Senescence is an endogenous trigger for microRNA-directed ...fulltext.calis.edu.cn/nature/ncb/14/3/ncb2443.pdf · followed by qPCR of CDC2, CDCA8, PCNA and CCNA2 promoters. Data are

ART I C L E S

Senescence is an endogenous trigger formicroRNA-directed transcriptional genesilencing in human cellsMoussa Benhamed1,2, Utz Herbig3, Tao Ye4, Anne Dejean1,2,5,6 and Oliver Bischof1,2,5,6

Cellular senescence is a tumour-suppressor mechanism that is triggered by cancer-initiating or promoting events in mammaliancells. The molecular underpinnings for this stable arrest involve transcriptional repression of proliferation-promoting genesregulated by the retinoblastoma (RB1)/E2F repressor complex. Here, we demonstrate that AGO2, RB1 and microRNAs (miRNAs),as exemplified here by let-7, physically and functionally interact to repress RB1/E2F-target genes in senescence, a process thatwe call senescence-associated transcriptional gene silencing (SA-TGS). Herein, AGO2 acts as the effector protein forlet-7-directed implementation of silent-state chromatin modifications at target promoters, and inhibition of the let-7/AGO2effector complex perturbs the timely execution of senescence. Thus, we identify cellular senescence as the an endogenous signalof miRNA/AGO2-mediated TGS in human cells. Our results suggest that miRNA/AGO2-mediated SA-TGS may contribute totumour suppression by stably repressing proliferation-promoting genes in premalignant cancer cells.

Cellular senescence represents a robust and essentially irreversibletumour-suppressive barrier that cells must overcome to develop intoa full-blown malignancy1. Previous studies have proposed that the irre-versibility of the senescence arrest is tightly associatedwith the senescentcells’ particular chromatin architecture, which is epitomized by theappearance of so-called senescence-associated heterochromatin foci(SAHFs; refs 2,3). SAHFs are a hallmark of senescent cells that containseveral common markers of transcriptionally repressed heterochro-matin and were hypothesized to silence genes important for cell prolif-eration, in particular those regulated by the E2F/RB1 repressor complex(for example cyclin A2 (CCNA2), cyclin E (CCNE) or proliferating cellnuclear antigen (PCNA); ref. 2). However, whether or not SAHFs reallycontain E2F-target genes and how the senescence-associated inactivechromatin state at E2F-target genes is implemented and maintainedremains unclear. Recent studies in model organisms have provided alink between argonaute (AGO) proteins, short interfering (si)RNA-guided heterochromatin formation and transcriptional gene silencing(TGS; ref. 4). In humans, there are four AGO proteins (AGO1–4, alsoknown as eukaryotic translation initiation factor 2C, 1–4 (EIF2C1–4)),and AGO1 and 2 have been previously implicated in TGS inducedby exogenous siRNAs and microRNAs (miRNAs) directed againstgene promoter transcripts through promotion of changes in histone

1Institut Pasteur, Nuclear Organisation and Oncogenesis Unit, Department of Cell Biology and Infection, F-75015 Paris, France. 2INSERM, U993, F-75015 Paris,France. 3New Jersey Medical School–University Hospital Cancer Center and Department of Microbiology and Molecular Genetics, UMDNJ-New Jersey Medical School,Newark, New Jersey 07103, USA. 4Institut de Génétique et de Biologie Moléculaire et Cellulaire (IGBMC), Department of Functional Genomics and Cancer, BPF-10142-67404 Illkirch Cedex, France. 5These authors contributed equally to this work.6Correspondence should be addressed to O.B. or A.D. (e-mail: [email protected] or [email protected])

Received 9 March 2011; accepted 16 January 2012; published online 26 February 2012; DOI: 10.1038/ncb2443

covalent modifications and DNA methylation5–7. Notwithstanding,many mechanistic details of this process remain poorly defined inhuman cells, and very little is known about the identity of possibleendogenous signals that may drive this process in human cells. Giventhe evolutionary conserved role of siRNAs and AGO proteins in TGSand heterochromatin formation, we set out to analyse their possibleinvolvement in senescence-associated repression of E2F-target genes.

RESULTSGenome-wide identification of AGO-bound E2F-target genes andAGO/heterochromatin-bound miRNAs in senescenceTo determine, in an unbiased manner, which genes might beunder the control of AGO proteins, we carried out genome-widepromoter profiling in senescent and presenescent controlWI38 primaryfibroblasts applying ‘ChIP (chromatin immunoprecipitation)-on-chip’technology using an anti-pan-AGO antibody. The enrichment valuefor each promoter was determined (Methods) and yielded 4,516potential AGO–promoter binding sites in senescent cells versus 2,619 inpresenescent cells. Of these binding sites, 702 were in common betweenthe two conditions; however, the false-discovery rate for AGO-bindingsites in senescent cells was several-fold lower than in control cells,indicating an enrichment for AGO proteins at the respective promoters

266 NATURE CELL BIOLOGY VOLUME 14 | NUMBER 3 | MARCH 2012

© 2012 Macmillan Publishers Limited. All rights reserved.

mailto:[email protected]

mailto:[email protected]

ART I C L E S

a b

702 1,9173,814

AGO-target genes

S C

0

1,000

2,000

3,000

4,000

5,000

S C320 2574,196 S

77 5002,542

E2FAGO

E2FAGO

C

Down150

Up65

c

fd

0

5

4

3

2

1

543210

e

0

1

2

3

4

0 1 2 3 4 50

1

2

3

4

5

0 1 2 3 4

Log 1

0 IP

AG

O c

pm

Σ A

GO

-tar

get

gene

s

Log 1

0 IP

AG

O c

pm

r2 = 0.74

r2 = 0.47

r2 = 0.788

r = 0.862 P < 0.00001

RAS log10 cpm

IP H

3K9

log 1

0 cp

m

r = 0.686 P < 0.000001

RAS log10 cpm IP H3K9 log10 cpm

r = 0.888 P < 0.000001

Figure 1 Identification of AGO-bound E2F-target genes andheterochromatin-bound miRNAs. (a–c) AGO ChIP-on-chip promoterprofiling. ChIP from RASV12-induced senescent (S) and presenescent,empty-vector control (C) cells was carried out using pan-anti-AGO antibody.Purified DNA fragments from matched genomic inputs and ChIP sampleswere amplified, fluorochrome labelled and hybridized to NimbleGenpromoter arrays. (a) Total number of AGO2-bound promoters in controland RASV12-induced senescent cells, and a Venn diagram depictingspecific and common AGO2–promoter targets. (b) Venn diagram depictingnumber of AGO-bound (AGO) and E2F-target promoters (E2F) in control

(C, upper panel) and senescent (S, lower panel) cells; P <0.001, chi-squaretest. (c) Correlation between AGO-bound E2F-target promoters andrespective E2F-target gene expression level; Up, AGO-bound E2F-targetgenes upregulated in senescence; Down, AGO-bound E2F-target genesdownregulated in senescence. (d–f) Degree of association between miRNAsequence counts observed in unfractionated, cellular RNA and AGO (d) orH3K9me2 (e) RIP and between AGO- and H3K9me2 RIPs (f). Normalizedlogarithmic counts per million (log10 cpm) for each miRNA from therespective sample were scatter-plotted (see also Supplementary Table S4);r2, correlation coefficient of determination; r , Pearson correlation coefficient.

in senescent cells (Fig. 1a, Supplementary Fig. S1a–f and Table S1). Toprobe a potential link betweenAGOproteins and E2F-target promoters,we inspected how many of the E2F-regulated promoters were AGObound. This analysis revealed, that of the top 577 E2F-responsivepromoters known at present8–11, 320 (that is 55.5%) were occupiedby AGO proteins in senescent cells, as opposed to only 77 (that is13.3%) in control cells (Fig. 1b). We then correlated AGO–promoteroccupancy of E2F-target genes with the comparative expression profileof these genes in senescent versus presenescent control cells usingmicroarray-based transcriptome data. We found that, of the 320AGO-bound E2F-target genes, 150 (that is∼46.5%) were down- and65 (that is ∼20.6%) were upregulated (Fig. 1c and SupplementaryTable S1). Functional annotation of the 150 downregulated E2F-targetgenes showed a clear enrichment in genes involved in cell-cyclecontrol (Supplementary Fig. S2a and Table S2), whereas no significantenrichment was found for upregulated E2F-target genes (data notshown). Together, these results suggest that AGO proteins may beinvolved in senescence-associated repression of E2F-target genes.Exogenous miRNAs with sequence complementarity to promoter

regions were shown to induce AGO-mediated TGS by implementing atranscriptionally repressive chromatin environment7. This led us to in-vestigate whether miRNAs might be involved in AGO-mediated repres-sion of E2F-target genes in senescence. To obtain a detailed picture ofAGO-immunoprecipitating miRNAs (RIPs) in senescent cells, we usednext-generation sequencing (NGS). Importantly, we included histoneH3 dimethylated on lysine 9 (H3K9me2) in this analysis to assign po-tential AGO2-interacting miRNAs to a repressive chromatin state andunfractionated, cellular RNA from senescent cells for normalization(Supplementary Fig. S2b). Small RNA sequences were determined by

NGS, producing between 3.2 and 4.5 million 33-nucleotide sequencesfor the three samples (Supplementary Table S3). To search for previ-ously known miRNAs within these results, we ran a string-matchingalgorithm (allowing one mismatch) to count the number of sequencesaligning with human mature miRNAs. Of the known 847 humanmature and mature-star miRNAs annotated in release 12.0 of miRBase,we identified a total of 451 unique miRNAs in unfractionated cellularRNA, 377 in RIP-AGO and 237 in RIP-H3K9me2 of senescent cells(Supplementary Table S4). After adjusting for sequencing depth by con-verting to logarithmic counts per million sequence reads (log cpm), wecomputed the degree of association (Pearson coefficient, r ; coefficientof determination, r2) between miRNA expression in unfractionated,cellular RNA and RIP-associated miRNAs and between AGO- andH3K9me2-associated miRNAs (Fig. 1d–f). We found that the degree ofassociation betweenmiRNA expression in unfractionated, cellular RNAandAGO-associatedmiRNAs is high (r=0.862; P<0.000001; Fig. 1d),whereas the degree of association between miRNA expression inunfractionated, cellular RNA andH3K9me2-associatedmiRNAs is onlymoderate (r = 0.686; P < 0.000001; Fig. 1e). Remarkably, however, thedegree of association betweenAGO- andH3K9me2-associatedmiRNAsis high (r = 0.888; P < 0.000001; Fig. 1f), thus there exists a strongpositive correlation between AGO-bound and heterochromatin-boundmiRNAs in senescent cells.

