a hairpin rna construct residing in an intron efficiently triggered rna-directed dna methylation in...

12
A hairpin RNA construct residing in an intron efficiently triggered RNA-directed DNA methylation in tobacco Athanasios Dalakouras 1,2 , Mirko Moser 1 , Miche ` le Zwiebel 1 , Gabi Krczal 1 , Ru ¨ diger Hell 2 and Michael Wassenegger 1,2,* 1 RLP AgroScience GmbH, AlPlanta-Institute for Plant Research, 67435 Neustadt, Germany, and 2 Heidelberg Institute for Plant Sciences, University of Heidelberg, 69120 Heidelberg, Germany Received 23 July 2009; revised 10 August 2009; accepted 12 August 2009; published online 14 September 2009. * For correspondence (fax +49 6321 671 1313; e-mail [email protected]). SUMMARY So far, conventional hairpin RNA (hpRNA) constructs consisting of an inverted repeat (IR) of target promoters directly introduced into an expression cassette have been used to mediate de novo DNA methylation. Transcripts of such constructs resemble mRNA molecules, and are likely to be exported to the cytoplasm. The presence of hpRNAs in the cytoplasm and the nucleus may account for the simultaneous activation of post- transcriptional gene silencing (PTGS) and RNA-directed DNA methylation (RdDM). We hypothesized that by retaining hpRNAs in the nucleus, efficient induction of only RdDM may be achieved. Thus, we introduced into tobacco a transgene containing an intron into which an IR of a target promoter was inserted. The intronic hpRNA initiated highly specific cis- and trans-methylation, but did not induce PTGS. No spreading of methylation into sequences flanking the region of homology between the hpRNA and the target DNA was detectable. The efficient methylation-directing activity of the intronic hpRNA may indicate a previously unrecognized role of introns, potentially regulating gene expression at the transcriptional level. Keywords: hairpin RNA, intron, RNA-directed DNA methylation (RdDM), RNA interference (RNAi), small RNAs. INTRODUCTION In plants, RNA-directed DNA methylation (RdDM) is a key process of nuclear RNA interference (RNAi) that is primarily associated with epigenetic gene silencing (Wassenegger, 2005; Huettel et al., 2007; Matzke et al., 2007, 2009; Wierzb- icki et al., 2008). RdDM is induced by double-stranded RNA (dsRNA) and, in concert with numerous proteins, leads to de novo cytosine methylation at symmetric CpG/CpHpG and asymmetric CpHpH sites (where H = A, T or G) (Matzke et al., 2009). RNA polymerase V (Pol V) is a central component of the de novo methylation complex. The SWI/SNF-like chromatin remodelling protein, DRD1, and the hinge domain-contain- ing protein, DSM3, appear to recruit Pol V to chromatin, although DRD1 does not physically interact with Pol V (Kanno et al., 2004, 2008; Wierzbicki et al., 2009). As Pol V does not utilize RNA templates, and is not specifically targeted to methylated DNA, it was suggested that the initiation of Pol V transcription occurs throughout the entire genome. Binding of small complementary RNAs to Pol V-produced transcripts would then lead to the sequence- specific directing of the de novo methyltransferase(s) to the target DNA (Pontier et al., 2005; Wierzbicki et al., 2008; ). Indeed, in plant nuclei, 24-nucleotide (24-nt) short interfering RNAs (siRNAs) are processed from dsRNA by Dicer-like 3 (DCL3), and are predominantly loaded onto Argonaute 4 (AGO4) (Zilberman et al., 2003; Chan et al., 2004). The 24-nt siRNAs are generally presumed to guide the RdDM machin- ery (Wierzbicki et al., 2009). In concert with the KOW domain-containing transcription factor 1 (KTF1), AGO4 could either directly interact with Pol V or would associate with Pol V via siRNA-mediated Pol V transcript binding (El-Shami et al., 2007; He et al., 2009). The hierarchical order of processes involved in the self- reinforcing feedback mechanism is unclear. An SNF2 domain-containing protein (CLSY1), Pol IV, Pol V, RNA- directed RNA polymerase 2 (RDR2), DCL3 and AGO4 were found, although to variable extents, to be connected with the accumulation of 24-nt siRNAs (Chan et al., 2004; Smith et al., 2007; Mosher et al., 2008). The latter three proteins and the NRPE1 subunit of Pol V were detectable in the nucleolus- associated Cajal bodies, indicating that they are siRNA processing centres (Li et al., 2006; Pontes et al., 2006). Using RNA templates, probably generated by Pol IV, RDR2 is producing dsRNA that is processed into secondary 24-nt siRNAs by DCL3. (Kanno et al., 2005; Zhang et al., 2007; Mosher et al., 2008). Yet, the substrate for Pol IV needs to be 840 ª 2009 The Authors Journal compilation ª 2009 Blackwell Publishing Ltd The Plant Journal (2009) 60, 840–851 doi: 10.1111/j.1365-313X.2009.04003.x

Upload: independent

Post on 02-Mar-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

A hairpin RNA construct residing in an intron efficientlytriggered RNA-directed DNA methylation in tobacco

Athanasios Dalakouras1,2, Mirko Moser1, Michele Zwiebel1, Gabi Krczal1, Rudiger Hell2 and Michael Wassenegger1,2,*

1RLP AgroScience GmbH, AlPlanta-Institute for Plant Research, 67435 Neustadt, Germany, and2Heidelberg Institute for Plant Sciences, University of Heidelberg, 69120 Heidelberg, Germany

Received 23 July 2009; revised 10 August 2009; accepted 12 August 2009; published online 14 September 2009.*For correspondence (fax +49 6321 671 1313; e-mail [email protected]).

SUMMARY

So far, conventional hairpin RNA (hpRNA) constructs consisting of an inverted repeat (IR) of target promoters

directly introduced into an expression cassette have been used to mediate de novo DNA methylation.

Transcripts of such constructs resemble mRNA molecules, and are likely to be exported to the cytoplasm. The

presence of hpRNAs in the cytoplasm and the nucleus may account for the simultaneous activation of post-

transcriptional gene silencing (PTGS) and RNA-directed DNA methylation (RdDM). We hypothesized that by

retaining hpRNAs in the nucleus, efficient induction of only RdDM may be achieved. Thus, we introduced into

tobacco a transgene containing an intron into which an IR of a target promoter was inserted. The intronic

hpRNA initiated highly specific cis- and trans-methylation, but did not induce PTGS. No spreading of

methylation into sequences flanking the region of homology between the hpRNA and the target DNA was

detectable. The efficient methylation-directing activity of the intronic hpRNA may indicate a previously

unrecognized role of introns, potentially regulating gene expression at the transcriptional level.

Keywords: hairpin RNA, intron, RNA-directed DNA methylation (RdDM), RNA interference (RNAi), small RNAs.

INTRODUCTION

In plants, RNA-directed DNA methylation (RdDM) is a key

process of nuclear RNA interference (RNAi) that is primarily

associated with epigenetic gene silencing (Wassenegger,

2005; Huettel et al., 2007; Matzke et al., 2007, 2009; Wierzb-

icki et al., 2008). RdDM is induced by double-stranded RNA

(dsRNA) and, in concert with numerous proteins, leads to

de novo cytosine methylation at symmetric CpG/CpHpG and

asymmetric CpHpH sites (where H = A, T or G) (Matzke et al.,

2009).

RNA polymerase V (Pol V) is a central component of the

de novo methylation complex. The SWI/SNF-like chromatin

remodelling protein, DRD1, and the hinge domain-contain-

ing protein, DSM3, appear to recruit Pol V to chromatin,

although DRD1 does not physically interact with Pol V

(Kanno et al., 2004, 2008; Wierzbicki et al., 2009). As Pol V

does not utilize RNA templates, and is not specifically

targeted to methylated DNA, it was suggested that the

initiation of Pol V transcription occurs throughout the entire

genome. Binding of small complementary RNAs to Pol

V-produced transcripts would then lead to the sequence-

specific directing of the de novo methyltransferase(s) to the

target DNA (Pontier et al., 2005; Wierzbicki et al., 2008; ).

Indeed, in plant nuclei, 24-nucleotide (24-nt) short interfering

RNAs (siRNAs) are processed from dsRNA by Dicer-like 3

(DCL3), and are predominantly loaded onto Argonaute 4

(AGO4) (Zilberman et al., 2003; Chan et al., 2004). The 24-nt

siRNAs are generally presumed to guide the RdDM machin-

ery (Wierzbicki et al., 2009). In concert with the KOW

domain-containing transcription factor 1 (KTF1), AGO4

could either directly interact with Pol V or would associate

with Pol V via siRNA-mediated Pol V transcript binding

(El-Shami et al., 2007; He et al., 2009).

