Repression of C. elegans microRNA targets at the initiation level of translation requires GW182 proteins

Translational blockade of endogenous let-7 target genes

We recently observed widespread genetic interaction between let-7 and the translational machinery in C. elegans (Ding et al, 2008). These findings prompted us to examine whether let-7 regulates its targets translationally in vivo. To this end, we fractionated whole animal lysates by sucrose density gradient ultracentrifugation to analyse the polyribosome association of endogenous let-7 targets in wild-type and let-7(n2853) mutant C. elegans at the L3 developmental stage, when let-7 activity is low, and at the late L4 stage, when let-7 activity is high (Reinhart et al, 2000) (Figure 1A; Supplementary Figure S1). As the two let-7 targets daf-12 and lin-41 (Slack et al, 2000; Grosshans et al, 2005) are expressed at very low levels in L4 stage larvae (Snow and Larsen, 2000; Bagga et al, 2005 and this study, below), we used reverse transcription–quantitative PCR (RT–qPCR) to quantify them. It is to be noted that all experiments were performed using random hexamer oligonucleotides to prime RT, to include even mRNA, the poly(A) tail of which might be short due to the action of the miRNA (Eulalio et al, 2008a; Filipowicz et al, 2008). Additional control experiments, described below, further confirmed that we are detecting full-length mRNAs rather than partially stable degradation fragments.

We found that both lin-41 and daf-12 mRNAs were moderately, but consistently depleted from the highly translated polysomal fractions in wild-type relative to let-7 mutant animals at the late L4 stage (Figure 1B and C; Supplementary Figure S1), in agreement with decreased translation initiation (Eulalio et al, 2008a). By contrast, ama-1 and act-1 mRNAs, which are not targeted by let-7, displayed similar translational profiles in both strains (Figure 1B and C; Supplementary Figure S1).

L3 stage animals express little or no let-7 (Reinhart et al, 2000); accordingly, we see no difference when comparing polysomal association of daf-12 and lin-41 mRNAs between let-7 mutant and wild-type animals at this stage (Figure 1B and C; Supplementary Figure S1). Moreover, in wild-type animals, polysome association of daf-12 and lin-41 is decreased at L4 compared with L3 stage, consistent with the establishment of an inhibitory mechanism affecting translation initiation as let-7 expression starts. A more moderate decrease of polysome association is also seen when performing this comparison for let-7 mutant animals, suggesting that the n2853 allele may provide residual let-7 activity or that alternative mechanisms, possibly the let-7 sister miRNAs mir-48, mir-84 and mir-241 (Abbott et al, 2005; Li et al, 2005), may contribute.

To exclude the possibility that the RNA that we analysed in our sucrose gradients was not representative for the total pool of cellular RNA, we performed the following control experiments. We used TRIzol to extract total RNA directly from ground worms, from the cleared lysate used for sucrose gradient centrifugation and from the pellet left behind upon lysate clearing. We found, first, that ∼90% of the RNA is in the lysate supernatant and will thus be loaded on the sucrose gradient. Second, composition of RNA in the supernatant and pellet is comparable, neither let-7 target nor control mRNAs are preferentially enriched in, or depleted from, the supernatant relative to RNA retained in the pellet (Supplementary Figure S2). Finally, although increasing sucrose concentration in total lysates decreased the yield of extracted RNA, RNA composition was largely unaffected (Supplementary Figure S2). To ensure comparable recovery from each fraction and greatest possible reproducibility, we equalized sucrose concentration in all fractions to 30% (w/v) prior to RNA extractions in all our experiments. This set of control experiments confirms that any results that we obtained in our analysis can be considered representative of the total pool of cellular RNA.