AGO2 accumulates in the nucleus of senescent cells to targetrepressed E2F-responsive promotersHaving established a link between AGO proteins and promoters ofrepressed E2F-target genes on the one hand as well as AGO-boundmiRNAs and a repressive chromatin state on the other hand, we

NATURE CELL BIOLOGY VOLUME 14 | NUMBER 3 | MARCH 2012 267


ART I C L E S

ba

c dAGO2 MacroH2A

DAPI Merge DAPI Merge

f

AGO2

Lamin B

GAPDH

Cytosol Nucleus

AGO2 MacroH2A

AGO2 DAPI Merge

S

C

CCNA2

Per

cent

age

of in

put

P

erce

ntag

e of

inp

ut

Per

cent

age

of in

put

P

erce

ntag

e of

inp

ut

C S

CDC2 CDCA8 PCNA

0

0.2

0.4

0.6

0.8

Per

cent

age

of in

put

0

0.2

0.4

0.6

0.8AGO2

CDC2 CDCA8 PCNA CCNA2

H3K27me3


0

0.2

0.4

0.6

0.8

1.0

1.2 H3K4me3

CCNA2PCNACDCA8CDC2

0

0.1

0.2

0.3

0.4 IgG

CCNA2PCNACDCA8CDC20

0.1

0.2

0.3

0.4 H3K9me2

Histone H3

K9ac

C S C S C S C S

K9me2K27me3

Co-IP e

IgG

AGO2

WCL 10%

C S C S C S

1:1 1:2.5 1:2.2 Ratio

Figure 2 AGO2 accumulates in the nucleus of senescent cells and isrecruited to promoters of repressed E2F-target genes. (a) Western-blotanalysis of cytosolic and nuclear fractions from RASV12-induced senescent(S) and presenescent, empty vector control (C) WI38 fibroblasts usinganti-AGO2, anti-lamin B and anti-GAPDH (glyceraldehyde 3-phosphatedehydrogenase) specific antibodies. Lamin B and GAPDH detection serve aspositive controls for nuclear and cytosolic fractions; WCL, whole-cell lysate.(b) Localization of AGO2 in detergent-pre-extracted presenescent, control(C) and senescent (S) cells as determined by indirect immunofluorescencemicroscopy. 4,6-diamidino-2-phenylindole (DAPI) was used to counterstainnuclear DNA; scale bar, 10 µm. (c,d) Co-localization of AGO2 and macroH2Ain nuclei of melanocytic nevus (c) and in cytosol of melanoma (d) asdetermined by indirect immunofluorescence; scale bar, 20 µm. (e) GlobalAGO2 enrichment in heterochromatin of senescent cells as detected by

histone-association assay. Presenescent, control (C) and senescent (S) cellswere fixed by paraformaldehyde, sonicated and lysed. Euchromatin- andheterochromatin-bound AGO2 was identified by co-immunoprecipitation(Co-IP) using anti-histone H3K9ac (euchromatin), H3K27me3, H3K9me2(heterochromatin) and IgG control antibodies. Immunoprecipitates wereWestern blotted with anti-AGO2 and anti-histone H3 antibodies. Ratios werecalculated densitometrically. (f) AGO2 and histone modification profiling ofE2F-responsive promoters. qChIP was carried out in presenescent, emptyvector control (C) and RASV12-induced senescent (S) cells using anti-AGO2,anti-H3K27me3, anti-H3K9me2, anti-H3K4me3 and IgG control antibodiesfollowed by qPCR of CDC2, CDCA8, PCNA and CCNA2 promoters. Dataare means ± s.d.; n = 3; P < 0.05. Experiments were carried out inquadruplicate. Uncropped images of blots are shown in SupplementaryFig. S9.

were then interested in whether a specific AGO–miRNA effectorcomplex is involved in SA-TGS of E2F-target genes. There are fourAGO proteins (AGO1–4) in humans. AGO2 was recently shown

to undergo nucleo-cytoplasmic shuttling12 and to be differentiallyexpressed in human cancers13–15. Therefore, we focused on AGO2in the further study.



ART I C L E S

a

HDAC1

IgG AGO2HDAC1

AGO2

Co-IP

RB1

WCL 10%

AGO1

AGO2

0

0.25

0.50

0.75

1.00

– 100 200 – 100 200 – –

– – – – – – 100 200

–

–2 3 4 5 6 7 8 91

0.210.12 0.06

0.20

0.530.63

1 1.02

0.33

RB1 – – – + + + – +–

E2F + + + + + + + +–

RLU

c

0

0.10

0.20

H3K27me3

CsiRNA siRNAAGO2

CDCA8CCNE2

RB1

C AGO2

0.05

0.15

0.25

0.30

Per

cent

age

of in

put

CDC2CCNA2

f

i

00.2

0.6

1.0

1.4

C RB1siRNA

Rel

. Rb

exp

ress

ion

g

0

0.05

0.10

0.15

0.20

0.25 CDCA8

CCNE2

C RB1

d

0

0.5

1.0

10

––

– – 100 200– – + + +

+++++

E2F1-wt E2F1-mt

–2 3 4 5 61

–

RLU

9.95

0.951 0.98

1.150.98

b

–RB1 +RB1

AGO2 AGO2RB1 RB1

WCL 10%–RB1+RB1

RB1AGO2

IP

WBMix

e

0123456

AGO2PCNA

CCNE2

CDCA8

CDC2

CCNA2

AGO1 siRNA/ControlAGO2 siRNA

Rel

. exp

ress

ion AGO2 0.27

RB1 1

AGO2C Ratio AGO2/CsiRNA:

AGO1

h Control siRNA RB1 siRNA

∗

1 2 3 4 1 2 3 4 5 6

*

Per

cent

age

of in

put

E2F1AGO2

RB1

Figure 3 AGO2 cooperates with RB1 to regulate E2F-target gene expression.(a,b) AGO2 and RB1 or HDAC1 were co-immunoprecipitated in celllysates prepared from RASV12-induced senescent WI38 fibroblasts (a) orRB1-inducible (Tet-Off) senescent Saos-2 (RB1-negative) cells (b). IgG,HDAC1 and AGO2 co-immunoprecipitates (Co-IP) were Western blottedwith anti-AGO2, anti-HDAC1 and anti-RB1 antibodies. Cellular lysates fromRB1-inducible (Tet-Off) Saos-2 (RB1-negative) cells were prepared andimmunoprecipitated with anti-AGO2 and anti-RB1 antibodies followed byWestern blotting (WB) using a mix of anti-AGO2 and anti-RB1 antibodies.WCL, whole cell lysate. −RB1, no RB1 expression, Tet+; +RB1, RB1,expression Tet−. Asterisk, IgG heavy chain; arrow, specific HDAC1 band.(c,d) AGO2 and RB1 co-repress cyclin E promoter. (c) Wild-type cyclin Epromoter–luciferase reporter assay in C33A cells. All indicated plasmidamounts in nanograms. (d) The same assay using a cyclin E promoterwith mutated E2F-recognition site: E2F1-wt, promoter reporter withE2F1-binding sites; E2F1-mt, promoter reporter with mutated E2F1-bindingsites. Data are means ± s.d.; n =5; P <0.05. Experiments were carriedout in triplicate. Expression of proteins was confirmed by Western blot (datanot shown). RLU, relative light units. (e,f) Depletion of AGO2 derepressesand releases RB1 from E2F-target genes. Cells were retrovirally infected

with pBabe-RASV12 and, 2 days post drug selection, cells were transientlytransfected with AGO1 siRNA, AGO2 siRNA or scrambled siRNA control(C; see also Supplementary Fig. S4c). (e) qrtPCR on total RNA preparedfrom respective samples at day 3 of siRNA treatment. Also shown is aWestern blot for AGO2 levels in siRNA-treated cells; RB1, loading control.Ratio calculated densitometrically. Data are means ± s.d.; n=3; P <0.05.Experiments were carried out in quadruplicate. (f) H3K27me3- andRB1-qChIP on CDCA8, CCNE2, CDC2 and CCNA2 promoters was carriedout at day 3 of siRNA treatment. Data are means ± s.d.; n =3; P <0.05.Experiments were carried out in quadruplicate. (g–i) Depletion of RB1 doesnot have an impact on AGO2 presence at E2F-target genes. Cells wereretrovirally infected with pBabe-RASV12 and 2 days post drug selection cellswere transiently transfected with scrambled siRNA control or RB1 siRNAfor 3 days. (g) AGO2-qChIP for E2F-target genes. Data are means ± s.d.;n=3; P <0.05. Experiments were carried out in quadruplicate. (h) Indirectimmunofluorescence of RB1 in siC- or RB1 siRNA-treated cells; scalebar, 20 µm. (i) qrtPCR on total RNA prepared from siRNA-treated samples.Data are means ± s.d.; n = 3; P < 0.05. Experiments were carried outin quadruplicate. Uncropped images of blots are shown in SupplementaryFig. S9.

First, we sought to determine whether AGO2 may accumulatein the nucleus of cells undergoing senescence using an anti-AGO2specific monoclonal antibody (Supplementary Fig. S2c). Senescencewas induced either by retroviral introduction of oncogenic RASV12

into WI38 primary human diploid fibroblasts, by treatment ofMCF-7 breast cancer cells with senescence-inducing concentrations ofdoxorubicin or by re-expression of RB1 in RB1-inducible Saos-2 (RB1-negative) cells. Biochemical fractionation (Fig. 2a and SupplementaryFig. S2d) and indirect immunofluorescence studies (Fig. 2b andSupplementary Fig. S2e,f) revealed that AGO2 accumulates in thenucleus of senescent cells when compared with presenescent controlcells, although the overall protein levels remain largely unchanged. Toextend the physiological significance of senescence-associated nuclearaccumulation of AGO2 from cell culture to human tissue samples, we

analysed localization of AGO2 in melanocytic nevi and melanomasby indirect immunofluorescence. Nevi are archetypes of benigntumours that frequently harbour oncogenic mutations and are highlyenriched in senescent melanocytes, whereas melanomas are malignanttumours essentially devoid of senescent melanocytes16–18. Strikingly,we observed a pronounced nuclear enrichment of AGO2 in senescentmelanocytic nevi, as shown by its co-localization with the senescencemarker macroH2A (refs 19,20; Fig. 2c and Supplementary Fig. S3a–d).Conversely, in melanomas AGO2 was essentially excluded frommelanocytic nuclei and predominantly localized to the cytosol (Fig. 2dand Supplementary Fig. S3e–g). Next, we quantitatively assessedAGO2’s partitioning between heterochromatin and euchromatin usinga histone association assay21. We established that, in senescent cells,AGO2 predominantly co-precipitates with facultative heterochromatin



ART I C L E S

Per

cent

age

of

Ed

U-p

ositi

ve n

ucle

i

0 10 20 30

AGO2 shRNA 1

AGO2 shRNA 2E7

Control shRNA

Time (d)

