As sequence-specific guides in RNA silencing in all eukaryotes, small RNAs play important and diverse roles in many biological processes (reviewed in 1, 2, 3, 4, 5). Small RNAs are classified into three major types, microRNAs (miRNAs), small interfering RNAs (siRNAs) and Piwi-interacting RNAs (piRNAs) based on differences in their precursors, biogenesis and protein partners 1, 4. miRNAs are produced from MIR gene transcripts that form stem-loop structures 1, 4, 6. Mature miRNAs are usually 21-24-nt-long and are generated through the processing of the stem-loop precursors by RNase III enzymes (Drosha and Dicer in animals; DICER-LIKE (DCL) in plants) 4, 6. miRNAs target endogenous genes for mRNA cleavage, decay and/or translational repression 7. siRNAs are derived from double-stranded RNA (dsRNA) precursors generated through sense and antisense transcription, transcription of inverted repeat elements, viral replication, or RNA-dependent RNA polymerase (RDR) activity that converts single-stranded RNAs into dsRNAs 2, 4, 8. siRNAs are 21-24-nt-long and their biogenesis also requires the endonuclease activity of Dicer or DCL 2, 4. siRNAs repress the expression of transposable elements (TEs) and transgenes, and mediate the cleavage of viral mRNAs 9, 10. piRNAs, found specifically in animals, are usually 24-32-nt-long and processed from presumably single-stranded RNA precursors in a Dicer-independent manner (11; reviewed in 12). Their biogenesis is composed of a primary processing pathway and a ping-pong amplification pathway 13, 14. piRNAs are derived from TEs, intergenic regions, and certain genes. piRNAs derived from TEs silence the TEs through the cleavage of TE-derived transcripts or DNA methylation of the genomic loci to cause transcriptional silencing 13, 14, 15, 16. While miRNAs and siRNAs are bound by the Argonaute (Ago) sub-clade of Ago proteins, piRNAs associate specifically with the Piwi sub-clade of Ago proteins 12.

Small RNA stabilization by methylation

The three types of small RNAs also vary in the requirement for the 2′-O-methyl modification on the ribose of their 3′ termini. HUA ENHANCER 1 (HEN1) was first identified in Arabidopsis as a small RNA methyltransferase (MTase) that methylates miRNA and siRNA duplexes 17, 18. HEN1 homologs were found to methylate small RNAs in other plants, piRNAs in animals and Ago2-associated small RNAs in Drosophila 11, 19, 20, 21, 22, 23, 24, 25. In general, 2′-O-methylation serves as a protective mechanism against 3′-5′ degradation and 3′ uridylation of small RNAs 17, 21, 26.

Plant miRNAs and siRNAs are methylated by HEN1

In plants, the biogenesis of small RNAs, including miRNAs and all types of siRNAs, involves 2′-O-methylation at their 3′ termini 17. The requirement for methylation was uncovered through the isolation of Arabidopsis hen1 mutants. miRNAs and siRNAs are methylated at their 3′ termini in wild-type Arabidopsis, but not in hen1 mutants 17 (Figure 1A). In hen1 mutants, miRNAs show reduction in abundance and heterogeneity in sizes 17, 26, the latter of which was reflected by a ladder of bands when total RNAs were resolved on high-resolution acrylamide gels and hybridized with a probe for specific miRNA species. Because primer extension studies demonstrated that the heterogeneous species have identical 5′ ends, the ladder of signals probably represented heterogeneity at the 3′ ends 26. Indeed, when particular miRNAs were cloned and sequenced from wild type and a hen1 mutant, it became clear that miRNAs tended to acquire an oligonucleotide tail enriched for U residues in the hen1 mutant 26 (Figure 2). The process of tailing miRNAs with uridine as the preferential nucleotide will be referred to as uridylation. miRNAs also become truncated from their 3′ ends in the hen1 mutant (Figure 2), and truncated miRNAs can also be uridylated 26. The uridylation of the unmethylated small RNAs perhaps leads to their 3′-5′ degradation, but this still needs to be confirmed genetically or biochemically. WAVY LEAF1 (WAF1), a HEN1 ortholog in rice, is also essential for the stabilization of small RNAs. The accumulation of miRNAs and trans-acting siRNAs, a type of endogenous siRNAs, is greatly reduced in waf1 mutants 27.

