Genes & Genetic Systems
Online ISSN : 1880-5779
Print ISSN : 1341-7568
ISSN-L : 1341-7568
Biogenesis of trans-acting siRNAs, endogenous secondary siRNAs in plants
Manabu Yoshikawa
Author information

2013 Volume 88 Issue 2 Pages 77-84


Trans-acting small interfering RNAs (tasiRNAs) are plant-specific endogenous siRNAs that control non-identical mRNAs via cleavage. The production of tasiRNAs is triggered by cleavage of capped and polyadenylated primary TAS transcripts (pri-TASs) by specific miRNAs. Following miRNA-directed cleavage, either 5′ or 3′ cleavage fragments are converted into double-stranded RNAs (dsRNAs) by RNA-DEPENDENT RNA POLYMERASE 6. The dsRNAs are processed to tasiRNAs by DICER-LIKE 4 in a phasing manner. There are two forms of pri-TASs; One has a single miRNA target site that is targeted by 22-nucleotide microRNAs, and the other has two miR390 target sites. Secondary siRNAs that are important for the amplification of RNA silencing are defined as siRNAs whose production is initiated by the cleavage of primary small RNA-containing RNA-induced silencing complexes. Thus, tasiRNA production is a model system of secondary siRNA production in plants. This review focuses on the production of tasiRNAs that are endogenous secondary siRNAs.


Phenomena that were previously described as RNA interference in nematodes, quelling in fungi, or co-suppression in plants can be attributed to small RNA (sRNA)-mediated sequence-specific gene regulation (Hannon, 2002). Two protein families, Dicer and Argonaute (AGO), play central roles in RNA silencing. Dicer proteins generate 20–25 nucleotide (nt) sRNA duplexes from double-stranded precursor RNAs. sRNAs produced by Dicer can be divided into two types based on the precursor RNAs: microRNAs (miRNAs) or small interfering RNAs (siRNAs). miRNAs are excised from single-stranded RNAs (ssRNAs) that form fold-back structures containing partially double-stranded regions. In contrast, siRNAs are processed from perfectly complementary long double-stranded RNAs (dsRNAs) that are formed between sense and antisense transcripts or are synthesized by RNA-dependent RNA polymerases (RDRs). Plant Dicers, termed DICER-LIKEs (DCLs) are phylogenetically grouped into the DCL1, DCL2, DCL3 or DCL4 clades on the basis of four Arabidopsis DCLs (Song et al., 2012; Mukherjee et al., 2013). Each subclade DCL protein can generate sRNA duplexes with distinct sizes from specific precursors. In A. thaliana, DCL1 exclusively produces miRNAs from long primary MIRNA transcripts via precursor miRNAs (Kurihara and Watanabe, 2004). DCL2, DCL3 and DCL4 generate 22 nt, 24 nt and 21 nt forms of siRNAs, respectively (Xie et al., 2004; Bouche et al., 2006; Henderson et al., 2006).

AGO family proteins associate with sRNAs and form silencing effector complexes containing single-stranded sRNA known as RNA-induced silencing complexes (RISCs). RISCs can act on target RNAs with complementary sequences and repress their targets post-transcriptionally (by cleavage or translational inhibition) or transcriptionally (by DNA methylation). The A. thaliana genome encodes 10 AGO genes that can be grouped into three clades (Vaucheret, 2008; Borges et al., 2011). Molecular analyses of these AGOs have revealed that AGOs function in gene regulation redundantly, specifically or antagonistically depending on the targets (Bouche et al., 2006; Katiyar-Agarwal et al., 2006; Baulcombe, 2007; Vaucheret, 2008). To achieve proper gene control, specific sRNA recognition and loading into specific AGOs from a myriad of sRNAs is critical. Although the forms of sRNA duplexes are major determinants for sRNA loading into AGOs in flies (Kawamata et al., 2009), the bases at the 5′ terminus of sRNAs dominate the specific sorting of individual sRNAs into specific AGOs in plants (Mi et al., 2008; Takeda et al., 2008). Multiple AGO genes are encoded in plants with large genomes (e.g., 18 in maize and 19 in rice), and among them, some AGOs are fine-tuned to respond to specific conditions (Nonomura et al., 2007; Wu et al., 2009; Qian et al., 2011; Singh et al., 2011).

