Requirement of CHROMOMETHYLASE3 for Maintenance of CpXpG Methylation

See allHide authors and affiliations

Science  15 Jun 2001:
Vol. 292, Issue 5524, pp. 2077-2080
DOI: 10.1126/science.1059745


Epigenetic silenced alleles of the Arabidopsis SUPERMANlocus (the clark kent alleles) are associated with dense hypermethylation at noncanonical cytosines (CpXpG and asymmetric sites, where X = A, T, C, or G). A genetic screen for suppressors of a hypermethylated clark kent mutant identified nine loss-of-function alleles of CHROMOMETHYLASE3(CMT3), a novel cytosine methyltransferase homolog. These cmt3 mutants display a wild-type morphology but exhibit decreased CpXpG methylation of the SUP gene and of other sequences throughout the genome. They also show reactivated expression of endogenous retrotransposon sequences. These results show that a non-CpG DNA methyltransferase is responsible for maintaining epigenetic gene silencing.

Cytosine methylation plays a major role in determining the epigenetic expression state of eukaryotic genes. This methylation is most often found at the symmetrical dinucleotide CG (or CpG sites). CpG methylation is maintained by the well-studied DNMT1 subfamily of methyltransferases, which includesArabidopsis MET1 (1–3). Methylation at sites other than CpG is also found in many organisms (4), but the mechanism by which this methylation is maintained is poorly understood. Arabidopsis can tolerate major disruptions in DNA methylation (2, 3, 5), making it useful for genetic analysis of methylation patterning. For unknown reasons, the floral development gene SUPERMAN (SUP) becomes densely hypermethylated and silenced in several mutants that display genome-wide hypomethylation. It occurs in plants expressing antisense RNA to the MET1 gene (6), in amet1 loss-of-function mutant (previously called ddm2) (7, 8), and in ddm1mutants (5, 7). In this way, SUPhypermethylation resembles a phenomenon observed in cancer cells, where genome-wide loss of methylation is frequently associated with hypermethylation and silencing of particular tumor suppressor genes (9).

SUP hypermethylation causes a floral phenotype similar to that of known loss-of-function sup mutants: an increased number of stamens and a defective gynoecium (female reproductive structure) (Fig. 1A). These hypermethylated SUP alleles (called the clark kent alleles) are recessive and heritable. They are associated with dense methylation at CpG sites, at CpXpG sites (X = A, T, C, or G), and at asymmetric sites (those cytosines not present in the symmetric CpG or CpXpG contexts). clark kent alleles that arise in an antisense-MET1 background or in themet1 mutant lack most CpG methylation but maintain the other types (6, 8), showing that non-CpG methylation is critical for the maintenance of SUP gene silencing.

Figure 1

CMT3 mutations. (A) clk-st flower containing 10 stamens and three incompletely fused carpels. (B) Flower from thecmt3-7 suppressor mutant, showing the normal number of six stamens and a normal gynoecium consisting of two fused carpels. (C) The CMT3 protein sequence determined from the Ler ecotype. Residues constituting the BAH (bromo-adjacent homology) domain and the chromodomain are underlined. Conserved methyltransferase catalytic motifs I, IV, VI, and VIII–X are marked. Asterisks denote highly conserved amino acids present in each motif (14), derived from alignments with the bacterial methylase Hha I. Residues mutated in the cmt3 mutants are in boldface. Single-letter abbreviations for amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.

To identify loci important for maintenance of methylation and silencing of SUP, we performed a mutant screen for suppressors of a nonreverting clark kent allele, clk-st, created by introducing an additional SUP locus into clark kent-3 plants (6, 10). clk-st seeds were mutagenized with ethylmethane sulfonate, and individual M2 families were screened for mutations that derepress SUP gene silencing, leading to plants with a wild-type floral phenotype (10). Sixteen independent recessive mutants were recovered and five were chosen for initial study. Of these, four completely reverted the clark kent phenotype to yield wild-type flowers (Fig. 1B), and one displayed partial reversion. Each of the five mutants failed to complement any of the others, indicating that they are loss-of-function alleles of the same gene (10).

One of these mutations was mapped to the bottom of chromosome I (10), near the CHROMOMETHYLASE3(CMT3) gene. CMT3 encodes a putative cytosine methyltransferase containing a chromodomain and a bromo-adjacent homology (BAH) domain (11–13). We crossed one of the strong suppressors to cmt3-2, a nonsense allele of CMT3isolated previously (13). These mutants failed to complement (10), showing that all five suppressor mutants are alleles of CMT3, here designated cmt3-3 (the partial suppressor), cmt3-4, cmt3-5, cmt3-6, and cmt3-7.

