Control of alternative splicing through siRNA-mediatedtranscriptional gene silencingMariano Allo1, Valeria Buggiano1, Juan P Fededa1, Ezequiel Petrillo1, Ignacio Schor1, Manuel de la Mata1,Eneritz Agirre2, Mireya Plass2, Eduardo Eyras2, Sherif Abou Elela3, Roscoe Klinck3, Benoit Chabot3 &Alberto R Kornblihtt1

When targeting promoter regions, small interfering RNAs (siRNAs) trigger a previously proposed pathway known astranscriptional gene silencing by promoting heterochromatin formation. Here we show that siRNAs targeting intronic or exonicsequences close to an alternative exon regulate the splicing of that exon. The effect occurred in hepatoma and HeLa cells withsiRNA antisense strands designed to enter the silencing pathway, suggesting hybridization with nascent pre-mRNA. Unexpectedly,in HeLa cells the sense strands were also effective, suggesting that an endogenous antisense transcript, detectable in HeLa butnot in hepatoma cells, acts as a target. The effect depends on Argonaute-1 and is counterbalanced by factors favoring chromatinopening or transcriptional elongation. The increase in heterochromatin marks (dimethylation at Lys9 and trimethylation at Lys27of histone H3) at the target site, the need for the heterochromatin-associated protein HP1a and the reduction in RNA polymeraseII processivity suggest a mechanism involving the kinetic coupling of transcription and alternative splicing.

Alternative splicing has a key role in generating protein diversity,affects the expression of nearly 70% of human genes and is implicatedin a wide variety of human diseases. The regulation of alternativesplicing depends on the interaction of splicing factors with regulatoryelements in the pre-mRNA, as well as on the rate and pausing oftranscriptional elongation1. The chromatin context affects RNA poly-merase II (Pol II) elongation. In mammalian cells, intragenic DNAmethylation initiates the formation of a closed chromatin structurethat reduces the efficiency of Pol II elongation2. Similarly, internalchromatin roadblocks for Pol II elongation facilitate alternative exoninclusion3, which is consistent with recent findings that intragenichistone acetylation affects Pol II processivity and alternative splicing4.

The kinetic model for coupling between alternative splicing andtranscription suggests that the rate of transcription elongation mod-ulates the outcome of two competing splicing reactions that occursimultaneously during transcription. Rapid, highly processive trans-cription favors exon skipping, whereas slower, less-processive trans-cription favors the inclusion process1.

Exogenously applied small interfering RNAs (siRNAs; 21–25nucleotides long) trigger transcriptional gene silencing (TGS) inhuman cells through heterochromatin formation at DNA targetsequences. The process involves recruitment of chromatin-modifyingenzymes, resulting in dimethylation of histone H3 Lys9 (H3K9me2),trimethylation of histone H3 Lys27 (H3K27me3), DNA methylationand histone deacetylation5–7.

Recent findings support the idea that endogenous siRNAs can alsoregulate gene expression in mammalian cells through histone mod-ifications and DNA methylation8–11. For example, the gene encodingendothelial nitric oxide synthase is epigenetically silenced by smallRNAs generated from the intron 4 of the gene; these RNAs are presentonly in cells expressing the gene (such as human aortic endothelialcells)10,11. Furthermore, the process of DNA methylation–dependentrepression of retrotransponsons has been linked to the presence ofsmall RNAs in mouse fetal male germ cells undergoing de novomethylation8. Moreover, many naturally occurring microRNAs (miR-NAs), such as miR-134, miR-296 and miR-470, target the codingsequences, including exon-exon junctions, of the genes encodingNanog, Oct4 and Sox2 in mouse embryonic stem cells12, and recentevidence indicates that human miRNAs can trigger TGS in HeLa andHEK293 cells13.

We hypothesized that siRNAs targeting gene sequences surroundingan alternative exon (that is, flanking introns) may generate a closedchromatin structure and prevent efficient Pol II elongation. This delayin elongation could, in turn, give more time to the splicing machineryfor the recognition of an alternative exon14. We report here thatsiRNAs targeting sequences located close to the fibronectin extradomain I (EDI) alternative exon can regulate its alternative pre-mRNA splicing in human cells. Knockdown of the Argonaute proteinsAGO1 and AGO2 indicates that these are needed for the effect onsplicing. The effect is abolished or reduced by factors that favor an

Received 30 October 2008; accepted 15 May 2009; published online 21 June 2009; doi:10.1038/nsmb.1620

1Laboratorio de Fisiologıa y Biologıa Molecular, Departamento de Fisiologıa, Biologıa Molecular y Celular, IFIBYNE-CONICET, Facultad de Ciencias Exactas y Naturales,Universidad de Buenos Aires, Ciudad Universitaria, Buenos Aires, Argentina. 2ICREA and Universitat Pompeu Fabra, Barcelona, Spain. 3Laboratoire de GenomiqueFonctionnelle, Faculte de Medecine et des Sciences de la Sante, Universite de Sherbrooke, Quebec, Canada. Correspondence should be addressed to A.R.K.(ark@fbmc.fcen.uba.ar).

NATURE STRUCTURAL & MOLECULAR BIOLOGY VOLUME 16 NUMBER 7 JULY 2009 7 1 7

ART IC L E S

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

open chromatin structure or increase transcriptional elongation. Themechanism involves an increase in facultative heterochromatin epi-genetic marks (H3K9me2 and H3K27me3) at the target site and theheterochromatin-associated protein HP1a.

RESULTSEffects of antisense intronic siRNAs in Hep3B cellsWe wondered whether siRNAs could affect alternative splicing of anexon such as fibronectin EDI, whose regulation by transcriptionalelongation is well documented14, through a mechanism involvingTGS. Accordingly, we designed various double-stranded siRNAstargeting intronic sequences around the EDI exon. These siRNAsare complementary to positions in introns 32 and 33 of the humanfibronectin 1 (FN1) gene (Fig. 1a and Supplementary Fig. 1a). Wedesigned the siRNAs to favor asymmetric loading by choosingsequences with much lower annealing DG values for the 5¢ end ofthe strand intended to enter the silencing pathway (SupplementaryFig. 1b)15. We transfected Hep3B cells with these siRNAs or with aluciferase-specific siRNA (siLuc) as a negative control. We mea-sured total fibronectin mRNA levels and relative abundances of EDIisoforms 72 h after transfection by quantitative real-time RT-PCR.Two different siRNAs targeting intron 33 sense sequences (I33*asand I33as), but not siRNAs targeting intron 32 (I32as) or theantisense strand of intron 33 (I33s), affected alternative splicing bypromoting a substantially greater inclusion-to-exclusion ratio ofEDI compared to the siLuc control (Fig. 1b). As a positivecontrol for efficient siRNA transfection, we examined fibronectinmRNA degradation by the canonical RNA interference (RNAi)pathway when an exonic siRNA (E34as) was transfected inthe same experiment (see below). We obtained similar resultsusing 32P-labeled radioactive RT-PCR (Supplementary Fig. 2a).We decided to pursue further experiments with I33as, given itsgreater effect on splicing.

