Roles of target site location and sequence complementarity in trans-acting siRNA formation in Arabidopsis


  • Changqing Zhang,

    1. Section of Molecular Cell and Developmental Biology and Institute for Cellular and Molecular Biology, The University of Texas at Austin, Austin, TX 78712, USA
    Search for more papers by this author
  • Danny W.-K. Ng,

    1. Section of Molecular Cell and Developmental Biology and Institute for Cellular and Molecular Biology, The University of Texas at Austin, Austin, TX 78712, USA
    Search for more papers by this author
  • Jie Lu,

    1. Section of Molecular Cell and Developmental Biology and Institute for Cellular and Molecular Biology, The University of Texas at Austin, Austin, TX 78712, USA
    Search for more papers by this author
  • Z. Jeffrey Chen

    Corresponding author
      (fax +1 512 471 2149; e-mail
    Search for more papers by this author

(fax +1 512 471 2149; e-mail


In plants, many mRNAs and non-coding RNAs are cleaved by RNA-induced silencing complexes. After cleavage, only a limited number of RNAs are processed into trans-acting siRNAs (tasiRNAs). One reason is that 22 nt small RNAs, but not the more common 21 nt small RNAs, can efficiently trigger tasiRNA formation. The characteristics of the target transcripts may also affect tasiRNA production. Here we report the effects of target site location and sequence complementarity on tasiRNA formation. A synthetic sequence that included a miR173 target site and two siRNAs targeting an endogenous mRNA encoding PHYTOENE DESATURASE3 was introduced into a protein-coding (GFP) gene in the coding region or 3′ UTR. tasiRNAs were generated in the transgenic seedlings, and the PDS3 mRNA level was reduced, leading to a photobleaching phenotype. It was found that tasiRNAs were most efficiently produced when the miR173 target site was placed immediately after the stop codon. Introducing premature stop codons caused a dramatic reduction of tasiRNAs and over-accumulation of 3′ cleavage products, suggesting positive effects of translation on processing the 3′ cleavage products into tasiRNAs. By systematically mutating the miR173 target site, we found that perfect complementarity between the 3′ end of miR173 and the 5′ end of the target sequence was crucial. Mismatches at that position abolished tasiRNA formation, but mismatches at the 5′ end of miR173 had less effect. These data suggest important roles for translation and specific sequence complementarity in tasiRNA formation, providing new insights into tasiRNA biogenesis as well as a strategy for improving the efficiency of RNA interference (RNAi) using tasiRNAs.


Trans-acting siRNAs (tasiRNAs) are 21 nt siRNAs that are generated from tasiRNA-generating (TAS) genes and function in regulating target mRNAs through mRNA degradation or translational repression (Peragine et al., 2004; Vazquez et al., 2004). Biogenesis of tasiRNAs is triggered by miRNA-guided cleavage of TAS transcripts. One of the cleavage products, usually the 3′ product, is converted to double-stranded RNA (dsRNA) by RNA-DEPENDENT RNA POLYMERASE6 (RDR6). Starting from the miRNA-guided cleavage site, the dsRNA is processed step by step (known as ‘phasing’) into 21 nt siRNA duplexes by DICER-LIKE4 (DCL4) (Peragine et al., 2004; Allen et al., 2005; Xie et al., 2005; Yoshikawa et al., 2005). One strand of the tasiRNA duplex is selectively incorporated into an RNA-induced silencing complex (RISC) that recognizes and cleaves target mRNA or represses translation.

The Arabidopsis thaliana genome contains eight TAS genes belonging to four families (Montgomery et al., 2008b). Transcripts of the four genes in the TAS1 and TAS2 families are targeted by miR173 to initiate formation of tasiRNAs, which target multiple mRNAs, including several that encode pentatricopeptide repeat proteins (Peragine et al., 2004; Vazquez et al., 2004; Allen et al., 2005; Axtell et al., 2006; Chen et al., 2007; Howell et al., 2007). Three TAS3 transcripts are targeted by miR390, and the resulting tasiRNAs control the mRNA levels of AUXIN RESPONSE FACTOR3 (ARF3) and ARF4, which regulate the juvenile-to-adult transition and leaf polarity (Allen et al., 2005; Williams et al., 2005; Adenot et al., 2006; Fahlgren et al., 2006; Garcia et al., 2006; Hunter et al., 2006). The TAS4 transcript is a target of miR828, and TAS4 tasiRNAs regulate the expression of several MYB transcription factor genes (Rajagopalan et al., 2006). More genes have been identified as potential TAS genes, and the triggers include both miRNAs and siRNAs (Lu et al., 2006; Ronemus et al., 2006; Chen et al., 2007).

In contrast to TAS transcripts, most target RNAs do not produce siRNAs following miRNA-guided cleavage. The ‘two-hit trigger’ model suggests that double-cleavage events trigger siRNA biogenesis (Axtell et al., 2006). Consistent with this model, all A. thaliana genes with two known target sites, including TAS3 transcripts, were found to form phased siRNAs. Alternatively, miR390 is specifically associated with ARGONAUTE7 (AGO7), suggesting a distinct role for miR390–AGO7 complexes in TAS3 tasiRNA biogenesis (Montgomery et al., 2008a).