The molecular lesions in the cmt3 mutants were identified by sequencing 5021 base pairs (bp) of the CMT3 gene from each homozygous mutant line. A single C/G to T/A transition mutation was found in each mutant, in every case altering the coding region ofCMT3 (Fig. 1C) (GenBank accession number AF364174). We used the DNA polymorphisms created by the cmt3-4, -5,-6, and -7 mutations to generate molecular markers, and we found that these markers perfectly cosegregated with the suppressor mutant phenotypes (10). The cmt3-5and cmt3-7 alleles contain stop codons terminating CMT3 after 95 or 27 amino acids, respectively, and thus they likely represent null alleles. The cmt3-3, cmt3-4, andcmt3-6 alleles are missense mutations within the methyltransferase segment of CMT3 (Fig. 1C) (14). We identified four additional cmt3alleles by sequencing the CMT3 gene from each of our remaining 11 mutants (Fig. 1C). cmt3-9 andcmt3-11 are phenotypically strong suppressors and contain nonsense mutations; cmt3-8 and cmt3-10 are phenotypically weak alleles and contain missense mutations in the methyltransferase segment. Thus, 9 of the 16 mutants isolated from our screen are alleles of CMT3.

We used bisulfite genomic sequencing (10) to determine the effect of CMT3 on methylation patterning. We compared the methylation profiles of three genotypes: line clk-st,cmt3-7 in the clk-st background, and a previously described met1 mutant line that had developed a clark kent phenotype (10). We cloned and sequenced individual polymerase chain reaction (PCR) products from the SUP gene, the long terminal repeat (LTR) of a pericentromeric Athila retrotransposon (15), and the 180-bp centromeric repeat sequence (Fig. 2) (10). Thecmt3-7 mutant showed a nearly complete loss of CpXpG methylation in all sequences tested, but it retained the majority of CpG methylation. In contrast, met1 showed a marked reduction in CpG methylation but had little effect on the level of CpXpG methylation. cmt3-7 displayed variable effects on asymmetric methylation, ranging from no reduction to nearly complete loss at the 5′ end of the SUP locus (Fig. 2). In this region, asymmetric methylation may depend on the presence of CpXpG methylation. Using direct sequencing of PCR products of bisulfite-treated DNA fromSUP and the Athila LTR, we found that three additionalcmt3 alleles (cmt3-4, cmt3-5, andcmt3-6) showed a pattern of methylation similar to that of cmt3-7.

Figure 2

Methylation profiles of the cmt3 andmet1 mutants. (A) Histograms represent the percentage of methylated cytosines found in different contexts inclk-st (blue), cmt3-7 (red), or met1(yellow), derived from cloned PCR products of bisulfite-treated genomic DNA of five regions: the 5′ and 3′ regions of SUP, two regions (A and B) of the Athila LTR, or the 180-bp centromeric repeat. (B) Diagram represents a 1028-nucleotide region of the top strand of the SUP gene, with the height of each bar representing the frequency of methylation found in different sequence contexts, within 15 cloned PCR products. For details of these experiments, see (10).

We analyzed the effect of cmt3-7 on methylation within the direct repeats present in the promoter of the FWA locus. These repeats were previously found to be methylated predominantly at CpG sites in wild-type plants, causing FWA expression to be silenced (16). When this methylation is lost, either spontaneously or in the ddm1 mutant, the FWA gene is overexpressed, causing a dominant late-flowering phenotype (16). Using direct sequencing of PCR products from bisulfite-treated genomic DNA, we found that the CpG methylation pattern was similar in line clk-st and in cmt3-7. However, this CpG methylation was lost in a met1 mutant line that had developed an fwa late-flowering mutant phenotype. Furthermore, no fwa-like late-flowering phenotypes have been observed in any of the cmt3 alleles, even after several generations of inbreeding. Thus, the cmt3 mutations do not appear to affect the CpG methylation or gene silencing at theFWA locus.