To determine further whether the fast-sedimenting mRNA was indeed associated with polyribosomes, we treated lysates with EDTA and observed that all four mRNAs were shifted to the top of gradients. Distributions became indistinguishable for late L4 wild-type and let-7(n2853) animals (Figure 1; Supplementary Figure S1). As EDTA also disrupts non-ribosomal ribonucleoprotein complexes, we further used puromycin to disassemble specifically polysomes by inducing premature termination of the elongating peptide chains. Puromycin treatment of extracts collapsed polysomes and shifted the mRNAs deeper into the gradient (Supplementary Figure S3), possibly reflecting aggregation of the mRNA and not further pursued by us. The resulting sedimentation patterns were indistinguishable for wild-type and let-7 animals and occurred for let-7 target as well as control mRNAs. The coincident loss of polysomes and shift of mRNAs demonstrates that our assay examines mRNAs associated with translation-competent ribosomes.

We conclude from these data that let-7 depletes its targets lin-41 and daf-12 from bona fide polysomes, consistent with blocking translation initiation on these mRNAs.

Translational repression requires let-7 complementary sites in the lin-41 3′UTR

The lin-41 3′ untranslated region (3′UTR) is necessary and sufficient to confer let-7-mediated regulation on an unrelated reporter gene (Slack et al, 2000). To verify that let-7 impaired lin-41 translation by binding to the lin-41 3′UTR, we employed transgenic animals expressing a lacZ reporter gene fused to the lin-41 3′UTR or a mutant variant thereof lacking the let-7-binding sites (Figure 2A). We expressed the transgene from the col-10 promoter to accumulate it specifically in the seam cells, where let-7 mediates lin-41 repression (Slack et al, 2000; Johnson et al, 2003).

Consistent with inhibition of translation initiation, we observed that only 40% of the reporter mRNA was associated with polysomes in wild-type animals, whereas this level reached almost 70% in let-7 mutant animals. Deletion of the validated let-7-binding sites in the reporter 3′UTR (Vella et al, 2004) relieved translational repression to the same extent in wild-type animals (Figure 2B and C). Consequently, let-7 mutation or deletion of its binding sites increased the average number of ribosomes per lacZ mRNA by more than two-fold relative to the wild-type situation, whereas the average number of ribosomes per act-1 mRNA stayed constant (Figure 2D; see Materials and methods for details on the calculation). This result shows that the interaction between let-7 and its binding sites in the lin-41 3′UTR mediates significant translational repression of the target mRNAs. This is confirmed by our finding that loss of let-7 regulation causes a ⩾5-fold derepression of the lacZ reporter (Supplementary Figure S4), although mRNA levels change less than two-fold (see below).

mRNA degradation does not correlate with translational repression

Bagga et al (2005) observed dramatic reduction of target mRNA levels in the presence of their cognate miRNAs in C. elegans. By RT–qPCR, we determined mRNA levels of the let-7 targets in total RNA that we prepared from aliquots of the same whole animal lysates that were used for the polysome profile experiments (Figure 3A). At the L3 stage, lin-41 and daf-12 mRNA levels are similar in wild-type and let-7(n2853) animals. However, at the late L4 stage, lin-41 mRNA is six-fold and daf-12 two-fold more abundant in let-7(n2853) relative to wild-type animals. Similar ratios were obtained when summing up the amounts of these mRNAs across all fractions of the sucrose gradients, further confirming that RNA extracted from the gradients is representative of total cellular full-length mRNAs. It is to be noted that the levels of lin-41 and daf-12 mRNAs are reduced by two-fold even in the let-7 mutant animals between the L3 and L4 stages.

For lin-41, our results are in agreement with those seen by Bagga et al (2005), and northern blot analysis of total RNA using a probe against lin-41 identified a single band, the intensity of which mirrored the signal obtained by RT–qPCR in the same backgrounds (Figure 3B). Although lin-41 mRNA levels in individual sucrose gradient fractions were below the limit of detection by northern blot analysis, these results essentially exclude the possibility that accumulation of lin-41 mRNA degradation products could bias our RT–qPCR results and confirm that we reliably quantify full-length mRNAs.

