Report

Inhibition of Translational Initiation by Let-7 MicroRNA in Human Cells

See allHide authors and affiliations

Science  02 Sep 2005:
Vol. 309, Issue 5740, pp. 1573-1576
DOI: 10.1126/science.1115079

Abstract

MicroRNAs (miRNAs) are ∼21-nucleotide-long RNA molecules regulating gene expression in multicellular eukaryotes. In metazoa, miRNAs act by imperfectly base-pairing with the 3′ untranslated region of target messenger RNAs (mRNAs) and repressing protein accumulation by an unknown mechanism. We demonstrate that endogenous let-7 microribonucleoproteins (miRNPs) or the tethering of Argonaute (Ago) proteins to reporter mRNAs in human cells inhibit translation initiation. M7G-cap-independent translation is not subject to repression, suggesting that miRNPs interfere with recognition of the cap. Repressed mRNAs, Ago proteins, and miRNAs were all found to accumulate in processing bodies. We propose that localization of mRNAs to these structures is a consequence of translational repression.

Initial studies of the function of lin-4 RNA during development of Caenorhabditis elegans indicated that this miRNA down-regulates protein accumulation without affecting mRNA concentrations and mRNA association with polysomes, suggesting that either translation is repressed at the step downstream of its initiation or proteins undergo synthesis but are rapidly degraded (1, 2). Subsequent work performed in Drosophila and mammalian cells has confirmed that decreased protein accumulation is not due to mRNA degradation (3-6). However, miRNAs can also affect metazoan gene expression in other ways, as indicated by the involvement of miR-16 in AU-rich element-mediated mRNA turnover (7).

To investigate how miRNAs inhibit protein accumulation in mammalian cells, we used mRNA reporters, whose translation is regulated by either the tethering of human Argonaute 2 (hAgo2) or the endogenous let-7 microribonucleoprotein (miRNP). The use of the tethering approach follows the observation that the repressive effect of miRNAs can be mimicked in HeLa cells by tethering of Ago proteins, established components of miRNPs, to the mRNA reporter (8). For assaying let-7 effects, we constructed Renilla reniformis (RL) and firefly (FL) luciferase reporters containing in the 3′ untranslated region (3′-UTR) either one (1xBulge) or three (3xBulge) sequences able to form bulged duplexes with the let-7 RNA; RL-Perf and FL-Perf contain a single site perfectly complementary to let-7 RNA (Fig. 1A and fig. S1A). We found that expression of RL-3xBulge and RL-Perf was inhibited up to 10-fold when compared with control RL mRNA (RL-Con) (fig. S1B). The inhibition was largely eliminated upon co-transfection of a 2′-O-Me oligonucleotide complementary to let-7 but not to control miR-122a RNA, consistent with the effect being mediated by let-7 miRNP (fig. S1C). The amount of RL-Perf mRNA was decreased about fivefold, whereas that of RL-3xBulge mRNA was reduced by only 20%; expression of RL-1xBulge mRNA was not decreased when compared with RL-Con mRNA (fig. S1). The data are in agreement with the findings that endogenous miRNAs can cleave an RNA containing a single perfectly complementary site and that translational repression generally requires several miRNA target sites (6).

Fig. 1.

The miRNA-mediated repression targets translational initiation. (A) Schematic representation of mRNA reporters used for most experiments. (B) Targeting of the protein to ER does not interfere with repression. Cells were transfected with RL reporters, which express RL that is either cytoplasmic (HA-RL) or targeted to the ER lumen (ER-HA-RL). Activity of reporters regulated by tethered NHA-hAgo2 (but not HA-hAgo2 or NHA-LacZ) and endogenous let-7 is shown in the left and right, respectively. (C and D) Polysomal distribution of RL mRNAs in extracts prepared from cells transfected with either pRL-Con (C) or pRL-3xBulge (D). RNA extracted from individual fractions was analyzed with probes specific for RL and endogenous β-actin mRNAs. (E) Quantification of mRNA distribution, expressed as a percentage of total radioactivity present in each fraction. (F and G) The shift of RL-3xBulge mRNA to the top of the gradient is specifically eliminated upon co-transfection of anti-let-7 oligonucleotide. (F) Transfected reporters and oligonucleotides are indicated. The bottom row represents analysis of RL-3xBulgeMut, containing nonfunctional let-7 sites. (G) Quantification of RL-3xBulge mRNA distribution in the presence of 2′-O-Me oligonucleotides.

