Nuclear Export of MicroRNA Precursors

See allHide authors and affiliations

Science  02 Jan 2004:
Vol. 303, Issue 5654, pp. 95-98
DOI: 10.1126/science.1090599


MicroRNAs (miRNAs), which function as regulators of gene expression in eukaryotes, are processed from larger transcripts by sequential action of nuclear and cytoplasmic ribonuclease III–like endonucleases. We show that Exportin-5 (Exp5) mediates efficient nuclear export of short miRNA precursors (pre-miRNAs) and that its depletion by RNA interference results in reduced miRNA levels. Exp5 binds correctly processed pre-miRNAs directly and specifically, in a Ran guanosine triphosphate–dependent manner, but interacts only weakly with extended pre-miRNAs that yield incorrect miRNAs when processed by Dicer in vitro. Thus, Exp5 is key to miRNA biogenesis and may help coordinate nuclear and cytoplasmic processing steps.

MicroRNAs (miRNAs) are ∼22-nucleotide (nt) single-stranded molecules that control post-transcriptional gene expression in eukaryotes. Several hundred distinct miRNAs exist in animals and plants (15), but few of their precise functions and mechanisms of action are known. miRNAs can silence gene expression through inactivation or degradation of mRNAs (613).

The biogenesis of miRNAs in mammalian cells involves both nuclear and cytoplasmic processing (14) catalyzed by ribonuclease III (RNase III)–like endonucleases that recognize double-stranded RNAs. First, nuclear Drosha cleaves long primary transcripts releasing 60- to 70-nt pre-miRNAs (15), which can be folded in silico into stem-loop hairpins. Then, cytoplasmic Dicer processes pre-miRNAs into ∼20- to 22-nt duplexes bearing two nucleotide single-stranded 3′ extensions (1517) (Fig. 1). Generally, only one strand of the duplex serves as the mature miRNA. Because the pre-miRNAs generated in the nucleus require further processing in the cytoplasm, nuclear export represents a critical, but hitherto uncharacterized, step in the biogenesis of miRNAs.

Fig. 1.

Pre-miRNA substrates for nuclear export and processing. (A) Pre-miR-31 has the 5′ end of mature miR-31 and 2 nt 3′ overhang. Pre-miR-31(71) and pre-miR-22(85) contain extended duplexes predicted from human genomic sequences (1). Sequences of mature miRNAs are bolded. Thick bars and brackets indicate approximate in vitro Dicer cleavage sites. (B) Products of in vitro Dicer cleavage of 32P-labeled pre-miRNA substrates. Arrow indicates ∼22-nt products. See fig. S1 for detailed analyses.

We studied three model human pre-miRNAs (Fig. 1A). Pre-miR-31 (57 nt) is the likely immediate precursor of miR-31, on the basis of sequence analyses of pre-miRNAs produced in vivo (9, 18) or in vitro (15). Pre-miR-31(71) and pre-miR-22(85) are putative precursor forms with extended stem structures (1, 16).

All three pre-miRNAs, synthesized in vitro (19), were processed by recombinant Dicer in vitro, albeit with different efficiencies, yielding 20- to 22-nt products (Fig. 1B). RNase T1 digestion of these processing products (fig. S1) showed that Dicer released RNA duplexes by cutting ∼20 to 22 nt from the bottom of the stem (Fig. 1A), consistent with cleavage patterns of perfectly double-stranded RNA substrates (20). Notably, pre-miR-31, with the miR-31 sequence at its very 5′ end, yielded the correct mature miRNA, whereas the two extended pre-miRNAs did not.

Export of pre-miRNAs from Xenopus oocyte nuclei was rapid, with >90% of the RNAs accumulating in the cytoplasm within 30 min after nuclear injection (Fig. 2A, panel 1). All RNAs appeared to use the same saturable export pathway, as they competed specifically, in a dosage-dependent manner, for export of pre-miRNA but not for export of U1ΔSm or tRNA (panels 2 to 4; fig. S2A). Thus, pre-miRNA export is carrier-mediated but unlikely to use the nuclear export receptors CRM1 or Exportin-t (Exp-t) (21).

Fig. 2.

