LETTERS

Control of female gamete formation by a small RNApathway in ArabidopsisVianey Olmedo-Monfil1, Noe Duran-Figueroa1, Mario Arteaga-Vazquez1{, Edgar Demesa-Arevalo1,Daphne Autran2, Daniel Grimanelli2, R. Keith Slotkin3{, Robert A. Martienssen3 & Jean-Philippe Vielle-Calzada1

In the ovules of most sexual flowering plants female gametogenesis isinitiated from a single surviving gametic cell, the functional mega-spore, formed after meiosis of the somatically derived megasporemother cell (MMC)1,2. Because some mutants and certain sexual spe-cies exhibit more than one MMC2–4, and many others are able to formgametes without meiosis (by apomixis)5, it has been suggested thatsomatic cells in the ovule are competent to respond to a local signallikely to have an important function in determination6. Here we showthat the Arabidopsis protein ARGONAUTE 9 (AGO9) controlsfemale gamete formation by restricting the specification of gameto-phyte precursors in a dosage-dependent, non-cell-autonomous man-ner. Mutations in AGO9 lead to the differentiation of multiplegametic cells that are able to initiate gametogenesis. The AGO9 pro-tein is not expressed in the gamete lineage; instead, it is expressed incytoplasmic foci of somatic companion cells. Mutations inSUPPRESSOR OF GENE SILENCING 3 and RNA-DEPENDENTRNA POLYMERASE 6 exhibit an identical defect to ago9 mutants,indicating that the movement of small RNA (sRNAs) silencing out ofsomatic companion cells is necessary for controlling the specificationof gametic cells. AGO9 preferentially interacts with 24-nucleotide

sRNAs derived from transposable elements (TEs), and its activity isnecessary to silence TEs in female gametes and their accessory cells.Our results show that AGO9-dependent sRNA silencing is crucial tospecify cell fate in the Arabidopsis ovule, and that epigenetic repro-gramming in companion cells is necessary for sRNA–dependentsilencing in plant gametes.

Large-scale transcriptional analysis indicated that a gene encodingan ARGONAUTE (AGO) protein (At5g21150 or ARGONAUTE 9) ishighly expressed in ovules and anthers of Arabidopsis (SupplementaryFig. 1). The in situ pattern of expression confirmed that AGO9 mes-senger RNA is localized in ovules throughout development(Supplementary Fig. 2). In both plants and animals, AGO proteinsare known to cleave endogenous mRNAs during either microRNA(miRNA) or short interfering RNA (siRNA)-guided post-transcrip-tional silencing7–9. To elucidate the function of AGO9 in Arabidopsis,individuals from three independent insertional lines harbouringT-DNA elements within the coding region of the AGO9 gene werephenotypically analysed at all stages of ovule development10 (Fig. 1aand Table 1). Whereas 94.2% of pre-meiotic ovules showed a singleMMC in wild-type plants (Fig. 1b and Supplementary Fig. 3), 5.8%

1Grupo de Desarrollo Reproductivo y Apomixis, Laboratorio Nacional de Genomica para la Biodiversidad y Departamento de Ingenierıa Genetica de Plantas, Cinvestav IrapuatoCP36500 Guanajuato, Mexico. 2Institut de Recherche pour le Developpement, UMR 5096, 34394 Montpellier, France. 3Cold Spring Harbor Laboratory, Cold Spring Harbor, New York11724, USA. {Present addresses: BIO5 Institute, University of Arizona, Arizona 85721-0240, USA (M.A.-V.); Department of Molecular Genetics, The Ohio State University, Columbus,Ohio 43210, USA (R.K.S.).

ago9-3ago9-4ago9-2

Ex18 Ex22 +5326

PAZ domain PIWI domain

a

b c ed

5′UTR 3′UTR

Figure 1 | Phenotypes of ago9 insertional mutants before meiosis.a, Genomic structure of the AGO9 gene in Arabidopsis; the location ofT-DNA insertions and the gene length (nucleotides) are indicated. UTR,untranslated region. b, Pre-meiotic wild-type (WT) ovules showing a singlesubepidermal MMC, adjacent to L1 cells. c, Pre-meiotic wild-type ovule

showing two MMCs. d, Pre-meiotic ago9-2 mutant ovule showing two larger(black arrows) and two smaller (white arrows) abnormal cells. e, Pre-meioticago9-3 ovule showing abnormally enlarged cells (arrows); one of them hasinitiated a nuclear division. Scale bars, 10 mm.

Vol 464 | 25 March 2010 | doi:10.1038/nature08828

628Macmillan Publishers Limited. All rights reserved©2010

exhibited two MMCs (Fig. 1c); however, only one of the latter under-went gametogenesis as twin female gametophytes were never observed.All ago9 insertional lines were fertile and did not show signs of ovule orseed abortion; however, in contrast to wild-type plants, the pre-meioticovule primordia of heterozygous ago9/1 individuals—including alleleago9-2 that was previously reported as having no defective pheno-type11—showed several abnormally enlarged sub-epidermal cells(Fig. 1d, e). In ago9/1 individuals, the ovules exhibited up to six cellscontaining a conspicuous nucleus and nucleolus at a frequency of30.29%, indicating that ago9 alleles are dominant and affect early celldifferentiation in the developing ovule. In homozygous ago9/ago9 indi-viduals, the percentage of ovule primordia showing more than oneenlarged cell was of 37.16% to 47.7%, depending on the allelic variant(Table 1). Triploid (3n) individuals that had two wild-type and onemutant ago9-3 allele showed 14.11% to 23.49% of abnormal ovules, avalue intermediate between diploid plants carrying a single ago9-3allele and wild type (Table 1). These results indicate that a dosage-dependent mechanism is responsible for the mutant ago9 phenotype.

