MicroRNA genes are transcribed by RNA polymerase II Yoontae Lee1, Minju Kim1, Jinju Han1, Kyu-Hyun Yeom1, Sanghyuk Lee2, Sung Hee Baek1 and V Narry Kim1,* 1 Institute of Molecular Biology and Genetics and School of Biological Science, Seoul National University, Seoul, Korea and 2 Division of Molecular Life Sciences, Ewha Womans University, Seoul, Korea MicroRNAs (miRNAs) constitute a large family of noncod- ing RNAs that function as guide molecules in diverse gene silencing pathways. Current efforts are focused on the regulatory function of miRNAs, while little is known about how these unusual genes themselves are regulated. Here we present the first direct evidence that miRNA genes are transcribed by RNA polymerase II (pol II). The primary miRNA transcripts (pri-miRNAs) contain cap structures as well as poly(A) tails, which are the unique properties of class II gene transcripts. The treatment of human cells with a-amanitin decreased the level of pri-miRNAs at a concentration that selectively inhibits pol II activity. Furthermore, chromatin immunoprecipitation analyses show that pol II is physically associated with a miRNA promoter. We also describe, for the first time, the detailed structure of a miRNA gene by determining the promoter and the terminator of mir-23aB27aB24-2. These data indicate that pol II is the main, if not the only, RNA polymerase for miRNA gene transcription. Our study of- fers a basis for understanding the structure and regulation of miRNA genes. The EMBO Journal (2004) 23, 4051���4060. doi:10.1038/ sj.emboj.7600385 Published online 16 September 2004 Subject Categories: chromatin & transcription RNA Keywords: cap Drosha microRNA promoter transcription Introduction Hundreds of microRNA (miRNA) genes have been found in animals, plants, and viruses (Bartel, 2004 Murchison and Hannon, 2004 Pfeffer et al, 2004), making them one of the largest gene families miRNAs are defined as single-stranded RNAs B22 nt in length (ranging 19���25 nt) generated from endogenous transcripts that can form local hairpin structures in silico (Ambros et al, 2003). MiRNAs act as guide mole- cules, by base-pairing with the target mRNAs leading to translational repression and/or mRNA cleavage. Recent studies revealed the key roles of miRNAs in diverse regulatory pathways, including development timing control, hematopoietic cell differentiation, apoptosis, cell prolifera- tion, and organ development (Bartel, 2004). MiRNAs and their targets seem to constitute remarkably complex regula- tory networks since a single miRNA can bind to and regulate many different mRNA targets and, conversely, several differ- ent miRNAs can bind to and cooperatively control a single mRNA target (Lewis et al, 2003). To dissect these complex networks operated by miRNAs, it would be critical to under- stand how miRNA genes themselves are regulated. The majority of miRNA genes are located in intergenic regions or in antisense orientation to annotated genes (Lagos- Quintana et al, 2001 Lau et al, 2001 Lee and Ambros, 2001 Mourelatos et al, 2002), indicating that they form indepen- dent transcription units (Lee et al, 2002). Most of the other miRNA genes are found in intronic regions, which may be transcribed as part of the annotated genes. In animals, miRNAs are transcribed as long primary transcripts (pri- miRNAs), which are cropped into the hairpin-shaped pre- miRNAs by nuclear RNase III Drosha (Lee et al, 2003 Kim, 2004). This cleavage event is important because it predeter- mines mature miRNA sequence and generates optimal sub- strate for the subsequent events (Lee et al, 2003 Lund et al, 2004). The processing intermediate, pre-miRNA, is exported out of the nucleus by exportin-5 (Exp5), member of the Ran- dependent nuclear transport receptor family (Yi et al, 2003 Bohnsack et al, 2004 Lund et al, 2004). Pre-miRNA is subsequently cleaved by cytoplasmic RNase III Dicer (Bernstein et al, 2001 Grishok et al, 2001 Hutvagner et al, 2001 Ketting et al, 2001 Knight and Bass, 2001) into B22-nt miRNA duplex. One strand of this short-lived duplex is degraded by an unknown nuclease, while the other strand remains as a mature miRNA. Which strand to select is determined by the relative internal stability of the two ends of the duplex, that is, the strand with the less stable 50 end (for instance, G:U pair versus G:C pair) usually survives (Khvorova et al, 2003 Schwarz et al, 2003). Our knowledge on miRNA biogenesis has been signifi- cantly advanced in recent years. However, little is known about transcription of miRNA genes although it is likely to be the key regulatory step in miRNA biogenesis. To understand the mechanism of miRNA gene regulation, the basic machin- ery for miRNA transcription needs first to be identified. RNA polymerase III (pol III) was initially believed to mediate miRNA transcription because it transcribes most small RNAs such as tRNAs and U6 snRNA. However, several circumstantial evidences suggest otherwise. Firstly, pri- miRNAs are sometimes over several kilobases long and contain stretches of more than four U���s, which would have terminated transcription by pol III (Lee et al, 2002). In addition, a number of chimeric transcripts containing miRNA sequences and pieces of adjacent mRNAs have been found from EST analyses (Smalheiser, 2003). Many of these ESTs contain poly(A) tails and are occasionally spliced, suggesting that these transcripts are produced by RNA polymerase II (pol II), although it is not clear from this Received: 29 June 2004 accepted: 9 August 2004 published online: 16 September 2004 *Corresponding author. Institute of Molecular Biology and Genetics and School of Biological Science, Seoul National University, Seoul 151-742, Korea. Tel.: �� 82 2 887 8734 Fax: �� 82 2 875 0907 E-mail: narrykim@snu.ac.kr The EMBO Journal (2004) 23, 4051���4060 | & 2004 European Molecular Biology Organization | All Rights Reserved 0261-4189/04 www.embojournal.org & 2004 European Molecular Biology Organization The EMBO Journal VOL 23 | NO 20 | 2004 EMBO THE E M B O JOURNAL THE EMBO JOURNAL 4051

