|Home | About | Journals | Submit | Contact Us | Français|
MicroRNAs (miRNAs) have emerged as one of the major regulatory mechanisms of gene expression. A major function of miRNAs involves the post-transcriptional regulation of target mRNAs, which is reported to occur primarily in the cytoplasm. However, there is a significant amount of evidence demonstrating the existence of small non-coding RNAs, including small-interfering RNA (siRNA), miRNA and Piwi-interacting RNA (piRNA) in the nucleus. In order to elucidate the potential subcellular localizations and functions of miRNAs, we have identified numerous miRNAs that are present in isolated nuclei from human colon cancer HCT116 cells. MicroRNA profiles were compared between cytoplasmic and nuclear fractions of the HCT116 cell line on the basis of multiple microarray analyses. MicroRNA species showing significant existence in isolated and highly purified populations of nuclei were selected and further tested with RT-PCR. The nuclear localization of the mature form of miRNAs was verified again by control RT-PCR excluding the detection of premature forms of miRNA, such as pri-miRNA or pre-miRNA. The elevated levels of representative miRNAs identified in purified nuclei were confirmed by northern blot analysis, supporting the notion that significant numbers of mature miRNAs exist not only in the cytoplasm but also in the nucleus. These results will likely provide a basis for further studies concerning the intracellular trafficking and nuclear location of miRNAs.
MicroRNAs (miRNAs) are a class of small (~21–24 nt long) noncoding RNAs, which are known to be involved primarily in the negative regulation of target messenger RNA (mRNA).1–4 Primary transcripts of miRNA (pri-miRNA) are processed by digestion with the RNase-III enzyme Drosha in the nucleus,5 generating pre-miRNAs of ~70 nt. The pre-miRNA is then exported from the nucleus to cytoplasm by Exportin-5, a Ran-GTP dependent transporter.6 Eventually, the transported pre-miRNA is cleaved by another RNase-III type enzyme Dicer that releases the mature miRNA duplex.7,8 The single strand of mature miRNA complementary to target mRNA sequence is incorporated into the RNA-induced silencing complex (RISC).9
The biogenesis of miRNA suggests that the mature form of miRNA remains in the cytoplasm after Exportin-5-mediated transport from the nucleus. This is also implicated by the cytoplasmic presence of mRNA, which is the main functional target of miRNAs. However, there have been sporadic reports to suggest the presence of mature miRNAs in the nucleus of animal cells. For example, a hexanucleotide element at the 3′ end of miRNA was suggested to direct the subcellular localization into the nucleus.10 At least two miRNAs have been identified in the nucleoli of rat myoblasts.11,12 The detection of nuclear Ago2, which is the essential component of RISC,9 also implies that there should be mature nuclear miRNAs, although the potential function(s) of such miRNAs in the nucleus is unknown.13 Furthermore, a novel nuclear-cytoplasmic shuttling system, employing another karyopherin CRM1, revealed endogenous miRNAs in the nucleus, although their biological function remains undefined.14
Despite growing evidence of mature forms of nuclear miRNAs, there has been no comprehensive investigation of the extent to which they reside in the nucleus. As an initial step towards understanding the potential intracellular trafficking of miRNAs, we have generated a list of miRNAs that are detected in highly purified nuclei from a human cancer cell line HCT116 at levels either comparable to or higher than those of the cytoplasm. Several rounds of microarray analyses collating miRNAs between nuclei and cytoplasmic fractions provided the basic candidate profiles of possible nuclear miRNAs. Selected miRNAs were further tested by RT-PCR, which also excluded the potential bias of the signals by nuclear precursor miRNAs such as pri-miRNA or pre-miRNA. Finally, northern blot analysis confirmed the presence of mature forms of the miRNAs in the isolated nuclei from HCT116 cells. Our results indicate that a considerable number of miRNAs may exist in cell nuclei, and provide the basis for future studies regarding their potential functions.