Receptor-mediated, RanGTP-dependent pre-miRNA export. (A) A mixture of 32P-labeled pre-miR-31, pre-miR-31(71), pre-miR-22(85), and U3 and U1ΔSm RNAs was injected into Xenopus oocyte nuclei (Inj). Export was monitored at 0.5, 1, and 2 hours in the absence (panel 1) or presence (panels 2 to 4) of unlabeled competitor pre-miRNAs. (B) Export of pre-miRNAs was monitored as in (A) in oocytes depleted of RanGTP upon preinjection of RanT24N. (C) Export of 32P-labeled tRNAiMet and pre-miRNA-31 (panel 1) was saturated by coinjection of unlabeled competitor RNAs (panels 2 to 5), and restoration of export was monitored after nuclear preinjection of recombinant export receptors (panels 3 to 5). RNA export was analyzed at 0.5 and 1 hour after RNA injection.

The affinities of most substrate-specific exportins for export cargos are greatly enhanced by complex formation with Ran guanosine triphosphate (RanGTP) (21). When RanGTP was depleted as a result of inhibition of the Ran guanine nucleotide exchange factor (22), export of pre-miRNA was greatly reduced (Fig. 2B), indicating that it is mediated by a RanGTP-binding export receptor.

All three pre-miRNA species were stable in the cytoplasm for several hours and were not processed into miRNA-sized products (23). This stability and lack of detectable processing suggests that oocyte cytoplasm is deficient in Dicer activity. Likewise, no RNA fragments resembling Dicer processing products were detected in the nucleus (23). However, when confined to the nucleus, both pre-miR-31(71) and pre-miR-22(85), but not pre-miR-31, gave rise to slightly shorter degradation products (Fig. 2, A and B).

Our competition experiments indicate that pre-miRNA uses an export receptor other than CRM1 or Exp-t. Exportin-5 (Exp5) (24) is a good candidate because it exports small, structured, minihelix-containing RNAs (25) and, less efficiently, tRNAs (26). As predicted, export of 32P-pre-miR-31, which was competed by excess unlabeled pre-miRNA, was effectively restored by exogenous Exp5 but not by CRM1 or Exp-t (Fig. 2C, panels 3 to 5; fig. S2B). Thus, pre-miRNA export is specifically stimulated by Exp5.

To determine whether Exp5 can bind pre-miRNAs directly in the absence of adaptor proteins, we assayed for formation of complexes between 32P-pre-miR-31 and recombinant Exp5 (Fig. 3A). As expected for an exportin/cargo interaction, the production of a stable complex was greatly enhanced by RanGTP and was competed by unlabeled pre-miR-31 (lanes 3 to 5). Complex formation was specific for Exp5, because neither Exp-t nor CRM1 bound pre-miR-31 (lanes 6 to 9). The direct binding of Exp5 to pre-miRNAs shows that adaptor proteins are not required for complex formation, but additional proteins might still promote interactions in vivo.

Fig. 3.

Direct and specific binding of pre-miRNA-31 to Exp5. (A) Complexes formed between 32P-pre-miR-31, recombinant 2z-tagged export receptors (2 μM), and RanQ69LGTP (5 μM) were analyzed by EMSA. Unlabeled competitor pre-miR-31 (lane 5) was present at 2.5 μM. (B) EMSA of 32P-pre-miR-31 in the absence (panel 1) or presence (panels 2 to 6) of 1, 2, or 4 μM of the indicated unlabeled competitor RNAs. (C) EMSA of 32P-pre-miR-31 incubated with HeLa cell extract plus Ranq69LGTP (lanes 3 to 7). Pre-miR-31/N-His-Exp5/RanQ69LGTP complex was a mobility marker (lane 2). Fab fragments of monospecific antibodies to human Exp5 or CRM1, or unlabeled competitor pre-miR-31 or U6Δss RNAs, were included as indicated (lanes 4 to 7).

The Exp5 complex can accommodate any of the three pre-miRNAs, but with varying efficiencies (Fig. 3B). Pre-miR-31 was the strongest competitor of both complex formation (panels 2 to 4) and pre-miRNA export (Fig. 2A; fig. S2, A and B), and pre-miR-22(85) was the weakest. Unrelated small, structured RNAs competed poorly for Exp5 complex formation (Fig. 3B, panels 5 and 6).

The selectivity of the interaction between pre-miRNA and Exp5 was demonstrated in total HeLa cell extract. Two complexes containing 32P-pre-miR-31 were observed by electrophoretic mobility shift assay (EMSA) (Fig. 3C). The slower migrating complex (lane 3), which resembles the one formed with purified factors (lane 2), was super-shifted by Exp5-specific but not unrelated control antibodies (lanes 4 and 5), demonstrating that it contained Exp5. The faster migrating complex (*) likely contains the nuclear antigen La, because its formation was competed by U6 small nuclear RNA (lane 7) and NL15 RNA (23), both of which bind La strongly (27, 28). This complex does not contain Exp5 (lane 4) and probably is not involved in export. Formation of both complexes was competed by unlabeled pre-miR-31 (lane 6).

