Higher eukaryotes employ extensive post‐transcriptional gene regulation to accomplish fine control of gene expression. The
microRNA (miRNA) family plays important roles in the post‐transcriptional gene regulation of broad networks of target mRNA
expression. Most miRNAs are generated by a conserved mechanism involving two RNase III enzymes Drosha and Dicer. However,
work from the past few years has uncovered diverse noncanonical miRNA pathways, which exploit a variety of other RNA processing
enzymes. In addition, the discovery of another abundant small RNA family, endogenous short interfering RNAs (endo‐siRNAs),
has also broadened the catalogs of short regulatory RNAs. This review highlights recent studies that revealed novel small
RNA biogenesis pathways, and discusses their relevance to gene regulatory networks. WIREs RNA 2012, 3:351–368. doi: 10.1002/wrna.113
This WIREs title offers downloadable PowerPoint presentations of figures for non-profit,
educational use, provided the content is not modified and full credit is given to the author
Canonical miRNA pathway. The majority of miRNA hairpins are transcribed as a part of pol II transcripts and they can locate either on an exon or an intron. Clustered miRNAs can also produced as polycistronic transcripts bearing multiple miRNA hairpins.The microprocessor complex (Drosha‐DGCR8 complex) liberate ∼60–70 nt pre‐miRNA hairpins. Pre‐miRNAs are exported to cytoplasm by Exportin 5 (Exp5) and processed by Dicer to produce small RNA duplexes called miRNA/miRNA∗︁ duplexes. One strand from the duplex is preferentially selected and loaded to the effector protein complex, Argonaute complex.
Typical structures of canonical miRNA, mirtron and tailed mirtrons. (a) Structure of a canonical miRNA, Drosophila bantam. Pre‐miRNA hairpins are accompanied by lower stem regions that are required for recognition by microprocessor complex. Pre‐miRNA hairpins include miRNA/miRNA duplex and terminal loop regions. Terminal loops are removed by Dicer cleavage. (b) Structure of a mirtron, Drosophila mir‐1003. Mirtrons lack lower stem regions but the 5′ and 3′ ends of the hairpins are precisely located at the 5′ and 3′ exon‐intron boundaries. The structures of mirtrons suggest that splicing reaction followed by lariat debranching can produce pre‐miRNA like hairpins. (c) Structure of a 3′ tailed mirtron, Drosophila mir‐1017. The 5′ end of this mirtron hairpin locates at the 5′ end of the intron. However, unlike typical mirtrons, the 3′ reads are produced from the middle of the intron. Because the nuclear export factor Exp5 is selective for hairpins with short 3′ overhangs, this structure suggests that 3′ tail has to be removed after splicing and debranching, in order to efficiently produce the mature miRNA products. A biochemical study has shown that RNA exosome is responsible for 3′ tail removal. (d) Structure of a 5′ tailed mirtron, mouse mir‐1982. In contrast to the Drosophila mir‐1017, all known mammalian tailed mirtrons are 5′ tailed. The 3′ end of this hairpin corresponds to the 3′ end of the intron, the 5′ end is predicted to have ∼11 nt tail. 5′ overhang is known to be inhibitory for export by Exp5, the 5′ tails have to be removed for efficient export. The precise mechanism of 5′ tail removal is not known.
Drosha independent miRNA pathways. (a) The mirtron pathway. Mirtrons are pre‐miRNA like hairpin introns that are directly produced by splicing followed by lariat debranching. The debranched mirtrons are further processed by Dicer to mature miRNA molecules. Mirtrons can have tails on either of the hairpin ends. Although the mechanism that removes 5′ tails is not known, 3′ tails are removed by the RNA exosome. (b) The endo‐shRNA pathway. In mammalian cells, a group of miRNA genes called endo‐shRNAs are proposed to be transcribed by RNA polymerase III to give rise to short hairpin resembling to the artificial shRNA, and directly processed by Dicer. (c) Small regulatory RNAs from abundant noncoding RNAs. Abundant noncoding RNAs such as tRNAs and snoRNAs can also be processed by Dicer. Specific fragments of these RNA molecules are processed by Dicer and are incorporated into Argonaute complexes. (d) miRNAs from viral genomes. Murine Herpes virus 68 (MHV68) encodes several miRNAs that are processed by a Drosha independent mechanism. These miRNA hairpins are accompanied by tRNA like moiety at the 5′ ends that direct cleavage by the tRNA 3′ processing enzyme tRNase Z at the 5′ ends of the miRNAs. 3′ ends of these hairpins are likely to be determined by transcriptional termination by RNA polymerase III. Although RNA viruses does not seem to encode natural miRNAs, miRNA hairpins artificially inserted in the genome of a cytoplasmic ssRNA virus, Sindbis virus can produce active miRNA‐class small RNAs in a Drosha‐independent manner. The processing mechanism of miRNAs from cytoplasmic RNA viruses has not been characterized.
