|Home | About | Journals | Submit | Contact Us | Français|
Memory storage and memory-related synaptic plasticity relies on precise spatiotemporal regulation of gene expression. To explore the role of small regulatory RNAs in learning-related synaptic plasticity we carried out massive parallel sequencing to profile the small RNAs of Aplysia californica. We identified 170 distinct miRNAs, 13 of which were novel and specific to Aplysia. Nine miRNAs were brain-enriched, and several of these were rapidly down-regulated by transient exposure to serotonin, a modulatory neurotransmitter released during learning. Further characterization of the brain-enriched miRNAs revealed that miR-124, the most abundant and well-conserved brain-specific miRNA, was exclusively present pre-synaptically in a sensory-motor synapse where it constrains serotonin-induced synaptic facilitation through regulation of the transcriptional factor CREB. We therefore present direct evidence that a modulatory neurotransmitter important for learning can regulate the levels of small RNAs and present a novel role for miR-124 in long-term plasticity of synapses in the mature nervous system.
miRNAs are a class of conserved, 20 to 23 nucleotide (nt) non-coding RNAs that depend on the RNAi machinery for maturation and function, and are able to mediate cleavage or translational repression of their target mRNAs by preferentially binding to their 3′UTRs (Filipowicz et al. 2008, Bartel 2009). Discovery of the first miRNAs in C. elegans led to an understanding of their regulatory role in cell lineage specification (Lee et al. 1993, Wightman et al. 1993, Reinhart et al. 2000). The subsequent development of methods for the large-scale identification of miRNAs (Lagos-Quintana et al. 2001, Lau et al. 2001, Lee et al. 2001) and the resulting functional studies revealed that miRNAs control many other cellular functions including proliferation, metabolism, apoptosis, immunity and more recently, neuronal growth and plasticity.
To obtain a more complete inventory of small RNAs that may have a role in learning-related synaptic plasticity, we used a large-scale identification approach to mine and functionally screen the small RNAs of the marine snail Aplysia. The large, stereotyped, easily identifiable neurons of Aplysia make it a practical choice for studying the distribution and function of small RNAs at the resolution of single cells and single synapses. Critical components of the neural circuitry of the Aplysia gill-withdrawal reflex can be reconstituted in cell culture, allowing for a well-controlled study of axonal growth, synapse formation, stabilization, and synaptic plasticity. The evolutionary closeness of Aplysia to vertebrates and mammals also makes Aplysia generally attractive as a model system for addressing the function of small RNAs (Moroz et al. 2007).
In Aplysia, both short-term memory lasting minutes and long-term memory lasting days have been well characterized in the gill-withdrawal reflex in response to sensitization, a simple form of learned fear (reviewed in Kandel 2001). In an Aplysia sensory-motor culture system (Montarolo et al. 1986), delivery of one pulse of serotonin (5HT), a modulatory neurotransmitter released in the intact animal by sensitizing stimuli, elicits PKA-dependent short-term facilitation lasting minutes. By contrast, five spaced pulses of serotonin cause both PKA and MAPK to translocate to the nucleus (Martin et al. 1997b), thereby releasing inhibition by the repressor CREB2 and activating CREB-dependent transcription, leading to long-term synaptic facilitation and growth of new synaptic connections. Thus in sensitization, as in many other forms of learning, nuclear activation of CREB is an important component of a general switch that converts short-term into long-term plasticity in both vertebrates and invertebrates (Dash et al. 1990, Barco et al. 2002). In addition, studies on both the gill-withdrawal reflex and the mammalian hippocampus has delineated the importance of local protein synthesis at the synapse in sustaining synapse activity independent from the distant cell body (reviewed by Sutton & Schuman 2006, Martin & Zukin 2006). Indeed, communication between the nucleus and the synapse, via the shuttling of mRNA and proteins by kinesin motors, serves as still another critical regulatory point in the induction of long-term facilitation (Puthanveetil et al 2008).
Since the spatio-temporal regulation of learning-related synaptic plasticity is extensive and complex, miRNAs appear well suited to serve as negative regulators. The ability of miRNAs to selectively (Farh et al. 2005) and reversibly (Bhattacharyya et al. 2006) silence mRNAs allows for precise control, possibly in a combinatorial fashion, of relevant subsets of the mRNA population recruited during plasticity. Moreover, their ability to form autoregulatory loops (Rybak et al. 2008, Johnston & Hobert 2003) suggests their potential involvement in either homeostatic or switch-like events during various phases of synaptic plasticity, an inherently multi-stable phenomenon. Several studies have demonstrated the involvement of brain-specific miRNAs in synapse formation and of miRNA ribonucleoprotein complexes (miRNPs) in controlling local protein synthesis associated with stable memory (reviewed in Schratt 2009). These findings have encouraged us to explore systematically the miRNA population of the Aplysia central nervous system to understand their functions during learning-related synaptic plasticity.
We identified small RNAs in neuronal and non-neuronal cell populations in Aplysia, which allowed the identification of brain-enriched miRNAs. We have functionally characterized the most highly abundant, well-conserved, brain-specific miRNA, Aplysia miR-124. This miRNA is specific to the pre-synaptic sensory neuron where it is rapidly down-regulated by serotonin. In the absence of serotonin regulation, miR-124 provides an inhibitory constraint on synaptic plasticity and long-term facilitation through the regulation of CREB, the transcriptional switch critical for converting short- to long-term facilitation.
