Cooperative recruitment of RDR6 by SGS3 and SDE5 during small interfering RNA amplification in Arabidopsis

Significance

In plants, RNA-induced silencing complexes that contain 22-nucleotide small interfering RNA (siRNA) and Argonaute 1 protein induce siRNA amplification from their targets. In this pathway, double-stranded RNAs (dsRNAs) are synthesized by RNA-dependent RNA polymerase proteins, and the dsRNAs are processed into siRNAs by Dicer-like proteins, leading to amplification of siRNAs to targets. Because dsRNA formation from nontarget RNAs causes nonspecific siRNA production, this pathway needs to be strictly regulated. By using an in vitro siRNA amplification system, we demonstrate that RNA-dependent RNA polymerase proteins are specifically recruited to targets in coordination with two plant-specific proteins. Based on the results, we discuss how specificity of siRNA amplification is secured.

Abstract

Small interfering RNAs (siRNAs) are often amplified from transcripts cleaved by RNA-induced silencing complexes (RISCs) containing a small RNA (sRNA) and an Argonaute protein. Amplified siRNAs, termed secondary siRNAs, are important for reinforcement of target repression. In plants, target cleavage by RISCs containing 22-nucleotide (nt) sRNA and Argonaute 1 (AGO1) triggers siRNA amplification. In this pathway, the cleavage fragment is converted into double-stranded RNA (dsRNA) by RNA-dependent RNA polymerase 6 (RDR6), and the dsRNA is processed into siRNAs by Dicer-like proteins. Because nonspecific RDR6 recruitment causes nontarget siRNA production, it is critical that RDR6 is specifically recruited to the target RNA that serves as a template for dsRNA formation. Previous studies showed that Suppressor of Gene Silencing 3 (SGS3) binds and stabilizes 22-nt sRNA–containing AGO1 RISCs associated with cleaved target, but how RDR6 is recruited to targets cleaved by 22-nt sRNA–containing AGO1 RISCs remains unknown. Here, using cell-free extracts prepared from suspension-cultured Arabidopsis thaliana cells, we established an in vitro system for secondary siRNA production in which 22-nt siRNA–containing AGO1-RISCs but not 21-nt siRNA–containing AGO1-RISCs induce secondary siRNA production. In this system, addition of recombinant Silencing Defective 5 (SDE5) protein remarkably enhances secondary siRNA production. We show that RDR6 is recruited to a cleavage fragment by 22-nt siRNA–containing AGO1-RISCs in coordination with SGS3 and SDE5. The SGS3–SDE5–RDR6 multicomponent recognition system and the poly(A) tail inhibition may contribute to securing specificity of siRNA amplification.

RNA silencing is a molecular mechanism of sequence-specific gene regulation by small RNAs (sRNAs) of 20 to 30 nucleotides (nt) in length (1). These regulatory sRNAs are produced through distinct pathways and are incorporated into Argonaute (AGO) family proteins, leading to the formation of effector complexes (2). In posttranscriptional gene silencing, an effector complex containing an sRNA and an AGO protein as its core components is known as the RNA-induced silencing complex (RISC). RISCs regulate target genes through cleavage or/and translational inhibition (3). Plant AGO proteins are divided into three clades based on phylogenetic analyses and have functions in the posttranscriptional or transcriptional gene silencing processes (4). AGO1 plays a range of critical roles in posttranscriptional gene silencing (5). It forms RISCs through specific loading of microRNAs (miRNAs) and small interfering RNAs (siRNAs) with 5′-uridine (6). These sRNA-AGO1–containing RISCs posttranscriptionally regulate numerous targets related to development, antiviral defense, or abiotic or biotic stress responses (7, 8). In plants and nematodes, gene regulation by RISCs induces de novo siRNA production from target RNA through double-stranded RNA (dsRNA) formation driven by RNA-dependent RNA polymerases (RDRs) (9). These newly generated siRNAs, designated secondary siRNAs, are further amplified through repeated entry into the pathway. In plants, secondary siRNAs play important roles in gene regulation and anti-viral defense (10).

Trans-acting siRNAs (tasiRNAs) among endogenous plant-specific siRNAs function, like miRNAs, in the posttranscriptional regulation of target genes (11–15). tasiRNA production is initiated by cleavage of primary transcripts from TAS loci by specific RISCs containing miR390-AGO7 or 22-nt miRNA-AGO1 (e.g., miR173-AGO1 in Arabidopsis thaliana) (12, 15). Based on genetic studies, Suppressor of Gene Silencing 3 (SGS3), Silencing Defective 5 (SDE5), RNA-dependent RNA polymerase 6 (RDR6), and Dicer-like 4 (DCL4) act on cleavage fragments in that order (12, 16). In a present model of tasiRNA production from the TAS2 locus in A. thaliana, the 5′-capped and 3′-poly(A)-tailed TAS2 primary transcript is cleaved by a miR173-AGO1–containing RISC (miR173-AGO1–RISC) and separated into two fragments. SGS3 stabilizes the 3′ cleavage fragment, which serves as a template for dsRNA synthesis by RDR6 in the SGS3–RISC–target RNA ternary complex (17). Although the precise function of SDE5 remains unclear, SDE5 protein shows partial amino acid sequence similarity to a human messenger RNA (mRNA) export factor TAP/NXF1 and has two domains, called DUF1771 and smr. SDE5 operates within the tasiRNA biogenesis pathway after SGS3 but before RDR6 (16, 18, 19). RDR6 converts the 3′ cleavage fragment into dsRNA (12). Finally, DCL4 excises 21-nt tasiRNAs from the dsRNA as siRNA duplexes (12, 15).