Unlike TAS3, transcripts of the TAS1, TAS2 and TAS4 families each contain only one miRNA target site. The trigger miRNAs miR173 and miR828 are specifically associated with AGO1 through a 5′ terminal uridine (Mi et al., 2008; Montgomery et al., 2008a). Montgomery et al. (2008b) proposed that the miR173–AGO1 complex possesses specialized functionality in initiating phased siRNA formation. Indeed, introduction of a single miR173 target site into a transcribed foreign sequence is sufficient to convert it into a TAS gene (Montgomery et al., 2008a; Felippes and Weigel, 2009). Similarly, a heterologous sequence flanked by the two authentic miR390 target sites found in TAS3 acquires the ability to produce siRNAs (Felippes and Weigel, 2009).

Recent studies have found that 22 nt small RNAs, rather than the more common 21 nt miRNAs, have the property to trigger tasiRNA formation (Chen et al., 2010; Cuperus et al., 2010). However, this cannot fully explain why most miRNA-targeted transcripts do not produce tasiRNAs. Transcripts targeted by the same 22 nt small RNA may not be equally efficient in generating tasiRNAs. Non-translatability may promote tasiRNA formation (Montgomery et al., 2008b), but the effect of translatability on tasiRNA formation has not been systematically tested. Using a modified synthetic tasiRNA (syn-tasiRNA) strategy (Montgomery et al., 2008a,b; Felippes and Weigel, 2009), we tested the efficacy of tasiRNA formation in a protein-coding gene by inserting the miR173 target site at various locations in the gene. We found evidence for positive effects of translation on tasiRNA formation. The alignments between miRNA sequences and corresponding TAS transcript sequences possessed perfect matches at the ends of alignments, and mismatches or ‘wobbles’ close to the canonical cleavage sites. This is in contrast to general miRNA and target alignments, which often show mismatches at either the 5′ or 3′ ends. By systematically mutating a functional target sequence to create different complementarity, we found that central mismatches/wobbles had little or no effect, but perfect matches at the 3′ end of miR173 were crucial for tasiRNA formation. These data provide new insights into tasiRNA biogenesis, and strategies for practical applications. As RNAi has become an important strategy in genetic engineering, such understanding is critical to the design of siRNAs or targets in order to avoid formation of tasiRNAs, which may cause downregulation of non-target genes. Moreover, the knowledge is also necessary for exploiting the full potential of tasiRNA as a new RNAi strategy.


Formation of syn-tasiRNAs from a protein-coding gene

To express tasiRNAs from a protein-coding gene, we created a 96 bp sequence, comprising 54 bp from TAS1c (At2g39675) and 42 bp of additional sequence (Figure 1a). The TAS1c sequence included the miR173 target site (22 bp) and 32 bp of downstream sequence. The additional 42 bp included the sequences of two synthetic tasiRNAs (syn-tasiRNAs), which were designed to target the mRNA encoding PHYTOENE DESATURASE 3 (PDS3) using an online program, WMD3 ( (Ossowski et al., 2008; Schwab et al., 2009). The sequences of these syn-tasiRNAs, which were designated PDS3-1 and PDS3-2, were arranged at DCL4 processing cycles 3 and 4 from the miR173-guided initiation site (Montgomery et al., 2008a). For convenience of discussion, this 96 bp sequence was termed a syn-tasiRNA (production) cassette. The syn-tasiRNA cassette was inserted at various locations in a transgene expressing an HTA6–GFP fusion protein under the control of a modified 35S promoter (Figure 1b). HTA6 is a core histone H2A from Arabidopsis. The resulting constructs were named on the basis of the insertion positions as A (between HTA6 and GFP coding sequences, in-frame fusion), B (1 bp downstream of stop codon), C (16 bp downstream of stop codon), and D (99 bp downstream of stop codon), respectively. The constructs and the vector control (V) were transformed into A. thaliana (Col-0) plants using the floral dip method (Clough and Bent, 1998).

Figure 1.

 Formation of syn-tasiRNAs from a protein-coding gene.
(a) A syn-tasiRNA production cassette. The sequences at positions 3′D1(+) and 3′D2(+) are from TAS1c. The sequences of two syn-tasiRNAs targeting PDS3 mRNA are located at 3′D3(+) and 3′D4(+).
(b) Gene structure. Arrows indicate positions in which the syn-tasiRNA cassette is inserted. The resulting constructs are designated A, B, C and D, respectively.
(c) Screening for T1 seedlings on an MS agar plate supplemented with 40 mg L−1 hygromycin. Left panel: a representative plate after 17 days of culture under reduced light intensity. Middle panel: enlarged view of transgenic plants expressing construct D and GFP signals detected in roots of plants expressing the empty vector (V) or construct A. Right panel: hygromycin-resistant seedlings transferred to a fresh plate.
(d) Quantification of hygromycin-resistant seedlings showing complete photobleaching (white), partial photobleaching (yellow) or no sign of photobleaching (green). The results include seedlings from six screening plates.
(e–g) Relative expression levels of syn-tasiRNA (e), PDS3 mRNA (f) and HTA6–GFP mRNA (g) after normalization to UBQ10 mRNA, as determined by quantitative RT-PCR. The level for the vector control is set to 1. Error bars represent the standard deviation of three replications.

