22-nucleotide RNAs trigger secondary siRNAbiogenesis in plantsHo-Ming Chena, Li-Teh Chena, Kanu Patelb, Yi-Hang Lia, David C. Baulcombeb, and Shu-Hsing Wua,1

aInstitute of Plant and Microbial Biology, Academia Sinica, Taipei 11529, Taiwan; and bDepartment of Plant Science, University of Cambridge, Cambridge CB23EA, United Kingdom

Edited by Detlef Weigel, Max Planck Institute for Developmental Biology, Tübingen, Germany, and approved June 22, 2010 (received for review February11, 2010)

The effect of RNA silencing in plants can be amplified if theproduction of secondary small interfering RNAs (siRNAs) is triggeredby the interaction of microRNAs (miRNAs) or siRNAs with a longtarget RNA. miRNA and siRNA interactions are not all equivalent,however; most of them do not trigger secondary siRNA production.Here we use bioinformatics to show that the secondary siRNAtriggers are miRNAs and transacting siRNAs of 22 nt, rather thanthe more typical 21-nt length. Agrobacterium-mediated transientexpression in Nicotiana benthamiana confirms that the siRNA-initiating miRNAs, miR173 andmiR828, are effective as triggers onlyif expressed in a 22-nt form and, conversely, that increasing thelength of miR319 from 21 to 22 nt converts it to an siRNA trigger.Wealsopredictedandvalidated that the 22-ntmiR771 is a secondarysiRNA trigger. Our data demonstrate that the function of small RNAsis influenced by size, and that a length of 22 nt facilitates the trigger-ing of secondary siRNA production.

gene silencing | microRNA | transacting siRNA

Small silencing RNAs (sRNAs) in plants and animals, includingmicroRNAs (miRNAs) and small interfering RNAs (siRNAs),

play important roles in the development and the response topathogens and stresses. These RNAs are also valuable tools infunctional genomics and biotechnology. The sRNAs associate withARGONAUTE (AGO) and other proteins in silencing effectorcomplexes, and they bind to a target nucleic acid viaWatson–Crickbase pairing. Inmost instances, the silencing is a direct consequenceof this interaction, and the AGO effector mediates RNA-mediatedDNA or histone methylation, endonucleolytic RNA cleavage, ortranslational inhibition. In a few instances, an sRNA interactionalso triggers the production of secondary siRNAs. The targetedRNA is converted into double-stranded RNA (dsRNA) by RNA-DEPENDENT RNA POLYMERASEs (RDRs), which is thencleaved into the secondary siRNAs by DICER-LIKE (DCL)nucleases (1). Several proteins are known to be required for thisprocess, but until now, the reason why most sRNA interactions donot result in secondary siRNA production was unclear.The transacting siRNA (tasiRNA) pathway in plants involves

secondary siRNA production (2). Noncoding transcripts encodedby TAS1–4 genes serve as the precursors of tasiRNAs (3–5). AftermiRNA-directed cleavage, part of the remaining transcript isconverted into dsRNA by RDR6. DCL4 then cleaves the dsRNAand generates tasiRNAs in a 21-nt phase relative to positions 10and 11 of the miRNA that defines the site of targeted cleavage.TAS1 and TAS2 are targets of miR173, and their tasiRNAs in turncan target mRNAs for pentatricopeptide repeat (PPR) proteins.In one instance, a small sRNA cascade is initiated by miR173 (6,7), because a TAS2-derived tasiRNA can itself initiate secondarysiRNA production on several PPRmRNAs. The initiator of TAS3tasiRNA is miR390 (3, 8), and the TAS3 targets are AUXINRESPONSE FACTOR mRNAs that influence the change fromjuvenile phase to adult phase, leaf morphology, and lateral rootgrowth (9–12). miR828 initiates TAS4 tasiRNAs that target themRNAs of MYB transcription factors (5).A “two-hit” model to explain tasiRNA production is based on

the finding that two miRNA target sites are critical for TAS3

tasiRNA production between the two sites (8, 13). One of thesesites, where RNA cleavage occurs, could be targeted by any ofseveral different miRNAs; however, the second noncleavable siteis functional only if it is the target of miR390 (13). This model isspecific for TAS3, and the two-hit requirement is probably relatedto the specific interaction of miR390 with AGO7 (12). In contrast,TAS1 and TAS2 have only a single miR173 target site, and theAGO protein involved is AGO1 rather than AGO7 (3, 14). RNAsother than TAS precursors also support secondary siRNA pro-duction if they have an miR173 target site (14, 15). This findingimplies that the determinant of secondary siRNA production isthe sRNA initiator rather than the targetedRNA, and it prompteda bioinformatics survey of common features in the miRNAs andtasiRNAs that are triggers of secondary siRNA production. Theresults reported here reveal that secondary siRNA triggers arerepresented as 22-nt sRNAs rather than the 21-nt size that ischaracteristic of most Arabidopsis miRNAs. We also describeexperimental results confirming that a size of 22 nt is necessary foran miRNA to trigger secondary siRNA biogenesis on the 3′cleavage products of their targets.