In order to generate a comprehensive profile of mature nuclear miRNAs, both nuclei and cytoplasmic fractions were simultaneously isolated from HCT116 human colorectal carcinoma cells. The purity of each fraction was assessed by immunoblot analyses (Fig. 1). The cytoplasmic marker, α-tubulin was appropriately detected only in the cytoplasmic fraction15 (Fig. 1, upper part). We detected DNMT3a, one of the mammalian de novo DNA methyltransferases,16 only in the nuclear fraction, confirming the clean separation of nuclear and cytoplasmic fractions (Fig. 1, middle part). It was also important to exclude additional contaminating subcellular organelles that lie in close proximity to the nuclear envelope. Specifically, we tested the fractions with calnexin, an endoplasmic reticulum-specific marker,17 and there was no detectable signal within the isolated nuclei (Fig. 1, lower part), further confirming the purity of each fraction derived from the HCT116 cells. Of note, an equivalent percentage (e.g., 7.5% of each total fraction from the same ~6 × 106 cells for DNMT3a) from each fraction was analyzed for a rigorous comparison.
After establishing a high level of purity for both nuclei and cytoplasm, total RNA was isolated from each of the fractions and analyzed on a microarray printed with probes for human miRNAs. Each probe was composed of a linked doublet of a complementary sequence to the target miRNA. To maintain the quantitative nature of the analysis, the equivalent percentage (~25%) of total RNA from the same proportions of each fraction (50% from ~6 × 106 cells) should be employed in the microarray analyses. However, to prevent saturation of the signal from cytoplasmic total RNA, we used reduced amounts, and then multiplied the signals from cytoplasmic RNA by proportional factors to achieve the equivalent comparisons. We summarized three independent analyses into a graphic representation (Fig. 2). Two analyses were carried out with nuclei isolated by additional treatment with MgCl2 and Triton X-100 for the removal of outer nuclear envelope.18,19 However, there was no significant difference between nuclei treated for outer membrane removal and those without such treatment in the overall pattern of detectable miRNAs. For analysis, the fluorescent signal of each miRNA was compared between the nuclear and cytoplasmic fractions; and the ratio of nuclear to cytoplasmic signal for each miRNA was averaged from three independent analyses. The overall average of this nuclear ratio across all miRNAs on the microarray was 0.471 ± 0.15, indicating miRNAs were detected in the nucleus at levels lower than half of those in the cytoplasm. We then selected miRNAs showing nuclear/cytoplasmic signals higher than the overall average 0.471, as miRNAs with potential nuclear localization (Fig. 2A). MicroRNAs such as let-7d, miR-19a, miR-19b and miR-200b, demonstrated nuclear localization at a level similar to that of cytoplasm. In contrast, miRNAs showing belowaverage (<0.47) nuclear signals and most probably cytoplasmic in localization, were also included in the other part of the diagram (Fig. 2B). It is noteworthy that miR-768s scored highest in levels of nuclear signal. Originally, the probes for miR-768-5p and miR-768-3p were included in the probe set for the microarray since they were annotated as human miRNAs. However, phylogenetic analysis revealed that miR-768 should be annotated as snoRNA HBII 239 (http://www.mirbase.org/cgi-bin/mirna_entry.pl?acc=MI0005117). The nucleolar origin of snoRNA HBII 239 clearly supports the microarray's capability to discriminate nuclear signals from cytoplasmic ones for small non-coding RNAs, further supporting the validity of these analyses.
Although microarray analysis provided a comprehensive profile of miRNAs with nuclear localization, we carried out RT-PCR to confirm the existence and abundance of the mature miRNA species in nuclei. The RT-PCR employed reverse transcription (RT) primers, which have a stem-loop structure and a short nucleotide sequence complementary to the 3′ end of target miRNA.20 We determined that RT primers combined with forward primers resulted in efficient amplification without the additional reverse primer complementary to the RT-primers. The expected size of the RT-PCR product from miRNA in a mature form was about 70 nt. A few miRNAs, which demonstrated higher levels in the nuclear fraction revealed nuclear localization at levels similar to those from cytoplasm (Fig. 3A). Repeatedly, the equivalent percentage of total RNA from each fraction was used as template for RT-PCR after identical number of dilutions to achieve the proper amount of template total RNA. Template amount and number of PCR cycles were optimized to generate products visible in the gel and still below the saturated levels (plateau) of amplification. RT-negative controls were run for most of the reactions without visible products. RT-PCR for U6 snRNA was used as a reference control for a nuclear-localized small non-coding RNA. RT-negative control of U6 RNA excluded the contamination of each fraction with genomic DNA.