The T1 seedlings transformed with the syn-tasiRNA constructs were extremely sensitive to light. Seedlings were bleached and stopped growing before the emergence of true leaves, indicating that the two syn-tasiRNAs targeting PDS3 mRNA functioned effectively. To make comparisons under uniform conditions, we screened transgenic seeds containing various constructs on the same plate and under reduced light intensities (Figure 1c and Figure S1). Seedlings transformed with the vector alone produced green cotyledons and true leaves at approximately 16 days after germination (Figure 1c, section V). In contrast, most seedlings transformed with syn-tasiRNA constructs had photobleached cotyledons (Figure 1c, sections A–D). The cotyledons first appeared purple and later turned white (Figure 1c, middle panel, section D), similar to the pds3 knockout mutant phenotype (Qin et al., 2007). Some seedlings had green cotyledons, but the emerging true leaves were photobleached. To confirm the presence of the transgenes, we examined GFP signals in roots of hygromycin-resistant seedlings. Strong signals were detected in the roots transformed with the vector alone (Figure 1c, middle panel, section V), but weak signals were found in the roots transformed with syn-tasiRNA constructs (Figure 1c, middle panel, section A). The transgenic seedlings were grouped on the basis of appearance and counted (Figure 1d). Of the four syn-tasiRNA constructs, B produced the highest frequency of photobleached seedlings, and showed the highest accumulation of syn-tasiRNAs and the most severe reduction of PDS3 mRNA (Figure 1e,f). The abundance of HTA6–GFP mRNA decreased to various levels in the various transgenic constructs (Figure 1g). No obvious difference in miR173 abundance was observed. The data suggest that tasiRNAs are efficiently produced from a protein-coding gene, and that the efficiency varies depending on the insertion site of the syn-tasiRNA cassette in the protein-coding gene.

Position effect of the syn-tasiRNA cassette on tasiRNA formation

Construct B produced the highest level of syn-tasiRNAs. To determine whether the flanking sequence or the position of the syn-tasiRNA cassette relative to the open reading frame (ORF) in construct B made it more efficient, construct B was modified to generate two constructs (Figure 2a). In construct B + uTAG, a new stop codon (TAG) was introduced at 732 bp upstream of the original stop codon. In construct B + dTAG, the original stop codon was removed to extend the ORF by 105 bp. Compared with construct B, transformation with construct B + uTAG produced significantly fewer white seedlings, and the frequency of white seedlings after transformation with construct B + dTAG also decreased slightly (Figure 2b and Figure S2). This reduction in the number of white seedlings correlated with reduced amounts of syn-tasiRNAs and increased levels of PDS3 mRNA (Figure 2c,d). As constructs B and B + uTAG had exactly the same flanking sequence around the cassette, the reduction of syn-tasiRNAs in B + uTAG transgenic plants was due to the increased distance between the end of the ORF and the syn-tasiRNA cassette.

Figure 2.

 Effect of position relative to ORF on tasiRNA formation.
(a) Construct B is the same as in Figure 1(b). Construct B + uTAG has a stop codon introduced at 732 bp upstream of the original stop codon. Construct B + dTAG has the original stop codon removed, resulting in an ORF extended by 105 bp.
(b) Quantification of hygromycin-resistant seedlings according to the same criteria as in Figure 1(d). The results include seedlings from three screening plates.
(c–e) Relative expression levels of syn-tasiRNA PDS3-1 (c), PDS3 mRNA (d) and HTA6–GFP mRNA (e) after normalization to UBQ10 mRNA, as determined by quantitative RT-PCR. The level for construct B is set to 1. Error bars represent the standard deviation of three replications.

The reduction in syn-tasiRNAs is accompanied by over-accumulation of 3′ cleavage products

Compared with construct B, the syn-tasiRNA construct B + uTAG produced a lower level of syn-tasiRNAs. This may be due to a lower level of HTA6–GFP mRNA or lower efficiency in processing this mRNA into tasiRNAs. As the stability of mRNA is often reduced when translation is terminated early by a nonsense codon (Muhlrad and Parker, 1994; van Hoof and Green, 1996), introduction of a premature stop codon in B + uTAG may lead to lower stability and thus a lower level of HTA6–GFP mRNA. However, the level of HTA6–GFP mRNA was slightly higher in B + uTAG transgenic plants than in B transgenic plants (Figure 2e), ruling out the possibility of mRNA instability. Alternatively, miR173-guided cleavage may be similarly efficient for both transcripts, but the 3′ cleavage products of B + uTAG may be less efficiently processed into tasiRNAs. This hypothesis suggests that more 3′ cleavage products will accumulate in B + uTAG transgenic plants than in B transgenic plants. In quantitative RT-PCR analysis using a pair of primers 3′ to the miR173 target site, thus detecting both full-length HTA6–GFP mRNA and 3′ cleavage products with poly(A) tails, more 3′ cleavage products were indeed observed in B + uTAG transgenic plants (Figure S3). Consistent with the quantitative RT-PCR results, RNA blot assays showed that the signal for 3′ cleavage products was approximately four times higher in B + uTAG transgenic plants than in B transgenic plants, whereas the signal for syn-tasiRNA in B + uTAG transgenic plants is approximately one-fifth of that in B transgenic plants (Figure 3a). The data suggest that the reduced level of syn-tasiRNAs in B + uTAG transgenic plants is caused by a lower efficiency of processing of the 3′ cleavage products into tasiRNAs.