Figure 1
figure 1

2′-O-methylation of small RNAs by HEN1. (A) Plant HEN1 methylates the 3′ termini of both strands of small RNA duplexes. (B) Drosophila Hen1 methylates Ago2-associated siRNAs, but not Ago1-associated miRNAs. (C) Animal Hen1 methylates piRNAs in germ cells. The ovals represent different Argonaute proteins.

Figure 2
figure 2

Methylation protects small RNAs from 3′ uridylation and truncation that lead to the degradation of the small RNAs.

Phylogenetic studies showed that the N-terminal three-fourth of Arabidopsis HEN1 is conserved only among the plant homologs, and the C-terminal domain, which constitutes approximately one-fourth of HEN1 and contains a recognizable S-adenosyl methionine (AdoMet)-binding motif 17, is conserved among many bacterial, fungal, and metazoan homologs 28 (Table 1). Bioinformatics analysis revealed that the MTase catalytic domain of HEN1 is closely related to small molecule MTases 28. In vitro activity assays showed that recombinant Arabidopsis HEN1 acts on 21-24-nt small RNA duplexes and deposits a methyl group onto the 2′ OH of the 3′-terminal nucleotides of each strand 18. The 2-nt overhang of the duplex, and the 2′ and 3′ OH of the 3′ nucleotide are two important features in the substrates of Arabidopsis HEN1 18.

Table 1 HEN1 homologs that methylate small RNAs

The mechanism of substrate recognition and methylation by HEN1 was revealed by the crystal structure of full-length HEN1 from Arabidopsis in complex with a 22-nt small RNA duplex. Arabidopsis HEN1 binds to the small RNA duplex substrate as a monomer 29. The plant-specific N-terminal portion of HEN1 contains two dsRNA-binding domains (dsRBD1 and dsRBD2) and a La-motif-containing domain (LCD) 28, 29. Both dsRBDs are involved in substrate recognition, and the substrate length specificity is determined by the distance between the MTase domain and the LCD, each interacting with one end of the small RNA duplex 29. HEN1 methylates the 2′ OH of the 3′-terminal nucleotide in a Mg2+-dependent manner 29. Kinetic studies illustrated that HEN1 is catalytically efficient in the absence of any supplementary proteins. The enzyme modifies individual strands in succession to complete the methylation of the duplex 30. Since the C-terminal portion of HEN1 (residues 666-942) efficiently modifies small RNA duplexes in vitro but exhibits weaker affinities for both small RNA duplexes and AdoMet (the methyl donor), the function of the N-terminal portion is to stabilize the catalytic complex 30.

Small RNA methylation by HEN1 in plants could be modulated in vivo. Several RNA-silencing suppressors from plant viruses, such as the Beet yellows virus 21 kDa protein (p21), the Tomato bushy stunt virus 19 kDa protein (p19) and the Turnip mosaic virus-silencing suppressor P1/HC-Pro interfere with small RNA methylation by HEN1 31, 32, 33. Interestingly, miRNA methylation defects in the weak hen1-2 mutant can be partially suppressed by mutations in the endogenous 24 nt siRNA biogenesis pathway 34. The suppression suggests that siRNAs compete with miRNAs for methylation when HEN1 function is partially compromised in Arabidopsis 34. In addition, differences in phenotypic severity of the same hen1 lesion in Col versus Ler accessions implicate the existence of a negative modifier of HEN1 activity or small RNA activity in the Columbia background 34.

Animal piRNAs and fly siRNAs are methylated by HEN1 homologs

piRNAs are expressed in animal germlines and guide Piwi proteins to silence TEs (reviewed in 12). piRNAs and Drosophila Ago2-associated siRNAs are 2′-O-methylated at their 3′ termini by HEN1 homologs in the animal kingdom 19, 21, 23, 25 (Figure 1B and 1C). Animal Hen1 proteins are less than half the size of the plant HEN1 proteins: they lack the plant-specific N-terminal portion 19, 21, 22, 23, 25, 35. Most animal Hen1 proteins have a similar accumulation pattern as the Piwi proteins 21, 23, 25. For example, mouse Hen1 is present specifically in testis as are piRNAs and Piwi proteins. Recombinant mouse Hen1 methylates single-stranded piRNAs in vitro 22, 23. The zebrafish Hen1 gene is expressed in both female and male germ lines, but is essential only for oocyte development and dispensable for testis development. The zebrafish Hen1 protein is localized to nuage, a germ cell-specific structure where Ziwi (one of the Piwi proteins in zebrafish) is localized 21. In the testes of hen1 mutants, retrotransposon-derived piRNAs become uridylated and reduced in abundance, and the retrotransposons are mildly de-repressed. Therefore, Hen1-mediated methylation stabilizes piRNAs for transposon silencing in ovary and testis 21.