We next investigated whether the classical post-transcriptionalgene silencing (PTGS) pathway could be affecting pre-mRNA levelsand selectively favoring the degradation of the exclusion iso-form through an unknown mechanism. We measured total fibronectinmature mRNA levels by quantitative real-time RT-PCR and observedthat they were not affected by transfection with I33as (Fig. 1c, belowleft) or any of the other intronic siRNAs tested (SupplementaryFig. 2b). Similarly, quantitative real-time RT-PCR with primersflanking the I33as target site showed that I33as does not causecleavage of fibronectin pre-mRNA (Fig. 1c, below right), confirmingthat PTGS is not responsible for the observed effect on alter-native splicing.

Sequence specificity of the intronic siRNA effectsGiven the still-limited number of TGS examples in mammalian cells, itbecame crucial to rule out putative off-target effects of I33as transfec-tion. We used a strategy different from the asymmetry rule topreferentially direct one of the siRNA strands to the silencing complex.Accordingly, similar results were obtained using Stealth (ST; Invitro-gen) siRNAs chemically modified to favor the entry of one or theother strand to the silencing complex. I33as(ST) had the same effecton EDI alternative splicing as did I33as (designed as above; Fig. 1d).Consistently, I33s(ST) had no effect. This control is also importantbecause although I33as and I33s map to the same target region, theyare not totally complementary, whereas I33as(ST) and I33s(ST) arefully complementary (Supplementary Fig. 1a). Although the exactnature of the RNA chemical modifications in the Stealth oligonucleo-tides are proprietary, the Stealth strategy has been validated bycotransfections of mammalian cell lines with Stealth duplexes designedto favor the entry of the antisense strand into the silencing complexalong with lacZ reporters expressed from an eukaryotic promoter, ineither the forward or reverse orientation. Under these conditions,b-galactosidase activity was inhibited by 90% when the vector was

Promoter Exon 1 Exon 21

E21as

Hep3B

Hep3B

siLuc l32as

siLuc

l33as

l33as

(ST)

l33s(

ST)

SCR l33asl33*as

**

**

** **

l33as l33s

ED

l+/E

Dl–

ED

l+/E

Dl–

ED

l+/E

Dl–

ED

l+/E

Dl–

FN

1 To

tal m

RN

A

Targ

et p

re-m

RN

A

1.25

2.51.25

siLuc l33as siLuc l33as

1.00

0.75

0.50

0.25

1.25

1.00

0.75

0.50

0.25

2.0

1.5

1.0

0.5

1.00

0.75

0.50

0.25

0

0

l32as l33*as l33as

l33s

E34as

Exon 32 EDI(Exon 33)

Exon 34B C D

Exon 35 Exon 36

EDI P1

P3 P4l33as

P2

A

2.5

2.0

**

1.5

1.0

0.5

0

2.01.61.20.8

5 nM

** ** **

10 nM 50 nM

0.4

0

a

c

d

e

f

b

Figure 1 Intronic siRNAs affect fibronectin alternative splicing in Hep3B cells. (a) Endogenous human FN1 gene regions studied in the present work. The

target sites for double-stranded siRNAs whose antisense (as) or sense (s) strands are designed to preferentially enter the small RNA pathway are indicated15.

EDI, alternatively spliced extra domain I exon; A–D, amplicons used for native ChIP analysis. (b) Effects of intronic siRNAs in Hep3B cells on endogenous

EDI alternative splicing. Ratios were measured in triplicate for four independent experiments. **Po 0.01. Error bars indicate s.d. (c) Above, primers P1 and

P2 were used for mature mRNA RT-PCR; P3 and P4 were used for pre-mRNA containing the target site of I33as. Below, I33as siRNA does not alter the

levels of fibronectin mature mRNA (left) or pre-mRNA (right). Results represent four experiments with three independent samples per experiment. (d) Effectsof Stealth (ST) siRNAs, chemically modified to favor the entrance of one or the other strand to the silencing complex, on EDI alternative splicing. I33as(ST)

has the same effect as I33as. (e) An siRNA with a scrambled I33as sequence (SCR) has no effect on EDI alternative splicing. (f) Dose-response of I33as

(black) or siLuc control (white) on EDI alternative splicing. RNAs were quantified by quantitative real-time RT-PCR (b, c and f) or radioactive RT-PCR (d and

e). Horizontal dashed lines in b, d and f indicate mean value of the siLuc negative control.

ART IC L E S

71 8 VOLUME 16 NUMBER 7 JULY 2009 NATURE STRUCTURAL & MOLECULAR BIOLOGY

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

expressed in the sense orientation, but by only 10% when expressed inthe reverse orientation (K. Wiedrholt, personal communication).

An siRNA with a scrambled I33as sequence had no effect on EDIalternative splicing (Fig. 1e). We observed the change in EDI alternativesplicing with I33as concentrations as low as 5 nM, and its magnituderemained constant up to 50 nM (Fig. 1f), providing additionalevidence against off-target or indirect effects. Previous studies haveindicated that the effects of RNAi are not entirely specific and that off-target effects are more prevalent with siRNA concentrations above50 nM. The effect we observed was insensitive to increasing concen-trations of the siRNA, unlike other off-target effects16,17.

We confirmed the regional specificity of the TGS effect within theFN1 gene by the lack of effect of I32as (Fig. 1b) and by showing thatalternative splicing patterns of another fibronectin alternative exon,IIICS, were not altered by I33as transfection (Supplementary Fig. 3a).The gene specificity was supported by the lack of effect on sevenindependent alternative splicing events (ASEs) corresponding to sevendifferent genes randomly chosen from the cancer-related RT-PCRpanel used below (Supplementary Fig. 3b).

Both sense and antisense intronic siRNAs act in HeLa cellsTo assess the cell specificity of the I33as effect, we transfected HeLa cellswith the same siRNAs. As in Hep3B cells, I33as induced an increase inEDI exon inclusion (Fig. 2a) that was also gene specific (Supplemen-tary Fig. 3c). In HeLa cells, we unexpectedly observed that the sensestrand, I33s, stimulated EDI inclusion (Fig. 2a). Only the antisensestrands of siRNAs targeting the HIV-1 promoter and those of the genesencoding EF52 and EF1A can trigger histone modifications and TGSinduction in human 293T cells18. Moreover, in the case of EF1A,transcription of the promoter sequences is required for TGS. Thisstrongly suggests that RNA-RNA interactions between the siRNA andnascent transcripts are required to trigger TGS19.

The effect of I33s could therefore be explained if an antisensefibronectin transcript, able to hybridize with I33s, existed in HeLacells. Using quantitative PCR with a sense primer for reverse trans-cription (P5), we detected an antisense RNA encompassing the targetsite of I33s in both HeLa and Hep3B cells. In agreement with ourprediction, the proportion of this antisense RNA with respect to the

sense pre-mRNA was 20 times higher in HeLa than in Hep3B cells(Fig. 2b). Most notably, an Ensembl database search revealedthe existence of an antisense expressed sequence tag (EST;ENSESTT00000078687) whose precursor transcript would span atleast one-third of the FN1 gene, including the alternative splicingregions EDI and EDII (Fig. 2c), and the target site of I33s. Becausethere is partial overlap between exons 2 and 3 of the EST and exons 22and 26 of the FN1 gene, we used primers mapping to the overlappingsequences to simultaneously amplify the antisense processed RNA(asFN1) and mature fibronectin mRNA sequences by RT-PCR. InHep3B cells, the asFN1 PCR product was undetectable compared tothe mRNA product; in HeLa cells, both PCR products were almostequally abundant (Fig. 2d), explaining why the sense and antisensesiRNAs were equally effective in these cells.