No molecular marker exclusively expressed in the MMC has beenreported, but the pattern of callose deposition is a reliable method todetermine cell identity at pre-meiotic stages12. To determine whetherone or several of the enlarged cells present in ago9-3 ovules are capable

of undergoing meiosis, we analysed callose deposition in wild-typeand homozygous ago9-3 ovules. In agreement with previous descrip-tions, wild-type ovules showed patches of callose in the MMC beforethe initiation of meiosis (Fig. 2a). After meiosis, callose was depositedin transverse walls between the functional megaspore and its de-generated sister cells (Fig. 2b). In pre-meiotic ago9-3 ovules, less than10% of abnormally enlarged cells showed patches of callose deposits(Fig. 2c, d). During meiosis, callose was only detected in the inter-mediate walls of a single cell and the degenerated neighbouring cells,but not in the closely associated abnormally enlarged cells (Fig. 2e, f).This pattern persisted following meiosis (Fig. 2g, h). These results showthat several enlarged cells differentiate before meiosis in ago9-3 ovules,but that a single one undergoes meiosis and gives rise to a functionalhaploid megaspore, indicating that the activity of AGO9 is necessary torestrict differentiation to a single sub-epidermal cell in the pre-meioticovule.

After meiosis, ago9-3 ovules showed persistent enlarged cells adjacentto meiotic products, including the three degenerated megaspores andthe functional megaspore (Fig. 2i–k). To determine the identity andassess the developmental potential of extranumerary enlarged cells inmutant ovules, we examined the expression of pFM2, a markerexpressed in the functional megaspore and the developing femalegametophyte, but not in the MMC or in the three meiotically-deriveddegenerated megaspores (Fig. 2l, m). In ago9-3 ovules, pFM2 expres-sion was initially observed following meiosis in the functionalmegaspore, but also in a cluster of adjacent cells that forms the nucellusand includes the abnormal gamete precursors (Fig. 2n). In all ago9-3ovules observed, more than four cells showed strong reporter geneexpression (b-glucuronidase or GUS) at post-meiotic stages, indicatingthat at least some of the cells that express pFM2 have a somatic origin.pFM2 expression was absent at pre-meiotic stages, indicating that defec-tive ago9-3 individuals differentiate other cells that persist in the

Table 1 | Genetic analysis of insertional ago9 mutants in Arabidopsis.

Allele Genotype Single MMC Abnormally enlarged cells

ago9-3 ago9-3/ago9-3 208 123 (37.16%)ago9-3/1 214 93 (30.29%)ago9-3m/1p/1p

286 47 (14.11%)1m/1m/ago9-3p

241 74 (23.49%)ago9-4 ago9-4/ago9-4 139 118 (45.9%)ago9-2 ago9-2/ago9-2 162 148 (47.7%)Wild type 1/1 292 18 (5.8%)

a b dc e f g h

i

o p

j k l m n

Figure 2 | Meiotic and post-meiotic phenotype of the ago9 mutant.a–h, Callose deposition (a, b, d, f, h) and morphology (c, e, g) of wild-typeand ago9-3 ovules. i–k, ago9-3 post-meiotic ovules showing abnormalgametic cells (AGC) adjacent to degenerated (asterisk and DM) andfunctional (FM) megaspores. l, Absence of pFM2-driven GUS expression ina pre-meiotic wild-type ovule. m, pFM2-driven GUS expression in the

functional megaspore (dashed) of a post-meiotic wild-type ovule. n, pFM2-driven GUS expression in the functional megaspore and adjacent cells of apost-meiotic ago9-3 ovule. o, ago9-3 ovule containing a 2-nuclear (2NFG)and a 1-nuclear (1NFG) female gametophyte. p, pFM2-driven GUSexpression in an ago9-3 ovule containing two female gametophytes. Scalebars: a–h, 5mm; i–n, 10 mm; o and p, 25mm.

NATURE | Vol 464 | 25 March 2010 LETTERS

629Macmillan Publishers Limited. All rights reserved©2010

developing ovule adjacent to the meiotic products and subsequentlyacquire a functional megaspore identity without undergoing meiosis.At subsequent stages of development, ago9-3 individuals exhibited anunusual phenotype of two independent female gametophytes develop-ing in the same ovule at a frequency of 44.03% (n 5 243; Fig. 2o).Crosses of ago9-3 plants with individuals expressing the pFM1 (ref. 13)or pFM2 marker showed that both acquire a female gametophyteidentity (Fig. 2p and Supplementary Fig. 4). These results indicatethat abnormal somatic cells are able to differentiate into gametic cellsand initiate gametogenesis without undergoing meiosis.

