMed]
  • Davis, Squire LR. Protein synthesis and memory: A review. Psychological Bulletin. 1984;96:518–559. [PubMed]
  • Feng J, Zhou Y, Campbell SL, Le T, Li E, Sweatt JD, Silva AJ, Fan G. Dnmt1 and Dnmt3a maintain DNA methylation and regulate synaptic function in adult forebrain neurons. Nat Neurosci. 2010 Apr;13(4):423–30. [PMC free article] [PubMed]
  • Fiore R, Khudayberdiev S, Christensen M, Siegel G, Flavell SW, et al. Mef2-mediated transcription of the miR379-410 cluster regulates activity-dependent dendritogenesis by fine-tuning Pumilio2 protein levels. Embo J. 2009;28:697–710. [PubMed]
  • Girard A, Sachidanandam R, Hannon GJ, Carmell MA. A germline-specific class of small RNAs binds mammalian Piwi proteins. Nature. 2006 Jul 13;442(7099):199–202. [PubMed]
  • Grivna ST, Beyret E, Wang Z, Lin H. A novel class of small RNAs in mouse spermatogenic cells. Genes & Dev. 2006;20:1709–1714. [PubMed]
  • Guang S, Bochner AF, Burkhart KB, Burton N, Pavelec DM, Kennedy S. Small regulatory RNAs inhibit RNA polymerase II during the elongation phase of transcription. Nature. 2010 Jun 24;465(7301):1097–101. [PMC free article] [PubMed]
  • Han JH, Kushner SA, Yiu AP, Cole CJ, Matynia A, Brown RA, Neve RL, Guzowski JF, Silva AJ, Josselyn SA. Neuronal competition and selection during memory formation. Science. 2007;316:457–60. [PubMed]
  • Hayer A, Bhalla US. Molecular switches at the synapse emerge from receptor and kinase traffic. PLoS Comput Biol. 2005 Jul;1(2):137–54. [PMC free article] [PubMed]
  • Hobert O. Gene regulation by transcription factors and microRNAs. Science. 2008 Mar 28;319(5871):1785–6. [PubMed]
  • Kirino Y, Mourelatos Z. Mouse Piwi-interacting RNAs are 2′-O-methylated at their 3′ termini. Nat Struct Mol Biol. 2007 Apr;14(4):347–8. [PubMed]
  • Krol J, Loedige I, Filipowicz W. The widespread regulation of microRNA biogenesis, function and decay. Nat Rev Genet. 2010 Sep;11(9):597–610. [PubMed]
  • Krol J, Busskamp V, Markiewicz I, Stadler MB, Ribi S, Richter J, Duebel J, Bicker S, Fehling HJ, Schübeler D, Oertner TG, Schratt G, Bibel M, Roska B, Filipowicz W. Characterizing light-regulated retinal microRNAs reveals rapid turnover as a common property of neuronal microRNAs. Cell. 2010 May 14;141(4):618–31. [PubMed]
  • Kuramochi-Miyagawa S, Watanabe T, Gotoh K, Totoki Y, Toyoda A, Ikawa M, Asada N, Kojima K, Yamaguchi Y, Ijiri TW, Hata K, Li E, Matsuda Y, Kimura T, Okabe M, Sakaki Y, Sasaki H, Nakano T. DNA methylation of retrotransposon genes is regulated by Piwi family members MILI and MIWI2 in murine fetal testes. Genes Dev. 2008 Apr 1;22(7):908–17. [PubMed]
  • Lee EJ, Banerjee S, Zhou H, Jammalamadaka A, Arcila M, Manjunath BS, Kosik KS. Identification of piRNAs in the central nervous system. RNA. 2011 Jun;17(6):1090–9. [PubMed]
  • Lisman JE. A mechanism for memory storage insensitive to molecular turnover: a bistable autophosphorylating kinase. Proc Natl Acad Sci U S A. 1985 May;82(9):3055–7. [PubMed]
  • Liu RY, Fioravante D, Shah S, Byrne JH. cAMP response element-binding protein 1 feedback loop is necessary for consolidation of long-term synaptic facilitation in Aplysia. J Neurosci. 2008 Feb 20;28(8):1970–6. [PubMed]
  • Ma DK, Jang M, Guo JU, Kitabatake Y, Chang M, Pow-anpongkul N, Flavell RA, Lu B, Ming G, Song H. Neuronal Activity–Induced Gadd45b Promotes Epigenetic DNA Demethylation and Adult Neurogenesis. Science. 2009;323( 5917):1074–1077. [PMC free article] [PubMed]
  • Miller CA, Gavin CF, White JA, Parrish RR, Honasoge A, Yancey CR, et al. Cortical DNA methylation maintains remote memory. Nature Neuroscience. 2010;13:664–666. [PMC free article] [PubMed]
  • Miller CA, Sweatt JD. Covalent modification of DNA regulates memory formation. Neuron. 2007;53:857–869. [PubMed]
