One-Way Control of FWA Imprinting in Arabidopsis Endosperm by DNA Methylation

See allHide authors and affiliations

Science  23 Jan 2004:
Vol. 303, Issue 5657, pp. 521-523
DOI: 10.1126/science.1089835


The Arabidopsis FWA gene was initially identified from late-flowering epigenetic mutants that show ectopic FWA expression associated with heritable hypomethylation of repeats around transcription starting sites. Here, we show that wild-type FWA displays imprinted (maternal origin–specific) expression in endosperm. The FWA imprint depends on the maintenance DNA methyltransferase MET1, as is the case in mammals. Unlike mammals, however, the FWA imprint is not established by allele-specific de novo methylation. It is established by maternal gametophyte–specific gene activation, which depends on a DNA glycosylase gene, DEMETER. Because endosperm does not contribute to the next generation, the activated FWA gene need not be silenced again. Double fertilization enables plants to use such “one-way” control of imprinting and DNA methylation in endosperm.

DNA methylation is a key epigenetic determinant that controls parent of origin–specific gene expression (imprinting) in mammals, where the methylation is erased and reestablished in each generation (1). In contrast, epigenetic states of gene expression in flowering plants are often inherited unchanged over many generations. Epigenetic mutations affecting plant development have been identified in laboratories and in natural populations (29). For example, the Arabidopsis late-flowering mutant fwa-1 does not have a change in the nucleotide sequence of the responsible gene FWA; instead, the phenotype is due to ectopic FWA expression associated with heritable loss of methylation (5). Although the loss of DNA methylation induces the late-flowering phenotype, there has been no evidence that FWA methylation is developmentally regulated in the wild type. Nor does this gene seem to control flowering time during normal development, because loss-of-function mutations of FWA do not affect flowering time (5). To understand the role of DNA methylation in plant development, we examined the expression of FWA during normal development.

FWA is not expressed in wild-type adult tissues, but the FWA transcripts are detectable in the silique and in 4-day imbibed seeds (5). We first examined FWA expression by reverse transcription polymerase chain reaction (RT-PCR) in various organs and dissected seeds (Fig. 1A). In the dissected seeds, FWA transcripts were not detected in the embryo fraction but were detected in the fraction containing endosperm and the seed coat. Examination of other organs suggests that FWA expression is confined to the developing endosperm or seed coat.

Fig. 1.

FWA expression in central cell and endosperm. (A) RT-PCR analysis of FWA expression in various organs. Total RNA was isolated from the indicated tissues. The APETALA2 (AP2) gene was used as a control. (B to F) FWA-GFP fusion protein localization was analyzed by confocal laser microscopy. (B) The fusion protein localizes to the diploid central cell before fertilization. (C) The triploid endosperm nucleus at 6 hours after pollination (HAP). (D) The four-nuclei stage of endosperm at 12 HAP. (E) The eight-nuclei stage of endosperm at 24 HAP. (F) GFP fluorescence has disappeared at 48 hours after pollination. Chlorophyll autofluorescence is shown in red. Scale bars, 20 μm.

We monitored FWA expression in plants with a pFWA::FWA-GFP transgene that express an FWA–green fluorescent protein (GFP) fusion protein under the control of the FWA promoter. Using confocal laser microscopy, we localized the FWA-GFP fusion protein to the central cell nucleus before fertilization, which is the progenitor of endosperm in the mature ovule (Fig. 1B). After fertilization, GFP fluorescence was observed in the fertilized central cell and the developing endosperm up to the 8- to 16-nuclei stage (Fig. 1, C to F). GFP fluorescence was not detected in the egg cell or embryo in these two constructs. We also confirmed endosperm-specific expression of FWA by in situ hybridization (fig. S1, G and H). Taken together, these results suggest that FWA expression is confined to the central cell of the female gametophyte and the endosperm. The transcripts were not detected in the embryo or vegetative organs.

Ectopic expression of the FWA gene in the fwa-1 hypomethylated epigenetic allele is accompanied by the loss of DNA methylation of the direct repeats of the 5′ region of the gene (5). To determine whether endosperm-specific expression is also correlated with loss of DNA methylation, we examined FWA methylation in various seed tissues and in pollen by the bisulfite sequencing procedure. We dissected the seed into three parts— embryo, seed coat, and endosperm (fig. S2)—and isolated DNA from each part. The overall DNA methylation level in the FWA 5′ direct repeats was markedly reduced in the endosperm (Fig. 2). By contrast, FWA methylation was not reduced in the embryo, seed coat, leaf, and pollen (Fig. 2).

Fig. 2.

Demethylation of FWA in endosperm. Percent methylation at CpG, CpNpG, and asymmetric sites of the 5′ direct repeats of the FWA gene was determined by bisulfite sequencing with individual 6-12 clones. DNA from each tissue was isolated from Col-0 accession.

Endosperm is the only tissue in which parental imprinting has been reported in flowering plants (1014). We used allele-specific RT-PCR analysis to test for FWA imprinting. A C/A polymorphism between strains Col-0 and Ler in exon 7 was used to distinguish the transcripts from the maternal and paternal FWA alleles. In F1 seeds resulting from reciprocal interstrain crosses, only the transcripts derived from the maternal allele were detected in the endosperm plus seed coat fractions, whereas both the maternal and paternal alleles were silent in the embryo (Fig. 3). The FWA imprinting in endosperm was also confirmed in other interstrain crosses between strains Col-0 and Ws (fig. S3A).