We prepared small RNA cDNA libraries from isolated central nervous system (CNS), and from the whole animal with CNS removed. Within the CNS, we also generated small RNA libraries from dissected abdominal and pleural ganglia. The libraries from the whole animals and CNS were sequenced using 454 sequencing technology yielding a total of about 250,000 sequence-reads for each library. The abdominal and pleural libraries were sequenced by traditional Sanger sequencing until approximately 2000 reads were collected for each library.
Because we lacked an assembled genome, we first built an Aplysia-specific annotation database to distinguish miRNAs from turnover of abundant and conserved non-coding RNAs such as rRNAs, tRNAs, or snRNAs. The total content of rRNAs, tRNAs and snRNAs taken together varied between 5 and 25% per library (Supplemental Table 1). To be considered a miRNA the residual sequences needed to satisfy the following three criteria: (1) Precise 5′ end processing: Length variants of members of a sequence family preferentially aligned to the 5′ end; (2) Fold-back precursor structure: A fold-back structure had to be identified comprising a genomic fragment retrieved from unassembled trace archives; (3) Cloning of the miR* sequence: Since double-stranded miR/miR* processing intermediates are assembled in an asymmetric fashion, capture of miR* that generates short 3′ overhangs when paired to the mature miRNA is further evidence for prototypical miRNA biogenesis.
The Aplysia genome trace sequence archives do not yet cover the full Aplysia genome, we therefore considered certain clone sequences that did not map to trace sequences as miRNAs, if we could map them to miRNA precursors annotated in other species. We identified 170 distinct miRNAs in Aplysia, of which 157 were orthologs to known miRNAs in other species and 13 were specific to Aplysia. The new discoveries are catalogued in Table 1. All miRNAs are catalogued in Supplemental Table 2. 60 sequences that were abundantly cloned and that demonstrated good 5′ processing, but that were neither conserved nor mapped, were designated miRNA candidates (Supplemental Table 3). The evidence for each miRNA is summarized in Supplemental Table 4. The overall abundance of miRNAs in the small RNA libraries ranged from 50 to 80%, consistent with the miRNA content in small RNA libraries prepared from other species (Aravin et al. 2003, Landgraf et al. 2007).
A phylogenetic analysis of the Aplysia transcriptome revealed that Aplysia is closer in evolutionary distance to the vertebrates than are C. elegans and D. melanogaster (Moroz et al. 2006). We similarly find that Aplysia miRNAs more closely resemble vertebrate miRNAs both in sequence similarity of individual genes and in the abundance of shared miRNA genes. We grouped the 170 distinct Aplysia miRNAs into 103 miRNA gene families based primarily on seed sequence similarity (Supplemental Table 5), of which 41 families are conserved specifically in vertebrates, whereas only 13 map specifically to invertebrates (Figure 1A). When we fit the observed miRNA gene gains and losses onto various phylogenetic trees, we find that our data best fits a model where Aplysia is a very ancient ancestor to the invertebrates, lies outside the D. melanogaster/C. elegans clade, and more directly straddles the invertebrate and vertebrate lineages (Figure 1B). The orthology relationships used to construct the phylogeny are given in Supplemental Table 6. A salient feature in support of this model is the presence of 46 miRNAs that are preserved from vertebrates to Aplysia, but subsequently lost in D. melanogaster and C. elegans (Figure 1B). Taken together, these findings illustrate that Aplysia miRNAs are ancient and well conserved, with relatively few losses or gains of genes, which makes it a distinctive model organism among invertebrates that shares important genomic similarities with vertebrates and mammals.
We observe one striking aspect of evolutionary history regarding the miR-9/79 gene family. The invertebrate-specific miR-79, and the vertebrate- and Drosophila-specific miR-9 are expressed in equal proportions in Aplysia, and are in fact star sequences of each other, found on opposite strands of the same precursor hairpin (Figure 1C). It is likely, then, that Aplysia mir-9/79 is a single gene that displays symmetric maturation patterns for both strands, whereas in other species, the gene has duplicated to give rise to multiple gene copies with asymmetric strand preference producing either miR-79 in other invertebrates or miR-9 in vertebrates.
Deep sequencing revealed the expression of over 100 distinct miRNA genes expressed in the Aplysia CNS. The miRNAs comprising the top 95% of clones are shown in order of their abundance in Figure 2 with their enrichment in the CNS versus the rest of the body. Nine miRNAs are either brain-specific or brain-enriched by cloning (Figure 2, and confirmed with Northern blotting in Figure 3A), three of which are miRNAs unique to Aplysia. In general, there was good overlap of the miRNAs of Aplyisa CNS with the miRNAs of the human and rodent brain but notable exceptions include the high abundance and brain enrichment of miR-22c (Figure 3A), miR-184 (Figure 3A), miR-34b, and miR-190, where studies in other species have not found CNS- enrichment for these miRNAs (Chen et al. 2005, Ruby et al. 2007, Landgraf et al. 2007). The multicopy cistrons of miR-1/133a and miR-206/133b, which are muscle-specific in vertebrates and D. melanogaster, were abundantly expressed in Aplysia CNS. Finally, the low expression of miR-9 and the complete absence of miR-128 in Aplysia CNS is noteworthy because both are highly abundant, and brain-specific, in vertebrates.