ResultssiRNA Triggers Are Predominantly 22 nt in Size. Through a combi-nation of literature search and analysis of the Arabidopsis sRNAsequencing data with a phasing algorithm (6), we identified eightmiRNAs and one tasiRNA as triggers of secondary siRNAproduced from 3′ cleaved transcripts of their targets (Table S1).We also identified miR825* as the potential siRNA trigger atAt5g38850, which encodes a TIR-NBS-LRRprotein and producesphased siRNA (7).Using Arabidopsis sRNA sequencing datasets derived by 454 or

Illumina technology (5, 8, 16–18), we found that the putativemiRNA triggers of secondary siRNA production are representedas 22-nt species rather than the normal 21-nt form (Fig. 1A). TheMIR168,MIR173, andMIR393 loci also produce other size classesin the 19- to 24-nt range, but for all of these except MIR168, themost abundant form was 22 nt in size (Fig. 1A). Among 163Arabidopsis MIR loci, only 24 have a dominant 22-nt form (Fig. 1Band Table S2); of these, MIR173 and MIR168 have the highestexpression (Fig. 1A and Table S2). Most of the 22-nt dominantmiRNAs, including miR780, miR771, and miR848, have beenidentified recently and lack predicted or validated targets.Among the Arabidopsis tasiRNAs, tasiR2140 [also known as

TAS2 3′D6(-)] is the only one known to be associated with sec-ondary siRNA production (6–8). The predicted structure of this

Author contributions: H.-M.C. and S.-H.W. designed research; H.-M.C., L.-T.C., K.P., andY.-H.L. performed research; H.-M.C., D.C.B., and S.-H.W. analyzed data; and H.-M.C., D.C.B.,and S.-H.W. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

See Commentary on page 14945.1To whom correspondence should be addressed. E-mail: shuwu@gate.sinica.edu.tw.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1001738107/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1001738107 PNAS | August 24, 2010 | vol. 107 | no. 34 | 15269–15274

PLANTBIOLO

GY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Aug

ust 3

0, 2

020

TAS2-derived tasiR2140, based on the phase set by miR173, is21 nt with a 5′ A (3). However, in the AGO1-associated sRNAs(19), tasiR2140 reads were detected mostly in the 22-nt form, withan additional U at the 5′ end relative to the predicted 21-nttasiR2140. This 22-nt tasiR2140 is likely a functional form basedon a 5′ RACE assay of tasiR2140 cleavage products on the targetmRNAs (At1g63130 and At1g63080). The sRNA cleavage sitesare normally opposite positions 10 and 11 of the sRNA guidemolecule, and this position corresponds to the 22-nt form ratherthan the 21-nt form of tasiR2140 (4, 6). Two rice miRNAs,miR2118 and miR2775, that may direct the 21-nt or 24-nt phasedsiRNA production in developing inflorescences are also 22 nt insize (Table S1) (20). Thus, a size of 22 nt is a common featureshared by many diverse triggers of secondary siRNA production.The 22-nt putative triggers of secondary siRNA all have U in

the 5′ position (Fig. 1C). However, this is a common feature ofmiRNAs associated with their loading to AGO1 (19) and is notspecific to triggers of secondary siRNAs. Similarly, we found thata preference for A in position 10 of the miRNAs is also a commonfeature by analyzing 163 Arabidopsis miRNAs, but that thispreference is not specific to triggers of secondary siRNAs. Sevenof the 12 triggers of secondary siRNAs have C at the 3′ terminalnucleotide, but others have G or A in this position, indicating thatthis feature is not essential for initiation of secondary siRNAs.

Thus, we hypothesized that a size of 22 nt is the primary featureassociated with the initiation of secondary siRNA production.

A 22-nt miRNA Triggers Secondary siRNA Production. To test this22-nt hypothesis, we engineered miRNA precursors to change thesize of the processed miRNA. WithMIR173, this change involvedadding an additional nucleotide in miR173* to pair with a bulgedbase in the WT miR173 duplex. The WT MIR173 precursor isnormally cleaved by DCL1, which leaves 21 paired bases between5′ and 3′ cleavage sites to release the 22-nt miR173 (Fig. 2A).However, with the modified structure, the 21–paired-base rulewould cause DCL1 to release a 21-nt miR173; in Fig. 2A, thismodified construct is termed MIR17321, and the unmodified con-struct is termedMIR17322.To test the biological activity of the modifiedMIR173, we used

a transient assay system based on infiltration of Agrobacteriumcultures transformed with these constructs in Nicotiana ben-thamiana, in which miR173 is not produced. The MIR constructswere coexpressed with a sensor construct that would be tran-

100000

19 nt 20 nt 21 nt 22 nt 23 nt 24 nt

10

100

1000

10000 re

ads

miR16

8

miR17

3

m i R39

3

m i R44

7

m i R 4 7

2

miR77

3

miR82

5*

m i R82

8

m i R85

6

tasiR

2 140

miRNA/tasiRNA

1

100%19 nt

0%

10%

50%19 nt

20 nt

21 nt

22 nt

23 nt

24 nt

miR168 UCGCUUGGUGCAGGUCGGGAAC miR173 UUCGCUUGCAGAGAGAAAUCAC miR393 UCCAAAGGGAUCGCAUUGAUCC

0%

22class (24)

21class (116)