We investigated whether let-7 miRNP or tethered hAgo2 inhibits the translation process per se or induces the proteolysis of nascent polypeptides. Toward this we tested the effect of targeting the reporter RL proteins to the endoplasmic reticulum (ER), arguing that co-translational insertion of the signal sequence-containing, hemagglutinin-tagged RL (ER-HA-RL) to the ER should make nascent proteins inaccessible to proteolysis. Inhibition of ER-HA-RL expression was similar to that of HA-RL by both let-7 and the hAgo2 tethering (Fig. 1B). Western analysis performed with purified ER fractions indicated that, as expected, the ER-HA-RL protein was targeted into the ER lumen (fig. S2). We also tested the effect of the proteasome inhibitors MG132 and Z-Leu-Leu-Leu-al. Neither relieved the repression (9). Together, these results suggest that translation itself is the target of the inhibition.

To determine which translation step is inhibited by miRNAs, we analyzed polysome profiles of reporter mRNAs undergoing repression by let-7 (Fig. 1, C to G) or tethered hAgo2 (fig. S3A). In both cases the repression was accompanied by a strong shift of reporter mRNAs toward the top of the gradient, similar to that seen upon addition of NaF or harringtonine, known inhibitors of translation initiation (fig. S3, B and C). The shift was largely eliminated when cells were co-transfected with anti-let-7 but not control oligonucleotide or when RL-3xBulgeMut [this RNA contains nonfunctional let-7 sites (fig. S9B)] was used as a reporter (Fig. 1, F and G). The distribution of endogenous β-actin mRNA remained unchanged, arguing for a specificity of the effect. These results suggest that miRNAs in mammalian cells inhibit the loading of mRNA to polysomes, possibly resulting from a block in initiation.

To investigate the mechanism of miRNA repression in more detail, we synthesized RNA reporters bearing different 3′-UTRs in vitro and studied their expression in transfected HeLa cells. We found that activity of the m7GpppN-capped RL-Perf or RL-3xBulge reporters, with or without a polyadenylate [poly(A)] tail, was about 10-fold lower than that of control RL RNAs (Fig. 2A). Similar results were obtained with FL reporters (fig. S4A). Co-transfection of the anti-let-7 but not of control oligonucleotide markedly increased the activity of 3xBulge RNAs, indicating that the effect was specific (fig. S4, B and C). When tested for translation in reticulocyte or wheat germ extracts [these extracts do not recapitulate the let-7-mediated repression (10)], the FL- and RL-3xBulge, FL- and RL-Perf, and FL- and RL-Con RNAs showed similar activity (fig. S4, D to G). We conclude that the let-7-mediated repression of capped mRNAs can be reproduced in HeLa cells transfected with RNA and that a poly(A) tail is not required for the repression.

Fig. 2.

Cap-dependent but not cap-independent translation is repressed by let-7. The values represent means of three transfections ± SD. (A) Translation of the in vitro-transcribed capped RL-RNA reporters, either poly(A)+ (200 ng) or poly(A)- (400 ng), in transfected HeLa cells. (B) IRES-mediated translation is immune to repression by let-7. HeLa cells were transfected with 200 ng of the indicated noncapped EMCV-FL or -RL RNAs or capped FL RNA. Oligonucleotides were included as indicated. (C) Translation driven by the tethered initiation factor eIF4E or eIF4GI is immune to let-7 repression. Cells were transfected with 100 and 500 ng of plasmids expressing dicistronic reporters and indicated NHA fusions. Activities in co-transfections performed with pFL-2BoxB-RL-Con and pNHA-eIF4E were set to 1.

We examined whether translation initiated at the internal ribosome entry site (IRES) is subject to the repression. Translation starting at the IRES is not dependent on the 5′-terminal m7G cap, and its requirements for initiation factors differ from the cap-initiated reaction (11). In contrast to cap-dependent translation, the activity of transfected RNAs containing the IRES of the encephalomyocarditis virus (EMCV), EMCV-RL, and EMCV-FL, or the IRES of the hepatitis C virus (HCV-FL) was not down-regulated by the insertion of let-7 sites. Moreover, co-transfection of the anti-let-7 oligonucleotide had no effect on the activity of IRES-containing mRNAs bearing let-7 sites (Fig. 2B and fig. S5A). We also compared the effect of let-7 site insertion on EMCV-IRES- and cap-mediated translation by co-transfecting both classes of RNA. For both the RL and FL reporters, only activity of the capped RNA was inhibited (fig. S5, B to H). These results argue against the possibility that resistance of IRES-containing mRNAs to the inhibition is due to saturation of the available let-7 miRNP.

