LIN-28 and the poly(U) polymerase PUP-2 regulate let-7 microRNA processing in Caenorhabditis elegans.

The let-7 microRNA (miRNA) is an ultraconserved regulator of stem cell differentiation and developmental timing and a candidate tumor suppressor. Here we show that LIN-28 and the poly(U) polymerase PUP-2 regulate let-7 processing in Caenorhabditis elegans. We demonstrate that lin-28 is necessary and sufficient to block let-7 activity in vivo; LIN-28 directly binds let-7 pre-miRNA to prevent Dicer processing. Moreover, we have identified a poly(U) polymerase, PUP-2, which regulates the stability of LIN-28-blockaded let-7 pre-miRNA and contributes to LIN-28-dependent regulation of let-7 during development. We show that PUP-2 and LIN-28 interact directly, and that LIN-28 stimulates uridylation of let-7 pre-miRNA by PUP-2 in vitro. Our results demonstrate that LIN-28 and let-7 form an ancient regulatory switch, conserved from nematodes to humans, and provide insight into the mechanism of LIN-28 action in vivo. Uridylation by a PUP-2 ortholog might regulate let-7 and additional miRNAs in other species. Given the roles of Lin28 and let-7 in stem cell and cancer biology, we propose that such poly(U) polymerases are potential therapeutic targets.

Small RNAs regulate gene expression in many eukaryotes including plants, animals and fungi. miRNAs are endogenous short RNAs that modulate gene expression by blocking translation and/or destabilizing target mRNAs1,2. In animals miRNAs are transcribed as long precursors (pri-miRNAs), which are processed in the nucleus by the RNase III enzyme complex Drosha-Pasha/DGCR8 to form ~80 nt pre-miRNAs, or are derived directly from introns3. pre-miRNAs are exported from the nucleus and processed by the RNase III enzyme Dicer, and incorporated into an Argonaute-containing RNA-induced silencing complex (RISC). The first identified miRNAs, the products of the C. elegans genes lin-4 and let-7, control cell fates during larval development4. When either lin-4 or let-7 is inactivated, specific epithelial cells fail to differentiate and undergo additional divisions. lin-4 acts during early larval development, regulating lin-14 and lin-28 mRNAs4,5,6,7,8. let-7 acts during late larval development, regulating lin-41, hbl-1, daf-12 and pha-4 mRNAs9,10,11,12. As such, the time of appearance of these miRNAs must be tightly controlled. In C. elegans and other animals the expression of let-7 is developmentally regulated, but the mechanisms underlying this regulation remain unknown13. Post-transcriptional regulation of specific miRNAs has recently been uncovered14. let-7 biogenesis is blocked by Lin28 at either the Drosha15,16 or Dicer17,18 step in mammalian cell culture. Lin28 is a conserved RNA-binding protein, which in mammals controls stem cell lineages and inhibits let-7 miRNA processing in vitro15,16,17,18,19. However, the mechanism and in vivo significance of this activity are unclear.

RESULTS

An in vivo assay of let-7 miRNA function reveals developmental regulation

To study the mechanism of miRNA action in vivo, we established a quantitative miRNA reporter assay based on let-7 in C. elegans (Fig 1a,b and Supplementary Methods). We generated two transgenes comprising the promoter of myo-2, the coding sequences of either GFP or mCherry and the 3′UTR of either lin-41 or unc-54 (myo-2::gfp::lin-41 and myo-2::mcherry::unc-54; hereafter referred to as the let-7 sensor; Fig. 1a). The myo-2 promoter confers expression exclusively in the pharyngeal muscle, the food pump of C. elegans20, and lin-41 is a genetically identified target of the let-7 miRNA9, whereas the unc-54 3′UTR is not known to be regulated by any miRNA. Transgenic animals carrying an intrachromosomal array of the let-7 sensor expressed both GFP and mCherry strongly throughout larval development (Fig. 1c). As expected, a transgene expressing let-7 (myo-2::let-7) silenced GFP, but not mCherry (Fig. 1d). Surprisingly, this effect was developmentally regulated; inhibition of GFP is markedly stronger in adults than L1 larvae (Fig. 1d). As the let-7 transgene does not contain the let-7 promoter, this regulation must occur post-transcriptionally. In addition to the qualitative analysis of fluorescent protein expression using microscopy we quantified the activity of the let-7 sensor using flow cytometry of whole animals. We used a COPAS Biosort instrument to quantify GFP and mCherry expression along the body axis of thousands of individual animals at different stages during development. let-7 sensor silencing was least efficient at the L1 larval stage, and reached maximal efficiency during L3 (Supplementary Methods and Supplementary Fig. 1). This correlates with the temporal expression pattern of let-7, which begins to accumulate during the L3 stage21. Since let-7 is being driven by a promoter active at all stages, but appears active only at later larval stages, these data may reflect a mechanism that post-transcriptionally regulates let-7 during development.

