MicroRNA-206 colocalizes with ribosome-rich regions in both the nucleolus and cytoplasm of rat myogenic cells

To detect and localize the very small targets represented by miRNAs, we exploited the higher specificity and hybridization efficiency of LNA probes (6–8). Fig. 1 shows typical in situ hybridization patterns obtained in a rat myogenic cell line with LNAs directed against two miRNAs, miR-206 and let-7, as well as one directed against signal recognition particle (SRP) RNA. The miR-206 and let-7 results shown here were detected in a dual hybridization experiment; similar results were obtained in experiments in which the probes were hybridized separately. It can be seen that the probe for miR-206 hybridized most extensively to nucleolar and cytoplasmic sites, with some nucleoplasmic signal, whereas the probe for let-7 gave a different pattern with a more uniform intracellular distribution overall. The hybridization pattern of SRP RNA with its LNA probe produced the same intranuclear pattern as was obtained by using peptide nucleic acid probes (SRP PNA in Fig. 1 Lower Right; also see ref. 15) and phosphodiester backbone probes (16), but hybridization to SRP RNA was often stronger using the LNA probe (Fig. 1, compare Upper Right and Lower Right). In both the miR-206 and SRP LNA hybridizations, the ratio of nucleolar:nuclear signal varied from ≈2:1 to almost 1:1, whereas in cells probed for let-7 the nucleolar signal levels never were higher than nucleoplasmic levels.

We next assayed the levels of miR-206 during myogenic differentiation. L6 myoblasts were grown in 2% horse serum to induce myotube formation (17) and subjected to in situ hybridization for miR-206 at various times. The amount of signal present in the cytoplasm, the nucleoplasm, and the nucleolus was quantified after digital images were captured. As can be seen in Fig. 2A, the amount of miR-206 in the cytoplasm increased significantly during differentiation, but the amounts in the nucleoplasm and the nucleolus remained more constant. When a scrambled sequence LNA probe was used, a low, rather constant level of signal was obtained over the course of differentiation (Fig. 2B). The observed increase in cytoplasmic miR-206 during myogenesis (Fig. 2A) is consistent with earlier results showing that total cellular levels of miR-206 increase during embryonic muscle differentiation (10) and also increase during differentiation of another myogenic cell line (13, 18).

These results demonstrated that the LNA probe hybridized as expected to miR-206 in the L6 myogenic cell culture system, but we wanted to further substantiate the specificity of miR-206 probe hybridization in the nucleolus. We considered it very unlikely that this localization pattern was due to cross-hybridization with other nucleolar RNAs because the results of BLAST (www.ncbi.nlm.nih.gov/BLAST and www.psb.ugent.be/rRNA/blastrrna.html) and Ensembl (www.ensembl.org) searches showed that the miR-206 LNA probe is not complementary to any rat ribosomal RNAs (including 28S, 18S, 5.8S, or 5S sequences) or any known expressed small nucleolar RNAs. (These searches did locate an 11-nt complementary sequence in a U1 snRNA pseudogene with no upstream promoter sequences.) However, we also tested the fidelity of the hybridization directly with experiments in L6 myoblasts in which we varied the stringency of the hybridization and then compared the ratio of nucleolar to cytoplasmic signal present in digital images. Hybridization of LNA oligos has been shown to be affected by ionic strength and temperature in a manner similar to hybridization with standard phosphodiester backbone DNA oligos (19), so if there were differences in hybridization fidelity between the nucleolar and cytoplasmic targets, we would expect to see differences in the ratio under altered hybridization stringencies. However, we observed no significant difference in the nucleolus:cytoplasmic signal ratios when hybridization was carried out at twice the ionic strength (ratios were 1:1 at both 4× SSC and 2× SSC) or when a very high-stringency wash was included (nucleolar:cytoplasmic ratios were 1:1 after washes at 1× SSC and 0.1× SSC). These results indicate that the hybridization to the detected target in each of these compartments was equally specific. Therefore, it is highly likely that the LNA probe is hybridizing to bona fide miR-206 sequence in the nucleolus.

Although the foregoing results indicated that the LNA probe was indeed hybridizing to miR-206 RNA, we did not know whether we were detecting the mature and/or the precursor forms of this RNA (3, 14). We therefore designed a LNA probe to the loop region of the precursor hairpin of miR-206 (see Materials and Methods), which is expected to hybridize to both the primary transcript (pri-miR-206) and the immediate miR-206 precursor (pre-miR-206). This loop region of various pre-miRNAs, including pre-miR-206, has been shown to be preferentially available for hybridization in microarray assays (20). Fig. 3 shows the results of an experiment in which a probe complementary to mature miR-206, labeled with cy3 (Fig. 3A), and a second probe complementary to the loop region of the precursors of miR-206, labeled with fluorescein, (Fig. 3B) were simultaneously hybridized to L6 cells. The precursor-specific probe displayed very little signal (Fig. 3B) compared with the probe for mature miR-206 (Fig. 3A), and no specific signal was observed in the nucleolus. However, the cytoplasmic levels for the precursor probe were slightly above those obtained by using the scrambled control probe (Fig. 3F), especially in the perinuclear region. We conclude that the signal we observed in the nucleolus using the probe to the mature sequence is very likely to reflect hybridization to the mature form of miR-206.