To learn which miRNAs might function in a compartment-specific manner, we developed a protocol (described in methods) for in situ hybridization of miRNAs in Aplysia using synthetic DNA probes. We dissected a functional circuit containing a sensory and motor neuron from Aplysia ganglia, placed them in co-culture, and examined the in situ hybridization patterns for various miRNAs (Table 2). We found that miR-124 stained much more intensely in the sensory neuron compared with the motor neuron (Figure 3B), and a 4 nt mismatch containing control probe showed no signal. We also consistently observed both a perinuclear density for mir-124, as well as punctuate staining in the processes (Figure 3C). Further in situ hybridization studies of the more abundant miRNAs in CNS revealed several other miRNAs (such as miRs 22c, 125c, let-7a, 100, and 8b) that were specifically expressed in the sensory neuron compared to the motor neuron (Figure 3B), and some that were enriched in either the cell body alone (miR-1) or in the neurite processes alone (miR-100001) (Figure 3C). The differential expression of miRNAs between sensory and motor neurons is also apparent from miRNA clone frequencies of abdominal versus pleural ganglia, the latter of which contain many more sensory neurons.
miR-124 has been shown to be important during neuronal differentiation and in specifying neuronal identity (Lim et al. 2005, Makeyev et al. 2007, Visvanathan et al. 2007, Cheng et al. 2009). Our finding that miR-124 is relatively absent in the motor neuron of a sensory-motor co-culture gave us the first indication that miR-124 may not be present in all neurons and may have functions in addition to maintaining neural identity. We therefore asked: might miR-124 be regulated by synaptic activity? Specifically, we wanted to know whether it might be modulated by serotonin, a neurotransmitter important for learning. We found, by Northern analysis, that already within one hour of exposure to five spaced pulses of serotonin the levels of miR-124 were consistently reduced by two-fold (Figure 4A). These findings were corroborated by in situ hybridization analysis, which also showed a drop in miR-124 levels in both the sensory neuron cell body and neurite processes within one hour after washout from five pulses of serotonin (Figure 4B). No change in miR-124 levels was observed in cells treated with just one pulse of serotonin (Supplemental Figure 1). To determine how long it takes for miR-124 levels to return to baseline after exposure to five pulses of 5HT, we performed a time course analysis and found that miR-124 levels continue to drop even two hours after 5HT treatment, but then slowly re-accumulate, returning to baseline by 12 hours (Figure 4C). To better understand the mechanism underlying the serotonin-induced regulation of miR-124, we tested whether the miR-124 precursor levels were also affected by 5HT, and found by real time PCR, that pre-miR-124 levels remained unaffected in sensory neurons after exposure to five pulses of 5HT (Figure 4D). This indicated that the regulation of miR-124 occurs downstream to the biogenesis of the precursor species, either at the level of the RNase III Drosha processing or turnover of the Argonaute-bound miRNA complex. Since 5HT is known to activate several downstream signaling pathways, including PKA (Castellucci et al. 1980), MAPK (Martin et al. 1997), PKC (Sacktor and Schwartz 1990), and the proteasome (Hegde et al. 1997), we applied inhibitors of each of these molecules, in the presence of 5HT, to determine which, if any, contributes most to the regulation of miR-124. We found, by Northern analysis, that a MAPK inhibitor (U0126) almost fully abolished the 5HT-induced down-regulation of miR-124. By contrast, inhibitors of PKC (Bisindolylmaleimide), and the proteasome (MG-132) had no effect, and a PKA inhibitor (KT5720) showed a modest, but insignificant, attenuation of the 5HT-induced miR-124 effect (Figure 4E).
In screening other miRNAs for serotonin dependent regulation, we found one miRNA (miR-184) that had an even more pronounced, 3-fold, reduction, and others that either showed modest (miR-125c) or no (miR-2b) regulation by serotonin (Figure 4A). This is the first demonstration that a synaptic neurotransmitter can regulate miRNA levels. In the case of miR-124 we find that this occurs rapidly, is sustained for many hours, occurs through MAPK signaling, and affects only the mature miRNA levels, not the precursor form. The ability of serotonin to modulate miR-124 levels is of specific interest because its previously known function in neuronal differentiation suggested constitutive expression in mature neurons to maintain neuronal identity.
To determine the physiological relevance of the 5-HT-induced changes in miR-124 levels, we altered miR-124 levels in sensory neurons and analyzed the effect on the 5-HT-induced long-term facilitation (LTF) of the synapses between the sensory and motor neuron (Figure 5A, B). Injection of a duplex miR-124 mimic (Dharmacon, Inc.), designed to increase the levels of miR-124 in sensory neurons, caused a significant reduction in LTF as measured at 24 and 48 hrs after exposure to five pulses of 5-HT (n=10), when compared to un-injected controls in the same co-culture (n=9, F(1,17) = 5.27, p < 0.05, two-way ANOVA with one repeated measure (time); Figure 5D, E). Conversely, injection of the single stranded antisense miR-124 inhibitor (Dharmacon, Inc.), designed to reduce the levels of miR-124, caused a significant increase in synaptic facilitation of the 5HT treated synapses (n=14) with respect to untreated controls (n=14) as measured at 24 and 48 hrs (F(1,26) = 4.70, p < 0.04, two-way ANOVA with one repeated measure; Figure 5J, K). Control experiments with the injection of scrambled miR mimic (n=16) and scrambled miR inhibitor (n=16) did not show significant changes in LTF (p > 0.05, two-way ANOVA with one repeated measure; Figure 5G, H, M, N). The observed differences in LTF among the different treatments were not due to differences in the basal strength of the synaptic connections in the different experimental groups (Figure 5F, I, L, O). In situ hybridization confirmed that the miR-124 mimics and inhibitors were able to increase or decrease respectively the levels of miR-124 in sensory neurons (Figure 5C).