RDR6-dependent secondary siRNAs are also produced from virus-related RNAs and transcripts of transgenes driven by strong promoters (20–22). Although the primary siRNA that acts as the trigger in these pathways remains unclear, it has been suggested that the 22-nt siRNA produced by DCL2 is important for initiation of secondary siRNA production (23, 24). In tasiRNA and secondary siRNA production, it is necessary that RDR6 is exactly recruited to a target RNA that serves as a template for dsRNA formation through a specific mechanism, because nonspecific RDR6 recruitment causes unnecessary siRNA production from dsRNAs of nontargets. However, what proteins are involved in strict RDR6 recruitment remains unidentified. Here, we established an in vitro system of secondary siRNA production using cell-free extracts prepared from suspension-cultured A. thaliana cells. Using this experimental system, we revealed that SGS3 and SDE5 cooperatively recruit RDR6 to a specific RNA that serves as a template for dsRNA formation.

Results

tasiRNA Production Is Induced by miR173-AGO1–Containing RISCs Contained in Cell-Free Extracts of Suspension-Cultured A. thaliana Cells.

The suspension-cultured A. thaliana cell line (YG1, named here) was previously established from wild-type (Col-0 ecotype) seedlings (25). It expresses miR173 and miR173-triggered tasiRNAs from TAS1a–c and TAS2 (SI Appendix, Fig. S1A), suggesting that the cells contain the essential components needed for tasiRNA production. Therefore, we tested whether tasiRNA is produced from TAS2 RNA in cell-free extracts of vacuole-removed YG1 protoplasts (YG1 extracts). Wild-type TAS2 RNA and three additional TAS2-related RNAs were used for analyses (SI Appendix, Fig. S1B). TAS2-miR173M RNA contains multiple mutations at the miR173 target site. The TAS2m1 and TAS2m1m2 RNAs contain mutations of one and two nt at the 5′ end of the miR173 target site, respectively. These mutations do not affect cleavage by miR173-AGO1–RISCs but abolish the accumulation of the 3′ cleavage fragment and tasiRNAs in vivo (26) and interfere with stabilization of the 3′ cleavage fragment through the interaction between AGO1 and SGS3 in vitro (17). These TAS2-related RNAs (TAS2-miR173M, TAS2, TAS2m1, and TAS2m1m2, respectively) were mixed with YG1 extracts or cell-free extracts of the suspension-cultured Nicotiana tabacum BY-2 protoplasts (BYL) (27), and the levels of target cleavage and tasiRNA production were analyzed (SI Appendix, Fig. S1C). Neither cleavage nor tasiRNA production were detected in BYL, as N. tabacum BY-2 cells do not express miR173 (SI Appendix, Fig. S1A). As expected, TAS2, TAS2m1, and TAS2m1m2 were cleaved at similar levels, while TAS2-miR173M was not cleaved in YG1 extracts. As we previously reported, the 3′ cleavage fragments from wild-type TAS2 were most highly accumulated among fragments from TAS2m1 and TAS2m1m2 (SI Appendix, Fig. S1C, Upper lanes 3, 7, 11, and 15), reflecting stabilization of the uncapped 3′ cleavage fragment by SGS3 (17). When tasiRNA production was examined, tasiRNAs migrating near 21-nt siRNA duplexes were detected in the reaction mixture with TAS2 (SI Appendix, Fig. S1C, Lower lane 7). No such siRNA duplexes were detected in the reaction mixtures with TAS2-miR173M, TAS2m1, and TAS2m1m2. These results suggest that YG1 extracts contain essential components for tasiRNA production and that tasiRNA is produced from TAS2 by miR173-AGO1–RISCs in the extracts.

Because SDE5 operates in this pathway next to SGS3 whose role was uncovered by previous analyses (16, 17), we investigated the function of SDE5 using our in vitro system. For this purpose, we added bacterially expressed A. thaliana SDE5 protein into YG1 extracts (SI Appendix, Fig. S2A) and analyzed the cleavage of TAS2-related RNAs and tasiRNA production. Although addition of SDE5 did not affect the cleavage efficiency of TAS2, TAS2m1, and TAS2m1m2 by miR173-AGO1–RISCs, it markedly enhanced tasiRNA production from TAS2 (SI Appendix, Fig. S1C, Lower lane 8). We tested whether in vitro–translated and affinity-purified SDE5 can bind to four forms of RNA described in Fukunaga and Doudna (28) (SI Appendix, Fig. S2B), but no obvious RNA binding activity was detected (SI Appendix, Fig. S2C).

Secondary siRNA Production Is Induced by 22-nt siRNA-AGO1–containing RISCs in YG1 Extracts.