The siRNA effect needs AGO1 and involves Pol II elongationPrevious work has shown that human AGO2 is necessary for the TGSand PTGS pathways in T47D and HeLa cells, but AGO1 seems to beparticularly linked to TGS5,20,21, being associated with Pol II andrecruited to the DNA target region5,7. Consistent with this previouswork, knockdown of AGO1 or AGO2 abolished the effect of I33as onsplicing, indicating that the TGS pathway is involved (Fig. 3a).

Given that Pol II elongation has been implicated in alternativesplicing, we examined the activity of I33as on minigenes with eithera constitutive promoter (Fig. 3b, left) or a tetracycline-induciblepromoter whose transcription can be activated by the transactiva-tors Sp1 (which promotes initiation) or VP16 (which promotesinitiation and elongation)22. The use of transiently transfectedminigenes as TGS templates was based on evidence showing thatthey assemble in a physiological chromatin context23 with typicalpatterns of dynamic epigenetic marks24,25. When transcription offibronectin minigenes was transactivated by Sp1, I33as increasedEDI inclusion by four-fold. This effect was significantly decreasedwhen the VP16 transactivator was used (Fig. 3b, center). SW6,a VP16 mutant with point mutations in four Phe residues,stimulates transcriptional initiation but is defective in supportingelongation26. Activation by wild-type VP16 abolished the effect ofI33as on alternative splicing, whereas activation by SW6 did not

7.25 2.5

2.0

1.5

1.0

0.5

HeLa

siLuc l32as l33*as

EDll

EDl

Hep3B

FN1Gene

pre-mRNA

617 bp FN1asFN1

Mature RNAs

172 bpExon 22

MHep

3B

HeL

a

Exon 23

mRNA

Exon 24

P7 P8617 bp

Exon 24 Exon 26

(Exon 25) Exon 26

EDll

EDll

EDl lllCS

HeLaP5

**Precursor RNA

Ant

isen

se/s

ense

P3 P4l33as

172 bp

E1E2E3

E3 E2 E1Processed asFN1 RNA

(ENSESTT00000078687)

Unprocessed asFN1 RNA

P7 P8

P6

EDl lllCSFN1

l33as l33s

** ****

ED

l+/E

Dl– 5.00

3.75

2.50

1.25

0 0

a b

c d

Figure 2 Effects of intronic siRNAs on EDI

alternative splicing of the endogenous FN1 gene

in HeLa cells. (a) Unlike in Hep3B cells, I33sin HeLa cells stimulates EDI exon inclusion.

**P o 0.01. Error bars indicate s.d. Horizontal

dashed line indicates mean value of the siLuc

negative control. (b) HeLa cells contain a higher

ratio of antisense to sense RNA at the I33as

target site. RNAs were detected by quantitative real-time RT-PCR using primers P5 and P6 to reverse-transcribe antisense and sense RNAs, respectively, and

P3 and P4 for both PCR reactions. RNAs in a and b were quantified by quantitative real-time RT-PCR. (c) Human FN1 gene. Indicated are the antisense

EST (above), the position of its unspliced precursor (above middle), a portion of the FN1 pre-mRNA (below middle) and corresponding spliced mRNA

spanning exons 22–26 (below). (d) Radioactive RT-PCR of mature sense and antisense fibronectin RNAs present in Hep3B and HeLa cells.

ART IC L E S

NATURE STRUCTURAL & MOLECULAR BIOLOGY VOLUME 16 NUMBER 7 JULY 2009 7 1 9

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

(Fig. 3b, right), confirming that it is the elongation activity of VP16that reduces the effects of I33as on alternative splicing.

Measuring changes in Pol II elongation or processivity on a givengene and under a particular stimulus is a technically difficult task. Anindirect way to assess Pol II processivity is to measure the abundanceof distal versus proximal pre-mRNAs in a given region, with theassumption that most pre-mRNA is a cotranscriptional intermedi-ate14. To do this, we isolated RNA from nuclei of cells transfected withthe tetracycline-inducible minigene, activated by either Sp1 or VP16and cotransfected with I33as or the siLuc control. We determined theamounts of minigene pre-mRNAs accumulated from proximal anddistal regions with respect to the transcription start site by quantitativereal-time RT-PCR. We synthesized cDNA with a different primer foreach region, then amplified PCR regions P and D. I33as caused2.8-fold higher accumulation of proximal over distal pre-mRNAscompared to the siLuc control, but only when transcriptionwas activated by Sp1 (Fig. 3c). These results are consistent withreduced Pol II processivity upon I33as transfection when transcriptionis activated by Sp1 but not by VP16. At the same time, thisevidence must be taken with the caveat that the chromatin status

acquired by the transfected minigenes does not necessarily reflectthat of the endogenous chromosomal gene.

The siRNA effect, histone modifications and DNA methylationThe need for AGO1 and for antisense transcription in HeLa cells washighly supportive of a mechanism similar to TGS underlying thesiRNA effects on alternative splicing. To further support this, wereasoned that if I33as promotes alterations in chromatin, treatmentsthat prevent these specific modifications should reduce or abolish theI33as effect. Confirming this prediction, the effect was reversed bytreatment with the histone deacetylase inhibitor trichostatin A and theDNA methyltransferase inhibitor 5-azadeoxycytidine (Fig. 4a). More-over, the histone methyltransferase inhibitor BIX 01294, which isspecific for H3K9me2 (ref. 27), an epigenetic mark characteristic of

1.00a

c

b0.75

ED

I+/E

DI–

P/D

ED

l+/E

Dl–

ED

l+/E

Dl–

0.50

siRNA (50 nM) siLuc

100 ± 5

100 ± 2

7.6 ± 0.5 105 ± 3

8.5 ± 1

99 ± 4

102 ± 3 125 ± 6

5.9 ± 0.7 103 ± 3

10 ± 2120 ± 7

siLuc siLucsiAGO1 siAGO2 siAGO1 siAGO2siLuc siLuc l33as l33as l33as

siRNA (10 nM)

0.25

0

AGO1 mRNA relative levels

2.0

1.6

1.2

0.8

0.4

SP1 siLuc l33as SP1 VP16 SW6 VP16VP16

**0.60

0.6**

**

**

**

**

0.4

0.2 ED

l+/E

Dl–

0.6

EDI+

HIV–Gal4

EDI

EDI–

EDI+

EDI–

EDI+

FN1 promoter PhCMV–1 pUHC–EDA DP

tet operators7X

EDIEDl

EDI–

0.4

0.2

00

0.45

0.30

0.15

00

AGO2 mRNA relative levels

Figure 3 Factors influencing the siRNA effect on EDI alternative splicing. (a) Above, knockdown of either AGO1 or AGO2 is equally effective in abolishing

the effect of I33as on EDI alternative splicing. Below, AGO1 and AGO2 knockdown was controlled by quantifying their respective mRNAs using quantitative

real-time RT-PCR. Data represent mean ± s.d. of four independent experiments. Horizontal dashed line indicates mean value of the siLuc negative control.

