TUT4 in Concert with Lin28 Suppresses MicroRNA Biogenesis through Pre-MicroRNA Uridylation Inha Heo,1,2 Chirlmin Joo,1,2 Young-Kook Kim,1,2 Minju Ha,1 Mi-Jeong Yoon,1 Jun Cho,1 Kyu-Hyeon Yeom,1 Jinju Han,1 and V. Narry Kim1,* 1Creative

Research Center and School of Biological Sciences, Seoul National University, Seoul, Korea 151-742 authors contributed equally to this work *Correspondence: [email protected] DOI 10.1016/j.cell.2009.08.002 2These

SUMMARY

As key regulators in cellular functions, microRNAs (miRNAs) themselves need to be tightly controlled. Lin28, a pluripotency factor, was reported to downregulate let-7 miRNA by inducing uridylation of let-7 precursor (pre-let-7). But the enzyme responsible for the uridylation remained unknown. Here we identify a noncanonical poly (A) polymerase, TUTase4 (TUT4), as the uridylyl transferase for pre-let-7. Lin28 recruits TUT4 to pre-let-7 by recognizing a tetra-nucleotide sequence motif (GGAG) in the terminal loop. TUT4 in turn adds an oligouridine tail to the pre-let-7, which blocks Dicer processing. Other miRNAs with the same sequence motif (miR107, -143, and -200c) are regulated through the same mechanism. Knockdown of TUT4 and Lin28 reduces the level of stem cell markers, suggesting that they are required for stem cell maintenance. This study uncovers the role of TUT4 and Lin28 as specific suppressors of miRNA biogenesis, which has implications for stem cell research and cancer biology. INTRODUCTION MicroRNAs (miRNAs) are small RNAs of 22 nt that regulate target mRNAs through complementary base-pairing (Kim et al., 2009). MiRNAs are generated via two-step processing. In the nucleus, the primary transcript of miRNA gene (pri-miRNA) is cropped into the hairpin-structured precursor (pre-miRNA) by nuclear RNase III Drosha (Lee et al., 2003) and its cofactor DGCR8/Pasha (Denli et al., 2004; Gregory et al., 2004; Han et al., 2004; Landthaler et al., 2004). Pre-miRNA is exported to the cytoplasm by exportin 5 (Bohnsack et al., 2004; Lund et al., 2004; Yi et al., 2003) and gets further processed into a mature miRNA of 22 nt by another RNase III, Dicer (Bernstein et al., 2001; Grishok et al., 2001; Hutvagner et al., 2001; Ketting et al., 2001; Knight and Bass, 2001). One strand of the duplex is loaded onto the RNA-induced silencing complex (RISC) that contains

696 Cell 138, 696–708, August 21, 2009 ª2009 Elsevier Inc.

the Argonaute protein as the core component (Hammond et al., 2001; Mourelatos et al., 2002; Tabara et al., 1999). Regulation of miRNA biogenesis can be achieved at multiple levels. Some miRNAs are reported to be controlled posttranscriptionally, but the detailed mechanisms remain largely unknown. let-7 miRNA was initially discovered as a heterochronic gene during larval development in C. elegans (Grosshans et al., 2005; Meneely and Herman, 1979; Pasquinelli et al., 2000; Reinhart et al., 2000). By targeting multiple oncogenes such as RAS, HMGA2, and c-MYC, let-7 also functions as a tumor suppressor, most prominently in lung cancer (Johnson et al., 2005; Lee and Dutta, 2007; Mayr et al., 2007; Sampson et al., 2007). Reduced let-7 expression is strongly associated with increased tumorigenicity and poor patient prognosis. Previously, we and other groups reported that mature let-7 appears only after differentiation in embryonic stem (ES) cells, while the levels of pri- and prelet-7 are comparable between undifferentiated and differentiated ES cells, suggesting a posttranscriptional control in let-7 biogenesis (Suh et al., 2004; Thomson et al., 2006; Wulczyn et al., 2007). Recently the Lin28 proteins have been identified as the regulatory factors (Heo et al., 2008; Newman et al., 2008; Rybak et al., 2008; Viswanathan et al., 2008). Mammals have two Lin28 homologs, Lin28a and Lin28b, which are indistinguishable from each other in their biochemical activities (Heo et al., 2008; Viswanathan et al., 2008). Lin28a is mainly expressed in undifferentiated stem cells while Lin28b is present in stem cells and certain cancer cells (Balzer and Moss, 2007; Guo et al., 2006; Polesskaya et al., 2007; Richards et al., 2004; Yang and Moss, 2003). Lin28 has been reported to interfere with Drosha processing of pri-let-7 (Newman et al., 2008; Viswanathan et al., 2008) and Dicer processing of pre-let-7 (Heo et al., 2008; Rybak et al., 2008). Furthermore, we observed that Lin28 mediates the terminal uridylation of pre-let-7 in the cytoplasm of liver cancer cells and embryonic stem cells (Heo et al., 2008). The uridylated pre-let-7 (up-let-7) fails to be processed by Dicer and undergoes decay. Lin28 itself does not show any uridylation activity and does not contain a catalytic domain. Thus, Lin28 may recruit a ribonucleotidyl transferase, which can add ribonucleotides to the 30 end of RNA substrates in a template-independent manner (Martin and Keller, 2007; Stevenson and Norbury, 2006; Wilusz and Wilusz, 2008). Ribonucleotidyl transferases contain DNA polymerase

b-like nucleotidyl transferase domains that are highly conserved. Best known members of this family are the canonical poly (A) polymerases (PAPs), which function in the polyadenylation of pre-mRNA in the nucleus. In the past few years, a new family of PAPs has emerged. This noncanonical PAP family is also known as the Cid1 family because the cytoplasmic PAP, Cid1, was the first identified member of noncanonical PAPs in fission yeast (Wang et al., 2000). Noncanonical PAPs are conserved across eukaryotes (Stevenson and Norbury, 2006). In mammals, seven noncanonical PAPs have been predicted (Martin and Keller, 2007; Stevenson and Norbury, 2006; Wilusz and Wilusz, 2008). Some noncanonical PAPs can utilize uridine triphosphate (UTP) instead of adenine triphosphate (ATP). Therefore, they are also called poly (U) polymerases (PUPs) or terminal uridylyl transferases (TUTases or TUTs) (Kwak and Wickens, 2007; Mullen and Marzluff, 2008; Rissland et al., 2007; Wickens and Kwak, 2008). The cytoplasmic PAP, GLD2 (also known as TUTase2 and PAPD4), functions in the polyadenylation of neuronal mRNAs at synapses (Rouhana et al., 2005). In addition, mouse GLD2 induces the translation of specific dormant mRNAs by adding poly (A) tails to the mRNAs during oogenesis (Nakanishi et al., 2006). Recently, GLD2 was found to monoadenylate and stabilize mature miR-122 in mammalian liver cells (Katoh et al., 2009). TUTase1 (also known as mtPAP, PAPD1, and Hs4) was reported to adenylate mitochondrial pre-mRNAs (Nagaike et al., 2005; Tomecki et al., 2004). TUTase3 (PAPD5 and TRF4-2) and TUTase1 uridylate histone mRNAs, leading to their decay at the end of early S phase (Mullen and Marzluff, 2008). U6 TUTase (TUTase6, also known as PAPD2, Hs5, and TUT1) can add up to three uridyl groups to the 30 end of U6 snRNAs and function in the recycling of U6 snRNA during splicing (Trippe et al., 2003, 2006). Here, we report that a noncanonical PAP, TUTase4 (TUT4, also known as ZCCHC11, Hs3, and PAPD3), functions as the uridylyl transferase for pre-miRNA. Lin28 interacts with pre-let-7 through a conserved sequence motif of GGAG in the terminal loop and recruits TUT4 to the pre-miRNA. TUT4 then adds a uridine tail to the 30 end of pre-miRNA. Through this activity, TUT4 represses miRNA biogenesis in embryonic stem cells and consequently affects the maintenance of ES cells. RESULTS TUT4 Associates with pre-let-7 MicroRNA We have previously reported that Lin28 induces the terminal uridylation of pre-let-7 (Heo et al., 2008). Because Lin28 itself does not possess uridylating activity, we suspected the existence of another factor that participates in let-7 suppression. To find this enzyme, we carried out RNA affinity purification by incubating immobilized pre-let-7a-1 with cell lysates from HEK293T cells transfected with the Lin28a expression plasmid (Figure 1A). We previously showed that the Lin28a-expressing HEK293T cells can support uridylation of pre-let-7 (Heo et al., 2008), implying that the uridylating enzyme is present in HEK293T cells. The interacting proteins were identified with liquid chromatography-tandem mass spectrometry (LC-MS/ MS) (Table S1 available online). As expected, Lin28a was efficiently pulled down. In addition, pre-let-7 precipitated another