The hierarchical order of processes involved in the self-

reinforcing feedback mechanism is unclear. An SNF2

domain-containing protein (CLSY1), Pol IV, Pol V, RNA-

directed RNA polymerase 2 (RDR2), DCL3 and AGO4 were

found, although to variable extents, to be connected with the

accumulation of 24-nt siRNAs (Chan et al., 2004; Smith et al.,

2007; Mosher et al., 2008). The latter three proteins and the

NRPE1 subunit of Pol V were detectable in the nucleolus-

associated Cajal bodies, indicating that they are siRNA

processing centres (Li et al., 2006; Pontes et al., 2006). Using

RNA templates, probably generated by Pol IV, RDR2 is

producing dsRNA that is processed into secondary 24-nt

siRNAs by DCL3. (Kanno et al., 2005; Zhang et al., 2007;

Mosher et al., 2008). Yet, the substrate for Pol IV needs to be

840 ª 2009 The AuthorsJournal compilation ª 2009 Blackwell Publishing Ltd

The Plant Journal (2009) 60, 840–851 doi: 10.1111/j.1365-313X.2009.04003.x

identified. The currently available data appear to be incon-

sistent (Pikaard et al., 2008). There is good evidence that

Pol IV is essential for the biosynthesis of the majority of

endogenous 24-nt siRNAs. However, if 24-nt siRNAs guide

the de novo methylation machinery, the RdDM-initiating 24-

nt-siRNAs (primary 24-nt siRNAs) cannot be processed from

precursors that are produced by Pol IV-mediated transcrip-

tion of de novo methylated DNA. Thus, primary 24-nt siRNA

biosynthesis must be upstream of Pol V-dependent Pol IV

transcription. Alternatively, one may speculate that RNA

molecules other than 24-nt siRNAs are involved in the

de novo RdDM step.

So far, there is no direct evidence demonstrating that 24-

nt siRNAs guide the RdDM machinery. Daxinger and co-

workers (2009) reported that in an Arabidopsis thaliana dcl3

mutant, RdDM was impaired but not completely lost.

However, even the full inhibition of RdDM would not provide

clear evidence for a direct involvement of the 24-nt siRNAs in

guiding the RdDM machinery. One may speculate that,

instead, they are essential in an intermediate step for the

production of RdDM-guiding molecules. It is likely that the

central role of the 24-nt siRNAs is to reinforce rather than

initiate RdDM. This view is supported by the observation

that inverted repeat (IR)-mediated initiation of de novo DNA

methylation was unaffected in an A. thaliana ago4 mutant

(Zilberman et al., 2004). In ago4-1, maintenance of methyl-

ation was severely impaired, suggesting that AGO4 and

thereby 24-nt siRNAs play a vital role in the maintenance of

RdDM. However, at that time the redundant activity of AGO6

was not known (Zheng et al., 2007). Although AGO6

appeared to generally play only a minor role in RdDM, that

AGO4 was partially replaced by AGO6 in the ago4-1 line

cannot be ruled out. Notably, asymmetric DNA methylation

at a few genomic loci was reported to take place in the

absence of 24-nt siRNAs, implying a mechanism that is

independent of these RNAs (Henderson et al., 2006). How-

ever, it should be noted that diverse RdDM pathways are

likely to exist. Defined chromatin structures (hetero-/euchro-

matin), DNA arrangements (tandem repeats/IRs) and the

sequence context (CpG-rich/CpG-depleted) may require the

involvement of specific RdDM components (Fischer et al.,

2008; Luo et al., 2009).

Initiation of RdDM has been most extensively studied

using transgenic plants. Viroid and virus infections led to the

de novo methylation of transgenes sharing sequence

homology with the pathogen RNAs. Because replication of

the pathogens proceeded without the involvement of DNA,

it became evident that de novo DNA methylation was

initiated by pathogen-derived RNA molecules (Wassenegger

et al., 1994a; Jones et al., 1998). In plants, post-transcrip-

tional gene silencing (PTGS) was found to be generally

accompanied by de novo methylation of the transgene

coding region, indicating that both processes are related

(Wassenegger and Pelissier, 1998). In fact, similarly to PTGS,

RdDM is initiated by dsRNA (Mette et al., 2000). Various

processes can lead to the production of dsRNA, including

RNA virus or viroid replication and transcription of IRs.

However, to initiate RdDM, the trigger RNA needs to be

present in the nucleus.

In this study, we examined the feasibility of using an

intronic hairpin RNA (int-hpRNA) to mediate RdDM. Intron

sequences that are removed during the splicing process

were reported to remain in the nucleus (Qian et al., 1992).

Thus, we speculated that an int-hpRNA that is retained in the

nucleus could be efficiently processed into RdDM-initiating

24-nt siRNA, but not into PTGS-inducing 21-nt siRNAs.

However, only a few tobacco plant lines expressing the int-

hpRNA transgene construct produced detectable quantities

of int-hpRNA-derived 24-nt siRNAs. In addition, accumula-

tion of transgenic siRNAs of other size classes was not

detected in any lines. The absence of the 21-nt siRNAs was in

agreement with the failure of the int-hpRNA to initiate PTGS

of a sensor transgene. Despite the low accumulation of 24-nt

siRNAs the transgene containing the int-hpRNA construct

became heavily de novo methylated. Hypermethylation was

specifically found in the region that was homologous with

the int-hpRNA. Moreover, the introduction of a sensor

transgene into the int-hpRNA-expressing plants revealed

efficient trans-methylation. As was found for the trigger

transgene, only the part of the sensor transgene sharing

homology with the int-hpRNA became de novo methylated.

No spreading of methylation into flanking regions, as

previously reported, was observed (Kanno et al., 2008;

Daxinger et al., 2009). Our data show that int-hpRNA

constructs could serve as a tool to initiate efficient and

highly specific de novo DNA trans-methylation. Further-

more, they indicate that in plants, a natural intron-based

transcriptional gene silencing (TGS) mechanism may exist.

RESULTS

Generation of the GFP-int-hpRNACMPS construct and

introduction into tobacco

We hypothesized that a system predominantly enabling the

initiation of nuclear but not cytoplasmic RNAi would require

an RdDM trigger that is retained in the nucleus. As spliced

intron sequences are proposed to generally stay in the

nucleus (Qian et al., 1992), we inserted an hpRNA into an

intron. However, it was unknown whether an hpRNA would

impair the splicing process, and if int-hpRNAs could be

processed by the nuclear RNAi machinery.

The primary transgene construct, GFPint, comprised the

Solanum esculentum RNA-directed RNA polymerase 1

(SeRDR1) intron 3 inserted into the GFP5 cDNA (Haseloff

et al., 1997) (Figure S1). As an RdDM trigger, an hpRNA

targeting the promoter of a reporter gene was selected. The

hpRNA construct contained a 240-bp antisense and a 215-bp

sense fragment of the Cestrum yellow leaf curling virus

Intronic hairpin-mediated RdDM 841

ª 2009 The AuthorsJournal compilation ª 2009 Blackwell Publishing Ltd, The Plant Journal, (2009), 60, 840–851

(CmYLCV) CMPS promoter (PCMPS) (Stavolone et al., 2003).

The sense and antisense PCMPS fragments sharing 214 bp of

homology were separated by an inverted 177-bp fragment of

the SeRDR1 intron 3, giving the hairpin PCMPS construct

(hpRNACMPS). The hpRNACMPS was inserted into the intron

of the GFPint to give the GFP-int-hpRNACMPS construct

(Figures 1a and S2).

The GFP-int-hpRNACMPS construct was cloned into the

expression cassette of the pPCV702SM binary vector

(Wassenegger et al., 1994b). The corresponding plasmid,

pPCV-GFP-int-hpRNACMPS, was introduced into Agrobacte-

rium tumefaciens, producing the GV-GFP-int-hpRNACMPS

strain that was used to transform tobacco plants. Indepen-

dent transformants were analyzed for the presence of the

T-DNA by PCR, and correct splicing of the intron in planta

was confirmed by northern blot analysis. Total RNA from

individual SR1-GFP-int-hpRNACMPS lines was hybridized

against a GFP5 cDNA probe. In all lines, RNA was detected

with a size corresponding to the RNA of the SR1GFP15E line

expressing the same GFP5 cDNA that was present in the

GFP-int-hpRNACMPS constructs, but lacked an intron (Vogt

et al., 2004). In a low (L) and a moderate (M) GFP-expressing

SR1-GFP-int-hpRNACMPS line, RNA was detected with the

size expected for a spliced transgene RNA (Figure 1b). No

additional RNA was detected with the probe, demonstrating

that the GFP pre-mRNA was fully spliced. Finally, western

blot analysis of the L and M SR1-GFP-int-hpRNACMPS lines

was performed. A GFP-specific antibody detected a single

(a)

(c)

(d)

(b)

Figure 1. Expression profile of the GFP-int-

hpRNACMPS construct.

(a) Physical map of the GFP-int-hpRNACMPS

construct. LB, T-DNA left border; P35S, Cauli-

flower mosaic virus 35S promoter; G, 5¢ part of

the GFP coding region; In-, 5¢ part of Solanum

esculentum intron 3 fragment; CMPS(–), frag-

ment of the CmYLCV CMPS promoter in anti-

sense orientation; spacer, middle part of

S. esculentum intron 3 in reverse orientation;

CMPS(+), fragment of the CmYLCV CMPS pro-

moter in sense orientation; -tron, 3¢ part of

S. esculentum intron 3; -FP, 3¢ part of the GFP

coding region; TNOS, terminator of the nopaline

synthase gene; RB, T-DNA right border.

