Hypermethylated SUPERMAN Epigenetic Alleles in Arabidopsis

See allHide authors and affiliations

Science  22 Aug 1997:
Vol. 277, Issue 5329, pp. 1100-1103
DOI: 10.1126/science.277.5329.1100


Mutations in the SUPERMAN gene affect flower development in Arabidopsis. Seven heritable but unstable supepi-alleles (the clark kent alleles) are associated with nearly identical patterns of excess cytosine methylation within theSUP gene and a decreased level of SUP RNA. Revertants of these alleles are largely demethylated at theSUP locus and have restored levels of SUP RNA. A transgenic Arabidopsis line carrying an antisense methyltransferase gene, which shows an overall decrease in genomic cytosine methylation, also contains a hypermethylated supallele. Thus, disruption of methylation systems may yield more complex outcomes than expected and can result in methylation defects at known genes. The clark kent alleles differ from the antisense line because they do not show a general decrease in genomic methylation.

DNA methylation is emerging as an important component of cell memory, the process by which dividing cells inherit states of gene activity. In mammals, methylation appears to play a key role in processes such as genomic imprinting and X-chromosome inactivation, and in plants methylation is correlated with a number of phenomena, including silencing of duplicated regions of the genome (1).

Arabidopsis mutants at the DDM1 andDDM2 loci have a reduced overall level of cytosine methylation and display a number of developmental defects (2). Transgenic Arabidopsis plants expressing an antisense cytosine methyltransferase RNA also exhibit abnormalities including a number of floral defects resembling the phenotypes of known floral homeotic mutants (3, 4). These experiments suggest a direct cause and effect relation between DNA methylation and proper regulation of developmentally important genes. We describe here a class of epi-mutations in Arabidopsis that appear to be caused by overmethylation of the flower development geneSUPERMAN (SUP).

Seven independent mutants were identified [clark kent(clk) 1 through 7] with phenotypes similar to but weaker than that of the known sup mutants (5,6). Wild-type Arabidopsis flowers (Fig.1A) contain six stamens (the male reproductive organs) and two central carpels that fuse to form the female reproductive structure. The sup-5 allele (Fig. 1B) (7), which contains a nearly complete deletion of theSUP gene (8), produces an increased number of stamens [12.3 ± 0.3 (mean ± SE)] and carpels (2.9 ± 0.1) on the first 10 flowers produced on the plant. Theclk-3 allele (Fig. 1C) has an average of 7.8 ± 0.3 stamens and 3.4 ± 0.1 carpels, whereas the weakerclk-1 allele has an average of 6.4 ± 0.1 stamens and 3.2 ± 0.1 carpels.

Figure 1

(A) Wild-type Landsberg erecta (Ler) flower showing the normal number of six stamens and the central gynoecium, which consists of two fused carpels. (B) Flower from a sup-5 homozygote containing 11 stamens and three incompletely fused carpels. (C) Flower from a clk-3 homozygote containing nine stamens and three incompletely fused carpels. (D) An F1 flower from a cross between a sup-5 homozygote and aclk-3 homozygote containing 10 stamens and three incompletely fused carpels. (E) Flower from a transgenicclk-3 plant containing a 6.7-kb wild-type SUPgenomic fragment (12), containing the normal number of six stamens and two fused carpels. (F) An F1 flower from a cross between the AMT line and a clk-3 homozygote containing 10 stamens and three incompletely fused carpels.

In F1 complementation tests, clk mutants fail to complement sup mutants (Fig. 1D). However, these tests are complicated by the fact that both clk and supmutants are semidominant (9). Also, in F2progeny of these crosses, about 1 to 3% of the plants are wild type. These data suggested that CLK might define a separate gene linked to SUP. However, further analysis shows that theclk mutants are SUP alleles.

First, clk and sup are very closely linked.clk-3 perfectly segregated with a Hind III restriction fragment length polymorphism detected with a 5.5-kb Eco RISUP genomic fragment in 45 meiotic progeny (10,11). In addition a cis-trans test was performed; ifsup and clk are different genes, they should have the same double-heterozygote phenotype in cis or trans. clkand sup homozygotes were crossed to make trans clk-3 +/+ sup-5 heterozygotes, which were then crossed to wild-type plants. These F1 progeny were analyzed for plants with sup-like cis clk sup/+ + heterozygote phenotypes. However, all 7863 F1 progeny from this cross appeared wild type. Because this number of plants represents a recombinant every 0.025 cM, or one about every 3.3 kb, the results of this test suggest either that clk and sup are allelic, or that if clk and sup are two different genes, they are very closely spaced in the genome.