Figure 1

A quantitative assay reveals post-transcriptional regulation of the let-7 miRNA by lin-28

(a,b) Schematic of the pharynx based assay of let-7 activity. (c) Fluorescence images of animals carrying the let-7 sensor transgene at L1 larval and adult stages. Both GFP and mCherry are strongly expressed.

d Fluorescence images of animals carrying both the let-7 sensor and myo-2::let-7 transgenes at L1 larval and adult stages. GFP is specifically and robustly downregulated in adults, but not in L1 larvae. Scale bar shows 20 μm.

e Fluorescence images showing that lin-28 mutants downregulate let-7 sensor GFP at the L1 stage in a myo-2::let-7 dependent fashion. This effect is not reversed in a lin-46 mutant background. Scale bar shows 20 μm.

f Fluorescence images showing that a myo-2::lin-28::unc-54 transgene is sufficient to block let-7 activity in adults carrying let-7 sensor and myo-2::let-7 transgenes. Scale bar shows 20 μm.

LIN-28 regulates pre-let-7 processing

Next we carried out forward genetic and RNAi screens to identify factors regulating let-7 activity in vivo. Knockdown of lin-28 by RNAi resulted in reduced GFP at L1 and L2 stages, in a manner dependent on the myo-2::let-7 transgene (Supplementary Fig. 2a,b and data not shown). We confirmed these results using a lin-28 loss-of-function mutant (Fig. 1e and data not shown). Mutations in lin-46 completely suppress the developmental timing defect of lin-28 mutants22, but do not restore developmental regulation of let-7 (Fig. 1e). Thus deregulation of let-7 activity in lin-28 mutants is not an indirect consequence of developmental timing defects. Other heterochronic genes, including lin-14 and lin-42 did not affect the let-7 sensor (Supplementary Fig. 2c and data not shown). Next we tested if LIN-28 was sufficient to inhibit let-7 activity. Ectopic expression of LIN-28 in the pharynx from an extrachromosomal array resulted in inhibition of let-7 in adults, which do not normally express LIN-288 (Fig. 1f). Mosaic expression of the extrachromosomal array within the pharynx indicated that LIN-28 acts cell autonomously. We concluded that LIN-28 is required and sufficient to inhibit let-7 activity in C. elegans.

We then used miRNA microarrays and northern blotting to confirm that LIN-28 regulates endogenous let-7 accumulation in L2 larvae (Supplementary Data 1; Supplementary Fig. 3)23. Further, expression of other let-7 family members was not increased in lin-28 mutants, whereas three unrelated miRNAs, including the developmentally regulated miRNA miR-85, showed increased expression in lin-28 mutant L2s (Supplementary Fig. 3a,b).

Whether Lin28 regulates let-7 processing at the Drosha15,16 or Dicer17,18 step in mammalian cells is unresolved. We addressed this in vivo in C. elegans. We used northern blotting and qRT-PCR to compare expression of let-7 and its processing intermediates from the myo-2::let-7 transgene in otherwise wild-type and lin-28 mutant L2 larvae. lin-28 mutants expressed higher levels of let-7 compared to wild type, indicating increased processing efficiency; this was accompanied by a slight reduction in the level of pre-let-7, and no change in pri-let-7 levels; these data are consistent with increased efficiency of Dicer-mediated processing (Supplementary Fig. 3c,d). We obtained similar results for endogenous let-7, although levels of pri-let-7 were decreased in lin-28 mutants, suggesting an indirect effect on the let-7 promoter (Supplementary Fig. 3e,f). Interestingly, miR-85 also appears to be regulated at the Dicer step in a lin-28-dependent fashion (Supplementary Fig. 3b,g). Consistent with these findings a functional LIN-28-GFP translational fusion is localised in the cytoplasm8 (Supplementary Fig. 4a and data not shown). Taken together, these data suggest that LIN-28 blocks Dicer-mediated processing of let-7 and possibly other developmentally regulated miRNAs.