(b) Effects of I33as siRNA (black bars) or siLuc control (white bars) on alternative splicing of the fibronectin EDI exon in Hep3B cells, expressed from

transfected reporter minigenes under the control of various promoters and transcriptional activators. Left, fibronectin promoter (pSVEDA/FN; ref. 44). Center,

tetracycline-inducible promoter (pUHC-EDA) transactivated by tTA-Sp1 or tTA-VP16 (ref. 14). Right, Gal4-HIV promoter (pSVEDA/Gal5-HIV-2) transactivated

by Gal4-SW6 or Gal4-VP16 (ref. 45). EDI+/EDI- ratios were determined using radioactive RT-PCR. Data represent means ± s.d. from three independent

experiments. (c) I33as affects Pol II processivity, producing an accumulation of pre-mRNA at proximal regions (P) with respect to the transcription start site

against distal regions (D), as measured by quantitative real-time RT-PCR. Hep3B cells were transfected with the minigene pUHC-EDA activated with tTA-Sp1

or tTA-VP16 and I33as (black) or siLuc (white) as control14. DNA amounts resulting from contamination of RNA preparations with transfected template

constructs were estimated in reactions in which reverse transcriptase was omitted and subtracted from total templates calculated by real time RT-PCR.

**P o 0.01. Error bars indicate s.d.

10.00a c

b

**

**

**

7.50

ED

I+/E

DI–

5.00

2.50

0

0

10.00

1.4

1.2

1.0

0.8

0.6

0.4

0.2

siRNA (50 nM)

siRNA (10 nM)

HP1 mRNArelative levels

siLuc siLuc I33as I33as

siLuc HP1α siLuc HP1α

7.50

ED

I+/E

DI–

ED

I+/E

DI–

5.00

2.50

0

Contro

lTSA

Contro

lBIX

5aza

dC TSA

+ 5a

zadC

95 ± 7 20 ± 2 110 ± 15 23 ± 3

Figure 4 Importance of the chromatin state in the TGS-AS pathway. (a) The

effect of I33as (black) or siLuc control (white) on EDI alternative splicing is

reduced by a histone deacetylase inhibitor (TSA) and completely abolished

by a DNA methyltransferase inhibitor (5azadC), indicating that these

chromatin-modifying enzymes are required for the effect on alternativesplicing. (b) An inhibitor of histone methyltransferase (BIX)26 specific for

H3K9me2 modification suppresses I33as splicing alteration. Results of

transfections with siLuc (white) and I33as (black) are shown. (c) Hep3B

cells were transfected with siLuc or I33as in cells depleted of HP1a.

Knockdown of HP1a also abolishes the effect of I33as. RNAs in a–c were

quantified by quantitative real-time RT-PCR. Data in a–c show means ± s.d.

of triplicate repeats. **P o 0.01.

ART IC L E S

72 0 VOLUME 16 NUMBER 7 JULY 2009 NATURE STRUCTURAL & MOLECULAR BIOLOGY

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

facultative heterochromatin, also reversed the effects of I33as onsplicing (Fig. 4b). Consistently, knockdown of HP1a, which is presentexclusively in heterochromatin, also abolished the I33as effect(Fig. 4c). SWI6, a homolog of HP1, is a key factor in the well-characterized TGS mechanism in yeast28,29. Taken together, these datasupport the importance of the chromatin context for the observedeffect on alternative splicing.

I33as siRNA promotes heterochromatin epigenetic marksTo investigate whether histone modifications involved in facultativeheterochromatin formation are present upon siRNA transfection7,13,18,we analyzed H3K9me2 and H3K27me3 levels in four regions of theendogenous FN1 gene by native chromatin immunoprecipitation(nChIP)30 in HeLa cells transfected with I33as. I33as caused noenrichment for H3K9me2, relative to the siLuc control, either nearthe promoter (region A in Fig. 1a) or further downstream of intron 33(region D); however, we observed a two- to four-fold enrichment ofH3K9me2 closer to the I33as target sequence (regions B and C;Fig. 5a). Moreover, we found an approximately three-fold enrichmentof H3K27me3 at region B that spread downstream over regions C andD (Fig. 5b). These results, together with the effect of HP1a depletion(Fig. 4c), show that I33as leads to modification of the chromatincontext of DNA target sequences located inside the coding region ofactive genes and that this modification is associated with the ability ofI33as to modulate alternative splicing of the nascent transcripts.

Intron- versus exon-targeted siRNAsTo determine whether TGS regulation of alternative splicing isrestricted to intronic siRNAs, we investigated the effects of exonicsiRNAs on EDI splicing targeting exon 21 (E21as) and exon 34(E34as). Both exonic siRNAs decreased fibronectin mature mRNAlevels by more than 80% (Fig. 5c), as expected for a classical siRNAknockdown, but only E34as concomitantly increased inclusion of EDI(Fig. 5d). We reasoned that the differential behavior of E21as andE34as was due to the fact that, similar to I33as, E34as maps closer tothe EDI exon in a region where TGS-like epigenetic modificationscould affect alternative splicing. To measure the effect of the putativeTGS pathway in the presence of an active PTGS pathway, wetransfected E34as together with a tetracycline-inducible minigene atvarious times22. Forty-eight hours after transfection, there was no

detectable change in mRNA levels, but there was a considerable effecton alternative splicing. At 72 h, the mRNA was mostly degraded, butthe effect on splicing was still observed on the remaining mRNA(Supplementary Fig. 4). This argues that E34as acts on splicingindependently of its ability to promote mRNA degradation.

Depletion of AGO1 or Dicer affects many ASEsIn view of the similarity of the above results to previously definedcharacteristics of TGS, we propose referring to this new pathway astranscriptional gene silencing–regulated alternative splicing (TGS-AS).The physiological relevance of TGS has been highlighted by recentfindings that endogenous siRNAs and miRNAs can regulate geneexpression in mammalian cells through histone modifications andDNA methylation8–11. We sought to examine whether these or othersmall RNAs can act as the raw material for an endogenous TGS-ASpathway. Notably, we found that 137,108 of 294,058 endogenous shorttranscribed RNAs previously identified in a HeLa cell database31

overlap intronic regions, of which 37,727 flank ASEs and 9,713flank alternative cassette exons, such as EDI. In contrast, our bioinfor-matics search revealed that 289 miRNA precursors (from a total of 689miRNAs in the UCSC database (http://genome.ucsc.edu; wgRNA tablefrom sno/miRNA track) are located in introns, of which 110 areflanking ASEs. This figure would be much higher if we took intoaccount recent evidence from 293T and HeLa cells that miRNAs couldtrigger TGS even with a few mismatches with their target sequences13.

Unfortunately, the generality of TGS-AS can only be investigated bythe use of sequence-specific siRNAs for various alternative splicingregions of candidate genes, analyzed individually. To overcome thisdifficulty, we speculated that if some of the above interactionsmodulate alternative splicing endogenously through a mechanismsimilar to TGS, an AGO1 knockdown would affect them. To testthis, we used an automated high-throughput RT-PCR platform toscreen 96 cancer-related ASEs32, in cells transfected with siLuc orsiAGO1. Isoform patterns were significantly affected by AGO1 knock-down in 35% and 37% of the ASEs in Hep3B and HeLa cells,respectively (Supplementary Table 1), providing initial evidence formore widespread regulation by TGS-AS.

Additionally, we reasoned that if any miRNA or endogenoussiRNA10,11,33–37 affects alternative splicing through TGS, then knock-down of Dicer (DCR) should affect alternative splicing patterns in thesame way. Using the same RT-PCR panel, we investigated how manyof the AGO1-dependent ASEs are also affected by DCR knockdown.We found that 18% of siAGO1-modified ASEs in HeLa cells and 29%in Hep3B cells changed their patterns in a similar way after DCRdepletion (Supplementary Table 2).