In vivo analyses suggested that DCL2-dependent 22-nt siRNAs could initiate RDR6-dependent secondary siRNA production (23, 24). Therefore, we explored whether AGO1 programmed in vitro with a 22-nt siRNA could trigger secondary siRNA production from a target in YG1 extracts. For this purpose, 21- and 22-nt siRNAs complementary to a portion of Green Fluorescent Protein (GFP) mRNA (28, 29) were used (21- and 22-nt siR-gf698, respectively). As the target of siR-gf698, Luciferase (Luc) mRNA of which 681th to 704th positions were swapped with the complementary sequence to siR-gf698 (Luc-gf698) was prepared (Fig. 1A). Initially, we investigated whether FLAG-tagged A. thaliana AGO1 (hereafter, AGO1) could be synthesized with in vitro–translated YG1 extracts and whether AGO1 could cleave Luc-gf698 after mixing the in vitro–translated mixtures with 21-nt siR-gf698 duplexes. However, target cleavage was hardly detected despite AGO1 protein was synthesized (SI Appendix, Fig. S3 A and B), implying that siRNA loading into AGO1 in YG1 extracts is not as efficient as that in BYL. Therefore, we used BYL to prepare 21- or 22-nt siR-gf698-AGO1 complexes. The siR-gf698-AGO1 complexes were immunoprecipitated using an anti-FLAG affinity gel, eluted from the gel with FLAG peptide–containing elution buffer, and then the siR-gf698-AGO1–containing eluted fractions were mixed with YG1 extracts and the Luc-related RNAs (Luc or Luc-gf698) (Fig. 1B). Neither Luc nor Luc-gf698 were cleaved in the control reaction mixtures without siR-gf698 and AGO1 (Fig. 1C, Upper lanes 1, 2, 5, and 6). Although Luc was not cleaved in reaction mixtures containing 21- or 22-nt siR-gf698-AGO1, Luc-gf698 was cleaved in both mixtures (Fig. 1C, Upper lanes 1 through 8). The efficiency of target RNA cleavage was a little better with 21-nt siRNA-programmed AGO1, and this is consistent with the fact that loading efficiency of 21-nt siRNA in AGO1 is better than that of 22-nt siRNAs in this system (30, 31). Migration of siRNAs near 21-nt siRNA duplexes was observed from Luc-gf698 in the reaction mixture of 22-nt siR-gf698-AGO1 but not in reaction mixtures with Luc, lacking siR-gf698 or AGO1, or those with Luc-gf698 and 21-nt siR-gf698-AGO1 (Fig. 1C, Lower lanes 1 through 8). This shows that 22-nt siRNA–programmed AGO1-RISCs that mimic DCL2-dependent siRNAs can trigger secondary siRNA production in vitro. When recombinant SDE5 was added to the YG1 extracts, the production of secondary siRNAs by 22-nt siR-gf698-AGO1 was also markedly enhanced and that by 21-nt siR-gf698-AGO1 was weakly detected (Fig. 1C, Lower lanes 15 and 16). This suggests that SDE5 is a critical factor in this experimental system. To confirm that emerged ³²P-RNAs that migrate near 21-nt siRNA duplexes on native Polyacrylamide gel electrophoresis (PAGE) are siRNA duplexes, ³²P-RNAs that were produced in the mixture supplemented with 22-nt siRNA-AGO1 and SDE5 were analyzed by denaturing PAGE along with 21-, 22-, and 24-nt single-stranded RNAs (SI Appendix, Fig. S4). On this gel system, ³²P-RNA signals were detected at positions around 21- and 22-nt single-stranded RNA markers. Together with the fact that single-stranded siRNAs migrate faster than siRNA duplexes of the same length on native PAGE (SI Appendix, Fig. S4), we conclude that 21- or 22-nt secondary siRNA duplexes were produced from the targets.

Fig. 1.

Secondary siRNAs are produced from targets cleaved by 22-nt-siRNA AGO1-RISCs in YG1 extracts. (A) Diagrams of Luc-related RNAs, siR-gf698, and miR-gf698 duplexes. The upper strands of siR-gf698 and miR-gf698 were loaded into AGO1 as guide strands or miRNAs, whereas the lower strands were removed from AGO1 as passenger strands or miRNA*s. (B) Scheme of the experimental procedure for in vitro RISC formation in BYL and secondary siRNA production in YG1 extracts. (C) Secondary siRNA production by in vitro–programmed 22-nt-siRNA AGO1-RISCs. Luc-related RNAs (2.0 nM) that were ³²P-labeled, 5′-capped, and 3′-poly(A60)-tailed were incubated with fractions from FLAG peptide elution and YG1 extracts in the presence or absence of 100 nM recombinant SDE5. Upper and Lower show target cleavage and secondary siRNA production, respectively. siRNA duplexes of 21 and 24 nt were ³²P-labeled and used as size markers (lane M). Immunopurified AGO1 proteins were detected by Western blot analysis using anti-FLAG antibody. (D) Secondary siRNA production by AGO1-D762A. Luc or Luc-gf698 (2.0 nM) that were ³²P-labeled, 5′-capped, and 3′-poly(A60)-tailed were incubated with the fractions from FLAG peptide elution containing AGO1 or AGO1-D762A in YG1 extracts either in the presence or absence of 100 nM recombinant SDE5. Immunopurified AGO1 proteins were detected by Western blotting analysis using anti-FLAG antibody. (E) Involvement of RDR6 in secondary siRNA production in YG1 extracts. A recombinant RDR6 protein (10 nM) was added to YG1 extracts. Luc-gf698 (2.0 nM) that was ³²P-labeled, 5′-capped, and 3′-poly(A60)-tailed was incubated with the fractions from FLAG peptide elution containing AGO1 and YG1 extracts in the presence or absence of 100 nM recombinant SDE5. Immunopurified AGO1 proteins were detected by Western blot analysis using anti-FLAG antibody.

To characterize the distribution of secondary siRNAs produced from Luc-gf698, sRNA sequencing (sRNA-seq) was performed (SI Appendix, Figs. S5–S11 and Table S1). We mixed Luc-gf698 and YG1 extracts with AGO1 alone or with 21- or 22-nt siR-gf698-AGO1 in the presence or absence of recombinant SDE5 (SI Appendix, Fig. S5), extracted RNA from each reaction mixture, and carried out sRNA-seq analyses. Reads per million values of 18- to 30-nt RNAs and six endogenous miRNAs from YG1 extracts were consistent among the six samples (SI Appendix, Fig. S6 A and B). Because 21- and 22-nt siRNAs derived from Luc-gf698 were more abundant than siRNAs of the other sizes (SI Appendix, Table S1), we focused on these siRNAs for mapping onto the target RNA, excluding siRNAs corresponding to siR-gf698 duplexes (SI Appendix, Fig. S6C, S7 A and B and Table S1). In the AGO1 alone and AGO1+SDE5 samples, 21- and 22-nt siRNAs were observed across the entire target RNA region at low levels (SI Appendix, Fig. S7 A and B, Upper). When 21-nt siRNA-AGO1 was added, 21- and 22-nt siRNAs were produced from the entire target RNA region, and those derived from the 3′ cleavage fragment were more abundant than those derived from the 5′ cleavage fragment (SI Appendix, Fig. S7 A and B, Third panel from the top). When 22-nt siRNA-AGO1 was added, 21- and 22-nt siRNA from the target RNA was observed in much greater abundance and made the preference for the 3′ cleavage fragment region more obvious (SI Appendix, Fig. S7 A and B, fifth panel from the top). An addition of SDE5 to reaction mixtures with 21- or 22-nt siRNA-AGO1 markedly enhanced siRNA production from the target RNA, and the preference for the 3′ region was more obvious (SI Appendix, Fig. S7 A and B, fourth and sixth panels from the top). These results indicate that cleavage by a 22-nt siRNA-AGO1–containing RISC (22nt siRNA-AGO1–RISC) preferentially induces secondary siRNA production from the 3′ cleavage fragment, as is the case for TAS1a–c and TAS2 in miR173-AGO1–RISC systems (15). Next, we examined whether phasing patterns were observed in 21-nt and 22-nt secondary siRNAs (SI Appendix, Figs. S8–S10). The result clearly showed that the secondary siRNAs were produced predominantly in a 21- or 22-nt phased manner from the cleavage positions.