Next, we tested whether LIN-28 directly interacts with pre-let-7. We performed pull-down assays using streptavidin beads and biotinylated pre-let-7. LIN-28-GFP from transgenic worm extracts was retained on streptavidin beads if the synthetic pre-let-7 RNA was biotinylated, but not using a non-biotinylated control (Supplementary Fig. 4b). We tested whether this interaction was direct by native gel mobility shift assay. pre-let-7 and GST-LIN-28 interact with an estimated Kd of 2 μM (Supplementary Fig. 4c and data not shown). We conclude that LIN-28 binds pre-let-7 to prevent Dicer processing. Experiments in mammalian cells suggested that the loop of the pre-let-7 hairpin is required for the interaction with Lin2815,24. However, the pre-let-7 loop is not conserved in C. elegans. We therefore tested a number of pre-let-7 loop mutants in vivo using the let-7 sensor. We found that the pre-let-7 loop is not required for the normal developmental regulation of let-7 activity (see Supplementary Methods and Supplementary Fig. 5).

PUP-2 regulates pre-let-7 processing in a lin-28 dependent fashion

Our results so far were consistent with a LIN-28 blockade of pre-let-7 processing, but we were puzzled that pre-let-7 accumulation in L2 larvae differed little in wild-type compared to lin-28 mutant animals (Supplementary Fig. 3). We reasoned that LIN-28 might target pre-let-7 for degradation. Recent work by Kim and colleagues demonstrated that Lin28 promotes pre-let-7 uridylation and subsequent degradation in mammalian cell lines, although the enzyme(s) involved are unknown18. We inspected published high-throughput sequencing data of C. elegans small RNA libraries25,26, and found frequent modification of the 3′ end of let-7* with 1 or 2 untemplated uracil residues (C. elegans let-7 resides on the 5′ arm of the hairpin; Fig. 2a). These species are likely to arise from Dicer processing of partially uridylated intermediates, and indicate in vivo uridylation of let-7. Therefore we carried out an RNAi screen against 15 potential poly(U) polymerases (Supplementary Table 1) assaying let-7 and pre-let-7 abundance in myo-2::let-7 transgenic L2 larvae. RNAi against pup-2 resulted in increased pre-let-7 levels (Fig. 2b,c; P = 7.5 × 10−5), and a small but significant increase in mature let-7 levels (Fig. 2b,d; P = 0.029). This effect is specific to pup-2, no other poly(U) polymerases, including cid-1, a potential paralogue, had this effect27 (Fig. 2c,d). These data suggest that PUP-2 uridylation targets pre-let-7 for degradation, and is required for maximally efficient blockade of let-7 processing by LIN-28. We reasoned that LIN-28 might target uridylation of pre-let-7 by PUP-2, leading to degradation of the uridylated pre-let-7 and turnover of LIN-28/pre-let-7 complexes, ensuring efficient LIN-28 function. We examined the effect of pup-2 RNAi in situations were pre-let-7 is released from LIN-28 bockade. The effect of pup-2 RNAi on pre- and mature let-7 levels is abolished at the L4 stage (Fig. 2c,d P = 1 × 10−5 and P = 0.01 respectively). Further, L2 larvae exposed to both pup-2 and lin-28 RNAi show significantly reduced accumulation of pre-let-7 (Fig. 2c P = 0.0003). These effects are not due to reduced RNAi against pup-2 (Supplementary Fig. 6a). In contrast, the effect of lin-28 RNAi on mature let-7 levels is not altered in lin-28, pup-2 double RNAi L2 larvae (Fig. 2d P = 0.2). From these data we concluded that PUP-2 post-transcriptionally regulates let-7 in a LIN-28-dependent fashion.