24 nt

miR447 UUGGGGACGAGAUGUUUUGUUG miR472 UUUUUCCUACUCCGCCCAUACC miR773 UUUGCUUCCAGCUUUUGUCUCC miR825* UCAAGCACCAGCUCGAAGAAGC miR828 UCUUGCUUAAAUGAGUAUUCCA miR856 UAAUCCUACCAAUAACUUCAGC tasiR2140 UAUAUCCCAUUUCUACCAUCUG

Arabidopsis

miR2118 UUCUCGAUGCCUCCCAUUCCUA miR2775 UUUGGUUUCCUCCAAUAUCUCA Rice

A

B

C

Fig. 1. siRNA triggers are predominantly 22 nt. (A) Size distribution of Ara-bidopsis siRNA triggers. The abundance of siRNA triggers in size classes from19 to 24 nt is plotted according to the pooled data of several available sRNAsequencing datasets (5, 8, 16–18). (B) A heat map showing clusters of Arabi-dopsismiRNAs with sizes enriched mainly as 21 or 22 nt. Each vertical columnrepresents one miRNA species. A gradient of yellow to red represents thefrequency (%) of each size class for a given miRNA. The map is organizedaccording to the relative abundance of the sizes (19–24nt) for eachmiRNA. (C)Alignment of siRNA triggers. Identical nucleotides and conserved nucleotidesamong siRNA triggers are highlighted in yellow and blue, respectively.

A 5’MIR17322

3’

miR17322

miR17322*

5’ MIR17321

AcuUUCGCUUGCAGAGAGAA UCAC|||||||||||| |||||| ||||GAAAGCGAAUGUGUCUCUU_AGug

miR173target

B

3’

miR17321

miR17321*

cuUUCGCUUGCAGAGAGAAAUCAc|||||||||||| |||||||||||GAAAGCGAAUGUGUCUCUUUAgug

35Starget FP

FP-siR probe

/3’ CACUAAAGAGAGACGUUCGCUU 5’

173FP 5’ GUGAUUUCUCUCUGCAAGCGAA 3’||||||||||||||||||||||

7/8

miR17322

3’ ACUAAAGAGAGACGUUCGCUU 5’173FP 5’ GUGAUUUCUCUCUGCAAGCGAA 3’|||||||||||||||||||||

8/9

miR17321

C

FP-siRNAs

miR17322

20

22

20

5S rRNA/tRNA

173FP

25S rRNA

Fig. 2. A size of 22 nt is required for miR173 to trigger secondary siRNA. (A)Predicted foldback structures of MIR17322 and MIR17321 for expressing 22-and 21-nt miR173, respectively. Expected duplexes of miR173 and miR173*are highlighted. (B) Construct of 35S:173FP and cleavage sites identified by5′ RACE assay. The position of the cleavage site is indicated by an arrow withthe proportion of sequenced clones. FP-siR probe (horizontal line) indicatesa DNA oligonucleotide probe complementary to the 42-nt region down-stream of the predicted cleavage site directed by miR173 for the detection ofFP-derived siRNA shown below. (C) Northern blot analysis of miR173, FP-derived siRNA, and 173FP transcripts.

15270 | www.pnas.org/cgi/doi/10.1073/pnas.1001738107 Chen et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 3

0, 2

020

scribed from the 35S promoter to produce a transcript witha target site of miR173 located at the 5′ end of a truncated GFPcoding sequence (FP) (35S:173FP; Fig. 2B).A 5′ RACE analysis of the transient assay samples confirmed

that both miR17322 and miR17321 target the cleavage of 173FPmRNA at the same site (Fig. 2B) and that, correspondingly, ac-cumulation of the full-length 173FPmRNA was lower than when35S:173FP was expressed alone (Fig. 2C). Northern blot analysisalso confirmed that the two MIR173 constructs each producemiRNAs of the expected size; 35S:MIR17322 produces an RNA ofthe same size as the Arabidopsis endogenous miR173, whereas35S:MIR17321 produces a smaller species (Fig. 2C). The sequenceauthenticity of 21-nt miR173 was confirmed by deep-sequencinganalysis of an sRNA population inN. benthamiana infiltrated with35S:MIR17321 and 35S:173FP. Among 19,077 reads of 19- to 24-ntmiR173s, 96.9% of the sRNAs are 21 nt with a sequence identicalto that shown in the lower panel of Fig. 2A.Northern blot analysis revealed thatFP-derived siRNAcould be

detected only when 173FP was coinfiltrated with the MIR17322construct (Fig. 2C). This indicates that the targeted cleavage bybothmiR17322 andmiR17321 is not sufficient for the production ofsecondary siRNAs. This finding is consistent with our hypothesisthat only the 22-nt miR173 can both direct the cleavage of thesynthetic sensor RNA and trigger the production of secondarysiRNA(s) using the 3′ cleavage fragment of the sensorRNA.Thesefindings also indicate that themiR173precursor sequencedoes notgive miR17321 the ability to act as a trigger of secondary siRNA.We also tested the 22-nt hypothesis with miR828 that targets