Several classes of functional endogenous siRNAs have been defined in plants (Axtell, 2013). Heterochromatic siRNAs derived from transposons or repetitive genomic regions are 24 nt in length and they transcriptionally control the expression of target genes by DNA methylation, a process called RNA-directed DNA methylation (RdDM) (Qi et al., 2006). Cis-natural antisense siRNAs are produced from tail-to-tail 3′ overlapping transcripts between neighboring genomic loci and regulate either transcript (Borsani et al., 2005; Katiyar-Agarwal et al., 2006). Trans-acting siRNAs (tasiRNAs) that contain 21 nt are processed from RDR-synthesized dsRNAs that are triggered by specific miRNA-guided target cleavages. TasiRNAs regulate non-identical mRNAs via cleavage (Peragine et al., 2004; Vazquez et al., 2004; Allen et al., 2005).

TasiRNAs were originally discovered in two independent studies on A. thaliana. In one study, genetic screening of mutants whose phenotypes showed the abnormality in the juvenile-to-adult transition was performed (Peragine et al., 2004). It has been shown that these phenotypes result from upregulation of some auxin response transcription factor genes regulated by tasiRNAs generated from TAS3 (Table 1) (Adenot et al., 2006; Fahlgren et al., 2006; Garcia et al., 2006; Hunter et al., 2006). The other study analyzed transcripts of plants with mutant RDR6 and SUPRESSOR OF GENE SILENCING 3 (SGS3) genes that were defective in post-transcriptional gene silencing mediated by the sense-transgene (S-PTGS) (Vazquez et al., 2004). The production and function of tasiRNAs are unique in the following ways: (i) their production is triggered by specific miRNA cleavage of capped and polyadenylated primary TAS transcripts (pri-TASs), (ii) they are processed from either 5′ or 3′ miRNA-cleaved fragments in a 21 nt phasing manner via conversion of dsRNAs (refer to Fig. 1 in Allen and Howell, 2010 for the phased processing), and (iii) they regulate non-identical mRNAs as well as miRNAs.

Table 1.TAS loci in Arabidopsis thaliana
Locus namesTrigger miRNAsTrigger AGOsSource of tasiRNAsTarget mRNAsReferences
TAS1miR173AGO13′ cleavage fragmentPentatricopeptide repeat proteins Unknown proteinsVazquez et al. (2004)
Peragine et al. (2004)
TAS2miR173AGO13′ cleavage fragmentPentatricopeptide repeat proteinsYoshikawa et al. (2005)
Allen et al. (2005)
TAS3miR390AGO75′ cleavage fragmentARF transcription factorsAllen et al. (2005)
TAS4miR828AGO13′ cleavage fragmentMYB transcription factorsRajagopalan et al. (2006)
Fig. 1.

A model of secondary siRNA production. The initial step is the cleavage of targets by a primary sRNA-containing RISC (1a). Most cleavage fragments are subsequently degraded by RNases (1b). RDR synthesizes complementary RNAs using parts of cleavage fragments as templates (2a). Dicer processes dsRNAs into secondary siRNAs (3a). Emerged secondary siRNAs trigger further rounds of siRNA production, resulting in the amplification of RNA silencing (4a).