Two different forms of precursor miRNAs, pri-miRNA and pre-miRNA, are localized to the nucleus, and contain the target nucleotide sequences of mature miRNAs. Even though the RT-PCR amplification from precursors of miRNA was shown to be minimal when compared to the amplification from authentic mature miRNA,20 we elected to determine whether precursors could produce detectable signals from RT-PCR with similar primer sets used for miRNA RT-PCR reactions. If it was possible for stem-loop RT primers to bind target nucleotide sequences in the mid-region of precursor RNA molecules, enabling RT-PCR amplification, stem-loop RT primers designed to target midnucleotide sequences specific only to precursors should also result in detectable amplification. In addition, precursor-specific forward primers could be also tested in combination with authentic stem-loop RT primers. Control RT-PCR with stem-loop RT primers of which complementary target nucleotides existed only in the precursor forms (miR-16 and miR-222) resulted in no visible products even from nuclear fractions enriched in precursor forms (Fig. 3B). Another test with precursor-specific forward primer and ordinary stem-loop RT primer (miR-29c) also resulted in no visible products from nuclear fractions (Fig. 3B), authenticating the previous miRNA RT-PCR results. These RT-PCR analyses confirmed the presence of numerous miRNAs in the nucleus in their mature form.
Northern blot analysis can discriminate the precursor forms (>70 nt) from mature miRNA (~21–24 nt long) in size, eliminating the possibility of precursor detection in the previous analyses of nuclear miRNA. Two independent sets of RNA samples from nuclear and cytoplasmic fractions, respectively, were investigated by northern blot assays (Fig. 4). Again, equal percentages of total RNA from each fraction were collated in the blot. The blots were probed for two different miRNAs, miR-19b and miR-195. In the previous microarray analyses, miR-19b produced nuclear signals well above the average, while miR-195 demonstrated a near-average nuclear signal (Fig. 2). U6 snRNA was used as a control for non-coding small RNA enriched in nuclei and tRNALyS was also probed as a cytoplasmic control.10 Even though the relative strength of the signal from the nuclear fraction was still smaller than that of cytoplasmic for both miRNAs, it should be noted that the summary diagrams (Fig. 2) were based on the nuclear signal in comparison with the average nuclear vs. cytoplasmic signals. On the northern blot, mature forms of both miRNA were apparently detected in the nuclear fraction in two separate experiments, thereby supporting the conclusion that significant amounts of mature miRNAs exist in the nuclei of HCT116 cells. In addition, weak signals from high-molecular-weight precursors were detected on the northern blot when probed with miR-19b (Fig. 4), in particular from nuclear fractions, indicating the sensitivity of northern blot as well as the proper size discrimination on the gel.
In order to gain insight into the potential cellular functions of nuclear-associated miRNA, we generated a list of genes which were already shown to be under the control of miRNAs identified in this study. Target genes and their references are listed (Table 2) and provided as Supplementary data, respectively. A number of common targets were identified with respect to cell cycle regulation, apoptosis and transcription factors, suggesting potential cellular pathways that could be regulated by a subset of nuclear miRNAs (Table 3).
The major goal of this study was to establish a comprehensive profile of miRNAs localized to the nucleus of a human cell line. The quality of the subcellular fractions was critical and entirely validated by multiple immunoblot assays with representative protein markers. MicroRNAs showing potential nuclear localization at levels higher than average were selected by a summary of microarray analyses; and the existence of nuclear miRNA was subsequently verified by RT-PCR and northern blot analyses.
The current model or consensus is that the major cellular function of miRNA involves post-transcriptional regulation of target messenger RNA.9 Depending on the mode of interaction with target mRNA, miRNA could lead to cleavage of the mRNA or inhibition of translation from the target mRNA.3,4,21 Interestingly, several recent studies provide strong evidence for the nuclear localization of a certain class of miRNAs in a short mature form.10–14 Most of those studies suggested sporadic cases of nuclear detection of miRNA(s). In order to establish the biological functions of such nuclear miRNAs, it would be beneficial to generate a comprehensive list of miRNAs that localize to the nucleus of mammalian cells. The potential role of nuclear miRNAs in the regulation of gene transcription could uncover an entirely new and exciting application for this family of small non-coding RNAs.