To further support these observations, we also performed physiological experiments using an alternative knockdown method. To inhibit miR-124, we bath-applied antisense 2′-O-methyl-oligoribonucleotides conjugated with the peptide penetratin (Qbiogene, Inc.). The penetratin-conjugated inhibitor (200 nM) is capable of crossing the membrane of cultured Aplysia neurons and of inducing inhibition of miR-124, as determined by in situ hybridization (Figure 5P, also shown for inhibition in whole ganglia by Northern blotting in Figure 6A). To induce a significant inhibition, we incubated cells with the penetratin-conjugated inhibitor for 24 hrs before testing the basal amplitude of the EPSP and applying five pulses of 5-HT. The experiments with penetratin-conjugated miR-124 inhibitor confirmed that LTF was significantly enhanced following inhibition of miR-124 (+127.1 ± 16.36, n=9) as compared to controls treated with penetratin-conjugated to a miR-194 inhibitor (+67.35 ± 18.18, n=9, p < 0.01, Newman-Keuls post-hoc test after two-way ANOVA; Figure 5R). The inhibition of miR-124, within these temporal limits, did not affect basal synaptic transmission while interfering with 5HT induced LTF (Figure 5Q).
How does the down-regulation of miR-124 lead to long-term facilitation? To find potential targets of miR-124, we screened many genes relevant to plasticity and known to be regulated by serotonin, for increases in expression levels after inhibition of miR-124. De-sheathed pleural ganglia were incubated in antisense 2′-O-methyl-oligoribonucleotides conjugated with penetratin to inhibit miR-124 (confirmed by Northern blot Figure 6A), following which total protein was extracted and Western blotted. We found that inhibition of miR-124 led to a robust up-regulation in the Aplysia pleural ganglia of CREB1, the activator of transcription required for long-term facilitation (Figure 6B). This is consistent with the fact that not only the protein (Bartsch et al. 1998) but also the CREB1 mRNA levels are up-regulated by serotonin (Liu et al. 2008). We also found several genes, whose expression levels were unaffected by miR-124 inhibition (Figure 6B). Specifically, CREB2, the repressor that antagonizes CREB1 in synaptic depression, was unaffected by miR-124.
To be certain that miR-124 acts through CREB, we asked whether inhibition of miR-124 might affect the regulation of genes downstream to CREB. We observed that all three known immediate response genes, induced by serotonin in a CREB- dependent manner, increased in their level of protein (Figure 6B) and in their level of transcript (Figure 6D) after inhibition of miR-124. These three genes are: 1) ubiquitin C-terminal hydrolase (UCH) (Hegde et al. 1997), 2) CAAT enhancer binding protein (C/EBP) (Alberini et al. 1994), and 3) kinesin heavy chain (KHC) (Puthanveetil et al 2008). The increased protein and mRNA levels of these three genes were specific, because control antisense inhibitor did not alter levels of UCH, C/EBP, or KHC, and moreover, inhibition of miR-124 did not affect other plasticity related genes such as MAPK, neurexin, and tubulin (Figure 6B, D). The observed induction of protein levels of UCH, C/EBP, and KHC by inhibition of miR-124 was further enhanced by exposure to 5HT (Supplemental Figure 2). This suggests that the miRNA inhibition is just one of perhaps many parallel 5HT-mediated events that converge to activate CREB and its immediate early genes.
A conserved putative target site (Supplemental Figure 3) for miR-124 in the CREB1 3′UTR of vertebrates and mammals indicated that miR-124 might directly bind and inhibit the translation of CREB1 mRNA. To determine whether miR-124 directly binds and regulates Aplysia CREB, we cloned the full length 3′UTR of Aplysia CREB1 and found a putative miR-124 binding site (9-mer seed + GU Wobble) near the poly A signal (Figure 6C). To test whether this site is functional, we examined the effect of miR-124 over-expression, on a luciferase reporter fused to the CREB1 3′UTR. We found that miR-124 over-expression was able to repress the expression of the luciferase reporter by 45% (p < .01) when carrying the wildtype UTR, but had no significant effect on the reporter when the seed of the miR-124 binding site in the UTR was disrupted by a 2 nt mismatch, or when the reporter was fused to a truncated form of the UTR that no longer contained the miR-124 binding site (Figure 6C). In addition, the over-expression of an unrelated miRNA, let-7, had no significant effect on the reporter construct carrying the full length CREB1 UTR (Figure 6C). As a positive control, an siRNA targeting the luciferase gene was able to repress luciferase activity by 80% (Figure 6C). These data indicate that miR-124 directly regulates Aplysia CREB1 by binding to its UTR near the poly A signal.