Second, a 6.7-kb genomic clone containing only the SUPcoding region and ∼5 kb of upstream sequence, which has been shown to rescue the sup mutant phenotype (11), complements the clk-3 phenotype in transgenic plants (Fig. 1E) (12). However, other genomic clones spanning 40 kb on the distal side of SUP and 25 kb on the proximal side fail to complement clk-3 (13), apparently precluding the possibility that CLK is a separate gene very closely linked to SUP.

In situ hybridization experiments show that SUP RNA expression is reduced in clk-3. In wild type, expression ofSUP RNA occurs early during floral meristem development in the incipient stamen primordia (11). In clk-3homozygotes, however, this expression was reduced in some floral meristems and undetectable in others (Fig.2).

Figure 2

SUP RNA expression in wild-type and clk-3. Longitudinal sections of the inflorescences of 14-day-old plants of wild-type Ler or the clk-3 mutant were mounted on the same slide and hybridized with a SUPantisense probe as described in (11). Medial sections of stage 4 flower primordia (19) were photographed with bright-field–dark-field double exposure. Yellow spots represent silver grains exposed by the 35S-labeled SUPprobe, after a 7.5-week exposure. Bar, 120 μM.

Despite the evidence that clk and sup are allelic, sequencing of the SUP coding region fromclk-1, -2, -3, and -5, and the entire 6.7-kb SUP genomic region from theclk-1 and -3 alleles revealed no nucleic acid sequence differences from the wild type. In addition, the clonedSUP gene from a clk-3 genomic library complements the clk-3 and sup-5 mutants in transgenic plants (12), as if cloning the clk-3 allele restores it to wild type. Together these results suggest that the clkalleles represent an alternative epigenetic state of the SUPgene (14).

To examine whether the clk mutants exhibit the genetic instability that is characteristic of many other epigenetic phenomena, we constructed a clk-3 gl1-1 double mutant and analyzed the selfed progeny [gl1-1 maps 10 cM from SUP and eliminates epidermal hairs (15)]. Of 586 plants, 17 had a wild-type or nearly wild-type phenotype, but were still hairless. Three of these lines were complete revertants; the selfed progeny from these plants segregated 3:1 for wild type:clk plants, and in subsequent generations they segregated homozygous clkand wild-type lines. The other 14 lines appeared to contain partially reverted alleles. Analysis of one complete revertant (number 6) by in situ hybridization showed that wild-type levels of SUP RNA expression were restored (16).

We analyzed methylation patterns within the SUP gene in different genotypes using bisulfite genomic sequencing (17). Although there was no cytosine methylation detected in the wild-type or in a sup nonsense allele [sup-1(11)], extensive methylation was found in theclk alleles (shaded regions, Fig.3A), covering the start of transcription and most of the transcribed region. The clk-3 allele contained six more methylcytosines (a total of 211 detected) than the weaker clk-1 allele (Fig. 3B), possibly providing an explanation for the slight difference in phenotypic strength of these alleles. In clk-3 revertant 6, only 14 of the original 211 methylcytosines remained. Thus, phenotypic reversion is correlated with both a restoration of the wild-type RNA expression level and a decrease in cytosine methylation of the SUP DNA.