We next carried out a detailed analysis of the distribution of ASEsaffected or not by AGO1 and DCR knockdown (Fig. 6). Approxi-mately 53% and 51% of the ASEs showed no change upon knock-down of AGO1 or DCR in Hep3B and HeLa cells, respectively. TheASEs that were affected by AGO1 depletion only (18% in Hep3B and

5.00

a b

c d

**

** **

** **

**

**

H3K9me2 H3K27me3

3.75

2.50

1.25

0Amplicon

Rel

ativ

e en

richm

ent

5.00

3.75

2.50

1.25

0

Rel

ativ

e en

richm

ent

1.25

0.75

1.00

0.50

0.25

0

siLuc

E21as

FN

1 To

tal m

RN

A

0.75

1.00

0.50

0.25

0

0.75

1.00

0.50

0.25

0

ED

I+/E

DI–

1.25

0.75

1.00

0.50

0.25

0

siLuc

E34as

siLuc

E21as

siLuc

E34as

A B C D AAmplicon B C D

Figure 5 I33as transfection promotes histone methylation at the gene target

region. (a,b) Native ChIP analysis showing that I33as siRNA increases the

levels of H3K9me2 (a) and H3K27me3 (b) at regions downstream of its

target site, but not at exon 1, near the promoter. Cells were transfected

with siLuc (white) or I33as (black). Error bars represent s.e.m. Three

independent immunoprecipitation replicates were conducted per experiment.

(c) siRNAs targeting exons 21 and 34 downregulate total fibronectin mRNA

levels through PTGS. (d) Only an siRNA targeting exon 34 alters EDI

alternative splicing. RNAs in c and d were quantified by quantitative

real-time RT-PCR. Data represent means ± s.d. of two independent

experiments. **P o 0.01.

ART IC L E S

NATURE STRUCTURAL & MOLECULAR BIOLOGY VOLUME 16 NUMBER 7 JULY 2009 7 2 1

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

27% in HeLa) might reflect Dicer-independent pathways of TGS. TheASEs that were affected by DCR depletion only (12% in both Hep3Band HeLa) might reflect PTGS regulation of alternative splicingmRNA isoforms by miRNAs, a process that would be independentfrom AGO1. If changes in alternative splicing patterns upon AGO1depletion are caused by a TGS-AS mechanism, one would expect tosee similarities in the exon and intron organization around theaffected ASEs, reflecting perhaps similar constraints for sensitivity toPol II elongation. Indeed, introns located upstream of events that wereaffected by AGO1 depletion were significantly shorter than those ofnegative events in both HeLa (P¼0.0268) and Hep3B (P¼0.0201)cells. Upstream introns were also significantly shorter when consider-ing the HeLa and Hep3B data sets together (P ¼ 0.0064; Supplemen-tary Fig. 5a). We found no significant differences in the lengths of thedownstream introns between positive and negative events in HeLa orHep3B cells (Supplementary Fig. 5b). When we considered the ratiosbetween the lengths of the upstream and downstream introns for eachevent (Supplementary Fig. 5c), we found a significant differencebetween positive and negative events, but only for the set of events inHep3B cells (P¼0.0290). The similarity between intron and exonstructural features among the AGO1-sensitive ASEs is highly sugges-tive of a common alternative splicing regulatory mechanism.

DISCUSSIONWe show here that transfection of mammalian cells with siRNAstargeting a gene region located near an alternative exon can affectalternative splicing of that exon. We show that the underlying

mechanism does not involve classical PTGS,but is more similar to the recently character-ized genomic effects of siRNAs (TGS) inmammals, which silence gene expression atthe transcriptional level by a mechanism in-volving increased levels of facultative hetero-chromatin marks near the target sequence.Several of our experiments support a TGSmechanism and a subsequent reduction inelongation. First, the effect depends on thepresence of AGO1 (Fig. 3a), the Argonauteprotein shown to be necessary for TGS5.Second, the effect is reduced or abolishedby treatments that promote chromatinrelaxation (Fig. 4a) or increase Pol II elonga-tion (Fig. 3b). Third, siRNA transfectioncauses H3K9 dimethylation and H3K27 tri-methylation at the endogenous target gene(Fig. 5)—two epigenetic marks typicallyassociated with heterochromatin and genesilencing. It is notable that the silencingmarks are in the vicinity of the siRNA target

sites, well inside the FN1 gene, and not on its promoter. Previousreports of TGS in mammalian cells have been almost exclusivelyrestricted to promoters5–7,18,19. In contrast, our results indicate thatsiRNAs targeting internal gene regions can generate silencing markswithin the gene body that do not affect promoter activity but inhibitinternal elongation, subsequently affecting splice site selection (Fig. 7).The findings in Figure 3c support the elongation mechanism. We areaware of the limitations of the proximal-distal pre-mRNA accumula-tion approach used in those experiments, but the fact that the effectwas observed when transcription was stimulated at initiation, but notat elongation (Fig. 3b), reinforces the evidence. Further support for arole for elongation comes from the observation that the ASEs that

Hep3Ba b c d

35%

29%

37%

27%

51%

10%

12% 22%

18% 17%

12%

53%

No change∆φ siAGO1 only

∆φ siAGO1/siDCR

∆φ siDCR only

∆φ siDCR

∆φ siAGO1

No change∆φ siAGO1 only

∆φ siAGO1/siDCR

∆φ siDCR only

∆φ siDCR

∆φ siAGO1

Hep3B

120

100

Tota

l mR

NA

80

60

40**

**

**

**

20

AGO1 DCR0

120

100

Tota

l mR

NA

80

60

40

20

AGO1 DCR0

HeLa HeLa

Figure 6 Depletion of AGO1 and DCR affects many ASEs. We used an automated high-throughput RT-

PCR platform to screen 96 cancer-related ASEs. (a–d) Distribution of DC (percent of genes from the

panel) significantly affected or unaffected by AGO1 and/or DCR knockdown are shown for Hep3B (a)

and HeLa (c) cells. Total mRNA levels of AGO1 and Dicer (DCR) were measured for quantitation of

knockdown efficiency (b and d). Transfections with siLuc (white bars), siAGO1 (yellow bars) and siDCR

(green bars) are shown. **P o 0.01. Error bars indicate s.d.

Fast

HMT DMTHDAC

Inclusion

Splicing machineryH3K9ac

H3K9me2/H3K27me3

Constitutive exon

Alternative exon

SlowPol ll HP1α HP1α

FastPol ll

Pol ll

Exclusion siRNA

AGO1

AG

O1A

GO

1

a

b

c

Figure 7 Model for TGS-AS. (a) In the absence of siRNAs, a relaxed

nucleosomal organization characterized by acetylated H3K9 (H3K9ac;

yellow circles) allows for fast Pol II elongation and exclusion of the

alternative exon (red). (b) Transfection with siRNAs targeting the intron

downstream of the alternative exon promotes dimethylation andtrimethylation of H3K9 and H3K27, respectively (green circles), by histone

methyltransferases (HMT), probably after DNA methyltransferases (DMT) and

histone deacetylases (HDAC) have acted. (c) Methylation of H3K9 and

H3K27 are propagated, HP1a is recruited and the resulting condensed

chromatin structure slows down Pol II elongation, causing higher inclusion

of the alternative exon. The participation of HMT, DMT, HDAC and HP1a is

inferred from experiments in Figure 4.