While cloning the Aplysia CREB1 3′UTR, we discovered a novel and previously uncharacterized isoform of CREB in Aplysia, which differs from the canonical CREB1 in its last exon and 3′UTR (Supplemental Figure 5). This newly identified CREB isoform (which we term CREB1d) also bore a putative miR-124 target site, but showed no direct regulation by miR-124 on luciferase reporter assays (Supplemental Figure 4). The lack of regulation could be because the seed of this site is weak (six-mer seed + GU wobble), or because this site is in the ORF, which is considered to be functionally weaker than sites in the UTR (Bartel 2009).
CREB1 is a transcription factor that acts as a switch to convert short-term, protein-synthesis-independent facilitation (requiring one pulse of 5HT), into long-term, protein-synthesis dependent facilitation (requiring five pulses of 5HT). Therefore, neurons that over-express CREB1 require only one pulse, rather than five pulses, of 5HT for the induction of LTF (Bartsch et al. 1998). If CREB1 were indeed regulated in vivo by miR-124, the inhibition of miR-124 in sensory neurons, through its enhancement of CREB1, should require fewer pulses of 5HT to cause LTF. Indeed, in response to even a single pulse of 5-HT cells treated with miR-124 inhibitor showed a significant level of facilitation after 24 hrs (+42.66 ± 6.18, n=35; comparable to that observed with CREB over-expression in Bartsch et al. 1998) with respect to a control miR-194 inhibitor (+20.57 ± 6.37, n=22, p < 0.04, Newman-Keuls post-hoc test after two-way ANOVA; Figure 6E) and with respect to control cells treated with vehicle alone (+11.77 ± 8.18, n=12, p < 0.01, Newman-Keuls post-hoc test after one-way ANOVA). The observed differences in the facilitation between treated and untreated groups were not due to differences in the basal strength of the synaptic connections as tested before 5HT application (Figure 6E). Together with the previous observations, these data support the idea that the 5-HT-dependent down-regulation of miR-124, by allowing an increase in the levels of CREB and CREB-dependent transcription, is an important component of a switch that converts short-term to long-term synaptic plasticity.
In mining the miRNAs of Aplysia by deep sequencing, we have obtained what is perhaps the most comprehensive catalogue of the miRNA population in a central nervous system. The well-conserved nature of these miRNAs encourages their study in other nervous systems. In Aplysia it specifically allows study of miRNAs at the level of single cells and single synapses in processes ranging from neuronal development to synapse formation, stabilization, and plasticity. As more organisms are mined for their miRNAs, we are likely to gain a better understanding of how ancient and diverse miRNA gene families are, and what constraints they face during evolution.
In this initial study, we describe 170 distinct miRNAs present in Aplysia, of which 13 were previously unknown. Recent studies (Lu et al. 2008, Liu et al. 2008, Grimson et al. 2008) indicate that miRNA evolution has been dynamic and that most species have undergone dramatic changes in their miRNA gene content characterized by greater than normal rates of gene loss, gain, and duplication events. However, we find the Aplysia miRNA content to be particularly stable relative to its last common ancestor. Aplysia has gained only 13 miRNAs from its shared ancestry with vertebrates (though this number is likely to increase as the Aplysia genome coverage improves), and preserves over 40 miRNAs that are subsequently lost in D. melanogaster and C. elegans. Interestingly, the abundant miRNAs in Aplysia CNS appear to be as well conserved in invertebrates as vertebrates (Figure 2), whereas the whole animal miRNA population in Aplysia (Figure 1b) has a significant enrichment of shared vertebrate miRNAs, compared with invertebrate miRNAs, and many of these, such as miR-15/16, miR-145, and miR-221 are abundant and have important function in mammals. The similarity of genes between Aplysia and vertebrate systems is not entirely due to loss of genes in C. elegans and D. melanogaster. An analysis of well-conserved miRNAs shows that the vertebrate homolog is often more similar in sequence identity to the Aplysia homolog than it is to the homologs of C. elegans and D. melanogaster. The underlying similarity between Aplysia miRNAs and vertebrate miRNAs may also correlate with similarity in targets and in function, therefore strengthening the ability to use Aplysia as a model to understand the role of miRNAs in mammalian and even human neural function.
Expression analysis in cultured neurons of sensorimotor synapses revealed several miRNAs that were localized to distinct cells and sub-cellular compartments. Of the miRNAs that were screened, the striking enrichment of miR-124 in the sensory neuron with relative absence of expression in the motor neuron was most surprising. Earlier studies of miR-124 found that it had a ubiquitous and constitutive expression pattern in most neuronal cell types of the mammalian brain, which together with its lack of expression in progenitor cells, suggested a primary role for miR-124 in specifying and maintaining neuronal cell identity. Our studies of miR-124 in Aplysia revealed that, in addition to being non-uniformly expressed in adult neurons, miR-124 is rapidly and robustly regulated by the neurotransmitter serotonin, indicating additional roles for miR-124 in mature neurons. Several other miRNAs showed a similar down-regulation by serotonin, suggesting a general mechanism by which synaptic activity might relieve negative constraints on gene expression during learning-related plasticity.