Resistance of IRES-mediated translation to let-7 inhibition implied that the miRNA machinery targets an early step in protein synthesis, a step required for the cap- but not the IRES-dependent initiation. To substantiate this possibility, we constructed dicistronic reporters in which translation of the downstream RL cistron is driven by the initiation factors eIF4E or eIF4G directly tethered at the intercistronic region via either two or six BoxB hairpins (Fig. 2C and fig. S6). For tethering, mutant forms of both factors were expressed as fusions with the N peptide, which binds to BoxB hairpins. Prior experiments (12) have indicated that tethering of either factor promotes cap-independent translation. Consistently, we found that tethering of eIF4E or eIF4G, but not of lacZ used as a control, promoted translation of the downstream cistron without having any effect on the cap-dependent synthesis of FL. Most importantly, insertion of three let-7 sites down-regulated synthesis of FL by about three-fold (in a let-7-RNA-dependent process) (fig. S6A) but had no effect on a cap-independent translation started internally in response to eIF4E or eIF4G tethering (Fig. 2C and fig. S6). These data suggest that the miRNA machinery interferes with the step upstream of eIF4E recruitment of eIF4G during initiation, possibly with the recognition of the m7G cap by eIF4E.

We analyzed the intracellular localization of the miRNP components. Previous work has indicated that translationally repressed mRNAs can be sequestered in specific cellular structures such as germline (13) or stress (14) granules. Immunofluorescence analysis indicated that HA-tagged hAgo2-4 proteins, expressed in transfected HeLa and human embryonic kidney (HEK) 293 cells (fig. S7) or in stable HeLa cell lines (figs. S8, A to C, and S9A), localize to processing bodies (PBs; visualized as structures enriched in the marker proteins Lsm1, Dcp1a, and Xrn1), suborganelles identified as sites of mRNA degradation (15). Association of Ago proteins with PBs was further confirmed by co-immunoprecipitation experiments (fig. S8, D and E). Notably, the GFP-tagged Dcp1 was found to interact with HA-hAgo2 but not HA-hAgo2ΔPRP, a mutant of hAgo2 that does not function as a repressor (8) (fig. S8E).

Next, we analyzed the cellular distribution of reporter mRNAs by in situ hybridization in HeLa cells. The RL-3xBulge mRNA but not its mutant version, RL-3xBulgeMut, co-localized with the PB marker Dcp1a, expressed as a fusion with green fluorescent protein (GFP) (Fig. 3A and table S1). The co-localization of RL-3xBulge mRNA with PBs was let-7-dependent (table S2). Moreover, RL-3xBulgeMut RNA accumulated in PBs when co-transfected with the siRNA-like duplex, let-7Mut, which carries mutations restoring base pairing between let-7 and RL-3xBulgeMut RNAs (Fig. 3A and fig. S10). The RL-Perf mRNA could not be visualized in PBs, arguing against a possibility that RL-3xBulge or RL-3xBulgeMut signals in PBs originate from an mRNA targeted for degradation (fig. S11). Closer examination of the images provided additional evidence that RL-3xBulge RNA localization is not related to the degradative function of PBs. RL-3xBulge RNA was often found adjacent to Dcp1 foci and not overlapping with them (Fig. 3A, fig. S11, and table S3). Furthermore, we also observed that some Dcp1 foci did not contain detectable amounts of RL-3xBulge RNA and, conversely, that a few mRNA foci were negative for Dcp1 (9). These data point to some heterogeneity and/or compartmentalization of PBs.

Fig. 3.

Localization of translationally repressed mRNA and miRNAs to discrete foci adjacent to or overlapping with PBs. Insets represent enlargements of indicated regions, representative of localization of RL reporters, miRNAs, and Dcp1a. (A) HeLa cells were co-transfected with the indicated RL reporters and a plasmid expressing GFP-Dcp1a. Cells shown in a lowest row we additionally co-transfected with the let-7Mut duplex. RL mRNAs were detected by in situ hybridization with Cy3-conjugated probes (red), and Dcp1a was visualized by GFP fluorescence (green). The nucleus was stained with 4′,6′-diamidino-2-phenylindole (DAPI, blue). (B) Enrichment of endogenous miRNAs let-7 and miR-122a in dot-like structures. The indicated cells were stained with digoxigenin-labeled complementary probes. (C) Accumulation of let-7 RNA in foci adjacent to or overlapping with PBs. Nuclei of HeLa cells were microinjected with an in vitro-transcribed Cy3-labeled RNA (red), which self-cleaves to produce authentic pre-let-7 RNA (26). Cells were counterstained with anti-Dcp1a antibody to locate PBs (green). The Cy3 staining of the nucleus likely originates from the 5′ cleavage fragment of the ribozyme.