We speculate that the subcellular localization of miRNA to nuclei is regulated in a defined manner by quite specific mechanisms. It is interesting that after completion of the maturation process in the cytoplasm by Dicer and incorporation into RISC, the mature forms of certain miRNAs are trafficked back into the nucleus.10,14,22 Given the complicated and various mechanisms involved in the nucleocytoplasmic transport of RNA molecules, it would not be surprising that specific pathways exist to traffic mature miRNAs from cytoplasm back to nuclei. In fact, they may be in complexes with protein factors related to the function of miRNAs or their transport between nucleus and cytoplasm.23 Despite the small size of miRNAs, it is highly unlikely that the nuclear pore complex (NPC) allows significant random diffusion of regulator molecules into the nucleus. Although a small 9-nm channel remains open in the NPC, even small tRNAs cannot pass through without gated transport.24 As recently suggested, karyopherins such as CRM1 (Exportin-1) could be involved in the shuttling system of miRNAs back and forth between nucleus and cytoplasm.14 Additionally, Ago2, the major component of RISC was detected as a simple complex with a short RNA molecule in the nucleus besides the large 3-MDa complex of RISC in the cytoplasm.13 Identification of such a nucleocytoplasmic transport system for miRNA could provide clues into the potential function of nuclear miRNAs.
The target sequences of probes on a microarray are by nature included in the precursor forms of the miRNAs.5,6 Thus, it was critical to rule out signals from nuclear precursors, such as pri-miRNA or pre-miRNA in the detection of mature species. Control probes complimentary to the antisense strand of precursor miRNA (strand complementary to the mature miRNA sequence) could be used to evaluate the detection of such precursors in the microarray analysis since the antisense strand would only exist in the precursor forms. By subtracting the estimated signals from probes specific for precursor sequences, the fluorescent signals from the authentic miRNA probes could be corrected for nuclear fractions, which would include precursor elements. However, in the hybridization conditions of microarray analysis, miRNA sense strand sequences in the precursor would hybridize primarily with their antisense counterparts in the same molecule of precursors due to the close proximity between those complementary sequences in the single molecule. In addition, the pri- or pre-miRNA precursors in the nucleus appear to be in a rapid cycle of turnover during miRNA biogenesis since we were able to detect only barely discernible signals of the precursor forms on the northern blot where mature short forms of miRNA were easily detected in both fractions (Fig. 4). In support of this, precursor forms of miRNA could be only detected by RT-PCR following the knockdown of Drosha or DGCR8 that are components of pri-miRNA processing complex.25
Unexpectedly, we were not able to recapitulate the nuclear enrichment of human miR-29b as previously reported.10 This might have been been due to cell-specific regulation in the subcellular localization of a particular miRNA since we examined HCT116 cells and not HeLa cells, as originally reported for detection of nuclear miR-29b.10 Even though our microarray also included probes for other miRNAs that were very similar to miR-29b in sequence, including cytoplasmic miR-29a, the probe design in our microRNA microarray platform was able to discriminate 2-nt differences between similar sequences of miRNAs from the same family.
It is also interesting that nucleolar-enriched miRNAs identified in rat myoblasts12 do not overlap significantly with our list of potential nuclear miRNAs derived from human HCT116 cells. Again, this could be attributed to differences in the origin of the cell lines since our data were acquired from a human colorectal cancer cell line, in contrast to rat myoblasts that are replete with nucleoli. As exemplified by miR-206, a miRNA could be enriched both in the nucleoli and cytoplasm simultaneously,11 but the nuclear signal would be masked by strong cytoplasmic signal in our analysis.