(b) Northern blot analysis to detect GFP mRNA

levels in SR1-GFP-int-hpRNACMPS lines L and M.

SR1-WT and SR1-GFP15E were used as negative

and positive controls. 25S rRNA accumulation

served as a loading control.

(c) Western blot analysis of GFP protein levels in

SR1-GFP-int-hpRNACMPS lines L and M. SR1-WT

and SR1-GFP15E were used as negative and

positive controls.

(d) Northern blot analysis to detect CMPS siRNAs

in SR1-GFP-int-hpRNACMPS lines L and M. SR1-

WT was used as a negative control. An RNA

sample from a Potato Spindle Tuber Viroid

(PSTVd)-infected SR1-WT plant was used as a

24–21-nt size marker (the membrane was re-

hybridized against a PSTVd cDNA to visualize the

PSTVd siRNAs). 5S rRNA accumulation served as

the loading control.

842 Athanasios Dalakouras et al.

ª 2009 The AuthorsJournal compilation ª 2009 Blackwell Publishing Ltd, The Plant Journal, (2009), 60, 840–851

29-kDa protein corresponding to the GFP expressed in the

SR1GFP15E line (Figure 1c). In summary, the SeRDR1

intron 3 proved to be functional, and was efficiently and

accurately spliced, even though it contained the hpRNACMPS

fragment.

Processing of the GFP-int-hpRNACMPS into siRNAs

Total RNA from the SR1-GFP-int-hpRNACMPS L and M lines

was used for siRNA analysis. Northern blots using a PCMPS-

specific probe detected a signal in line M samples corre-

sponding to 24-nt-long siRNAs only (Figure 1d). No 21-nt

PCMPS siRNAs or siRNAs of other size classes were detect-

able. The 24-nt siRNAs were not detected in line L

(Figure 1d).

Transgene transcripts can induce the production of

siRNAs. However, expression of the GFP-int-hpRNACMPS

transgene represents a rare example in which a Pol II-

transcribed transgenic RNA is processed into 24-nt siRNAs,

but not into siRNAs of other size classes. Agroinfiltration of

SR1-GFP-int-hpRNACMPS plants using an A. tumefaciens

strain containing a ‘classical’ GFP-hairpin construct (pPCV-

GpG) produced both sizes of GFP siRNAs, indicating that

biosynthesis of 21-nt siRNA was not affected in SR1-GFP-int-

hpRNACMPS lines (Figure S3). Finally, the fact that no GFP

siRNAs were detectable in mock-infiltrated SR1-GFP-int-

hpRNACMPS lines demonstrated that GFP silencing was not

induced. This finding was in accordance with the absence

of GFP mRNA degradation products in SR1-GFP-int-

hpRNACMPS lines (Figure 1b). In summary, our data indicate

that processing of the GFP-int-hpRNACMPS into siRNAs was

exclusively catalyzed by the nuclear RNAi machinery.

Moreover, the absence of GFP siRNAs suggested that

nuclear RNAi was not associated with transitive silencing.

Notably, transgene intron-specific siRNAs, including loop-

specific siRNAs, were also absent, supporting the view that

no transitive silencing had taken place.

Initiation of RdDM in SR1-GFP-int-hpRNACMPS lines

In order to investigate the capability of the int-hpRNA to

trigger RdDM of cognate DNA, the methylation status of the

GFP-int-hpRNACMPS transgene was examined by Southern

blot analysis using the methylation-sensitive restriction

endonuclease Sau96I. A single Sau96I restriction site

(5¢-GGNCC-3¢) exists in the antisense and sense PCMPS se-

quences (Figure 2a, S4 and S5), and there are three addi-

tional Sau96I sites within the upstream P35S sequence

(Figure 2a, S1, S2 and S3). In the PCMPS sequences, the S4 and

S5 sites allowed the analysis of CpHpH-methylation. Thus,

Southern blot analysis of the PCMPS sequences with Sau96I

allowed the detection of only asymmetric methylation,

which is a hallmark of RdDM (Pelissier et al., 1999). In

addition to digestion with Sau96I, the genomic DNA was cut

with the methylation-insensitive restriction endonuclease

AseI. As a control for fully cleaved DNA, the pPCV-GFP-int-

hpRNACMPS plasmid DNA was used (Figure 2b). Hybridiza-

tion of the fully cleaved non-methylated plasmid DNA with a

P35S/G probe (Figure 2a, black bar) detected an 800-bp

fragment. In contrast, a major 1482-bp fragment was

detected in both SR1-GFP-int-hpRNACMPS lines (Figure 2b).

This fragment can be explained by cleavage inhibition of

the S4 and S5 sites, indicating that these two sites were

methylated.

In order to obtain detailed data on the methylation status

of the transgene, bisulfite sequencing was conducted.

Treatment of DNA with sodium bisulfite results in the

conversion of non-methylated cytosine to uracil, and during

(a)

(b)

Figure 2. Southern blot analysis of the methylation status of the GFP-int-

hpRNACMPS transgene.

(a) Physical map of the GFP-int-hpRNACMPS transgene, showing the expected

AseI/Sau96I restriction fragments of fully (FC) and partially (PC) cleaved DNA.

The two AseI and five Sau96I sites are indicated. The black bar specifies the

hybridization probe.

(b) Autoradiogram of the Southern blot.

Intronic hairpin-mediated RdDM 843

ª 2009 The AuthorsJournal compilation ª 2009 Blackwell Publishing Ltd, The Plant Journal, (2009), 60, 840–851

PCR amplification uracil is replaced by thymine. Thus,

sequences of PCR products from bisulfite-treated DNA

exhibit thymines for non-methylated cytosines. A mixed

plasmid/tobacco DNA sample and 2 lg of genomic DNA

pooled from the SR1-GFP-int-hpRNACMPS L and M lines

were subjected to bisulfite sequencing. Two transgene

fragments were analyzed: the 254-bp fragment (CMPS254)

containing parts of the 5¢ GFP5, the 5¢ intron and the PCMPS

antisense sequences, and the 360-bp fragment (CMPS360)

containing the 3¢ spacer, the PCMPS sense and a part of the 3¢intron sequences (Figure 3). Bisulfite sequencing is strand-

specific. The upper strand was analyzed using the primer

pairs BisF2/BisR2 for CMPS254 amplification and BisF3/BisR3

for CMPS360 amplification. Sequencing of cloned PCR prod-

ucts revealed that inserts originating from the pPCV-GFP-int-

hpRNACMPS plasmid that was not methylated were fully

converted (Figure 3, s). In contrast, sense and antisense

PCMPS sequences amplified from the plant genome-inte-

grated transgene were methylated to almost 100% (Figure 3,

). Importantly, methylation did not spread into sequences

flanking the PCMPS-specific sequences. All cytosines within

the intron, GFP5 and spacer were found to be converted in all

clones (Figure 3, h). It should be noted that in the first set of

experiments, DNA material from lines M and L was pooled

to avoid sample bias. However, methylation patterns were

identical when samples of the two lines were separately

analyzed.

To validate the bisulfite sequencing data, restriction

analysis of the PCR products obtained from amplified

bisulfite-treated DNA was also performed. If a cytosine

residue located in a restriction site of the authentic sequence

is converted, the site will be lost. Thus, PCR products

amplified from this DNA will lack the restriction site, and so

analysis of the PCR products will give an overview of the

methylation status of the tissue from which the genomic

DNA was extracted. The results of this analysis (Figure S4)

confirmed the Southern and bisulfite sequencing data,

and altogether clearly demonstrated that the GFP-int-

hpRNACMPS construct efficiently initiated RdDM. Methyla-

tion was highly specific for the sequences that shared

homology with the double-stranded region of GFP-int-

hpRNACMPS. Neither the spacer nor the intron or GFP5

sequences flanking the sense and antisense copy of the

PCMPS were targeted by RdDM.

GFP-int-hpRNACMPS efficiently triggered trans-methylation

of a sensor construct

GFP-int-hpRNACMPS induced the specific methylation of the

genome-integrated transgene region from which the hpRNA

was transcribed. To explore whether, in addition to this

cis-methylation, the GFP-int-hpRNACMPS had the capacity to

trigger trans-methylation, a ‘methylation-sensor’ construct

was introduced into tobacco. The construct comprised a 791-

bp red-shifted GFP (rsGFP) cDNA driven by the full-length

Figure 3. Bisulfite sequencing data of the GFP-int-hpRNACMPS transgene.

Physical map of the GFP-int-hpRNACMPS transgene, in which the regions analyzed by bisulfite sequencing are indicated. Abbreviations are as in Figure 1a. The 254-

bp-long (CMPS254) and the 360-bp-long (CMPS360) transgene fragments were separately analyzed. CMPS254 comprised parts of the 5¢ GFP, the 5¢ intron (underlined)

and the 5¢ CMPS(–) (bold) sequences. CMPS360 comprised parts of the spacer (lower case), the full-length CMPS(+) (bold) and the 3¢ intron (underlined) sequences.

