Control of Genic DNA Methylation by a jmjC Domain-Containing Protein in Arabidopsis thaliana

See allHide authors and affiliations

Science  25 Jan 2008:
Vol. 319, Issue 5862, pp. 462-465
DOI: 10.1126/science.1150987


Differential cytosine methylation of repeats and genes is important for coordination of genome stability and proper gene expression. Through genetic screen of mutants showing ectopic cytosine methylation in a genic region, we identified a jmjC-domain gene, IBM1 (increase in bonsai methylation 1), in Arabidopsis thaliana. In addition to the ectopic cytosine methylation, the ibm1 mutations induced a variety of developmental phenotypes, which depend on methylation of histone H3 at lysine 9. Paradoxically, the developmental phenotypes of the ibm1 were enhanced by the mutation in the chromatin-remodeling gene DDM1 (decrease in DNA methylation 1), which is necessary for keeping methylation and silencing of repeated heterochromatin loci. Our results demonstrate the importance of chromatin remodeling and histone modifications in the differential epigenetic control of repeats and genes.

Genomes of vertebrates and plants contain a substantial proportion of transposons and repeats (1). These potentially deleterious sequences are cytosine-methylated and inactivated (2, 3) to form heterochromatin (4, 5). Methylated heterochromatin, especially when dispersed within gene-rich regions, has the potential to spread by self-reinforcing mechanisms (6, 7) to flanking cellular genes and disrupt their expression. Mechanisms that confine the methylated regions remain enigmatic, despite their importance in maintaining the integrity of large genomes with a high proportion of dispersed transposons. Here, we identify a new pathway that excludes cytosine methylation from genic regions by histone modification and chromatin remodeling, thus ensuring proper plant development.

In the flowering plant Arabidopsis thaliana, and in plants in general, cytosine methylation is found in both CG and non-CG contexts. In Arabidopsis, methylation at CG sites is maintained by the DNA methyltransferase MET1, whereas methylation at non-CG sites requires the DNA methyltransferase CMT3 (812). Non-CG methylation is also controlled by methylation of histone H3 at lysine 9 (H3mK9) and by the RNA interference (RNAi) machinery (1316). DDM1, a chromatin-remodeling adenosine triphosphatase, is involved in maintenance of both CG and non-CG methylation (1719). Mutations in MET1 and DDM1 also result in a variety of developmental abnormalities by inducing heritable changes in other loci (810, 20, 21). One of the ddm1-induced abnormalities, called bonsai (bns), is caused by epigenetic silencing of a gene encoding a homolog of a cell cycle regulator, APC13 (22). The silencing of this gene, BONSAI (BNS), is associated with spreading of methylated heterochromatin from a flanking LINE retroelement (22). This LINE functions as methylated heterochromatin, which has a potential to spread to the flanking BNS gene (Fig. 1A).

Fig. 1.

Identification of ibm1 (increase in BONSAI methylation 1). (A) Structure of the BONSAI (BNS) locus. Exons of the BNS gene are shown by boxes with the coding region in black and the untranslated regions in white. The gray box shows the LINE. (B) Methylation of the BNS gene in the ibm1 mutant detected by methylation-sensitive restriction digestion. Four alleles of the ibm1 mutants are shown, together with wild-type Columbia (Col) and the bns epigenetic mutation (22). Genomic DNA was digested by Bgl II or Sau3A I and was subsequently amplified by the polymerase chain reaction (PCR). (C) Bisulfite analysis of the BNS locus in the ibm1 mutants. The percentage of methylated cytosine is indicated by vertical bars (black, CG; blue, CNG; red, asymmetric cytosine). (D) A schematic representation of structure of IBM1 gene and IBM1 protein. T-DNA insertions are shown by white triangles and the ibm1-1 base substitution by an asterisk.

To explore the mechanisms that exclude genic cytosine methylation in wild-type plants, we used methylation-sensitive restriction enzymes to screen a mutagenized population for individuals with ibm (increase in BONSAI methylation) phenotype (23). One of them, ibm1, is described in this report. The IBM1 gene (At3g07610) was identified by a map-based approach (23). The original ibm1-1 mutant has a base substitution causing an amino acid substitution (Gly672 → Glu). We subsequently tested three additional ibm1 alleles carrying T-DNA insertions and verified that these independent alleles also caused DNA hypermethylation of the BNS gene (Fig. 1B).