The highly specific interactions between Exp5 and pre-miRNAs implicate this exportin as an essential factor in miRNA biogenesis. If that is the case, depletion of cellular Exp5 by RNA interference (RNAi) should decrease the levels of mature miRNAs. Treatment of HeLa cells with three different Exp5-specific short interfering RNAs (siRNAs), but not control siRNAs against CRM1 or Exp-t, effectively reduced the levels of Exp5 (Fig. 4A, fig. S4). Cells depleted of Exp5 were also impaired for Exp5 function (Fig. 4B), as indicated by the nuclear accumulation of GFP-NLS-eEF1A (29). Normally, this reporter protein is exported to the cytoplasm by Exp5, using aminoacylated tRNA as an adaptor (26, 29).

Fig. 4.

Exp5-dependent production of miRNA in HeLa cells. (A) Total protein extracts from untreated, mock–treated, or siRNA transfected cells were immunoblotted with antibodies to Exp5, CRM1, or HuR. Note that the Exp-t siRNA reduces Exp-t levels by only 25% (23). (B) HeLa cell lines stably expressing GFP-NLS-eEF1A were mock treated or transfected with siRNAs and analyzed by fluorescence microscopy 48 hours later. (C) Total cellular RNAs from untreated, mock–treated, or siRNA transfected cells (52 hours) were assayed for levels of let-7a RNA and normalized to a constant amount of glyceraldehyde phosphate dehydrogenase mRNA with RNA Invader assays (Third Wave Technologies Inc., Madison, WI) (19). Error bars indicate the standard deviations of triplicate assays. Depletion of Exp5 resulted in similar reductions of let-7a RNA levels in eight independent RNAi experiments.

Upon depletion of Exp5 by RNAi for 48 to 72 hours, the levels of let-7a-1 and other mature miRNAs were reduced by 40 to 60% (Fig. 4C) (23). The residual levels of miRNA may indicate that low amounts of Exp5 are sufficient for continued pre-miRNA export (and hence miRNA production) or that mature miRNAs have long half-lives. The latter possibility is consistent with observations of comparable reductions of miRNA levels after 3 days of RNAi against Drosha or Dicer (15, 23). Thus, our results demonstrate a direct and central role of Exp5 in miRNA biogenesis and offer a possible explanation for the extensive developmental defects observed in strains of Arabidopsis thaliana carrying loss-of-function mutations in the ortholog of Exp5, HASTY (30).

RNAs that interact with Exp5 all have a high degree of double-stranded character (25) (Fig. 1). The slightly shorter pre-miRNAs that accumulated in nuclei when export was saturated (Fig. 2A) were poorly exported even when excess Exp5 was provided (fig. S2B). It is unclear whether this inefficient export results from changes in RNA secondary structure or from loss (or gain) of terminal nucleotides or phosphate groups. Variant pre-miRNA-31 with an unpaired 5′ extension was greatly impaired for export and Exp5-binding (fig. S3, B and C) (23), indicating that the precise ends of pre-miRNAs generated in vivo by Drosha (15) may contribute both to efficient export and to correct cytoplasmic processing by Dicer (Fig. 1B) (23).

The direct binding of RNAs to Exp5 allows aminoacylated tRNAs to function as export adaptors for eEF1A (Fig. 4B) (26, 29). Because tRNAs can be exported by both Exp5 and Exp-t, whereas pre-miRNAs are exported only by Exp5 (Fig. 2C), tRNA can compete for export of pre-miRNA (fig. S2C), but not vice versa (fig. S2A).

The spatial separation, and hence sequential action, of Drosha and Dicer, which are localized in the nucleus and cytoplasm of mammalian cells, respectively, appears to promote the correct and efficient processing of precursors in the generation of mature miRNAs (15). Exp5, which functions in the middle of this pathway, may facilitate miRNA biogenesis by monitoring the integrity of pre-miRNAs and by promoting efficient release of pre-miRNAs from Drosha in the nucleus, where the level of RanGTP is high. Conversely, in the cytoplasm, where the level of RanGTP is low, Exp5 would release pre-miRNAs to Dicer for further processing.

Supporting Online Material

Materials and Methods

Figs. S1 to S4


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article