Circles represent bisulfite sequencing data produced with pPCV-GFP-int-hpRNACMPS plasmid DNA, and squares represent those produced with SR1-GFP-int-

hpRNACMPS genomic DNA. The percentage levels of cytosine methylation are indicated by open circles and squares (0–10%), by half-filled symbols (40–60%) and by

filled symbols (90–100%). Detailed sequencing data for each clone are presented in Figure S8. The XbaI (5¢-TCTAGA-3¢) and HpyCH4IV (5¢-ACGT-3¢) restriction sites

used for restriction analysis of the bisulfite PCR product (Figure S4) are set in italic font, and are indicated by a gray and open box, respectively.

844 Athanasios Dalakouras et al.

ª 2009 The AuthorsJournal compilation ª 2009 Blackwell Publishing Ltd, The Plant Journal, (2009), 60, 840–851

406-bp PCMPS (PCMPS-rsGFP) (Figures 4 and S5). SR1-PCMPS-

rsGFP plant lines moderately expressing rsGFP were identi-

fied by northern blot analysis. No PCMPS siRNAs were

detectable in these lines. The SR1-PCMPS-rsGFP line 1 was

genetically crossed with the SR1-GFP-int-hpRNACMPS line M

to generate the SR1-PCMPS/int-hpRNACMPS line 1. T1 progeny

plants of this line, in which the presence of both transgenes

was verified by PCR, were used for further experiments.

In order to clarify if the presence of the GFP-int-

hpRNACMPS transgene led to the trans-methylation of the

PCMPS sequence, bisulfite sequencing was performed. A 226-

bp fragment (CMPS226) was analyzed, of which the 3¢ half

overlapped with the GFP-int-hpRNACMPS (Figure 4, upper

panel). DNA from three independent T1 plants of the

SR1-PCMPS/int-hpRNACMPS line 1 was pooled, bisulfite trea-

ted and amplified with the BisF1/BisR1 primer pair. The

pPCV800-PCMPS-rsGFP plasmid and genomic DNA from a T1

plant of the SR1-PCMPS-rsGFP line 1 served as controls. In the

controls, CMPS226 was virtually free of methylation (Fig-

ure 4, upper panel, s and h) whereas in the SR1-PCMPS/int-

hpRNACMPS progeny, the 3¢ half of CMPS226 overlapping

with GFP-int-hpRNACMPS was fully methylated (Figure 4,

upper panel, sequence in bold). Importantly, in the double

transformants, the non-overlapping part of PCMPS was not

methylated, demonstrating that the spread of methylation

into the 5¢ flanking region had not taken place. This finding

was validated by restriction analysis of PCR products from

bisulfite-treated and untreated DNA (Figure S6a).

So far, the spread of RdDM-mediated methylation along

promoter sequences was only found to proceed into

downstream sequences (Kanno et al., 2008; Daxinger et al.,

2009). Thus, we analyzed the junction between the PCMPS

and the rsGFP sequences of the PCMPS-rsGFP transgene by

bisulfite sequencing. The 354-bp CMPS354 fragment was

amplified using the BisF4/BisR4 primer pair from bisul-

fite-treated DNA of the pPCV800-PCMPS-rsGFP plasmid, the

SR1-PCMPS-rsGFP line 1 and from SR1-PCMPS/int-

hpRNACMPS T1 plants (Figure 4, lower panel). Sequencing

of individual clones containing PCR products from the

pPCV800-PCMPS-rsGFP plasmid and the SR1-PCMPS-rsGFP

line 1 revealed that the region corresponding to CMPS354

was not methylated. In contrast, the region of the CMPS354

sequence overlapping with int-hpRNACMPS was fully methy-

lated in SR1-PCMPS/int-hpRNACMPS plants. However, as was

found for cis-methylation of the GFP-int-hpRNACMPS trans-

gene, trans-methylation also appeared to decline immedi-

ately downstream of the int-hpRNACMPS/CMPS354 overlap.

Virtually no 5mC was detectable downstream of the junc-

tion between the PCMPS and the rsGFP sequence (Figure 4,

lower panel). This finding was again validated by restriction

analysis of PCR products from bisulfite-treated and

untreated DNA (Figure S6b,c).

Figure 4. Bisulfite sequencing data of the PCMPS-rsGFP sensor construct.

Physical map of the PCMPS-rsGFP transgene, in which the regions analyzed by bisulfite sequencing are indicated. The 226-bp-long (CMPS226) and the 354-bp-long

(CMPS354) transgene fragments were separately analyzed. CMPS226 comprised the middle part of the full-length CmYLCV CMPS promoter (PCMPS). CMPS354

comprised the 3¢ part of the PCMPS and the 5¢ rsGFP sequences. The bold sequences indicate regions of complementarity between int-hpRNACMPS (methylation

trigger) and PCMPS (methylation target). Circles represent bisulfite sequencing data produced with pPCV-PCMPS-rsGFP plasmid DNA, squares represent those

produced with SR1-PCMPS-rsGFP genomic DNA and triangles represent those produced with SR1-PCMPS/int-hpRNACMPS genomic DNA. The percentage levels of

cytosine methylation are indicated by open symbols (0–10%), and by filled symbols (90–100%). Detailed sequencing data for each clone are presented in Figure S8.

The guanidine of transcriptional start site (TSS) is boxed in dark gray. The BsgI (5¢-GTGCAG(N)16-3¢), Hpy188I (5¢-TCNGA-3¢) and EagI (5¢-CGGCCG-3¢) restriction sites

used for restriction analysis of the PCR products (Figure S6) are all set in italic font, and are underlined or indicated by a gray and open box, respectively.

Abbreviations are as in Figure 1a; rsGFP, red-shifted GFP.

Intronic hairpin-mediated RdDM 845

ª 2009 The AuthorsJournal compilation ª 2009 Blackwell Publishing Ltd, The Plant Journal, (2009), 60, 840–851

It has been proposed that the spread of 3¢ methylation

involves a nascent transcript of the target. To investigate

whether such a transcript existed, and whether it had an

effect on 3¢ spread, we mapped the transcriptional start site

(TSS) of the PCMPS-rsGFP sensor using the 5¢ RACE analysis.

The TSS was located 60-bp upstream from the end of the

overlap between the trigger and the target sequence,

demonstrating that the 3¢ region of the target was tran-

scribed (Figure 4). Nevertheless, no spreading of methyla-

tion into the 3¢ flanking region was observed.

GFP-int-hpRNACMPS failed to trigger the PTGS of a sensor

construct

A non-polyadenylated hpRNA could lead to RdDM, but not

to the PTGS of homologous sequences (Mette et al., 2001).

The GFP-int-hpRNACMPS construct also efficiently triggered

cis- and trans-methylation. However, it was not clear whe-

ther it would be able to mediate target RNA cleavage. DCL3-

produced 24-nt siRNAs were not active in directing RNA

cleavage (Deleris et al., 2006; Fusaro et al., 2006). We

detected only 24-nt PCMPS siRNAs, but it was recently shown

that 24-nt siRNAs produced from a Pol III-transcribed hpRNA

triggered mRNA degradation (Wang et al., 2008). In addi-

tion, we could not fully exclude the presence of another class

of siRNAs in the SR1-GFP-int-hpRNACMPS lines. For exam-

ple, 21-nt RNAs may have accumulated below the detection

limit but still have been sufficiently abundant to initiate

PTGS. To study if PTGS could be initiated, the pPCV-GUS-

CMPS plasmid was generated (Figures 5a and S7). This

RNAi sensor construct contained a transcriptional fusion of a

PCMPS fragment to the 3¢ end of GUS cDNA. The pPCV-GUS-

CMPS was introduced into A. tumefaciens, and the resulting

bacterial strain GV-GUS-CMPS was used for co-infiltration

assays of SR1-WT plants (see below).

In addition, a putative PTGS trigger construct was gener-

ated. A ‘classical’ hpRNA was produced, in which the hpRNA

construct was directly flanked by the cauliflower mosaic

virus 35S promoter (P35S) and the nopaline synthase poly-

adenylation signal sequence (TNOS), by cloning the

hpRNACMPS fragment into the expression cassette of the

pPCV702SM vector (Figure 5b). The resulting pPCV-

hpRNACMPS was introduced into A. tumefaciens, and

GV-hpRNACMPS was co-infiltrated with GV-GUS-CMPS,

which served as a control for the initiation of the PTGS of

the GUS-CMPS sensor gene. Our prediction was that in

contrast to the hpRNA molecules produced by the int-

hpRNACMPS, those generated by hpRNACMPS would be

mainly processed into 21-nt siRNAs that would lead to the

silencing of the GUS-CMPS mRNA. Infiltration of sense

constructs in wild-type plants generally results in the

accumulation of siRNAs (Szittya et al., 2003). However, as

was previously reported, a low infiltration OD would filter

out this ‘noise’, but would allow the trans-PTGS of the

sensor constructs to take place (Koscianska et al., 2005).

Tobacco plants were co-infiltrated with (i) GV-GUS-CMPS/

GV-D, (ii) GV-GUS-CMPS/GV-GFP-int-hpRNACMPS and

(iii) GV-GUS-CMPS/GV-hpRNACMPS. The GV-D contained

the pPCV702SM T-DNA, but lacked the P35S/TNOS expression

cassette. This construct was used to demonstrate that

co-infiltration per se did not enhance the initiation of PTGS.