Pop

ulat

ion

dou

blin

gs AGO2 shRNA 1

AGO2 sh

RNA 1

Control shRNA

Contro

l shR

NA

AGO2 shRNA 2

AGO2 sh

RNA 2

AGO2 sh

RNA 1

Contro

l shR

NAAGO2

shRNA 2

AGO2 sh

RNA 1

Contro

l shR

NA

AGO2 sh

RNA 1

0

30

60

Per

cent

age

of

SA

HF-

pos

itive

nuc

lei

90

AGO2 shRNA 2

Control shRNA

Magnified

Per

cent

age

of

Ed

U-p

ositi

ve n

ucle

i

AGO2 shRNA 1 AGO2 shRNA 2Control shRNA

0

20

40

Per

cnat

age

of S

A-β

-G

al-p

ositi

ve c

ells

60

0 25 50 75

AGO2 shRNA 1AGO2 shRNA 2

Control shRNA

Time (d)P

opul

atio

n d

oub

lings E7

0

1

2

3

4

0

5

10

15

0

2

4

6

8

05

101520253035

a b

de

f

g

c

AGO2 sh

RNA 1

Contro

l shR

NA

AGO2 sh

RNA 2

Figure 4 Depletion of AGO2 delays senescence arrest in W38 fibroblasts.(a) Proliferation curves of WI38 fibroblasts under physiological oxygenconcentration of 3% superinfected either with control shRNA andRASV12, AGO2 shRNA 1/2 and RASV12 or HPV16E7 and RASV12 (E7).(b–d) Analysis of cell populations expressing control shRNA withRasV12 and AGO2 shRNAs with RasV12 for different proliferation andsenescence markers at day 14 of lifespan study. (b) Representativephotomicrographs of cell density and morphology; scale bar, 10 µm.(c,d) Bar charts for fraction of EdU-incorporating cells (c) andpercentage of SAHF-positive cells (d). Also shown is SAHF formationas determined by fluorescence microscopy using DAPI counterstainingof DNA in cells expressing control shRNA with RASV12 and AGO2

shRNAs with RASV12; scale bar, 20 µM. (e) Proliferation curve ofWI38 fibroblasts at physiological oxygen concentration of 3% stablyinfected with either control shRNA, two independent AGO2 shRNAconstructs (AGO2 shRNA 1 and 2) or viral oncogene HPV16E7. Growthcurves were initiated when control shRNA vector cell population ceasedproliferation. (f,g) Analysis of control shRNA and AGO2 shRNA-silencedcell populations for different proliferation and senescence markers atday 21 of lifespan study. (f) Bar chart for fraction of EdU-incorporatingcells. (g) Representative micrographs showing SA-β-Gal staining andgraph for percentage of SA-β-Gal-positive cells; scale bar, 15 µM. Dataare means ± s.d.; n = 3; P < 0.05. Experiments were carried outin duplicate.

markers histone H3 trimethylated on lysine 27 (H3K27me3) andH3K9me2 when compared with presenescent control cells (Fig. 2e).By contrast, a comparable amount of AGO2 co-precipitated with theeuchromatin marker histone H3 acetylated at lysine 9 (H3K9ac) inboth cell states (Fig. 2e). We conclude that AGO2 accumulates inthe nucleus of senescent cells, where it shows an increased affinity toinactive chromatin when compared with presenescent cells.To validate AGO2-bound E2F-target promoters as revealed by

ChIP-on-chip, we used qChIP (ChIP quantified by real-time PCR)for a selected number of target genes. Simultaneously, we alsoexamined histone modifications at the respective promoters and thegene-expression statuses. We found that AGO2 levels were elevated∼2–4-fold on selected E2F-target promoters of CDC2 (also knownas CDK1, cyclin-dependent kinase 1), CDCA8 (cell division cycleassociated 8), PCNA and CCNA2 (Fig. 2f) as well as ORC6L, USP1,NEK2,MYC, FANCF, CDC6, BUB1, BRCA2, ATM, ASF1B and CCNE2(Supplementary Fig. S4a) in senescent cells when compared withpresenescent cells. The elevated physical presence of AGO2 at therespective promoters was accompanied by a substantial increase in the

negative methyl histone marks H3K27me3 (∼3–6-fold) and H3K9me2(∼2–4-fold) and by a decrease (∼2–6-fold) in the positive methylhistone mark histone H3 trimethylated on lysine 4 (H3K4me3; Fig. 2f)and paralleled by a decrease (∼2–5-fold) in expression of the respectivegenes (Supplementary Fig. S4b). Importantly, in senescent cells silencedfor AGO2 expression we detected a greatly diminished presence ofAGO2 at investigated promoters, thus underscoring AGO2 specificity(Supplementary Fig. S4c,d). Together, these results provide strongevidence for an enrichment of AGO2 at E2F-target promoters insenescent cells. Moreover, they reveal that the repressive histonemodification marks H3K27me3 and H3K9me2 coincide with thepresence of AGO2 at E2F-regulated cell-cycle genes in senescence arrest,thus hinting at an active contribution of AGO2 in senescence-associatedrepression of E2F-target genes.

Physical and functional interaction between AGO2 and RB1 inthe regulation of E2F-target genes in senescent cellsThe RB1 protein functions as a repressor of E2F-target genes insenescent cells supposedly by nucleating heterochromatin formation



ART I C L E S

C

AGO2 0 5 10 15

Pop

ulat

ion

dou

blin

gs

Time (d)

0

20

40

60

C AGO2

Per

cent

age

of E

dU

-pos

itive

cel

ls

Per

cent

age

of S

AH

F-p

ositi

ve c

ells

Magnified

C

AGO2

DAPIDAPI

AGO2

DAPI DAPIDAPI

AGO2

DAPI

DAPI

H3K9me3

GFP

Merge

0

2

4

6

8

10

0

25

50

75

a b

de

c

C AGO2

Figure 5 Overexpression of AGO2 induces proliferative arrestwith features of premature senescence. (a) Lifespan study offluorescence-activated cell-sorted GFP-positive WI38 fibroblaststransduced with vector control (C) or MigR–AGO2. (b-e) Analysis ofcell populations overexpressing empty vector control (C) or AGO2 fordifferent proliferation and senescence markers at day 14 of lifespanstudy. (b,c) Bar charts for percentage of EdU-incorporating cells (b)

and percentage of SAHF-positive cells (c). (d,e) SAHF formationin detergent-pre-extracted control and AGO2-overexpressing cells byindirect fluorescence using DAPI DNA counterstain to visualize SAHF (d)and in detergent-pre-extracted GFP-positive AGO2-overexpressing cellsby indirect H3K9me3 immunostaining to visualize SAHF (e); scale bars,20 µm. Data are means ± s.d.; n = 3; P < 0.05. Experiments werecarried out in duplicate.

at respective promoters2, at least partly through recruitment ofhistone deacetylase 1 (HDAC1) or histone methyl-transferase SUV39H(refs 22–25). We therefore considered whether AGO2 might beinvolved in RB1-mediated repression in senescence. To test thispossibility, we first carried out co-immunoprecipitation experimentsbetween endogenous AGO2, RB1 and HDAC1 in cellular lysatesprepared from RASV12-induced senescent fibroblasts. As shown inFig. 3a, HDAC1 and AGO2 co-immunoprecipitated with each other.Moreover, both HDAC1 and AGO2 efficiently co-immunoprecipitatedRB1. Similar results were obtained in Saos-2 cells undergoing RB1-induced senescence (Fig. 3b). Consistent with the physical interactionbetween AGO2 and RB1, AGO2 showed partial co-localization withSAHFs at their periphery, a staining pattern that was similar forthe heterochromatin marker H3K27me3 (Supplementary Fig. S5a)and RB1 (ref. 2) and is thus congruent with the peripheral SAHFlocalization of E2F-target genes3.Next, we investigated whether AGO2 could act as a co-repressor

with RB1 by carrying out promoter–reporter assays using the naturalE2F-responsive CCNE promoter26. Expression of RB1 repressedE2F-mediated transactivation of the promoter by 79% whereasincreasing amounts of AGO2 led to a 37–47% repression (Fig. 3c,columns 2–5). A co-repressive effect was observed when both AGO2and RB1 were present simultaneously (Fig. 3c, columns 6 and 7) andthis effect was further enhanced in the presence of HDAC1 (datanot shown). Co-repression between AGO2 and RB1 was specificfor AGO2 as its paralogue AGO1 had no effect (Fig. 3c, columns8 and 9). By contrast, RB1 and AGO2 did not affect a version ofthe cyclin E promoter in which the E2F1-recognition sites wereeliminated by mutation (Fig. 3d). The repressive function of AGO2 onexpression of endogenous E2F-target genes was then verified usingquantitative real-time PCR (qrtPCR) on total RNA isolated from

RASV12-induced senescent fibroblasts silenced for AGO2 expressionby RNA interference. RASV12-induced senescent fibroblasts treatedwith AGO2 siRNA, but not with AGO1 siRNA, showed derepressionof E2F-target genes (Fig. 3e and Supplementary Fig. S5b) and thiswas positively correlated with a decreased presence of H3K27me3and RB1 at target promoters (Fig. 3f). Combined, these results showthat AGO2 and RB1 cooperate in the transcriptional repressionof E2F-responsive promoters. Interestingly, AGO2 recruitment toE2F-target promoters seems to be largely independent from RB1, assiRNA-mediated depletion of RB1 only mildly impacts the physicalpresence of AGO2 at tested promoters (Fig. 3g–i).

AGO2 deficiency regulates the onset of senescenceGiven the repressive function of AGO2 on E2F-target genes in senes-cence, we tested whether AGO2 plays an active role in the establishmentof the senescent phenotype. To start to address this question, we firstoverexpressed RASV12 either in fibroblasts expressing control shorthairpin RNA, in fibroblasts in which AGO2 expression was stablysilenced by either of two distinct short hairpin RNAs (AGO2 shRNA 1and 2) or in fibroblasts overexpressing viral oncogene HPV16E7, thelatter serving as a positive control for senescence delay by interferingwith the RB1 pathway. Cells expressing control shRNA and RASV12

ceased proliferation and senesced within 7 days. By contrast, RB1- andAGO2-deficient RASV12 cells had a slightly postponed senescence onsetby∼2.5–3 population doublings but were unable to completely bypasssenescence (Fig. 4a,b). Consistently, in AGO2-deficient RASV12 cells thepercentage of 5-ethynyl-2′-deoxyuridine (EdU)-incorporating cellsincreased and the fraction of SAHF-positive cells decreased transientlywhen compared with control cells expressing RASV12 and controlshRNA (Fig. 4c,d and Supplementary Fig. S6). Similar results were ob-tained in cells treated withAGO2 siRNAmolecules (see Fig. 7a below).