piRNAs and Ago2-associated siRNAs in Drosophila are 2′-O-methylated at their 3′ termini 11, 19, 25. Saito et al. 25 showed that loss of Pimet (piRNA MTase), the HEN1 homolog in Drosophila, results in the loss of 2′-O-methylation of piRNAs. Recombinant Pimet methylates single-stranded small RNA oligos in vitro and physically interacts with the Piwi protein 25. The same HEN1 homolog in Drosophila was named DmHen1 by Horwich et al 19. DmHen1 methylates the 3′ termini of Piwi-associated piRNAs and Ago2-associated siRNAs, but not Ago1-associated miRNAs. Without DmHen1, the length and abundance of piRNAs are decreased, and the levels of piRNA targets are elevated 19. Horwich et al. 19 further showed that methylation occurs on single-stranded siRNAs in association with Ago2, which is also likely to be the case for piRNAs in vivo. In Drosophila, 2′-O-methylation of Ago2-associated endo-siRNAs protects the siRNAs from 3′ tailing and trimming 36. In the absence of Hen1 activities, endo-siRNAs generally become 3′ truncated and/or acquire a 3′ tail. The most abundant nucleotide added is uridine, followed by adenine. Long tails are rare but are nearly always uridine tails 36 (Figure 2).

A HEN1 homolog is also present in the ciliated protozoan Tetrahymena thermophila, which contains two classes of small RNAs 35, 37. The 28-29 nt scan (scn) RNAs are produced during sexual reproduction, and bind specifically to the Ago protein, Twi1p 38. The dsRNA precursors of scnRNAs are generated from bidirectional transcription of non-coding DNA 39 and processed by the Dicer-like protein, Dcl1p 40, 41. The 23-24 nt RNAs may also be generated from dsRNA precursors and map to the genome in clusters that are antisense to predicted genes 37. The scnRNAs are methylated by Hen1p (the Tetrahymena HEN1 homolog), which is co-expressed with Twi1p in vivo and physically interacts with Twi1p in vitro 35. Loss of Hen1p causes a reduction in the abundance and length of scnRNAs, defects in programmed DNA elimination, and inefficient production of sexual progeny 35. Recombinant Hen1p methylates single-stranded scnRNAs in vitro. Therefore, the stabilization of scnRNAs by Hen1p-mediated 2′-O-methylation ensures DNA elimination in Tetrahymena 35. Intriguingly, approximately half of the 23-24 nt sRNAs that resemble siRNAs from other organisms undergo non-templated 3′ nucleotide addition, with a single uridine as the most common addition 37. Is this reminiscent of the 3′ tailing of unmethylated Drosophila endo-siRNAs? The effects of uridylation on the accumulation and activity of these 23-24 nt sRNAs await further investigation.

Damaged bacterial RNAs are methylated by HEN1 in vitro

Bacterial Hen1 is encoded by a two-gene operon that also encodes polynucleotide kinase-phosphatase (Pnkp), an RNA repair enzyme 42, 43. Bacterial Hen1 methylates the 3′-terminal nucleotide of the 5′ fragment of a broken tRNA in a Mn2+-dependent manner to protect it against further damage by a transesterifying endonuclease; then the Pnkp in the Hen1-Pnkp complex repairs the RNA substrate by end healing and sealing 43, 44. The bacterial Hen1 MTase domain contains a core fold shared by other RNA and DNA MTases as well as motifs unique to bacterial Hen1 homologs. The interaction between bacterial Hen1 and their RNA substrates is likely similar to that of their eukaryotic counterparts 43.

Small RNA turnover mechanisms

Loss of HEN1 activity causes reduced accumulation of naturally methylated small RNAs, and studies of hen1 mutants helped uncover previously unknown mechanisms that degrade small RNAs 17, 21, 26, 35, 36, 45. These studies suggest that 3′ uridylation and 3′-5′ exonucleolytic degradation are two major mechanisms that turnover small RNAs when they lose methylation (Figure 2). Studies on naturally unmethylated small RNAs, such as miRNAs in animals, further support that 3′ uridylation generally marks small RNAs for degradation and reveal other 3′ tailing events that influence small RNA stability (Figure 3A). Moreover, 3′-5′ exonucleolytic degradation has been found to be a major mechanism to turnover mature small RNAs across organisms, although a 5′-3′ exonuclease was found to degrade miRNAs in Caenorhabditis elegans.