We used HeLa cells stably expressing either RL-3xBulge or RL-3xBulgeMut reporters (fig. S9, B and C) to further document the association of repressed mRNAs with cellular structures. The cells were permeabilized with digitonine, and the resulting S14 supernatant and pellet fractions were analyzed for the presence of mRNAs and PB and miRNP components. The RL-3xBulgeMut RNA was enriched in a soluble fraction, but the RL-3xBulge reporter was found almost exclusively in the pellet fraction containing cellular structures. Likewise, endogenous N-Ras and K-Ras mRNAs, recently established targets of let-7 RNA (16), were enriched in the pellet as were all tested PB and miRNP components (fig. S12). Importantly, treatment of cells with anti-let-7 oligonucleotide specifically released a substantial fraction of the pellet-associated mRNAs to the supernatant (fig. S12A).

Lastly, we investigated whether miRNAs localize to specific cellular structures. The let-7 and miR-122a probes stained dot-like structures in HeLa and human hepatoma Huh7 cells, respectively, but not in control cells known not to express these miRNAs (16, 17) (Fig. 3B). No staining was observed with the mutant or pre-miRNA-specific probes (fig. S13). Because the in situ hybridization was not compatible with the antibody labeling of PBs, we microinjected the in vitro-transcribed, Cy3-labeled pre-let-7 RNA into nuclei of HeLa cells and analyzed its distribution. Remarkably, exported let-7 RNA accumulated in PB foci also labeled with antibodies against Dcp1 (Fig. 3C). The presence of let-7 in the cytoplasm was inhibited by wheat germ agglutinin, indicating that export of pre-miRNA from the nucleus occurred by a physiological mechanism (fig. S14). As with the RL-3xBulge RNA foci, let-7 foci were frequently adjacent to, rather than exactly co-localizing with, Dcp1 foci. Some let-7 foci not labeled by Dcp1, and vice versa, were also observed (Fig. 3C and tables S1 and S3).

Our data indicate that, in mammalian cells, let-7 miRNP inhibits translation at the initiation step. The observation that the cap-independent translation is immune to the repression suggests that miRNPs target an early step of initiation, likely involving the m7G cap. Importantly, the results of experiments involving an entirely independent methodology—a direct tethering of hAgo2 to mRNA reporters, an approach that mimics the miRNP-mediated inhibition (8)—are consistent with the above interpretation. Inhibition of initiation by factors binding to the mRNA 3′-UTR is a common theme in translational regulation. Such inhibition may involve interference either with the recruitment of eIF4E to the m7G cap or with the eIF4E-eIF4G interaction (18-20). Our findings that the polysomal status of mRNAs repressed by let-7 in mammalian cells differs from that of mRNAs repressed by lin-4 in C. elegans (1, 2) suggest that the inhibition of productive translation by different miRNAs, or in different organisms, can follow diverse routes.

After initial submission of this work, other reports implicating PBs in miRNA-mediated repression have appeared (21, 22). Our data extend these findings by directly demonstrating that miRNAs and repressed mRNAs localize to PBs. We believe that relocalization of the repressed mRNA to PBs is a consequence rather than a cause of the repression. However, the relocalization could contribute to the repression by maintaining the mRNA in an environment unfavorable for translation (15, 23, 24). Inhibition of translation initiation leads to the appearance of stress granules and the concomitant recruitment of repressed mRNAs to these structures (24). PBs were originally identified as sites of mRNA degradation (15), but a role in mRNA storage has recently been suggested (23). Our results provide experimental evidence that PBs could act as a storage compartment for translationally repressed mRNAs. The observation that the repressed reporter mRNA, and also let-7 RNA, do not always precisely co-localize with a PB marker protein, Dcp1p, suggests that PBs may comprise two subcompartments, one dedicated to the storage of mRNA and the other to its degradation. Close proximity of the compartments might explain why mRNAs subjected to miRNP-mediated repression may also undergo limited degradation (25).

Supporting Online Material

www.sciencemag.org/cgi/content/full/1115079/DC1

Materials and Methods

Figs S1 to S14

Tables S1 to S3

References

References and Notes

View Abstract

Navigate This Article