Maternal Control of Embryogenesis by MEDEA, a Polycomb Group Gene in Arabidopsis

See allHide authors and affiliations

Science  17 Apr 1998:
Vol. 280, Issue 5362, pp. 446-450
DOI: 10.1126/science.280.5362.446


The gametophytic maternal effect mutant medea(mea) shows aberrant growth regulation during embryogenesis in Arabidopsis thaliana. Embryos derived from meaeggs grow excessively and die during seed desiccation. Embryo lethality is independent of the paternal contribution and gene dosage. Themea phenotype is consistent with the parental conflict theory for the evolution of parent-of-origin–specific effects.MEA encodes a SET domain protein similar to Enhancer of zeste, a member of the Polycomb group. In animals,Polycomb group proteins ensure the stable inheritance of expression patterns through cell division and regulate the control of cell proliferation.

The plant life cycle alternates between diploid and haploid generations, sporophyte and gametophyte, as the haploid spores undergo several cell divisions before the gametes finally differentiate and fuse to produce the diploid zygote. We identified an Arabidopsis thalianamutant, medea (mea), in which self-fertilization of the heterozygote produces 50% aborted seeds that collapse, accumulate anthocyanin, and do not germinate. This ratio of defective to normal seeds is consistent with a gametophytic control of the defect, because half the haploid gametophytes receive the mutant allele. Heterozygous embryos abort if the mutant allele is derived from the female (Fig. 1A), but develop normally if it is derived from the male (Fig. 1B and Table1). Embryos derived from mutant eggs abort irrespective of the paternal contribution (1). Thus, the mea mutant displays maternal-effect embryo lethality (2). In flowering plants, embryo development is affected by both the female gametophyte (3) and the sporophytic tissue of the parent plant (4). The survival of the resultant embryo depends on the presence of a wild-type MEA allele in the genome of the female gametophyte.

Figure 1

Seed and embryo development in meaplants. (A) Silique resulting from a cross betweenmea-1 (female) and a wild-type plant (male) of the Columbia (Col) ecotype. Seeds derived from mea female gametes turn white and collapse. (B) Silique resulting from a cross between Col (female) and mea-1(male). Histological analysis of embryos was conducted in heterozygous mea-1 plants using semithin sections (C and D) and cleared seeds (E and F). (C) Late globular wild-type embryo. (D) Late globular mea-1 embryo. (E) Late heart stage wild-type embryo. (F) Late heart stagemea-1 embryo. Magnification: bar is 215 μm in (A) and (B); 12 μm in (C) and (D); and 40 μm in (E) and (F).

Table 1

Seed phenotype of reciprocal crosses betweenmea and diploid or tetraploid wild-type plants. Green or dry siliques resulting from self pollination or out-crosses to Columbia (Col) or a tetraploid plant were opened and the seeds classified as unfertilized ovules, normal, or aborted seeds. In crosses involving tetraploids, fertility is reduced because of a large fraction of unfertilized ovules (29). The relative increase in unfertilized ovules in the “Col × 4n” cross is due to longer siliques and a larger number of ovules per silique in Col plants (Col: 65.1 ± 6.4; mea-1: 39.7 ± 5.9; 4n: 43.6 ± 3.2). Pollen viability in tetraploids is reduced such that only ovules at the top of the silique get fertilized. The average number of seeds per silique initiating development is similar (Col: 17.9 ± 2.1;mea-1: 22.1 ± 5.9; 4n: 26.1 ± 4.4). The “normal” class includes brown, quite regularly shaped seeds that are smaller than those of the wild type and are common in crosses involving plants of different ploidy. They do not show the collapsed, black mea phenotype. n (seeds), number of seeds scored.

View this table:

Fertilization of the egg and central cell generates the diploid zygote and the triploid primary endosperm. Endosperm resulting from a cross between a wild-type male and a mea-1 female carries two mutant mea-1 alleles and one wild-type MEAallele. To determine whether seed abortion is caused by a mutation in a dosage-sensitive gene, we introduced additional wild-typeMEA copies: In a cross between a mea-1heterozygous female and a wild-type tetraploid male, half of the fertilized seeds abort (Table 1) and the mutant mea-1 allele is not recovered in the progeny (0/41). In control crosses, seeds rarely abort (Table 1), and paternal mea-1 alleles are transmitted to half the offspring (66/124) suggesting that an additional paternal wild-type MEA allele is unable to rescue maternal-effect lethality. Thus, mea either affects a maternally produced cytoplasmic factor in the egg or central cell (or both), or disrupts an imprinted gene expressed from the maternal allele.