computational analysis whether these chimeric transcripts are indeed involved in miRNA biogenesis. Secondly, the expression profiles of miRNAs suggest the class II gene-like control over miRNA genes. Insertion of a pol II enhancer can induce miRNA in the case of bantem RNA in Drosophila (Brennecke et al, 2003). Temporal regulation of let-7 RNA in Caenorhabditis elegans is dependent on an enhancer element, termed temporal regulatory element (TRE) (Johnson et al, 2003). Thirdly, the analysis of the noncoding RNAs contain- ing miRNA sequence (miR-155 (BIC) and miR-172 (EAT)) showed that they are polyadenylated and spliced (Tam, 2001 Aukerman and Sakai, 2003). The miR-172 precursor, EAT, in Arabidopsis was suggested to contain the cap structure because the RACE protocol used in the study favors capped RNAs (Aukerman and Sakai, 2003). However, direct experimental evidence is still missing as to (1) whether pol II is indeed responsible for miRNA gene transcription and (2) how miRNA genes are structured. Here we present the first direct evidence that miRNA genes are transcribed by pol II. Results In order to address these questions directly, we decided to analyze the 50- and 30-end structures of various pri-miRNAs. First, RNAs containing 7-methyl guanosine cap were selec- tively enriched from total RNA by affinity purification using eIF4E, the high-affinity cap-binding protein, fused to gluta- thione-S-transferase (GST) (Figure 1). As a control, hnRNP A1, which is a generic pre-mRNA-binding protein, was included (Burd and Dreyfuss, 1994). RNase protection assay (RPA) was carried out then to detect pri-miR-23a from the bound (B) and unbound (UB) fractions. The fragment of 69 nt representing pri-miR-23a was detected from the GST-eIF4E- bound fraction but not from the GST-hnRNP A1-bound frac- tion (Figure 1A). Pre-miR-23a and mature miR-23a was not retained in the eIF4E-bound fraction, indicating that only pri- miR-23a, but not its processed products, contains the cap structure. To test whether the 50 cap is a general property of pri- miRNAs, we randomly chose seven different pri-miRNAs (miR-23aB27aB24-2, miR-30a, let-7a-1, let-7a-3, miR-21, miR-17B18B19aB20B19b-1, and miR-15aB16-1) and ana- lyzed them by RT���PCR (Figure 1B). Their genomic loci relative to the closest annotated genes are shown in Supplementary figure. Reverse transcription was carried out with gene-specific primers instead of oligo-dT primer. GAPDH pre-mRNA, 45S pre-rRNA, and pre-tRNAtyr were amplified as controls representing transcripts produced by RNA pol II, I, and III, respectively. All the pri-miRNAs tested in this study showed affinity to the GST-eIF4E column, indicating that they contain the cap structures. The affinities appear to vary, with pri-miR-15aB16-1 being the lowest. Assuming that all the GAPDH pre-mRNA molecules are capped, approximately 5���50% of pri-miRNAs seem to contain the cap structures, which implicates that additional processing event(s) may exist prior to Drosha-mediated cropping step and removes the cap from the primary transcript. A similar approach was then taken to verify the presence of poly(A) tail at the 30 end of pri-miRNA (Figure 2). Following selective enrichment of polyadenylated RNA using oligo-d(T) cellulose beads, RNA was extracted and analyzed by RPA (Figure 2A) and RT���PCR (Figure 2B). The results show that all of the pri-miRNAs tested in our study contain poly(A) tails although some of these signals may have come from internal priming at A stretches in the transcripts. 7 ��� pri-miR-23a pre-miR-23a Mature miR-23a M 1 2 3 4 5 6 Input (100%) B UB B UB RNase T1: + + + + + + GST- eIF4E GST- hnRNP A1 Yeast RNA Probe B 1 2 UB (100%) GST-eIF4E B (100%) pri-miR-21 GAPDH pre-mRNA pri-let-7a-1 pre-tRNAtyr pri-let-7a-3 pri-miR-30a pri-miR-23a~27a~24-2 pri-miR-15a~16-1 45S pre-rRNA pri-miR-17~18~19a~20~19b-1 A Figure 1 Pri-miRNAs contain the 50 cap structures. (A) Affinity purification of cap-containing RNA followed by RPA. Total RNA from HeLa cells was used for selective enrichment using the cap-binding protein eIF4E. RNA was extracted from the bound (B) or unbound (UB) fraction. Because the whole RNA from each fraction was used for the assay without further normalization, it corresponds to 100% of the input. (B) RT���PCR demonstrating that cap is present not only in pri-miR-23aB27aB24-2 but also in pri-let-7a-1, pri-let-7a-3, pri-miR-30a, pri-miR-21, pri-miR- 17B18B19aB20B19b-1, and pri-miR-15aB16-1. GAPDH pre- mRNA, 45S pre-rRNA, and pre-tRNAtyr were used as controls. The whole RNA extracted from each fraction (100% of the bound (B) or 100% of the unbound (UB) fraction) was used for the assay. MicroRNA gene structure and transcription Y Lee et al The EMBO Journal VOL 23 | NO 20 | 2004 & 2004 European Molecular Biology Organization 4052