A well conserved vertebrate miRNA, miR‐451 is processed by a Dicer‐independent manner. (a) A schematic drawing of the mir‐451 processing pathway. Pri‐mir‐451 transcripts contain a short hairpin that is recognized by the Drosha/DGCR8 complex and produces a 42 nt hairpin. This hairpin is directly loaded to Ago2 complex to be cleaved at the specific position in the 3′ arm by the catalytic activity of Ago2 protein. The resulted 30 nt product is further shortened by unknown mechanism to fully mature as ∼22–24 nt miR‐451 product. (b) Structures of human, mouse and fish mir‐451 hairpin. The mature region (Green) of mir‐451 is perfectly conserved from fish to humans. The 3′ end of this 5p miRNA is extended to the opposite arm, suggesting that this may not be a dicer product. Recent studies have shown that the ∼42 nt short hairpin is produced by Drosha cleavage and the slicer Argonaute, Ago2 cleaves the middle of 3p arm of this hairpin to produce the 30 nt products. The 3′ ends of slicing products are trimmed by an unknown nuclease to mature as a ∼22–24 nt miRNA.
Mechanisms of siRNA biogenesis in mammalian oocytes and Drosophila (a), C. elegans (b), S. pombe (c), and Neurospora (d). Dicer and related enzymes are shown in green, RNA dependent RNA Polymerases (RdRPs) in pink, exonucleases in light blue, Helicases in yellow and Argonautes in orange. (a) Endo‐siRNAs are generated from dsRNA molecules in mouse oocytes and insect cells. The protein components in the fly system are shown. Dcr‐2 and an isoform of dsRNA binding protein Loquacious (Loqs‐PD) process dsRNAs to generate 21 nt siRNAs, and the siRNAs are loaded to AGO2 complex to regulate mRNAs and TEs. Loading of endo‐siRNAs is dependent on Dicer‐2/R2D2 complex. Trans‐NAT siRNAs have not been identified in fly cells. The molecular mechanism of mammalian endo‐siRNA pathway is not well understood but Dicer is essential for endo‐siRNA production and mature siRNA products are loaded at least to Ago2. It is not clear whether loading is dependent on Dicer complex in mammalian cells. (b) In C. elegans, the ERI complex containing the Dicer (DCR‐1), the RdRP (RRF‐3) and the exonuclease (ERI‐1) produces ∼26 nt primary siRNAs called 26G RNAs that are 5′ monophosphorylated and enriched with 5′ Guanine. 26G RNAs are loaded to the primary Argonaute proteins ERGO1 or ALG‐3/‐4. Exogenous siRNAs are loaded into the RNAi Argonaute RDE‐1. These primary siRNA complexes recruit the RdRP complex containing the RdRP EGO‐1 or RRF‐1, the RNA helicase DRH‐3 and the Tudor domain protein EKL‐1 and initiate production of ∼22 nt secondary siRNAs called 22G RNAs against target mRNAs. 22G RNAs are triphosphorylated at their 5′ ends and their 5′ nucleotides are Guanines. These 22G RNAs are loaded into the worm specific Argonautes CSR‐1 or one of WAGO class Argonaute proteins. (c) In S. pombe, degradation products of transcripts from the repeat sequences (dg and dh) are loaded to Ago1 complex to mature as primal RNAs (priRNAs). This priRNA complex triggers generation of antisense RNAs by RdRP complex (RDRC) to produce dicer substrate dsRNAs and to amplify siRNAs from bidirectionally transcribed repeats. (d) In Neurospora crassa, DNA damage can induce aberrant RNA production from rDNA repeats by an RNA/ssDNA dependent RNA polymerase QDE‐1. The RecQ related helicase QDE‐3 is required for aberrant RNA production. Aberrant transcripts are processed by the Dicer‐like proteins DCL‐1/2 to produce mature qiRNAs loaded into the Argonaute QDE‐2. Several different hairpins also produce QDE‐2 interacting small RNAs, called milRNAs (microRNA‐like RNAs). Each milRNA exhibits distinct dependency on the small RNA biogenesis factors such as the RNase III enzymes DCL‐1/2 and MRPL3 and the exonuclease QIP.