Figure 1. Identification of TUT4 as the Interactor and Suppressor of pre-let-7 (A) Proteins interacting with pre-let-7 were pulled down from Lin28a-transfected HEK293T cells. pre-let-7a-1 with a short 50 extension was transcribed in vitro. A 30 -biotinylated adaptor DNA that is complementary to the 50 extension of pre-let-7 was used to immobilize pre-let-7 on streptavidin-coated agarose beads (left panel). pre-miR-16-1 was used as a negative control. The associated proteins were separated on 10% SDS-PAGE and analyzed by LC-MS/MS (right panel and Table S1). (B) Knockdown of TUT4 results in the upregulation of mature let-7. TUT4 was knocked down in mouse embryonic stem (mES) cell line, R1, for 48 hr. Level of mature miRNAs (left) and pri-miRNAs (middle) were determined by quantitative RT-PCR (qRT-PCR). The degree of TUT4 knockdown was measured by semiquantitative RT-PCR (right). The standard errors are from two independent experiments.

protein named TUT4. TUT4 belongs to a family of noncanonical PAPs and contains nucleotidyl transferase domains and C2H2and CCHC-types of zinc finger domains. Although TUT4 has been reported to exhibit PUP activity on synthetic oligonucleotide in vitro, its natural target substrates and its cellular role as a PUP remains unknown (Kwak and Wickens, 2007). To confirm the interaction between TUT4 and pre-let-7, we repeated the RNA affinity purification and the LC-MS/MS analysis with Huh7 cells where Lin28b is expressed endogenously. TUT4 was again identified in Huh7 cells as a protein interacting with pre-let-7 (data not shown). Considering its association with pre-let-7 and its PUP activity, we hypothesized that TUT4 is the uridylyl transferase for pre-let-7 elongation.

Cell 138, 696–708, August 21, 2009 ª2009 Elsevier Inc. 697

Figure 2. Lin28 Recruits TUT4 to pre-let-7 in the Cytoplasm (A) TUT4 interacts with pre-let-7 only when Lin28a is coexpressed. FLAG-TUT4 and/or V5-Lin28a proteins were expressed in HEK293T cells, and the RNA pulldown assay was carried out using pre-miR-16-1 or pre-let-7a-1 as the baits, as in Figure 1A. Precipitated proteins were visualized by western blotting using antiTUT4, anti-Lin28a, and anti-tubulin antibodies. Tubulin was detected as a control in (A) and (B). The amounts of RNA baits used are visualized by ethidium bromide staining (lanes 7 and 8). (B) TUT4 fails to interact with pre-let-7 when Lin28a is knocked down. The mouse ES cell (A3-1) was transfected with siRNA against Lin28a, and the total cell extract was used for RNA pull-down and western blotting as in (A). (C) V5-Lin28a is coimmunoprecipitated with FLAG-TUT4 only in the presence of pre-let-7. FLAG-TUT4 and V5-Lin28a were coexpressed in HEK293T cells. Immunoprecipitation (IP) was carried out using anti-FLAG antibody-conjugated agarose beads together with in vitro transcribed pre-let-7a-1 or pre-miR-16-1 (100 pmol). Anti-FLAG and anti-V5 antibodies were used for western blotting. (D) Cytoplasmic localization of TUT4. Fractionated nuclear (Nu) and cytoplasmic (Cyto) extracts were analyzed by western blotting using anti-TUT4 antibody. The efficiency of fractionation was monitored by detecting tubulin (cytoplasmic protein) and hnRNP C (nuclear protein). (E) TUT4 binds to pre-let-7 in the cytoplasmic fraction. Cytoplasmic or nuclear extracts were prepared from HEK293T cells that had been transfected with FLAGTUT4 and V5-Lin28a. The extracts were used for RNA pull-down assay and western blotting as described in (A). The fractionation efficiency was monitored by measuring the protein levels of tubulin (cytoplasmic protein) and lamin (nuclear protein).

TUT4 Suppresses the Biogenesis of let-7 MicroRNA Posttranscriptionally Because the uridylated pre-let-7 (up-let-7) is resistant to Dicer processing and subject to decay (Heo et al., 2008), it is expected that the uridylating enzyme will have an inhibitory effect on let-7 biogenesis. To test this, we transfected a mouse embryonic stem (mES) cell line R1 with siRNA against TUT4. After 48 hr of transfection, the miRNA levels were determined by quantitative RT-PCR (qRT-PCR). The mature miRNA levels of let-7a, let-7g, and let-7f increased by 2- to 4-fold upon TUT4 knockdown, demonstrating the suppressive role of TUT4 in let-7 biogenesis (Figure 1B). On the contrary, the level of mature miR-16 did not change, suggesting that regulation by TUT4 is specific to let-7

698 Cell 138, 696–708, August 21, 2009 ª2009 Elsevier Inc.

family miRNAs. The let-7-specific change was also observed with another siRNA targeting a different position in TUT4 mRNA, excluding a possibility of off-target effects by siTUT4 (data not shown). In contrast to the significant increase of mature let-7, the levels of pri-let-7 remained unaltered in TUT4-depleted cells (Figure 1B). This suggests that TUT4 regulates let-7 biogenesis posttranscriptionally and that TUT4 does not affect Drosha processing. The knockdown of TUT4 did not cause differentiation of the stem cells in the time range of our experiment (48 hr), excluding the possibility that the observed upregulation of let-7 is a secondary effect of cell differentiation (Figure S1).

Figure 3. TUT4 Specifically Uridylates pre-let-7 in a Lin28-Dependent Manner (A) Immunoprecipitated TUT4 uridylates pre-let7a-1 in vitro in the presence of recombinant Lin28a or Lin28b. FLAG-TUT4 wild-type (WT) and catalytically dead mutant (mt) were expressed in HEK293T cells and were immunoprecipitated using anti-FLAG beads. Comparable amounts of WT and mt TUT4 proteins (Figure S3A) were used for the reaction together with 0.5 mM recombinant Lin28 (rLin28), 0.25 mM UTP, and 50 -labeled synthetic pre-let-7a-1 (left) or pre-miR16-1 (right). (B) 50 -labeled pre-let-7g and pre-miR-30a were used for in vitro uridylation assay as in (A). (C) TUT4 favors UTP over ATP, cytosine triphosphate (CTP), and guanine triphosphate (GTP). The immunopurified TUT4 protein was incubated with 50 -labeled synthetic pre-let-7a-1, 0.5 mM recombinant Lin28b, and 0.025 mM of UTP, ATP, CTP, or GTP.