SiRNAs that arise from target RNAs beyond primary sRNA target sites are called secondary siRNAs or transitive siRNAs (Baulcombe, 2007; Voinnet, 2008). Secondary siRNAs have been reported in nematodes and plants, whose genomes encode multiple RDR genes (Sijen et al., 2001; Vaistij et al., 2002). This indicates that the conversion of primary sRNA-containing RISC-cleaved targets into dsRNAs by RDRs is an essential step in the production of secondary siRNAs. A current model of secondary siRNA production is shown in Fig. 1. First, a primary sRNA-programmed AGO-containing RISC cleaves a complementary target (step 1a in Fig. 1). In most cases, RISC-cleaved RNAs are degraded (step 1b). However, a subset of RISC-cleaved RNAs is involved in the production pathway of secondary siRNAs (step 2a). In plants, 22 nt forms of sRNAs and bulged forms of miRNA/miRNA* duplexes lead RISC-cleaved RNAs into the pathway (Mlotshwa et al., 2008; Chen et al., 2010; Cuperus et al., 2010). Their cleavage fragments are converted into dsRNAs by RDRs (step 2a), after which they are processed into siRNAs by Dicers (step 3a) and their siRNAs reinitiate further siRNA production (step 4a). Secondary siRNAs are important for amplification and reinforcement of RNA silencing against RNAs that are harmful to hosts, such as invading viruses (Ding, 2010). Because the cleavage of pri-TASs by miRNAs is essential for tasiRNA production, tasiRNAs belong to endogenous secondary siRNAs. Here, I review the steps in tasiRNA biogenesis in A. thaliana. Genes have been identified via extensive genetic screenings for susceptibility to viruses in Arabidopsis plants, the release of S-PTGS and developmental abnormality, and analyses have uncovered the molecular functions of those genes in tasiRNA production.


Pri-TASs are transcribed from TAS loci in the genome by DNA-dependent RNA polymerase II (RNA Pol II). Thus, pri-TASs have a 5′ cap structure and 3′ polyadenylation similar to mRNAs. Eight TAS loci (TAS1a-c, TAS2, TAS3a-c and TAS4) have been identified in A. thaliana and each pri-TAS is approximately 1 kb in length (Table 1) (Peragine et al., 2004; Vazquez et al., 2004; Allen et al., 2005; Yoshikawa et al., 2005; Rajagopalan et al., 2006; Howell et al., 2007). Two distinct forms of pri-TASs have been described using the nucleotide requirements of miRNA target sites for tasiRNA biogenesis; whereas “one-hit” forms of pri-TASs contain single miRNA target sites, “two-hit” forms contain two miRNA target sites (Axtell et al., 2006). Pri-TAS1a-c and pri-TAS2 contain the miR173 target site and pri-TAS4 contains the miR828 target site. Both miR173 and miR828 are incorporated into AGO1, which is the most extensively characterized AGO protein in plants (Baumberger and Baulcombe, 2005; Mi et al., 2008; Takeda et al., 2008). In contrast, pri-TAS3a-c, which are “two-hit” types, contain two target sites for miR390. The polarities of the miRNA cleavage fragments serving as sources of tasiRNAs also differ. TasiRNAs of pri-TAS1a-c, pri-TAS2 and pri-TAS4 are derived from the miRNA-cleaved 3′ fragments. However, pri-TAS3a-c are cleaved at the 3′ miR390 target site but not the 5′ miR390 target site, and tasiRNAs are derived from the 5′ cleavage fragments. While TAS3 and miR390 are well conserved in moss and higher plants, the other TAS loci and their trigger miRNAs are unique in Arabidopsis (Axtell et al., 2006).