To confirm that the detected miR-206 signal indeed represented hybridization to RNA, we treated fixed cells with RNase before carrying out in situ hybridization with the miR-206 probe. Only low level of signal was observed in these cases (Fig. 3 G and H), indicating that the probe was hybridizing to RNA in both the nucleus and the cytoplasm. When cells were treated with DNase before in situ hybridization with miR-206, the signal intensity and distribution remained unchanged (data not shown).

We next examined the nucleolar localization of miR-206 at higher spatial resolution. The nucleolus has been classically defined as having three components by ultrastructural criteria (21): the fibrillar centers (FCs), where the rDNA genes are located; the dense fibrillar component (DFC), which immediately surrounds the FCs and into which the prerRNA nascent transcripts extend and initial processing events take place; and the GC, where additional prerRNA processing and ribosome assembly take place. In addition to the classically identified ribosomal components, other proteins and RNAs that are not known to be linked to the ribosome assembly pathway have been detected in the nucleolus, including SRP RNA (15, 16), nucleostemin (22), telomerase components (23–25), and various cell cycle-related proteins (26, 27).

Fibrillarin, a protein complexed with small nucleolar RNAs involved in rRNA processing (28), is commonly used to demarcate the DFC. Fig. 4 shows the typical results of experiments in which L6 myoblasts were either transfected with plasmids coding for GFP-fibrillarin (Fig. 4 A–E and K) or immunostained with antibodies to fibrillarin (Fig. 4 F–J and L) and then subjected to in situ hybridization to detect miR-206. Fig. 4A shows the hybridization pattern for miR-206 after cells were transfected with GFP-fibrillarin and then fixed (Fig. 4B), and Fig. 4F shows the hybridization pattern for miR-206 after cells were fixed and then subjected to immunostaining for fibrillarin (Fig. 4G). Comparison of the miR-206 hybridization pattern in Fig. 4 A and F shows that, after cells have been immunostained and then subjected to in situ hybridization, much of the cytoplasmic and nucleoplasmic miR-206 signal is lost, whereas the amount of nucleolar miR-206 signal stays the same or slightly increases as compared with results from the in situ hybridization after transfection with the GFP-fibrillarin plasmid. We believe that this is because the small miR-206 target molecules are washed out of the cytoplasm and nucleoplasm during the immunostaining procedure, because similar experiments with SRP LNA probes did not show reduced detection of the (15-fold) longer SRP RNA (data not shown).

Images obtained after iterative deconvolution analysis (22, 29) of the intranucleolar distribution of miR-206 with respect to GFP-fibrillarin distribution are shown in Fig. 4 C–E, and similarly the results with fibrillarin antibody are shown in Fig. 4 H–J. Both detection methods revealed similar nucleolar distribution patterns, and it can be seen that there was very little overlap miR-206 (red) and fibrillarin (green) in either case. Fig. 4 K and L shows intensity linescans across the regions in Fig. 4 D and I, revealing the extent of miR-206 overlap with GFP-fibrillarin and fibrillarin antibody, respectively. These results show that miR-206 is not appreciably present in the DFC of the nucleolus.

We then looked to see whether miR-206 was present in the FCs, using as a fiduciary marker the upstream binding factor (UBF), a protein that specifically binds the rDNA across the entire rDNA repeat (30). Fig. 5A–H shows two examples of nucleoli stained with UBF antibody (green) and hybridized with the miR-206 LNA probe (red), with the intensity linescans for the color overlays in Fig. 5 B and E shown in Fig. 5 G and H, respectively. Again, little overlap between UBF and miR-206 was observed.

We next examined the localization of miR-206 within the GC by comparing its hybridization pattern to the distribution of nucleostemin within this compartment (Fig. 5 I–P). Our earlier work with nucleostemin has shown that this protein was a good marker for this region (22). Fig. 5 I–K and L–N show deconvolved midplanes from nucleoli of two different cells. Fig. 5 O and P shows intensity distributions along the lines depicted in the pseudocolored images in Fig. 5 J and M, respectively, and represent typical results. Linescans through some nucleolar regions showed virtually complete overlap (Fig. 5P) whereas others showed a much lesser degree of overlap (Fig. 5O), indicating that miR-206 partially, but not completely, overlaps with the regions of the GC occupied by nucleostemin. Taken together, these nucleolar mapping results demonstrate that the miR-206 is primarily localized to the GC.

To learn more about the possible associations of miR-206 in the nucleolar GC, as well as the rest of the cell, we next compared the intracellular localization of miR-206 to that of 28S rRNA. We performed double in situ hybridization experiments to detect miR-206 and 28S rRNA and found that miR-206 colocalizes to a significant extent with 28S rRNA in both the nucleolus and the cytoplasm. Fig. 6 shows both whole-cell deconvolved images, and enlarged images of nucleoli, with accompanying linescans of the combined color-coded images. The yellow in the pseudocolored images (Fig. 6 B, D, H, and J), as well as the presence of overlapping maxima and minima in the linescans (even though not always the same intensity; Fig. 6 E, F, K, and L), demonstrate the substantial colocalization of these two probes. A significant overlap between miR-206 and 28S rRNA in both the nucleolus and the cytoplasm was observed in all cells, and in some regions the extent of spatial overlap was almost complete (e.g., Fig. 6K). These results demonstrate that miR-206 extensively colocalizes with 28S rRNA in both the cytoplasm and the GC of the nucleolus.