Lin28 Mediates the Association between TUT4 and pre-let-7 In order to investigate the action mechanism of TUT4, we further analyzed the interaction between TUT4, Lin28, and pre-let-7. We first carried out RNA affinity purification as in Figure 1A. For this, either pre-let-7a-1 or pre-miR-16-1 was immobilized on beads and was incubated with extracts from HEK293T cells transfected with TUT4- and Lin28a-expression plasmids. We observed that TUT4 did not bind to pre-let-7 when it was expressed alone (Figure 2A, lane 3). But when Lin28 was coexpressed, TUT4 was precipitated with pre-let-7 (Figure 2A, lane 6). This suggests that TUT4 requires Lin28 to bind to pre-let-7. TUT4 did not associate with pre-miR-16 under any conditions. Lin28 can interact with pre-miR-16 to some degree but less avidly compared to pre-let-7. This weak interaction with pre-miR-16 appears to be nonspecific and nonfunctional because the inactive Lin28 mutant (Heo et al., 2008) can also bind to pre-miR-16 at a similar affinity (Figure S9B). To confirm the Lin28 dependency of TUT4, we performed a similar RNA pull-down experiment with mES cells in which both Lin28a and TUT4 are endogenously expressed (Figure 2B). Lin28a and TUT4 were precipitated with pre-let-7 (Figure 2B, lane 3). However, after knockdown of Lin28a, TUT4 failed to associate with pre-let-7 efficiently (Figure 2B, lane 6). This again demonstrates that TUT4 interacts with pre-let-7 in a Lin28dependent manner. We then asked whether Lin28 recruits TUT4 through direct protein-protein interaction. In order to test this, we transfected FLAG-TUT4 and V5-Lin28a expression plasmids into HEK293T cells, in which let-7 is transcribed at very low levels (Figure 2C). Immunoprecipitation was carried out using anti-FLAG antibody, and the precipitated proteins were detected using anti-FLAG and anti-V5 antibodies. FLAG-TUT4 could not precipitate V5Lin28, unless synthetic pre-let-7 was added into the extract (Figure 2C, lane 8–10). pre-miR-16 could not mediate the inter-

action between Lin28 and TUT4 (Figure 2C, lane 10). Thus, TUT4 specifically requires pre-let-7 to interact with Lin28. It was previously shown that purified recombinant Lin28 binds directly to pre-let-7 without any other protein cofactors (Piskounova et al., 2008). On the other hand, TUT4 can interact with prelet-7 and Lin28 only when they are present at the same time (Figure 2). These results suggest that Lin28 first recognizes pre-let-7 and, subsequently, TUT4 is recruited to the binary complex of Lin28 and pre-let-7. Association between TUT4, Lin28, and pre-let-7 Occurs in the Cytoplasm We then asked in which cellular compartment TUT4 localizes and functions. The nucleus and the cytoplasm were separated, and the protein levels were determined by western blotting (Figure 2D). TUT4 was detected mainly in the cytoplasmic fraction in both HEK293T and mES cells. The cytoplasmic localization of TUT4 was also confirmed by immunocytochemistry in Huh7 and HeLa cell lines using anti-TUT4 antibody (Figure S2). When the nuclear and the cytoplasmic extracts were incubated with immobilized pre-miRNA, we found that TUT4 binds efficiently to pre-let-7 in the cytoplasmic fraction but not in the nuclear fraction (Figure 2E). This is consistent with our previous observations that Lin28 acts mainly in the cytoplasm and that the pre-let-7 uridylating activity is detected in the cytoplasmic extract (Heo et al., 2008). Therefore, the data show that TUT4 and Lin28 act together to suppress let-7 biogenesis in the cytoplasm. TUT4 Uridylates pre-let-7 In Vitro In previous experiments, we recapitulated the uridylation reaction of pre-let-7 in vitro by incubating synthetic pre-let-7 and recombinant Lin28 protein with HEK293T cell extract (Heo et al., 2008). In order to examine whether TUT4 is indeed the factor that is responsible for the uridylating activity in the cell extract, TUT4 was immunopurified and incubated with recombinant

Cell 138, 696–708, August 21, 2009 ª2009 Elsevier Inc. 699

Figure 4. TUT4, but Not the Other Human TUTases, Can Uridylate pre-let-7 (A) Schematic representation of human TUTase proteins based on the information from InterPro database. The catalytic motif is composed of nucleotidyl transferase domain (red box) and PAP-associated domain (orange box). Hatched red box indicates a conserved nucleotidyl transferase domain that is inactive due to sequence variations. Blue and olive boxes represent C2H2- and CCHC-type zinc finger domains, respectively. Yellow box corresponds to the RRM (RNA recognition motif). (B) FLAG-tagged TUTase proteins were immunoprecipitated at comparable levels. Each FLAG-tagged TUTase protein was expressed in HEK293T cells and immunoprecipitated using anti-FLAG agarose beads. The proteins were visualized by western blotting using anti-FLAG antibody. Though the juxtaposed lanes are not contiguous, all of them are from a single gel, which is true of all the membranes with gray lines. (C) In vitro uridylation assay. The FLAG-TUTases immobilized on anti-FLAG-agarose beads were incubated with 50 -labeled pre-let-7a-1, 0.25 mM UTP, and 0.5 mM recombinant Lin28b.

Lin28 and synthetic pre-let-7a-1 (Figure 3A). The uridylated prelet-7a-1 (up-let-7a-1) band of 100 nt appeared only when both TUT4 and Lin28 were included in the reaction. This up-let-7a-1 band was similar in size to the band that had previously been detected in incubation with cell extract but was stronger in its intensity (Heo et al., 2008). A catalytically dead mutant of TUT4, which contains a point mutation at the nucleotidyl transferase domain (D1011A), failed to generate any elongated band (Figures 3A and S3A), ruling out the possibility that a contaminating protein was responsible for the reaction. The two human homologs of Lin28—Lin28a and Lin28b—show the equivalent effect on the uridylation of pre-miRNA by TUT4 (Figure 3A), which is consistent with our previous results (Heo et al., 2008). Furthermore, TUT4 can extend pre-let-7g but neither pre-miR-16-1 nor -30a, indicating that Lin28-mediated TUT4 activity is specific to let-7 family pre-miRNAs (Figures 3A and 3B). Next, to investigate the nucleotide specificity of TUT4, the immunopurified TUT4 was incubated with pre-let-7, recombinant Lin28b, and each of the four nucleotide triphosphates (NTPs) separately. TUT4 elongated pre-let-7 most efficiently with UTP (Figure S3B). At a lower concentration of NTPs (0.025 mM), the elongation activity with UTP was not reduced whereas the activ-

700 Cell 138, 696–708, August 21, 2009 ª2009 Elsevier Inc.

ities with other NTPs decreased (Figure 3C). Therefore, TUT4 prefers UTP to the other NTPs. This UTP preference is consistent with our previous results showing that the 30 -tails of up-let-7 are predominantly, but not exclusively, composed of uridine residues (Heo et al., 2008). We also compared the activities between three different isoforms of human TUT4. Although the isoforms a, b, and c (NP_001009881, NP_056084, and NP_001009882, respectively) are different in protein length (1645, 1644, and 1640 amino acids, respectively) due to alternative splicing, they are comparable in their uridylating activities (Figure S4). Isoform c was used in the experiments throughout this study.

The Other TUTases Do Not Regulate pre-let-7 To investigate whether the other TUTases can also uridylate prelet-7, seven human noncanonical PAPs including TUT4 were cloned and immunopurified (Figure 4A). Western blotting data showed that the seven TUTases were expressed and immunoprecipitated at comparable levels (Figure 4B). When these proteins were incubated with pre-let-7, however, only TUT4 generated up-let-7 (Figure 4C).

Figure 5. Knockdown of TUT4 and Lin28 Increases let-7 and Decreases Stem Cell Markers (A) MicroRNA microarray was performed with mES cells (R1) transfected with siRNAs against TUT4 or Lin28a. Four biological replicates were used for the assay. The microarray signals from each type of sample were normalized to the signals from siGFP-treated samples. The normalized microarray signals from Lin28a-depleted samples (horizontal axis) were plotted versus those from TUT4-depleted samples (vertical axis) in log scale. (B) Northern blotting analysis was carried out following the knockdown of TUT4 and Lin28a in mES cells (A3-1). For quantification of the northern blotting, the band intensity was measured by phosphoimager and normalized against signals from siGFP-treated sample. The standard errors are from two independent experiments, and the original northern blotting data are shown in Figure S6A. (C) Double knockdown of TUT4 and Lin28a in mES cells (R1). Equal amounts of siTUT4 (20 nM) and siLin28a (20 nM) were combined to knock down TUT4 and Lin28a simultaneously. For singlegene knockdown, 20 nM siGFP was added to adjust the total amount of siRNA to 40 nM. The levels of mature miRNAs and pri-miRNAs were analyzed by qRT-PCR. The standard errors are from three data sets. The knockdown efficiency was determined by western blotting in Figure S6B. (D) Knockdown of TUT4 and Lin28a affects the maintenance of mES cells (R1). The siRNA against either TUT4 or Lin28a was transfected into feederfree R1. The cells were incubated in the presence of LIF. One day later, identical numbers of cells were aggregated to make EBs using the hanging drop method (Wang and Yang, 2008) without LIF for 2 days and were transferred into the bacterial Petri dish with fresh differentiation medium. Two days later, RNA was extracted and analyzed by qRT-PCR to measure the mRNA levels of pluripotency markers, Oct4 and Nanog. The standard errors are from three independent experiments, and paired one-tailed t test was used to calculate the p value (*p < 0.1; **p < 0.05; ***p < 0.01).