Figure 3
figure 3

Mechanisms influencing small RNA stability. (A) 3′ nucleotide addition affects the stability of small RNAs. While U-tails probably lead to small RNA degradation, 3′ adenylation may have a protective role against degradation or have no effects on the stability of small RNAs. (B) Target transcripts can either enhance or suppress the degradation of the corresponding small RNAs in different situations. (C) Protein factors can stabilize or destabilize RISCs to influence the accessibility of nucleases to small RNAs. Yellow shapes indicate 3′-to-5′ exonucleases, orange shapes represent 5′-to-3′ exonucleases, and the purple oval and the green pentagon indicate RISC-interacting proteins.

3′ nucleotide modification affects the stability of small RNAs

Methylation protects plant small RNAs, animal piRNAs and fly Ago2-associated siRNAs against 3′ nucleotide addition 17, 21, 26, 35, 36, which probably stimulates a 3′-5′ exonucleolytic activity to degrade the tagged small RNAs. However, for these naturally methylated small RNAs, the nucleotidyl transferases modifying the small RNAs when they lose methylation have not yet been identified in most organisms (Table 2), such that the role of uridylation in small RNA turnover has not been genetically evaluated. Once these enzymes are identified, it would be interesting to determine whether they impact small RNA metabolism under the normal condition when HEN1 function is present.

Table 2 Nucleotidyl transferases affecting mature small RNAs

The characterization of a Chlamydomonas reinhardtii mutant (Mut-68) has shed light on the mechanisms of uridylation and degradation of miRNAs and siRNAs. In the alga C. reinhardtii, siRNAs and miRNAs are 2′-O-methylated at their 3′ termini as in angiosperms 46. MUT68, a terminal nucleotidyl transferase, was initially shown to add adenines to the 5′ cleavage fragments from mRNAs targeted by RNA-induced silencing complexes (RISCs) to lead to their efficient decay 47. MUT68 cooperates with the cytoplasmic exosome to degrade RISC-generated 5′ RNA cleavage products 47, and cooperates with the peripheral exosome subunit RRP6 to degrade small RNAs 48. In a Mut-68 mutant and an RRP6 depletion line, the levels of a subset of miRNAs and siRNAs are elevated 48. MUT68 plays a role in the uridylation of the 3′ ends of small RNAs in vivo and stimulates their degradation by RRP6 in vitro 48. MUT68 and RRP6 were unable to use a 2′-O-methylated miR912 oligonucleotide as a substrate in vitro, indicating that MUT68 only acts on unmethylated small RNAs 48. However, it is not clear if MUT68 and RRP6 compose the downstream pathway, degrading unmethylated small RNAs in hen1, since it has not been tested yet whether MUT68 loss of function or RRP6 depletion suppresses the molecular defects associated with HEN1 loss of function.

A 3′-nucleotide addition of small RNAs has been observed on a global scale in high throughput sequencing studies in different model organisms 49, 50, 51. The 3′ nucleotide addition influences small RNA stability and function differently, depending on the specific small RNAs examined and the nature of the added nucleotides (Figure 3A). A recent study showed that 3′ nucleotide addition to miRNAs is a common posttranscriptional event that exhibits selectivity for specific miRNAs ranging from C. elegans to humans 51. The modifications are predominantly mono-adenylation or mono-uridylation across different tissue types, disease states, and developmental stages 51. Multiple enzymes, including MTPAP, PAPD4, PAPD5, ZCCHC6, ZCCHC11, and TUT1, were found to be responsible for the 3′-nucleotide addition in an miRNA-specific manner 51.

Uridylation usually enhances the decay of the tailed small RNAs (Figure 3A). For example, in C. elegans, uridylation of the naturally unmethylated siRNAs in the CSR-1 pathway restricts these siRNAs to CSR-1, an Ago that is not involved in regulating mRNA levels but is required for proper chromosome segregation 52, 53, 54. Loss of CDE-1, the nucleotidyl transferase that exerts the uridylation activity, leads to an increase in the levels of CSR-1 siRNAs, defective chromosome segregation in mitosis and meiosis, and erroneous gene silencing 54. Therefore, 3′ uridylation regulates the accumulation of the CSR-1 siRNAs by destabilizing them.