A Slicer Defective AGO1 Can also Trigger Secondary siRNA Production.

To explore whether target cleavage is essential for secondary siRNA production, a slicer defective AGO1 mutant protein of which the 762nd aspartate was substituted to alanine (AGO1-D762A) was used. Slicer-defective AGO proteins cannot form RISCs with siRNA duplexes, but they can remove miRNA*s from miRNA/miRNA* duplexes containing a few mismatched base pairs and form miRNA-containing RISCs (28, 32–35). Based on this information, we introduced three mismatched bases into the 21- and 22-nt passenger strands of the siR-gf698 duplexes. Duplexes formed between the guide strands and modified passenger strands of siR-gf698 mimic miRNA/miRNA* duplexes. In this section, we define these duplexes as miR-gf698/miR-gf698* of which the guide strands of siR-gf698 and the modified passenger strands corresponds to miR-gf698 and miR-gf698*, respectively (Fig. 1A, Third and fourth duplexes from the top). Prior to analyses of secondary siRNA production by AGO1-D762A, we confirmed that 21- or 22-nt miR-gf698 could preferentially be loaded into AGO1-D762A from 21- or 22-nt miR-gf698/miR-gf698*, respectively (SI Appendix, Fig. S12). We investigated whether AGO1-D762A programmed with 21- or 22-nt miR-gf698 could induce secondary siRNA production (Fig. 1D). AGO1 programmed with 21-nt or 22-nt miR-gf698 cleaved Luc-gf698, whereas AGO1-D762A programmed with 21- or 22-nt miR-gf698 did not (Fig. 1D, Upper lanes 14, 15, 17, and 18). Secondary siRNAs were detected from AGO1 programmed with 22-nt miR-gf698 in the absence of exogenous SDE5, and their abundance was enhanced by adding SDE5, as was the case for 22-nt siR-gf698 (Fig. 1D, Lower lanes 9 and 15). Although secondary siRNA production did not occur with 22-nt miR-gf698-AGO1-D762A in the absence of recombinant SDE5, it was detected in the presence of the protein (Fig. 1D, Lower lanes 12 and 18). These results suggest that target cleavage is not required for secondary siRNA production and that the efficiency of secondary siRNA production by slicer-defective AGO1 is lower than that by wild-type AGO1. Previous research has indicated that slicer-defective mutants of AGO1 induce tasiRNA production in vivo (36), and the present results are consistent with that observation.

RDR6 Is Involved in Secondary siRNA Production In Vitro.

To confirm whether RDR6 plays roles in secondary siRNA production in YG1 extracts, purified recombinant RDR6 protein was added to secondary RNA reactions in vitro (37). Addition of RDR6 protein did not affect cleavage of Luc-gf698 by 21- or 22-nt siR-gf698-AGO1-RISCs (Fig. 1E, Upper lanes 1 through 6) but enhanced secondary siRNA production (Fig. 1E, Lower lanes 2, 3, 5, and 6). Furthermore, secondary siRNA production in the presence of recombinant SDE5 and RDR6 was dramatically elevated in the reaction mixture containing 22-nt siR-gf698-AGO1 (Fig. 1E, Lower lanes 3, 6, 12, and 15). This result indicates that RDR6 functions synergistically with SDE5 in secondary siRNA production. The addition of SDE5 and RDR6 also markedly enhanced secondary siRNA production by 21-nt siR-gf698-AGO1. (Fig. 1E, Lower lanes 2, 5, 11, and 14), implying that excessive SDE5 and RDR6 enhance secondary siRNA production from RISC targets. By contrast, upon addition of RDR6-CT, an RDR6 mutant of which RDR activity is abolished by substitution of the 867th aspartate with alanine located in the polymerase catalytic center, secondary siRNA production was inhibited even in the presence of excessive SDE5 (Fig. 1E, Lower lanes 3, 9, 12, and 18), indicating that RDR6-CT dominantly impairs RDR6 function in YG1 extracts. This result supports the hypothesis that RDR6 plays a role in secondary siRNA production in YG1 extracts.

SDE5 Facilitates dsRNA Formation from 3′ Cleavage Fragments of 22-nt siRNA-AGO1-RISCs.

