Address correspondence and reprint requests to Neil R. Smalheiser, Department of Psychiatry and Psychiatric Institute, MC912, University of Illinois at Chicago, 1601 W. Taylor Street, Chicago, IL 60612, USA. E-mail: email@example.com
We have characterized the expression of microRNAs and selected microRNA precursors within several synaptic fractions of adult mouse forebrain, including synaptoneurosomes, synaptosomes and isolated post-synaptic densities (PSDs), using methods of microRNA microarray, real time qRT-PCR, Northern blotting and immunopurification using anti-PSD95 antibody. The majority of brain microRNAs (especially microRNAs known to be expressed in pyramidal neurons) are detectably expressed in synaptic fractions, and a subset of microRNAs is significantly enriched in synaptic fractions relative to total forebrain homogenate. MicroRNA precursors are also detectable in synaptic fractions at levels that are comparable to whole tissue. Whereas mature microRNAs are predominantly associated with soluble components of the synaptic fractions, microRNA precursors are predominantly associated with PSDs. For seven microRNAs examined, there was a significant correlation between the relative synaptic enrichment of the precursor and the relative synaptic enrichment of the corresponding mature microRNA. These findings support the proposal that microRNAs are formed, at least in part, via processing of microRNA precursors locally within dendritic spines. Dicer is expressed in PSDs but is enzymatically inactive until conditions that activate calpain cause its liberation; thus, we propose that synaptic stimulation may lead to local processing of microRNA precursors in proximity to the synapse.
Since the pioneering report of Lagos–Quintana et al. (2002), it is known that hundreds of microRNAs are expressed in mammalian brain, and there is strong evidence that they regulate protein translation both during development and in the adult (reviewed in Tai and Schuman 2006; Kosik, 2006; Fiore et al., 2008). Furthermore, numerous components of the microRNA machinery, including dicer and the core RISC component eIF2c (an Argonaute homologue), are expressed within dendrites and are, in part, associated with post-synaptic densities (PSDs) (Lugli et al. 2005). Similarly, not only have polyribosomes and a broad range of mRNAs been detected within dendrites, but polyribosomes have been shown to be expressed within dendritic spines (especially after stimuli that elicit long-term potentiation) (Ostroff et al. 2002; Bourne et al. 2007) and mRNAs and translational regulatory components have been detected in proximity to PSDs as well (Asaki et al. 2003; Suzuki et al. 2007). Cytoplasmic bodies such as RNA granules, stress bodies and P bodies have also been identified within dendrites (Vessey et al. 2006; Ferrari et al. 2007).
A large body of evidence indicates that synaptic plasticity, dendritic spine growth, and learning are critically dependent upon regulation of specific protein synthesis near or within dendritic spines, and emerging evidence suggests that microRNAs play roles in these processes. For example, Schratt et al. (2006) showed in mouse that a specific microRNA, mir-134, regulates a specific dendritic mRNA, LIMK1, which is implicated in growth of dendritic spines. In Drosophila, Ashraf et al. (2006) showed that Armitage protein, a component of the RISC complex, was modified by proteolysis during long-term olfactory learning, and that learning was impaired in an armitage mutant. Fragile X mental retardation protein (FMRP), which regulates protein synthesis in dendritic spines and is critical for normal cognition, behavior, and spine development (reviewed in Vanderklish and Edelman, 2005) has recently been shown to interact with multiple components of the microRNA machinery which, at least in part, appear to mediate the effects of FMRP (reviewed in Qurashi et al. 2007).
Despite these advances, it is not clear whether a diverse population of microRNAs is expressed in dendrites and/or in dendritic spines, and whether they are as abundant near synapses as within the cell body compartment. Nor is it understood how microRNAs arrive at the synaptic compartment(s). There are at least three possible ways that microRNAs may arrive at dendrites (Tai and Schuman 2006; Kosik, 2007; Fiore et al., 2008): First, mature microRNAs may passively diffuse into dendrites, which seems to be the predominant mode within neurons that are still maturing (Kye et al. 2007). Second, mature microRNAs may be processed from precursors in the neuronal cell body and then actively transported to dendrites, either by themselves or by ‘piggybacking’ on their mRNA targets as the latter become transported to dendrites. Third, primary microRNA gene transcripts (pri-miRs) or small hairpin precursors (pre-miRs) may be actively transported to dendrites, and then processed to microRNAs locally. These scenarios are not mutually exclusive. However, because dicer, the Rnase III enzyme that cleaves precursors into mature microRNAs, is expressed within dendritic spines and is highly enriched at PSDs (Lugli et al. 2005), we hypothesized that local processing of pre-miRs occurs near synapses. At present, however, there is no published evidence that microRNA precursors are detectably expressed at all within dendrites or synaptic fractions of adult mammalian neurons.