Figure 3.

 Accumulation of syn-tasiRNA and 3′ cleavage products.
(a,b) RNA blot analysis of transgenic seedlings as shown in Figure 1(c) (b) or in Figure S2 (a).
Signal intensities of the 3′ cleavage products and syn-tasiRNA (PDS3-1), after normalization to that of U6, are shown below. The probe used for detecting syn-tasiRNA PDS3-1 is indicated in Figure 1(a).

The accumulation of 3′ cleavage products was also compared between plants expressing syn-tasiRNA constructs B, C and D, which had the syn-tasiRNA cassette inserted in the 3′ UTR, but at various distances from the stop codon. Plants expressing constructs C and D, which produced lower levels of syn-tasiRNAs than plants expressing construct B, showed greater accumulation of 3′ cleavage products (Figure 3b). Notably, construct B had the syn-tasiRNA cassette inserted 1 bp downstream of the stop codon, but it was inserted at 16 bp downstream of the stop codon in construct C. It was designed this way, because, for mammalian miRNA, target sites within the first 15 nt of the 3′ UTR are less effective than those located in further downstream regions, and the first 15 nt is believed to be cleared of RISCs by ribosomes entering stop codon (Grimson et al., 2007; Bartel, 2009). The fact that this subtle difference causes a dramatic effect suggests that translation may have a positive effect on processing the 3′ cleavage products into tasiRNAs.

A positive effect on tasiRNA formation exists when the miR173 target site is <10 nt downstream of the stop codon

Based on the above observations, we predicted a positive effect of translation on tasiRNA formation, and this effect was predicted to disappear when the miR173 target site is located at certain distance downstream of the stop codon. To test this, a series of constructs based on construct B were generated by inserting a stop codon (TAG) at 7, 10, 13, 16, 19 and 364 bp upstream of the miR173 target site. These constructs were named TAG-7, TAG-10, TAG-13, TAG-16, TAG-19 and TAG-364, respectively. The transgenic plants expressing these constructs were analyzed together with those expressing constructs B or B + uTAG, in which the stop codon (TAG) is located 1 and 736 bp, respectively, upstream of the miR173 target site (Figure 4a). In T1 seedlings transformed with the constructs B, TAG-7 and TAG-10, more severe photobleaching was observed than in those transformed with TAG-13, TAG-16, TAG-19, TAG-364 and B + uTAG (Figure S4). No obvious phenotypic difference was found among the latter group. Using quantitative RT-PCR, similar low levels of syn-tasiRNAs and high levels of PDS3 mRNA were detected in transgenic plants expressing the group of constructs that had the stop codon inserted 13 bp or more upstream of the miR173 target site (Figure 4c,d). A gradual decrease in syn-tasiRNA levels was observed in plants expressing constructs B, TAG-7 and TAG-10 (Figure 4c), but this trend was not observed when the same RNA samples were analyzed by small RNA Northern blotting (Figure 4b). This discrepancy may reflect the fact that quantitative RT-PCR detects only precisely processed (phased) tasiRNAs, while Northern blotting detects both phased and phase-shifted tasiRNAs. Overall, these data indicated a positive effect on tasiRNA formation when the miR173 target site is positioned close to the stop codon in the 3′ UTR (10 nt or fewer).

Figure 4.

 Effects of stop codons on syn-tasiRNA formation.
(a) Structure of a series of constructs with a stop codon inserted into construct B at 7, 10, 13, 16, 19 and 364 bp upstream of the miR173 target site. The resulting constructs were named TAG-7, TAG-10, TAG-13, TAG-16, TAG-19 and TAG-364, respectively.
(b) RNA blot analysis of transgenic seedlings as shown in Figure S5.
(c,d) Relative expression levels of syn-tasiRNA PDS3-1 and PDS3-2 (c) and PDS3 mRNA (d) after normalization to UBQ10 mRNA, as determined by quantitative RT-PCR. The level for construct B is set to 1. Error bars represent the standard deviation of three replications.