the TAS4 RNA and initiates tasiRNA production (5). However,

both miR828 and miR828* are 22 nt, and the modifications of theMIR828 sequence necessary to produce 21-nt miRNA are notobvious. Our alternative approach was to modify the MIR319afoldback (21) so that it would release a 21-nt miR828 without thenucleotide 22. The constructs in this assay, 35S:MIR82822 and35S:MIR82821 (Fig. S1A), were designed to produce 22-nt and21-nt miR828 in the transient expression assay, and the sensorconstruct was 35S:828FP (Fig. S1B). Transient expression of 35S:MIR82822 and 35S:MIR82821 led to production of the predicted22- and 21-nt miR828s (Fig. S1C). When these constructs werecoexpressed with 35S:828FP, the cleavage of the 35S:828FPtranscript was at the same site (Fig. S1B), but only miR82822initiated miRNA-dependent production of FP-siRNA (Fig. S1C).Thus, these results are consistent with the proposed role of 22-ntmiRNA in secondary siRNA production.In a third assay, we aimed to convert a 21-nt miRNA into a 22-nt

species with the expectation that this change would be associatedwith an increased potential to trigger secondary siRNA pro-duction. The miR319 is normally expressed as a 21-nt form, and itlacks the ability to trigger secondary siRNA production. By cre-ating a bulge in the miRNA/miRNA* duplex in the foldback se-quence ofMIR319, this construct should generate a 22-nt miR319(Fig. 3A). A 5′ RACE analysis confirmed that 319FP transcripts,when coexpressed with 35S:MIR31921 or 35S:MIR31922, arecleaved at the predicted miR319 target site (Fig. 3B). Northernblot analysis demonstrated that these two 35S:MIR319 constructsproduce different-sizedmiRNAs, as predicted (Fig. 3C). However,the electrophoretic migration of these RNAs is anomalous, as hasbeen reported previously (22). We used sequence analysis to

A 5’

3’

miR31921*

MIR31921

MIR319gagAGAGCUUCCUUGAGUCCAUUC

B

35S

miR319target FP

FP siR probe

19 nt

20 nt

21 nt

22 nt

23 nt5’

MIR31922

miR31921

MIR31921

cUCCCUCGAGGGAAGUCAGGUUag||| |||||||||| |||||| || FP-siR probe

319FP 5’ GAGGGAGCUCCCUUCAGUCCAA 3’|||||||||||||||||||||

9/13

0% 20% 40% 60% 80% 100%

23 nt

24 nt3’

MIR31922 319FP 5’ GAGGGAGCUCCCUUCAGUCCAA 3’

3’ UCCCUCGAGGGAAGUCAGGUU 5’

3’ CUCCCUCGAGGGAAGUCAGGUU 5’

miR31921

||||||||||||||||||||||

7/9

miR31922miR31922

gaGAGAGCUCCUUG_AGUCCAUUCCUCCCUCGAGGGAA UCAGGUUag||| ||||||||| |||||| ||

G

miR31922*

C

miR31922

4000

6000

4000

6000

DReg. 1-2

Reg. 3-21

MIR31921

P10M

)

Reg. 1-2

Reg. 3-21

MIR31922

P10M

)

22

20

FP-siRNAs

miR319

22

20

NbmiR159

2000

0

2000

2000

0

2000

S

AS

NA

abundance (

TP

AS

S

NA

abundance (

TP

5S rRNA/tRNA

319FP

25S RNA

0 100 200 300 400

-4000

-2000

0 100 200 300 400

-4000

-2000

21-n

t siR

N

21-n

t siR

N

r miR319-directed

cleavage siteposition

miR319-directed

cleavage siteposition

Fig. 3. A change in size from 21 to 22 nt is sufficient to convert miR319 to an siRNA trigger. (A) Predicted foldback structures of MIR31921 and MIR31922 forexpressing 21- and 22-nt miR319, respectively. Expected duplexes of miR319 and miR319* are highlighted. Sizes of miR319s produced from MIR31921 orMIR31922 were obtained by small RNA sequencing and shown as percentage (%) of each size class. (B) Construct of 35S:319FP and cleavage sites identified by5′ RACE assay. The position of the cleavage site is indicated by an arrow with the proportion of sequenced clones. FP-siR probe (horizontal line) indicatesa DNA oligonucleotide probe complementary to the 42-nt region downstream of the predicted cleavage site directed by miR319 for the detection ofFP-derived siRNA shown below. (C) Northern blot analysis of miR319, FP-derived siRNA, and 319FP transcripts. N. benthamiana miR159 (NbmiR159) has a highsequence similarity with miR319 and likely was cross-hybridized with the miR319 probe. (D) Normalized abundance (TP10M) of FP-derived 21-nt small RNAspecies was plotted for both sense (S) and antisense (AS) strands of 319FP in N. benthamiana coinfiltrated with 35S:319FP/35S:MIR31921 or 35S:319FP/35S:MIR31922. 21-nt sRNAs in registers (reg.) 1 and 2 corresponding to the miR319-directed cleavage site are highlighted in red.