The results of SI Appendix, Fig. S7 revealed that SDE5 specifically promotes secondary siRNA production from the 3′ cleavage fragment. To further address the molecular function of SDE5, the Flock House virus (FHV) B2 protein, which binds to long dsRNAs and inhibits siRNA production by Dicer, was used (Fig. 2) (38). An addition of recombinant B2 protein to the reaction mixture did not affect the efficiency of cleavage by 21- or 22-nt siR-gf698-AGO1–RISCs (Fig. 2A, Upper lanes 2, 3, 5, and 6). However, B2 repressed the secondary siRNA production induced by 22-nt siR-gf698-AGO1–RISCs, even in the presence of excess SDE5 (Fig. 2A, Lower lanes 3, 6, 9, and 12). To compare dsRNA accumulation among reaction mixtures, RNA was extracted from each mixture and treated with S1 nuclease, which degrades single-stranded RNAs but not dsRNAs (Fig. 2B). No accumulation of dsRNA was detected in the AGO1 alone reaction mixtures (Fig. 2B, lanes 2, 8, 14, and 20). Weak dsRNA accumulation was detected in the reaction mixture containing 21-nt siR-AGO1-gf698–RISCs, B2, and SDE5 (Fig. 2B, lane 22) but not in the reaction mixtures with 21-nt siR-gf698-AGO1–RISCs that lacked SDE5, B2, or both (Fig. 2B, lanes 4, 10, and 16). By contrast, dsRNA was detected in the reaction mixtures containing 22-nt siR-gf698-AGO1–RISCs (Fig. 2B, lanes 6, 12, 18, and 24). The dsRNA regions were longer in the presence of B2 than in its absence, and dsRNAs accumulated to the highest level in the presence of both SDE5 and B2 (Fig. 2B, lane 24). To confirm that antisense RNA to Luc-gf698 was synthesized in YG1 extracts, we carried out Northern blot hybridization. RNA extracts were treated with S1 nuclease, blotted, and hybridized with sense or antisense oligoDNAs corresponding to the 3′ cleavage fragments (Fig. 2C). The signal distribution pattern for antisense RNA was quite similar to that shown in Fig. 2B. This result indicates that the synthesis of antisense RNA of the 5′-uncapped and 3′-polyandeylated 3′ fragments was induced through target cleavage by 22-nt siR-gf698-AGO1–RISCs. The formation of dsRNA of the 3′ cleavage fragment is consistent with sRNA-seq data showing that abundant secondary siRNAs were produced from the 3′ fragments in 22-nt siR-gf698-AGO1 and 22-nt siR-gf698-AGO1+SDE5 reactions. Furthermore, the synthesis of antisense RNA increased with adding of SDE5. These data suggest that SDE5 promotes dsRNA synthesis by RDR6 rather than activating Dicer activity, resulting in overaccumulation of secondary siRNAs.

Fig. 2.

SDE5 enhances dsRNA synthesis from 3′ fragments cleaved by 22-nt-siRNA AGO1-RISCs. (A) Inhibition of secondary siRNA production by Flock House virus (FHV) B2 protein. Luc-gf698 (2.0 nM) that was ³²P-labeled, 5′-capped, and 3′-poly(A60)-tailed was incubated with the fractions from FLAG peptide elution containing AGO1 in YG1 extracts in the presence or absence of 100 nM B2. Upper and Lower show target cleavage and secondary siRNA production, respectively. Immunopurified AGO1 proteins were detected by Western blot analysis using anti-FLAG antibody. (B) dsRNA accumulation of Luc-gf698 in YG1 extracts. After the analyses shown in A, the remaining RNA was treated with S1 nuclease, separated on a denaturing 1.2% agarose gel, transferred onto a Hybond N+ membrane, and detected with autoradiography. S1 minus indicates RNAs that were sampled prior to S1 nuclease treatment. According to our calculations, the relative amounts of RNAs between S1 minus and S1 plus is ∼1 to 200. (C) Detection of antisense Luc-gf698 in YG1 extracts. Non–³²P-labeled Luc-gf698 (2.0 nM) that was poly(A60)-tailed was incubated with the fractions from FLAG peptide elution containing AGO1 and YG1 extracts in the presence or absence of 100 nM B2. RNA was extracted from the reaction mixtures and treated with S1 nuclease. RNA was separated on a denaturing 1.2% agarose gel and transferred onto a Hybond N+ membrane. Sense or antisense ³²P-labeled oligoDNA probes were used for detection of antisense RNA (Upper) or sense RNA (Lower) for the 3′ region of Luc-gf698, respectively. S1 minus samples (odd lanes) contain RNA that were sampled prior to S1 nuclease treatment. According to our calculations, the amount of RNA in S1 minus relative to S1 plus is ∼1 to 200. Immunopurified AGO1 proteins were detected by Western blot analysis using anti-FLAG antibody.

In the Fig. 2 experiments, antisense RNA to poly(A)-tailed target RNA was synthesized. However, a previous study showed that poly(A) tails at 3′ ends of target RNAs inhibit antisense RNA synthesis by RDR6 (37). We examined whether poly(A) tail affects initiation of dsRNA synthesis of the 22-nt AGO1–RISC 3′ cleavage fragments of the target RNA in YG1 extracts (SI Appendix, Fig. S13). When poly(A)-tailed Luc-gf698 RNAs were used, smear antisense RNA signals were detected in the S1 nuclease-treated sample containing 22-nt siRNA-AGO1, SDE5, and B2 as shown in Fig. 2C (SI Appendix, Fig. S13, lane 24). In contrast, when poly(A)-less targets (Luc-gf698-A₀) were used, discreet antisense RNAs of ca. 1,000 nt were observed in the S1 nuclease-treated sample containing 22-nt siRNA-AGO1, SDE5, and B2 (SI Appendix, Fig. S13, Upper lane 16). When sense RNAs were detected, discrete sense RNA of ca. 1,000 nt was also observed in the S1 nuclease-treated sample containing Luc-gf698-A₀, 22-nt siRNA-AGO1, SDE5, and B2 (SI Appendix, Fig. S12, Lower lane 16). The size of the sense and antisense RNA was close to that of the 3′ cleavage fragments, suggesting that RDR6 in YG1 extracts initiates conversion of poly(A)-less 3′ cleavage fragments to dsRNAs from a specific position, probably their 3′ ends, whereas RDR6 initiates the synthesis of antisense RNA conversion to poly(A)-tailed 3′ cleavage fragment from nonspecific regions upstream of poly(A) tails.

SDE5 and SGS3 Cooperatively Recruit RDR6 to a Target and Promote Its dsRNA Synthesis.