ART I C L E S

c

CDCA8

CYP

CDC2

C let-7f Ratio let-7f/C

Antisense CDC2

g

0

1

2

3

4

0 2 4 6 8

C

let-7fCel

l num

ber

× 1

05

Time (d)

a

let-7f200 bp

100 bp

b

Sense CDCA8

miR-let-7f

miR-let-7f

f

Per

cent

age

of in

put

siC +

let-7

0

0.1

0.3

0.5

0.7

AGO2 siR

NA + le

t-7

h 100

80

60

40

20Per

cent

age

of

pos

itive

cel

ls

Ki67 SA-β-Gal

C let-7f

d

Fold

enr

ichm

ent

AGO2

CDC2

CDCA8

CDC2

CDCA8

CDC2

CDCA8

C let-7f 0

0.5

1.0

1.5

2.0

2.5

3.03.5

H3K27me3

C let-7f C let-7f

IgG

Biotin–let-7f

1. Anti-AGO2

2. SA

1. Anti-AGO2

2. SA

1. Anti-AGO2

2. SA

Biotin–let-7f

1. Control

2. SA

e

0

1

2

3

4

5

Fold

enr

ichm

ent

siC let-7f

6

7

RIP: H3K9 AGO2 IgG Input

CDC2 ∗

∗

miR target

UA

–

CDCA8

–388/410 –143/138

– –

– – GA

A

A

–220/241 –49/42

U U U

U

UUAUGU UG GG G GA

UUAUGGAUGGGC G

GGU A

UU UGA

C UUAUGUU UGA

UUAUGU UGA

G G

U

A

AG

A

U G

G

AUGAUG G

AU UG GGA

0.13

0.10

miR target

Figure 6 AGO2 and let-7f cooperate to induce TGS of E2F-target promoters.(a) Interaction of AGO2 and H3K9me2 (H3K9) with let-7f was validated bycombined RIP and srcDNA cloning using a miR-let-7f-specific primer. Thisresult is representative of three independent experiments. IgG, non-specificantibody. bp, base pairs. (b) Alignment of let-7f to CDC2 and CDCA8target promoters. Identical nucleotides are shown in red. Asterisk, TSS;the E2F-binding site is depicted as a grey box with respect to the TSS;the arrows to the right and left of the miRNA target are primers usedfor qChIP or qrtPCR; the numbers on top indicate the miRNA-target-siteposition with respect to the TSS. (c–f) let-7f and AGO2 cooperate toimplement epigenetic gene silencing at E2F-target genes. (c) Exogenouslet-7f or scrambled miRNA (C) were transiently transfected at 100nM intopresenescent WI38 fibroblasts. Transcription of CDC2, CDCA8 and CYP(cyclophylin) mRNAs was measured by nuclear run-on (left) or qrtPCR(right). CYP served as a normalization control. Shown is a representativeexperiment of three experimental repeats. (d) qChIP was carried out inlet-7f-treated cells using anti-AGO2, anti-H3K27me3 and IgG controlantibodies followed by qPCR using primers amplifying CDCA8 and CDC2promoter regions. Data are means ± s.d.; n=4; P <0.05. Experiments were

carried out in quadruplicate. (e) Simultaneous presence of let-7f and AGO2on CDCA8 and CDC2 promoters was determined by combined qChIP–ChOPassay. Presenescent WI38 fibroblasts were transfected with 100nMbiotin-labelled let-7f, control miRNA (siC) or let-7 followed by successiverounds of control-IgG-ChIP or AGO2-ChIP and streptavidin-affinity (SA)precipitation. Streptavidin-affinity chromatin precipitates were analysedby qPCR with primers detecting E2F-responsive promoters CDCA8 andCDC2. Data are means ± s.d.; n = 4; P < 0.001. Experiments werecarried out in duplicate. (f) Presence of let-7 at target promoters is AGO2dependent. Presenescent WI38 were co-transfected with AGO2 siRNA or siC(100nM) and biotin-labelled let-7f (100nM) followed by streptavidin-affinityprecipitation. Data are means ± s.d.; n = 3; P < 0.05. Experimentswere carried out in duplicate. (g,h) Exogenous let-7f induces cellularsenescence. Presenescent WI38 fibroblasts were transiently transfectedwith 100nM let-7f or control miRNA (C). (g) Cell numbers were determinedat the indicated times. (h) Number of cells staining positive for Ki67 andSA-β-Gal at day 8 post treatment. Data are means ± s.d.; n =3; P <0.05.Experiments were carried out in duplicate. Uncropped images of blots areshown in Supplementary Fig. S9.

To further corroborate the functional requirement for AGO2 incellular senescence, we investigated its role in replicative senescence.Accordingly, we silenced AGO2 expression in medium-passage fibrob-lasts with shRNAs and determined their replicative potential using

cells overexpressing viral oncogene HPV16E7 as a lifespan exten-sion yardstick for RB1-compromised cells. In AGO2-deficient cellsreplicative lifespan increased by∼3–4 population doublings whereasE7-expressing cells showed a lifespan extension of ∼5–6 population



ART I C L E S

a

0

20

40

60

80

100

0 2 4 6 8

Control siRNA

AGO2 siRNA

Anti-miR

AGO1 siRNA

Cel

l num

ber

× 1

03

Time (d)

b

0

10

20

30

40

Per

cent

age

of

Ki6

7-p

ositi

ve n

ucle

i

Contro

l/AGO1

siRNA

AGO2 siR

NA

Anti-m

iR

c

00.20.40.60.81.01.21.4

CDC2CDCA8

C

Per

cent

age

of in

put

Anti-m

iR

d

01234567

CDC2CDCA8

5

1

Rel

ativ

e ex

pre

ssio

n

5.8

1.81.85

C/AGO1

siRNA

Anti-m

iR

AGO2 siR

NA

**

**

Figure 7 Inhibition of let-7f perturbs timely execution of senescence andSA-TGS. (a,b) WI38 fibroblasts undergoing RASV12-induced senescencewere treated with 100nM AGO1 siRNA, AGO2 siRNA, controlscrambled siRNA (siC) or let-7f antagomirs (Anti-miR). Cell proliferativecapacity was measured by growth curves (a) and by immunostainingusing anti-Ki67 antibody (b). Data are means ± s.d.; n=3; P <0.04.Experiments were carried out in duplicate. (c) AGO2-qChIP wascarried out on RASV12-senescent cells treated with 100nM control

(C) or let-7f antagomirs on CDCA8 and CDC2 promoters. Data aremeans ± s.d.; n = 3; P < 0.01. Experiments were carried out induplicate. (d) qrtPCR was carried out on total RNA prepared fromWI38 fibroblasts undergoing RASV12-induced senescence treated with100nM AGO1 siRNA, control (C), AGO2 siRNA or let-7f antagomirsto measure CDCA8 and CDC2 mRNA levels. Asterisk, statisticalsignificance P . Data are means ± s.d.; n=3; P <0.05. Experimentswere carried out in quadruplicate.

doublings (Fig. 4e). In line with this finding, the fraction of AGO2-deficient cells incorporating EdU increased and senescence-associatedbeta-galactosidase (SA-β-Gal) activity decreased transiently when com-pared with shC control cells (Fig. 4f,g). Next, we overexpressed AGO2in early-passage WI38 fibroblasts. Strikingly, AGO2-overexpressingcells induced an abrupt proliferative arrest (Fig. 5a) with features ofsenescence as shown by a decrease in the number of cells incorporatingEdU (Fig. 5b) and an increase in the percentage of SAHF-positivecells (Fig. 5c–e). Altogether, these data indicate that AGO2 activelycontributes to the proper execution of the senescence response.

AGO2–miRNA complexes are involved in SA-TGS ofE2F-target genesWe hypothesized that AGO-bound miRNAs with sequence comple-mentarity to promoter regions may induce AGO-mediated SA-TGS byimplementing a transcriptionally repressive chromatin environment.On closer inspection of all detected miRNAs, we found that the topten AGO2-interacting miRNAs were identical to the top ten H3K9me2heterochromatin-interactingmiRNAs andwere allmembers of the let-7family, with let-7a and let-7f having the highest normalized sequencecounts (cpm; Supplementary Table S5). The interaction between eitherH3K9me2 or AGO2 and let-7f was confirmed by RNA immunopre-cipitation combined with small RNA complementary DNA (srcDNA)cloning and sequencing27 in senescent cells (Fig. 6a). Of note, pre-miR-let-7f2 shows the strongest upregulation among the let-7 family mem-bers in senescent cells when compared with control cells (Supplemen-tary Fig. S7a). In the further analysis we therefore focused our attentionon let-7f function in SA-TGS. To assess whether let-7f directs TGS insenescent cells, we carried out a bioinformatic search for potential bind-ing sites in AGO2-repressed E2F-target promoters using RNAhybrid28.We found two imperfectly matched, putative let-7f-target sites showingvery low minimal free energy (mfe) values of hybridization, one in theantisense direction of the CDC2 promoter at position−388 to−410(mfe= 19.9 kcalmol−1) and the other in the sense direction of theCDCA8 promoter at position−220 to−241 (mfe= 29.4 kcalmol−1)relative to the respective transcriptional start sites (TSSs; Fig. 6b).

To elucidate whether let-7 and the putative let-7-binding site inthe E2F-responsive CDC2 promoter29 are involved in RB1/AGO2-mediated transcriptional repression, we initially carried out pro-moter–reporter assays using a wild-type version of the CDC2 promoteror a derivative thereof in which the putative let-7-binding site wasmutated (Supplementary Fig. S7b-d). Expression of RB1 repressed boththe wild-type and let-7-mutated promoter (∼1.8 versus ∼1.4-fold).When we co-transfected cells with either RB1/let-7, RB1/AGO2 orRB1/AGO2/let-7 and the respective promoter–reporter construct,we observed an increasing co-repressive effect between let-7, AGO2and RB1 only for the wild-type (RB1:let-7, ∼3-fold; RB1:AGO2,∼4-fold; RB1:AGO2:let-7, ∼5-fold) and not the let-7-mutated pro-moter–reporter construct (RB1:let-7, ∼1.9-fold; RB1:AGO2, ∼1.9-fold; RB1:AGO2:let-7,∼2.2-fold). These data indicate that let-7, AGO2and RB1 cooperate in the repression of the E2F-responsive CDC2promoter and that the let-7 site at position−388 to−410 is importantfor optimal repression.On the basis of the promoter–reporter data, we then assessed

the susceptibility of CDCA8 and CDC2 genes to miRNA-mediatedTGS by transiently transfecting let-7f or control miRNA into early-passage WI38 human fibroblasts and monitoring CDCA8 and CDC2transcript levels either by nuclear-run-on transcription assay orqrtPCR. Transcript abundance was significantly reduced for thetwo genes in the presence of let-7f (Fig. 6c) or pre-miR-let-7f(Supplementary Fig. S8a) in both assays. The nuclear-run-on data,however, indicated that let-7f-induced gene silencing occurs directlyat the transcriptional level. To examine whether silencing leads torecruitment of AGO2 and changes in histone marks at the respectivepromoters, we carried out qChIP of let-7f- and pre-miR-let-7f-treatedcells. We observed a pronounced let-7f-dependent enrichment ofAGO2 and H3K27me3 at the CDCA8 and CDC2 promoters (Fig. 6dand Supplementary Fig. S8b). Importantly, AGO2 and let-7f weresimultaneously present at the CDCA8 and CDC2 promoters as shownby a dual-pulldown assay (qChIP–ChOP, chromatin-oligo-affinityprecipitation30; Fig. 6e). Using this assay, we also identified ORC6Las an AGO2/miR-185-bound E2F-response gene (Supplementary



ART I C L E S

RB1

E2F

HMTsHDACs

AGO2

H3K9me2 H3K27me3

pRNA

RB1

E2FAGO2

Senescence

ncAGO2

cyAGO2

miR Cytosol

Nucleus

Passenger

Senescence

Senescence

E2F targets E2F targets

HMTsHDACs

RNA Pol IIRNA Pol II

H3K27me3H3K9me2

Figure 8 Model for AGO2 and miRNA function in SA-TGS of E2F-targetgenes. In cells undergoing senescence, cytosolic (cy) miRNA/AGO2complexes are retained in the nucleus (nc). We favour a model inwhich the complex is guided by the miRNA (red) to the respective targetpromoter(s), where it associates with promoter-associated RNA(s) (pRNA)(left), which has/have been shown to be critical for TGS (ref. 30). Indeed,our preliminary data indicate that RNA Pol II-dependent TSS-proximalpRNAs are produced at E2F targets in proliferating cells, whereasin senescent cells these transcripts become almost extinct. Thus, itremains to be seen whether or not AGO2 slices pRNAs after association

with the respective promoters. Alternatively, the complex may directlytarget complementary target promoter sequences (right). In both cases,the miRNA/AGO2 complex ultimately blocks productive RNA Pol IIengagement and cooperates with E2F/RB1 to repress E2F-targetpromoters through recruitment of (further) co-repressor activitiesincluding histone methyl transferases (HMTs) and deacetylases (HDACs)and/or by stabilizing a pre-existing E2F/RB1-associated repressorcomplex. Targeted promoters are ultimately stably repressed by aninactive chromatin state characterized by methylated lysines 9 and 27on histone H3 as cells undergo senescence.