Effects of complementarity between miRNA and target sequences on tasiRNA formation

The alignments between TAS transcripts and miRNAs showed perfect complementarity at the ends, but contained mismatches/wobbles close to the cleavage sites (Figure S5). To determine how these features influence tasiRNA formation, we created a series of constructs based on construct B (Figure 5a). In construct 173mTa, the mismatch at position 9 and the G:U wobble at position 15 were fixed, resulting in perfect complementarity between miR173 and the target sequence. In construct 173mTb, mismatches were introduced at both ends, and the central mismatch and wobble pair remained unchanged. Construct 173mTc combined the mutations in the first two constructs, resulting in a target site with perfect complementarity in the middle and mismatches at the ends. A control construct (control sequence 1), designated Cseq1, was produced by reshuffling the target sequence to drastically reduce complementarity with miR173. These five constructs (including construct B) were examined for syn-tasiRNA formation. As expected, plants expressing control construct Cseq1 produced green seedlings (Figure S6a,d). Plants expressing construct 173mTa produced similar levels of photobleached seedlings to those expressing construct B, suggesting that the central mismatch and wobble do not have an obvious effect on tasiRNA formation. In contrast, plants expressing constructs 173mTb and 173mTc produced mostly green seedlings and no white seedlings. The syn-tasiRNA levels determined by quantitative RT-PCR and RNA blotting were higher in the plants expressing constructs B and 173mTa than in those expressing 173mTb and 173mTc, whose levels were similar to that of plants expressing Cseq1 (Figure 5b, left panel, and Figure S6b). Consistently, PDS3 mRNA levels decreased dramatically only in plants expressing constructs B and 173mTa (Figure S6c). To determine whether the mutations introduced into the target site prevented cleavage, a ligation-mediated 5′ RACE assay was performed for all constructs. The results indicated that cleavage events predominantly occur at the canonical position within the miR173 target site or variants, but this was not the case for the control Cseq1, which had no significant complementarity with miR173 (Figure 5c). This suggests that the mutations do not prevent the targets from being cleaved, although the cleavage efficiencies varied. The reduction of syn-tasiRNA production by 173mTb and 173mTc was not accompanied by over-accumulation of 3′ cleavage products (Figure 5b, left panel). Construct 173mTb had mismatches introduced to both ends of the target site. To determine whether the 5′ end or the 3′ end was critical for tasiRNA formation, we generated two additional constructs, 173mTd and 173mTe, which retained mismatches only at the 5′ end and only at the 3′ end, respectively (Figure 5a). T1 seedlings transformed with these two constructs were compared with those transformed with constructs B and 173mTb (Figure S7d). Similarly to construct 173mTb, plants expressing construct 173mTd produced mostly green seedlings and no white seedlings, while plants expressing construct 173mTe produced mostly white and yellow seedlings (Figure S7a). In RNA blot analysis, syn-tasiRNAs were detected in plants expressing construct 173mTe, but not in plants expressing construct 173mTd (Figure 5b, right panel). When measured using quantitative RT-PCR, the syn-tasiRNA level in plants expressing construct 173mTe was slightly lower than in those expressing construct B, while plants expressing constructs 173mTb or 173mTd showed similar low levels (Figure S7b). Correspondingly, similar high levels of PDS3 mRNA were detected in plants expressing constructs 173mTb or 173mTd (Figure S7c). Collectively, these data show that perfect complementarity between the 5′ end of the target site and the 3′ end of miR173 is crucial for tasiRNA formation.

Figure 5.

 Effects of complementarity between miR173 and target site on tasiRNA formation.
(a) Alignments between miR173 and its target sequence (found in TAS1c) or mutated sequences (indicated by italicized bold letters). Open circle: wobble base pair; arrows: cleavage site.
(b) RNA blot analyses of transgenic seedlings shown in Figure S7(d) (left panel) and Figure S8(d) (right panel).
(c) Ethidium bromide-stained 5′ RACE products. The amplicons of Actin2/8 cDNA are shown in the bottom panel as an input control.


In this study, we used a modified synthetic tasiRNA (syn-tasiRNA) system to examine the characteristics of TAS transcripts that are important for tasiRNA formation. Non-translatability was thought to promote tasiRNA formation, and non-coding sequences were used to generate artificial TAS genes (Montgomery et al., 2008b). However, how translation affects tasiRNA formation had not been studied systematically. By inserting a syn-tasiRNA cassette into the coding region and various locations in the 3′ UTR of a protein-coding gene (HTA6–GFP), we determined how translation suppresses tasiRNA formation. The results were in contrast with the previous prediction, and suggested a positive effect of translation on tasiRNA formation. The highest efficiency of tasiRNA formation was observed when the syn-tasiRNA cassette was placed immediately after the stop codon (construct B). In comparison, placing the cassette at 16 nt downstream of the stop codon (construct C) caused considerable reduction in tasiRNA formation. It was predicted that RISCs bound to the first 15 nt of 3′ UTR are still on the path of ribosomes. (Grimson et al., 2007; Bartel, 2009). In that sense, ribosomes would reach the miR173-RISCs associated with transcripts of construct B, while for transcripts of construct C, the ribosomes would dissociate before reaching the edge of miR173 target site. Although the exact mechanism is unknown, the interaction between the translation machinery and the miR173-associated RISCs may have a positive effect on processing the 3′ cleavage products into tasiRNAs. Evidence for the existence of such an interaction came from the study using a series of constructs bearing premature stop codons. The efficiency of tasiRNA formation decreases with increasing distance between the stop codon and the miR173 target site, and reaches a steady level when the distance is 13 nt or more. This is slightly different from the distance of 16 nt or more previously found to be required (Grimson et  al., 2007; Bartel, 2009). A recent study using genome-wide ribosome footprint analysis (Ingolia et al., 2009) indicated that ribosomes could reach approximately 7–10 nt after the stop. These new data are consistent with our observations.