Immunoblots hybridized with a polyclonal antibody against AGO9detected a protein of the expected 100.5 kDa size in developing wild-type gynoecia but not in 1-week-old seedlings, developing rosetteleaves or developing siliques (Fig. 3a). Immunolocalizations showedthat the AGO9 protein was initially expressed in somatic cells of theepidermal (L1) layer located in the apical region of the pre-meioticovule, but not in the MMC (Fig. 3b). Interestingly, we observed AGO9in cytoplasmic foci reminiscent of P-bodies or stress granules presentin the cytoplasm of animal cells (Fig. 3c–e). AGO9 did not localize inthe haploid megaspores or the developing female gametophyte beforeof after cellularization. In ovules containing a female gametophyte atthe four-nuclear stage, AGO9 was localized in the outer integumentarycells, but also in the periphery of the endothelium, at the sporophyte-gametophyte cellular boundary (Fig. 3f). In anthers, AGO9 was loca-lized in the cytoplasm of microsporocytes following meiosis, and laterin the cytoplasm of the vegetative cell but not in the sperm cells(Supplementary Fig. 5a–d). Ovules or pollen of ago9-3 individualsdid not show AGO9 expression (Supplementary Fig. 5d, e), confirm-ing that the antibody exclusively recognized AGO9. Overall, theseresults indicate that AGO9 is preferentially expressed in reproductivecompanion cells but not in the associated male or female gametes ortheir precursors.

In Arabidopsis, trans-acting siRNAs (ta-siRNAs) are known tomove as signal molecules and cause gene silencing beyond their cel-lular sites of initiation14–16. Their biogenesis depends on transcriptionby RNA-DEPENDENT RNA POLYMERASE 6 (RDR6) that convertstheir single-stranded RNA precursors into double-stranded RNA in apathway that is also dependent on the function of the putative RNAbinding protein SUPRESSOR OF GENE SILENCING 3 (SGS3)17,18.The extent of gene silencing movement outside their site of initiationalso depends on the activity of RDR6 (ref. 19). To determine whetherthe function of AGO9 could be associated with a non-cell-autonom-ous pathway, we examined ovule development in homozygous sgs3-11 and rdr6-11 individuals. Although both sgs3 and rdr6 mutantsshow seedling and floral defects characterized by leaf curling andlimited stamen elongation17, their possible role during gamete forma-tion has not been investigated. Both sgs3-11 and rdr6-11 plantsshowed an identical phenotype to ago9 mutants with additional gam-etic cells differentiating in the pre-meiotic ovule (Fig. 4a–d). In rdr6-11 plants, post-meiotic ovules showed two independently developingfemale gametophytes at a frequency of 43.3% (n 5 224). Crosses of

rdr6-11 plants to individuals expressing the pFM2 marker indicatethat both acquire a female gametophyte identity (Fig. 4e). Theseresults support the hypothesis that AGO9 controls gametic cell com-mitment by acting in a non-cell-autonomous sRNA-dependent path-way in the developing ovule of Arabidopsis.

To identify the nature of AGO9-associated sRNAs, wild-typedeveloping gynoecia were isolated and used for total protein extrac-tion, immunoprecipitation with the AGO9 antibody, and elution ofthe associated sRNA fraction. After sequencing, 2,508 sRNA sequences(98% of total) could be mapped to the Arabidopsis nuclear genome andcategorized based on their location and function (Supplemen-tary Tables 1 and 2). Although most are 24 nucleotides in length(79.1%), 8.9% are 21 to 22 nucleotides long. Most 24-nucleotidesequences derive from TEs belonging to distinct families of retrotran-sposons: Gypsy (23%) Athila (9.3%), CACTA (5.5%), and less fre-quently LINE or Mutator. All sequences mapping to Gypsy TEsbelong to the AtGP1 sub-family, and 3% of all sequences mappingto retrotransposons correspond to siRNAs shown to be dependenton RNA polymerase IV (PolIV) for their biogenesis20. A further17.4% of the total maps to genomic signatures assigned to other fam-ilies containing nested components of Gypsy, Athila or CACTA TEs. Incontrast, 21-nucleotide sRNAs preferentially derive from previouslycharacterized miRNAs (3.2%), including MIR167 that is known toact in the ovule21, and protein-coding genes (14.5%). These resultsshow that primary targets of AGO9-dependent silencing in the ovuleof Arabidopsis are TEs.

Previous studies have shown that some TEs that are active in maturepollen grains are not expressed in developing or fully differentiatedovules of Arabidopsis22. To determine whether AGO9 is necessary forthe inactivation of these TEs in the ovule, we crossed lines containingenhancer traps that tagged specific TEs to homozygous ago9 indivi-duals. In agreement with previous results, no GUS expression wasobserved in the ovule of enhancer trap lines present in a wild-typegenetic background (Fig. 4f). By contrast, heterozygous ago9/1 indi-viduals containing an enhancer trap within either an Athila, LINE orAtlantys retrotransposon showed strong GUS staining in the egg andsynergid cells of the mature female gametophyte before pollination(Fig. 4g–i). These results not only confirm that AGO9 is necessary forTE inactivation in the ovule, but also show that one of its targets is theegg and synergid cells (the egg apparatus) before fertilization.