Chen et al. PNAS | August 24, 2010 | vol. 107 | no. 34 | 15271

PLANTBIOLO

GY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Aug

ust 3

0, 2

020

confirm the size of the miR319 species; 87.3% of the sequencereads in the 35S:MIR31921 samples were 21 nt, and 61.4% of 35S:MIR31922 were 22 nt (Fig. 3A). Although the 319FP transcript wascleaved in the presence of both 21- and 22-nt miR319, significantFP-siRNA could be detected only when 319FP transcripts werecoexpressed with a 22-nt miR319 (Fig. 3C). Deep-sequencinganalysis also revealed an obvious increase in FP-siRNAs phased inthe register set by miR31922 cleavage site (register 1) or in the 1-ntforward register (register 2) (Fig. 3D). For the first 231 bp down-stream of the cleavage site, the normalized abundance [transcriptsper 10 million (TP10M)] for phased FP-siRNAs is 9,120 whencoexpressed with miR31922 and 1,506 when coexpressed withmiR31921. Phase analysis of FP-siRNAs using a previously de-scribed algorithm (6) showed a significant P value of 0.00074 whencoexpressed with a 22-nt miR319, compared with the P value of0.00655 when coexpressed with a 21-nt miR319. FP-siRNA pro-duction is also triggered by a 22-nt miR319 expressed from theMIR173 and MIR771 foldback constructs 35S:MIR31922(173) and35S:MIR31922(771) (Figs. S2 and S3). Thus, by adding one extranucleotide to the 3′ terminus, the 22-ntmiR319 acquires the abilityto trigger secondary siRNA production. Therefore, the 22-nt sig-nature alone determines whether anmiRNA can act as a trigger ofsecondary siRNA.

miR771 Is a New Trigger of Secondary siRNA Production. miR771 isone of the 24ArabidopsismiRNAs that are present predominantlyas 22-nt species, and we predicted that it would trigger siRNA

production. As shown in Fig. 4, this prediction is confirmed by thefinding that the 22-nt miR771 can induce site-specific cleavage ofthe 771FP transcript and FP-siRNAs are produced. Thus, the 22-nt miR771 is a new trigger of siRNA production.

DiscussionA Size of 22 nt Is the Key Determinant of siRNA Triggers. Based ona bioinformatics analysis of sRNA databases, we hypothesized thatmiRNA or siRNA triggers of secondary siRNA production mighthave a lengthof 22 nt rather than themore common21nt. The 22-ntmiRNAs or siRNAs would direct the cleavage of a target RNA, andthe secondary siRNAs would be produced in an RDR-dependentmanner from the 3′ cleavage products. We confirmed this hypoth-esis experimentallybydemonstrating that the22-nt form,butnot the21-nt form, of miR173 and miR828 can initiate siRNA productionon the 3′ cleavage product of a sensor RNA, although both candirect cleavage (Fig. 2 and Fig. S1). We further confirmed thisfinding by demonstrating that an additional nucleotide at the 3′ endconvertsmiR319 into an siRNA trigger (Fig. 3 and Figs. S2 and S3).We identified eight Arabidopsis miRNA families with 22-nt

forms that are known triggers of siRNA production. An additional13 miRNA families with >50% of the sequence reads as 22-ntspecies are candidate triggers of secondary siRNA. Our findingthat one of these, the 22-nt miR771, is capable of initiatingsecondary siRNAs confirms that these additional miRNAs arepotential siRNA triggers (Fig. 4). Presumably, the secondarysiRNAs from these additional 22-nt miRNAs were not detectedpreviously because the target RNA is not expressed in the tissuesused to generate the sRNA sequence libraries.

Biogenesis and Action of 22-nt sRNAs. The biogenesis of 22-ntcompared with 21-nt sRNAs might be influenced by the DCLprotein involved or by the structure of the sRNA precursor.Examples in which the RNA structure is involved include miR771and the 22-nt miRNAs listed in Table S1 (except miR828; seebelow). In all of these examples, the miRNA duplex containsa bulge (miRBase: http://www.mirbase.org/), so that DCL1 gen-erates a 22-nt miRNA and a 21-nt miRNA*.In other examples, DCL2 generates 22-nt siRNAs from per-

fectly duplexed precursors. These 22-nt siRNAs might be triggersof secondary siRNAs (23). miR828 is derived from perfectlyduplexed 22-nt precursors and might be produced by DCL2,although this possibility has not been tested experimentally.There is good evidence that DCL2 and secondary siRNAs areinvolved in transgene and viral RNA silencing, however (23–26).We consider a causal connection likely and suggest that the 22-ntsmall RNAs produced by DCL2 in these systems are the triggersof secondary siRNA production.The likely involvement of DCL2 in the initiation of secondary

siRNA biogenesis could explain why dcl1 dcl4 and dcl1 dcl3 dcl4mutants grow poorly in greenhouse conditions, whereas dcl1 dcl2dcl4 and dcl1 dcl2 dcl3 dcl4 mutants are viable (27). In dcl1 dcl4and dcl1 dcl3 dcl4 genotypes, the DCL2 might generate siRNAtriggers using dsRNAs that are normally substrates of DCL1,DCL3, or DCL4. This overproduction of 22-nt sRNAs could thentrigger massive overproduction of secondary siRNAs, leading touncontrolled posttranscriptional silencing in the plant and sub-sequent impaired plant growth.The 22-nt forms of miR173, miR828, miR319, miR771, and

tasiR2140 all trigger secondary siRNA production after targetedRNA cleavage. These RNAs all have a 5′U and, like miR173, arelikely loaded into a complex with the AGO1 that also is associatedwith the cleavage-only interactions (14, 19). To explain the differ-ent effects of 21- and 22-nt miRNAs bound to AGO1, we proposethat the AGO1 protein can adopt different conformations. Whenbound to the 21-nt miRNA, the protein has a structure that pro-motes cleavage of the targeted RNA, whereas when the 22-ntmiRNA is present, the AGO1 structure promotes target cleavage