A recent study found that the last approximately 50 codons of eukaryotic mRNAs are the most efficiently translated (Tuller et al., 2010). This suggests that the ribosomes have better opportunities to interact with the miR173-associated RISC complex if the target site is placed in this region. If the miR173 target is placed in the middle of the ORF (as in construct A), there are fewer chances for the ribosomes to interact with the miR173-associated RISC complex. This speculation may help to explain why the efficiencies were lower when the syn-tasiRNA cassettes were placed in the middle of the ORF (constructs A and B + dTAG). The question of why endogenous TAS genes are all non-coding prompted us to analyze the four TAS transcripts that are targeted by miR173. We found that each transcript had a short ORF (132–177 nt), and the miR173 target site was close to the stop codon (14–23 nt upstream of the stop codon) (Figure S8). This suggests that translation is involved in miR173-induced tasiRNA formation from endogenous TAS transcripts. However, there is no experimental evidence that these endogenous short ORFs are translated. It remains to be tested whether translation has an impact on any endo-genous TAS genes. In addition, the over-expression system used in our study is likely to induce secondary siRNA formation. Future experiments should be designed to test a direct link between tasiRNA formation and ribosome/translation activities.

Introduction of premature stop codons dramatically reduced tasiRNA formation. This observation is remarkably similar to a phenomenon observed in studies of sense co-suppression in petunia (Napoli et al., 1990; Que et al., 1997). Sense co-suppression refers to the phenomenon that over-expression of a coding sequence homologous with an endogenous gene causes silencing of both the transgene and the endogenous gene. The authors found that early nonsense codons do not affect the mRNA abundanceof the transgene, but they do dramatically reduce the frequency and degree of co-suppression. It is possible that the transgene mRNAs are targeted by unidentified miRNAs or siRNAs to induce formation of tasiRNAs, which could target and silence the endogenous gene. When early nonsense codons are introduced, tasiRNA formation is impaired, leading to a reduced level of co-suppression. In support of this possibility, phased siRNAs were detected in petunia flowers undergoing co-suppression (De Paoli et al., 2009).

An intriguing observation is that the reduced tasiRNA formation in transgenic seedlings expressing the constructs C, D or B + uTAG relative to seedlings expressing construct B correlates with over-accumulation of 3′ cleavage products. This suggests that miR173-guided cleavage and processing the 3′ cleavage products into tasiRNAs are separate events. Cleavage by the miR173-associated RISC does not necessarily route the 3′ products into a tasiRNA formation pathway. The accumulation of 3′ cleavage products may also indicate protection of the 5′ end from digestion by the 5′→3′ exonuclease XRN4 (Souret et al., 2004; Valencia-Sanchez et al., 2006). This protection effect is correlated with the observation that functional target sites show perfect complementarity with the ends of miR173. When the target site is mutated to introduce mismatches with miR173 at the ends, as in 173mTb and 173mTc, neither 3′ cleavage products nor tasiRNAs accumulate. This may be explained by the possibility that the 3′ cleavage products are not protected due to the mismatches, and are quickly degraded, thus preventing them from being processed into tasiRNAs. It was shown previously that perfect complementarity between the 3′ end of miR171 and its target was required for the 5′ cleavage products to be processed into secondary siRNAs (Moissiard et al., 2007). The alignments between most plant miRNAs and their targets possess mismatches, which are highly conserved (Ronemus et al., 2006). These mismatches are more frequent at the ends than the middle of the alignments. A possible role for the conserved mismatches may be to prevent the cleavage products from being processed into siRNAs. In addition to the requirement for 22 nt sRNA triggers, this explains why most miRNA target transcripts do not produce tasiRNAs. These observations may have implications for secondary siRNA formation in hybrids and allopolyploid plants formed between species. In such plants with diverged genomes, secondary siRNA formation may be disrupted due to sequence variation in the trigger siRNAs or target transcripts (Ha et al., 2009).

Syn-tasiRNA has been proposed to be a useful tool for RNAi (Montgomery et al., 2008b), but its efficiency needs to be improved. We used the same miR173 target sequence as reported by Montgomery et al., (2008b) in the syn-tasiRNA cassette and the resulting tasiRNAs target the same PDS3 mRNA. However, the efficiency in our study is dramatically higher in terms of photobleaching phenotype and reduction of PDS3 mRNA. Most T1 seedlings were completely bleached before the emergence of true leaves, resembling pds3 knockout mutants (Qin et al., 2007). Three possible factors are believed to contribute to the higher efficiency: (i) a stronger 35S promoter, (ii) a protein-coding gene that employs the translational leader sequence of tobacco etch virus to enhance translation (Carrington and Freed, 1990), and (iii) more efficient syn-tasiRNAs. Together, the results obtained from this study will be valuable for designing efficient RNAi strategies using tasiRNAs.