We also knocked down each protein in mES cells and carried out qRT-PCR to measure let-7 levels. As the knockdown of TUTase6 was previously reported to be lethal and TUTase6 is known to uridylate U6 snRNA, we did not test TUTase6 in this experiment (Trippe et al., 2006). Although the TUTases were knocked down efficiently, the mature let-7 level did not change considerably (Figure S5). Therefore, our data indicate that TUT4 may be the sole enzyme responsible for pre-let-7 uridylation. In the case of TUTase7, which is similar to TUT4 in its domain organization, a weak smeared band appeared in the reaction (Figure 4C, lane 14). Therefore, although depletion of TUTase7 did not show a significant derepressive effect on let-7 levels (Figure S5), we cannot formally rule out the possibility that TUTase7 has weak redundancy with TUT4.

let-7 Family Is a Major Target of Both TUT4 and Lin28 Next, by performing microarray experiments on the RNA samples from mES cells (R1) depleted of either TUT4 or Lin28a, we investigated whether TUT4 can regulate miRNAs other than let-7 (Table S2). The signals of let-7 family members increased significantly upon Lin28a knockdown (Figure 5A, horizontal axis); the same effect was observed with TUT4 knockdown (Figure 5A, vertical axis). Notably, it was only the let-7 family— let-7a, let-7d, let-7f, let-7g, and let-7i—that was upregulated by over 2-fold in both TUT4- and Lin28a-depleted cells (Figure 5A, top right). The other members of the let-7 family—let7b, let-7c, let-7e, and miR-98—were expressed at levels too low for reliable quantitation. Levels of most miRNAs including miR-16 did not change significantly, with some exceptions in which the changes were modest but reproducible (discussed

Cell 138, 696–708, August 21, 2009 ª2009 Elsevier Inc. 701

Figure 6. Conserved Sequence Motif Triggers Lin28 Binding and TUT4 Uridylation Activity (A) (left) Shown is the schematic representation of a hybrid form of pre-miRNA. (middle) Electrophoretic mobility shift assay (EMSA) was carried out with 50 endlabeled synthetic pre-let-7a-1, pre-miR-16-1, and their hybrid. ‘‘Free’’ represents RNA free of proteins and ‘‘bound’’ represents RNA bound to Lin28. The concentrations of rLin28b used are 0, 3, 10, 30, 100, and 300 nM. (right) In vitro uridylation was carried out with FLAG-TUT4-overexpressed HEK293T cell extracts and 150 nM rLin28b, which is the same with all the other uridylation reactions in this figure. The experiment was repeated twice (Figure S7). (B) Aligned are the sequences of human pre-let-7 miRNAs. Highlighted at left and right are 5p and 3p strands. Boxed in the middle is the conserved GGAG sequence. (C) GGAG and GGAG-like motifs were searched in the terminal loop near the 3p strand with all pre-let-7 miRNAs in bilateral animals (miRBase, release 13.0). ‘‘X’’ stands for a random nucleotide sequence. The numbers indicate pre-let-7 miRNAs counted in total. The numbers in parentheses indicate miRNAs counted in invertebrates only. N/C stands for ‘‘not conserved.’’

702 Cell 138, 696–708, August 21, 2009 ª2009 Elsevier Inc.

below). We confirmed this observation using another mES cell line, A3-1 (Figures 5B and S6A). Consistent with the microarray data, the levels of all the let-7 family members detected by northern blotting increased by more than 2-fold in both TUT4and Lin28a-depleted cells. These results indicate that let-7 miRNAs are the major targets of both TUT4 and Lin28. Then, we tested the effect of double knockdown of TUT4 and Lin28 on let-7 biogenesis. We found that the levels of mature let-7 increased dramatically in mES cells depleted of both TUT4 and Lin28a at the same time (Figure 5C). The degree of accumulation was even greater than the changes observed in the samples with knockdown of either TUT4 or Lin28a alone, suggesting that both TUT4 and Lin28 are required for the suppression of let-7 biogenesis. The level of miR-16 did not change significantly in the double knockdown cells. The efficiency of knockdown was measured by western blotting (Figure S6B). We reproducibly observed that RNA interference (RNAi) of TUT4 is less efficient than that of other genes including Lin28, which may explain why let-7 is upregulated less prominently in TUT4-depleted cells than in Lin28-depleted cells. Further investigation on the knockdown of TUT4 is described later (see Discussion). TUT4 and Lin28 Contribute to the Maintenance of ES Cells let-7 has been implicated in cell differentiation in mammals (Bussing et al., 2008) and nematodes (Roush and Slack, 2008). It has also been shown that Lin28a overexpression facilitates the formation of inducible pluripotent stem (iPS) cells (Yu et al., 2007). Because TUT4 suppresses let-7, it is conceivable that TUT4 is also involved in stem cell formation and/or maintenance. To further examine the effect of TUT4 and Lin28a on ES cell maintenance, we transfected mES cells (R1) with siRNAs and incubated the cells for 5 days under conditions that induce the formation of embryoid bodies (EBs). It is known that mES cells pose for differentiation and begin to lose the expression of pluripotency markers when the cells are grown to form cell aggregates (EBs) in the absence of leukemia inhibitory factor (LIF). Figure 5D shows that the pluripotency markers, Oct4 and Nanog, decrease more rapidly in cells depleted of TUT4 or Lin28a than in the control cells. Notably, more significant reduction was observed when both TUT4 and Lin28a were knocked down. This is consistent with the increased effect of double knockdown on the level of mature let-7 (compare Figures 5C and 5D). Taken together, our results demonstrate that TUT4 and Lin28a are required for ES cells to maintain pluripotency. Lin28 Interacts with a Motif in the Terminal Loop of pre-let-7 Finally, the high selectivity of TUT4 and Lin28 toward let-7 observed in Figure 5A promoted us to address the reason for this specificity. It has been suggested that Lin28 specifically

recognizes the terminal loop of pre-let-7 (Newman et al., 2008; Piskounova et al., 2008). To confirm this, we generated a hybrid form of pre-miR-16 by replacing its loop with that of pre-let-7a-1 and carried out electrophoretic mobility shift assay (EMSA) with 50 end-labeled RNA and recombinant Lin28 protein (Figure 6A, middle). The hybrid pre-miR-16 with the let-7 loop effectively interacted with Lin28. Furthermore, the hybrid RNA was as efficiently uridylated by TUT4 as pre-let-7 was (Figure 6A, right and Figure S7). Therefore, the pre-let-7 loop is sufficient for Lin28 binding of pre-miRNA and subsequent TUT4 activity. To identify potential sequence motifs in the loop of pre-let-7, which may be recognized by Lin28, we aligned the pre-miRNA sequences of all human let-7 members (Figure 6B). While the loop sequences exhibit diversity among the let-7 members, there are short common sequences near both 5p and 3p strands. To determine which of these regions may function as a recognition motif, we generated pre-let-7a-1 mutants with a shortened terminal loop that is equivalent to only a half of the full-length loop (Figure S8A). When we carried out in vitro uridylation with these constructs, the mutant containing the shortened loop sequence from the 30 half was uridylated (Figure S8B), indicating that a recognition motif exists in this region. Lin28 proteins contain two CCHC-type zinc fingers that are critical for uridylation (Heo et al., 2008). When we performed EMSA with the Lin28b zinc finger mutant (Figure S9A), its binding affinity toward pre-let-7 decreased dramatically compared to that of the wild-type Lin28b, whereas the weak and nonspecific binding toward pre-miR-16 did not (Figure S9B). Together, these data suggest that it may be the CCHC-type zinc fingers that mediate the specific interaction of Lin28 with pre-let-7 and the subsequent uridylation. It had been reported that the CCHC-type zinc finger domains recognize a specific RNA sequence motif of ‘‘GGAG’’ or ‘‘GGUG’’ in the case of an HIV nucleocapsid protein (Amarasinghe et al., 2000; De Guzman et al., 1998; Pappalardo et al., 1998). Interestingly, the sequence of GGAG is present in the terminal loop near the 3p strand of pre-let-7 (Figure 6B). The sequence and its position are highly conserved throughout all vertebrates (Figures 6C and 6D). To explore the potential importance of the GGAG motif, we introduced a mutation (GuAu) and carried out EMSA and in vitro uridylation. The affinity of Lin28 toward the mutant was lower compared to that toward the wild-type pre-let-7 (Figure S10A). In addition, uridylation was not observed in the mutant (Figures 6E and S10B), showing that the GGAG motif is indeed essential for Lin28 binding and TUT4 action. To further test the importance of the GGAG motif, we asked whether the introduction of GGAG into other unrelated premiRNAs is sufficient for the binding to Lin28 and subsequent TUT4 activity. We generated three artificial pre-miR-16-1 constructs, each of which has the GGAG motif in its terminal loop near the 3p strand (Figures 6F and S11). Remarkably, the