To determine whether loss of CpXpG methylation in thecmt3 mutants is genome-wide, we performed Southern blot analysis with methylation-sensitive restriction enzymes. Bothcmt3-5 and cmt3-7 showed an increased level of enzyme digestion at the Athila LTR sequences with the enzymes Eco RII [which is inhibited by methylation of the inner cytosine within its recognition site CC(A/T)GG] and Msp I (inhibited by methylation of the outer cytosine of its recognition site CCGG). However, thecmt3 mutants showed a level of digestion equal to the wild type with the enzymes Hpa II and Hha I, which are inhibited by CpG methylation in their recognition sites (Fig. 3). Using similar restriction enzyme analyses, we found that cmt3 mutants exhibit decreased CpXpG methylation, but not CpG methylation, at the centromeric 180-bp repeat sequence (5) and at the Ta3 retrotransposon sequence (17). We also analyzed cmt3-2, a strong CMT3 allele in the Nossen genetic background (13). This allele showed increased digestion with Msp I, but not with Hpa II, using both a 180-bp centromeric repeat probe (Fig. 3) and an Arabidopsisribosomal DNA probe (5). In summary, the cmt3mutants showed decreased CpXpG methylation at all sequences examined.

Figure 3

Southern blot analysis of cmt3 mutants. (A) Genomic DNA of wild-type Ler (left) andcmt3-5 (right) digested with the indicated restriction enzyme. Blot was probed with an Athila LTR probe (10). (B) Genomic DNA of wild-type Nossen (left) andcmt3-2 (right) digested with the indicated restriction enzyme. Blot was probed with a 180-bp centromeric repeat probe (5). Note the more intense lower molecular weight bands in the Msp I digest of the cmt3-2 mutant.

We tested the role of CMT3 in the silencing of endogenous Arabidopsis retrotransposons. We analyzed the expression of an Athila sequence that was previously shown to be heavily methylated and silenced in wild-type plants but transcriptionally activated in several Arabidopsis silencing mutants, including met1 (18). Figure 4A shows that 2.5-kb and 1.2-kb Athila-related transcripts are indeed activated in thecmt3-7 mutant line. These transcripts are similar in size to those activated by other silencing mutants (18). We also tested for expression of the Ta3 element, a copia-like retrotransposon previously found to be transcriptionally silent in both the wild type and the ddm1 mutant (17, 19). A 5.3-kb transcript was easily detected in cmt3-7, but no expression was observed in the wild-type line clk-st (Fig. 4B). Using similar analyses, we did not see activation of two additional retrotransposons: Evelknievel (11), or Tar17, which was previously shown to be reactivated in the ddm1 mutant (19). Together, these results demonstrate thatCMT3 is required for maintaining gene silencing at a subset of retrotransposon sequences.

Figure 4

Retrotransposon expression in cmt3mutants. Blots containing 40 μg of total RNA from whole shoots of line clk-st (left) or cmt3-7 (right) were hybridized with either an Athila probe (A) or a Ta3 probe (B) (10). The positions of molecular size markers (in kilobases) are indicated.

Our results suggest that CMT3 is specific for CpXpG methylation—a specificity different from that of the DNMT1/MET1 class of methyltransferases. Because cmt3 mutants show a loss of CpXpG methylation in a background that is wild type forMET1, MET1 cannot substitute for the function of CMT3 at these sites. This corroborates earlier observations of distinct CpG and CpXpG methylases that could be purified from plant extracts (20), and is consistent with observations (21) suggesting that a mutation of a maize CMT3 homolog,Zmet2, causes a specific reduction in CpXpG methylation (22). CMT genes have thus far only been found in the plant kingdom (23), which agrees well with the observation that plants have a much higher incidence of CpXpG methylation than do other organisms such as mammals (4).

The differential reactivation of gene expression observed in thecmt3 mutants suggests a model where different loci may depend preferentially on either CpXpG or CpG methylation as the main mechanism of gene silencing. For instance, SUP and the Ta3 retrotransposon appear to depend more heavily on CpXpG methylation, whereas FWA and possibly Tar17 rely more on CpG methylation. Athila sequences require both types of methylation, because Athila-related transcripts are activated in both cmt3 and met1mutants.

Despite a nearly complete loss of genomic CpXpG methylation, nullcmt3 mutants are morphologically normal, even after five generations of inbreeding. In contrast, met1 mutants exhibit severe developmental abnormalities (3, 7). One explanation for this is that CpXpG and CpG methylation may act in a partially redundant fashion to silence most genes. Viability despite severe loss of genomic methylation makes Arabidopsis an ideal model system for elucidating the roles of DNA methylation in epigenetic and developmental processes.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: jacobsen{at}


View Abstract

Stay Connected to Science

Navigate This Article