Based on the results of Fig. 2, we considered whether SDE5 facilitates the recruitment of RDR6 to a target that serves as a template for dsRNA synthesis. To address this, we performed a tethering assay using a λN peptide and BoxB sequence-containing RNA (39, 40). In this assay, λN peptide fusion proteins specifically bind to RNA containing the BoxB sequence through the λN peptide. Because SGS3 binds to a target associated with a 22-nt sRNA-AGO1–RISC (17), we tethered SGS3 to BoxB sequence-containing RNA. SGS3, to which a λN peptide and FLAG tag at the N terminus and a myc tag at the C terminus were attached (λN-FLAG-SGS3) or that to which a FLAG tag at the N terminus and a myc tag at the C terminus were attached (FLAG-SGS3), FLAG-tagged SDE5 (FLAG-SDE5), and FLAG-tagged RDR6 (FLAG-RDR6) were prepared via in vitro translation using BYL and immunopurified with anti-FLAG antibody gels. We also constructed GFP template DNAs containing three consecutive BoxB sequences at the 5′ region adjacent to the coding region of GFP mRNA or lacking a BoxB sequence (3×BoxB-GFP or GFP, respectively) (Fig. 3A). Considering that RDR6 selects poly(A)-less RNA over polyadenylated RNA as a template (37), these GFP mRNAs were prepared through in vitro transcription as nonpolyadenylated or poly(A30)-tailed RNAs (A₀ or A₃₀, respectively). After the immunopurified fractions and the GFP mRNAs were mixed and incubated, RNA was extracted and analyzed through Northern blot hybridization using the sense or antisense oligoDNA probes to detect GFP-related RNAs (Fig. 3B). When antisense RNA was examined, two bands were detected only in the reaction mixture containing λN-FLAG-SGS3, FLAG-SDE5, FLAG-RDR6, and 3×BoxB-GFP-A₀ (Fig. 3B, Upper lane 9), indicating that antisense GFP RNA was synthesized only under this combination of conditions. Because the upper band corresponded to approximately double the size of 3×BoxB-GFP-A₀, we assumed these RNAs are fold-back (back-priming) extension products of FLAG-RDR6 (41). The lower band corresponded to RNA with a size similar to 3×BoxB-GFP-A₀, and we assumed that these RNAs were synthesized from the 3′ end of 3×BoxB-GFP-A₀ by FLAG-RDR6. No antisense RNA synthesis was detected in reaction mixtures that lacked λN-FLAG-SGS3 or FLAG-SDE5 (Fig. 3B, Upper lanes 1 and 13). Antisense RNA synthesis to 3×BoxB-GFP-A₀ was also observed in the combined presence of λN-peptide– and FLAGtag-fused SDE5 (λN-FLAG-SDE5), FLAG-SGS3, and FLAG-RDR6 (Fig. 3C). These results suggest that RDR6 is recruited to a template RNA onto which SGS3 and SDE5 bind and that AGO1 is not essential for RDR6 recruitment. Unlike in the YG1 extract–containing system, polyadenylation inhibited antisense RNA synthesis in this system utilizing purified proteins (Fig. 3 B and C, Upper lane 11).

Fig. 3.

SGS3 and SDE5 cooperatively recruit RDR6 to a target that serves as a template for dsRNA synthesis. (A) Structures of GFP-related RNAs used for the experiments presented in B and C. (B) Tethering assay using λN-FLAG-SGS3 and 3×BoxB-GFP RNA. λN-FLAG-SGS3, FLAG-SGS3, FLAG-SDE5, and FLAG-RDR6 were prepared through in vitro translation using BYL. Mock indicates the mock translated BYL, to which no mRNA was added. After immunopurification with anti-FLAG antibody gels, the fractions from FLAG peptide–eluted fractions containing each protein were incubated with GFP-related RNA (2.0 nM). RNA was extracted, separated on a denaturing 1.2% agarose gel, transferred onto a Hybond N+ membrane, and analyzed via Northern blotting. Antisense and sense GFP RNA was detected with ³²P-labeled sense or antisense oligoDNA probes (Upper and Lower, respectively), respectively. Antisense GFP-containing RNA (1,004-nt) was used as a control. Immunopurified FLAG-tagged proteins were detected by Western blotting analysis using anti-FLAG antibody. (C) Tethering assay using λN-FLAG-SDE5 and 3×BoxB-GFP RNA. The experiment was performed as described in B, except that SDE5 was tethered. Immunopurified FLAG-tagged proteins were detected by Western blotting analysis using anti-FLAG antibody.

RDR6 Cannot Convert a Poly(A)-tailed RNA into a dsRNA Even in the Presence of 22-nt sRNA–Containing AGO1, SGS3, and SDE5.

RDR6-dependent secondary siRNAs were produced from polyadenylated 3′ cleavage fragments by 22-nt siRNA-AGO1–RISC in YG1 extracts, but purified RDR6 could not synthesize dsRNA from polyadenylated template RNA even in the presence of purified SGS3 and SDE5 (Fig. 3 B and C and SI Appendix, Fig. S6). A possible explanation for this contradiction is that AGO1 is essential for dsRNA synthesis from poly(A)-tailed RNAs by RDR6. To test this possibility, we constructed an additional GFP RNA (GFP-M) of which the gf698 target site was replaced with a 24-nt Luc sequence shown in Fig. 1A (Fig. 4A). We also constructed another GFP-M RNA that contains the gf698 target site in the 5′ region outside of the coding sequence (gf698-GFP-M). These GFP-M RNAs were prepared with or without poly(A30) tail (A₃₀ or A₀, respectively) and then mixed with λN-FLAG-SGS3, FLAG-SDE5, and FLAG-RDR6 in the presence of AGO1-D762A or 22-nt miR-gf698-AGO1-D762A (Fig. 4B). Antisense GFP RNAs from 3×BoxB-GFP-A₀ were detected in the presence of λN-FLAG-SGS3, FLAG-SDE5, and FLAG-RDR6 as shown in Fig. 3 (Fig. 4B, Upper lane 1). No antisense GFP RNA was detected in reaction mixtures containing AGO1-D762A that were not programmed with 22-nt miR-gf698. However, antisense RNAs were detected from gf698-GFP-M-A₀ in reaction mixtures containing 22-nt miR-gf698-AGO1-D762A but not detected from gf698-GFP-M-A₃₀ under the same conditions (Fig. 4B, Upper lanes 9 and 11). These results suggest that RDR6 is specifically recruited to an RNA to which SGS3 or SDE5 binds and that poly(A) tail inhibits dsRNA synthesis by RDR6 even in the copresence of 22-nt siRNA-AGO1, SGS3. and SDE5.