Fig. S8c). The physical interaction between let-7 and the respectivepromoters was found to beAGO2dependent, as in AGO2-depleted cellslet-7 detectability at the respective promoters was greatly diminished(Fig. 6f).Moreover, let-7 was undetectable at the E2F-responsive AGO2target promoter of CDC6, which shows no sequence complementarityto let-7 (Supplementary Fig. S8d). Strikingly, introduction of let-7f orpre-miR-let-7f into presenescent WI38 fibroblasts led to proliferationarrest (as did treatment of cells with CDCA8 siRNA) with cellularfeatures reminiscent of senescence, whereas AGO2 siRNA/let-7 treatedcells proliferated normally (Fig. 6g,h and Supplementary Fig. S8e–g).Conversely, senescent fibroblasts treated with a let-7f antagomirshowed enhanced proliferative capacity as manifested by an increase inboth cell proliferation and positive Ki67 immunostaining (Fig. 7a,b).However, the observed effects were less prominent than in senescentcells treated with AGO2 siRNA, indicating that AGO2 depletionhas a more widespread effect on cell proliferation. Furthermore,let-7f-antagomir-treated senescent cells had decreased amounts (abouthalf) of AGO2 at CDCA8 and CDC2 promoters (Fig. 7c), whereasmRNA levels were increased (Fig. 7d). Collectively, these resultssupport the conclusion that let-7f, AGO2 and RB1 cooperate tonucleate a transcriptionally repressive chromatin state at selected targetpromoters during senescence, thus promoting the proper execution ofthe senescence response.

DISCUSSIONSeveral studies have linked synthetic siRNAs and endogenous miRNAstargeted to promoter regions to heterochromatin formation and TGSin human cells; however, an endogenous trigger for this phenomenonhas not been identified so far5. Moreover, the global association ofendogenous RNA interference components with chromatin and theirrole in TGS has so far not been studied in detail. Here, we reporta genome-wide chromatin-binding analysis for endogenous AGOproteins and miRNAs in human cells and provide direct evidencefor cellular senescence as a physiological signal for miRNA-mediatedTGS to occur. The results of our study strongly suggest that inparticular nuclear AGO2 acts as the effector protein for miRNA-guidedSA-TGS and cooperates with RB1 in the stable repression of certainE2F-target genes in senescence (Fig. 8). The roles AGO2 and let-7 (plusother miRNAs) play in SA-TGS are critical for the implementationof inactive chromatin marks at target promoters and may thusplay an important role for the timely execution of the senescencearrest, as exemplified here by their ability to repress expressionof the CDC2 and CDCA8 oncogenes. These findings, therefore,extend the well-established role of miRNAs and AGO2 in cytosolicpost-transcriptional gene silencing to nuclear TGS, although the exactcontribution of miRNA-mediated TGS and post-transcriptional genesilencing for senescence remains to be determined.



ART I C L E S

What might be the underlying mechanism for AGO2/miRNA-mediated SA-TGS? Our data imply that miRNA/AGO2 complexesmay bring in further co-repressor functions such as HDACs orhistone methyl transferases to stably silence gene expression andmay constitute an effective roadblock for productive engagementof RNA polymerase II (Pol II) at targeted promoters. Alternatively,and not mutually exclusively, they could also enhance the stability ofthe RB1-associated repressor complex(es). An important question thatneeds to be addressed in the future concerns the mechanism by whichAGO2/miRNA complexes accumulate in the nucleus of senescentcells. Another important question that awaits answers regards themechanism by which AGO2/miRNA complexes are tethered to theirrespective target promoters. In conclusion, we provide strong evidencethat nuclear translocation/retention and function of miRNA/AGO2complexes involved in SA-TGS may represent a decisive parameterfor tumour suppression. Further studies are now needed to identifyother nuclear AGO/miRNA complexes regulating both E2F- andnon-E2F-responsive genes during senescence. �

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

Note: Supplementary Information is available on the Nature Cell Biology website

ACKNOWLEDGEMENTSWe would like to thank A. Verdel and M. Yaniv for discussions and critical readingof the manuscript. We are grateful to P. Adams for providing the anti-macroH2Aantibody to U.H. and N. Mirani for technical help with histopathology as well asS. Volinia and C. Croce for miRNA profiling and J. Doudement (GenomeQuest,France) for bioinformatics analysis. This work was supported by grants from LigueNationale Contre le Cancer (Equipe labellisée), Association for International CancerResearch, Agence Nationale de la Recherche, Association pour la Recherche sur leCancer (ARC), OdysseyRe and the New Jersey Commission on Cancer Research09-1124-CCR-EO to U.H. O.B. is a CNRS (Centre National de la RechercheScientifique) fellow, A.D. Institut National de la Santé et de la Recherche Médicale(INSERM)/Institut Pasteur and M.B. was supported by ARC.

AUTHOR CONTRIBUTIONSO.B. and M.B. conceived the project. M.B. carried out experiments. O.B., M.B.and A.D. analysed the data. U.H. carried out immunohistochemical staining onnevi and melanomas. T.Y. carried out bioinformatic analysis. O.B. and A.D. wrotethe manuscript.

COMPETING FINANCIAL INTERESTSThe authors declare no competing financial interests.

Published online at http://www.nature.com/naturecellbiologyReprints and permissions information is available online at http://www.nature.com/reprints

1. Collado, M. & Serrano, M. Senescence in tumours: evidence from mice and humans.Nat. Rev. Cancer 10, 51–57 (2010).

2. Narita, M. et al. Rb-mediated heterochromatin formation and silencing of E2F targetgenes during cellular senescence. Cell 113, 703–716 (2003).

3. Adams, P. D. Remodeling chromatin for senescence. Aging Cell 6, 425–427 (2007).4. Moazed, D. Small RNAs in transcriptional gene silencing and genome defence.

Nature 457, 413–420 (2009).5. Morris, K. V. RNA-mediated transcriptional gene silencing in human cells. Curr. Top.

Microbiol. Immunol. 320, 211–224 (2008).6. Janowski, B. A. et al. Involvement of AGO1 and AGO2 in mammalian transcriptional

silencing. Nat. Struct. Mol. Biol. 13, 787–792 (2006).7. Kim, D. H., Saetrom, P., Snove, O. Jr. & Rossi, J. J. MicroRNA-directed

transcriptional gene silencing in mammalian cells. Proc. Natl Acad. Sci. USA 105,16230–16235 (2008).

8. Rabinovich, A., Jin, V. X., Rabinovich, R., Xu, X. & Farnham, P. J. E2F in vivo bindingspecificity: comparison of consensus versus nonconsensus binding sites. GenomeRes. 18, 1763–1777 (2008).

9. Xu, X. et al. A comprehensive ChIP–chip analysis of E2F1, E2F4, and E2F6 in normaland tumor cells reveals interchangeable roles of E2F family members. Genome Res.17, 1550–1561 (2007).

10. Ross, J. F., Naar, A., Cam, H., Gregory, R. & Dynlacht, B. D. Active repressionand E2F inhibition by pRB are biochemically distinguishable. Genes Dev. 15,392–397 (2001).

11. Bieda, M., Xu, X., Singer, M. A., Green, R. & Farnham, P. J. Unbiased locationanalysis of E2F1-binding sites suggests a widespread role for E2F1 in the humangenome. Genome Res. 16, 595–605 (2006).

12. Rudel, S., Flatley, A., Weinmann, L., Kremmer, E. & Meister, G. A multifunctionalhuman Argonaute2-specific monoclonal antibody. RNA 14, 1244–1253 (2008).

13. Zhou, Y. et al. High-risk myeloma is associated with global elevation ofmiRNAs and overexpression of EIF2C2/AGO2. Proc. Natl Acad. Sci. USA 107,7904–7909 (2010).

14. Kim, M. S. et al. Somatic mutations and losses of expression of microRNAregulation-related genes AGO2 and TNRC6A in gastric and colorectal cancers.J. Pathol. 221, 139–146 (2010).

15. Li, L., Yu, C., Gao, H. & Li, Y. Argonaute proteins: potential biomarkers for humancolon cancer. BMC Cancer 10, 38 (2010).

16. Michaloglou, C. et al. BRAFE600-associated senescence-like cell cycle arrest ofhuman naevi. Nature 436, 720–724 (2005).

17. Dhomen, N. et al. Oncogenic Braf induces melanocyte senescence and melanomain mice. Cancer Cell 15, 294–303 (2009).

18. Gray-Schopfer, V. C. et al. Cellular senescence in naevi and immortalisation inmelanoma: a role for p16? Br. J. Cancer 95, 496–505 (2006).

19. Sporn, J. C. et al. Histone macroH2A isoforms predict the risk of lung cancerrecurrence. Oncogene 28, 3423–3428 (2009).

20. Zhang, R. et al. Formation of MacroH2A-containing senescence-associatedheterochromatin foci and senescence driven by ASF1a and HIRA. Dev. Cell 8,19–30 (2005).

21. Ricke, R. M. & Bielinsky, A. K. Easy detection of chromatin binding proteins by theHistone Association Assay. Biol. Proced Online 7, 60–69 (2005).

22. Nielsen, S. J. et al. Rb targets histone H3 methylation and HP1 to promoters. Nature412, 561–565 (2001).

23. Luo, R. X., Postigo, A. A. & Dean, D. C. Rb interacts with histone deacetylase torepress transcription. Cell 92, 463–473 (1998).

24. Brehm, A. et al. Retinoblastoma protein recruits histone deacetylase to represstranscription. Nature 391, 597–601 (1998).

25. Magnaghi-Jaulin, L. et al. Retinoblastoma protein represses transcription byrecruiting a histone deacetylase. Nature 391, 601–605 (1998).

26. Ohtani, K., DeGregori, J. & Nevins, J. R. Regulation of the cyclin E gene bytranscription factor E2F1. Proc. Natl Acad. Sci. USA 92, 12146–12150 (1995).

27. Ro, S. et al. Cloning and expression profiling of testis-expressed piRNA-like RNAs.RNA 13, 1693–1702 (2007).

28. Krueger, J. & Rehmsmeier, M. RNAhybrid: microRNA target prediction easy, fast andflexible. Nucleic Acid Res. 34, W451–W454 (2006).