A

5’ MIR7715

3’

miR771

miR771*

miR771B35S

miR771target FP

FP-siR probe

22

20miR771

C

20

FP-siRNAs

5S rRNA/tRNA

22

20

771FP

25S rRNA

Fig. 4. miR771 is a new siRNA trigger. (A) Predicted foldback ofMIR771. Theexpected duplexes of miR771 and miR771* are highlighted. (B) Construct of35S:771FP and cleavage sites identified by 5′ RACE assay. The position of thecleavage site is indicated by an arrow with the proportion of sequencedclones. FP-siR probe (horizontal line) indicates a DNA oligonucleotide probecomplementary to the 42-nt region downstream of the predicted cleavagesite directed by miR771 for the detection of FP-derived siRNAs shown below.(C) Northern blot analysis of miR771, FP-derived siRNA, and 771FP transcripts.

15272 | www.pnas.org/cgi/doi/10.1073/pnas.1001738107 Chen et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 3

0, 2

020

and recruitment of proteins involved in siRNA biogenesis, in-cluding SUPPRESSOR OF GENE SILENCING3 (SGS3) andRDR6. This conformational change is likely to occur because the 5′and 3′ ends of AGO-bound sRNAs are tethered within definedstructural domains of the AGO proteins (28). Thus, an AGO1bound to 22-nt miRNA is likely to be distorted relative to the formwith an associated 21-nt miRNA.The miR2775 is likely to trigger 24-nt rather than 21-nt phased

siRNA production in developing inflorescences of rice (20), but itnevertheless conforms to the 22-nt rule and has a 5′U(Fig. 1C). Anexception to the 22-nt rule is miR390, a 21-nt miRNA, which ini-tiates TAS3 tasiRNA production. Unlike the 22-nt miRNAs thathave single site targets, miR390 is effective only when there are twomiRNA target sites in TAS3 RNA (8, 13). The TAS3 cleavageproduct on the 5′ side of the miR390 target site is the secondarysiRNA precursor, whereas with TAS1 and TAS2, the siRNAs areproduced from the 3′ side of the cleavage site directed by thetrigger miRNA. AGO7, rather than AGO1, binds to miR390, andwe infer that the variant mechanisms of TAS1 and TAS2 as op-posed to TAS3 secondary siRNA biogenesis are somehow relatedto the different properties of the AGO1 and AGO7 proteins.Production of 22-nt miRNAs does not always trigger secondary

siRNAs; for example, there are both 21-nt and 22-nt species ofmiR165, miR166, and miR167, but phased secondary siRNAsfrom their target mRNAs have not been detected. The 22-ntforms of these miRNAs are associated with AGO1 (19) and aremore abundant than most known siRNA triggers; however, theproportion of 22-nt species is relatively lower than that of 21-ntspecies (Table S2). This suggests that the ratio of 22-nt to 21-ntmiRNAs, rather than the absolute abundance of 22-nt species ofthese miRNAs, influences the triggering of secondary siRNAs.The cell- or tissue-specific triggering of secondary siRNA pro-

duction also may be regulated by the discrete expression of 21- or22-nt miRNA species in different cells or tissues. For example,MIR168a generates predominantly a 21-nt mature miRNA and hasbroad expression in most tissue types (29). In contrast,MIR168b isspecifically expressed at the shoot apical meristem and producessimilar levels of 21- and 22-nt species (29). The level of 22-nt or theratio of 22- to 21-nt miR168 may reach the threshold in meristemiccells for miR168 to function as an siRNA trigger.

Future Perspectives. We previously determined that silencing cas-cades can be established in plants when miRNAs and siRNAstrigger secondary siRNA production on a long RNA target (6).However, these cascades are of limited extent, and propagation ofthe cascade is likely constrained. Our present analysis has shownthat only a few miRNAs and siRNAs can trigger secondarysiRNAs, and that these must be 22 nt long and with a 5′U so thatthey associate with AGO1, or otherwise, as in the exceptionalexample of miR390, they must associate with AGO7.Further research is needed to establish how a subtle difference

between a 21-nt miRNA and a 22-nt miRNA can affect thestructural configuration of AGO1 and the recruitment of factorsfor the production of secondary siRNAs. The biological role of thesiRNA cascades is also of interest. In terms of growth and de-velopment, some early data suggest that the production of sec-ondary siRNAs might be associated with signaling of positionalinformation (30). These cascades also are likely involved in anti-viral defense, given that the 22-nt viral siRNAs are abundant inplants infected with some viruses (31, 32). In these virus-infectedplants, the amplification associated with secondary siRNA pro-duction is likely important for the effectiveness of RNA silencingfor protection against replicating viruses. This amplification con-

tains a potential penalty, however, because misregulation of en-dogenous gene silencing cascades could have a damaging effect onthe gene expression profile in plants. Presumably, plants havemechanisms to ensure that the cascades are effective only at theappropriate time; however, the potential for amplification couldaid in the development of efficient gene silencing technologies forvirus resistance and manipulation of gene expression in bio-technology and for research applications.