Total RNA was isolated at 4 days post-inoculation (4 dpi),

and was then used for RNA analysis (Figure 5c). The level of

GUS-CMPS mRNA was clearly decreased in (iii) when

compared with (i) and (ii). The presence of mature GFP

mRNA in (ii) demonstrated that after agroinfiltration, the

(a)

(b)

(c)

Figure 5. Northern blot analysis to study the capacity of the GFP-int-

hpRNACMPS to trigger PTGS of the GUS:CMPS sensor construct.

(a) Physical map of the GUS:CMPS transgene. Abbreviations are as in

Figure 1a; CMPS, CMPS promoter fragment in sense orientation with

homology to int-hpRNACMPS, as indicated.

(b) Physical map of the ‘classical’ hpRNACMPS transgene construct that was

used as a PTGS inducer. Abbreviations are as in Figure 1a.

(c) Northern blot analysis of GUS-CMPS mRNA, GFP mRNA, GUS siRNAs and

CMPS siRNAs, after co-agroinfiltration of GV-GUS:CMPS with GV-D, GV-GFP-

int-hpRNACMPS and GV-hpRNACMPS, respectively. 25S rRNA accumulation

served as the loading control.

846 Athanasios Dalakouras et al.

ª 2009 The AuthorsJournal compilation ª 2009 Blackwell Publishing Ltd, The Plant Journal, (2009), 60, 840–851

expression and processing of GFP-int-hpRNACMPS was not

impaired. Abundant PCMPS siRNAs were detected only when

the ‘classical’ hpRNACMPS construct, but not the GFP-int-

hpRNACMPS construct, was introduced. Accumulation of

GUS-specific secondary siRNAs was detectable only after

the co-infiltration of GV-GUS-CMPS with GV-hpRNACMPS,

but not with GV-GFP-int-hpRNACMPS. Our co-infiltration data

demonstrated that, in contrast to hpRNACMPS, GFP-int-

hpRNACMPS failed to trigger sensor mRNA degradation.

DISCUSSION

In plants, some miRNAs are known to reside within introns.

However, in contrast to our construct, they only triggered

PTGS (Voinnet, 2009). Compared with this, an intronic

transposon within FLC alleles is targeted by, and is probably

a source of, siRNAs that lead to heterochromatinization

(Mylne et al., 2006). Apart from these examples, a brief

bioinformatics-based analysis of the A. thaliana genome

provided no evidence for int-hpRNAs existing in nature.

However, so far, plants possessing larger genomes have not

been investigated. Recently, it was reported that several

thousand unique sequences exist in maize that could form

hairpins, but they were not miRNA precursors (Wang et al.,

2009). Whether such hairpins were located in introns is not

known.

Here, we demonstrate that, in tobacco, int-hpRNA trig-

gered the de novo cis- and trans-methylation of transgenes.

The efficiency of this system to trigger RdDM was remark-

able. Full methylation patterns were established already in

the primary transformants (cis-RdDM), and in the first

generations after crossing the trigger with a sensor

construct (trans-RdDM). This was in contrast to previous

findings, where more than three generations were required

(Aufsatz et al., 2002; Zilberman et al., 2004). Full methylation

patterns were homogeneous along the entire target

sequence, unlike previous reports where methylation was

most efficient near the IR center (Stam et al., 1998). Finally,

no spread of cis- and trans-methylation into 5¢ or 3¢ flanking

sequences took place, as observed previously (Daxinger

et al., 2009). After splicing, intron sequences are generally

not exported to the cytoplasm (Ernst et al., 1997; Zhou et al.,

2000). It is reasonable to assume that, as a part of the intron,

int-hpRNACMPS was also retained in the nucleus. The high

efficiency of the intronic system may thus be based on the

fact that int-hpRNACMPS was predominantly accessible to

the nuclear RNAi machinery.

Plant introns do not have absolute requirements for

branch sites and polypyrimidine tracts. An A/T-rich base

composition, a minimal size of 70-nt and the presence of

5¢-GT- and 3¢-AG- splice sites are sufficient for intron

functionality (Goodall and Filipowicz, 1991). We have ran-

domly chosen the site where the int-hpRNACMPS cDNA was

inserted into the intron. Secondary structures inside introns

could inhibit pre-mRNA splicing in dicotyledons and yeast.

Hairpin structures of only 18 or 24 bp had a strong inhibitory

effect on splicing in Nicotiana plumbaginifolia (Liu et al.,

1995). In the GFP-int-hpRNACMPS construct, the IR was

inserted 45-bp downstream of the 5¢ splice site. On the basis

of the currently available data on pre-mRNA processing, one

would have expected that the int-hpRNACMPS could interfere

with the splicing process. However, the secondary structure

of the hpRNACMPS had no obvious effect on splicing

accuracy or efficiency of the GFP-int-hpRNACMPS pre-mRNA.

Little is known about the decay of intron sequences after

splicing. Thus, it was not clear whether int-hpRNACMPS

would be accessible to the nuclear RNAi machinery, or

whether intron degradation would proceed faster than

int-hpRNACMPS processing by nuclear RNAi enzymes. The

observation that the int-hpRNACMPS of the SR1-GFP-int-

hpRNACMPS line M was processed into detectable quantities

of 24-nt siRNAs argued for the former case. However, the

concentration of int-hpRNACMPS 24-nt siRNAs in line M was

relatively low, and no int-hpRNACMPS siRNAs were found in

line L (Figure 1d). From our data, we could not conclude

whether the int-hpRNACMPS 24-nt siRNAs were primary

siRNAs directly generated from the int-hpRNACMPS by DCL3,

or whether they derived from a secondary siRNA production

pathway that may involve Pol IV and RDR2 activity.

Apart from 24-nt siRNAs, int-hpRNACMPS was apparently

not processed into 21-nt siRNAs or any other siRNA size

class. The absence of 21–22-nt siRNAs was in accordance

with the observation that the PTGS sensor was not silenced

in the presence of the GFP-int-hpRNACMPS construct. In

contrast, expression of the ‘classical’ hpRNACMPS led to the

degradation of the PTGS sensor mRNA, indicating that the

CMPS sequence was a target of RNAi (Figure 5c). In contrast

to 24-nt siRNAs that were derived from Pol II-transcribed

hpRNA, those produced from a Pol III-transcribed hpRNA

were PTGS competent (Fusaro et al., 2006; Wang et al.,

2008). This indicated that two types of 24-nt siRNAs exist

and/or that 24-nt siRNAs associate with diverse silencing

complexes, which could involve different AGO proteins.

AGO4, which is mainly associated with 24-nt siRNAs, has a

cleavage activity that is required for de novo methylation at

some loci (Qi et al., 2006). However, no cytoplasmic AGO4

activity has been reported so far. Thus, it is likely that 24-nt

siRNA-mediated PTGS involves siRNA loading onto a

different AGO protein.

The bisulfite data showed that in the SR1-GFP-int-

hpRNACMPS lines, transgene sequences sharing homology

with the double-stranded part of the hpRNACMPS displayed

essentially identical de novo methylation patterns. Accu-

mulation of different 24-nt siRNA levels, as found in these

lines, had no apparent effect on the density of target

sequence methylation. It is generally assumed that 24-nt

siRNAs guide the de novo DNA methylation machinery

(Chan et al., 2004; Wierzbicki et al., 2008; Daxinger et al.,

2009; Huang et al., 2009). Our findings raise the question of

Intronic hairpin-mediated RdDM 847

ª 2009 The AuthorsJournal compilation ª 2009 Blackwell Publishing Ltd, The Plant Journal, (2009), 60, 840–851

whether RNA molecules other than siRNAs direct methyl-

ation. As we observed here for line L, other reports

describe cases of RdDM in the absence of detectable

quantities of 24-nt siRNAs (Melquist and Bender, 2003;

Henderson et al., 2006; Lister et al., 2008). However, it can

not be excluded that siRNAs were produced, but to levels

that were under the detection limit. Ultra-deep sequencing

of the A. thaliana smRNAome revealed that siRNAs were

only associated with approximately a third of all genomic

cytosine methylation (Lister et al., 2008). In maize, many

siRNAs mapped to transposable elements (TEs) that were

not methylated, wereas TEs in which DNA was highly

methylated were relatively devoid of siRNAs (Wang et al.,

2009). These findings would argue against the requirement

of siRNAs for RdDM at some genomic loci. However, from

the available data it was not clear whether methylation of

regions that were not associated with siRNAs was based on

RdDM or the MET1-mediated maintenance of methylation.

Appearance of asymmetric methylation, the hallmark of

RdDM, within these regions would support the view that

siRNAs are not generally needed for the initiation of RdDM.

If 24-nt siRNAs guide the RdDM machinery, we would have

expected a denser methylation pattern in the SR1-GFP-int-

hpRNACMPS line M, accumulating 24-nt siRNAs, than in the

SR1-GFP-int-hpRNACMPS line L, where no siRNAs were

detected.