The BNS sequence was hypermethylated in the first generation in which the ibm1 mutant allele became homozygous (Fig. 1B). This feature was different from the BNS hypermethylation in the ddm1 mutant, which is slow and detectable only after several generations of self-pollination in the mutant background (22). Bisulfite sequencing revealed that cytosine methylation occurred at the BNS gene in ibm1 and that non-CG sites are the main targets of the BNS methylation (Fig. 1C and table S1). Unlike ddm1, the ibm1 mutation did not affect methylation in repeat sequences, such as centromeric satellites or 5S rDNA (fig. S1). These features suggest that low-copy sequences are the primary target of the IBM1 gene function.

In addition to cytosine hypermethylation at the low-copy BNS locus, ibm1 plants exhibited a variety of developmental abnormalities (Fig. 2 and table S2). The morphological phenotypes become apparent in the F3 generation. Most of the ibm1 plants formed small, narrow leaves and exhibited arrested flower development and reduced fertility (Fig. 2, A to C, and table S2). Homozygous ibm1 plants produced few viable seeds, whereas seed set in heterozygous IBM1/ibm1 was normal (Fig. 2D and table S3), indicating that the seed phenotype depends on the parental sporophytic genotype rather than the genotype of seed or gametophyte. Silencing of the BNS gene alone cannot account for the pleiotropic phenotypes exhibited by the ibm1 mutation; knockdown of BNS transcripts by RNAi results in the bonsai phenotypes but not the other phenotypes found in the ibm1 mutants, such as leaf deformation, abnormal flower, and reduced fertility (22). Multiple target loci seem to mediate the developmental defects of the ibm1 mutations, suggesting a general role of IBM1 in the control of genome integrity.

Fig. 2.

Developmental abnormalities of the ibm1 mutants. (A to C) Phenotypes of rosette leaves (A), inflorescence (B), and flowers (C) in wild-type (WT) Col, ibm1-1, and ibm1-4 plants. (D) Abnormal seed development in an ibm1-1 mutant (–/–) compared to an IBM1/ibm1-1 heterozygote (+/–). (E and F) Pollen grains of WT Col (E) and ibm1-1 (F). Arrowheads indicate degenerated pollen grains. (G and H) Plants showing bonsai (bns)–like phenotype found in ibm1-4 (G) and ibm1-1 (H). All the ibm1 mutant plants were homozygous for two generations (F3) except those shown in (G), which were homozygous for three generations. Scale bars, 1 cm (A), 2 mm (C).

Several considerations suggest that IBM1 might affect DNA methylation through H3mK9. The IBM1 gene encodes a jmjC (Jumonji C) domain–containing protein in the JHDM2 family, which is constituted of demethylases of H3mK9 (Fig. 1D and fig. S2) (24, 25). The Jumonji domain of IBM1 shows conservation of amino acid residues for Fe(II) and α-ketoglutarate (αKG) binding that are critical for H3K9 demethylase activity (fig. S2B) (25). In addition, H3mK9 is known to direct cytosine methylation by CMT3 at non-CG sites (13, 14). We therefore hypothesized that the hypermethylation of the non-CG sites at the BONSAI locus in the ibm1 mutantsismediated by ectopic H3mK9 accumulation. In fact, chromatin immunoprecipitation experiments revealed that the H3mK9 level in the BONSAI locus increased in the ibm1 mutant, especially in the 3′ region near the LINE (Fig. 3A), which correlates with cytosine methylation at non-CG sites there (Fig. 1C).

Fig. 3.

ibm1 phenotypes depend on H3K9 methylation. (A) Histone modification of wild-type Col (WT) and the ibm1-4 mutant detected by chromatin immunoprecipitation (IP). INPUT is the sample before IP; No AB denotes samples after IP procedure without antibody. H3mK4 and H3mK9 are samples after IP with monoclonal antibodies against H3 with dimethylation at the corresponding lysine. Amplified regions around the BONSAI locus are indicated in Fig. 1A. A flanking gene at1g73180, LINE downstream of BNS, and Ta3 and CACTA1 transposons in other loci are used for controls. Serial dilution of the INPUT before the PCR showed that the amount of the chromatin used here was almost identical in the wild type and the ibm1 mutant (not shown). Three independent preparations of the samples from different individuals gave essentially the same results. (B) The kyp mutation suppressed the BNS hypermethylation phenotype of the ibm1 mutant. Lanes 3 to 14 show plants homozygous for the indicated genotype segregating in self-pollinated progeny of an IBM1/ibm1-4 KYP/kyp double heterozygote. (C) Abnormal leaf phenotypes in 3-week-old plants. More plants are shown in fig. S3B. (D) Reduced fertility phenotype. (E) Pollen defect phenotype. Plants shown in (C) to (E) are F3 generation, as in fig. S3B. Phenotype scoring of all the families is shown in tables S2 and S3.