Uridylation can be triggered by the interaction with target RNAs (Figure 3B). One implication comes from the uridylation of piRNAs derived from RNA-based but not DNA-based TEs in a zebrafish hen1 mutant 21. In Drosophila, miRNAs typically associate with Ago1 to repress the translation and/or cause the decay of their target mRNAs 55. miRNAs are complementary to their targets only at the seed sequence region, with few additional base pairs tethering the two RNAs together 56. The abundance of the corresponding miRNAs decreases when transgenes containing one or more highly complementary sites are introduced 36. Extensive base pairing between Ago1-associated miRNAs and their target mRNAs triggers miRNA 3′ tailing and trimming that could result from the activities of a terminal nucleotidyl transferase and a 3′-5′ exonuclease, respectively. Consistently, structural studies on the Thermus thermophiles Ago in complex with the guide sequence and the target strands of varying lengths show that the presence of a long, but not a short, stretch of sequence complementarity between the guide and the target induces the 3′ end of the guide to be dislodged from the Ago PAZ domain 57. This would presumably allow a nucleotidyl transferase or an exonuclease to access the 3′ end of the small RNA. The nucleotidyl transferase and the exonuclease acting upon Ago1-associated miRNAs maybe the same as those acting upon Ago2-bound, unmethylated siRNAs in the hen1 mutant 36. In fact, tailing and truncation of siRNAs in the Drosophila hen1 mutant may also be induced by the interactions with their endogenous target mRNAs, which are highly complementary to the siRNAs.

Target-dependent small RNA turnover is conserved between flies and mammals. Highly complementary targets also induce 3′ trimming and tailing of miRNAs in human cells, which is demonstrated by both synthetic target RNA analogs, “antagomirs”, and in vitro transcribed-target mRNAs containing sites fully complementary to miRNAs 36. A kinetic analysis of miR-233 decay in human cells shows that the presence of target transcripts, especially those with perfect complementarity to the miRNA, increased the decay of miRNA-233 58. Deep sequencing unveiled that the frequency of miR-233 with 3′ mono-uridylation increased during decay, suggesting that the addition of a uridine may be a destabilizing modification 58. The frequency of miR-233 with one or two adenines added to its 3′ end was also elevated in the cells co-expressing a target with mismatches. However, the levels of these variants did not change during decay 58.

Adenylation usually has no effect on the stability of modified small RNAs, or increase their stability (Figure 3A). For example, mature miR-122 was shown to be adenylated by the RNA nucleotidyl transferase PAPD4/GLD-2 in humans and mice, which resulted in an increase in the stability of this miRNA 59. Considering the low in vitro specificity of PAPD4/GLD-2, either some specificity factors recruit PAPD4/GLD-2 to the 3′ terminus of miR-122 or a broader set of miRNAs is regulated by PAPD4/GLD-2 through adenylation in vivo 59. In Populus trichocarpa, adenylation was observed for both full length and truncated miRNAs 60. An in vitro decay assay showed that replacement of the natural 3′ nucleotide with an adenine resulted in a slower miRNA degradation in P. trichocarpa extracts 60.

These observations imply that 3′ nucleotide addition, predominantly uridylation and adenylation, is a widespread modification on small RNAs. Interestingly, the A- or U-tails in animals are usually only 1- or 2-nt-long, whereas plant miRNAs acquire a tail of 1-7 nt in a hen1 mutant. Uridylation and adenylation on other RNAs, such as U6 RNA and 5S rRNA, occur competitively, and a single adenylation prevents further oligouridylation 61. It has not been tested whether this is true for small RNAs in plants or animals.

The stability and activities of miRNAs can be regulated through active 3′ trimming during the maturation of miRNAs in RISCs 62, 63. In Drosophila, many miRNAs undergo 3′ trimming by the 3′-5′ exoribonuclease Nibbler after the miRNA* strands have been removed from Ago1. Depletion of Nibbler results in reduced miRNA activities as well as the accumulation of longer miRNA isoforms. This suggests that Nibbler renders miRNAs smaller and perhaps a better fit for Ago1 to enhance their activities. Intriguingly, depletion of Nibbler in S2 cells results in miR-34 species that are longer than the lengths of Dicer products 62, suggesting that the untrimmed species are subjected to 3′ tailing and perhaps degradation. It is likely that the active 3′ trimming of miRNAs is conserved in other organisms because both C. elegans and humans have homologs of Nibbler, and miRNAs in humans also exhibit 3′ heterogeneity 62. Given that Ago1-associated miRNAs are accessible to Nibbler in Drosophila, other exonucleases or terminal nucleotidyl transferases may also be able to modify small RNAs when they are in association with Ago proteins.