Fig. 4.

RDR6 does not convert a poly(A)-tailed target into a dsRNA even in the presence of AGO1, SGS3, and SDE5. (A) Structures of GFP-related RNAs used for the experiments presented in B. (B) dsRNA synthesis in the presence of AGO1, SGS3, SDE5, and RDR6. Loading of 22-nt miR-gf698 to AGO1-D762A was carried out as described in Fig. 1D. λN-FLAG-SGS3, FLAG-SDE5, and FLAG-RDR6 were prepared as shown in Fig. 3. Fractions from FLAG peptide elution containing each protein were incubated with GFP-M RNAs (2.0 nM). RNA was analyzed as shown in Fig. 3. Immunopurified FLAG-tagged proteins were detected by Western blot analysis using anti-FLAG-antibody.

Discussion

Here, we established an in vitro system of tasiRNA or secondary siRNA production using YG1 extracts, in which secondary siRNAs were produced predominantly from the 3′ fragment of the cleavage product by AGO1-RISCs with 22-nt siRNA, as observed in tasiRNA production by miR173-AGO1-RISCs in vivo. In terms of dsRNA processing, both 21- and 22-nt siRNAs were produced at similar levels, implying that DCL2 and DCL4 are active in YG1 extracts. This system paved the way for further elucidation of the mechanisms of secondary siRNA production. We first confirmed that 22-nt siRNA is important for the induction of secondary siRNA production. The importance of 22-nt siRNAs has been suggested based on the observation that DCL2, which generates 22-nt siRNAs, is important for the induction of secondary siRNA production (23, 24). However, there was no direct evidence showing that 22-nt siRNA–containing AGO1-RISCs induce secondary siRNA production from RNA targets. In the present study, we demonstrated that 22-nt siRNA–programmed AGO1 triggers secondary siRNA production from 3′ cleavage fragments, as is the case for miR173-AGO1–RISCs. In this experimental system, 22-nt siRNA-AGO1–RISCs but not 21-nt siRNA-AGO1–RISCs produced secondary siRNA, and base-pairing between the 3′ nt of miR173 and the 5′ end of its target site was mandatory for tasiRNA production. Together with our previous results showing that SGS3 specifically interacts with sRNA-AGO1–RISC–bound target RNA in part via a 5′-overhanged RNA duplex formed between 22-nt sRNA and the target (17), the present study indicates that formation of this ternary complex is essential to secondary siRNA and tasiRNA production.

We found that adding purified SDE5 and λN-fused SGS3 or that of SGS3 and λN-fused SDE5 but not λN-fused SGS3 or λN-fused SDE5 alone to BoxB-containing RNA enabled RDR6-dependent complementary RNA synthesis. This indicates that SGS3, SDE5, and RDR6 are necessary and sufficient for synthesis of complementary RNA for an RNA target onto which SGS3 or SDE5 is tethered. An addition of recombinant SDE5 or RDR6 into YG1 extracts increased secondary siRNA production, and an addition of both proteins synergistically enhanced it (Fig. 1E, lanes 6, 12, and 15). Furthermore, addition of RDR6-CT dominantly inhibited secondary siRNA production even in the presence of excess SDE5 (Fig. 1E, lane 18). We speculate that RDR6-CT is recruited onto the 22-nt siRNA-AGO1 cleavage fragment by SGS3 and SDE5, stays without initiating the synthesis of dsRNA, and outcompetes endogenous RDR6 in YG1 extracts. We propose that in a natural context, SGS3 binds to 22-nt siRNA-AGO1–RISC-cleaved target RNA to form a ternary complex to which RDR6 is recruited using SDE5, enabling specific RDR6 recruitment to the 3′ cleavage fragment of the target RNA. This situation, in which the copresence of SGS3 and SDE5 is required for RDR6 recruitment, may increase the specificity of RDR6 recruitment to target RNA compared to a hypothetical situation in which SGS3 directly recruits RDR6 without SDE5. It was reported that miR173-RISC–cleaved fragments of TAS1a and TAS2 transcripts accumulate in plants in the order of rdr6 > sde5 > wild-type > sgs3 (16, 19). In addition to involvement of SDE5 in stabilization of those cleavage fragments, we now propose that overaccumulation of cleavage fragments in sde5 plants results from loss of RDR6 requirement as in rdr6 plants (i.e., in sde5 plants, cleaved fragments are somehow stabilized by SGS3 alone but remain to be converted into dsRNA and diced into sRNAs), while in wild-type plants, the cleaved fragments are converted into siRNAs.