For daf-12 mRNA, its low abundance prevents detection by northern blotting even in unfractionated, total RNA without prior selection of polyadenylated mRNA (Snow and Larsen, 2000). Therefore, to confirm that our RT–qPCR assay similarly measures the levels of full-length daf-12 mRNA, we tested a second set of primers, and obtained comparable results as expected (Supplementary Figure S5A). Finally, we examined the expression levels of both daf-12 and lin-41 using cDNA obtained through oligo-dT-primed RT. Again, we obtained comparable results (Supplementary Figure S5B and C), arguing against the detection of a stable degradation product and suggesting that any residual poly(A) tail on these mRNAs is sufficient to support priming through oligo-dT oligonucleotides. In summary, we confirm by several independent methods that our assays quantify full-length mRNAs, and we find that the daf-12 and lin-41 mRNAs are not only translationally repressed by let-7 but also subject to degradation.

Translational repression of daf-12 is at least equal to that of lin-41 but the decrease of daf-12 mRNA levels is more modest, suggesting that translational repression and transcript degradation may not be directly linked. Indeed, although the lacZ∷lin-41 reporter mRNA is very efficiently repressed translationally, mRNA levels differed by less than two-fold in wild-type relative to let-7(n2853) animals (Figure 3C). Although these findings strongly argue against a scenario where lower abundance of an mRNA diminishes its access to the translational machinery, we wished to exclude the possibility further that the translational effects that we observed were due to altered mRNA levels. ugt-63 and vit-1 are differentially expressed in synchronized late L4 wild-type and let-7(n2853) animals but are not direct targets of let-7 (B Hurschler and HG, unpublished data). Although vit-1 was four-fold less abundant in let-7(n2853) than in wild-type animals, and ugt-63 was two-fold more abundant, the translational profiles of both genes were similar in wild-type and let-7(n2853) (Supplementary Figure S6). Thus, altered mRNA levels per se do not appear to influence the efficiency of translation initiation.

Multiple miRNAs function by preventing translation initiation

The finding that let-7 mediates repression of translation initiation on its targets in C. elegans was unexpected, as C. elegans lin-4 was previously reported to repress these mRNAs at a step downstream of translation initiation (Olsen and Ambros, 1999; Seggerson et al, 2002). To determine whether repression of translation initiation is specific for let-7 or a more general mechanism, we tested whether lin-4 repressed translation initiation of lin-14 and lin-28, two experimentally validated targets (Wightman et al, 1993; Moss et al, 1997). lin-4 is first expressed in the mid-L1 stage and represses lin-14 by late L1/early L2 and lin-28 one stage later (Olsen and Ambros, 1999; Seggerson et al, 2002). When we compared extracts from late L2 stage wild-type and lin-4(e912) mutant animals, we observed that both mRNAs were shifted into the polysomal fraction in the mutant (Figure 4A and B). This shift is particularly pronounced for lin-28, where the effect is highly statistically significant (Figure 4A). By contrast, polysome association of the control mRNAs act-1 and ama-1 and the let-7 target daf-12 is identical in lin-4(e912) and wild-type animals (Figure 4A and B). We conclude that lin-4, similar to let-7, can repress its target at the level of translation initiation.

lin-14 and lin-28 transcript levels are increased in lin-4 mutants compared with wild-type animals, whereas daf-12, act-1 and ama-1 mRNA levels remain unchanged (Figure 4C). The observation that lin-4 induces a stronger translational blockade of lin-28 than of lin-14 and conversely a more pronounced degradation of lin-14 than of lin-28 further suggests that translational repression and target mRNA degradation are not directly linked mechanisms.

Translational repression and degradation of miRNA targets require the GW182 proteins AIN-1 and AIN-2

Having established that miRNAs mediate both target mRNA degradation and translational repression in vivo, we sought to identify the factors mediating these mechanisms. Good candidates were the GW182 homologues AIN-1 and AIN-2 (Ding et al, 2005; Zhang et al, 2007), as depletion of GW182 causes upregulation of miRNA target genes in various systems (Ding and Han, 2007). However, although mRNA degradation is readily prevented upon GW182 depletion, derepression of those miRNA targets that are not strongly regulated by degradation is typically well below that seen with AGO depletion (Behm-Ansmant et al, 2006; Eulalio et al, 2007), consistent with the proposal that GW182 proteins might enhance miRNA activity by targeting repressed mRNAs to P-bodies for degradation (Ding et al, 2005).