Fig. 3.

Imprinting of FWA in endosperm. (A) Allele-specific RT-PCR shows no FWA expression in embryo. (B) Maternal allele–specific expression in endosperm at 6 and 8 days after pollination (DAP), corresponding to torpedo and early maturation stages of embryo development. The nonimprinted αVPE gene was used as a control.

We next examined whether the tissue-specific and parent of origin–specific FWA expression depends on DNA methylation. Wild-type Col-0 was reciprocally crossed to mutants of DNA methyltransferases (Fig. 4, A and B) (fig. S3B) and the effect of each mutation on FWA expression was examined by allele-specific RT-PCR using dissected F1 seeds. Paternally derived FWA transcripts were detected when the male parent had a mutation in MET1, the maintenance methyl-transferase for CpG sites (15, 16) (Fig. 4, A and B). The met1 mutation in the female parent did not induce paternal FWA expression in the embryo or endosperm plus seed coat fractions (fig. S3B), which suggests that the loss of paternal silencing occurred before fertilization. On the other hand, the imprinting was not affected by mutation of CMT3 (CHROMOMETHYLASE3) or by mutation of the DRM (DOMAINS REARRANGED METHYLTRANSFERASE) de novo methylase (Fig. 4, A and B) (fig. S3B). CMT3 has been shown to be important for methylation of non-CG sites (17, 18). DRMs are structurally similar to mammalian Dnmt3 de novo methylases (19, 20), and DRM2 is necessary for the de novo methylation induced by transgenes (21). The met1 mutation also induces FWA expression in the embryo (Fig. 4A). These results suggest that maintenance of endosperm-specific and parent of origin–specific FWA expression depends on MET1.

Fig. 4.

Trans mutations affecting FWA imprinting. (A and B) Col-0 females were crossed with drm1 drm2 (Ws), cmt3-7 (Ler), met1-1 (Ler), or fwa-1 (Ler) mutants. The polymorphic site of Ws is the same as for Ler. Total RNA was isolated from dissected embryo and endosperm plus seed coat fractions at 7 days after pollination (corresponding to the walking-stick stage of embryo development) and was subjected to the allele-specific RT-PCR analysis. (C) RT-PCR analysis of FWA transcripts in wild type and dme-1 mutant. The ACTIN (ACT) gene was used as a control. (D and E) Fluorescence images of pFWA::FWA-GFP expression in wild type (D) and dme-1 homozygous mutant (E). Scale bars, 20 μm.

Because the DRM de novo methyltransferases did not affect FWA imprinting, a remaining important question is how the specific DNA methylation and expression patterns are established in the endosperm. We next examined the effect of DME, which has been shown to activate expression of the maternal MEDEA (MEA) allele in central cells before fertilization (22). DME encodes a protein with a DNA glycosylase domain, and the product has a DNA glycosylase activity in vivo. We tested the effect of the dme-1 mutation on expression of FWA. RT-PCR revealed that FWA transcripts did not accumulate in homozygous dme-1 mutant ovules, whereas they were detectable in control wild-type ovules (Fig. 4C). We also examined the effect of the dme-1 mutation on FWA promoter activity. A pFWA::FWA-GFP transgenic line was crossed to dme-1 mutant plants. In the F2 progeny from this cross, GFP fluorescence was detected in the central cell nucleus of wild-type DME ovules (Fig. 4D), but no signal was observed in homozygous mutant dme-1 ovules (Fig. 4E). These results suggest that the maternal-specific pFWA::FWA-GFP expression depends on a functional DME allele in the female gametophyte. Although no DME product is detectable after fertilization (22), its effect on FWA expression is prolonged after fertilization, which suggests that DME affects a heritable epigenetic mark on FWA, as is the case for MEA (22).

Our results indicate that the maintenance of FWA imprinting depends on the maintenance DNA methylation machinery, a situation comparable to mammalian imprinting (23). Unlike mammals, however, the maternal-specific expression of FWA is not established by a paternal-specific de novo methylation, but it is established by maternal-specific activation that is dependent on the DME DNA glycosylase. Thus, the silent methylated state is the default for this class of imprinted genes (22). It would be important to know how general such a controlling mechanism is in plants. The control by DME is conserved between FWA and MEA. However, it has been reported that loss of MET1 activity with a paternally transmitted transgene with MEA promoter does not induce its activation (14). On the other hand, MET1 regulates MEA expression in the female gametophyte in an antagonistic manner to DME (24). Thus, control of imprinting by MET1 and DME might be a general mechanism. In any case, a unique feature of imprinting in flowering plants is that the epigenetic state in the endosperm does not need to be reprogrammed again. Because the endosperm degenerates during seed maturation, it does not transmit genetic or epigenetic information to the next generation. In this sense, the endosperm is functionally analogous to mammalian extra embryonic membrane. Establishment of imprinting in the central cell and its subsequent maintenance in endosperm after double fertilization enables plants to use such simple one-way control of imprinting. When methylation is lost in the embryonic lineages (e.g., by the met1 mutation), the fwa epigenetic mutation and its associated late-flowering phenotype can be stably inherited over many generations.

Supporting Online Material

Materials and Methods

Figs. S1 to S3


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article