The regions of the SR1-GFP-int-hpRNACMPS lines and

SR1-PCMPS/int-hpRNACMPS progeny plants that became

methylated corresponded exclusively and precisely to

sequences sharing homology with the double-stranded

region of hpRNACMPS (Figures 3 and 4). This finding pointed

to a kind of ‘molecular ruler’ that would enable the

measurement of regions to be methylated at the nucleotide

level. An accurate ruler would be the double-stranded region

of int-hpRNACMPS or a-single-stranded RNA corresponding

to this region. The highly specific boundaries of methylation

observed here strongly argue against a function of 24-nt

siRNAs as rulers. It is unlikely that they precisely cover the

full length of the hpRNACMPS, as was also shown for a

transgenic nopaline synthase promoter fragment (Papp

et al., 2003). If the 24-nt siRNAs function as rulers, it must

be hypothesized that DCL3 processing started at the first and

stopped at the last nucleotide of the double-stranded region

of int-hpRNACMPS. In addition, accumulation of a homoge-

neous 24-nt siRNA population would help to explain the

equal distribution and density of methylation along the

targeted DNA. However, deep sequencing revealed that

thymine (T) is highly over-represented at position –1 of 24-nt

siRNA precursors, indicating that DCL3 processing occurs at

preferential sites. In addition, adenine (A) is the most

common first base of 24-nt siRNAs (Lister et al., 2008; Wang

et al., 2009). The preference for an A in the first position is in

accordance with the finding that in A. thaliana, 79% of

AGO4-associated 24-nt siRNAs exhibit a 5¢ terminal A (Mi

et al., 2008). These data indicated that 24-nt siRNAs are

unlikely to homogeneously map to the DNA that is targeted

by RdDM. Therefore, if 24-nt siRNAs were required to guide

the de novo DNA methylation machinery, one would expect

a more heterogeneous methylation pattern of the int-

hpRNACMPS target DNA. In summary, it is still unclear which

type of RNA molecules direct the de novo DNA methylation

machinery. In contrast to the general assumption, 24-nt

siRNAs may not be involved in the triggering of this process,

but in its amplification and maintenance.

So far, we have only generated a single int-hpRNA

construct. Thus, it can not be excluded that GFP-int-

hpRNACMPS-induced RdDM is a unique and artificial excep-

tion. The functionality of additional introns, RdDM triggers

and intron insertion sites must be tested to generalize our

findings. Currently, we are investigating the effect of a

nucleus-retained intronic hairpin on the induction of TGS of

transgenes and endogenous genes. In addition, we are

interested in further examining our finding of putative

selective Dicer processing of the intronic hairpin by subcel-

lular fractionation and siRNA/dsRNA detection. Finally, deep

sequencing of siRNAs and characterization of the GFP-int-

hpRNACMPS construct in a dcl2/dcl3/dcl4 triple mutant back-

ground will help to gain additional information about the

significance of 24-nt siRNAs in RdDM.

EXPERIMENTAL PROCEDURES

Plant transformation and agroinfiltration

All pPCV702SM binary vector derivatives were introduced into theA. tumefaciens strain GV3101. Nicotiana tabacum (cv. Petit HavanaSR1) plants were transformed using the leaf disc transformationmethod, as previously described (Wassenegger et al., 1994a), and/or were used for infiltration of N. tabacum leaves at an OD of 0.5, asdescribed previously (Koscianska et al., 2005).

Bisulfite analysis

Tobacco genomic DNA was isolated with the DNEasy Plant Mini Kit(Qiagen, http://www1.qiagen.com), and 2 lg of DraI-digested DNAwas bisulfite-treated with the EZ DNA Methylation Gold kit (ZYMOResearch, http://www.zymoresearch.com). The recovered materialwas amplified by PCR with AmpliTaq Gold polymerase (AppliedBiosystems, http://www.appliedbiosystems.com) under the follow-ing conditions: 10 min at 95�C, 45 amplification cycles (95�C for30 sec, 50�C for 30 sec and 72�C for 30 sec), and 72�C for 10 min.The CMPS226 was amplified with the BisF1/BisR1 primers, theCMPS254 with the BisF2/BisR2 primers and the CMPS360 with theBisF3/BisR3 primers (for primer sequences see Table S1). As apositive control for full cytosine conversion, 2 lg of SR1 wild-type(WT) DNA was mixed with approximately 40 pg of DraI-cleavedplasmid DNA, corresponding to the genome-integrated T-DNA. Theresulting PCR products were cloned into the pGEM-T Easy vector(Promega, http://www.promega.com), and 10 clones weresequenced from each experiment.

Plasmid construction

The details of all cloning strategies and the sequences of the primersused are shown in Figures S1–S7 and in Table S1, respectively. The

848 Athanasios Dalakouras et al.

ª 2009 The AuthorsJournal compilation ª 2009 Blackwell Publishing Ltd, The Plant Journal, (2009), 60, 840–851

pPCV-GFPint was generated by introducing a modified SeRDR1 in-tron 3 (accession number AC209589; base pairs 102578–103212) intothe GFP5 between nucleotides 375 and 376 of the cDNA (Figure S1).The intron was modified by introducing unique BglII and XbaI sites,which were subsequently used to insert the hairpin CMPS construct(hpRNACMPS). The hpRNACMPS was produced by amplifying a senseand an antisense CmYLCV CMPS promoter (PCMPS) fragment fromthe pNOV2820 plasmid (Syngenta, http://www.syngenta.com)(Figure S2). The two fragments were separated by a 120-bp frag-ment that was amplified from the SeRDR1 intron 3. The assembledhpRNACMPS was cloned into pT3T7-lac (Roche, http://www.roche.com), giving pTPCR-hpRNACMPS. From this plasmid, the hpRNACMPS

was released and introduced into pPCV-GFPint.The pPCV-GpG binary vector expressed an IR GFP5 construct

containing the second 143-bp Sau3AI fragment of the GFP5 cDNA(accession number U87973; base pairs 246–389) in sense andantisense orientation. The two fragments were separated by a 90-bpspacer that was derived from a Potato spindle tuber viroid cDNA.

The pPCV800-PCMPS-rsGFP plasmid was generated by modifyingthe pPCV702SM binary vector (Figure S5), whereby an additionalexpression cassette was introduced, comprising the full-lengthPCMPS, the red-shifted GFP (rsGFP; accession number FM883229)and the polyadenylation signal sequence of the A. tumefaciens tmrgene (pA-tmr) (accession number X00010; base pairs 1363–1755).

pPCV-GUS-CMPS was produced by fusing one of the PCMPS

fragments of the hpRNACMPS construct in sense orientation to the3¢ end of GUS cDNA of the pBI121 plasmid (Clontech, http://www.clontech.com). The GUS-CMPS transcriptional fusion productwas then introduced into the P35S/TNOS expression cassette ofpPCV702SM (Figure S7).

Northern, Southern and Western blot analysis

Total RNA was extracted from N. tabacum leaves with Tri-ReagentSolution (Ambion, http://www.ambion.com). For Northern blotanalysis, 10 lg of total RNA was separated on a 1.2% agaroseformaldehyde gel, capillary transferred onto a positively chargedmembrane BioBond Plus (Sigma-Aldrich, http://www.sigmaaldrich.com) and UV312nm-cross-linked (300 mJ cm)2). Random-primed[a-32P]dCTP-labeled PCR fragments (Random Primed DNA LabellingKit; Roche) were used as probes. The PerfectHyb Plus 1x (Sigma-Aldrich) was used for overnight hybridization at 64�C. Membraneswere washed at 64�C with buffer 1 (2x SSC, 0.1% SDS, w/v) for30 min, and with buffer 2 (0.5x SSC, 0.1% SDS, w/v) for 15 min.Membranes were exposed to FujiFilm Imaging Plates (FujiFilm,http://www.fujifilm.com) for 24 h and scanned using PharosFXPlus PhosphorImager (BioRad, http://www.bio-rad.com). For siRNAblot analysis, total RNA was extracted as above, and 20 lg wasseparated on a 20% Tris-Borate-EDTA (TBE)-acrylamide gel (Ana-med, http://www.anamed-gele.com) at 80 V for 4 h. RNA wastransferred onto positively charged nylon membranes (Ambion) byelectro-blotting at 300 mA for 1 h. Semidried membranes were UV-cross-linked as above. The appropriate PCR product for eachexperiment was labeled and purified as above. The hybridizationtemperature was 42�C, and membranes were washed once withbuffer 1 (2x SSC, 0.1% SDS, w/v) at 42�C for 30 min.

The DNEasy Plant Maxi kit (Qiagen) was used for plant DNAextraction, and Southern blot analyses were performed as previ-ously described (Vogt et al., 2004).

Western blot analysis was performed essentially as described bySambrook et al. (1989). Protein samples were separated on 9%SDS-polyacrylamide gels (Anamed) and transferred onto a polyv-inyldidene fluoride (PVDF) membrane (Roche). Membranes wereprobed with rabbit polyclonal anti-GFP (Santa-Cruz, http://

www.scbt.com). Primary antibodies were detected using a horse-radish peroxidase-labeled goat anti-rabbit IgG secondary antibody(Santa-Cruz).