To determine whether depletion of the GW182 family members AIN-1 and AIN-2 permitted uncoupling of translational repression and degradation of miRNA targets, we performed polysome profile analyses on L4 stage wild-type and ain-2(RNAi); ain-1(ku322) animals and analysed various targets of multiple miRNAs: the let-7 targets daf-12 and lin-41, the lin-4 targets lin-14 and lin-28, the lsy-6 targets cog-1 (Johnston and Hobert, 2003) and hbl-1, which is targeted by mir-48, mir-84, mir-241, let-7 and lin-4 (Abrahante et al, 2003; Lin et al, 2003; Abbott et al, 2005). As predicted, depletion of AIN-1/-2 increased lin-41 and daf-12 transcript levels (Figure 5A). However, to our surprise, translational repression of both let-7 targets was also efficiently relieved (Figure 5B and C). In fact, the relief of both modes of let-7 target repression was more extensive in ain-2; ain-1 than in let-7(n2853) mutant animals, possibly suggesting that remaining let-7 activity or distinct miRNAs, perhaps of the let-7 family, contribute to residual repression of lin-41 and daf-12 in let-7(n2853) animals. Consistent with this idea, mRNA levels of the two let-7 targets daf-12 and lin-41 are upregulated in miR-48 miR-241; miR-84 triple mutant animals (Supplementary Figure S7).

Translational repression was also relieved for lin-14, lin-28, cog-1 and hbl-1, although the results for lin-28 and hbl-1 narrowly missed statistical significance (lin-28, P=0.053; hbl-1, P=0.056). We also analysed genes not known to be miRNA targets (act-1, tbb-2, ama-1 and eft-2). We observed no effect on total mRNA levels and no consistent trend of translational upregulation in response to AIN-1/-2 depletion (Figure 5). Low abundance of the investigated miRNA target mRNAs in late L4 wild-type animals (see Figure 3B) prevented us from performing northern blot analysis on polysome profile fractions. However, consistent results were obtained by RT–qPCR with multiple lin-14 primer pairs (Supplementary Figure S8) and by semiquantitative classical RT–PCR (Supplementary Figure S9) confirming our observation that translational repression of miRNA target is relieved in AIN-1/-2 depleted animals.

Taken together, these data reveal that translational control is a mechanism that is widely used by miRNAs in vivo. Equally significant, our results show that AIN-1/-2 have a general and important function in the process. Notably, although transcript levels of lin-14, lin-28 and hbl-1 increased in ain-2; ain-1 mutant relative to wild-type animals, cog-1 mRNA levels remained unchanged (Figure 5C), demonstrating that translational repression can occur independently of target mRNA degradation.

C. elegans strains and RNAi

Wild-type N2, MT7626: let-7(n2853), MH2385: ain-1(ku322) and DR721: lin-4(e912) strains were provided by the CGC; CT5a: N2;Is[goa-1∷gfp; col-10∷lacZ∷lin-41] (Caudy et al, 2003) by R Plasterk; VT1066: mir-48 mir-241(nDf51); mir-84(n4037) (Abbott et al, 2005) by V Ambros. The HW211: let-7(n2853);Is[goa-1∷gfp; col-10∷lacZ∷lin-41] strain was obtained by crossing CT5a with MT7626. The HW390: N2;xeIs11[rol-6(su1006); col-10∷lacZ∷lin-41ΔLCS 1-3] strain was generated by genomic integration of an extrachromosomal array made of pFS1031 and rol-6(su1006) (Vella et al, 2004), followed by several rounds of backcrossing to N2. As ain-2(tm1863); ain-1(ku322) (Zhang et al, 2007) double mutant animals grow very poorly, we exposed ain-1(ku322) animals to ain-2(RNAi), starting with L1 larvae and using a published RNAi feeding construct and protocol (Kamath et al, 2003; Grosshans et al, 2005). Enhancement of the alae defect phenotype from 21% penetrant in the ain-1(ku322) single mutant to 100% in the double mutant confirmed efficient ain-2 knockdown (n⩾17 each). To circumvent reduced brood size associated with the lin-4(e912) allele and obtain a sufficiently large synchronized population, lin-4(e912) animals were expanded at 20°C on lin-14(RNAi). Following synchronization by hatching in M9 buffer, animals were shifted back to standard non-RNAi food (OP50) and grown to the desired stage for extract preparation. The reappearance of the lin-4(e912) phenotypes (long, egg-laying defective) on these animals excluded the possibility that lin-14(RNAi) had prolonged effects.