Experimental procedures

Plasmid construction

A plasmid vector, pCMIRT15, was modified to create the syn-tasiRNA expression constructs used in this study. It expresses an HTA6–GFP fusion protein (Sheen et al., 1995; Zhang et al., 2005) under the control of a modified 35S promoter, which includes two 35S enhancers and the tobacco etch virus translation leader sequence (Sheen et al., 1995). The 35S terminator sequence was modified to create a multiple cloning site at 92 bp downstream of the stop codon. A hygromycin resistance gene (HPT) under the control of the mannopine synthase promoter was included to facilitate selection of transgenic seedlings. Sequence information for pCMIRT15 has been deposited in GenBank under accession number HQ540318. A mixture of two DNA oligos, 5′-GTGATTTTTCTCTACAAGCGAATAGACCATTTATCGGTGGATCTTAGAAAATTATCTAC-3′ and 5′-GCGCAACTTAATTCTAGGATAAACGCTTTGTATGCCAGTAGATAATTTTCTAAGATCCAC-3′ was used as template for PCR amplification of the 96 bp syn-tasiRNA cassette. Primer pairs A-F/A-R, B-F/B-R, C-F/C-R and D-F/D-R were used to amplify the cassette, and the amplicons were digested and inserted into pCMIRT15 at positions A, B, C and D of the transgene (Figure 1b). The resulting plasmids were named pCMIRT15-A, -B, -C and -D. To produce construct B + uTAG, a 56 bp DNA fragment was generated by annealing two DNA oligos, B + uTAGf and B + uTAGr, followed by filling-in of 5′ overhangs using AccuPrime Pfx DNA polymerase (Invitrogen, The DNA fragment was then digested using PmlI and NcoI and ligated into pCMIRT15-B to replace the sequence between restriction sites PmlI and NcoI, resulting in plasmid pCMIRT15-B + uTAG. To produce construct B + dTAG, a 489 bp sequence was amplified from pCMIRT15-B using primers B + dTAGf and BstR and digested using BamHI and XhoI. The digested product was ligated into pCMIRT15-B to replace the sequence between restriction sites BamHI and XhoI, resulting in plasmid pCMIRT15-B + dTAG. To generate syn-tasiRNA constructs 173mTa, 173mTb, 173mTc, Cseq1, 173mTd and 173mTe, forward primers 173mTaF, 173mTbF, 173mTcF, Cseq1F, 173mTdF and 173mTeF were used in combination with reverse primer BstR to amplify 487 bp sequences from pCMIRT15-B. The amplicons were digested using XbaI and XhoI, and inserted into pCMIRT15-B between restriction sites XbaI and XhoI. To generate constructs TAG-n (where = 7, 10, 13, 16, 19 or 364), two pairs of primers qHGFPF/TAGnR and TAGnF/RT8DN1 were used to amplify two DNA fragments from pCMIRT15-B. Then primer pair qHGFPF/RT8DN1 and a mixture of the two fragments as template were used to amplify a 1352 bp sequence. The amplicon was digested using NcoI and XbaI and inserted into pCMIRT15-B between restriction sites NcoI and XbaI. All syn-tasiRNA constructs were introduced into Agrobacterium tumefaciens strain GV3101. Primer sequences are listed in Table S1.

Plant transformation and screening of T1 seeds

The floral dip method (Clough and Bent, 1998) was used to introduce syn-tasiRNA constructs into A. thaliana (Col-0). In order to increase transformation frequency, relatively old plants (7–9 weeks old) were used, and all siliques and open flowers were trimmed off immediately before dipping. The seeds harvested from floral-dipped plants (T1 seeds) were screened on Murashige and Skoog (MS) agar medium (M9274; Sigma-Aldrich, supplemented with 40 mg L−1 hygromycin B (Invitrogen) and 50 mg L−1 cefotaxime (Plant Media, In order to compare seedlings transformed with the various syn-tasiRNA constructs under uniform conditions, T1 seeds were sown on the same plate. Plates were maintained at 4°C for 2 days, and then kept at room temperature for approximately 5 days to allow seeds to germinate. Hygromycin-resistant seedlings grew taller than the rest due to the relatively weak light intensity. Plates were then transferred to a growth chamber under a 16 h day/8 h night illumination regime with temperatures of 22°C (day) and 20°C (night). A layer of Whatman 3MM paper ( was placed on top of the plates to reduce incident light intensity. After 12 more days of culture, all hygromycin-resistant seedlings were transferred to fresh plates. Seedlings were arbitrarily grouped into white seedlings (completely bleached), yellow seedlings (either cotyledons or true leaves show signs of green), and green seedlings (most parts of the seedling are green). All seedlings that were transformed using the same construct and from the same selection plate were collected as one sample for RNA extraction. The total number of seedlings in each sample varied between 29 and 125.

RNA extraction and cDNA synthesis

Total RNA was extracted using plant RNA reagent (Invitrogen) and treated with RQ1 DNase (Promega, according to the manufacturers’ instructions. For cDNA synthesis, approximately 3 μg of total RNA was reverse-transcribed using Superscript III reverse transcriptase (Invitrogen) and a mixture of oligo(dT) and three stem-loop reverse transcription primers, which were designed to specifically recognize miR173 (5′-GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACGTGATT-3′) and syn-tasiRNAs PDS3-1 (5′-GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACAACGCT-3′) and PDS3-2 (5′-GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACGCGCAA-3′) (Chen et al., 2005). The reaction was performed in a PCR machine running the following program: 16°C for 5 min, 25°C for 5 min, 50°C for 30 min, 55°C for 30 min and 70°C for 15 min, then held at 4°C. The cDNA was treated with RNase H (Invitrogen) at 37°C for 20 min.