ART IC L E S

72 2 VOLUME 16 NUMBER 7 JULY 2009 NATURE STRUCTURAL & MOLECULAR BIOLOGY

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

were affected by AGO1 depletion in the alternative splicing platformshare the common feature of shorter upstream introns (Supplemen-tary Fig. 5). We speculate that in AGO1-depleted cells, the ASEs withrelatively short upstream introns were more sensitive to Pol IIelongation changes in response to epigenetic modifications. Never-theless, explanations other than an effect on elongation remainpossible. For example, an internal silencing mark could affect cotran-scriptional alternative splicing by promoting local recruitment ofsplicing factors.

TGS-AS was triggered not only by intronic siRNAs but also byexonic siRNAs such as E34as (Fig. 5). We interpreted the effects ofE34as as a combined action of TGS and PTGS. The proximity of theE34as target site to the EDI exon affects its inclusion by TGS-AS,similar to the mode of action of I33as. This nuclear event might befollowed by export of both mRNA isoforms (EDI+ and EDI–) to thecytoplasm, where E34as triggers their degradation through PTGS. Theabsence of pre-mRNA degradation by the intron-targeted siRNA isconsistent with this idea.

The mechanism by which siRNA application leads to formation oflocal heterochromatin marks is unclear. The need for RNA-RNAhybridization between the small RNA guide strand embedded in thesilencing complex and a nascent transcript has been proposed6. Insupport of this, we found that in Hep3B cells, only I33as affectsalternative splicing, whereas both I33as and I33s are effective in HeLacells. This differential behavior can be explained if one assumes that theeffect of I33as involves hybridization with a nascent sense fibronectinpre-mRNA present in both cell types, whereas the effect of I33s is theresult of hybridization with an antisense nascent RNA whose existenceis shown in Figure 2. Consistently, the relative abundance of thisantisense transcript is much higher in HeLa than in Hep3B cells. Thepresence of an antisense EST (Fig. 2c) covering the affected regionsupports the previous assumption and provides further independentand unpredicted support for a TGS-mediated mechanism38,39.

The use of the cancer-related panel for alternative splicing in cellsdepleted of AGO1 and Dicer sheds light on a possible endogenousTGS-AS pathway (Fig. 6). AGO1 has been implicated in TGS and isbelieved to guide the small RNA to its genomic target region,triggering the chromatin modifications. In contrast, the role ofDicer in chromatin structure modification was observed in a previousstudy in which upregulation of b-globin intergenic transcripts wasfound in cells depleted of Dicer, accompanied by chromatin relaxationand histone tail modifications40. In this context, AGO1 is a strongcandidate to mediate TGS-AS, and Dicer is a key enzyme to generateat least some of the physiological effectors of TGS-AS. In regard to thebest candidates for physiological effectors, the results of our panelscreen suggest that the effectors are not limited to double-strandedsmall RNAs generated by Dicer, such as endo-siRNAs, as some ASEsare AGO1 dependent and Dicer independent. Indeed, RNAs fromdifferent genomic sources have been reported to trigger transcriptionalgene silencing, including noncoding RNAs, sense-antisense transcrip-tion couples, repetitive sequences and other small RNA families41,42.Among these, some noncoding RNAs43 and PIWI small RNAs affectchromatin structure independent of Dicer, which could explain thepanel screen results.

In summary, our results reveal a link between alternative splicingand the RNAi pathway and define a mechanism for the regulation ofalternative splicing. Our data also suggest the presence of two differentand independent silencing mechanisms that, acting in the cell simul-taneously, may affect the relative abundance of protein isoforms: thecytoplasmic PTGS pathway and a nuclear TGS pathway, thelatter affecting not only transcription but also pre-mRNA processing.

Moreover, because mutations that affect alternative splicing patternsare frequent causes of hereditary disease and cancer, intronic siRNAsmay represent a new therapeutic tool to correct gene- and exon-specific alternative splicing defects through TGS-AS.

METHODSMethods and any associated references are available in the onlineversion of the paper at http://www.nature.com/nsmb/.

Note: Supplementary information is available on the Nature Structural & MolecularBiology website.

ACKNOWLEDGMENTSWe thank P. Bertucci, M. Blaustein, F. Pelisch, M. Munoz, A. Srebrow, G. Risso,L.G. Acuna, M.G. Herz, N. Tilgner, I. Listerman, M. Buhler, J. Martınez,K. Neugebauer and K. Morris for their support and useful discussions; andBoehringer Ingelheim Pharmaceuticals (Biomolecular Screening, Department ofMedicinal Chemistry) for the gift of BIX-01294. This work was supported bygrants to A.R.K. from the Fundacion Antorchas, the Agencia Nacional dePromocion de Ciencia y Tecnologıa of Argentina, the University of Buenos Airesand the European Alternative Splicing Network. S.A.E., R.K. and B.C.acknowledge support from Genome Canada and Genome Quebec. M.A. is therecipient of a fellowship and A.R.K. is a career investigator from the ConsejoNacional de Investigaciones Cientıficas y Tecnicas of Argentina. B.C. holds aCanada Research Chair in functional genomics. A.R.K. is an international researchscholar of the Howard Hughes Medical Institute.

AUTHOR CONTRIBUTIONSM.A. proposed, designed and conducted most of the experiments and preparedthe manuscript; V.B., J.P.F., E.P., I.S. and M.d.l.M. conducted some experiments;E.A., M.P. and E.E. conducted the bioinformatics analysis; S.A.E., R.K. and B.C.conducted the PCR panel experiment; and A.R.K. coordinated the work andprepared the manuscript.

Published online at http://www.nature.com/nsmb/

Reprints and permissions information is available online at http://npg.nature.com/

reprintsandpermissions/

1. Kornblihtt, A.R. Chromatin, transcript elongation and alternative splicing. Nat. Struct.Mol. Biol. 13, 5–7 (2006).

2. Lorincz, M.C., Dickerson, D.R., Schmitt, M. & Groudine, M. Intragenic DNA methyla-tion alters chromatin structure and elongation efficiency in mammalian cells. Nat.Struct. Mol. Biol. 11, 1068–1075 (2004).

3. Batsche, E., Yaniv, M. & Muchardt, C. The human SWI/SNF subunit Brm is a regulatorof alternative splicing. Nat. Struct. Mol. Biol. 13, 22–29 (2006).

4. Schor, I.E., Rascovan, N., Pelisch, F., Allo, M. & Kornblihtt, A.R. Neuronal celldepolarization induces intragenic chromatin modifications affecting NCAM alternativesplicing. Proc. Natl. Acad. Sci. USA 106, 4325–4330 (2009).

5. Kim, D.H., Villeneuve, L.M., Morris, K.V. & Rossi, J.J. Argonaute-1 directs siRNA-mediated transcriptional gene silencing in human cells. Nat. Struct. Mol. Biol. 13,793–797 (2006).

6. Morris, K.V., Chan, S.W., Jacobsen, S.E. & Looney, D.J. Small interfering RNA-inducedtranscriptional gene silencing in human cells. Science 305, 1289–1292 (2004).