Figure 2

pup-2 regulates let-7 processing in a lin-28-dependent fashion

a let-7 is uridylated in vivo. Frequency of unmodified and modified let-7* molecules identified by high-throughput sequencing.

b Representative northern blot showing pup-2-dependent regulation of pre-let-7. 5 μg of total RNA from control, lin-28(RNAi), and pup-2(RNAi) myo-2::let-7 L2 larvae was loaded. U6 was used as a loading control.

c,d Quantification of relative pre-let-7 (c), and let-7 (d) abundance in lin-28(RNAi), pup-2(RNAi) and cid-1(RNAi) myo-2::let-7 L2 and L4 larvae from northern blotting experiments. Mean fold change relative to empty vector control samples is shown. P-values from Students’ t-tests indicated; n = 4. Error bars show standard error of the mean.

e Fluorescence image showing the seam cell defect observed in pup-2(RNAi) adults. A DLG-1-mCherry fusion marks seam cell boundaries. Upper panel; wild-type with continuous seam. Lower panel; pup-2(RNAi) with incompletely fused seam. Arrows indicate sites of failed fusion. Scale bar shows 20 μm.

PUP-2 contributes to LIN-28-dependent regulation of let-7 during development

Next we sought to determine if PUP-2 is required for regulation of let-7 during development. Misregulation of let-7 results in altered timing of larval development, defects in differentiation of a hypodermal stem cell lineage required for the formation of adult-specific lateral alae4, and defects in vulval morphogenesis 21. Lateral seam cells differentiate and fuse into a syncytium in wild-type adults, but this fusion is defective if pup-2 or lin-28 is knocked down, consistent with a role in regulating let-7 (Fig. 2e, Supplementary Table 2). pup-2 RNAi in a lin-28 null mutant background does not increase seam cell fusion defects suggesting this activity of pup-2 is lin-28 dependent (Table 1). Next we tested whether pup-2 genetically interacts with let-7. let-7(n2853ts) animals show reduced let-7 expression and temperature sensitive vulval bursting21. At 15°C, vulval bursting of let-7(n2853ts) animals was suppressed by pup-2 RNAi, whereas lin-28 RNAi suppressed vulval bursting at both 15°C and 20°C (Table 1). Weaker suppression of let-7(n2853ts) by pup-2 compared to lin-28 is consistent with a role for pup-2 as a lin-28 modifier. Taken together with the effect of pup-2 on let-7 processing, these data indicate pup-2 ensures efficient activity of lin-28 by targeting blockaded pre-let-7 molecules for destruction. This might occur via LIN-28 dependent uridylyl-transferase activity of PUP-2 on pre-let-7; we sought to test this hypothesis in vitro.

Table 1

Genetic interactions of pup-2 and lin-28 with let-7 in vulval development

genotype	% burst at 20°C (n)	% burst at 15°C (n)
let-7(n2853); empty vector RNAi	97 (119)	47 (86)
let-7(n2853); lin-28(RNAi)	16 (115)	0 (28)
let-7(n2853); pup-2(RNAi)	97 (120)	28 (114)

RNAi by feeding.

PUP-2 uridylates pre-let-7 in a LIN-28 dependent fashion in vitro

We expressed HA-FLAG tagged PUP-2 and SBP tagged LIN-28 in a HEK293T human embryonic kidney cell line. Immunoprecipitation of HA-FLAG-PUP2 using anti-FLAG antibodies specifically co-precipitated SBP-LIN-28 (Fig. 3a). This interaction was in absence of C. elegans pre-let-7 and likely direct. Indeed, the addition of excess exogenous pre-let-7 to the cell extract did not enhance the interaction. We also confirmed this interaction in GST pull-down experiments. In vitro translated PUP-2 directly interacts with GST-LIN-28 (Fig. 3b). PUP-2 was previously shown to polyuridylate an artificially tethered RNA in Xenopus oocytes, but was inactive without tethering27. Therefore, we tested if LIN-28 might be able to recruit PUP-2 to mediate pre-let-7 uridylation (Fig. 3c). We incubated anti-FLAG immunoprecipitates from cell extracts expressing HA-FLAG-PUP2 and/or SBP-LIN-28 with radiolabeled pre-let-7 and radiolabelled UTP. We find that HA-FLAG-PUP2 uridylated pre-let-7 only in the presence of SBP-LIN-28 (Fig. 3c). We also confirmed LIN-28 dependent uridylation of pre-let-7 by PUP-2 in vitro (Supplementary Fig. 6b). Finally, we attempted to identify in vivo uridylated pre-let-7 directly by cloning, but we were unable to do so. We conclude that rapid degradation of uridylated pre-let-7 prevents accumulation of these species in vivo, as has been postulated in human cell lines18.