We have presented evidence that a miRNA involved in muscle development concentrates not only in the cytoplasm, but also in the nucleolus of cultured rat myogenic cells. Although nucleolar localization has not been reported for any other miRNA, it is too soon to know whether others visit the nucleolus because very little high-resolution work has been done in this relatively new field of miRNA cell biology. Indeed, while this manuscript was in preparation, endogenous siRNAs were reported to localize to the nucleolus in Arabidopsis (31), and there is evidence that microinjected siRNAs visit and sometimes concentrate in the nucleolus (32).

We found that miR-206 concentrates in the nucleolar GC, and not in the other two major compartments of the nucleolus, the FCs or the DFC. The GC is the region of the nucleolus where rRNA processing and ribosomal subunit assembly are completed and where a number of nonribosomal components, including nucleostemin, also concentrate (22, 33). Dual in situ hybridization in the present study showed that a substantial fraction of nucleolar miR-206 colocalizes with the 28S rRNA in the GC. Furthermore, a substantial portion of cytoplasmic miR-206 was observed to be colocalized with 28S rRNA. This cytoplasmic colocalization is consistent with the current hypothesis that miRNAs control gene expression by interfering with translation (1, 2, 4), and one recent study has suggested that at least some miRNAs may associate with polysomes (34) and thus directly interfere with translation at the elongation stage. Therefore, although an association of miR-206 with polysomes has yet to be examined in cell fractionation studies, our finding that miR-206 colocalizes with 28s rRNA in the cytoplasm suggests that it may be regulating target mRNAs in this way.

However, our discovery that the miR-206 is also localized in the nucleolus finds no obvious explanation in any previous or contemporary notions about miRNA biogenesis or function. As mentioned above, a recent study showed that endogenous siRNAs, which are closely related to miRNAs, localize to the nucleolus in Arabidopsis (31). It was proposed that siRNAs go to the nucleolus for modification and assembly, similar to other small RNAs that visit the nucleolus to be modified and/or assemble into RNP complexes (16, 35). At least some mammalian miRNAs are subject to A-to-I editing by adenosine deaminases that act on RNA (36, 37), and RNA editing enzymes have been found associated with the nucleolus in cultured mammalian cells (38). It has been proposed that adenosine deaminases that act on RNA compete with Drosha and thus regulate the processing of miRNAs in the nucleus (36, 39). It has also been shown that the 3′ terminal nucleotides of Arabidopsis miRNAs are subject to ribose-2′-O-methylation (40), but there are no reports so far of this modification in human miRNAs, much less any reports describing the intracellular compartment at which this might occur.

Although the above explanation for the nucleolar localization of miR-206 is reasonable, it does not explain why a substantial portion of miR-206 colocalizes with 28S rRNA in the nucleolus. It seems unlikely that miR-206 is involved in regulating transcription of the genes for rRNA because it did not colocalize with the known sites of these genes in the nucleolus, the FCs. This possibility was not trivial, because recent studies have implicated small noncoding RNAs, produced as intergenic transcripts within the rDNA repeats, in the transcription regulation of these genes (41). Furthermore, it is not likely that miR-206 is directly targeting rRNA transcripts in the way that miRNAs are predicted to target messenger RNAs, because we found no 7- or 8-nt sequence elements complementary to the 8-nt seed sequence of miR-206 in rat 28S, 5.8S, or 5S rRNAs. (The seed sequence is the region of miRNAs that is thought to hybridize most strongly to the target mRNA.) We did find one 18S site that contained 7/8 complementary nucleotides to the miR-206 seed sequence; however, there was very little (upstream) complementarity to the rest of miR-206. Nevertheless, we cannot rule out the possibility that miR-206 hybridizes with different affinity to rRNA or associates with certain nascent ribosomal subunits through interactions other than hybridization to the rRNA.

A final intriguing possibility is that miR-206 could be targeting messenger RNAs that transit through the nucleolus, examples of which have been reported (42, 43), and perhaps miR-206 even moves with these messages to the cytoplasm. Certain proteins have been proposed to associate with messenger RNAs in the nucleolus and then move as a messenger ribonucleoprotein complex to the cytoplasm for translation (44), and it is possible that certain miRNAs may do the same. Of course, an experimental demonstration of the functional significance of the nucleolar concentration of miR-206 awaits further investigation.

We are grateful to Kevin Fogarty and Larry Lifshitz (Biomedical Imaging Facility, University of Massachusetts Medical School) for help with deconvolution analysis and Faba Neumann and Jan Ellenberg (European Molecular Biology Laboratory, Heidelberg, Germany) for suggesting the sequence for the nonhybridizing control probe. This work was supported by National Institutes of Health Grant GM-60551 and National Science Foundation Grant MCB-0445841.