7. Suzuki, K. et al. Closed chromatin architecture is induced by an RNA duplex targetingthe HIV-1 promoter region. J. Biol. Chem. 283, 23353–23363 (2008).

8. Kuramochi-Miyagawa, S. et al. DNA methylation of retrotransposon genes is regulatedby Piwi family members MILI and MIWI2 in murine fetal testes. Genes Dev. 22,908–917 (2008).

9. Gonzalez, S., Pisano, D.G. & Serrano, M. Mechanistic principles of chromatinremodeling guided by siRNAs and miRNAs. Cell Cycle 7, 2601–2608 (2008).

10. Zhang, M.X. et al. Effect of 27nt small RNA on endothelial nitric-oxide synthaseexpression. Mol. Biol. Cell 19, 3997–4005 (2008).

11. Zhang, M.X. et al. Biogenesis of short intronic repeat 27-nucleotide small RNAfrom endothelial nitric-oxide synthase gene. J. Biol. Chem. 283, 14685–14693(2008).

12. Tay, Y., Zhang, J., Thomson, A.M., Lim, B. & Rigoutsos, I. MicroRNAs to Nanog, Oct4and Sox2 coding regions modulate embryonic stem cell differentiation. Nature(2008).

13. Kim, D.H., Saetrom, P., Snove, O. Jr. & Rossi, J.J. MicroRNA-directed transcriptionalgene silencing in mammalian cells. Proc. Natl. Acad. Sci. USA 105, 16230–16235(2008).

14. de la Mata, M. et al. A slow RNA polymerase II affects alternative splicing in vivo. Mol.Cell 12, 525–532 (2003).

15. Schwarz, D.S. et al. Asymmetry in the assembly of the RNAi enzyme complex. Cell115, 199–208 (2003).

ART IC L E S

NATURE STRUCTURAL & MOLECULAR BIOLOGY VOLUME 16 NUMBER 7 JULY 2009 7 2 3

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

16. Kim, D.H. et al. Synthetic dsRNA Dicer substrates enhance RNAi potency and efficacy.Nat. Biotechnol. 23, 222–226 (2005).

17. Persengiev, S.P., Zhu, X. & Green, M.R. Nonspecific, concentration-dependent stimu-lation and repression of mammalian gene expression by small interfering RNAs(siRNAs). RNA 10, 12–18 (2004).

18. Weinberg, M.S. et al. The antisense strand of small interfering RNAs directs histonemethylation and transcriptional gene silencing in human cells. RNA 12, 256–262(2006).

19. Han, J., Kim, D. & Morris, K.V. Promoter-associated RNA is required for RNA-directedtranscriptional gene silencing in human cells. Proc. Natl. Acad. Sci. USA 104,12422–12427 (2007).

20. Meister, G. et al. Human Argonaute2 mediates RNA cleavage targeted by miRNAs andsiRNAs. Mol. Cell 15, 185–197 (2004).

21. Janowski, B.A. et al. Involvement of AGO1 and AGO2 in mammalian transcriptionalsilencing. Nat. Struct. Mol. Biol. 13, 787–792 (2006).

22. de la Mata, M. & Kornblihtt, A.R. RNA polymerase II C-terminal domain mediatesregulation of alternative splicing by SRp20. Nat. Struct. Mol. Biol. 13, 973–980(2006).

23. Tong, W., Kulaeva, O.I., Clark, D.J. & Lutter, L.C. Topological analysis of plasmidchromatin from yeast and mammalian cells. J. Mol. Biol. 361, 813–822(2006).

24. Wang, G., Liu, T., Wei, L.N., Law, P.Y. & Loh, H.H. DNA methylation-related chromatinmodification in the regulation of mouse delta-opioid receptor gene. Mol. Pharmacol.67, 2032–2039 (2005).

25. Yokoyama, R., Pannuti, A., Ling, H., Smith, E.R. & Lucchesi, J.C. A plasmid modelsystem shows that Drosophila dosage compensation depends on the global acetylationof histone H4 at lysine 16 and is not affected by depletion of common transcriptionelongation chromatin marks. Mol. Cell. Biol. 27, 7865–7870 (2007).

26. Blau, J. et al. Three functional classes of transcriptional activation domain. Mol. Cell.Biol. 16, 2044–2055 (1996).

27. Kubicek, S. et al. Reversal of H3K9me2 by a small-molecule inhibitor for the G9ahistone methyltransferase. Mol. Cell 25, 473–481 (2007).

28. Cam, H.P., Chen, E.S. & Grewal, S.I. Transcriptional scaffolds for heterochromatinassembly. Cell 136, 610–614 (2009).

29. Moazed, D. Small RNAs in transcriptional gene silencing and genome defence. Nature457, 413–420 (2009).

30. Umlauf, D., Goto, Y. & Feil, R. Site-specific analysis of histone methylation andacetylation. Methods Mol. Biol. 287, 99–120 (2004).

31. Kapranov, P. et al. RNA maps reveal new RNA classes and a possible function forpervasive transcription. Science 316, 1484–1488 (2007).

32. Klinck, R. et al. Multiple alternative splicing markers for ovarian cancer. Cancer Res.68, 657–663 (2008).

33. Kim, V.N., Han, J. & Siomi, M.C. Biogenesis of small RNAs in animals. Nat. Rev. Mol.Cell Biol. 10, 126–139 (2009).

34. Okamura, K. & Lai, E.C. Endogenous small interfering RNAs in animals. Nat. Rev. Mol.Cell Biol. 9, 673–678 (2008).

35. Tam, O.H. et al. Pseudogene-derived small interfering RNAs regulate gene expressionin mouse oocytes. Nature 453, 534–538 (2008).

36. Watanabe, T. et al. Endogenous siRNAs from naturally formed dsRNAs regulatetranscripts in mouse oocytes. Nature 453, 539–543 (2008).

37. Yang, N. & Kazazian, H.H. Jr. L1 retrotransposition is suppressed by endogenouslyencoded small interfering RNAs in human cultured cells. Nat. Struct. Mol. Biol. 13,763–771 (2006).

38. Schwartz, J.C. et al. Antisense transcripts are targets for activating small RNAs. Nat.Struct. Mol. Biol. 15, 842–848 (2008).

39. Morris, K.V., Santoso, S., Turner, A.M., Pastori, C. & Hawkins, P.G. Bidirectionaltranscription directs both transcriptional gene activation and suppression in humancells. PLoS Genet. 4, e1000258 (2008).

40. Haussecker, D. & Proudfoot, N.J. Dicer-dependent turnover of intergenic transcriptsfrom the human beta-globin gene cluster. Mol. Cell. Biol. 25, 9724–9733 (2005).

41. Tufarelli, C. et al. Transcription of antisense RNA leading to gene silencing andmethylation as a novel cause of human genetic disease. Nat. Genet. 34, 157–165(2003).

42. Yu, W. et al. Epigenetic silencing of tumour suppressor gene p15 by its antisense RNA.Nature 451, 202–206 (2008).

43. Rinn, J.L. et al. Functional demarcation of active and silent chromatin domains inhuman HOX loci by noncoding RNAs. Cell 129, 1311–1323 (2007).

44. Kadener, S. et al. Antagonistic effects of T-Ag and VP16 reveal a role for RNA pol IIelongation on alternative splicing. EMBO J. 20, 5759–5768 (2001).