Exonucleolytic degradation of small RNAs

The discovery of a small RNA 3′ truncation activity in Arabidopsis hen1 mutants prompted a search for the 3′-5′ exonuclease responsible for this activity. A family of 3′-5′ exoribonucleases named small RNA-degrading nuclease (SDN) was found to turnover mature miRNAs and siRNAs in Arabidopsis 64. In vitro assays showed that SDN1 acts specifically on single-stranded small RNA oligonucleotides that are longer than 8 nt, and is deterred but not completely inhibited by the 2′-O-methyl modification on the 3′ termini of small RNAs 64. There are four closely related SDN family exoribonucleases sharing overlapping functions in Arabidopsis; only simultaneous knockdown of multiple SDN genes results in elevated miRNA levels in vivo and pleiotropic developmental defects 64. Interestingly, the efficiency for SDN1 to degrade U-tailed miR173 is lower than that for miR173 in vitro, which suggests that SDN1 may not be the enzyme that degrades the uridylated small RNAs in the absence of HEN1 activity 64. However, it cannot be excluded that the SDN proteins degrade U-tailed miRNAs in vivo with the assistance of other proteins. Whether SDNs are responsible for the 3′ truncation of small RNAs in hen1 mutants has yet to be evaluated by knocking down multiple SDN genes in hen1 mutants and observing the status of miRNA 3′ truncation in this background. Another pressing question is whether SDNs degrade small RNAs released from RISCs or small RNAs bound by Ago proteins. SDN homologs are universally present in eukaryotes, but whether they exert similar functions in animals is currently unknown.

XRN-2 was found to affect miRNA homeostasis in C. elegans by promoting the dislodging of miRNAs from RISC and degrading the free miRNAs through its 5′-3′ exoribonucleolytic activity 65 (Figure 3C). miRNAs are efficiently released from the RISC complex and degraded by XRN-2 in a C. elegans larval lysate 65. The presence of wild-type target mRNAs appears to stabilize miRNAs in C. elegans. Both the release and the degradation of miRNAs can be blocked by adding miRNA target mRNAs. In vivo, the steady-state miRNA levels are increased in the presence of target mRNAs 65. However, it is currently unknown how the dislodging of miRNAs from RISC by XRN-2 is achieved or how target mRNAs stabilize miRISCs. It is also not known whether the XRN-2-dependent turnover and target mRNA-mediated stabilization of miRNAs exist in other organisms. The plant homologs of XRN-2 are involved in miRNA biogenesis in that they degrade the non-miRNA fragments, which are by-products of pri-miRNA processing, but do not affect mature miRNA levels 66.

Both exonucleases mentioned above (SDN in Arabidopsis and XRN-2 in C. elegans) act on mature, single-stranded small RNAs. C. elegans Eri-1 represses RNA-mediated silencing by degrading siRNA duplexes 67. eri-1 encodes a conserved protein containing nucleic acid-binding and exonuclease domains 67.

Additional factors that influence small RNA stability

The life spans of small RNAs in an organism are not homogenous or constant. There are mechanisms that influence the stability of individual small RNAs. Factors that influence small RNA stability include target transcripts, cis-acting elements and trans-acting proteins.

Target RNAs affect the stability of small RNAs

In addition to the highly complementary-target-induced miRNA trimming and tailing in flies and humans, and natural-target-mediated blocking of miRNA degradation in C. elegans, targets with mismatches at the slicing sites promote the degradation of miRNAs in plants (Figure 3B). In Arabidopsis, miR399 is regulated by IPS1 (Induced by Phosphate Starvation1), which encodes a non-coding RNA containing a short motif highly complementary to miR399 68. The IPS1 sequence contains a 3-nt insertion at the cleavage site, which prevents endonucleolytic cleavage of IPS1 transcripts by miR399 RISC, resulting in the sequestration of miR399 and reduction of miR399 activity 68. Based on this principle, artificial target mimics have become a popular tool to downregulate miRNA activity for functional studies in Arabidopsis 69, 70. Todesco et al. 70 generated a large set of artificial target mimics for Arabidopsis miRNAs. In all target mimic lines they examined, the levels of the corresponding miRNAs were decreased, suggesting that unproductive interaction of miRISC with a decoy target affects miRNA stability, even though such an effect was not observed in the interaction between IPS1 and miR399 70. Does the interaction with the target mimics trigger tailing and trimming, which are the consequences of exposing Drosophila miRNAs to highly complementary targets? The mechanism underlying target mimic-mediated degradation has not been examined. Taken together observations in plants and animals, target mRNAs can serve as an additional layer of regulation for the activity and turnover of specific miRNAs. This can potentially be exploited for intervention of miRNA expression.