Figure 3

LIN-28 interacts with PUP-2 and promotes uridylation of pre-let-7 by PUP-2

a Co-Immunoprecipitation of PUP-2 and LIN-28 expressed in HEK293T cells.

b GST pull-down assay demonstrating a direct interaction of GST-LIN-28 and PUP-2 in vitro.

c In vitro uridylation assay showing that PUP-2 uridylates pre-let-7 in a LIN-28 dependent fashion.

DISCUSSION

We have developed a quantitative assay of let-7 miRNA function in C. elegans. This assay is highly sensitive and amenable to high-throughput experiments. We have isolated new mutants in known miRNA pathway components through mutagenesis screens using this assay; analysis of novel miRNA function defective mutants should provide insights into the miRNA mechanism. In addition, this assay could be modified to study post-transcriptional regulation or target specificity of other miRNAs.

Here we demonstrate that LIN-28 regulates C. elegans pre-let-7 (see Supplementary Fig. 7a for a model). These results provide a molecular basis for the genetic link between lin-28 and let-7 in controlling developmental timing. In C. elegans this pathway determines the behaviour of epithelial stem cells. In mammals let-7 and Lin28 might regulate primordial germ cell differentiation and other stem cell lineages28. Therefore, the specific interaction of a structured RNA (pre-let-7) with a protein (Lin28) constitutes an ultraconserved switch regulating stem cell differentiation. The let-7/lin-28 switch might be as conserved as let-7 itself. For example, pre-let-7 processing is developmentally regulated in the sea urchin Strongylocentrotus purpuratus13, which also expresses a Lin28 orthologue (data not shown).

Our finding that the terminal loop of pre-let-7 is dispensable for regulation by LIN-28 is at odds with two previous studies15,16, but is consistent with competition experiments carried out by Rybak et al.17. Our approach has been to assess let-7 function in vivo, whereas previous work was based on in vitro interaction studies. All 22 nucleotides of mature let-7 are conserved in bilateria, whereas for many other miRNAs only the “seed” sequence (nucleotides 2 to 8) appears to be under evolutionary constraint. In contrast, there is little sequence similarity in the terminal loops of let-7 in different species. It is therefore tempting to speculate that nucleotides corresponding to mature let-7 contribute to LIN-28 recognition. Similar RNA-protein interactions might impose evolutionary constraint on the sequences of other ultraconserved miRNAs.

Here we show that LIN-28 recruits the poly(U) polymerase PUP-2 to uridylate C. elegans pre-let-7. We speculate that mammalian PUP-2 orthologues might similarly regulate let-7 in stem cells (Supplementary Fig. 7b). Indeed, the mouse Tut4/Zcchc11 uridylyl transferase regulates let-7 in embryonic stem cells29. let-7 is a candidate tumour suppressor21,30,31,32 and LIN28 is a potential proto-oncoprotein28,33. Therefore TUT4/ZCCHC11 might be an important novel target for anti-cancer therapy. Our data suggest that miRNAs are regulated through pre-miRNA sequestration and uridylation-dependent pre-miRNA degradation. This situation appears to be analogous to two-step regulation of the activity of proteins through sequestration and targeted degradation, for example in the case of cadherin34. Uridylation-dependent degradation of RNA has been observed previously and U tails have been shown to recruit either 5′ to 3′ or 3′ to 5′ exonucleases35,36. High-throughput sequencing suggests additional miRNAs and/or pre-miRNAs are subject to uridylation (data not shown), so regulation in this way may be widespread. Further uncovering the mechanisms underlying this pathway will be of great interest.

Supplementary Material

Supplementary Information This information is provided as a collated PDF file including Supplementary Tables and Figures, and a 2 separate XLS data files.