  • Moroz LL, Edwards JR, Puthanveettil SV, Kohn AB, Ha T, Heyland A, Knudsen B, Sahni A, Yu F, Liu L, Jezzini S, Lovell P, Iannucculli W, Chen M, Nguyen T, Sheng H, Shaw R, Kalachikov S, Panchin YV, Farmerie W, Russo JJ, Ju J, Kandel ER. Neuronal transcriptome of Aplysia: neuronal compartments and circuitry. Cell. 2006 Dec 29;127(7):1453–67. [PMC free article] [PubMed]
  • Pavlopoulos E, Trifilieff P, Chevaleyre V, Fioriti L, Zairis S, Pagano A, Malleret G, Kandel ER. Neuralized1 Activates CPEB3: A Novel Function of Ubiquitination in Synaptic Plasticity and Memory Storage. Cell. 2011 Dec 9;147(6):1369–83. [PMC free article] [PubMed]
  • Price JC, Guan S, Burlingame A, Prusiner SB, Ghaemmaghami S. Analysis of proteome dynamics in the mouse brain. Proc Natl Acad Sci U S A. 2010 Aug 10;107(32):14508–13. [PubMed]
  • Rai K, Huggins IJ, James SR, Karpf AR, Jones DA, Cairns BR. DNA demethylation in zebrafish involves the coupling of a deaminase, a glycosylase, and gadd45. Cell. 2008;135:1201–1212. [PMC free article] [PubMed]
  • Rajasethupathy P, Fiumara F, Sheridan R, Betel D, Puthanveettil SV, Russo JJ, Sander C, Tuschl T, Kandel E. Characterization of small RNAs in Aplysia reveals a role for miR-124 in constraining synaptic plasticity through CREB. Neuron. 2009 Sep 24;63(6):803–17. [PMC free article] [PubMed]
  • Rybak A, Fuchs H, Smirnova L, Brandt C, Pohl EE, Nitsch R, Wulczyn FG. A feedback loop comprising lin-28 and let-7 controls pre-let-7 maturation during neural stem-cell commitment. Nat Cell Biol. 2008;10:987–93. [PubMed]
  • Saito K, Siomi MC. Small RNA-mediated quiescence of transposable elements in animals. Dev Cell. 2010 Nov 16;19(5):687–97. [PubMed]
  • Serrano P, Friedman EL, Kenney J, Taubenfeld SM, Zimmerman JM, Hanna J, Alberini C, Kelley AE, Maren S, Rudy JW, Yin JC, Sacktor TC, Fenton AA. PKMzeta maintains spatial, instrumental, and classically conditioned long-term memories. Plos Biol. 2008;6:2698–2706. [PMC free article] [PubMed]
  • Si K, Choi YB, White-Grindley E, Majumdar A, Kandel ER. Aplysia CPEB can form prion-like multimers in sensory neurons that contribute to long-term facilitation. Cell. 2010 Feb 5;140(3):421–35. [PubMed]
  • Si K, Lindquist S, Kandel ER. A neuronal isoform of the Aplysia CPEB has prion-like properties. Cell. 2003 Dec 26;115(7):879–91. [PubMed]
  • Song H, Smolen P, Av-Ron E, Baxter DA, Byrne JH. Dynamics of a minimal model of interlocked positive and negative feedback loops of transcriptional regulation by cAMP-response element binding proteins. Biophys J. 2007 May 15;92(10):3407–24. [PubMed]
  • Verdel A, Jia S, Gerber S, Sugiyama T, Gygi S, Grewal SI, Moazed D. RNAi-mediated targeting of heterochromatin by the RITS complex. Science. 2004 Jan 30;303(5658):672–6. [PMC free article] [PubMed]
  • Wassenegger M. The role of the RNAi machinery in heterochromatin formation. Cell. 2005 Jul 15;122(1):13–6. [PubMed]
  • Watanabe T, Takeda A, Tsukiyama T, Mise K, Okuno T, Sasaki H, Minami N, Imai H. Identification and characterization of two novel classes of small RNAs in the mouse germline: Retrotransposon-derived siRNAs in oocytes and germline small RNAs in testes. Genes & Dev. 2006;20:1732–1743. [PubMed]
  • Wayman GA, Davare M, Ando H, Fortin D, Varlamova O, et al. An activity-regulated microRNA controls dendritic plasticity by down-regulating p250GAP. Proc Natl Acad Sci U S A. 2008;105:9093–9098. [PubMed]
  • Weaver IC, Cervoni N, Champagne FA, D’Alessio AC, Sharma S, Seckl JR, Dymov S, Szyf M, Meaney MJ. Epigenetic programming by maternal behavior. Nat Neurosci. 2004 Aug;7(8):847–54. [PubMed]
  • Won J, Silva AJ. Molecular and cellular mechanisms of memory allocation in neuronetworks. Neurobiol Learn Mem. 2008;89:285–92. [PMC free article] [PubMed]