It should be noted that Xist RNA was detected in a complex with RISC and miR-210,26 (Table 2). Xist is a long non-coding RNA which is well known for its crucial function in the mammalian X-chromosome inactivation and has been proven to be localized in the nucleus,27 painting the inactive X-chromosomes.28 This observation suggests that certain non-coding RNAs in the nucleus could be potential targets of nuclear miRNAs. Currently, most computational methods of miRNA target prediction use stringent seed pairing at the potential target 3′-UTR, as well as evolutionary conservation of the target pairing sites.29 With such predictive algorithms, it is impossible to identify potential noncoding nuclear RNA targets in the absence of associated UTRs. MicroRNA binding to Xist RNA was established by immunoprecipitation and biochemical analysis.26 Thus, the identification of nuclear factors, including proteins, RNA and DNA, which can interact with microRNAs, could provide valuable clues in understanding the potential function(s) of nuclear-associated miRNAs.
In summary, we have provided strong evidence for the nuclear localization of numerous miRNAs in a human cell line. These data will undoubtedly provide a basis for future studies to investigate the possible functions of nuclear miRNAs in mammalian cells. Further investigations such as in situ hybridization in a comprehensive scale and identification of nuclear (or nuclear envelope components) protein factors associated with mature miRNAs could provide valuable clues to elucidating their mode of nuclear trafficking and potential biological functions.
HCT116, a human colon adenocarcinoma cell line (American Type Culture Collection (ATCC); ATCC# CCL 247) was cultured in McCoy's 5a media with L-glutamine (10-050-CV, Mediatech, Manassas, VA), supplemented with 10% FBS and antibiotic-antimycotic mixture. The cell line was maintained in a humidified chamber with 5% CO2 at 37°C.
The overall procedure was based on an established protocol with few modifications.30 Briefly, cells were harvested by trypsin treatment and washed twice with ice-cold PBS. Cells were completely resuspended in the NP-40 lysis buffer [10 mM Tris-HCl, pH 7.4; 10 mM NaCl; 3 mM MgCl2; 0.5% (v/v) NP-40; protease inhibitors (Roche Applied Sciences); 100 U/mL RNasin (Promega)] by gentle vortexing. The lysis of the cells and nuclei were verified on light field microscopes. Nuclei were centrifuged for 5 min at 500x g. The supernatant cytoplasmic fraction was collected and centrifuged again at 5,000x g for 10 min to completely remove any contaminating nuclei. Nuclei were once again washed in the NP-40 lysis buffer, centrifuged and finally resuspended in the NP-40 lysis buffer. The outer nuclear envelope was stripped away by incubating the purified populations of nuclei with 0.5% Triton X-100 buffer containing 10 mM Tris-HCl, 5 mM MgCl2 and 0.25 M sucrose.
Discontinuous SDS-polyacrylamide gel electrophoresis (8%) was carried out31 with modifications. Lysates or fractions were denatured by incubation at 95°C for 5 min in the sample loading buffer composed of 50 mM Tris-HCl (pH 6.8), 2% SDS, 15% glycerol, 75 mM dithiothreitol and 0.05% bromophenol blue. Equivalent portions of nuclei and cytoplasmic fractions were used for quantitative comparison. For α-tubulin, 0.5% of each fraction from ~6 × 106 cells was analyzed. The protein levels of DNMT3a and calnexin were detected with 7.5% and 1% of each fraction, respectively, from the same number of cells. After gel electrophoresis, proteins were transferred onto a nitrocellulose membrane (Hybond ECL, GE Healthcare Life Sciences) by Mini Trans-Blot® Electrophoretic Transfer Cell (170-3930, BioRad) in 25 mM Tris-HCl and 192 mM glycine. The blocking and incubation steps were performed with standard TBS (20 mM Tris-HCl, pH 7.6; 150 mM NaCl) buffer containing proper blocking agent(s) for each target protein (for DNMT3a, 3% bovine serum albumin and 2% non-fat skim milk; for α-tubulin and calnexin, 5% non-fat milk). Dilution factors for primary antibodies were 1/2,000 for anti-α-tubulin antibody (T-6199, Sigma); 1/500 for anti-DNMT3a antibody (H-295, sc-20703, Santa Cruz Biotechnology); and 1/2,000 for anti-calnexin antibody (ab22595, Abcam). The membranes were washed in TBS-T (20 mM Tris-HCl, pH 7.6; 150 mM NaCl; and 0.1% Tween 20) for 5 min × 3. The dilution factor for secondary antibodies was ~1/12,000. After chemiluminescent reaction with SuperSignal West Pico Substrate (PIERCE), the blots were exposed to X-ray film for 10 to 30 min.