cis elements affect small RNA stability

cis elements in miRNAs have been shown to influence the stability of specific miRNAs. The identified cis elements vary in length, sequence and position within the miRNAs. Sequences near the 3′ end of miRNAs have been shown to affect miRNA stability 71; this is not surprising since 3′ modification/degradation seems to be a major mode of miRNA turnover such that 3′ sequences may affect the accessibility of the decay machinery to the miRNAs. Elements in the middle of the small RNA as well as the overall percentage of AU or UA dinucleotides also serve as destabilization signals 72, 73, 74. However, the enzymes responsible for sequence-specific decay of small RNAs have not been identified in most cases.

The 7-nt motif in miR-382 is an example of a cis element located at the 3′ end. miR-382, a miRNA that contributes to Human immunodeficiency virus-1 (HIV-1) latency in resting CD4+ T lymphocytes, is unstable in cells 71, 75. Bail et al. showed that the 3′-terminal 7 nt (GGAUUCG) were necessary for the low stability of miR-382 in a cell-free system that is able to recapitulate the differential stability of miRNAs 71. miR-382 decay was primarily due to the exosome with a more modest contribution by Xrn1 and no detectable contribution by Xrn2 71. However, it was not determined whether these nucleotides were sufficient to impart instability to other miRNAs.

Central motifs that influence miRNA stability have been identified in both humans and Drosophila. For example, miR-200c and miR-141, differing by 5 nt and encoded by the miR-200c_141 cluster, show varying relative abundance in different human cell lines 72. Strikingly, cell detachment induced the decay of miR-141 but not miR-200c. A sequence motif (UGUCU) in the central region was identified to be necessary for the destabilization of miR-141 by cell detachment 72. Several other miRNAs, miR-200a, miR-429 and miR-34a, which also contain this motif, were also destabilized upon cell detachment. But the motif was not able to render let-7 unstable 72. Zhang et al. 74 measured the lifespan of miRNAs in human cell lines using the pulse-chase approach. They illustrated that uridines at nucleotide positions 9-11 of miR-29b are necessary for its rapid decay in a nucleotide- and position-specific manner. In addition, a substantial subset of miRNAs with the uridine-rich sequence, such as miR-29c and fly bantam, tends to turnover faster. Different from the UGUCU motif, the destabilization effect of the uridine motif in miR-29b is conferred upon the miRNA/miRNA* duplex rather than upon mature miRNAs. Therefore, the nucleases that respond to the two motifs likely have different substrate specificities. Both studies show that the motifs are only necessary but not sufficient to confer instability 72, 74. Therefore, these motifs may rely on the sequence context of miRNAs or additional trans factors that coexist with the miRNAs.

RNA editing may affect the stability of small RNAs through altering the identity of nucleotides in small RNAs. Adenosine deaminase acting on RNA (ADAR) proteins edit adenosines to inosines in dsRNAs 76, 77. Non-coding transcripts are the most common targets of ADARs 77. It has been demonstrated that certain miRNA precursors undergo A-I editing in vivo, such as miRNA-22 and miRNA-142 precursors 78, 79. Edited pri-miR-142 is poorly processed by Drosha and degraded by Tudor-SN, a ribonuclease specific to inosine-containing dsRNAs 79. These data illustrate that RNA editing regulates the biogenesis of miRNAs. However, the effects of RNA editing on the stability of small RNA duplexes or mature miRNAs have not been evaluated thoroughly, and the function of Tudor-SN as a component of RISC is not clear either.

In summary, some miRNAs contain sequence information to influence their own turnover, which might be a mechanism that enables adjustments of miRNA levels in response to environmental or endogenous cues. Understanding the effects of cis elements that influence miRNA stability will benefit the design and utilization of small RNAs for genetic manipulations.