Materials and methods
Two-month old male C57Bl/6 mice were employed in the present study. Each prep consisted of a pool of 3–4 forebrains (including cortex and hippocampus). Synaptoneurosomes, synaptosomes and isolated PSDs were prepared using standard methods (modified to preserve RNA integrity) and each prep was characterized to ensure that it had the expected enrichment of specific protein and RNA components, as previously described (Smalheiser and Collins 2000; Lugli et al. 2005). Forebrain was rapidly dissected, placed in RNAlater (Ambion, Austin, TX, USA) for the time of the dissection of 3–4 mice and each pool was immediately homogenized using a Dounce pestle in ice-cold homogenization buffer (HB) containing a cocktail of protease and Rnase inhibitors [50 mM Hepes, pH 7.5, 125 mM NaCl, 100 mM sucrose, 2 mM K acetate, 10 mM EDTA, 2 mM phenylmethylsulfonyl fluoride, 10 mM N-ethylmaleimide, 10 μg/mL leupeptin, 1 μg/mL pepstatin A, 2 μg/mL aprotinin, 160 U/mL Superase-In (Ambion), 160 U/mL Rnase-OUT (Invitrogen Life Technologies, Carlsbad, CA, USA)]. Synaptoneurosomes were prepared as in Lugli et al. (2005), except that the synaptoneurosomes were quickly pelleted and rinsed twice in 10–20 × volume of homogenization buffer containing all inhibitors prior to extracting RNA. Synaptosomes were prepared as in Smalheiser and Collins (2000) but sucrose was treated with RNAsecure (Ambion) as per the manufacturer’s instructions before making the sucrose gradient, and the sucrose solution was supplemented with Rnase-OUT (Invitrogen) 320 U/mL.
Total RNA was isolated with Trizol reagent (Invitrogen) according to manufacturer’s directions with a few modifications to maximize yield of small RNAs. Glycoblue 20 μg (Ambion) was added to the RNA precipitation step which was allowed to proceed overnight at −20°C. The RNA pellet was spun down at 20 000 g for 25 min at 4°C; rinsed with 80% ethanol in DEPC-treated water (Invitrogen); resuspended and treated with RNAsecure (Ambion); and treated with Dnase I using DNA-free (Ambion). Each prep was characterized for purity by OD 260/280 ratio and for integrity by running on agarose gels.
Synaptoneurosomes were lysed with 1% Nonidet P-40, and the pellet was spun-down at 20 000 g for 20 min and rinsed twice with HB buffer prior to preparing total RNA. The RNA was heated (100°C, 3 min) in loading buffer type II (Ambion) and loaded (3.8 μg RNA for miR-124a, 0.5 μg RNA for BC1) in a 10% Tris-borate-EDTA (TBE) UREA criterion gel (Bio-Rad, Hercules, CA, USA). The gel ran in 1X TBE buffer at 120 V for 1 h and the RNA was transferred in 0.5X TBE buffer to Hybond-N+ (Amersham Bioscences, Piscataway, NJ, USA) for 30 min at 20 V. Membranes were UV-cross-linked for 2 min and baked at 80°C for 45 min, incubated in ULTRAhyb hybridization buffer (Ambion) for 45 min at 68°C on a shaker, then biotin-linked oligonucleotide probes were added (BC1 probe at 0.1 nM and miR-124a at 0.3 nM) with incubation overnight at 42°C on a shaker. Membranes were rinsed and signals were detected using the BrightStar BioDetect kit (Ambion) with 1 min exposure to Hyperfilm ECL (Amersham Bioscences). Probe sequences were as follows: BC1: aaaggttgtgtgtgccagttaccttgtttt; mir-124a: tgGcaTtcAccGcgTgcCttAa (lower case is DNA, upper case is Locked Nucleic Acid; probes were modified with biotin at both 5′ and 3′ ends; probes were prepared by IDT DNA, Inc., Coralville, IA, USA).
RT-PCR and real time qRT-PCR measurements
See Supplementary material File 1 for a list of all primer sequences.
Housekeeping and synaptic RNAs (BC1, GAPDH, CAMK2A, ribosomal 18S RNA and U6)
Total RNA (0.5 μg) was reverse transcribed with short gene-specific primers with Tm around 38–42°C using Superscript II (Invitrogen) as per the manufacturer’s protocol with a few modifications. Denaturation was performed at 80–90°C for 5 min, followed by 5 min at 33°C, 30 min at 36°C and 30 min at 42°C. Real-time quantitative PCR was performed on a Stratagene MX 3005P real-time PCR Instrument (Stratagene, La Jolla, CA, USA). Each sample was run in duplicate or triplicate. All products were confirmed and distinguished from primer dimers by examining melting curves and by running the PCR product on 3.5% agarose gels. As negative controls, parallel samples lacked reverse transcriptase, or lacked added RNA sample. The PCR mix, 20 μL, contained 10 μL of SYBER GREEN PCR master mix (Applied Biosystems, Foster City, CA, USA) 0.4 μM of each pair of primers, and 5 μL of dilute cDNA (1 : 50). PCR was performed for 20 s at 95°C, 25 s at 58°C and 25 s at 72°C for 40 cycles followed by the thermal denaturation protocol.