45. Nogues, G., Kadener, S., Cramer, P., Bentley, D. & Kornblihtt, A.R. Transcriptionalactivators differ in their abilities to control alternative splicing. J. Biol. Chem. 277,43110–43114 (2002).

ART IC L E S

72 4 VOLUME 16 NUMBER 7 JULY 2009 NATURE STRUCTURAL & MOLECULAR BIOLOGY

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.

ONLINE METHODSTransfections and treatments. We transfected HeLa and Hep3B cells with 2 mg

of total plasmid DNA using Lipofectamine 2000 according to the manufac-

turer’s protocol (Invitrogen). We used intronic or exonic siRNAs at a range of

5–50 nM for each experiment. We used siAGO1, siAGO2 and siHP1a at 10 nM.

We cotransfected Hep3B cells with a tetracycline-controlled transactivator

(tTA-Sp1 or tTAVP16) and a tetracycline-inducible reporter22. We added tetra-

cycline (1 mg ml–1) shortly after transfection. We induced the cells 24 h later by

washing out tetracycline and adding tetracycline-free medium. We collected the

cells 48 h after induction and used them for subsequent assays. HIV-Gal4

plasmid contains the a-globin–fibronectin minigene reporter for alternative

splicing of the EDI exon under the control of the HIV-2 promoter, fused to

five copies of the target site for the DNA binding domain of Saccharomyces

cerevisiae Gal4. This plasmid generates transcripts with the Tar sequence at their

5¢ ends, which is a binding site for HIV-1 or HIV-2 Tat. The construct contains a

SV40 enhancer-origin located at B600 bp with respect to the transcriptional

start site. Expression vectors for Gal4 fusion proteins Gal4-VP16-(410–490) and

Gal4-SW6 and transfection conditions were described previously45.

For treated cells, 24 h after transfection we added 5-azadeoxycytidine (5 mM)

or BIX-01294 (10 mM) to the medium. Treatments were maintained for 72 h.

We added trichostatin A (300 nM) to cells in culture 24 h before the end of

the experiment.

The sequences of the siRNAs used for transfections are described in

Supplementary Methods.

RNA analysis. We isolated RNA with TRIzol according to the manufacturer’s

protocol (Invitrogen) and carried out reverse transcription using oligo-dT

according to the M-MLV reverse transcriptase protocol (Invitrogen). For FN1

sense-antisense RNA, we used both oligo-dT and random primers (Invitrogen).

We carried out radioactive PCR amplification of cDNA splicing isoforms using

reported specific primers for EDI, both endogenous and reporter minigene, as

previously described44, and we used the same conditions for IIICS control and

Bcl-X. PCR for specificity controls (APAF1, CCNE1, BCL2L1, CHEK2, PCSK6

and BPM4) was done with an initial incubation of 2 min at 95 1C, followed by

35 cycles of 30 s at 94 1C, 30 s at 55 1C and 60 s at 72 1C. The amplification was

completed by a 2-min incubation at 72 1C. We electrophoresed radioactive RT-

PCR products in 6% (w/v) polyacrylamide native gels and detected them by

autoradiography. We measured the radioactivity in the bands with a scintilla-

tion counter according to the Cerenkov method46.

Real-time PCR. We conducted real-time PCR using an Eppendorf Mastercycler

Realplex machine. We measured EDI isoforms in a final volume of 25 ml using

20 mM Tris-HCl (pH 8.0), 50 mM KCl, 4 mM MgCl2 (alternative exon

inclusion) or 5 mM MgCl2 (alternative exon exclusion), 0.2 mM deoxynucleo-

tides, 0.4 mM primer, 1� SYBR Green (Roche) and 0.75 U of Taq DNA

polymerase (Invitrogen). The program used was 2 min at 95 1C and 35 cycles of

15 s at 95 1C and 45 s at 64 1C. We completed each run with a melting curve to

confirm the specificity of amplification and lack of primer dimers. In parallel,

we analyzed the amplification products by gel electrophoresis.

We monitored total human fibronectin mRNA levels with PCR for 2 min at

95 1C and 35 cycles of 15 s at 95 1C, 15 s at 62 1C and 25 s at 72 1C. We

measured I33as target sequence pre-mRNA levels under the same conditions as

for the EDI isoforms.

We measured AGO1 and AGO2 mRNAs using the same SYBR Green Master

Mix as for the EDI isoforms, with 4 mM MgCl2. The program used was 2 min

at 95 1C and 35 cycles of 15 s at 95 1C, 15 s at 68 1C and 30 s at 72 1C.

We amplified fibronectin sense and antisense cDNA simultaneously using

the SYBR Green Master Mix with primers P7 and P8 and 3 mM MgCl2. The

program used was 3 min at 95 1C and 35 cycles of 30 s at 95 1C, 30 s at 57 1C

and 45 s at 72 1C. We used similar program conditions to measure proximal

and distal regions in the pUHC-EDI minigene. We carried out reverse

transcription with specific primers for each region.

We used the housekeeping genes ACTB (b-actin) and HSP90AB1 (Homo

sapiens heat shock 90-kDa protein 1, beta) as controls. We used 4 mM MgCl2with both primer sets. For ACTB, we used the same program as for EDI

isoforms. For HSP90AB1, we used 2 min at 95 1C and 35 cycles of 15 s at 95 1C,

30 s at 63 1C and 20 s at 68 1C.

We measured all fibronectin genomic regions (A–D) with a PCR mix

prepared by Biodynamics Argentina. We used 2 min at 95 1C and 35 cycles

of 15 s at 95 1C and 45 s at 66 1C.

Antisense quantitative real-time RT-PCR. We reverse-transcribed TRIzol-

purified mRNA from cells transfected with siLuc or I33as for 50 min at

42 1C using 2 pM of specific primers: RTFN1sense, RTasFN1 (antisense) and

RTHP1 (unspecific), with (+RT) or without (–RT) M-MLV. We calculated the

relative quantities by real-time PCR with P3 and P4 primers (I33as target

sequence) using the equation FN1sense(+RT) – FN1sense(–RT) – HP1 (non-

specific reverse transcription). We did the same with asFN1 and then measured

the ratio of antisense to sense for each cell line.

Sequences of primers used in PCR experiments are described in Supple-

mentary Methods.

Native chromatin immunoprecipitation. We performed mononucleosome

preparation and ChIPs assays as described30. We used 20 mg of input mono-

nucleosomes for immunoprecipitation. We recovered the DNA from unbound

and bound fractions with a QIAquick PCR purification kit according the

manufacturer’s protocol (Qiagen) and analyzed it by real-time PCR as

described previously. We calculated relative enrichments by dividing the ratio

(amount of DNA product bound to specific antibody:amount bound to

control IgG) by the ratio (amount of DNA product not bound to specific

antibody:amount not bound to control IgG). Final results were normalized to

exon 1 values.

Antibodies and conditions for the bioinformatics search, high-throughput

RT-PCR and capillary electrophoresis are described in Supplementary Methods.

46. Cramer, P. et al. Coupling of transcription with alternative splicing: RNA pol IIpromoters modulate SF2/ASF and 9G8 effects on an exonic splicing enhancer. Mol.Cell 4, 251–258 (1999).

NATURE STRUCTURAL & MOLECULAR BIOLOGY doi:10.1038/nsmb.1620

©20

09 N

atu

re A

mer

ica,

Inc.

All

rig

hts

res

erve

d.