(D) The location of the conserved motifs is measured based on their distance from the 50 end of the 3p strand, throughout all bilateral animals. (E) In vitro uridylation was performed with the wild-type and the mutant of pre-let-7a-1. The GGAG motif was mutated into GuAu. This experiment was repeated twice (Figure S10B). Lin28 binding affinity toward the mutant RNA is shown in Figure S10A. (F) (top) Shown are the sequences of pre-miR-16-1 and its mutants. The GGAG motif introduced is highlighted in the middle. (left) EMSA was carried out with 30 nM rLin28b and was repeated twice (Figure S11A). (right) In vitro uridylation was carried out and repeated twice (Figure S11B).

Cell 138, 696–708, August 21, 2009 ª2009 Elsevier Inc. 703

introduction of the short sequence motif of GGAG enhanced Lin28 binding affinity. Furthermore, ‘‘GGAG@4’’ mutant was as effectively uridylated as pre-let-7, suggesting that when located in the terminal loop with a proper context, the GGAG motif is sufficient for Lin28 binding and TUT4 action. A Set of Pre-miRNAs with GGAG Are Subject to Uridylation in a Lin28-Dependent Manner The essential role of the GGAG motif suggested that other premiRNAs with the same motif might be regulated by Lin28 and TUT4. Searching human miR-1 through miR-400 revealed 15 such candidates, 12 of which contain the motif near the 3p strand (miR-9, 107, 132, 139, 142, 143, 149, 152, 200c, 204, 324, and 363) (Figure S12). The GGAG motifs in 9 of the 12 miRNAs are conserved in mammals (miR-107, 139, 143, 149, 152, 200c, 204, 324, and 363). We examined Lin28-dependent uridylation in a subset of the pre-miRNAs selected above (5 out of the final 9; miR-107, 143, 200c, 324, and 363; randomly chosen) (Figure 7A, top). UpmiRNA was observed with all of the five miRNAs—the highest efficiency with pre-miR-107, 143, and 200c. The up-miRNA population was not observed with randomly chosen pre-miRNAs that do not contain the GGAG motif (Figure 7A, bottom). Lin28 binds to the three miRNAs (pre-miR-107, 143, and 200c) (Figures 7B and S13A). The qRT-PCR and the microarray data indicate that the mature miRNA levels of the three miRNAs are modestly upregulated upon Lin28 and TUT4 knockdown in mES cells (Figures 7C and S13B). Whereas the functions of these miRNAs in stem cells are largely unknown, they are more abundantly observed in differentiated cells than in undifferentiated cells (Bar et al., 2008; Esau et al., 2004; Peter, 2009) as in the case of let-7 (Figure S13C). These data support the role of Lin28 and TUT4 in stem cell maintenance through suppression of a set of miRNAs including let-7. DISCUSSION Here, we identified a terminal uridylyl transferase, TUT4, as a novel regulator of miRNA biogenesis. The action mechanism is summarized in Figure 7D. Following the nuclear export of premiRNA, Lin28 binds to pre-miRNA in the cytoplasm. Lin28 recognizes its substrate through a conserved sequence motif in the RNA loop. Subsequently, TUT4 recognizes the binary complex of Lin28 and pre-miRNA and adds an oligouridine tail (U tail) of 10–30 nt to the 30 terminus of pre-miRNA. The uridylated pre-miRNA (up-miRNA) is resistant to Dicer processing since Dicer is unable to cleave hairpin RNAs with such long 30 extensions. Because uridylyl groups are known to recruit 30 /50 exonucleases (Mullen and Marzluff, 2008; Shen and Goodman, 2004), the U tail may facilitate decay of up-miRNA. The identity of the nuclease(s) responsible for the degradation of up-miRNA is currently unknown. We have previously shown that Lin28 acts mainly after Drosha processing and after nuclear export steps (Heo et al., 2008). This conclusion was based on the following evidence: (1) Lin28 is localized mainly to the cytoplasm, (2) the pri-let-7 levels do not change significantly in either Lin28-depleted or Lin28-overexpressing cells, (3) nuclear export of pre-let-7 is not affected in

704 Cell 138, 696–708, August 21, 2009 ª2009 Elsevier Inc.

Lin28-overexpressing cells, (4) Lin28 associates with pre-let-7 but not with pri-let-7 in vivo, and (5) Lin28 induces uridylation of pre-let-7 when incubated with cytoplasmic extract. Our current discovery of TUT4, which confirms and extends our previous model, suggests that let-7 biogenesis is suppressed mainly in the cytoplasm by the concerted action of the two cytoplasmic proteins Lin28 and TUT4. From the biochemical analyses, we learn how Lin28 and TUT4 recognize their substrates. The association of TUT4 with premiRNAs is strongly dependent on Lin28 (Figure 2). Lin28 was reported to interact with the terminal loop of pre-let-7 (Newman et al., 2008; Piskounova et al., 2008), and we show that Lin28 specifically recognizes a sequence motif of GGAG in the terminal loop of pre-let-7 (Figure 6). The GGAG motif is conserved throughout vertebrates (Figure 6C). It is notable that the motif may come in variable forms of GGAG (GAAG, GGUG/GGCG, and UGXG) in other miRNAs such as human let-7b, miR-98, and let-7a-3. Lin28 may recognize these variable forms as well, considering that the fourth G is the most essential for the interaction between CCHC-type zinc fingers and RNA (De Guzman et al., 1998). The mutations retaining the fourth G intact (GagG and aaAG) weaken but do not fully disrupt the interaction between Lin28 and pre-let-7 (Figure S10). It is noted that the sequence motif of GGAG is not conserved in invertebrates (Figures 6C and 6D). Lin28 homologs of invertebrates may have different sequence specificity. Lin28 binding does not always lead to effective uridylation by TUT4 (Figure S10, GagG and aaAG mutants; Figure 6F, GGAG@5). This result is in parallel with the observation that some pre-miRNAs containing the GGAG motif are not effectively uridylated (Figure 7A; miR-324 and 363). We suspect that the location of the tetra-nucleotide motif determines the proper context for Lin28 binding and TUT4 action, considering that its location is highly conserved in all vertebrates (Figure 6D). It is also possible that TUT4 may discriminate the pre-miRNAs to some degree, contributing to the substrate specificity, when it binds to the binary complex of Lin28 and pre-miRNA. TUT4 is expressed broadly while Lin28 is restricted to undifferentiated stem cells and certain cancer cells. Therefore, while TUT4 regulates some miRNAs together with Lin28 in Lin28abundant cells, TUT4 may also regulate other miRNAs in a Lin28-independent manner in different cell types. It is conceivable that different specificity factors assist TUT4 in recognizing distinct pre-miRNAs. It is also plausible that other TUTases may participate in the miRNA pathway by modifying the 30 end of other pre-miRNAs. It is worthwhile to mention that we encountered technical difficulties in the RNAi of TUT4. We have tried six different siRNAs and five small hairpin RNAs for RNAi of TUT4 without significantly improving the knockdown efficiency. When we carried out knockdown for various durations, we realized that TUT4 mRNA level is rapidly restored 24 hr after transfection whereas Lin28 mRNA is depleted effectively for a longer period (Figure S14). We observed a similar pattern of TUT4 recovery in a hepatocellular carcinoma line, Huh7 (data not shown). This fast restoration of TUT4 implicates a strong feedback control that maintains TUT4 at homeostatic levels in these cells. The mechanism of this feedback regulation requires further investigation.