RNA (1 μg) was employed for each reaction. Primer pairs specific to each precursor were employed, following the protocol described in Jiang et al. (2005), employing mouse sequences. The PCR mix was as described above, and PCR was performed for 20 s at 95°C, and 50 s at 60°C for 50 cycles followed by the thermal denaturation protocol. In our hands, microRNA precursors were more ‘finicky’ to measure using RT-PCR than other types of RNAs, and for some pre-miRs (especially those forming relatively tight hairpins) it was necessary to modify the Schmittgen protocols by changing the primer sequences or by raising the denaturation temperature. (We were never able to analyze pre-miR-143 successfully, presumably because it forms a very tight hairpin structure.) Besides confirming PCR products by melting curves and by running the PCR products on agarose gels, selected pre-miRs were also tested to verify that they were cut by specific restriction enzymes giving pieces of the expected sizes (Fig. 1). Note that the PCR method does not distinguish between primary microRNA gene transcripts and microRNA precursors; thus, additional size selection experiments were carried out to distinguish pri-miRs from pre-miRs (see Results).
Reverse transcription was performed in 20 ng of total RNA using specific TaqMan® MicroRNA Reverse Transcription and Assay kits (Applied Biosystems) following the manufacturer’s protocol. MicroRNA specific PCR employed the TaqMan miRNA specific assay kit from Applied Biosystems. This recognizes dimeric PCR product in a sequence-specific manner and thus does not detect primer dimers. The looped primers are designed to avoid detecting pre-miRs. Although it is conceivable that some pre-miRs are recognized under some conditions (i.e., when the mature microRNA derives from the extreme end of the pre-miR), this would not have a practical effect on measuring mature microRNA levels because the steady-state levels of mature microRNAs are typically 20–50 times higher than that of the corresponding pre-miRs (e.g., Jiang et al. 2005; Schmittgen et al. 2008).
Each microarray measured a single paired prep (total forebrain homogenate vs. synaptoneurosome fraction, SYN). Total RNA (10 μg) was size-selected and labeled with Alexa dye A3 (total homogenate) or A5 (SYN) by Invitrogen staff. The labeled RNA was mixed together and hybridized to a NCODE multi-species microRNA chip (Invitrogen version 1.0) using Maui hybridization. Invitrogen version 1.0 chips contain 2 spots for each annotated microRNA across multiple species – since many mouse, rat, and human microRNAs have identical sequences, up to six spots per array could be averaged for certain microRNAs. We removed from consideration any individual spots with high internal standard deviation (i.e., SD > 3X the median intensity value in either the red or green channel) as well as outlier spots (i.e., cases in which 1 or 2 spots had intensity values <2X background, and the remaining spots for that microRNA had intensity values > 3X background in either red or green channel). Raw intensity values were adjusted by subtracting local background values from each spot (any spot with resulting intensity <0 was assigned the value 1). The microarray intensity values, as well as the results of real-time RT-PCR measurements, indicate that the overall concentration of microRNAs relative to total RNA was similar in both the SYN fraction and the total homogenate (Fig. 2); the overall profile of intensity values was also similar across the three preps (data not shown). The intensity values were normalized further as described in Results.
Immunopurification with anti-PSD95 antibodies
Protein-G and Protein-A agarose beads were rinsed twice with HB buffer containing the cocktail of protease and Rnase inhibitors and once with HB diluted 1 : 1 with radioimmunoprecipitation (RIPA) buffer containing inhibitors. Purified synaptosomes (2.3 mg/mL) were lysed by diluting the suspension 1 : 1 with RIPA buffer (pre-treated with RNAsecure, Ambion and containing the protease and RNAase inhibitor cocktail) and pre-cleared with Protein-A agarose beads (Amersham) or with Protein-G agaorse beads (Amersham) for 60 min at 4C. Anti-PSD95 monoclonal antibody (University of California-Davis, clone 28/86, 1 mg/mL) or anti-synapsin I affinity purified polyclonal antibody (Chemicon, Temecula, CA, USA; AB1543P, 0.1 mg/mL) was added to pre-cleared synaptosomes and incubated over night at 4°C with gentle agitation. These were then mixed with protein-G beads (anti-PSD95 sample) or protein-A beads (anti-synapsin I sample) for 4 h, then the beads were rinsed twice with HB (diluted 1 : 1 with RIPA containing the inhibitor cocktail), once with HB, and once with 50 mM Tris pH 7.0. To analyze proteins associated with the immunoprecipitates, 40 μL Laemmli loading buffer was added, the beads were heated at 100°C for 5 min, and dithiothreitol was added (final concentration 1%). As negative controls, assays were conducted in the absence of primary antibodies and in the absence of synaptosomes. To analyze RNA associated with the immunoprecipitates, Trizol 1 mL (Invitrogen) was added to rinsed beads and total RNA was extracted.