29. Dalton, S. Cell cycle regulation of the human cdc2 gene. EMBO J. 11,1797–1804 (1992).

30. Han, J., Kim, D. & Morris, K. V. Promoter-associated RNA is required for RNA-directed transcriptional gene silencing in human cells. Proc. Natl Acad. Sci. USA104, 12422–12427 (2007).



http://www.nature.com/naturecellbiology/

http://www.nature.com/naturecellbiology

http://www.nature.com/reprints

http://www.nature.com/reprints

METHODS DOI: 10.1038/ncb2443

METHODSVectors, viruses, cell culture and antibodies. Retroviral vector pBabe-RASV12 and LXSN-HPV16E7 were as reported31. MigR and MigR–AGO2 werefrom A. Tarakhovsky. Lentiviral AGO2 shRNA constructs TRCN0000007864,TRCN00000011203 and shControl SHC002 were from Sigma. Lentiviral infectionwas carried out by standard procedures. The CDC2 wild-type and mutant promoterconstructs were cloned by PCR into pGL3 (Promega). Low-passage, primary humandiploid fibroblast strain WI38 and breast cancer cell line MCF-7 were purchasedfrom ATCC. Tetracycline (Tet)-Off RB1-inducible Saos-2 cell line was from L.Zhu32. Culturing of cells and infection of primary human diploid fibroblasts strainWI38 by retroviral-mediated gene transfer were carried out at physiological oxygenconcentration of 3% as previously described31. The following antibodies were usedin this study: mouse or rabbit polyclonal antibodies anti-H3K9-dimethyl (Milliporeno 07-441 and Abcam no 1220), anti-H3K27-trimethyl (Upstate and Milliporeno 07-449), anti-histone 3 (Millipore no 09-838), anti-pan-AGO (Milliporeno 07-590), anti-AGO2 9E8.2 (Millipore no 04-642), normal rabbit andmouse IgGs(Millipore no 12-370 and no 12-371), anti-lamin B (Ab-1, Calbiochem, no NA12),GAPDH and HDAC1 antibodies (Abcam, no ab8245 and no ab7028, respectively),mouse monoclonal anti-RB1 (clones G3-245 ref 554136 and XZ-55 ref 554144,Pharmingen), rabbit polyclonal Ki67 (Vectorlabs, noVP-K451) and anti-macroH2Aantibody (P. Adams). All fluorochrome-tagged secondary antibodies were fromMolecular Probes.

Senescence analysis. Senescence inMCF-7 cells was induced by 1 µMdoxorubicintreatment for 24 h (ref. 33). Subsequently, drug was washed out and cells were leftin fresh medium for 4–5 days, after which they became senescent. RB1-inducibleTet-Off was used to induce senescence in Saos-2 cells. Tet was washed out fromthe medium and Tet-free fresh medium was added. Senescence of cells wasobtained within 4–5 days. Senescence was assessed using several assays as previouslypublished31,34. Briefly, proliferative capacity was determined by growth curves,indirect immunofluorescence using anti-Ki67 antibody staining (Boehringer)or EdU incorporation using Click-iT (Invitrogen) as per the manufacturer’sinstructions. Cells were also co-stained for SA-β-Gal activity.

Subcellular fractionation. Cells were collected and lysed in CSK buffer on ice(0.5M sucrose; 15mM Tris at pH 7.5; 60mM KCl; 0.25mM EDTA; 0.125mMEGTA; 0.5mM spermidine; 0.15mM spermine; 1mMdithiothreitol). Subsequently,NP-40 was added to a final concentration of 0.5% and lysates were centrifuged toobtain a cytosolic supernatant and nuclear pellet. Nuclei were further purified bycentrifugation at 300g through a 0.88M sucrose cushion.

Co-immunoprecipitation and Western blotting. Cells were extracted in300mM NaCl, 10mM Tris at pH 8.0, 0.5% NP-40, 1mM EDTA and proteaseinhibitors. For immunoprecipitations, equal amounts of lysate were incubated with1–2 µg of respective antibodies overnight at 4 ◦C. Precipitates were prepared andanalysed by Western blotting according to standard procedures using indicatedantibodies at a dilution of 1:1,000. When immunoprecipitation was not carriedout, Western blotting of total protein lysates was carried out according tostandard procedures.

ChIP and histone association assay. ChIP assay on chromatin from senescentand presenescent, control cells was carried out using 1–2 µg of respective antibodiesas described35. Calculation of the amount of immunoprecipitated E2F-targetpromoter DNA relative to that present in total input chromatin was calculatedaccording to ref. 36. When the histone association assay was carried out, crosslinkedcellular lysates were immunoprecipitated with the respective histone antibodiesaccording to ref. 21. Co-immunoprecipitates were analysed by Western blot usingindicated antibodies.

Dual affinity purification (ChIP–ChOP). Presenescent WI38 fibroblasts weretransfected in multiple replicates with 100 nM antisense 5′-biotin-tagged miR-let-7f(MWG). Cultures were collected 72 h after transfections and ChIP–ChOP assay wascarried out according to ref. 30 with minor modifications.

Nuclear run-on. 50,000 presenescent fibroblasts were transfected with indicatedmiRNAs. Nuclear run-on was carried out as described in ref. 37. A dot blot assay ofpulled-down RNA was carried out using radiolabelled CYP for normalization andCDC2 and CDCA8 reverse primers as probes and autoradiography for visualization.

Immunofluorescence. Cells were prepared and immunolabelled with primary(1:200 dilution) and secondary (1:800 dilution) antibodies as previously described34.Where noted cells were pre-extracted first with CSK buffer (10mM PIPES at pH6.8, 100mMNaCl, 300mM sucrose, 3mMMgCl2, 1mMEGTA, 0.5%Triton-X100)and cytoskeleton stripping buffer (10mM Tris-HCl at pH 7.4, 10mM NaCl, 3mMMgCl2, 1% Tween 40 (v/v), 0.5% sodium deoxycholate (v/v)) followed by fixationwith paraformaldehyde at 4%. Cells were blocked with 0.5% BSA, 0.2% gelatin in

PBS ×1 and probed with respective primary and secondary antibodies. Labelledcells were viewed by epifluorescence microscopy. For immunohistochemistrydeparaffinized 4 µm tissue sections were incubated in sodium citrate buffer (10mMNa citrate, 0.05% Tween 20 at pH 6) at 95 ◦C for 45min to retrieve antigens. Tissuesections were subsequently rinsed with water and incubated with block buffer (4%BSA in PBS+0.1%Tween 20, PBST) for 30min. Primary antibodies were incubatedovernight at 4 ◦C in block buffer (macroH2A 1:5,000; Ago2 1:200). Following 2×5minwashes with PBST, tissue sectionswere incubatedwith secondary antibodies asindicated (1:1,000 in block buffer) for 1 h at room temperature. Slides were washed(3× 5min) with PBST, rinsed with water and mounted using DAPI-containingmounting medium. Archival and paraffin-embedded tumour tissue were obtainedfrom the tumour tissue bank at New JerseyMedical School University Hospital withthe approval of the local institutional review board committee.

Transfection, reporter assays, siRNA, miRNA and antagomir. Cells wereco-transfected with 0.1 µg wild-type or mutant cyclin E–luciferase26 or CDC2promoter–reporter constructs andmammalian expression vectors for pCMV-β-Gal(10 ng), pCMV-RB1 (300 ng), pCMV-E2F1/DP1 (10 ng), let-7 (100 nM) and pCMV-AGO1 and 2 at the indicated amounts in nanograms using Lipofectamine 2000(Invitrogen). β-galactosidase and luciferase activities were measured using Galacto-Star (Tropix) and luciferase (Promega) luminescent assay kits according to thesuppliers’ instructions. Luciferase activities were normalized with β-galactosidaseactivities. Expression of proteins was confirmed by Western blot (data not shown).For siRNA, miRNA and antagomir transfection experiments, cells were transfectedat 100 nM. All siRNA duplexes were SMARTpools (Dharmacon). Pre-miR-let-7f,miR-let-7f and miR-let-7f antagomir were purchased from Exiqon. Respectivescrambled sequences were used as negative control.

RNA isolation, qrtPCR analysis and qChIP-primers. Total RNA and cDNApreparation were carried out by standard procedures. Real-time PCRwas conductedusing Qiagen QuantiTec primer pairs and designed primer pairs for indicatedgenes or promoter-specific primers for qChIP analysis and nuclear-run-on assay(Supplementary Table S6) with Roche LightCycler and Roche Absolute QPCR SYBRGreen Capillary Mixes. In all assays CYP served as normalization control.

RNA immunoprecipitation (RIP). Senescent cells induced by oncogenic RAS orreplicative exhaustion were crosslinked by ultraviolet radiation at 100–400mJ cm−2

and immediately collected afterwards on ice. Cell pellets were lysed in radioim-munoprecipitation assay lysis buffer on ice, adjusted to a protein concentration of10mgml−1 and digested extensively with DNase I for 15min at 37 ◦Cfollowed bydigestion with RNase T1 (1000Uml−1) for 15min at 23 ◦C. Lysates were cleared andincubated with 2 µg of pan-anti-AGO, anti-H3K9me2 or IgG-antibodies (Upstate).Immunoprecipitates were collected with protein A/G beads, washed extensivelyand incubated again with 5,000Uml−1 RNase T1 at 23 ◦C to rid the precipitatesof any unspecifically attached RNA molecules. Immunocomplexes were proteinaseK-digested and purified RNA molecules ethanol-precipitated and used as probes indownstream applications.

Next-generation sequencing of small RNAs and analysis of sequencing data.Small RNA samples were prepared from RIPs or unfractionated, cellular RNAof senescent cells using a Small RNA-Seq Sample Prep Kit (Illumina) followedby the standard Illumina RNA sequencing protocol using the Genome Analyser.Bioinformatic analysis of RNA-Seq results was carried out by GenomeQuest France(www.genomequest.com). Beginning with raw reads obtained from Illumina G1sequencing the adaptor sequence was removed. All reads smaller than 18 nucleotidesafter removing the adaptor sequence were discarded, allowing one mismatchwithin the adaptor sequence. Small RNA sequences were mapped to the complete,unmasked, National Center for Biotechnology Information version of the humangenome (build 36). The alignment algorithm used aligns the entire read, allowingone mismatch, insertion or deletion. Each sequence read was classified as either hitor no hit and was annotated with the number of hits found. All reads mapping to asingle contiguous position in the genome were then clustered. For each genomiccluster the genomic sequence was extracted and put into a separate sequencedatabase (the genomic cluster database). The cluster database was mapped tomiRBase (v12, http://www.mirbase.org/), resulting in an miRNA annotated clusterdatabase. miRNA reads were normalized to one million small RNAs (cpm). Allmappings used the GenomeQuest KERR proprietary implementation. The KERRalgorithm belongs to a class of approximate string-matching algorithms. Given twosequences S1 and S2, KERR computes the minimal number of differences betweenS1 and S2, by trying to optimally fit the shorter sequence into the longer one.Differences in sequence alignments arise from mismatches or gaps. This strategyassures finding all the local alignments allowing a given number of mismatches,insertions or deletions. For all comparisons one error was allowed. For moreinformation about the GenomeQuest KERR implementation, please refer to http://wiki.genomequest.com/index.php/LSPMUL_HELP#KERR_.28aka_GenePAST.29.