Figure 7. Pre-miRNAs with the GGAG Sequence Motif Are Regulated by Lin28 and TUT4 (A) In vitro uridylation was carried out with six different premiRNAs that contain the GGAG sequence near 3p strand in their terminal loop (top) and five pre-miRNAs without GGAG (bottom). It was carried out with FLAG-TUT4 overexpressed HEK293T cell extracts, 150 nM rLin28b, and 0.25 mM UTP. All the pre-miRNAs are from human except miR-7b, which is from mouse. The standard deviations are from three (top) and two (bottom) independent experiments. (B) EMSA was carried out with 50 end-labeled synthetic pre-miR107, 143, and 200c as well as pre-let-7a-1 and pre-miR-16-1 with 30 nM rLin28b. ‘‘Free’’ represents RNA free of proteins and ‘‘bound’’ represents the slowly migrating RNA bound to Lin28. The standard deviations are shown in Figure S13A. (C) The mature miRNA levels of miR-107, 143, and 200c were measured by qRT-PCR with the RNA samples used in Figure 5C. The standard errors are from three independent experiments. (D) A model for the suppression of miRNA by Lin28 and TUT4. PrimiRNA is cropped into pre-miRNA by Microprocessor consisting of Drosha and DGCR8. The pre-miRNA is exported to the cytoplasm by exportin 5 (EXP5). After the export, the pre-miRNA is bound to Lin28, which interferes with Dicer processing. Lin28 recruits TUT4, which uridylates the pre-miRNA. The uridylated pre-miRNA (up-miRNA) fails to be processed by Dicer and gets degraded by nuclease(s).

Cell 138, 696–708, August 21, 2009 ª2009 Elsevier Inc. 705

let-7 is a critical regulator of cell differentiation and cell proliferation (Bussing et al., 2008). By suppressing let-7, Lin28 participates in stem cell maintenance and tumorigenesis (Bussing et al., 2008; Chang et al., 2009). Our current study suggests that TUT4 is important in the maintenance of the undifferentiated state (Figure 5D) and possibly in the formation of iPS cells. It would be interesting to investigate the involvement of TUT4 in human cancers since abnormal expression of TUT4 may perturb miRNA populations. Lung, liver, and ovarian cancers in which let-7 is downregulated are particularly interesting tumor types to investigate. Furthermore, Lin28 and TUT4 are promising new targets for drug discovery. The uridylation assay developed in our study may provide a way to screen for chemicals that either boost or inhibit let-7 biogenesis, which may be useful for stem cell engineering and cancer therapy. EXPERIMENTAL PROCEDURES Cell Culture HEK293T cell was grown in DMEM (WelGENE) supplemented with 10% fetal bovine serum (WelGENE). Mouse embryonic stem cell line R1 was maintained in GlutaMAX DMEM (GIBCO, Invitrogen) supplemented with 10% fetal bovine serum (Hyclone), 100 unit/ml penicillin, 100 mg/ml streptomycin (WelGENE), nonessential amino acids (GIBCO, Invitrogen), 100 mM 2-mercaptoethanol (Amresco), and 10 ng/ml leukemia inhibitory factor (Sigma). R1 cell was cultured on a feeder layer of mouse CF-1 cells treated with mitomycin C (Sigma), and the media were changed daily. Another mouse embryonic stem cell line A3-1 was cultured on gelatin-coated dishes with DMEM (WelGENE) containing 20% FBS (Hyclone), 100 units/ml penicillin, 100 mg/ml streptomycin, nonessential amino acids (GIBCO, Invitrogen), 13 nucleosides mix (Sigma), 100 mM 2-mercaptoethanol (Amresco), and 0.15% LIF. The LIF was from the conditioned media from a CHO cell line secreting human LIF. To make EB from a mES cell (R1), hanging drop method was used (Wang and Yang, 2008). In summary, transfected cells were trypsinized and diluted to a concentration of 100,000 cells per 1 ml of LIF-free differentiation medium. The 20 ml drops (500 cells/one drop) were suspended on the inside surface of the up-turned lids of 150 mm tissue culture dishes. The dishes were incubated for 2 days in the 37 C incubators to make cell aggregates (EBs). Two days later, EBs were transferred into bacterial Petri dishes with fresh differentiation medium. After 2 days of incubation, EBs were collected for RNA extraction. Transfection To prepare R1 cells for transfection, feeder cells were removed, and 300,000 cells were seeded onto gelatin-coated 60 mm dish. The 30 nM siRNA duplexes and Lipofectamine 2000 (Invitrogen) were used for the transfection. In the double knockdown experiment, a total 40 nM siRNA (20 nM for each gene) was used. The sequences of siRNA are listed in the Supplemental Data.

and incubated with constant rotation for 12 hr at 4 C. After washing with buffer I, the associated proteins were separated on 10% SDS-PAGE. The nuclear and cytoplasmic extracts from subcellular fractionation were diluted in buffer L and used for the same procedure as above. Immunoprecipitation and In Vitro Uridylation For immunoprecipitation of FLAG-TUTases, HEK293T cells grown on 10 cm dishes were collected 48 hr after transfection of FLAG-TUTase expression plasmids. The cells were incubated with lysis buffer (500 mM NaCl, 1 mM EDTA, 10 mM Tris [pH 8.0], 1% Triton X-100) for 20 min followed by sonication on ice and centrifugation twice for 10 min at 4 C. The supernatant was incubated with 5 ml of anti-FLAG antibody-conjugated agarose beads (anti-FLAG M2 affinity gel, Sigma) with constant rotation for 2 hr at 4 C. The beads were washed three times with lysis buffer and then four times with 200 mM KCl buffer D (200 mM KCl, 10 mM Tris [pH 8.0], 0.1 mM EDTA). The reaction was performed in a total volume of 30 ml in 3.2 mM MgCl2, 1 mM DTT, 0.25 mM NTP (Ambion), 50 end-labeled pre-miRNA of 1 3 104–1 3 105 cpm, and 15 ml of immunopurified proteins in buffer D. The reaction mixture was incubated at 37 C for 20 min. The RNA was purified from the reaction mixture by phenol extraction and analyzed on 12.5% urea polyacrylamide gel. pre-let7a-1, pre-let-7g, pre-miR-16-1, and pre-miR-30a were synthesized by Samchully Pham. The pre-miRNAs were labeled at the 50 end with T4 polynucleotide kinase (Takara) and [g-32P] ATP. The sequences of pre-miRNAs are listed in the Supplemental Data. SUPPLEMENTAL DATA Supplemental Data include Supplemental Experimental Procedures, 14 figures, and 2 tables and can be found with this article online at http://www. cell.com/supplemental/S0092-8674(09)00964-7. ACKNOWLEDGMENTS We are grateful to the members of our laboratory, particularly Hyeshik Chang for helpful discussion and Ahyoung Cho and Je-Keun Rhee for technical help. We also thank Yong Joon Huh and Dr. Dong-Wook Kim for the gift of mES cells (R1) and their advice on stem cell culture; Kyung Hoon Kim and Dr. Se Won Seo for their advice on protein purification; and Drs. Guhung Jung and Seung Hwan Hong for the gift of cell lines. This work was supported by the Creative Research Initiatives Program (20090063603) through the National Research Foundation of Korea (NRF) and the BK21 Research Fellowships (I.H., C.J., Y.-K.K., J.C., K.-H.Y., and J.H.) from the Ministry of Education, Science and Technology of Korea. Received: April 6, 2009 Revised: June 16, 2009 Accepted: August 3, 2009 Published: August 20, 2009 REFERENCES