Western blotting was performed as in Lugli et al. (2005). Blots were blocked in 1% nonfat dry milk for 1 h, room temperature, incubated with primary antibody anti-PSD95 monoclonal antibody (Clontech, 1 : 30 000) or anti-Synapsin I affinity purified polyclonal antibody (Chemicon, AB1543P, 0.1 μg/μL) overnight with rocking at 4°C and rinsed; goat anti-mouse IgG (peroxidase conjugated; Chemicon) or goat anti-rabbit IgG (peroxidase conjugated; Sigma A0545) was added at 1 : 30 000–50 000 for 2 h and rinsed; finally, blots were incubated in ECL-Plus reagent (Amersham) and exposed to film (Hyperfilm ECL, Amersham).
Synaptoneurosomes are small vesicular structures that are prepared by filtration of total forebrain homogenate through mesh of progressively decreasing size, followed by low-speed centrifugation (discarding the pellet) and higher-speed centrifugation (keeping the pellet). This fraction is thought to be enriched in dendritic spines. They were prepared by a modification of Weiler et al. (1997) as described in our previous study (Lugli et al. 2005), adding a cocktail of Rnase inhibitors to preserve RNA species. Synaptoneurosomal preparations were characterized by western blotting of PSD95 and by real-time RT-PCR measurements of BC1 and CAMK2a to verify that these components were enriched relative to total forebrain homogenate (see below). In the experiments reported here, each sample [synaptoneurosomes (‘SYN’) vs. total forebrain homogenate, ‘total’] was pooled from 3–4 forebrains, and three independent pairs of samples were tested in parallel on microRNA microarrays. An additional three pairs of samples were prepared and measured for follow-up experiments.
About three-quarters (178 of 232) of the mouse microRNAs present on the microarrays were detectably expressed in the forebrain homogenate using the threshold criterion of having average intensity values ≥2.5 X background. Of the microRNAs expressed in forebrain, 78% were detectable within the SYN fraction using the same criterion. Normalizing the intensity values to the mean intensity values of the microRNA population as a whole (i.e., the mean value was chosen for one dye channel on one chip and the others normalized relative to that one) had only a slight effect on the distribution of intensity values (Fig. 2). However, this did improve the linearity of the assay and adjusted for possible variations among preps as well as for possible differences in the red versus green dye channels. Thus, this was the preferred method of normalizing microarray data.
Individual microRNAs showed a wide range of relative enrichment ratios (Fig. 3). The lowest ratio was observed with mir-143, which was ∼ 7-fold less abundant in the SYN fraction than in the total homogenate, whereas on the other hand, mir-200c was ∼ five-fold more abundant in the SYN fraction. Thirty-seven microRNAs showed an enrichment ratio of two-fold or greater in the SYN fraction, which in this experiment, was similar to the number of microRNAs (31) which achieved statistical significance at p ≤0.05 (Table 1). Numerous microRNAs still achieved significance at more stringent criteria (Table 1).
Table 1. Number of microRNAs (enriched in synaptoneurosomes) measured by microarray that achieve statistical significance at different stringency criteria
# Observed microRNAs
# Expected by chance
Statistics were computed using the t-test (two-sided). Note that the Bonferroni correction for multiple testing is not appropriate for these data, since different microRNAs are not expressed independently of each other; see text.
The synaptic enrichment was also analyzed by quantile ranking of the microRNAs. This is a non-parametric method in which each microRNA in a sample is simply ranked in order of its average intensity value, from greatest to least, and thus is assigned a quantile value (from 100% to 0%). Then, for each microRNA, one computes the difference between its rank value in synaptoneurosomes versus its rank value in the total forebrain homogenate. As shown in Fig. 4, there was a very good agreement between the findings obtained using enrichment ratios and obtained using rank differences.
Real-time RT-PCR measurements were made of selected microRNAs chosen because they covered a range of intensities and enrichment ratios (Table 2). MiR-124a, an abundant microRNA which is thought to be ubiquitously expressed in neurons (but not glial cells), had an enrichment ratio of 1.04, which is similar to the value estimated by microarray (0.88). Thus, miR-124a is neither enriched nor depleted from the SYN fraction, which is consistent with the lack of enrichment for other neuronal microRNAs measured by microarray (e.g., miR-128a, enrichment ratio = 0.85, Supplementary material File 2). The enrichment ratios measured by RT-PCR generally showed close agreement with the values determined by microarray for most microRNAs (Table 2), including that for miR-143 which was strongly depleted from the SYN fraction (0.11 vs. 0.15; Tables 2 and 3). A quantitative discrepancy between RT-PCR and microarray enrichment values was observed for miR-339, which showed less magnitude of enrichment ratio by RT-PCR (1.46, Table 2) than by microarray (4.07; Table 3). Similarly, miR-350 showed an enrichment ratio of 1.48 by RT-PCR in contrast to a measured value of 3.44 by microarray (Tables 2 and 3). This suggests that the microarray system did not maintain linearity throughout the measured range. Nevertheless, the qRT-PCR method did confirm that both miR-339 and miR-350 were significantly SYN-enriched (Table 2).