The ago9 phenotype was also identified in homozygous mutantsfor RNA-dependent RNA polymerase 2 (rdr2), dicer-like 3 (dcl3), andthe double mutant nrpd1a nrpd1b that is defective in the activity ofboth polymerase IV and polymerase V, but not in dicer-like 1 anddicer-like 4 (dcl4) that are essential for the generation of miRNAs andta-siRNAs, respectively, indicating that AGO9-dependent TE inac-tivation restricts female gametogenesis to a single gametic cellthrough an endogenous 24-nucleotide siRNA biosynthetic pathway23

(Supplementary Figs 6 and 7). The consistent identification of asingle cell undergoing meiosis and several cells acquiring a functionalmegaspore identity in the post-meiotic ovule, combined to the

a b d e fc1 2 3 4

130kDa

95

13095725543

FM

DML1

Figure 3 | AGO9 protein expression in developing ovules. a, Immunoblotanalysis of AGO9 in wild-type seedlings (1), developing gynoecia (2) leaves(3), and 7-day-old siliques (4). b–d, AGO9 is expressed in cytoplasmic foci(green) of companion cells but absent from the MMC (outlined in b) or thefunctional megaspore (arrow in nuclei are counterstained with

49,6-diamidino-2-phenylindole (DAPI) (b and d). e, Diagram showingAGO9 localization (green) in a wild-type ovule at the end of meiosis. f, AGO9expression in companion somatic cells but not within a 4-nuclear femalegametophyte (FG). Scale bars: b, 5 mm; c and d, 10 mm; f, 20 mm.

LETTERS NATURE | Vol 464 | 25 March 2010

630Macmillan Publishers Limited. All rights reserved©2010

presence of two developing female gametophytes separated by severalsomatic cells, provides strong evidence for the initiation of femalegametophytes from two non-sister cells, one of which is somatic inorigin.

By preferentially interacting with sRNAs derived from TEs andsilencing their activity in the female gametophyte, the function ofAGO9 is reminiscent of the PIWI subclass of ARGONAUTE proteinsthat are necessary to maintain transposon silencing in the germlinegenome of invertebrates and mammals24. Some maternal siRNAsequences found in the endosperm20 and 24-nucleotide siRNA foundin pollen22 resemble AGO9-interacting sRNAs, raising the possibilitythat AGO9 may also contribute to these populations in a non-autonomous way. The ago9 mutant phenotype is reminiscent ofapospory, a component of asexual reproduction through seeds (apo-mixis) prevailing in many flowering species that produce unreducedfemale gametes from somatic cells5. Our findings provide opportun-ities to investigate the genetic basis and molecular mechanisms thatcontrol cell fate, offering new possibilities to explore the epigeneticinduction of apomixis in sexual plants.

METHODS SUMMARYMaterial and growth conditions. We used Arabidopsis thaliana of ecotype

Columbia 0 (Col-0). Insertional mutant lines were ago9-2 (SALK_112059),

ago9-3 (SAIL_34_G10) and ago9-4 (SAIL_260_A03) (ago9-1 showed an identical

phenotype but was not quantified). Seeds were sterilized with 100% ethanol andgerminated under stable long day (16 h light/8 h dark) conditions at 22 uC.

Seedlings were planted and grown under controlled greenhouse conditions

(24 uC). For a detailed description of mutant stocks, enhancer trap lines, transgenic

lines and DNA constructs, see Methods.

Histological analysis. Cleared ovules and histochemical GUS analysis was per-

formed as described25. Callose analysis was performed as described26 with minor

modifications described in Methods.

Immunoblot and immunoprecipitation. Amino acids N-SSRNHAGNDTNDA

DRK were used to generate a specific AGO9 antibody (Invitrogen). Immuno-

purification of AGO9–sRNA complexes was performed as described27, with

modifications described in Methods.

Cloning and genomic analysis of small RNAs. After sequencing, sRNA reads

were filtered and sequences were mapped to the Arabidopsis genome (http://

www.arabidopsis.org). Details of the sRNA annotation procedure are provided

in Methods.

Full Methods and any associated references are available in the online version ofthe paper at www.nature.com/nature.

Received 16 September 2009; accepted 11 January 2010.Published online 7 March; corrected 25 March 2010 (see full-text HTML version fordetails).

1. Walbot, V. & Evans, M. M. S. Unique features of the plant life cycle and theirconsequences. Nature Rev. Genet. 4, 369–379 (2003).

2. Maheswari, P. An Introduction to the Embryology of the Angiosperms (McGraw-Hill,1950).

3. Sheridan, W. F., Avalkina, N. A., Shamrov, I. I., Batygina, T. B. & Golubovskaya, I. N.The mac1 gene: controlling the commitment to the meiotic pathway in maize.Genetics 142, 1009–1020 (1996).

4. Nonomura, K. et al. The MSP1 gene is necessary to restrict the number of cellsentering into male and female sporogenesis and to initiate anther wall formationin rice. Plant Cell 15, 1728–1739 (2003).