Our finding that small RNAs can be regulated by conventional neurotransmitters extends further the scope of neurotransmitter actions. Neurotransmitters were first appreciated in the context of their ability to (1) regulate gating of ion channels, and subsequently to (2) covalently modify protein substrates by activating second messenger pathways. Subsequently, it was found that transmitters (3) also regulate transcription (reviewed in Kandel 2001). We now describe a fourth function of neurotransmitter action, the regulation of small RNAs. These considerations raise the further question: How are the miRNAs regulated by serotonin? Recent studies have uncovered two major stages of regulation during miRNA biogenesis, one at the Drosha cleavage step that converts the primary transcript into a miRNA precursor, and the second at the Dicer cleavage step that converts the precursor to the mature form (Obernosterer et al. 2006, Heo et al. 2008, Michlewski et al. 2008, Viswanathan et al. 2008). The ability of serotonin to selectively affect mature miR-124 levels, without affecting its precursor, argues for regulation during Dicer processing, or during RISC incorporation and stabilization by Argonaute, or even perhaps is the result of passive turnover of the miRNA in response to increased turnover of their target mRNAs.
The study by Ashraf et al. 2006 was the first to show learning-dependent changes in RISC, and that this was dependent on the proteasome. In light of their finding, we reasoned that the proteasome may also be involved in the serotonin regulation of miR-124, especially since the changes in miRNA levels are rapid and may mean enhanced degradation rather than impeded maturation. However, we found that inhibiting the proteasome had no effect on the serotonin-induced down-regulation of miR-124. Instead, we did observe that a MAPK inhibitor almost fully abolished the ability of serotonin to regulate miR-124. MAPK is one of the major signaling molecules downstream of serotonin that is known to activate CREB (Martin et al. 1997, Impey et al. 1998), and our data would suggest that one way it does so is by relieving miR-124 inhibition of CREB. The implication of MAPK in miR-124 regulation is of further interest because a GO term analysis of potential neuronal miR-124 targets (Supplemental Table 7) indicates that several of these genes are MAPK regulated. The dissection of the precise mechanism by which MAPK down-regulates miR-124 will require first an understanding of the MAPK substrates in the RNAi pathway, and then an exploration of how phosphorylation events, say on Dicer or Argonaute, may lead to the destabilization of the mature miRNAs.
We find that miR-124 serves as a negative constraint on serotonin-induced long-term facilitation, since increased or decreased miR-124 levels in sensory neurons leads to a significant inhibition or enhancement respectively of synaptic facilitation. In particular, the inhibition of miR-124 confers to sensori-motor synapses a greater sensitivity for serotonin, since just one pulse of serotonin is sufficient to cause long-term facilitation. These physiology data also suggest that miR-124 inhibition is just one of many 5HT-mediated events that activates CREB to induce long-term facilitation, since the inhibition of miR-124 alone, in the absence of 5HT, does not lead to long-term facilitation. Therefore, while the observed effects of the miR-124 manipulations on LTF are of a significant magnitude, it is likely that these effects would be even greater if there were a coordinated manipulation of several miRNAs that act together in parallel pathways during synaptic plasticity. The observation that miR-124 levels affect facilitation both at 24 and 48 hrs after exposure to spaced pulses of serotonin suggests that miR-124 regulation is required not only for the induction phase, but that it is also critical for the maintenance phase of synaptic facilitation. Since miR-124 levels return back to baseline within 12 hours after exposure to serotonin, the initial drop in miR-124 during this time window appears to be sufficient enough to up-regulate the relevant transcripts to allow for facilitation for up to 48 hours after exposure to serotonin. Indeed, the up-regulation of many plasticity related transcripts are transient and fall into this initial time-window. The data also suggest that miR-124 does not significantly affect or contribute to serotonin-independent processes such as basal and constitutive synaptic activity. However, since all of our experiments were conducted on several day old cultures, at which point the cells and synapses are fully mature and stable, our studies leave open the possibility that miR-124 contributes to serotonin-independent processes in immature neurons such as neurite out-growth and synapse formation.
The negative constraint that miR-124 imposes on synaptic facilitation is mediated, at least in part, by its direct regulation of CREB. The fact that miR-124 inhibition significantly and specifically increases CREB1 levels, along with immediate downstream genes such as UCH, C/EBP and KHC, that miR-124 serotonin kinetics parallels the CREB1 serotonin kinetics, and that miR-124 inhibition can provide the switch necessary to convert short-term facilitation into long-term facilitation, all strongly support the conclusion that miR-124 can tightly control CREB and CREB-mediated signaling during plasticity. CREB has been extensively studied over the years for its regulation by kinase dependent post-translational modifications, such as phosphorylation by PKA and MAPK. The present study, however, is one of the first to address post-transcriptional regulation of CREB. While this additional level of regulation might appear redundant, for example by paralleling the function of CREB2, it is likely that miR-124 inhibition allows for more rapid and transient control over CREB expression, as well as the opportunity for CREB to be drawn into various distinct downstream pathways once activated. We also noticed that CREB, in turn, may be able to regulate miR-124 expression levels since there are several putative CREB binding sites in the presumed promoter region upstream of the Aplysia mir-124 gene (Supplemental Figure 2). Although Aplysia and mammalian systems have clear differences in the complexities of their CNS, and also even in the types of neurotransmitters used during long-term memory processes, the underlying calcium induced signaling pathways (including cAMP, PKA, MAPK, and CREB) and their functions are very much shared (reviewed by Kandel 2001). It is therefore very likely that miR-124 is activity-regulated in the mammalian hippocampus, and regulates CREB in much the same way as observed here, especially in light of the fact that the mammalian CREB1 UTR bears a conserved miR-124 target site as predicted by targetscan (Lewis et al. 2003), which was recently confirmed as a site directly bound by Argonaute in mouse brain (Chi et al. 2009).