Table 2. Enrichment ratios (synaptoneurosomes/total homogenate) of selected mature microRNAs measured by real-time qRT-PCR
Mean ± SEM
MicroRNAs were measured in 3–6 independent preps (three used in the microarray experiments and/or in three others prepared similarly). *Significantly enriched (or depleted) at a confidence level of p <0.01 or better.
1.04 ± 0.13
1.08 ± 0.22
1.70 ± 0.59
0.11* ± 0.022
1.46* ± 0.13
1.48* ± 0.07
1.79* ± 0.22
Table 3. microRNAs most and least enriched in synaptoneurosomes as measured by microarray
C. elegans homolog
C. elegans homolog
C. elegans homolog
As shown in Fig. 5, SYN-enriched microRNAs exhibited a wide range of intensity values, and were NOT predominantly found among the least abundant microRNAs (this was checked since small changes in intensity values for microRNAs near threshold could potentially cause enrichment ratios to be artifactually inflated). This finding, together with the agreement between enrichment ratios and rank differences (Fig. 4), confirms that the observed enrichment of certain microRNAs in the SYN fraction is robust.
To analyze the biological features of SYN-enriched microRNAs, the top 20-most enriched microRNAs were compared to the bottom 20 (least-enriched) microRNAs (Table 3). The top 20 include eight microRNAs that were cloned originally from cultured cortical neurons (miR-322–352, Kim et al., 2004), whereas none of these were found among the bottom 20. This is consistent with the expectation that synaptoneurosomes are enriched in pinched-off dendritic spines, which are derived from pyramidal neurons. Conversely, many of the bottom 20 microRNAs are expressed in many tissues and have been widely studied as being important in growth, differentiation and cancer (e.g., let-7b, miR-125a, 126, 143, 145, etc.). Several of the bottom 20 microRNAs have been reported to be expressed predominantly in astrocytes (miR-29b and 29c, Smirnova et al. 2005). The two sets of microRNAs differed in their evolutionary histories as well: The top 20 includes 11 microRNAs that have only been described in rodents and/or other mammalian species, and none have homologues in invertebrates. In contrast, the bottom 20 includes only 1 microRNA, their expression is restricted to mammals; most are expressed in fish and at least three have homologues in C. elegans (Table 3). Moreover, several of the bottom 20 microRNAs arise from multiple precursor genes (data not shown); it is likely that this reflects microRNA gene duplication and divergence over evolutionary time. In conclusion, the microRNAs that are enriched in synaptoneurosomes tend to be expressed predominantly in pyramidal neurons, and tend to be evolutionarily newer than forebrain microRNAs that are relatively depleted from the synaptic fraction.
In many cases, two microRNAs encoded on the same chromosome at adjacent sites (presumably transcribed by the same primary transcript) showed very similar SYN enrichment ratios. For example, this was observed for three pairs of microRNAs within the top 20 (miR-322 and 351; miR-182 and 183; miR-429 and 200b) and four pairs within the bottom 20 (miR-29b and 29c; miR-34b and 34c; miR-126-5p and 126-3p; and miR-143 and 145). This underscores the consistency of the microarray results. Biologically, the phenomenon may simply reflect the fact that a primary transcript will be processed within the same cell type as the resulting mature microRNAs. Regardless, these cases are common and are important to note for data analysis, because they emphasize that the expression of any one microRNA cannot be assumed to be independent of the others.
Using real time qRT-PCR, the expression of selected microRNA precursors was compared in synaptoneurosomes versus the total forebrain homogenate. As shown in Fig. 6, microRNA precursors were readily detected within the SYN fraction, at levels that were comparable to the total homogenate. Because the primers derive from precursor sequences lying beyond the mature microRNA, this method should not detect mature microRNAs at all. On the other hand, the method detects both primary microRNA gene transcripts (pri-miRs) as well as small hairpin precursors (pre-miRs), as considered further in the section Distinguishing pre-miRs from pri-miRs in synaptic fractions.
Synaptosomes and post-synaptic densities
Synaptosomes are a well-studied synaptic fraction that is thought to consist predominantly of axon terminals with adherent PSDs. Detergent extraction of synaptosomes using Triton X-100 results in isolated PSDs. Synaptosomes and isolated synaptic densities were prepared and characterized as described in our previous publications (Smalheiser and Collins 2000; Lugli et al. 2005; see also Fig. 7a). Synaptosomes and isolated PSDs are more extensively rinsed than synaptoneurosomes and their isolation involves additional steps (e.g., sucrose gradients and detergent extraction) that should remove most contaminants. On the other hand, losses of RNA may occur because of the longer time involved in preparing synaptosomes and PSDs and/or to their extensive rinsing. We noted that levels of all RNAs measured, including ribosomal 18S RNA, U6, CAMK2a and BC1, as well as mature microRNAs and their precursors, were lower in synaptosomes relative to that observed in synaptoneurosomes. Because the components that were soluble (see below) were disproportionately lower in synaptosomes relative to synaptoneurosomes, it is likely that the losses predominantly reflect the more extensive rinsing of these preps (data not shown). Nevertheless, microRNA precursors were readily detected within synaptosomes and PSDs and were amenable to study.