5. Bicknell, R. A. & Koltunow, A. M. Understanding apomixis: recent advances andremaining conundrums. Plant Cell 16, S228–S245 (2004).

6. Grossniklaus, U. & Schneitz, K. The molecular and genetic basis of ovule andmegagametophyte development. Semin. Cell Dev. Biol. 9, 227–238 (1998).

7. Baumberger, N. & Baulcombe, D. C. Arabidopsis ARGONAUTE1 is an RNA Slicerthat selectively recruits microRNAs and short interfering RNAs. Proc. Natl Acad.Sci. USA 102, 11928–11933 (2005).

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

9. Pham, J. W., Pellino, J. L., Lee, Y. S., Carthew, R. W. & Sontheimer, E. J. A. Dicer-2-dependent 80s complex cleaves targeted mRNAs during RNAi in Drosophila. Cell117, 83–94 (2004).

10. Alonso, J. M. et al. Genome-wide insertional mutagenesis of Arabidopsis thaliana.Science 301, 653–657 (2003).

11. Takeda, A., Iwasaki, S., Watanabe, T., Utsumi, M. & Watanabe, Y. The mechanismselecting the guide strand from small RNA duplexes is different among Argonauteproteins. Plant Cell Physiol. 49, 493–500 (2008).

12. Webb, M. C. & Gunning, B. E. S. Embryo sac development in Arabidopsis thaliana. I.Megasporogenesis, including the microtubular cytoskeleton. Sex. Plant Reprod. 3,244–256 (1990).

13. Huanca-Mamani, W., Garcia-Aguilar, M., Leon-Martınez, G., Grossniklaus, U. &Vielle-Calzada, J. P. CHR11, a chromatin-remodeling factor essential for nuclearproliferation during female gametogenesis in Arabidopsis thaliana. Proc. Natl Acad.Sci. USA 102, 17231–17236 (2005).

14. Chitwood, D. H. et al. Pattern formation via small RNA mobility. Genes Dev. 23,549–554 (2009).

15. Schwab, R. et al. Endogenous tasiRNAs mediate non-cell autonomous effects ongene regulation in Arabidopsis thaliana. PLoS One 4, e5980 (2009).

16. Voinnet, O. Non-cell autonomous RNA silencing. FEBS Lett. 579, 5858–5871(2005).

a

f g h i

b c d e

Figure 4 | Phenotype of the rdr6 mutant and activation of transposableelements in the ago9 mutant. a–d, Ovules of rdr6-11 or sgs3-11 showingabnormal gametic cells adjacent to the functional megaspore and two femalegametophytes (arrows) separated by the L1 cell layer or degenerated cells(arrowheads). e, pFM2-driven GUS expression in a rdr6-11 ovule.

f, Enhancer trap ET13889 inserted in a AtLINE3 transposon shows no GUSexpression in mature wild-type ovules. g–i, GUS expression conferred byenhancer traps ET13889 (g), ET11075 (h) and ET10306 (i) in the eggapparatus of ago9-3/1 ovules. Scale bars: a, 10mm; b–i, 25 mm.

NATURE | Vol 464 | 25 March 2010 LETTERS

631Macmillan Publishers Limited. All rights reserved©2010

17. Peragine, A., Yoshikawa, M., Wu, G., Albrecht, H. L. & Poethig, R. S. SGS3 andSGS2/SDE1/RDR6 are required for juvenile development and the production oftrans-acting siRNAs in Arabidopsis. Genes Dev. 18, 2368–2379 (2004).

18. Yoshikawa, M., Peragine, A., Park, M. Y. & Poethig, R. S. A pathway for thebiogenesis of trans-acting siRNAs in Arabidopsis. Genes Dev. 19, 2164–2175 (2005).

19. Himber, C., Dunoyer, P., Moissiard, G., Ritzenthaler, C. & Voinnet, O. Transitivity-dependent and -independent cell-to-cell movement of RNA silencing. EMBO J. 22,4523–4533 (2003).

20. Mosher, R. A. et al. Uniparental expression of PolIV-dependent siRNAs indeveloping endosperm of Arabidopsis. Nature 460, 283–286 (2009).

21. Wu, M. F., Tian, Q. & Reed, J. W. Arabidopsis microRNA167 controls patterns ofARF6 and ARF8 expression, and regulates both female and male reproduction.Development 133, 4211–4218 (2006).

22. Slotkin, R. K. et al. Epigenetic reprogramming and small RNA silencing oftransposable elements in pollen. Cell 136, 461–472 (2009).

23. Chen, X. Small RNAs and their roles in plant development. Annu. Rev. Cell Dev. Biol.25, 21–44 (2009).

24. Klattenhoff, C. & Theurkauf W.. Biogenesis and germline functions of piRNAs.Development 135, 3–9 (2008).

25. Vielle-Calzada, J. P., Baskar, R. & Grossniklaus, U. Delayed activation of thepaternal genome during seed development. Nature 404, 91–94 (2000).

26. Siddiqi, I., Ganesh, G., Grossniklaus, U. & Subbiah, V. The dyad gene is required forprogression through female meiosis in Arabidopsis. Development 127, 197–207(2000).