pri-TAS1a-c, pri-TAS2 and pri-TAS4

The miRNA/miRNA* duplexes are loaded into AGO1 via the molecular chaperone HSP90 accompanied by the co-chaperone SQUINT/CYCROPHILIN 40 (Iki et al., 2010, 2012; Earley and Poethig, 2011). After incorporation of miRNA/miRNA* duplexes into AGO1, miRNA*s are released and miRNAs remain in AGO1, resulting in the formation of RISCs. Although most plant miRNAs are 21 nt in length, miR173 and miR828, which trigger tasiRNA production, are 22 nt. In fact, 21 nt forms of modified miR173 and miR828 lose the ability to initiate tasiRNA production in vivo (Chen et al., 2010; Cuperus et al., 2010). On the other hand, bulged miRNA/miRNA* duplexes of different sizes are competent enough to trigger secondary siRNA production. Contrary to the strict requirements for miRNA length, target transcripts have no necessity beyond the 22 nt miRNA target sites (Montgomery et al., 2008b; Felippes and Weigel, 2009). Experiments using an AGO1 defective in slicer activity (the target cleavage activity) have demonstrated that the RISC-directed cleavage event is essential for the tasiRNA production (Carbonell et al., 2012).


The TAS3a locus has been identified as the tasiRNA-generating locus initiated by miR390-guided cleavage (Allen et al., 2005), and pri-TAS3 from the moss Physcomitrella patens contains two miR390 target sites. These dual miR390 complementary sites are well conserved in higher plants (Axtell et al., 2006). Axtell et al. (2006) also confirmed that both miR390 target sites are indispensable for tasiRNA production, and then proposed a “two-hit trigger model.” Interestingly, Montgomery et al. (2008a) showed that the 5′ target site is not cleaved by miR390-AGO7 and cannot be swapped to other miRNA target sites, while the 3′ target site is cleaved by miR390-AGO7 and can be swapped to other AGO1-bound miRNA target sites. This observation implies that the cleavage event of the 3′ target site by a miRNA-guided RISC is important regardless of RISCs containing AGO7 or AGO1, but the recognition of miR390-binding AGO7 at the 5′ target site is essential for tasiRNA production. An additional difference between tasiRNA production from pri-TAS3 and other pri-TASs is that miR390 is 21 nt in length and is exclusively associated with AGO7 (Montgomery et al., 2008a). An experiment using the substitution of the 5′ nucleotide of miR390 suggested that specific loading of miR390 into AGO7 is not affected by the 5′ terminal nucleotides and the structure of the primary MIR390 transcript. The molecular mechanism by which specific sorting of miR390 to AGO7 occurs remains unknown.

THO/TREX complex

The accumulation of tasiRNAs decreases in loss-of-function mutants of the THO/TREX complex (Jauvion et al., 2010; Yelina et al., 2010). The THO/TREX complex is composed of multiple subunits that participate in the transportation of mRNAs from the nucleus to the cytosol (Finnegan et al., 2003). In the case of tasiRNA production, the molecular phenotypes of tho mutants indicated a low abundance of tasiRNAs; over-accumulation of pri-TAS1 and pri-TAS2, but a low abundance of their miR173-cleaved fragments; an abundance of miR173 to similar levels as that in wild-type. These observations strongly suggest that the THO/TREX complex acts on pri-TASs at a step after transcription but before miR173-AGO1 cleavage. However, the function of this complex in tasiRNA production remains unclear.


Zhang et al. (2012) examined the effects of translation on tasiRNA production using modified GFP mRNA constructs that contained an miR173-triggering synthetic tasiRNA cassette before or after stop codons. They found that a greater abundance of synthetic tasiRNAs were produced from constructs whose cassettes were located < 10 nt downstream of the stop codon, unlike those in which the cassettes were located > 10 nt downstream of the stop codon. They discussed the possibility that the interaction of ribosomes with miR173-programmed RISCs affects the efficiency of tasiRNA production.