To test whether the increase in the H3mK9 level is responsible for the ectopic DNA methylation and developmental defects, we examined the effect of a mutation in the H3K9 methylase gene KYP/SUVH4 on the ibm1 mutant. Figure 3B shows that the kyp/suvh4 mutation suppressed the hypermethylation phenotype of the ibm1 mutation, suggesting that H3mK9 mediates the hypermethylation. The hypermethylation was also suppressed by a mutation in the non-CG methylase CMT3 gene (fig. S3A). In addition to suppression of the hypermethylation, both the kyp/suvh4 and cmt3 mutations suppressed all of the detectable developmental phenotypes of ibm1, such as leaf deformation, abnormal flowers, pollen defects, and reduced fertility (Fig. 3, C to E, and tables S2 and S3), demonstrating that these diverse developmental defects are also mediated by ectopic deposition of H3mK9 and non-CG methylation.

The ddm1 mutation results in cytosine hypomethylation and derepression of transposons and repeats. Paradoxically, the ddm1 mutation also induces hypermethylation at some low-copy loci, such as BNS and SUPERMAN (22, 26). To see whether the hypermethylation effect of the ddm1 mutation overlaps with that of ibm1, we generated double mutants of ddm1 and ibm1. The ddm1 ibm1 double mutants showed strong enhancement of the developmental phenotypes (Fig. 4), with small leaves, slow growth, and complete sterility. These results suggest that the effects of the ddm1 mutation overlap with that of the ibm1 mutation. Although ddm1 single mutants exhibited BNS hypermethylation and other developmental phenotypes only after repeated self-pollinations (2022), the effect of the ddm1 mutation in the ibm1 mutant background was apparent in the first generation (Fig. 4), which suggests that the role of the DDM1 gene in low-copy regions may be more important than previously thought. Although the ddm1 mutation resulted in loss of H3mK9 in repeated sequences, overall level of H3mK9 was not severely affected (27). In the ddm1 single mutant, its effect on the H3mK9 in genic regions may be compensated by the IBM1 gene function.

Fig. 4.

Developmental phenotypes of the ibm1 mutation are enhanced by the ddm1 mutation. This panel shows F2 homozygotes segregating in self-pollinated progeny of an IBM1/ibm1-4 DDM1/ddm1-1 double heterozygote. This double heterozygote was derived from a cross between IBM1/ibm1-4 and DDM1/ddm1-1 backcrossed repeatedly to remove any heritable effect of the mutations (20). Six-week-old plants are shown. Developmental abnormalities in the ibm1 single mutants were still mild in the segregating population. Within 189 plants in the F2 segregating population, all 12 ddm1/ddm1 ibm1/ibm1 plants exhibited the phenotype as shown in this figure; such a severe phenotype was not found in any of the other 177 plants, which includes 31 ddm1/ddm1 IBM1/– plants and 31 DDM1/ibm1/ibm1 plants. Scale bar, 1 cm.

We have shown that the IBM1 and DDM1 proteins, by repressing ectopic non-CG methylation in genic regions, are required for normal Arabidopsis development. Neither the kyp nor the cmt3 single mutant shows developmental phenotypes (1114), possibly because the primary target of H3K9 methylation and non-CG methylation is nongenic sequences in wild-type plants (28). In the ibm1 mutant background, however, the H3mK9 and non-CG methylation machinery has pronounced effects on Arabidopsis development. Non-CG methylation and CG methylation are both important for silencing the repeats and transposons (28, 29). At the same time, cellular genes need to be protected from the spreading of methylated heterochromatin. Coordination of the cytosine methylation in repeats and genes, through histone modifications and chromatin remodeling, will likely have an important impact on the genome integrity.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

Tables S1 to S4


References and Notes

View Abstract

Navigate This Article