27. Qi, Y., Denli, A. M. & Hannon, G. J. Biochemical specialization within ArabidopsisRNA silencing pathways. Mol. Cell 19, 421–428 (2005).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements We thank N. Sanchez for sharing the pFM2 marker,E. Demunck for technical assistance during cloning and sequencing, S. Poethig,J. Carrington, T. Lagrange and the Arabidopsis Stock Center for providing mutants,J. Mendiola and C. Alvarez for help with genetic and bioinformatic analysis, andR. Jorgensen for critically reading the manuscript. This work was supported byIRD-France and ANR (D.A. and D.G.), NIH and NSF (R.K.S. and R.A.M), ConsejoNacional de Ciencia y Tecnologıa (V.O.-M., N.D.-F., M.A.-V., E.D.-A. andJ.-P.V.-C.), Consejo Estatal de Ciencia y Tecnologıa de Guanajuato (J.-P.V.-C.), andthe Howard Hughes Medical Institute (J.-P.V.-C.).

Author Contributions J.-P.V.-C. and V.O.-M. designed the research, V.O.-M.generated the phenotypic analysis, performed the histological and expressionanalysis, and conducted the genetic experiments, N.D.-F. designed the antibodyand performed the immunoprecipitations and sRNA analysis, M.A.-V. conductedthe bioinformatic expression analysis, E.D.-A. performed immunolocalizationexperiments, D.G. contributed ideas and performed immunolocalizationexperiments, R.K.S. and D.A. provided unpublished materials, R.A.M. contributedideas and J.-P.V.-C. wrote the paper.

Author Information sRNA sequences are deposited in the EMBL NucleotideSequence Database (FN649764 to FN650107). Reprints and permissionsinformation is available at www.nature.com/reprints. The authors declare nocompeting financial interests. Correspondence and requests for materials shouldbe addressed to J.P.V.-C. (vielle@ira.cinvestav.mx).

LETTERS NATURE | Vol 464 | 25 March 2010

632Macmillan Publishers Limited. All rights reserved©2010

METHODSPlant material and growth conditions. We used A. thaliana ecotype Columbia

(Col-0) for wild-type plants, chemical homozygous mutant sgs3-11 and inser-

tional lines CS24285 (rdr6-11), SALK005512 (dcl3-1) (ref. 28), GABI160G05

(dcl4-2) (ref. 29), SAIL-1277H08 (rdr2-1) (ref. 28), a double mutant nrpd1a-2

nrpd1b-11 (SALK_128428 and SALK_029919, respectively), SAIL_34_G10

(ago9-3), SAIL_260_A03 (ago9-4), SALK_112059 (ago9-2) and A. thaliana eco-

type Landsberg erecta (Ler) for Enhancer Trap lines ET13889, ET11075 and

ET10306. Seeds were surface-sterilized by washing three times with 100%

ethanol and plated on Murashige and Skoog (MS) medium. The pFM2 plasmid

construction was generated by amplifying the pFM2 genomic regulatory region

using primers 59-GCGTGACACGCCACTACAACACACCAA-39 (sense) and

59-GCGGATCCAGGAAGCCATCGTCAGACAG-39 (antisense); a 564-base-

pair (bp) genomic fragment was subsequently cloned in front of the uidA gene

using the pBI101.2 plasmid. Transformation was in Col-0. In all cases MS med-

ium plates containing seeds were placed in full darkness for 3 days at 4 uC, and

subsequently germinated in a growth chamber at 22 uC under a 16 h light/8 h

dark photoperiod, transferred to soil, and grown in the greenhouse under long-

day 16 h light/8 h dark controlled conditions.

In situ hybridization. A specific 149-bp fragment corresponding to the AGO9

39UTR was PCR amplified by using primers ago9isS2 (59-TCCAGTCCAC

ACGATAGCT-39) and ago9isAS2 (59-ATTCTGTCGGTTTTTGTGGG-39) and

cloned in TOPO-PCRII (Invitrogen). The resulting plasmid was linearized with

BamH1 (sense) and NotI (antisense) and used for generating digoxigenin-

labelled RNA-probes (DIG RNA labelling kit SP6/T7; Roche). Developing flower

buds were fixed in 4% paraformaldehyde and embedded in tissue-prep paraffin

(Fisher Scientific). Sections of 10–12-mm thickness were generated using a Leica

microtome and mounted on ProbeOnPlus slides (Fisher Biotech). Hybridization

was performed as described30; for whole mount in situ hybridization, anthers and

ovules were fixed, mounted in acrylamide-covered slides, and hybridized as

described31.

PCR with reverse transcription (RT–PCR). Total RNA was extracted from

leaves, roots, stems, inflorescences, mature flowers, gynoecia, ovules and seed-

lings with TRIzol (Invitrogen). Total RNA (2mg) from each tissue was used to

synthesize first-strand cDNA by using 20mer-oligo dT (Sigma), 0.25 mM dNTPs

and SuperScript III Reverse Transcriptase (Invitrogen), and incubating at 42 uCfor 2 h. cDNA (100 ng) was used to amplify a 369-bp AGO9 fragment by using

primers gntpS2 (59-TCCCCAATCAAAGGAAAATGG-39) and gntpAS2 (59-TC

TTGGAATTGTGACTCAGTGCA-39). Amplification of a 96-bp fragment cor-

responding to ACTIN2 mRNA was used as a control32. PCR was performed with

an initial denaturation step at 94 uC for 1 min 30 s, followed by 30 cycles of

denaturation at 94 uC for 30 s, annealing at 60 uC, and extension at 72 uC for 30 s.