In general, miRNA-cleaved fragments are unstable due to the lack of polyadenylation or cap structures. In plants, the 3′ fragments generated by miRNA cleavage are promptly degraded by XRN4, which is a 5′ to 3′ exonuclease, and other unidentified RNases (Souret et al., 2004). Thus, the stabilization of the cleavage fragments generated by miR390 or 22 nt miRNA-containing RISCs is important for subsequent steps. The 5′ and 3′ miR173-cleaved fragments accumulate more in the rdr6 mutant than in wild-type, while the 3′ fragments disappear and the 5′ fragments decrease in the sgs3 rdr6 double mutant similar to the sgs3 mutant (Yoshikawa et al., 2005). Given that SGS3 and RDR6 are genetically epistatic, SGS3 may act on the miR173-cleaved fragments before RDR6, and SGS3 may stabilize the miR173 cleavage fragments. An extensive characterization of SGS3 protein expressed in Escherichia coli indicated that SGS3 can specifically bind to a dsRNA containing a 5′-overhang, but cannot bind to an ssRNA or a dsRNA with a 3′ overhang (Fukunaga and Doudna, 2009). These results suggest that SGS3 acts on dsRNA with 5′ overhangs formed during tasiRNA production, but contradicts the prediction that SGS3 stabilizes the single-stranded miR173-cleaved fragments. Using RISC prepared from extracts of evacuolated tobacco protoplasts (BYL), my group recently found that SGS3 is recruited on miR173-containing RISCs that bind to the TAS2 RNA and form a complex containing RISC and the cleavage fragments (Yoshikawa et al., 2013). The formation of the complex is disturbed by a 1 nt deletion at the 3′ terminus of miR173 or a mismatch at the 3′ end of the miR173 target site on TAS2 RNA, resulting in the degradation of the 3′ fragments. Furthermore, the depletion of endogenous tobacco SGS3 in BYL causes instability in the 3′ miR173-cleaved fragments. These results indicate that the 22 nt form of miR173 and SGS3 are important for the stabilization of the miR173 cleavage fragments. The effects of base-pairing at the 3′ end of miR173 on SGS3 imply that SGS3 binds to the dsRNA, resulting in the formation of a 5′ overhang between miR173 and its target. Although the function of SGS3 remains unclear in terms of the production of tasiRNAs from pri-TAS3, SGS3 co-localizes with AGO7 (Jouannet et al., 2012). Thus, SGS3 may also stabilize miR390-guided RISC-cleaved fragments in concert with an RISC that contains miR390-AGO7.

RDR6 shows dsRNA synthesis activity that is complementary to an ssRNA in vitro (Curaba and Chen, 2008). In fact, RDR6-dependent complementary RNAs to the 3′ fragment from the TAS2 transcript have been detected in vivo (Yoshikawa et al., 2005). However, how RDR6 is recruited to the miRNA-cleaved TAS RNA fragments remains unknown. SGS3 and RDR6 co-localize to a specific compartment in the cytosol, designated the SGS3/RDR6 body. SGS3/RDR6 bodies never overlap with P-bodies that contain AGO proteins and decapping enzymes (Kumakura et al., 2009). However, some SGS3/RDR6 bodies are observed close to P-bodies. This might indicate the functional linkage between the SGS3/RDR6 body and the P-body. In contrast, AGO7 co-localizes with the SGS3/RDR6 body (Jouannet et al., 2012).


Genetic screenings of PTGS release have recovered sde5 mutants (Hernandez-Pinzon et al., 2007; Jauvion et al., 2010). TasiRNA accumulation is abolished in the sde5 null mutant and pri-TAS1 and pri-TAS2 accumulate to similar level in wild-type, rdr6 and sgs3. The cleavage fragments from pri-TAS1 and pri-TAS2 accumulate more in the sgs3 mutant and less in the rdr6 mutant (Jauvion et al., 2010). These observations suggest that SDE5 functions in tasiRNA production after pri-TASs by miRNA cleavage but before dsRNA formation by RDR6 in concert with SGS3 and/or RDR6 after miRNA cleavage. Although SDE5 is partially homologous to the human mRNA export factor TAP, the biochemical activity of SDE5 remains uncharacterized (Hernandez-Pinzon et al., 2007).