In summary, we have identified a comprehensive set of brain-enriched miRNAs in Aplysia, many of which can be regulated by the neuromodulator serotonin, signifying potential roles in learning-related synaptic plasticity. Specifically, we demonstrated that brain-specific miR-124 responds to serotonin by de-repressing CREB and enhancing serotonin-dependent long-term facilitation. This initial study compels the exploration of how neuromodulators act through small RNAs during various forms of plasticity and whether some act locally at synapses. We also have evidence that some 5HT regulated Aplysia miRNAs regulate plasticity-related genes involved in local protein synthesis at the synapse (Fiumara, Rajasethupathy, and Kandel, in preparation). The likelihood of a coordinated set of miRNAs combinatorially regulating events at the synapse makes possible a new and rich layer of computational complexity that could be responsible for the emergence of discrete and long-lasting states of activity at the synapse.
All animals were obtained from the NIH/University of Miami National Resource for Aplysia. Prior to dissection, animals were anesthetized by injection of isotonic MgCl2 (337 mM) at a volume of 50%–60% of their body weight. RNA was isolated from dissected tissue according to the standard Trizol (Invitrogen) protocol, with additional extractions with acidic phenol:chloroform:isoamyl alcohol, and finally again with chloroform before precipitation in 3 volumes of ethanol. Starting RNA amounts for each library were as follows: whole animal – 250 μg; CNS – 90 μg; pleural ganglia – 45 μg; abdominal ganglia – 90 μg. Small RNA cloning was performed as described (Hafner et al. 2008). Pre-adenylated 3′adapters were used, along with a truncated T4 RNA ligase, Rnl2 (Ho et al, 2004) to avoid circularization of the microRNAs during 3′ adapter ligation. 5′ adapter ligation was carried out at standard conditions with T4 RNA ligase (Fermentas Life Sciences) in the presence of ATP. The adapter sequences were as follows: 3′ adapter - AppTTTAACCCGGCACCCTC; 5′ adapter – ATCGTaggcaccugaaa. After both ligation steps, and following RT-PCR, the markers were removed from the samples by PmeI digest. The samples were again PCR amplified, concatenated, and then cloned into the commercially available TOPO 2.1 vector as described (Hafner et al. 2008). A total of about 250,000 reads each were obtained for the whole animal and CNS libraries by 454 sequencing (454 Life Sciences, Connecticut, USA). Traditional Sanger sequencing was used (Columbia Genome Center, New York, USA) to obtain approximately 15000, 20000, and 30000 reads each from the abdominal, pleural, and CNS libraries respectively.
For annotation, various databases were used (referenced in supplemental methods) to obtain a collection of known annotated RNA sequences. We generated and added to this, a set of Aplysia tRNAs and rRNAs (explained further in supplemental methods). All cloned sequences were aligned to this annotation database and sequences with no annotation, but with appropriate length and processing requirements were considered miRNA candidates. These candidates were aligned to Aplysia trace archives, around which contigs were assembled, and the resulting miRNA precursors were entered into the annotation database for miRNA annotation. Sequenced clones were again aligned to the database for annotation and this recursive process continued until all miRNAs were catalogued and manually curated. All procedures pertaining to sequence extraction, annotation, contig assembly, building of miRNA precursors, families, and orthology tables are further discussed in supplemental methods.
To prepare cultured neurons for in situ hybridization, cells were fixed in 4% paraformaldehyde in artificial seawater for 15 minutes at room temperature, and then washed with PBS. Cells were then permeabilized with 0.1% Triton X-100 in PBS for 10 minutes, and then endogenous peroxidase activity was quenched using 3% H2O2 for 20 minutes, following which a 10 minute acetylation step was performed, all at room temperature, with quick 1x PBS washes between each step. Pre-hybridization was carried out for 1 hr at 42°C in 50% formamide, 5x SSC, 5x Denhardt’s solution, and 0.1 mg/ml each of salmon sperm DNA and yeast tRNA. Hybridization was then carried out for 4 hrs, at 42°C, with 60ng of 3′end-labeled digoxigenin (DIG) probe per 150 μl of hybridization solution. The first wash was done using 5x SSC for 20 minutes at 42°C, and two subsequent washes were done with 0.5x SSC for 10 minutes each at the same temperature. The probes were then blocked for 1 hr in 10% (in TBS) heat inactivated sheep serum at room temperature, incubated in 1:1000 dilution of anti-DIG-POD antibody (roche) in 1% sheep serum (in TBS) overnight at 4°C, then labeled for detection with TSA-Plus FITC system (PerkinElmer) according to the manufacturer’s instructions.