Synaptosomes were extracted with non-ionic detergent and measurements were made of the soluble extract versus the insoluble residue (i.e., the PSD fraction). As shown in Fig. 7b, the microRNA precursors were all predominantly associated with the PSD fraction. In contrast, the mature microRNAs were predominantly detected in the Triton-soluble fraction (Fig. 7b). The small RNA BC1 was almost entirely soluble as well. These results were confirmed for mature mir-124a and BC1 using Northern blotting of synaptoneurosomal lysates as well (Fig. 8).
To verify further that microRNA precursors are not simply contaminants of synaptic fractions that are co-isolated during subcellular fractionation, synaptosomes were lysed with RIPA detergent and immunopurification was carried out under stringent conditions using anti-PSD95 antibodies (Vinade et al. 2003). As a negative control, an antibody was employed against synapsin I, which is associated with pre-synaptic vesicles in this prep. Other negative controls included omitting primary antibody, and omitting synaptosomal lysate. The anti-PSD95 and anti-synapsin I antibodies did, as expected, bring down the respective proteins (Fig. 9a). As shown in Fig. 9b, anti-PSD95 brought down a striking amount of microRNA precursors (67.9-fold over baseline) and detectable, though small amounts of mature microRNAs (1.62-fold over baseline) both of which are significant at p <0.05. Synapsin I antibody did not bring down significant amounts of either microRNAs or precursors (Fig. 9b).
Distinguishing pre-miRs from pri-miRs in synaptic fractions
The real time RT-PCR primers employed here do not distinguish between small microRNA hairpins (pre-miRs), which are ∼ 70–110 nt. in length, and primary gene transcripts (pri-miRs), which are generally several kilobases long (Saini et al. 2007). Although pri-miRs are generally thought to be processed within the nucleus, there are some reports that pri-miRs can be transported intact to the cytoplasm under some conditions (Eis et al. 2005; Barthelson et al. 2007). As well, a noncoding RNA localized to the rat neuromuscular junction (Velleca et al. 1994) was identified as encoding a primary microRNA transcript (Rodriguez et al. 2004). Thus, pri-miRs might possibly be expressed in the cytoplasm or within synaptic compartments in the CNS as well.
It would have been desirable to pick primer pairs that are specific for selected pri-miRs from within the sequences flanking the pre-miRs, but the genomic structure of the microRNAs studied here have not been well characterized. As an alternative strategy, we employed the mirVana Paris kit (Ambion) to separate total synaptosomal RNA into pools of large and small RNAs, with a stated cut-off of ∼200 nt. Large versus small RNA pools were measured by real time RT-PCR. Pre-miRs were chosen that have a simple relationship with the mature microRNA (i.e., multiple precursors do not exist for the same microRNA), that were readily measured by qRT-PCR, and that represented a wide range of SYN enrichment values, being on the top-20 list (miR-429, 146 and 100) or the bottom-20 list (miR-125a, 126, 153, and 301).
Using this method, miR-146 (a top-20 microRNA) and miR-125a (a bottom-20 microRNA) showed clean size separation, with the mature microRNAs being found predominantly in the small fraction and the pre-miRs predominantly in the large fraction. The pool of small RNAs contained 82% of the total amount of U6, 80% of the precursor activity for miR-146 and 88% of the precursor activity for miR-125a. This indicates that synaptosomes express precursors of miR-146 and 125a predominantly in the form of pre-miRs but little or no pri-miRs. In contrast, the pool of small RNAs within the total forebrain homogenate (which contains nuclear as well as cytoplasmic RNA) still contained most (86%) of the U6 RNA, but only 30% of the measured precursor activity for miR-146 and only 14% of the measured precursor activity for miR-125a. This suggests that, in the forebrain tissue as a whole, the pri-miRs for miR-146 and 125a are more abundant than are the pre-miRs.
The size selection method was evidently affected by RNA sequence and secondary structure; for example, we found that synthetic pre-miR-122a failed to partition into the small RNA pool both when tested by itself and when it was spiked into total forebrain RNA (data not shown). Similarly, we were not able to estimate the relative expression of pre-miRs versus pri-miRs in synaptosomes for some of the pre-miRs because they partitioned predominantly into the large RNA fraction, or showed partial sorting into both large and small fractions. Nevertheless, it was possible to estimate the relative synaptic enrichment of the entire set of selected pre-miRs by measuring their relative abundance in the small RNA pools of synaptoneurosomes versus total forebrain homogenate. The pre-miRs whose corresponding mature microRNAs showed high SYN enrichment (pre-mir-429, 146 and 100) exhibited significantly higher SYN enrichment as a group than did the pre-miRs whose corresponding mature microRNAs showed low SYN enrichment (pre-miR-125a, 126, 153, and 301; p <0.05). This is consistent with the hypothesis that mature microRNAs expressed in synaptic fractions are likely to arise from local processing of their precursors.