Materials and MethodsIdentification of siRNA Triggers. Loci producing phased siRNAs were identifiedthrough a literature search or a phasing analysis of Arabidopsis sRNA se-quencing datasets using an algorithm described previously (6). Only 21-ntsRNA sequences were subjected to the phasing analyses, and a cutoff of P ≤0.001 was applied. Phased siRNA-producing loci not reported to be targetsof known miRNAs or tasiRNAs were searched against known miRNAs andmiRNAs*. sRNAs that can direct the cleavage upstream of phased siRNAclusters and set the phase consistent with or 1 nt forward from that ofphased siRNAs were collected for further analyses.

Construction of MIRs and Sensors. For transient expression of miRNAs,sequences containing foldback structures of MIR173, MIR319, MIR828, andMIR771were amplified from Arabidopsis genomic DNA and inserted into thepBIN61 vector containing the 35S promoter.MIR foldbacks were modified bysite-directed mutagenesis to express 21-nt miR173, 21-nt miR828, and 22-ntmiR319. For miRNA targets, a 22-nt sequence complementary to the miRNAsequencewas fused to a 5′ truncatedGFP sequence (FP) and used as the sensorconstruct. The sequences of MIR foldbacks and sensors are listed in Table S3.

RNA Structure Prediction. Structures of MIR foldbacks were predicted bymfold v3.2 (33) (http://mfold.bioinfo.rpi.edu/cgi-bin/rna-form1.cgi).

Nicotiana benthamiana Transient Expression. The procedures for transientexpression assay have been described previously (34). Leaves of 3- to 4-wk-oldN. benthamiana plants were infiltrated with Agrobacterium strain C58C1carrying MIR and/or sensor constructs. The input amount of each constructwas adjusted according to its expression level. The total input of Agro-bacterium was brought to OD600 = 1.0 with use of the empty vector. Infil-trated tissues were collected after 2–3 d for total RNA extraction.

Northern Blot Analysis. Total RNA of infiltrated tissues was extracted usingTrizol reagent (Invitrogen). For the sRNANorthernblot analysis, 3–5 μgof totalRNA was separated by 15% denaturing polyacrylamide TBE-urea gels (Invi-trogen). DNA oligonucleotides complementary to the miRNAs or the 42-ntregion immediately downstream of the predicted cleavage site directed bymiRNAs were used as probes to detect the expression of miRNAs or siRNAsfrom the sensor. For the Northern blot analysis of sensors, 2 μg of total RNAwas separated by 1% agarose gel. A 23-nt DNA oligonucleotide comple-mentary to the truncatedGFPwas used as the probe. The sequences of probesused in this study are listed in Table S3. Probe labeling, hybridization,washing, and signal detection were performed as described previously (6).

sRNA Sequencing. Total RNA was isolated using the mirVana miRNA Iso-lation Kit (Ambion) to generate sRNA libraries for sequencing by theIllumina platform.

Validation of Sensor Cleavage. Modified 5′ RACE with the GeneRacer Kit(Invitrogen) was adopted to validate the cleavage sites on the sensors. Theprimers used for 5′ RACE are in Table S3.

ACKNOWLEDGMENTS. We thank the members of the S.-H.W. laboratory andKrys Kelly for helpful discussions. This work was supported by research grantsfrom Academia Sinica (Foresight Project L20-2) and the National ScienceCouncil (NSC 97-2311-B-001-003-MY3) (to S.-H.W.). H.-M.C. is the recipient ofpostdoctoral fellowships from the National Science Council (098-2811-B-001-030) andAcademiaSinica. TheworkatCambridgewas supportedby theGatsbyCharitable Foundation and European Research Council Advanced InvestigatorGrant 233325 REVOLUTION. D.C.B. is a Royal Society Research Professor.

1. Voinnet O (2008) Use, tolerance and avoidance of amplified RNA silencing by plants.

Trends Plant Sci 13:317–328.2. Vaucheret H (2005) MicroRNA-dependent trans-acting siRNA production. Sci STKE

2005:pe43.

3. Allen E, Xie Z, Gustafson AM, Carrington JC (2005) MicroRNA-directed phasing during

trans-acting siRNA biogenesis in plants. Cell 121:207–221.4. Yoshikawa M, Peragine A, Park MY, Poethig RS (2005) A pathway for the biogenesis

of trans-acting siRNAs in Arabidopsis. Genes Dev 19:2164–2175.

Chen et al. PNAS | August 24, 2010 | vol. 107 | no. 34 | 15273

PLANTBIOLO

GY

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Aug

ust 3

0, 2

020

5. Rajagopalan R, Vaucheret H, Trejo J, Bartel DP (2006) A diverse and evolutionarilyfluid set of microRNAs in Arabidopsis thaliana. Genes Dev 20:3407–3425.

6. Chen HM, Li YH, Wu SH (2007) Bioinformatic prediction and experimental validationof a microRNA-directed tandem trans-acting siRNA cascade in Arabidopsis. Proc NatlAcad Sci USA 104:3318–3323.