Total RNA was isolated from equivalent amounts of each fraction (e.g., 50%) with Trizol® LS reagent (Invitrogen). For microarray analysis, a set of miRNA probes was acquired from Invitrogen on the basis of the Sanger miRBase Sequence Database Release 9.0 (October 2006) (http://microrna.sanger.ac.uk/sequences/). The set had approximately 1,140 oligonucleotide probes, which targeted C. elegans, Drosophila, zebrafish, mouse, rat and human miRNAs, and also contained internal and negative controls. The oligonucleotides were printed on Corning GAPSII coated slides by the Microarray Facility at the University of Minnesota. Total RNA was labeled by ligation to 0.5 µg of a synthetic linker, pCU-DY547 (Dharmacon), as previously described.32 Reference DNA oligonucleotides were labeled with a ULYSIS Alexa Fluor 647 kit (Invitrogen). Labeled RNA and DNA were then mixed and hybridized to microarray slides. Slides were then scanned with a ScanArray 5000 machine (Perkin Elmer), and BlueFuse (BlueGenome) was used to quantify pixel intensities.
RT-PCR analysis of miRNAs was based on the previously reported stem-loop RT-PCR protocols with minor modifications.20 Stem-loop RT primers were combined with forward primers for each miRNA species (Table 1). The amount of template total RNA was adjusted for each miRNA species so that the final RT-PCR products were easily visible on the gel, but still maintaining the quantitative nature of RT-PCR and short of saturation. For a comparison between nuclei and cytoplasmic fractions, equivalent portions (e.g., 10% total RNA from each fractions isolated from the same number of cells) of total RNA were diluted with the same factor and equal volumes of diluted total RNA were used for each fraction. Titan One Tube RT-PCR System (Roche Applied Science) was used for RT-PCR with 1.5 µM of each primer. The reverse transcription step was followed by PCR cycles with the entire procedure consisting of 16°C for 30 min without the RT-PCR enzyme mix, 42°C for 40 min with the enzyme mix and 30 alternating cycles of 95°C for 20 sec and 60°C for 1 min. The number of PCR cycles was adjusted between 17 and 30 in order to make the reactions semi-quantitative. The negative control for reverse transcription was run with 2.5 U of Expand High Fidelity PCR enzyme mix (Roche Applied Science) instead of the RT-PCR enzyme cocktail and subjected to the same RT-PCR processes. RT-PCR products were analyzed by 12% non-denaturing polyacrylamide gel in 0.5X TBE buffer.
An equivalent ~18% of total RNA from the same number of cells (~6 × 106) was compared between nuclei and cytoplasmic fractions. Thus, about 50 µg of cytoplasmic and 16 µg of nuclear total RNA was denatured by incubation with equal volume of 95%-formamide containing dyes at 80°C for 10 min and loaded in a 15% polyacrylamide/8 M urea/1X TBE gel with 1.5 hrs pre-run. For tRNA detection, ~10% of total RNA from each fraction was applied. Gelelectrophoresis was run at 250 V for an hour and the gel was stained with EtBr. RNA was electrophoretically transferred onto a nylon membrane (BrightStar®-Plus, Ambion) in 0.5X TBE. The membrane was irradiated to fix the RNA with 250 mJ energy in a UV crosslinker (Stratalinker® UV Crosslinker, Stratagene). Probe DNA oligomer sequences 5′-GAG GCC AAT ATT TCT GTG CTG CTA-3′ (miR-195); 5′-AGT CAG TTT TGC ATG GAT TTG CAC A-3′ (miR-19b); 5′-CTG ATG CTC TAC CGA CTG AGC TAT CCG GGC-3′ (tRNALys); and 5′-ACG AAT TTG CGT GTC ATC CTT GCG-3′ (U6) were used. Probe labeling, hybridization and detection of radioactive signals were performed as previously described.33 The data image was obtained by PhosphorImager.
The authors would like to thank Thomas J. Bedor for technical assistance. This research was supported, in part, by ARRA NIH Grant R01 DK081865-01 (to C. J. Steer).
Previously published online: www.landesbioscience.com/journals/rnabiology/article/13215