Figure 3

(above). (A) Summary of the methylation pattern observed in theSUP genomic region from the clk-1 andclk-3 alleles and from the AMT line. Gene diagram shows the start of transcription (arrow), the single intron (shaded area), the putative start of translation (M), the stop codon, and the poly(A) addition site (11). Boxes below the gene show the regions of either the top or bottom strand analyzed by bisulfite genomic sequencing; open boxes denote no methylation detected in all three genotypes, solid boxes indicate that all of the cytosines in the region were methylated, and hatched boxes indicate that only some of the cytosines in the region were methylated. Numbers indicate the length in nucleotides (nt) of each region analyzed. (B) Exact pattern of methylation detected in six different genotypes: clk-1(1), clk-3 (3), the AMT line (M), revertant 6 (R), wild-type Ler, and wild-type C24. Open circles indicate that cytosine methylation was not detected in any of the six genotypes. Boxes with no accompanying symbol indicate that methylation was detected at this site in clk-1, clk-3, and the AMT line, but not in either of the wild-type controls or in revertant 6. All exceptions to this general pattern are indicated by symbols above or below the box. For example, 3,1,R indicates that methylation was detected inclk-3, clk-1, and revertant 6, but not in the other genotypes. Because the data were obtained by direct sequencing of PCR-amplified genomic DNA (17), an average level of methylation was determined for each cytosine. Thus, solid boxes indicate that more than 50% of the cytosine at this position was methylated, whereas half-shaded boxes indicate that less than 50% of the cytosine was methylated. The region between the vertical lines on the top strand was analyzed from other genotypes as described in the text. Arrow indicates the beginning of the SUP RNA as deduced from the longest cDNA detected and corresponds to nucleotide −202 relative to the putative start codon as described in (11). Methylation at the two underlined GATC restriction sites was confirmed with the methylation-sensitive restriction enzyme Sau 3AI and its methylation-insensitive isoschizomer Mbo I.

The methylation pattern in clk was dense and essentially non–sequence specific; both symmetric (CG and CXG) and nonsymmetric cytosines were methylated (Table 1). However, the pattern of methylation was nearly identical in differentclk alleles (Fig. 3B). For example, clk-1 andclk-3 shared 204 methylcytosines; seven sites were methylated in clk-3 but not clk-1, and one site was methylated in clk-1 but not clk-3. Also, in the most densely methylated region on the top strand (Fig. 3B),clk-2, clk-5, and clk-6 had a hypermethylation pattern very similar to that seen in clk-1and clk-3. The reproducibility of this pattern in independently isolated clk lines suggests that the mechanism by which these sequences become methylated is rather specific. In addition, the sequences methylated in the SUP gene appear to be single copy (18). This suggests that the SUPhypermethylation could be produced or maintained by a mechanism other than that responsible for most of the methylation in plants, which is mostly found at repetitive sequences and primarily consists of symmetric sites (1, 2).

Table 1

Sequence context of the methylated cytosines in theSUP region inclk-3.

View this table:

A number of Arabidopsis antisense cytosine methyltransferase (AMT) lines exhibit phenotypes resembling sup mutants (3, 4). Crosses between an AMT homozygote exhibiting a sup floral phenotype (line 10) (3) and either clk-3 or sup-5 plants yielded F1 plants with a sup phenotype (Fig. 1F), whereas F1 plants resulting from crosses of the AMT line to a wild-type Ler plant had a wild-type floral phenotype (16). This suggests that the AMT line contains a defective SUPallele. Hypermethylation of the SUP gene was found in the AMT line in a pattern similar to that seen in the clk lines. One hundred and eighty-six methylcytosines were detected, all but three of which were in the same positions as those in clk-3 (Fig.3B). Thus, although overall methylation in this AMT line is decreased by up to 90% (3), the SUP gene has become hypermethylated. These results challenge the original interpretation of the AMT phenotype, in that the various phenotypes in these lines were predicted to be caused by demethylation of specific genes (3, 4).

To examine whether the clk lines have general demethylation defects similar to those seen in the AMT lines, we analyzed the methylation status of the 180–base pair (bp) centromeric repeat (Fig.4) and the ribosomal DNA loci (16) by DNA blot analysis (2). Whereas in the AMT lines these sequences were hypomethylated (Fig. 4) (3,4), five clk alleles showed normal methylation of these repetitive genes, suggesting that the defects in theclk lines are different from those in the AMT line. One possible interpretation of these results is that the AMT lines cause misregulation of a component of the methylation pathway other than the methyltransferase, resulting in hypermethylation of some genomic regions. If this hypothesized component were mutated in theclk lines, this might cause only a portion of the AMT phenotype, namely, hypermethylation of SUP.

Figure 4

(right). Repetitive DNA methylation in clk. Genomic DNA of the AMT line (1), wild-type Ler (2), and clk-3 (3) was digested with the methylation-sensitive enzyme Hpa II, fractionated on an agarose gel, and probed with a 180-bp centromeric repeat clone (2). Undermethylated DNA in the AMT line is detected as a low molecular weight ladder, not present in the wild type or clk-3.

Note added in proof: We have found a clk-like pattern of methylation at the SUP locus in fonl-2 and fonl-3 [see (14)].


View Abstract

Navigate This Article