Polysome profile analysis

A detailed description can be found in the Supplementary data. Briefly, lysates of synchronized worms were layered on linear sucrose gradients (15–60% w/v) and centrifuged for 3 h at 39 000 r.p.m., 4°C, using a SW-40 rotor and an Optima L-80 XP Ultracentrifuge (Beckman Coulter). The gradients were fractionated in 12 fractions of equal volume while absorbance at 254 nm was recorded. The entire gradient was fractionated so that any pelleted material would be recovered with the last fraction. However, we typically found very little RNA in this fraction, suggesting that RNAs did not substantially occur in heavy particles or compartments under our experimental conditions. After adjusting sucrose concentration in each fraction to 30% (w/v), RNA was extracted using TRIzol (Invitrogen) and RNA integrity confirmed on ethidium bromide-stained agarose gels before proceeding to RT. RNA recovery was quantitative under these conditions reaching up to 89% of input RNA. See Results section for a discussion of control experiments confirming that RNA extracted from the lysate is representative of the composition of total cellular RNA.

RT–qPCR

RNA RT was performed using the ImProm-II™ Reverse Transcription System (Promega) with random hexamer primers, according to the manufacturer's recommendations using equal amounts of RNA (400 or 800 ng) for each sample to avoid saturating the RT reactions in fractions with high concentrations of RNA. For polysome profile fractions, relative transcript levels quantified by qPCR (below) were subsequently corrected for the total amount of RNA extracted from each fraction and expressed as a percentage of the total amount recovered for the gradients. Identical results were obtained using oligo-(dT)₁₅ primers and equal volumes of RNA without applying any correction, validating the method and establishing that qPCR following RT by random hexamer is unlikely to detect stable degradation fragments. This was also directly examined for lin-14 and daf-12 using distinct qPCR primer sets, which yielded comparable results to the original primers.

qPCR reactions were performed in technical duplicate using the ABsolute™ QPCR SYBR^® Green ROX Mix (Thermo Fisher Scientific), according to the manufacturer's recommendations, on an ABI Prism 7000 Sequence Detection System coupled to ABI Prism 7000 SDS 1.0 Software (Applied Biosystems). Relative transcript levels were calculated using the 2^{−Δ C′_t} method (Livak and Schmittgen, 2001). For all primer pairs (Supplementary data), amplification efficiencies were determined to be equal or superior to 1.8. Control reactions lacking either the reverse transcriptase or template mRNA confirmed specificity of the amplification reaction.

Northern blot

RNA electrophoresis and transfer were performed as described earlier (Bagga et al, 2005). UV crosslinked membranes were hybridized using ULTRAhyb hybridization buffer (Ambion) according to the manufacturer's recommendations with randomly radiolabelled probes prepared from PCR-amplified DNA (see Supplementary data for oligonucleotide sequences). Radioactive signals were detected and quantified using a Storage Phosphor Screen and a Typhoon 9400 with the Imagequant TL software (all GE Healthcare).

Calculation of average number of ribosomes per mRNA

To calculate the average number of ribosomes per mRNA, each gradient fraction was assigned an average number of ribosomes by counting the peaks of the polysome profile at 254 nm. This number was multiplied with relative amount of the mRNA detected in this specific fraction. The sum of this product over all the fractions yielded the average number of ribosomes per mRNA.

Statistical methods

All statistical significances were calculated using the paired one-tailed Student's t-test. Note that the statistical significance in this stringent test not only depends on the average and standard deviation of the data sets but also on the variation of the difference between the paired values so that error bars will not fully reflect the statistical significance obtained through this test.