Whole CNS or pleural ganglia were dissected in ice-cold sea water, de-sheathed, and kept in L-15 supplemented with glutamine for 24 hrs at 18°C. Serotonin treatment was performed with 5 pulses of 10 μM 5HT for five minutes each at 20 minute intervals. All drug treatments were done at a concentration of 10 μM in L-15 and bath applied to CNS for thirty minutes prior to treatment with five pulses of 5HT. The inhibitors used in this study are as follows: KT5720 (PKA inhibitor, Calbiochem), U0126 (MAPK inhibitor, Sigma), MG-132 (Proteasome inhibitor, Sigma), and Bisindolylmaleimide (PKC inhibitor, Calbiochem). Inhibition of miRNAs was carried out using penetratin conjugated 2′-O-methyl antisense oligonucleotides. These oligonucleotides were ordered (Dharmacon, Inc.) with 5′ thiol modification and incubated, with equimolar concentrations of activated Penetratin (Qbiogene, Inc. PENA0500) featuring an N-terminal pyridyl disulfide functional group, for 15 min at 65°C, then 1 hour at 37°C. The penetratin conjugated antisense oligonucleotides were checked for conjugation efficiency by Coomassie staining on 17% polyacrylamide gels, and knockdown efficiency by Northern blot. 150 μl of 200 nM penetratin conjugated oligonucleotides were then applied to de-sheathed pleural ganglia in Eppendorf tubes, for a minimum of 4 hrs before washout, and kept in L-15 with glutamine for a minimum of 24 hrs before harvesting RNA or protein.
Northern blot analysis was performed as described (Landgraf et al. 2007). Between 20 and 40 μg of total RNA was loaded per lane, the probes were 5′32P-radiolabeled 21- or 22-nt oligodeoxynucleotides complementary to the predominantly cloned miRNA sequence, and the hybridization was done at 42°C. To monitor equal loading of total RNA, the blots were reprobed with 5′-TGGAGGGGACACCTGGGTTCGA-3′ to detect tRNA.
For Western blotting, protein was isolated from de-sheathed pleural ganglia by incubating and rotating in SDS-urea lysis buffer (50 μl per 2 pleural ganglia) for 15 minutes at room temperature followed by centrifugation at 13000 RPM for 10 minutes, and collection of the supernatant. Protein samples were then quantified using the BCA kit (Pierce Biotechnology) and 15 μg were loaded for Western blot analysis. The following commercial antibodies were used: CREB1 (New England Biolabs) 1:1000, MAPK (Cell Signaling Technology) 1:1000, C/EBP (Santa Cruz Biotechnology, Inc.) 1:1000, UCH (Biomolecules) 1:1000, Tubulin (Sigma-Aldrich) 1:10000. KHC and CREB2 were rabbit polyclonal antibodies raised in the laboratory. Following incubation with primary antibodies, a 1:10000 dilution of either anti-rabbit or anti-mouse antiserum was used to detect protein bands by chemiluminescence (Amersham Biosciences).
Ganglia were dissected, maintained, and treated as described above. RNA was isolated according to the traditional Trizol (Invitrogen) method. After the isopropanol precipitation, the pellet was washed with 70% ethanol, and converted into cDNA using random hexamer priming and Superscript III (Invitrogen Life Sciences). Primers were selected using the Primer Express software (Applied Biosystems) and chosen to ensure no significant amplification of DNA. The primer sequences were as follows:
HEK293 cells were transfected using Lipofectamine 2000 (Invitrogen) in 96-well plates (250,000 cells/well) at 50% confluency with 100 ng psicheck-2 dual promoter plasmid (Promega), with renilla bearing the synthetic UTRs, and firefly serving as the internal transfection control. Cells were simultaneously transfected with or without 5pmol miRNA duplex. Firefly and Renilla luciferase activities were measured 36 hrs after transfection with Dual-luciferase assay (Promega).
Cell cultures of Aplysia neurons were prepared as previously reported (Schacher and Proshansky 1983; Montarolo et al. 1986). For intracellular injections, miRNA mimic and inhibitors (Dharmacon, Inc.) were re-suspended in nuclease-free water (Ambion, Inc.) to obtain a final concentration of 5.0-5.5 μM, aliquoted and stored at −20°C. Before injection, the solutions were thawed and combined with 10% v/v 2M KCl and 5% v/v of a saturated fast green solution to monitor the intracellular injection under both electrophysiological and visual control. 2-5 μl of each solution were loaded into the tip of beveled sharp glass microelectrodes, and after impalement, sensory neurons were injected by 7-10 pressure pulses (1-10 psi; 300-500 ms) delivered through a pneumatic picopump (PV820; WPI). In some experiments, Aplysia sensorimotor co-cultures were pretreated with penetratin-conjugated inhibitors before electrophysiological recording and 5-HT application. To this aim, the existing culture medium was gently exchanged with a 200 μl solution of a penetratin-conjugated miRNA inhibitor that has been diluted in L-15 to a final concentration of 200 nM. The cells were then maintained for 24 hrs at 18°C and subsequently tested electrophysiologically after adding 0.5-1 ml of fresh L-15 into the culture dish.
P.R would like to thank Amanda Rice for an introduction to the small RNA cloning procedure, and Praveen Sethupathy, Jeff Chen, and Li-Chun Cheng for many helpful discussions throughout. We thank Vivian Zhu and Edward Konstantinov for technical assistance with cell cultures, and Minchen Chien and David B. Weir for sequencing the Sanger libraries. We are grateful to David Bartel and Eric Lai for sharing unpublished data. This work is supported by an HHMI grant P50 HG002806 and NIH grants R01 MH075026 and P01 GM073047.
F.F is also supported by the Italian Academy for Advanced Studies in America at Columbia University.