NATURE CELL BIOLOGY


www.genomequest.com

http://www.mirbase.org/

http://wiki.genomequest.com/index.php/LSPMUL_HELP#KERR_.28aka_GenePAST.29

http://wiki.genomequest.com/index.php/LSPMUL_HELP#KERR_.28aka_GenePAST.29

DOI: 10.1038/ncb2443 METHODS

srcDNA cloning. srcDNA cloning was conducted according to ref. 27. Primersused for PCR were 5′-CGAATTCTAGAGCTCGAGGCAGG-3′ and let-7f-specificprimers (5′-GAGGTAGTAGATTGTATAGT-3′).

ChiP-on-chip assay and bioinformatic analysis. We used a NimbleGen 385KPromoter two-array set (UCSC HG18), which consists of 59,357 human promoters(see also www.nimblegen.com for details concerning array platform). Preparationof ChIP samples was per supplier’s instructions. The labelling and hybridizationof DNA from input and immunoprecipitation fractions for ChIP-on-chip analysiswere carried out by NimbleGen Systems. Raw fluorescence-intensity data wereobtained from scanned images of the oligonucleotide tiling arrays usingNimbleScan2.2 extraction software (NimbleGen Systems). Only peak data with a false-discovery-rate score of 0.05 or less were used for analysis. Scaled log 2-ratio data wereviewed using SignalMap. Assembly of top E2F-target genes was from refs 8–11. Functional annotations were carried out using the program Database forAnnotation, Visualization and Integrated Discovery (DAVID) 2009 (ref. 38).

miRNA expression profiling Microarray analysis was carried out as described inref. 39.

Statistical analysis All statistical analyses were carried out in Excel. Results areshown asmeans ± s.d. as indicated. P values were calculated by Student’s two-tailedt -test or chi-squared test. All qPCR experiments were carried out in quadruplicate,reporter and senescence assays in triplicate and growth curves in duplicate.

Accession numbers. The ChIP-on-chip data used in this study may beviewed under GSE33998 at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE33998.

The data from the genome-wide expression analysis used in this study maybe viewed under GSE33710 at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE33710.

The RIP-Seq data used in this study may be viewed under GSE34494 athttp://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE34494.

31. Bischof, O. et al. The E3 SUMO ligase PIASy is a regulator of cellular senescenceand apoptosis. Mol. Cell 22, 783–794 (2006).

32. Ji, P. et al. An Rb-Skp2-p27 pathway mediates acute cell cycle inhibition by Rb andis retained in a partial-penetrance Rb mutant. Mol. Cell 16, 47–58 (2004).

33. Chang, B. D. et al. A senescence-like phenotype distinguishes tumor cells thatundergo terminal proliferation arrest after exposure to anticancer agents. CancerRes. 59, 3761–3767 (1999).

34. Bischof, O, Nacerddine, K. & Dejean, A. Human papillomavirus oncoprotein E7targets the promyelocytic leukemia protein and circumvents cellular senescence viathe Rb and p53 tumor suppressor pathways.Mol. Cell Biol. 25, 1013–1024 (2005).

35. Kumar, P. P. et al. Functional interaction between PML and SATB1 regulateschromatin-loop architecture and transcription of the MHC class I locus. Nat. CellBiol. 9, 45–56 (2007).

36. Frank, S. R., Schroeder, M., Fernandez, P., Taubert, S. & Amati, B. Binding ofc-Myc to chromatin mediates mitogen-induced acetylation of histone H4 and geneactivation. Genes Dev. 15, 2069–2082 (2001).

37. Meng, L., Bregitzer, P., Zhang, S. & Lemaux, P. G. Methylation of the exon/intronregion in the Ubi1 promoter complex correlates with transgene silencing in barley.Plant Mol. Biol. 53, 327–340 (2003).

38. Huang, D. W., Sherman, B. T. & Lempicki, R. A. Systematic and integrativeanalysis of large gene lists using DAVID bioinformatics resources. Nat. Protocols4, 44–57 (2008).

39. Liu, C. G., Spizzo, R., Calin, G. A. & Croce, C. M. Expression profiling of microRNAusing oligo DNA arrays. Methods 44, 22–30 (2008).

NATURE CELL BIOLOGY


www.nimblegen.com

http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE33998





S U P P L E M E N TA RY I N F O R M AT I O N

WWW.NATURE.COM/NATURECELLBIOLOGY 1

DOI: 10.1038/ncb2443

Supplementary Information, Fig. S1 a. b.

c. d.

e. f.

g.

Figure S1 a-g, Visualisation of AGO2 ChIP-on-chip data obtained from Nimblegen arrays. a, AGO2 promoter association as detected on chromosome (chr) 16. Note the enrichment of AGO2 on promoters in senescence compared to control. b, AGO2 binding region on PCNA promoter. c, AGO2 binding region on CCNE2 (cyclin E2) promoter. d, AGO2 binding region

on CDCA8 promoter. e, AGO2 binding region on CDC2 promoter. f, AGO2 binding region on CCNA2 (cyclin A2) promoter, g, AGO2 binding region on CDC6 promoter . Red peaks represent false discovery rate (FDR) score ≤ 0.05. Orange peaks FDR score ≤ 0,1; Yellow peaks 0,1 < FDR score ≤ 0,2; Grey peaks represent the lowest probability of a peak with FDR score > 0.2.



2 WWW.NATURE.COM/NATURECELLBIOLOGY


e.

d.

C S C S C S Input cytoplasm nucleus

AGO2

Lamin B

GAPDH

Supplementary Information, Fig. S2

f.

c.

DAPI AGO2 MERGE

Control

Doxorubicin

Rb-off (TET+)

Rb-on (TET-)

DAPI AGO2 MERGE Rb

a. b.

Adaptor-Ligation

Cellular RNA AGO-RIP

Next Generation Seq

Bioinformatic Analysis

Senescent Cells

H3K9me2-RIP

FH-AGO AGO 1 2 3 4

α-AGO2

α-HA 95

Nuclear mitotic apparatus protein1 Cylindromatosis Geminin mutL homolog1 H2AX Cyclin-dependent Kinase Inhibitor 3 MCM2 Nuclear Protein, ATM locus Ubiquitin-Conjugating Enzyme E2S Ligase I Vacuolar Protein Sorting 24 Homolog 5 (S. Cerevisiae) CDC28 Protein Kinase Regulatory Subunit 1B Rac GTPase Activating Protein 1 Protein Regulator of Cytokinesis 1 Ubiquitin-Conjugating Enzyme E2C Kinesin Family Member C1 Budding Uninhibited by Benzimidazoled 3 Homolog (yeast) Kinesin Family Member 11 CDC25B Homolog (S. pombe) CDCA8 Sperm-Associated Antigen 5 CDC6 Homolog (S. Cerevisiae) Cyclin A2 CDC23 homolog (S. Cerevisiae) CDCA3 Opa-Interacting Protein 5 Nucleolar and Spindle-Associated Protein1 Budding Uninhibited by Benzimidazoled 1 Homolog (yeast) NIMA-Related Kinase 2 Telomeric-Repeat Binding Factor (NIMA-Inetracting) 1 Kinesin Family Member 2C Chromosome 21 Open Reading Frame 45 CENP-F (Mitosin) Budding Uninhibited by Benzimidazoled 1 Homolog Beta (yeast) CENP-E M

itosis C

ell Divison

Cell C

ycle

Figure S2 a, Functional annotation clustering of AGO-bound E2F target genes downregulated in senescence. AGO-binding to downregulated E2F targets shows predilection for cell cycle control genes (colored in blue). Enrichment score is 21,04 and p-value < 0,01 using DAVID gene ontology tool (see also Supplementary Information, Table S2). b, Experimental design for small RNA sequence aquisition using Next-Generation Sequencing (Next-Generation-Seq) and bioinformatic analysis; RIP, RNA immunoprecipitation. c, Anti-Ago2 antibody (9E8.2) is AGO2-specific. Flag/HA(FH)-tagged Ago genes (FH-Ago1 through FH-Ago4) were expressed in Hela cells and cell lysates analysed by immunoblot using mouse monoclonal anti-AGO2 9E8.2 or anti-HA antibodies. d-f, AGO2 accumulates in nucleus of senescent

cancer cells. MCF-7 breast cancer cells were left untreated or treated with 1mM doxorubicin for 24 hrs (MCF-7). Subsequently, doxorubicin was washed out and cells were left unperturbed for 5 days after which cells aquired a senescence phenotype. Tetracyclin was washed out from TET-off Rb-inducible SAOS-2 osteosarcoma cancer cells (-TET) and cell cultures were replenished with fresh medium and left unperturbed for 5 days at which point cells underwent senescence. d, Localisation of AGO2 as determined by preparation of cytosolic and nuclear fractions from cell lysates of MCF-7 cells as described in Material and Methods. Indirect immunofluorescence of AGO2 in pre-extracted e, doxorubicin-induced MCF-7 (scale bar 20mm) and f, Rb-induced SAOS-2 senescent cells 5 days post-treatment; scale bar 15mm.




Nevus Melanoma


AGO2 macroH2A1.2 AGO2/macroH2A1.2 DAPI MERGE b.

c.

d.

e.

f.

g.

a.

Figure S3 Histopathological and indirect immunofluorescence characterisation of benign melanocytic nevi and malignant melanomas. a, Hematoxylin and eosin stained tissue section from a melanocytic nevus and malignant melanoma that were used for the AGO2 localisation

studies in Figures 2c and d. b-d, Nuclear colocalisation of AGO2 and macroH2A in melanocytic nevi (n=3) and e-f, cytosolic colocalisation in melanomas (n=3) as determined by indirect immunofluorescence; scale bar, 20mm.





b.

a.

c.

C S

0 0,2 0,4 0,6 0,8

1 1,2 1,4

Rel

. Exp

ress

ion

0

% o

f inp

ut

C S

0,1

0,2

0,3

0,4

0,5

-4 -3 -2 -0 +2 +5 +13

Infection Selection

siRNAs miRs

Antagomir Assays

Time (d)

0

0,4

0,8

1,2

1,6 siC siAGO2


Fold

-enr

icm

ent

d.

Figure S4 AGO2 targets promoters of repressed E2F-response genes in senescence. a, AGO2 qChIP analysis of indicated E2F target genes in pre-senescent, empty vector control (C) and RasV12-induced senescent (S) WI38 fibroblasts. b, mRNA transcript levels were measured by qRT-PCR in pre-senescent, empty vector control (C) and RasV12-induced senescent (S) WI38 fibroblasts for the indicated genes. c, Experimental

design and reference time frame for siRNA, miR and antagomir treatments of cells. d, Silencing of AGO2 diminishes its binding to E2F-target genes. AGO2 qChIP was performed in RasV12-induced senescent cells treated with scramble siRNA (siC) or siAGO2. Data are means ± s.d.; n=3; P < 0,05. Experiments were performed in duplicates.