5¢ RACE

The 5¢ cDNA ends were amplified using the GeneRacer Kit (Invi-trogen, http://www.invitrogen.com), following the manufacturer’sinstructions. The PCR reaction was performed using the GeneRacer5¢ primer and the gene-specific reverse primer RACE-R (Table S1).The PCR products were excised from the gel, cloned to pGEM-TEasy and sequenced.

ACKNOWLEDGEMENTS

We are grateful to Gunther Buchholz for providing the pMVA4fplasmid containing the red-shifted GFP cDNA, and to Detlef Weigeland Stefan Henz for providing the bioinformatics data on A. thali-ana. We thank Arno Putz, Milena Kitova and Marc Fullgrabe forassistance, Olivier Voinnet and Patrice Dunoyer for fruitful discus-sions, and Mike Haydon for critically reading the manuscript. Theexcellent care of plants in the glasshouse was provided by HeikoHerrmann. This work was supported by the grants of the Sixth Re-search Framework Programs of the European Union, Project LSHG-CT-2006-037900 (SIROCCO) and LSHG-CT-2004-005120 (FOSRAK).

SUPPORTING INFORMATION

Additional Supporting Information may be found in the onlineversion of this article:Figure S1. Cloning strategy for pPCV-GFPint.Figure S2. Cloning strategy for pPCV-GFP-int-hpRNACMPS.Figure S3. GV-GpG siRNAs.Figure S4. Verification of the CMPS254 and CMPS360 bisulfitesequencing data by restriction analysis of PCR products.Figure S5. Cloning strategy for pPCV800-PCMPS-rsGFP.Figure S6. Verification of the CMPS226 and CMPS354 bisulfitesequencing data by restriction analysis of PCR products.Figure S7. Cloning strategy for pPCV-GUS-CMPS.Figure S8. Detailed bisulfite sequencing data for CMPS254, CMPS360,CMPS226 and CMPS354.Table S1. Primers used for bisulfite sequencing and plasmidconstruction.Please note: Wiley-Blackwell are not responsible for the content orfunctionality of any supporting materials supplied by the authors.Any queries (other than missing material) should be directed to thecorresponding author for the article.

REFERENCES

Aufsatz, W., Mette, M.F., van der Winden, J., Matzke, A.J. and Matzke, M.

(2002) RNA-directed DNA methylation in Arabidopsis. Proc. Natl. Acad. Sci.

USA, 99(Suppl. 4), 16499–16506.

Chan, S.W., Zilberman, D., Xie, Z., Johansen, L.K., Carrington, J.C. and

Jacobsen, S.E. (2004) RNA silencing genes control de novo DNA methyl-

ation. Science, 303, 1336.

Daxinger, L., Kanno, T., Bucher, E., van der Winden, J., Naumann, U., Matzke,

A.J. and Matzke, M. (2009) A stepwise pathway for biogenesis of 24-nt

secondary siRNAs and spreading of DNA methylation. EMBO J. 28, 48–

57.

Deleris, A., Gallego-Bartolome, J., Bao, J., Kasschau, K.D., Carrington, J.C.

and Voinnet, O. (2006) Hierarchical action and inhibition of plant Dicer-like

proteins in antiviral defense. Science, 313, 68–71.

El-Shami, M., Pontier, D., Lahmy, S., Braun, L., Picart, C., Vega, D., Hakimi,

M.A., Jacobsen, S.E., Cooke, R. and Lagrange, T. (2007) Reiterated WG/

GW motifs form functionally and evolutionarily conserved ARGONA-

UTE-binding platforms in RNAi-related components. Genes Dev. 21,

2539–2544.

Intronic hairpin-mediated RdDM 849

ª 2009 The AuthorsJournal compilation ª 2009 Blackwell Publishing Ltd, The Plant Journal, (2009), 60, 840–851

Ernst, R.K., Bray, M., Rekosh, D. and Hammarskjold, M.L. (1997) A structured

retroviral RNA element that mediates nucleocytoplasmic export of intron-

containing RNA. Mol. Cell. Biol. 17, 135–144.

Fischer, U., Kuhlmann, M., Pecinka, M., Schmidt, R. and Mette, F. (2008) Local

DNA features affect RNA-directed transcriptional gene silencing and DNA

methylation. Plant J. 53, 1–10.

Fusaro, A.F., Matthew, L., Smith, N.A. et al. (2006) RNA interference-inducing

hairpin RNAs in plants act through the viral defence pathway. EMBO Rep. 7,

1168–1175.

Goodall, G.J. and Filipowicz, W. (1991) Different effects of intron nucleotide

composition and secondary structure on pre-mRNA splicing in monocot

and dicot plants. EMBO J. 10, 2635–2644.

Haseloff, J., Siemering, K.R., Prasher, D.C. and Hodge, S. (1997) Removal of a

cryptic intron and subcellular localization of green fluorescent protein are

required to mark transgenic Arabidopsis plants brightly. Proc. Natl. Acad.

Sci. USA, 94, 2122–2127.

He, X.J., Hsu, Y.F., Zhu, S., Wierzbicki, A.T., Pontes, O., Pikaard, C.S., Liu, H.L.,

Wang, C.S., Jin, H. and Zhu, J.K. (2009) An effector of RNA-directed DNA

methylation in arabidopsis is an ARGONAUTE 4- and RNA-binding protein.

Cell, 137, 498–508.

Henderson, I.R., Zhang, X., Lu, C., Johnson, L., Meyers, B.C., Green, P.J. and

Jacobsen, S.E. (2006) Dissecting Arabidopsis thaliana DICER function in

small RNA processing, gene silencing and DNA methylation patterning.

Nat. Genet. 38, 721–725.

Huang, L., Jones, A.M., Searle, I., Patel, K., Vogler, H., Hubner, N.C. and

Baulcombe, D.C. (2009) An atypical RNA polymerase involved in RNA

silencing shares small subunits with RNA polymerase II. Nat. Struct. Mol.

Biol. 16, 91–93.

Huettel, B., Kanno, T., Daxinger, L., Bucher, E., van der Winden, J., Matzke,

A.J. and Matzke, M. (2007) RNA-directed DNA methylation mediated by

DRD1 and Pol IVb: a versatile pathway for transcriptional gene silencing in

plants. Biochim. Biophys. Acta, 1769, 358–374.

Jones, A.L., Thomas, C.L. and Maule, A.J. (1998) De novo methylation and

co-suppression induced by a cytoplasmically replicating plant RNA virus.

EMBO J. 17, 6385–6393.

Kanno, T., Mette, M.F., Kreil, D.P., Aufsatz, W., Matzke, M. and Matzke, A.J.

(2004) Involvement of putative SNF2 chromatin remodeling protein DRD1

in RNA-directed DNA methylation. Curr. Biol. 14, 801–805.

Kanno, T., Huettel, B., Mette, M.F., Aufsatz, W., Jaligot, E., Daxinger, L., Kreil,

D.P., Matzke, M. and Matzke, A.J. (2005) Atypical RNA polymerase

subunits required for RNA-directed DNA methylation. Nat. Genet. 37, 761–

765.

Kanno, T., Bucher, E., Daxinger, L., Huettel, B., Bohmdorfer, G., Gregor, W.,

Kreil, D.P., Matzke, M. and Matzke, A.J. (2008) A structural-maintenance-

of-chromosomes hinge domain-containing protein is required for RNA-

directed DNA methylation. Nat. Genet. 40, 670–675.

Koscianska, E., Kalantidis, K., Wypijewski, K., Sadowski, J. and Tabler, M.

(2005) Analysis of RNA silencing in agroinfiltrated leaves of Nicotiana

benthamiana and Nicotiana tabacum. Plant Mol. Biol. 59, 647–661.

Li, C.F., Pontes, O., El-Shami, M., Henderson, I.R., Bernatavichute, Y.V., Chan,

S.W., Lagrange, T., Pikaard, C.S. and Jacobsen, S.E. (2006) An ARGONA-

UTE4-containing nuclear processing center colocalized with Cajal bodies in

Arabidopsis thaliana. Cell, 126, 93–106.

Lister, R., O’Malley, R.C., Tonti-Filippini, J., Gregory, B.D., Berry, C.C., Millar,

A.H. and Ecker, J.R. (2008) Highly integrated single-base resolution maps

of the epigenome in Arabidopsis. Cell, 133, 523–536.

Liu, H.X., Goodall, G.J., Kole, R. and Filipowicz, W. (1995) Effects of secondary

structure on pre-mRNA splicing: hairpins sequestering the 5¢ but not the 3¢splice site inhibit intron processing in Nicotiana plumbaginifolia. EMBO J.

14, 377–388.

Luo, C., Durgin, B., Watanabe, N. and Lam, E. (2009) Defining the functional

network of epigeentic regulators in Arabidopsis thaliana. Mol. Plant, 2, 661–

674.

Matzke, M., Kanno, T., Huettel, B., Daxinger, L. and Matzke, A.J. (2007)

Targets of RNA-directed DNA methylation. Curr. Opin. Plant Biol. 10,

512–519.