RNA Pull-Down pre-let-7a-1 and pre-miR-16-1 with 50 extension were produced by in vitro transcription. PCR products were used as direct templates for in vitro transcription. A 30 -biotinylated adaptor DNA complementary to the 50 extension was incubated with streptavidin-conjugated agarose beads (Pierce) in buffer I (20 mM Tris-Cl [pH 7.5], 100 mM KCl, 2 mM EDTA) at 4 C for 1 hr. The beads were washed twice with 300 mM KCl buffer D (20 mM Tris-Cl [pH 8.0], 300 mM KCl, 0.2 mM EDTA) and incubated with the in vitro transcribed pre-miRNA in buffer D with 100 U/ml of RNase inhibitor (Takara) for 2–3 hr at room temperature. The beads were washed twice with buffer I and incubated with cell extract. HEK293T and A3-1 cells were collected and incubated in buffer L (50 mM Tris-HCl [pH 8.0], 150 mM KCl, 1 mM DTT, 10% glycerol, 0.5% Triton X-100) on ice for 30 min. The suspended cells were passed through a 23-guage syringe 20 times. After removing cell debris by centrifugation, the total cell extract was added to the pre-miRNAs immobilized on the streptavidin beads

706 Cell 138, 696–708, August 21, 2009 ª2009 Elsevier Inc.

Amarasinghe, G.K., De Guzman, R.N., Turner, R.B., Chancellor, K.J., Wu, Z.R., and Summers, M.F. (2000). NMR structure of the HIV-1 nucleocapsid protein bound to stem-loop SL2 of the psi-RNA packaging signal. Implications for genome recognition. J. Mol. Biol. 301, 491–511. Balzer, E., and Moss, E.G. (2007). Localization of the developmental timing regulator Lin28 to mRNP complexes, P-bodies and stress granules. RNA Biol. 4, 16–25. Bar, M., Wyman, S.K., Fritz, B.R., Qi, J., Garg, K.S., Parkin, R.K., Kroh, E.M., Bendoraite, A., Mitchell, P.S., Nelson, A.M., et al. (2008). MicroRNA discovery and profiling in human embryonic stem cells by deep sequencing of small RNA libraries. Stem Cells 26, 2496–2505. Bernstein, E., Caudy, A.A., Hammond, S.M., and Hannon, G.J. (2001). Role for a bidentate ribonuclease in the initiation step of RNA interference. Nature 409, 363–366.

Bohnsack, M.T., Czaplinski, K., and Gorlich, D. (2004). Exportin 5 is a RanGTPdependent dsRNA-binding protein that mediates nuclear export of premiRNAs. RNA 10, 185–191.

Lee, Y., Ahn, C., Han, J., Choi, H., Kim, J., Yim, J., Lee, J., Provost, P., Radmark, O., Kim, S., et al. (2003). The nuclear RNase III Drosha initiates microRNA processing. Nature 425, 415–419.

Bussing, I., Slack, F.J., and Grosshans, H. (2008). let-7 microRNAs in development, stem cells and cancer. Trends Mol. Med. 14, 400–409.

Lee, Y.S., and Dutta, A. (2007). The tumor suppressor microRNA let-7 represses the HMGA2 oncogene. Genes Dev. 21, 1025–1030.

Chang, T.C., Zeitels, L.R., Hwang, H.W., Chivukula, R.R., Wentzel, E.A., Dews, M., Jung, J., Gao, P., Dang, C.V., Beer, M.A., et al. (2009). Lin-28B transactivation is necessary for Myc-mediated let-7 repression and proliferation. Proc. Natl. Acad. Sci. USA 106, 3384–3389.

Lund, E., Guttinger, S., Calado, A., Dahlberg, J.E., and Kutay, U. (2004). Nuclear export of microRNA precursors. Science 303, 95–98.

De Guzman, R.N., Wu, Z.R., Stalling, C.C., Pappalardo, L., Borer, P.N., and Summers, M.F. (1998). Structure of the HIV-1 nucleocapsid protein bound to the SL3 psi-RNA recognition element. Science 279, 384–388. Denli, A.M., Tops, B.B., Plasterk, R.H., Ketting, R.F., and Hannon, G.J. (2004). Processing of primary microRNAs by the Microprocessor complex. Nature 432, 231–235. Esau, C., Kang, X., Peralta, E., Hanson, E., Marcusson, E.G., Ravichandran, L.V., Sun, Y., Koo, S., Perera, R.J., Jain, R., et al. (2004). MicroRNA-143 regulates adipocyte differentiation. J. Biol. Chem. 279, 52361–52365. Gregory, R.I., Yan, K.P., Amuthan, G., Chendrimada, T., Doratotaj, B., Cooch, N., and Shiekhattar, R. (2004). The Microprocessor complex mediates the genesis of microRNAs. Nature 432, 235–240. Grishok, A., Pasquinelli, A.E., Conte, D., Li, N., Parrish, S., Ha, I., Baillie, D.L., Fire, A., Ruvkun, G., and Mello, C.C. (2001). Genes and mechanisms related to RNA interference regulate expression of the small temporal RNAs that control C. elegans developmental timing. Cell 106, 23–34. Grosshans, H., Johnson, T., Reinert, K.L., Gerstein, M., and Slack, F.J. (2005). The temporal patterning microRNA let-7 regulates several transcription factors at the larval to adult transition in C. elegans. Dev. Cell 8, 321–330. Guo, Y., Chen, Y., Ito, H., Watanabe, A., Ge, X., Kodama, T., and Aburatani, H. (2006). Identification and characterization of lin-28 homolog B (LIN28B) in human hepatocellular carcinoma. Gene 384, 51–61.

Martin, G., and Keller, W. (2007). RNA-specific ribonucleotidyl transferases. RNA 13, 1834–1849. Mayr, C., Hemann, M.T., and Bartel, D.P. (2007). Disrupting the pairing between let-7 and Hmga2 enhances oncogenic transformation. Science 315, 1576–1579. Meneely, P.M., and Herman, R.K. (1979). Lethals, steriles and deficiencies in a region of the X chromosome of Caenorhabditis elegans. Genetics 92, 99–115. Mourelatos, Z., Dostie, J., Paushkin, S., Sharma, A., Charroux, B., Abel, L., Rappsilber, J., Mann, M., and Dreyfuss, G. (2002). miRNPs: a novel class of ribonucleoproteins containing numerous microRNAs. Genes Dev. 16, 720–728. Mullen, T.E., and Marzluff, W.F. (2008). Degradation of histone mRNA requires oligouridylation followed by decapping and simultaneous degradation of the mRNA both 50 to 30 and 30 to 50 . Genes Dev. 22, 50–65. Nagaike, T., Suzuki, T., Katoh, T., and Ueda, T. (2005). Human mitochondrial mRNAs are stabilized with polyadenylation regulated by mitochondria-specific poly(A) polymerase and polynucleotide phosphorylase. J. Biol. Chem. 280, 19721–19727. Nakanishi, T., Kubota, H., Ishibashi, N., Kumagai, S., Watanabe, H., Yamashita, M., Kashiwabara, S., Miyado, K., and Baba, T. (2006). Possible role of mouse poly(A) polymerase mGLD-2 during oocyte maturation. Dev. Biol. 289, 115–126.

Hammond, S.M., Boettcher, S., Caudy, A.A., Kobayashi, R., and Hannon, G.J. (2001). Argonaute2, a link between genetic and biochemical analyses of RNAi. Science 293, 1146–1150.

Newman, M.A., Thomson, J.M., and Hammond, S.M. (2008). Lin-28 interaction with the Let-7 precursor loop mediates regulated microRNA processing. RNA 14, 1539–1549.

Han, J., Lee, Y., Yeom, K.H., Kim, Y.K., Jin, H., and Kim, V.N. (2004). The Drosha-DGCR8 complex in primary microRNA processing. Genes Dev. 18, 3016– 3027.

Pappalardo, L., Kerwood, D.J., Pelczer, I., and Borer, P.N. (1998). Threedimensional folding of an RNA hairpin required for packaging HIV-1. J. Mol. Biol. 282, 801–818.

Heo, I., Joo, C., Cho, J., Ha, M., Han, J., and Kim, V.N. (2008). Lin28 mediates the terminal uridylation of let-7 precursor MicroRNA. Mol. Cell 32, 276–284.

Pasquinelli, A.E., Reinhart, B.J., Slack, F., Martindale, M.Q., Kuroda, M.I., Maller, B., Hayward, D.C., Ball, E.E., Degnan, B., Muller, P., et al. (2000). Conservation of the sequence and temporal expression of let-7 heterochronic regulatory RNA. Nature 408, 86–89.