trans factors affect small RNA stability

Components of the RISC or RISC-associated factors may influence the stability of small RNAs (Figure 3C). Most endogenous miRNAs are in RISCs, and only a very small proportion is free in cells 80. Antagomirs that are complementary to miRNAs are able to dissociate the targeted miRNAs from RISC, causing the loss of protection of the miRNAs by RISC 80. The antagomir-induced miRNA degradation may be similar to highly complementary-target mRNA-mediated miRNA degradation, and the two may share common mechanisms, such as the displacement of the miRNAs from RISCs or the dislodging of the 3′ end of the small RNA from the PAZ domain of Ago. In the mouse, Ago proteins promote mature miRNA accumulation posttranscriptionally, in that overexpression of Ago genes elevates the levels of miRNAs and Ago2 knockout reduces the levels of miRNAs 81. Moreover, half-lives of multiple endogenous miRNAs are correlated with the cellular Ago2 levels 82. Similar stabilization effects of miRNAs by Ago proteins were observed in Arabidopsis. Loss of function of AGO1, the major miRNA effector in Arabidopsis, results in reduced levels of most miRNAs 83. Arabidopsis AGO2 is highly induced by the bacterial pathogen Pseudomonas syringae. Meanwhile, the accumulation of several miRNA star strands that are bound by AGO2 is also elevated 84. Intriguingly, Arabidopsis AGO10 specifically associates with miR165/166, but the accumulation of miR165/166 is increased, rather than reduced, in ago10 mutants, implicating that AGO10 possibly represses the level of miR165/166 instead of stabilizing this miRNA 85, 86, 87. Therefore, different Ago proteins could affect the stability of associated small RNAs differently. Besides physical association and stabilization, the slicer activity of Ago proteins can influence the stability of associated small RNAs. For example, in the T. thermophila strain containing slicer defective Twi1p, the removal of the passenger strand of scnRNA is blocked 88. The amount of scnRNAs is drastically reduced probably due to the degradation of the double-stranded scnRNAs in Twi1p by an exonuclease 88, implicating that efficient removal of passenger strands is necessary for the activity as well as the accumulation of small RNAs.

Non-Ago proteins can potentially affect small RNA stability by stabilizing or destabilizing RISCs to influence the accessibility of the small RNA-degradation machinery to small RNAs. Few such proteins have been identified. Translin, a nucleic acid-binding protein that is known to bind several mRNAs, a non-coding RNA and miR-122a in germ cells of mouse testis 89, is one example. The binding of Translin to miR-122a increases its in vivo stability through an unknown mechanism 89. Translin proteins are components of C3PO that functions in RISC activation by degrading Ago-nicked passenger strands 90. Other non-Ago proteins that may influence small RNA stability include the ADARs. Mouse ADAR1 and ADAR2 bind siRNA duplexes without editing the RNAs 91. The binding by ADARs represses the activity of siRNAs, in that siRNAs are significantly more effective in ADAR1 null mutant mouse cells than in wild-type cells 91. It was proposed that the formation of ADAR-siRNA complexes interferes with the assembly of siRNA-RISCs 91. However, it was not reported whether or how ADARs affect the turnover of siRNAs.

Conclusions and perspectives

Precise and faithful regulation of small RNA levels is critical for diverse biological processes in various organisms. Either reduction or elevation in small RNA levels results in pleiotropic developmental defects in Arabidopsis, as illustrated by the phenotypes of plants with hen1 mutations or SDN depletion 17, 64. In C. elegans, when the excess siRNAs from the CSR-1 pathway are fed into other RNAi pathways, chromosome segregation is disrupted 54. The stability of small RNAs contributes to their steady-state levels and impacts target gene expression. Understanding small RNA turnover is becoming a new focus in the RNA-silencing field. Future investigations in this area include identification of major components of the small RNA turnover machinery, and understanding how they are recruited to specific types of small RNAs and how their activities are regulated.

Understanding the mechanisms regulating small RNA stability will benefit our utilization of RNAi. For example, the stability of fork-siRNAs (a type of siRNAs with designed mismatches in the duplexes to enhance asymmetry in Ago loading) was significantly improved by methylating the nuclease-sensitive sites. As a result, the duration of the gene-silencing activity of the modified fork-siRNAs doubled that of the unmethylated analogs 92. Modulation of miR-382 stability could potentially impact the control of HIV-1 latency 75, which is a major barrier for the eradication of the virus in patients on suppressive highly active anti-retroviral therapy (reviewed in 93). Moreover, novel genetic manipulations could be devised based on the knowledge of highly complementary-target-mediated miRNA instability to target miRNAs whose overexpression is associated with human disorders. The stability of individual miRNAs can be altered after we understand how the specificity of miRNA turnover or stabilization is achieved. A recent report showed that plant miRNAs can be obtained by animals through food intake 94. This finding enlightens scientists on the production and the delivery of miRNAs into patients for therapeutic purposes.