Quantitative RT-PCR analysis

First-strand cDNA (20 μl) was diluted tenfold, and 2 μl was used for quantitative PCR in a 15 μl reaction using FastStart Universal SYBR Green Master Mix (Roche, in an Applied Biosystems 7500 real-time PCR system ( The quantitative PCR primers used were 5′-AACGGCACTTCGCTTGCAGA-3′ and 5′-TATCCAGTGCAGGGTCCGAGGTATT-3′ (SLR) for miR173, 5′-ACCTGCACGTCTACTGGCATACAA-3′ and SLR for PDS3-1, 5′-CGGTGCGGGTATCCTAGAATTAAG-3′ and SLR for PDS3-2, 5′-ATCCTCCGGAAAGGCTTTGTATGC-3′ and 5′-TAAGCGTCTCCTTCGACAGTGCTT-3′ for PDS3 (At4g14210) mRNA, 5′-GTCTGCCACTAAACCAGCTGAAGA-3′ and 5′-ACACGCTGAACTTGTGGCCGTTTA-3′ for HTA6–GFP mRNA (note primers are located upstream or flanking the cleavage sites), and 5′-ACGTGAAAGCCAAGATCCAGGACA-3′ and 5′ CTTCAAGTTGCTTTCCGGCGAAGA-3′ for ubiquitin10 (UBQ10, At4g05320) mRNA. The Ct values were normalized against UBQ10 mRNA. The abundance of mRNAs or small RNAs was expressed as relative to controls, with control values set to 1. The error bars represent the standard deviation of technical repeats.

Small RNA Northern analysis

A previously published protocol (Lackey et al., 2010) was used for small RNA Northern analysis with modifications. In brief, a smaller amount of total RNA (2 μg) was analyzed than in the published protocol. A small RNA-specific probe and a control probe (U6) were labeled separately. Purified probes were added to the same hybridization reaction. After hybridization, membranes were washed more vigorously: twice with 1 × SSC + 0.5% SDS each for at least 30 min at 40°C, and twice with 0.2 × SSC + 0.1% SDS each for at least 30 min and at an elevated temperature (42°C). Membranes were exposed to a phosphor imaging plate (Fuji, for 2–8 days before collection of hybridization signals using a Typhoon scanner (GE Healthcare, The probes used were 5′-AACGCTTTGTATGCCAGTAGA-3′ for detecting PDS3-1, and 5′-TTCTCGATTTATGCGTGTCATCCTTGCGCAGGGGCCATGCTAATCTTCTCTGTATCGTTCCAATTTTATC-3′ for detecting U6. Signal intensities of hybridization bands were measured using ImageJ (Abramoff et al., 2004).

5′ RACE analysis

DNase I-treated total RNA (approximately 3 μg) was used in a 10 μl ligation reaction including 0.5 μg of synthetic RNA oligo (5′-GACACCUCAGGACGGACCGAAUUCGAAA-3′), 1 μl of RNaseOut RNase inhibitor (Invitrogen), 1 μl of T4 RNA ligase buffer and 1 μl of T4 RNA ligase (M105A, Promega). The reaction was incubated at 37°C for 1 h. RNA was extracted once using phenol/chloroform/isoamyl alcohol (25/24/1), and precipitated by adding a one-tenth volume of 3 m NaOAc (pH 5.3) and two volumes of ethanol. RNA was dissolved in water and reverse-transcribed into cDNA using oligo(dT) primers and Superscript III reverse transcriptase (Invitrogen). First-strand cDNA was treated with RNase H and diluted tenfold, and 2 μl were used for a 20 μl PCR reaction amplifying the 5′ ends. The primers used were 5′-GAATAGACACCTCAGGACGGACC-3′ (forward), 5′-ACACCTCAGGACGGACCGAAT-3′ (nested forward), 5′-GTAACGGTACCGAGCTCAAGCTTG-3′ (reverse) and 5′-CGGGATCCAAGCAAATATCATGCGATCA-3′ (nested reverse). Amplicons from second-round PCR were cloned into the pGEM-T vector (Promega) and sequenced using Sp6 primer 5′-GATTTAGGTGACACTATAG-3′. As a control, Actin2/8 (At3g18780 and At1g49240) cDNA was amplified using primers 5′-CCTATTGAGCATGGTGTTGTTAGCAAC-3′ and 5′-TGTGAGACACACCATCACCAGA-3′.


We thank Marisa Miller and other members of the Chen lab for valuable suggestions to improve the manuscript. We thank Chris Sullivan (Molecular Genetics & Microbiology, The University of Texas at Austin) for valuable suggestions on the experiments and Vikram Agarwal for inspiring discussions during the initial stage of this work. C.Z. thanks Jonathan Seely and Bert Brown for proofing the manuscript. The work is supported by grants from the National Science Foundation (MCB1110957) to Z.J.C..