Hutvagner, G., McLachlan, J., Pasquinelli, A.E., Balint, E., Tuschl, T., and Zamore, P.D. (2001). A cellular function for the RNA-interference enzyme Dicer in the maturation of the let-7 small temporal RNA. Science 293, 834–838. Johnson, S.M., Grosshans, H., Shingara, J., Byrom, M., Jarvis, R., Cheng, A., Labourier, E., Reinert, K.L., Brown, D., and Slack, F.J. (2005). RAS is regulated by the let-7 microRNA family. Cell 120, 635–647. Katoh, T., Sakaguchi, Y., Miyauchi, K., Suzuki, T., Kashiwabara, S., and Baba, T. (2009). Selective stabilization of mammalian microRNAs by 30 adenylation mediated by the cytoplasmic poly(A) polymerase GLD-2. Genes Dev. 23, 433–438. Ketting, R.F., Fischer, S.E., Bernstein, E., Sijen, T., Hannon, G.J., and Plasterk, R.H. (2001). Dicer functions in RNA interference and in synthesis of small RNA involved in developmental timing in C. elegans. Genes Dev. 15, 2654–2659.

Peter, M.E. (2009). Let-7 and miR-200 microRNAs: guardians against pluripotency and cancer progression. Cell Cycle 8, 843–852. Piskounova, E., Viswanathan, S.R., Janas, M., LaPierre, R.J., Daley, G.Q., Sliz, P., and Gregory, R.I. (2008). Determinants of microRNA processing inhibition by the developmentally regulated RNA-binding protein Lin28. J. Biol. Chem. 283, 21310–21314. Polesskaya, A., Cuvellier, S., Naguibneva, I., Duquet, A., Moss, E.G., and Harel-Bellan, A. (2007). Lin-28 binds IGF-2 mRNA and participates in skeletal myogenesis by increasing translation efficiency. Genes Dev. 21, 1125–1138.

Kim, V.N., Han, J., and Siomi, M.C. (2009). Biogenesis of small RNAs in animals. Nat. Rev. Mol. Cell Biol. 10, 126–139.

Reinhart, B.J., Slack, F.J., Basson, M., Pasquinelli, A.E., Bettinger, J.C., Rougvie, A.E., Horvitz, H.R., and Ruvkun, G. (2000). The 21-nucleotide let-7 RNA regulates developmental timing in Caenorhabditis elegans. Nature 403, 901–906.

Knight, S.W., and Bass, B.L. (2001). A role for the RNase III enzyme DCR-1 in RNA interference and germ line development in Caenorhabditis elegans. Science 293, 2269–2271.

Richards, M., Tan, S.P., Tan, J.H., Chan, W.K., and Bongso, A. (2004). The transcriptome profile of human embryonic stem cells as defined by SAGE. Stem Cells 22, 51–64.

Kwak, J.E., and Wickens, M. (2007). A family of poly(U) polymerases. RNA 13, 860–867.

Rissland, O.S., Mikulasova, A., and Norbury, C.J. (2007). Efficient RNA polyuridylation by noncanonical poly(A) polymerases. Mol. Cell. Biol. 27, 3612–3624.

Landthaler, M., Yalcin, A., and Tuschl, T. (2004). The human DiGeorge syndrome critical region gene 8 and Its D. melanogaster homolog are required for miRNA biogenesis. Curr. Biol. 14, 2162–2167.

Rouhana, L., Wang, L., Buter, N., Kwak, J.E., Schiltz, C.A., Gonzalez, T., Kelley, A.E., Landry, C.F., and Wickens, M. (2005). Vertebrate GLD2 poly(A) polymerases in the germline and the brain. RNA 11, 1117–1130.

Cell 138, 696–708, August 21, 2009 ª2009 Elsevier Inc. 707

Roush, S., and Slack, F.J. (2008). The let-7 family of microRNAs. Trends Cell Biol. 18, 505–516. Rybak, A., Fuchs, H., Smirnova, L., Brandt, C., Pohl, E.E., Nitsch, R., and Wulczyn, F.G. (2008). A feedback loop comprising lin-28 and let-7 controls pre-let-7 maturation during neural stem-cell commitment. Nat. Cell Biol. 10, 987–993. Sampson, V.B., Rong, N.H., Han, J., Yang, Q., Aris, V., Soteropoulos, P., Petrelli, N.J., Dunn, S.P., and Krueger, L.J. (2007). MicroRNA let-7a down-regulates MYC and reverts MYC-induced growth in Burkitt lymphoma cells. Cancer Res. 67, 9762–9770.

Trippe, R., Guschina, E., Hossbach, M., Urlaub, H., Luhrmann, R., and Benecke, B.J. (2006). Identification, cloning, and functional analysis of the human U6 snRNA-specific terminal uridylyl transferase. RNA 12, 1494–1504. Viswanathan, S.R., Daley, G.Q., and Gregory, R.I. (2008). Selective blockade of microRNA processing by Lin28. Science 320, 97–100. Wang, S.W., Toda, T., MacCallum, R., Harris, A.L., and Norbury, C. (2000). Cid1, a fission yeast protein required for S-M checkpoint control when DNA polymerase delta or epsilon is inactivated. Mol. Cell. Biol. 20, 3234–3244.

Shen, B., and Goodman, H.M. (2004). Uridine addition after microRNAdirected cleavage. Science 306, 997.

Wang, X., and Yang, P. (2008). In vitro differentiation of mouse embryonic stem (mES) cells using the hanging drop method. J. Vis. Exp. 17. http://www.jove. com/index/details.stp?id=825.

Stevenson, A.L., and Norbury, C.J. (2006). The Cid1 family of non-canonical poly(A) polymerases. Yeast 23, 991–1000.

Wickens, M., and Kwak, J.E. (2008). Molecular biology. A tail tale for U. Science 319, 1344–1345.

Suh, M.R., Lee, Y., Kim, J.Y., Kim, S.K., Moon, S.H., Lee, J.Y., Cha, K.Y., Chung, H.M., Yoon, H.S., Moon, S.Y., et al. (2004). Human embryonic stem cells express a unique set of microRNAs. Dev. Biol. 270, 488–498.

Wilusz, C.J., and Wilusz, J. (2008). New ways to meet your (30 ) end oligouridylation as a step on the path to destruction. Genes Dev. 22, 1–7.

Tabara, H., Sarkissian, M., Kelly, W.G., Fleenor, J., Grishok, A., Timmons, L., Fire, A., and Mello, C.C. (1999). The rde-1 gene, RNA interference, and transposon silencing in C. elegans. Cell 99, 123–132. Thomson, J.M., Newman, M., Parker, J.S., Morin-Kensicki, E.M., Wright, T., and Hammond, S.M. (2006). Extensive post-transcriptional regulation of microRNAs and its implications for cancer. Genes Dev. 20, 2202–2207. Tomecki, R., Dmochowska, A., Gewartowski, K., Dziembowski, A., and Stepien, P.P. (2004). Identification of a novel human nuclear-encoded mitochondrial poly(A) polymerase. Nucleic Acids Res. 32, 6001–6014. Trippe, R., Richly, H., and Benecke, B.J. (2003). Biochemical characterization of a U6 small nuclear RNA-specific terminal uridylyltransferase. Eur. J. Biochem. 270, 971–980.

708 Cell 138, 696–708, August 21, 2009 ª2009 Elsevier Inc.

Wulczyn, F.G., Smirnova, L., Rybak, A., Brandt, C., Kwidzinski, E., Ninnemann, O., Strehle, M., Seiler, A., Schumacher, S., and Nitsch, R. (2007). Post-transcriptional regulation of the let-7 microRNA during neural cell specification. FASEB J. 21, 415–426. Yang, D.H., and Moss, E.G. (2003). Temporally regulated expression of Lin-28 in diverse tissues of the developing mouse. Gene Expr. Patterns 3, 719–726. Yi, R., Qin, Y., Macara, I.G., and Cullen, B.R. (2003). Exportin-5 mediates the nuclear export of pre-microRNAs and short hairpin RNAs. Genes Dev. 17, 3011–3016. Yu, J., Vodyanik, M.A., Smuga-Otto, K., Antosiewicz-Bourget, J., Frane, J.L., Tian, S., Nie, J., Jonsdottir, G.A., Ruotti, V., Stewart, R., et al. (2007). Induced pluripotent stem cell lines derived from human somatic cells. Science 318, 1917–1920.