Histological analysis. For phenotypic analysis of ovules, inflorescences from

wild-type and mutant plants were fixed in FAA (formaldehyde 10%, acetic acid

5%, ethanol 50%), for 12 h and subsequently dehydrated in 70% ethanol.

Gynoecia at different developing stages were dissected with hypodermic needles

(1-ml insulin syringes), cleared in Herr’s solution (phenol:chloral hydrate:85%

lactic acid:xylene:clove oil in a 1:1:1:0.5:1 proportion), and observed using a

DMR Leica microscope with Nomarski optics. Histochemical localization of

GUS activity was performed by incubating dissected gynoecia in GUS staining

solution (10 mM EDTA, 0.1% Triton X-100, 0.5 mM potassium ferrocyanide,

0.5 mM potassium ferricyanide and 1 mg ml21 5-bromo-4-chloro-3-indolyl-b-

D-glucoronic acid in 50 mM sodium phosphate buffer, pH 7.0) for 24 h as

described25. Callose was detected by incubating floral buds in aniline blue stain-

ing solution (0.1% aniline blue, 100 mM Na2HPO4, pH 7.4) for 12–24 h in

darkness; for each developmental stage and sample, at least 100 ovules were

dissected on a slide and mounted in 30% glycerol. Observations were conducted

in an Olympus BX60 (Model BX60F5; Olympus Optical) microscope using

epifluorescence ultraviolet filters (365 nm excitation, 420 nm emission).

Micrographs were acquired using Image Pro-Plus Software, version 4.0.

Protein analysis and immunoblots. The amino-terminal sequence of 16

N-SSRNHAGNDTNDADRK-C (16 amino acids) was selected after three-

dimensional modelling (HHpred, available at http://toolkit.tuebingen.mpg.de)

to generate a rabbit polyclonal peptide antibody (Invitrogen). After affinity

purification, the same antibody was used for Immunoblot analysis (1:500 dilu-

tion) or immunolocalization (1:100 dilution). Immunoblots were generated with

the WesternBreeze Chemiluminescent Detection Kit (Invitrogen) using 5mg of total

protein for each assay (1-week-old seedlings, developing gynoecia, developing ros-

ette leaves, and siliques 7 days after pollination). Proteins were stained with SYPRO

Ruby protein gel stain (Invitrogen) and with a Silver Staining Kit (Invitrogen).

Immunolocalization in sectioned specimens. For immunolocalization experi-

ments, flowers at different developmental stages were fixed in 4% paraformalde-

hyde in PBS (10 mM KH2PO4, 150 mM NaCl, pH 7) for 12 h at room

temperature, gradually dehydrated in an ethanol series (10%), and embedded

in LR White Resin (Electron Microscopy Sciences). Sections (0.5mm) were

generated with a ultramicrotome (Leica Ultracut R) and placed on

ProbeOnPlus (Fisher Biotech) slides. After washing twice with PBS, sections

were blocked for 2 h with 5% BSA and 0.05% Tween 20 in PBS, and incubated

with the AGO9 antibody (1:100 in 0.1% BSA in PBS) for 2 h at room tem-

perature. After washing with PBS, slides were incubated with Alexa Fluor 488

goat anti-rabbit (Invitrogen) 1:50 dilution during 2 h at room temperature,

washed with PBS and counterstained with 1mg ml21 DAPI (Sigma). The slides

were mounted with ProLong Gold antifade reagent (Invitrogen). Fluorescence

was visualized using a Leica DM 6000B epifluorescence microscope, using filter

cubes I3 (excitation 450–490 nm, emission 510 nm) and ultraviolet filter A

(excitation 340–380 nm, emission 400 nm). Images were acquired by using

Leica QWin Standard V3.4.0 (Leica Microsystems).

Immunolocalization in whole-mounted specimens. Pistils and siliques at

various developmental stages were fixed overnight at 4 uC in 4%

paraformaldehyde:PBS:2% Triton fixative, washed three times in PBS, and dis-

sected to isolate the ovules and early seeds. The dissected ovules and seeds were

embedded in acrylamide as described33 to facilitate manipulation and maintain

the three-dimensional architecture of the tissues. Samples were digested in an

enzymatic solution (1% driselase, 0.5% cellulase, 1% pectolyase, 1% BSA, all

from Sigma) for 25 min to 1 h at 37 uC, depending on the developmental stage,

subsequently rinsed three times in PBS, and permeabilized for 2 h in PBS:2%

Triton. They were then incubated overnight at 4 uC with primary antibodies used

at a dilution of 1:100 for AGO9 and 1:400 otherwise. The slides were washed for a

day in PBS:0.2% Triton, and coated overnight at 4 uC with secondary antibodies

(Alexa Fluor 488 or 568 conjugate, Molecular Probes) used at 1:400 dilution.