H3K27me3 DAPI MERGE

AGO2 H3K9me3 DAPI MERGE a.

b.

siC siAGO2

Rel

. Gen

e E

xpre

ssio

n

0 0,5

1 1,5

2 2,5

3

USP1 NEK2 MYC FANCF CDC6 BUB1 BRCA2 ATM ASF1B ORC6L

3,5 c.

Figure S5 a, Partial Colocalisation of AGO2, H3K9me2 and H3K27me3 at periphery of SAHF. Indirect immunostaining for a, AGO2 and H3K9me2 and b, H3K27me3 in cells undergoing RASV12-induced senescence using DAPI as DNA counterstain; scale bars, 20mm c, Depletion of AGO2 derepresses E2F

target genes. Cells were retrovirally infected with pBABE-RasV12 and 2 days post-drug selection cells were transiently transfected with siScramble control (siC) or siAGO2 for 3 days followed by qRT-PCR for indicated genes. Data are means ± s.d.; n=3; P < 0,02. Experiments were performed in quadruplicates.




Ras

shA

GO

2-2

shA

GO

2-1

shC

Tub

RasV12

C


0 0,5

1 1,5

Rel

. AG

O2

Exp

ress

ion

AGO2

Tub

shA

GO

2-2

shA

GO

2-1

shC

a.

b.

c.

d.

0 0,5

1 1,5

Rel

. AG

O2

Exp

ress

ion

Figure S6 Depletion of AGO2 delays the onset of RasV12-induced and replicative senescence. a, Immunoblot analysis of lysates from empty vector control cells (C), RasV12-infected cells expressing shControl (shC) or the indicated shAGO2s (as used in Figs. 3e-g) probed for Ras and b-tubulin (Tub) at day 14 of life span study. (b-c), Stable down-regulation of AGO2 by shRNAs expression in cells undergoing RasV12-induced senescence (used

in Figs. 3e-g) as determined by b, qRT-PCR and c, Western blot at day 14 of life span study. Tubulin (Tub) is used as loading control. d, Relative gene expression of AGO2 in control (shC) and AGO2-depleted (shAGO2-1/2) cell populations undergoing replicative senescence as determined by qRT-PCR. Data are means ± s.d.; n=3; P < 0,05. Experiments were performed in duplicates.




p-value FDR

Geom mean of intensities

Control

Geom mean of intensities

Senescence

Ratio geom means S/C Unique id

0,02 0,21 272,5 1671,4 6,1 hsa-let-7f-2-prec2

0,003 0,16 1107,6 4248,7 3,8 hsa-let-7a-1-prec

0,005 0,16 1083 3848,1 3,5 hsa-let-7c-prec

0,006 0,16 1268,8 4447,1 3,5 hsa-let-7a-2-precNo2

0,007 0,16 846,6 2947,1 3,4 hsa-let-7g-precNo1

0,004 0,16 766,7 2565,8 3,3 hsa-let-7d-prec

0,008 0,16 367,7 1106,5 3,0 hsa-let-7d-v1-prec

0,009 0,16 342,1 983,1 2,8 hsa-let-7f-1-precNo2

0,01 0,17 846,2 2407,4 2,2 hsa-let-7a-3-prec

0,03 0,25 1048,7 2428,9 2,3 hsa-let-7iNo1

0,04 0,34 367,7 765,7 2,0 hsa-let-7d-v1-prec

0,005 0,14 342,1 677,1 1,9 hsa-let-7f-1-precNo2

0,01 0,31 4672,2 3325,7 0,7 hsa-let-7b-prec

0,01 0,17 789,5 276,3 0,35 hsa-let-7d-v2-precNo1

0,006 0,14 789,5 193,9 0,24 hsa-let-7d-v2-precNo1

0,02 0,25 952,3 186,1 0,19 hsa-let-7iNo2


CDC2 -450 Chr 10q21

TSS +1

Luciferase -450 let-7 Wildtype

+51

-410 -388

E2F

-143 -138

Luciferase -386 let-7

+51

E2F

b.

c.

d.

e.

0 1 2 3 4 5 6

C Rb Rb-let7 Rb-Ago2 Rb-Ago2-let7

Fold

-Rep

ress

ion

1 1,1 1,82 1,47

2,97 1,91

4,22

1,91

5,11

2,29

let-7 wt

1,53 1,25

let-7

a.

Figure S7 a, Expression profile of let-7 family members. Total RNA was prepared from RasV12-induced senescent and pre-senescent, control WI38 fibroblasts and hybridised to microarray as described in experimental procedures. b, MiR-let-7 corepresses the E2F-responsive CDC2 promoter-reporter gene together with AGO2 and Rb. b, CDC2 promoter-luciferase reporter assays were performed in C33A cells using either a wild-type version of the CDC2 promoter or a derivative there-of in which the putative let-7 site is deleted. Promoter-reporter was stimulated by 10ng

of E2F1/DP1 expression. Rb was used at 300ng, AGO2 at 100ng and let-7 at 100nM. Data are means ± s.d.; n=3; P < 0,05. Experiments were performed in triplicates. Expression of proteins was confirmed by Western blot (data not shown). c, Genomic locus of the human CDC2 gene and its upstream promoter fragment used in this study. (d-e), Schematic representation of the pGL3 luciferase CDC2 promoter reporters d, containing the let-7 site or e, deleted for the let-7 site. TSS, transcriptional start-site.




f.

% p

ositi

ve c

ells

100 80 60 40 20

Ki67 SA-β-Gal

C Pre-miR-let-7f

a.

0

0,5

1

1,5 mRNA

C Pre-mir-let-7f

CDCA8

CDC2

Fold

Enr

ichm

ent

b.

H3K27me3

C Pre-miR-let-7f 0 1 2 3 4 5 AGO2

C Pre-mir-let-7f

CDCA8

CDC2

Fold

Enr

ichm

ent

g.

100

200

300

0

400

Cel

l Num

ber x

103

siC siC+siAGO2 let-7+siAGO2 let-7+siC

e.

0

100

200

0 5 10

C

pre-miR-let-7f

Cel

l Num

ber x

103

300

time (days)


0

0,5

1

1,5 Biotin-let-7 ChOP

CDCA8 CDC6

Fold

-enr

ichm

ent

d.

siCDCA8

Biotin-miR-185 qChIP-ChOP

0 1 2 3 4

1. α-Control 2. SA

1. α-AGO2 2. SA

c.

Fold

Enr

ichm

ent

Sense ORC6L

ORC6L * miR-target -257/236 -38/31

AGGAGGGAAAGGCAGTGTGTGA!

TGGAGAGAAAGGCAGTTCCTGA!

Figure S8 AGO2 and pre-miR-let-7f induce TGS of E2F target promoters. Exogenous pre-miR-let-7f or control pre-miR (C) was transiently transfected at 100nM into pre-senescent WI38 fibroblasts. a, Abundance of CDCA8 and CDC2 mRNAs was measured by qRT-PCR of total RNA prepared from the respective samples. Data are means ± s.d.; n=3; P < 0,05. Experiments were performed in quadruplicates. b, qChIP was performed in pre-miR-let-7f-treated cells using anti-AGO2 and anti-H3K27me3 antibodies followed by qPCR using primers amplifying CDCA8 and CDC2 promoter regions. Data are means ± s.d.; n=3; P < 0,05. Experiments were performed in quadruplicates. c, Simultaneous presence of miR-185 and AGO2 on ORC6L promoter was determined by combined qChIP-ChOP assay. Biotin-labeled miR-185 (100nM) was transfected into pre-senescent WI38 fibroblasts followed by successive rounds of AGO2-ChIP or control-ChIP and streptavidin-affinity (SA) precipitation. SA-chromatin-precipitates were analyzed by qPCR with primers detecting E2F-responsive promoter ORC6L. Experiment was performed in quadruplicates and the s.d. is indicated. Also shown is the alignment of miR-185 to ORC6L target promoter. Identical nucleotides are shown in red.

Asterisk, transcription start site (TSS); E2F-binding site is depicted as grey box with respect to TSS ; Arrows to the right and left of miR-target are primers used for qChIP or qRT-PCR; numbers on top of indicate miR target site position with respect to TSS. d, Biotin-labeled let-7f (100nM) was transfected into pre-senescent WI38 fibroblasts followed by streptavidin-affinity (SA) precipitation. SA-chromatin-precipitates were analyzed by qPCR with primers detecting E2F-responsive promoters CDCA8 and CDC6. Data are means ± s.d.; n=3; P < 0,05. Experiments were performed in duplicates. (e-f), Exogenous pre-miR-let-7f induces cellular senescence. Pre-senescent WI38 fibroblasts were transiently transfected with 100nM pre-miR-let-7f, control pre-miR (C) or siCDCA8 and cell proliferation was determined at the indicated time points. Experiments were performed in duplicates. e, Growth curve and f, Number of cells staining positive for Ki67 and senescence-associated beta-galactosidase (SA-b-Gal) at day 7 post-treatment. Data are means ± s.d.; n=3; P < 0,05. Experiments were performed in duplicates. g, Cell number of pre-senescent WI38 fibroblasts 96 hrs after cotransfection with indicated siC, siAGO2 and let-7 combinations. Shown is the mean of quadruplicates.




AGO2

Histone H3

AGO2

Lamin B

GAPDH

Supplementary_Information_ Fig_S9: Full scans of Figures

Figure 2a

Figure 2e

Figure 3e

AGO2

Rb

RB AGO2

Figure 3b

HDAC1

AGO2

RB

55 72 95

55

95

55 72 95

55

72 95

Figure 3a

55

72 95

55

72 95

AGO2

Lamin B

GAPDH

Supplementary Figure S2b

Ras

Tub

Supplementary Figure S6a

AGO2

Tub

Supplementary Figure S6c

55

16,5

32

55

72 95

55

72

16,,5

25 32

55

72 95

55

72

95

55

72 95

55

72

95

14

32 32

55

72 95

Figure S9 Full scans




Supplementary Tables

Supplementary Table S1: List of genome-wide AGO2-bound- and AGO2-bound E2F target promoters in pre-senescent, control and oncogenic RASV12-induced senescent cells and their respective expression statuses (up-down; only E2F-targets are shown).

Supplementary Table S2. Annotated gene ontology analysis using DAVID. Depicted is the functional annotation cluster including cell cycle, mitosis and cell division for AGO-bound E2F target genes repressed in senescence showing the highest enrichment score.

Supplementary Table S3. Summary of ILLUMINA Next-Generation Sequencing (NGS). Usable sequence reads are reads after trimming of adapter and selection of reads greater than 17 nucleotides.

Supplementary Table S4. Summary of miRs identified by NGS for all samples analysed.

Supplementary Table S5. Top 10 AGO and H3K9me2-bound miRs in senescent cells. Next-generation sequencing was performed on small RNAs purified by AGO2- and H3K9me2 immunoprecipitation (RIP) from RASV12-induced senescent cells. Shown are cpm obtained for the respective top 10 miRs.

Supplementary Table S6. List of oligonucleotides used in this study.


Documents

Senescence is an endogenous trigger for microRNA-directed ...fulltext.calis.edu.cn/nature/ncb/14/3/ncb2443.pdf · followed by qPCR of CDC2, CDCA8, PCNA and CCNA2 promoters. Data are