Association of microRNAs and their precursors with protein complexes
Mature microRNAs are known to bind to the Argonaute homologue eIF2c as part of the RISC complex, and to be associated with FMRP as well (Caudy et al. 2002; Ishizuka et al. 2002). Several previous studies have also noted an association of FMRP with pre-miRs (Jin et al. 2004; Plante et al. 2006; Wulczyn et al. 2007), which is very interesting because FMRP or its homologues may potentially help to transport pre-miRs to dendrites (cf. Davidovic et al. 2007; who found that FMRP is a molecular adaptor between RNA granules and kinesin motors), and/or may potentially sequester pre-miRs from being cleaved by dicer. To investigate the association of microRNA precursors with proteins in brain in vivo, we prepared a low-speed S1 cytoplasmic fraction of forebrain homogenate and carried out immunoprecipitations using antibodies specific to a variety of microRNA pathway components, including dicer, eIF2c, FMRP, and PACT, as well as antibodies against PSD95, synapsin I, and MECP2. The immunoprecipitates were extracted for total RNA and measured for a series of selected mature microRNAs and microRNA precursors. (Note that the S1 fraction should exclude nuclear RNA where the majority of pri-miRs are thought to reside.) The immunoprecipitations were carried out under stringent conditions, i.e., in the presence of RIPA buffer. Some biologically significant interactions are disrupted by RIPA; for example, the previously reported association of dicer with eIF2c (Lugli et al. 2005) was not detectable in the presence of RIPA, whereas the association of dicer with FMRP was preserved (data not shown). Thus, binding that is detected in RIPA is likely to be specific and strong.
As shown in Fig. 10, FMRP showed a striking association with both microRNA precursors and mature microRNAs. Dicer showed a strong association with microRNA precursors as well, but no significant interaction with mature microRNAs.
The present paper demonstrates that most of the microRNAs expressed within the adult mouse forebrain are readily detectable within synaptic fractions, at levels that are comparable to those seen within the total forebrain homogenate. In fact, 37 microRNAs were enriched two-fold or greater in synaptoneurosomes (SYN) as measured by microarray (Fig. 3), and the enrichment was confirmed for selected microRNAs by real time qRT-PCR (Table 2). The magnitude of SYN microRNA enrichment estimated by microarray (2–5-fold) was comparable to the enrichment observed with two known synaptic RNAs, BC1, and CAM kinase II alpha mRNA (Fig. 6). The forebrain microRNAs that were enriched in synaptoneurosomes were biologically quite distinct from microRNAs that were relatively depleted from this fraction, both in their tissue expression patterns (many were expressed predominantly in pyramidal neurons) and in their evolutionary histories (SYN-enriched microRNAs tended to be evolutionarily new, often mammalian-specific or rodent-specific). These findings strongly suggest that a broad, diverse, yet biologically coherent population of microRNAs is expressed within dendrites and within dendritic spines, where they may be expected to contribute to the regulation of local protein synthesis.
MicroRNA precursors were also readily detected in synaptoneurosomes at levels similar to or greater than that found in the total forebrain homogenate (Fig. 6). Within synaptosomes, the microRNA precursors were predominantly associated with PSDs (Fig. 7). This was further confirmed by co-immunoprecipitation with PSD95 protein (Fig. 9). Although the real time RT-PCR method used here does not distinguish between primary microRNA gene transcripts (pri-miRs) and small hairpin precursors (pre-miRs), size fractionation experiments indicate that, at least in the case of miR-146 and 125a, pre-miRs are the predominant form within synaptic fractions. The available data do not exclude that pri-miRs may be expressed near synapses in some cases. However, we conclude that microRNA precursor forms (and not merely mature microRNAs) are prominent synaptic components.
The findings reported here support our earlier proposal that pathways related to RNA interference regulate long-term gene expression in the mature brain (Smalheiser et al. 2001). They also extend the observations that the RNAse III enzyme dicer is expressed within dendritic spines and is enriched in PSDs, in a form that is enzymatically inactive (Lugli et al. 2005; see also Fig. 7a). It is unlikely that dicer bound to PSDs lacks a necessary co-factor needed for processing, since the RNAse III enzymatic assay (Lugli et al. 2005) employed exogenous dsRNA which does not require any co-factors. As well, at least one known dicer co-factor, PACT, is detectably associated with isolated PSDs (unpublished data). NMDA stimulation of hippocampal slices or raising intracellular calcium levels within synaptoneurosomes causes the calpain-dependent liberation of enzymatically active dicer (Lugli et al. 2005). Thus, dicer and pre-miRs are both located at PSDs but are prevented from processing further under resting conditions, apparently because of the presence of endogenous inhibitor(s) of dicer activity and/or pre-miR processing (cf. Leuschner and Martinez 2007) that are removed under stimulated conditions (e.g., by calpain, by another calcium-dependent protease or by a phosphorylation event).