After washing in PBS:0.2% Triton for a minimum of 6 h, the slides were incu-

bated with DAPI (1mg ml21 in PBS) for 1 h, washed for 2 h in PBS, and mounted

in PROLONG medium (Molecular Probes). Complete 3D ovule or seed images

were captured on a laser scanning confocal microscope (Leica SP2) equipped

with 405 nm (DAPI), 488 nm (green) or 568 nm (red) excitation and either 340

or 363 objectives. Projections of selected optical sections were generated for this

report, and edited using Graphic Converter (LemkeSOFT). At least 50 ovules

were scored for each developmental stage.

Immunoprecipitation and analysis of small RNAs. Immunopurification

AGO9 and its associated sRNAs was conducted as described34, with some mod-

ifications. Owing to extremely low protein yields obtained in preliminary experi-

ments conducted with hundreds of female reproductive organs, protein

extraction was conducted with a total of 12,000 wild-type developing gynoecia

containing ovules at mixed developmental stages (four-nuclear stage of game-

togenesis to unpollinated mature). Protein extract (0.5 ml) from 12,000 wild-

type gynoecia was pre-cleared by incubation with 10 ml Protein A-Sepharose

(Invitrogen, Cat. No. 10-1041) at 4 uC for 30 min. Pre-cleared extracts were then

incubated either with AGO9 antibody or AGO9 pre-immune serum as a negative

control, and 30 ml Protein A-Sepharose at 4 uC overnight. The immunoprecipi-

tates were washed three times (15 min each) in extraction buffer (50 mM Tris-

HCl, pH 7.5, 150 mM NaCl, 10% glycerol, 0.2% NP-40, 5 mM MgCl2, 5 mM

dithiothreitol, one tablet of Roche Protease Inhibitor Cocktail for each 10 ml).

Commercial columns (Ambion) were used to isolate sRNAs from the purified

AGO9 complex. Small RNAs were resolved on a 12.5% denaturing PAGE 7 M

urea gel, and stained with SYBR-gold (Invitrogen). Before cloning, gel slices

within the range of 18–30 nucleotides were excised, and the RNAs were eluted

and purified using DTR Gel Filtration Cartridges (EdgeBio). A detailed protocol

of the immunoprecipitation and elution procedure is available on request.

After elution and gel-purification, sRNAs were ligated with adaptors at their 59

and 39 ends, converted to cDNA products, and subsequently cloned and

sequenced by Sanger methods. Whereas immunopurifications conducted with

the pre-immune serum did not yield any bacterial clones containing endogenous

Arabidopsis sequences, we obtained a total 2,552 sequences representing 344 dis-

tinct small RNAs with the AGO9 antibody. Cloning of small RNAs was performed

with the miRCat Small RNA Cloning Kit (Integrated DNA Technologies) follow-

ing manufacturer instructions. Individually cloned products were sequenced with

the BigDye Terminator v3.1 Cycle Sequencing kit (Applied Biosystems) in a 3730xl

DNA Analyzer (Applied Biosystems,). Sequences were quality-checked with

sequence Scanner 1.0 (Applied Biosystems). Sequences were filtered and mapped

to the Arabidopsis genome (http://www.arabidopsis.org). Annotation of sRNAs

was performed using information from TAIR9 (ftp://ftp.arabidopsis.org/

Sequences/blast_datasets/TAIR9_blastsets), and miRBase (http://microrna.sanger.

ac.uk/sequences).

28. Xie, Z. et al. Genetic and functional diversification of RNA pathways in plants. PLoSBiol. 2, e104 (2004).

doi:10.1038/nature08828

Macmillan Publishers Limited. All rights reserved©2010

29. Xie, Z., Allen, E., Wilken, A. & Carrington J. C.. DICER-LIKE 4 functions in trans-acting small interfering RNA biogenesis and vegetative phase change inArabidopsis thaliana. Proc Natl Acad Sci USA 102, 12984–12989 (2005).

30. Vielle-Calzada, J.-P. et al. Maintenance of genomic imprinting at the Arabidopsismedea locus requires zygotic DDM1 activity. Genes Dev. 13, 2971–2982 (1999).

31. Garcıa-Aguilar, M., Dorantes-Acosta, A., Perez-Espana, V. & Vielle-Calzada, J.-P.Whole-mount in situ mRNA localization in developing ovules and seeds ofArabidopsis. Plant Mol. Biol. Rep. 23, 279–289 (2005).

32. Kasahara, R. D., Portereiko, M. F., Sandaklie-Nikolova, L., Rabiger, D. S. & Drews,G. N. MYB98 is required for pollen tube guidance and synergid cell differentiationin Arabidopsis. Plant Cell 17, 2981–2992 (2005).

33. Bass, H. W. et al. Evidence for the coincident initiation of homolog pairing andsynapsis during the telomere-clustering (bouquet) stage of meiotic prophase. J.Cell Sci. 113, 1033–1042 (2000).

34. Qi, Y., Denli, A. M. & Hannon, G. J. Biochemical specialization within ArabidopsisRNA silencing pathways. Mol. Cell 19, 421–428 (2005).

doi:10.1038/nature08828

Macmillan Publishers Limited. All rights reserved©2010