7. Howell MD, et al. (2007) Genome-wide analysis of the RNA-DEPENDENT RNAPOLYMERASE6/DICER-LIKE4 pathway in Arabidopsis reveals dependency on miRNA-and tasiRNA-directed targeting. Plant Cell 19:926–942.

8. Axtell MJ, Jan C, Rajagopalan R, Bartel DP (2006) A two-hit trigger for siRNAbiogenesis in plants. Cell 127:565–577.

9. Adenot X, et al. (2006) DRB4-dependent TAS3 trans-acting siRNAs control leafmorphology through AGO7. Curr Biol 16:927–932.

10. Fahlgren N, et al. (2006) Regulation of AUXIN RESPONSE FACTOR3 by TAS3 ta-siRNAaffects developmental timing and patterning in Arabidopsis. Curr Biol 16:939–944.

11. Garcia D, Collier SA, Byrne ME, Martienssen RA (2006) Specification of leaf polarity inArabidopsis via the trans-acting siRNA pathway. Curr Biol 16:933–938.

12. Marin E, et al. (2010) miR390, Arabidopsis TAS3 tasiRNAs, and their AUXIN RESPONSEFACTOR targets define an autoregulatory network quantitatively regulating lateralroot growth. Plant Cell 22:1104–1117.

13. Montgomery TA, et al. (2008) Specificity of ARGONAUTE7-miR390 interaction anddual functionality in TAS3 trans-acting siRNA formation. Cell 133:128–141.

14. Montgomery TA, et al. (2008) AGO1-miR173 complex initiates phased siRNAformation in plants. Proc Natl Acad Sci USA 105:20055–20062.

15. Felippes FF, Weigel D (2009) Triggering the formation of tasiRNAs in Arabidopsisthaliana: The role of microRNA miR173. EMBO Rep 10:264–270.

16. Lister R, et al. (2008) Highly integrated single-base resolution maps of the epigenomein Arabidopsis. Cell 133:523–536.

17. Gregory BD, et al. (2008) A link between RNA metabolism and silencing affectingArabidopsis development. Dev Cell 14:854–866.

18. German MA, et al. (2008) Global identification of microRNA-target RNA pairs byparallel analysis of RNA ends. Nat Biotechnol 26:941–946.

19. Mi S, et al. (2008) Sorting of small RNAs into Arabidopsis argonaute complexes isdirected by the 5′ terminal nucleotide. Cell 133:116–127.

20. Johnson C, et al. (2009) Clusters and superclusters of phased small RNAs in thedeveloping inflorescence of rice. Genome Res 19:1429–1440.

21. Schwab R, Ossowski S, Riester M, Warthmann N, Weigel D (2006) Highly specific genesilencing by artificial microRNAs in Arabidopsis. Plant Cell 18:1121–1133.

22. Palatnik JF, et al. (2007) Sequence and expression differences underlie functionalspecialization of Arabidopsis microRNAs miR159 and miR319. Dev Cell 13:115–125.

23. Mlotshwa S, et al. (2008) DICER-LIKE2 plays a primary role in transitive silencing oftransgenes in Arabidopsis. PLoS One 3:e1755.

24. Deleris A, et al. (2006) Hierarchical action and inhibition of plant Dicer-like proteins inantiviral defense. Science 313:68–71.

25. Moissiard G, Parizotto EA, Himber C, Voinnet O (2007) Transitivity in Arabidopsis canbe primed, requires the redundant action of the antiviral Dicer-like 4 and Dicer-like 2,and is compromised by viral-encoded suppressor proteins. RNA 13:1268–1278.

26. Wang XB, et al. (2010) RNAi-mediated viral immunity requires amplification of virus-derived siRNAs in Arabidopsis thaliana. Proc Natl Acad Sci USA 107:484–489.

27. Bouché N, Lauressergues D, Gasciolli V, Vaucheret H (2006) An antagonistic functionfor Arabidopsis DCL2 in development and a new function for DCL4 in generating viralsiRNAs. EMBO J 25:3347–3356.

28. Parker JS (2010) How to slice: Snapshots of Argonaute in action. Silence 1:3.29. Vaucheret H (2009) AGO1 homeostasis involves differential production of 21-nt and

22-nt miR168 species by MIR168a and MIR168b. PLoS ONE 4:e6442.30. Chitwood DH, et al. (2009) Pattern formation via small RNA mobility. Genes Dev 23:

549–554.31. Akbergenov R, et al. (2006) Molecular characterization of geminivirus-derived small

RNAs in different plant species. Nucleic Acids Res 34:462–471.32. Xie Z, et al. (2004) Genetic and functional diversification of small RNA pathways in

plants. PLoS Biol 2:e104.33. Zuker M (2003) Mfold web server for nucleic acid folding and hybridization

prediction. Nucleic Acids Res 31:3406–3415.34. Voinnet O, Rivas S, Mestre P, Baulcombe D (2003) An enhanced transient expression

system in plants based on suppression of gene silencing by the p19 protein of tomatobushy stunt virus. Plant J 33:949–956.

15274 | www.pnas.org/cgi/doi/10.1073/pnas.1001738107 Chen et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 3

0, 2

020