Baeg et al. showed that RDR6 alone can synthesize RNA complementary to added template RNA and that its complementary RNA synthesis is inhibited by more than eight polyadenylations at the 3′ terminus of the template RNA (37). The present study revealed that purified RDR6 could also synthesize dsRNA from the 3′ terminus of poly(A)-less RNAs in the presence of purified AGO1-RISC, SGS3, and SDE5 but could not synthesize dsRNA from poly(A)-tailed RNA templates. Meanwhile, tasiRNAs from TAS1a–c and TAS2 are produced by miR173-AGO1–RISC from the 5′-uncapped and 3′-polyadenylated 3′ cleavage fragment in A. thaliana cells (15). In YG1 extracts, RDR6-dependent secondary siRNA production occurs from polyadenylated 3′ cleavage fragments generated by 22-nt siRNA-AGO1–RISC. Furthermore, antisense RNAs of Luc-gf698 produced by 22-nt siRNA-AGO1–RISC were detected in the presence of B2 in YG1 extracts. These findings suggest that poly(A)-tailed RNA serves as a template for dsRNA synthesis by RDR6 in vivo and in cell extracts. To explain these differences, we assume that an additional factor assists synthesis of dsRNA by RDR6 from polyadenylated 3′ cleavage fragments generated by 22-nt siRNA-AGO1–RISCs. The band patterns of dsRNA for Luc-gf698 after S1 nuclease treatment in the presence of B2 were smeared, and the longest dsRNAs were slightly shorter than the 3′ cleavage fragments (including poly(A)). These observations suggest that RDR6 does not begin synthesizing dsRNA from the 3′ poly(A) end of the target RNA or a specific position but from broad regions upstream of the poly(A) tail. In Caenorhabditis elegans, which is equipped with RDR-dependent siRNA amplification triggered by RISCs, the endoribonuclease RDE-8 promotes siRNA amplification by RDR (42). Although there is no RDE-8 homolog in plants, a similar endoribonuclease may be involved in this pathway. Alternatively, an unidentified factor may mediate internal entry of RDR6 upstream of the poly(A) tail on a temple for dsRNA formation without removal of the poly(A) tail. In either case, RDR6 is specifically recruited to a 22-nt-siRNA-RISC-targeted poly(A)-containing RNA via SGS3 and SDE5 and synthesizes its complementary RNA, probably at low efficiency. Once secondary siRNA is produced, the secondary siRNA-loaded RISCs cleave the target RNA, generating truncated, poly(A)-deficient target RNA. RDR6 then synthesizes its complementary RNA more efficiently, accelerating the amplification of secondary siRNA. Thus, the inefficiency of synthesis of the first complementary RNA by RDR6 may act as a safeguard against nonspecific siRNA amplification on nontarget mRNA.

Materials and Methods

Preparation of YG1 Extracts.

The suspension-cultured A. thaliana YG1 cell line was previously established from seedlings of the wild-type Col-0 ecotype and was maintained as described previously (25). To prepare protoplasts, the cells were treated with 1% Cellulase Onozuka RS (Yakult Pharmaceutical) and 0.2% Pectolyase Y-23 (Kyowa Chemical Products) in a solution of 3 mM MES-KOH (pH5.8), 0.6 M mannitol, 0.1 M sucrose, 5 mM CaCl₂, 0.1 mg/mL myo-inositol, 1 μg/mL thiamine, 0.2 μg/mL 2,4-dichlorophenoxyacetic acid, and 1×Murashige and Skoog Plant Salt Mixture (FUJIFILM Wako Pure Chemical) for 2 to 3 h at room temperature. After washing three times with mannitol washing buffer (3 mM MES-KOH [pH 5.8] and 0.7 M mannitol), the protoplasts were overlaid on a gradient of 2 mL at 70% (vol/vol), 4 mL at 34% (vol/vol), 3 mL at 34% (vol/vol) and 3 mL at 10% (vol/vol) Percoll (GE Healthcare) in 5 mM Pipes-KOH (pH 7.0), 0.7 M mannitol, and then centrifuged at 10,000 × g at 25 °C for 1 h. The protoplasts were recovered from the interface between 34% and 70% Percoll solutions. After washing with mannitol washing buffer, the protoplasts were mixed with the translation reaction (TR) buffer (30 mM Hepes-KOH [pH 7.4], 80 mM potassium acetate, 1.8 mM magnesium acetate, 2 mM dithiothreitol, and EDTA-free cOmplete protease inhibitor), homogenized with a Dounce homogenizer, and centrifuged at 800 × g at 4 °C for 10 min. The supernatant was recovered and stored at −80 °C. YG1 extracts were obtained from the fresh supernatant after centrifugation of the stored-supernatant with 30,000 × g at 4 °C for 15 min

In Vitro siRNA Loading into FLAG-AGO1.

In vitro siRNA loading into FLAG-AGO1 was performed using BYL as previously described (28) but with incubation for 60 min at 25 °C after adding sRNA duplexes. After incorporation of siRNAs into FLAG-AGO1, the mixtures were combined with 10 μL of EZview Red ANTI-FLAG M2 Affinity Gel (Sigma-Aldrich) and 10 μL of TR buffer and processed at 4 °C for 2 h on a rotating wheel. The gels were washed with 100 μL of TR buffer three times. The FLAG peptide elution was carried out using TR buffer containing 170 μg/mL 3×FLAG at 4 °C for 30 min.

Analysis of In Vitro Secondary siRNA Production.

A representative 20-μL reaction mixture contained ∼8.5 μL of FLAG peptide–eluted fraction, 3.8 μL of YG1 extract, 1 μL of recombinant SDE5, and 1 μL of 40 nM ³²P-labeled RNA. The reaction solution contained 30 mM Hepes-KOH (pH 7.4), 80 mM potassium acetate, 2.7 mM magnesium acetate, 1 mM ATP, 0.5 mM GTP, 0.5 mM CTP, 0.5 mM UTP, 0.2 μg/μL creatine kinase (Roche), 20 mM creatine phosphate, and 0.8 unit/ μl of RNasin Plus RNase Inhibitor (Promega). After incubation for 20 min at 25 °C, RNA was extracted using Tris-EDTA–saturated phenol, mixed with 1.6 volumes of urea dye (12.5 M urea, 0.2 mg/mL bromophenol blue [BPB], and 0.2 mg/mL xylene cyanol [XC]) or a quarter-volume of native dye solution (2.5 × TBE, 0.5 mg/mL BPB, 0.5 mg/mL XC, and 25% [vol/vol] glycerol) for analyses of target cleavage or secondary siRNA production, respectively. Target cleavage and secondary siRNA production were analyzed with 5% 7M urea denaturing polyacrylamide gels and 15% native polyacrylamide gels, respectively, using 0.5 × TBE as the running buffer. Radiolabeled products were detected using the Typhoon FLA 7000 imaging system (GE Healthcare).

Sequences of Proteins and Targets.

The representative sequences of proteins and targets used in this study are shown in SI Appendix, Table S4.

Data Availability

The accession number for the sRNA-seq data were deposited in the DNA Data Bank of Japan (http://www.ddbj.nig.ac.jp/) Sequence Read Archive under the accession number DRA010397. All other data supporting this study are included in this article and/or SI Appendix.