Matzke, M., Kanno, T., Daxinger, L., Huettel, B. and Matzke, A.J. (2009) RNA-

mediated chromatin-based silencing in plants. Curr. Opin. Cell Biol. 21, 1–10.

Melquist, S. and Bender, J. (2003) Transcription from an upstream promoter

controls methylation signaling from an inverted repeat of endogenous

genes in Arabidopsis. Genes Dev. 17, 2036–2047.

Mette, M.F., Aufsatz, W., van der Winden, J., Matzke, M.A. and Matzke, A.J.

(2000) Transcriptional silencing and promoter methylation triggered by

double-stranded RNA. EMBO J. 19, 5194–5201.

Mette, M.F., Matzke, A.J. and Matzke, M.A. (2001) Resistance of RNA-medi-

ated TGS to HC-Pro, a viral suppressor of PTGS, suggests alternative

pathways for dsRNA processing. Curr. Biol. 11, 1119–1123.

Mi, S., Cai, T., Hu, Y. et al. (2008) Sorting of small RNAs into Arabidopsis

argonaute complexes is directed by the 5¢ terminal nucleotide. Cell, 133,

116–127.

Mosher, R.A., Schwach, F., Studholme, D. and Baulcombe, D.C. (2008) PolIVb

influences RNA-directed DNA methylation independently of its role in

siRNA biogenesis. Proc. Natl. Acad. Sci. USA, 105, 3145–3150.

Mylne, J.S., Barrett, L., Tessadori, F., Mesnage, S., Johnson, L., Bernatavi-

chute, Y.V., Jacobsen, S.E., Fransz, P. and Dean, C. (2006) LHP1, the

Arabidopsis homologue of HETEROCHROMATIN PROTEIN1, is required

for epigenetic silencing of FLC. Proc. Natl. Acad. Sci. USA, 103, 5012–

5017.

Papp, I., Mette, F., Aufsatz, W., Daxinger, L., Schauer, S., Ray, A., van der

Winden, J., Matzke, M. and Matzke, A. (2003) Evidence for nuclear pro-

cessing of plant micro RNA and short interfering RNA precursors. Plant

Physiol. 132, 1382–1390.

Pelissier, T., Thalmeir, S., Kempe, D., Sanger, H.L. and Wassenegger, M.

(1999) Heavy de novo methylation at symmetrical and non-symmetrical

sites is a hallmark of RNA-directed DNA methylation. Nucleic Acids Res. 27,

1625–1634.

Pikaard, C.S., Haag, J.R., Ream, T. and Wierzbicki, A.T. (2008) Roles of RNA

polymerase IV in gene silencing. Trends Plant Sci. 13, 390–397.

Pontes, O., Li, C.F., Nunes, P.C., Haag, J., Ream, T., Vitins, A., Jacobsen, S.E.

and Pikaard, C.S. (2006) The Arabidopsis chromatin-modifying nuclear

siRNA pathway involves a nucleolar RNA processing center. Cell, 126, 79–

92.

Pontier, D., Yahubyan, G., Vega, D., Bulski, A., Saez-Vasquez, J., Hakimi,

M.A., Lerbs-Mache, S., Colot, V. and Lagrange, T. (2005) Reinforcement of

silencing at transposons and highly repeated sequences requires the

concerted action of two distinct RNA polymerases IV in Arabidopsis. Genes

Dev. 19, 2030–2040.

Qi, Y., He, X., Wang, X.J., Kohany, O., Jurka, J. and Hannon, G.J. (2006) Dis-

tinct catalytic and non-catalytic roles of ARGONAUTE4 in RNA-directed

DNA methylation. Nature, 443, 1008–1012.

Qian, L., Vu, M.N., Carter, M. and Wilkinson, M.F. (1992) A spliced intron

accumulates as a lariat in the nucleus of T cells. Nucleic Acids Res. 20,

5345–5350.

Sambrook, J., Fritsch, E.F. and Maniatis, T. (1989) Molecular cloning: a lab-

oratory manual, 2nd edn. New York: Cold Spring Harbor Laboratory Press.

Smith, L.M., Pontes, O., Searle, I., Yelina, N., Yousafzai, F.K., Herr, A.J., Pik-

aard, C.S. and Baulcombe, D.C. (2007) An SNF2 protein associated with

nuclear RNA silencing and the spread of a silencing signal between cells in

Arabidopsis. Plant Cell, 19, 1507–1521.

Stam, M., Viterbo, A., Mol, J. and Kooter, J. (1998) Position-dependent

methylation and transcriptional gene silencing in inverted T-DNA repeats:

implications for posttranscriptional silencing of homologous host genes in

plants. Mol. Cell. Biol. 18, 6165–6177.

Stavolone, L., Kononova, M., Pauli, S., Ragozzino, A., de Haan, P., Milligan, S.,

Lawton, K. and Hohn, T. (2003) Cestrum yellow leaf curling virus (CmYLCV)

promoter: a new strong constitutive promoter for heterologous gene

expression in a wide variety of crops. Plant Mol. Biol. 53, 663–673.

Szittya, G., Silhavy, D., Molnar, A., Havelda, Z., Lovas, A., Lakatos, L., Banfalvi,

Z. and Burgyan, J. (2003) Low temperature inhibits RNA silencing-medi-

ated defence by the control of siRNA generation. EMBO J. 22, 633–640.

Vogt, U., Pelissier, T., Putz, A., Razvi, F., Fischer, R. and Wassenegger, M.

(2004) Viroid-induced RNA silencing of GFP-viroid fusion transgenes does

not induce extensive spreading of methylation or transitive silencing. Plant

J. 38, 107–118.

Voinnet, O. (2009) Origin, biogenesis, and activity of plant microRNAs. Cell,

136, 669–687.

Wang, M.B., Helliwell, C.A., Wu, L.M., Waterhouse, P.M., Peacock, W.J. and

Dennis, E.S. (2008) Hairpin RNAs derived from RNA polymerase II and

polymerase III promoter-directed transgenes are processed differently in

plants. RNA, 14, 903–913.

Wang, X., Elling, A.A., Li, X., Li, N., Peng, Z., He, G., Sun, H., Qi, Y., Liu, X.S.

and Deng, X.W. (2009) Genome-Wide and Organ-Specific Landscapes of

850 Athanasios Dalakouras et al.

ª 2009 The AuthorsJournal compilation ª 2009 Blackwell Publishing Ltd, The Plant Journal, (2009), 60, 840–851

Epigenetic Modifications and Their Relationships to mRNA and Small RNA

Transcriptomes in Maize. Plant Cell, 21, 1053–1069.

Wassenegger, M. (2005) The role of the RNAi machinery in heterochromatin

formation. Cell, 122, 13–16.

Wassenegger, M. and Pelissier, T. (1998) A model for RNA-mediated gene

silencing in higher plants. Plant Mol. Biol. 37, 349–362.

Wassenegger, M., Heimes, S., Riedel, L. and Sanger, H.L. (1994a) RNA-di-

rected de novo methylation of genomic sequences in plants. Cell, 76, 567–

576.

Wassenegger, M., Heimes, S. and Sanger, H.L. (1994b) An infectious viroid

RNA replicon evolved from an in vitro-generated non-infectious viroid

deletion mutant via a complementary deletion in vivo. EMBO J. 13, 6172–

6177.

Wierzbicki, A.T., Haag, J.R. and Pikaard, C.S. (2008) Noncoding transcription

by RNA polymerase Pol IVb/Pol V mediates transcriptional silencing of

overlapping and adjacent genes. Cell, 135, 635–648.

Wierzbicki, A.T., Ream, T.S., Haag, J.R. and Pikaard, C.S. (2009) RNA poly-

merase V transcription guides ARGONAUTE4 to chromatin. Nat. Genet. 41,

630–634.

Zhang, X., Henderson, I.R., Lu, C., Green, P.J. and Jacobsen, S.E. (2007) Role

of RNA polymerase IV in plant small RNA metabolism. Proc. Natl. Acad.

Sci. USA, 104, 4536–4541.

Zheng, X., Zhu, J., Kapoor, A. and Zhu, J. (2007) Role of Arabidopsis AGO6 in

siRNA accumulation, DNA methylation and transcriptional gene silencing.

EMBO J. 26, 1691–1701.

Zhou, Z., Luo, M.J., Straesser, K., Katahira, J., Hurt, E. and Reed, R. (2000) The

protein Aly links pre-messenger-RNA splicing to nuclear export in meta-

zoans. Nature, 407, 401–405.

Zilberman, D., Cao, X. and Jacobsen, S.E. (2003) ARGONAUTE4 control of

locus-specific siRNA accumulation and DNA and histone methylation.

Science, 299, 716–719.

Zilberman, D., Cao, X., Johansen, L.K., Xie, Z., Carrington, J.C. and Jacobsen,

S.E. (2004) Role of Arabidopsis ARGONAUTE4 in RNA-directed DNA

methylation triggered by inverted repeats. Curr. Biol. 14, 1214–1220.

Intronic hairpin-mediated RdDM 851

ª 2009 The AuthorsJournal compilation ª 2009 Blackwell Publishing Ltd, The Plant Journal, (2009), 60, 840–851