RDR6-synthesized dsRNAs are processed into 21 nt tasiRNA duplexes with a 2 nt overhang at both 3′ termini by a heterodimer composed of DCL4 and the dsRNA-binding protein DOUBLE-STRANDED RNA BINDING PROTEIN 4 (DRB4) (Hiraguri et al., 2005; Adenot et al., 2006; Nakazawa et al., 2007; Fukudome et al., 2011). The processing of the dsRNAs by DCL4/DRB4 begins from the miRNA cleavage site in a stepwise manner (Xie et al., 2005). To regulate registered target mRNAs, the 21 nt processing that begins at the miRNA cleavage sites is critical for the production of defined tasiRNAs from each TAS locus (Allen et al., 2005; Yoshikawa et al., 2005). DCL4 and DRB4 localize to the nucleus (Xie et al., 2004; Hiraguri et al., 2005), whereas the SGS3/RDR6 body localizes to the cytosol (Kumakura et al., 2009; Jouannet et al., 2012). These observations imply that the dsRNAs synthesized in the SGS3/RDR6 body are imported into the nucleus. Two recent studies found that the tasiRNA pathway partially participates in crosstalk with the RdDM pathway that occurs in the nucleus (Wu et al., 2012; Kanno et al., 2013). They found that the downstream genomic region of the miR173 target site on the TAS1c locus and the spanning genomic region between the 5′ and 3′ miR390 target site on the TAS3a locus are methylated in CG, CHG, and CHH contexts. These methylation patterns are consistent with the miRNA-cleaved fragments that are converted into dsRNAs by RDR6. In fact, these methylations are affected by loss-of function mutations in SGS3 and RDR6. Interestingly, these types of methylations are needed for NUCLEAR RNA POLYMERASE E1 (the largest subunit of DNA-dependent RNA polymerase V), AGO4 and AGO6, but not DCL2, DCL3, DCL4 and RNA-DEPENDENT RNA POLYMERASE 2, implying that RDR6-synthesized dsRNAs are processed by DCL1. DCL1-produced siRNAs are bound to AGO4 and AGO6, and the RISCs containing siRNA-bound AGO4 or AGO6 can induce methylation on the corresponding TAS loci. Thus, because DCL1 localizes to the nucleus, these observations might support that the hypothesis the processing of dsRNAs synthesized from pri-TASs occurs in the nucleus.


TasiRNAs are associated with AGO1, AGO2 or AGO5 following the sRNA-sorting rule of the 5′ terminal bases (Mi et al., 2008; Takeda et al., 2008). These AGO proteins are preferentially localized to the cytoplasm (Takeda et al., 2008; Borges et al., 2011; Wang et al., 2011). If tasiRNAs are loaded into AGO proteins in the cytoplasm, the export of tasiRNA duplexes to the cytosol is necessary (Fig. 2). Plant miRNA/miRNA* duplexes are exported from the nucleus to the cytoplasm by HASTY (HST), which is an ortholog to Exportin-5 in animals (Park et al., 2005). Loss-of-function mutations in HST affect the accumulation of most miRNAs, but do not affect the accumulation of tasiRNAs from TAS1 and TAS2 (Park et al., 2005; Yoshikawa et al., 2005). This observation implies that an unknown protein exports tasiRNA duplexes to the cytoplasm. Alternatively, tasiRNA-containing RISCs might act on target mRNAs in the nucleus. In fact, there is some evidence that PTGS occurs in the plant nucleus (Xie et al., 2003; Hoffer et al., 2011). Future studies are required to elucidate this issue.

Fig. 2.

A proposed model of tasiRNA production in Arabidopsis thaliana. Pri-TASs are transcribed by RNA pol II and the 22 nt miRNA- or miR390-guided RISCs cleave pri-TASs in the cytosol. Either the 5′ or 3′ fragments are converted into dsRNAs by RDR6. The dsRNAs are processed by the DCL4/DRB4 complex. The orange dashed line reveals a nuclear membrane. The black lines and the black dashed lines reveal probable and putative steps, respectively.


I thank Drs Tatsuo Kanno and Yoshihisa Ueno for helpful comments on the manuscript. M.Y. is supported in part by the Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

© 2013 by The Genetics Society of Japan