The present findings also support the suggestion that the FMRP may be involved in the processing arm of the microRNA pathway (associated with transport and/or processing of pre-miRs) as well as the effector arm (associated with mature microRNAs, RISC, and target mRNAs). MicroRNA precursors were strongly associated with both dicer and FMRP in cytoplasmic extracts, and dicer interacted with FMRP, as shown by co-immunoprecipitation with specific antibodies under stringent conditions (Fig. 10). Thus, it is likely that pre-miRs, dicer and FMRP are part of a single complex in vivo. Studies of whole brain in neonatal fmr1 knockout mouse have shown no deficits in overall microRNA levels (Landgraf et al. 2007), and our preliminary studies have confirmed that no global changes in microRNA levels are seen either in adult mouse forebrain or within synaptoneurosomes (G. Lugli et al., in preparation). However, this does not rule out a role for FMRP in pre-miR processing, because FMRP homologues FXR1P and/or FXR2P may also interact with mature microRNAs and microRNA precursors (Jin et al. 2004; Plante et al. 2006; Duan and Jin 2006). As well, we have observed alterations in specific microRNAs in fmr1 knockout mouse that vary according to brain region and that can affect synaptic fractions differently than whole tissues (Lugli et al., in preparation). Thus, further studies are needed to learn whether FMRP homologues may play a role in pre-miR processing in conjunction with dicer at PSDs, and especially under conditions of localized synaptic stimulation that elicit long-term plasticity.
Finally, how might the microRNA machinery described here contribute to the regulation of protein translation near and within dendritic spines, which has been intensively studied by neurobiologists as a fundamental mechanism of synaptic plasticity? Certainly, localized protein synthesis occurs within dendritic spines: mRNAs and translational regulatory components have been detected in close proximity to PSDs (Asaki et al. 2003; Suzuki et al. 2007), and intense stimuli that elicit long-term potentiation (LTP) have been shown to cause some polyribosomes to move into dendritic spines where they are situated very close to active synapses (Ostroff et al. 2002; Bourne et al. 2007). Such conditions are also associated with localized increases of calcium and activation of calpain. We propose that synaptic stimulation may lead to local processing of pre-miRs by dicer near the post-synaptic density, leading to the loading of RISC with mature microRNAs that bind to any available target mRNAs that are restricted to the activated synapse.
However, it is not clear whether microRNAs should be expected to repress translation or to enhance it: Whereas microRNAs do appear to repress translation of mRNAs within neurons under resting conditions (Kosik 2006; Schratt et al. 2006), it has been recently discovered that RISC can actually ENHANCE translation of a target mRNA as a function of the cell cycle, depending on the facultative binding of a FXR1 isoform to RISC (Vasudevan et al. 2007). It is not too hard to conceive that a similar switch might occur reversibly in the context of intense synaptic stimulation. Indeed, although FMRP (like microRNAs) is thought of as a translational repressor, FMRP is required for a transient burst of translation of specific synaptic target mRNAs to occur upon stimulation by mGluR5 agonist (e.g., Weiler et al. 1997; Muddashetty et al. 2007). This apparent paradox may be explained if FMRP works together with microRNAs in a concerted but transient switch from repression to enhancement of target mRNA translation. Upon mGluR5 stimulation, FMRP undergoes a rapid but transient dephosphorylation that might be related to this transient burst of translation (Narayanan et al. 2007). Further research is needed to learn whether processing of pre-miRs near synapses, expected to accompany LTP-eliciting stimuli, will have transient or long-lasting effects on their target mRNAs, and whether the effects will be repressive or stimulatory.
It is even less certain what the fate may be of synaptic microRNAs that become loaded onto RISC but that do not encounter target mRNAs. Are these rapidly destroyed? Or do they persist for hours (or longer) within dendritic spines? If they have a significant lifetime, then potentially they may bind mRNAs that are being transported down dendrites and ‘trap’ them to the vicinity of previously activated synapses. This could potentially serve as a type of synaptic tagging and capture (cf. Martin and Kosik 2002; Barco et al. 2008).
We thank Chris Adams and Bradley Love for supervising the microarray labeling and hybridizations and for helpful discussions. Supported by NIH Grants MH81099, DC 05793 and LM07292. This Human Brain Project/Neuroinformatics research was funded jointly by the National Library of Medicine and the National Institute of Mental Health. The contents of this paper are solely the responsibility of the authors and do not necessarily represent the official views of NIH.