During early stages of embryogenesis the development ofmea-1 embryos is indistinguishable from wild-type siblings in cleared or sectioned specimens (5). Visible differences between wild-type and mea-1 embryos began at the late globular stage (Fig. 1, C and D). Globular mea-1 embryos show excess cell proliferation and enlarge radially symmetrical. When wild-type embryos reach the mid to late heart stage, siblingmea-1 embryos are still globular and contain small vacuolated cells with curvilinear cell walls and sometimes irregular cell divisions in the ground tissue and procambium. Suspensor and hypophysis are normal, and cotyledons initiate synchronously as in the wild type. Thus, despite increased cell proliferation and occasional irregular cytokinesis, morphogenetic progression is normal. However, each stage is prolonged and includes more division cycles, and morphogenesis is delayed. As a consequence, giant heart stagemea-1 embryos (Fig. 1, E and F) are present along with late torpedo or cotyledonary stage wild-type embryos. mea-1 heart stage embryos have supernumerary cell layers (mea-1: 19.6 ± 1.1; wild type: 13.0 ± 0.9). When wild-type siblings are fully differentiated, mostmea-1 embryos have reached the late heart stage and are up to 10 times larger than normal. mea embryos degenerate during desiccation. These results suggest that mea controls cell proliferation during embryogenesis, allowing morphogenesis to progress normally, albeit slowly.

Endosperm development in mea-1 seeds is indistinguishable from that of the wild type at early stages. When cellularization begins normally in wild-type seeds at the transition from the globular to the heart stage, no cellularization is observed in sibling mea-1seeds. Although nuclear divisions take place more slowly inmea-1 endosperm, the distribution of endosperm nuclei is as in the wild type. Partial cellularization occurs at the micropyle whenmea-1 embryos reach the late heart stage in desiccating seeds, but because fewer nuclei have been generated, most of the central cell is devoid of nuclei. Thus, in mea-1 seeds, the development of both fertilization products is delayed but morphogenesis proceeds normally, and it appears that the embryo shows increased cell proliferation at the expense of the endosperm.

To determine whether mea-1 is a gain-of-function or loss-of-function mutation we introduced mea-1 into a tetraploid background, which produces diploid gametophytes carrying either none, one, or two mea-1 alleles (6). Because tetraploids carrying one (simplex) or two (duplex) mutant alleles could have been recovered, we considered models for both possibilities with mea-1 as either recessive or dominant (Table 2). The observed inheritance of seed abortion and kanamycin resistance is consistent with a simplex recessive model (X2 < χ2 0.05[1] = 3.84). The recessive nature ofmea-1 is confirmed by the high transmission frequency of kanamycin resistance (247/331 = 75%). If mea-1 was dominant, it would be exclusively transmitted through pollen at a frequency of about 47% in a simplex tetraploid (7). These results suggest that mea is a loss-of-function mutation. Thus, the wild-type function of MEA is to restrict cell proliferation during embryogenesis.

Table 2

Segregation of embryo lethality and kanamycin resistance in a 4n mea line. The tetraploid nature of this plant was confirmed at the cytogenetic level (35). Dry siliques were opened and their seeds classified as normal or aborted seeds displaying the mea phenotype. The progeny of this plant was tested for sensitivity (KanS) or resistance to kanamycin (KanR) linked to mea-1. Expected values of the two phenotypes scored are given for four different models [(1) to (4)]. For the calculation of expected progeny classes, a spontaneous embryo abortion rate of 1.5%, as determined for the parental tetraploid, was included. The coefficient of double reduction was taken as c = 0.1 (29).

View this table:

We cloned and characterized the MEA gene as mea-1was tagged with derivatives of the maize Ac/Dstransposon system (8): mea-1 and theDs element cosegregated without detectable recombination. This mutation was unstable in the presence of Actransposase, and large revertant sectors could be identified in plants heterozygous for both mea-1 and Ac. Using genomic fragments flanking the Ds element, we identified and sequenced a MEA cDNA (Fig. 2A) (9). TheMEA gene is similar to Enhancer of zeste[E(z)], a Drosophila melanogaster protein involved in the regulation of homeotic genes. The highest similarity between the two proteins (55% amino acid identity) is found at the COOH-terminus (Fig. 2B), in the SET domain, which was named after the three founding members of the family inDrosophilaSuppressor of variegation 3-9[Su(var)3-9], E(z), and trithorax(trx) (10, 11). Although the function of the SET domain is unknown, members of this family are thought to regulate gene expression by associating with chromatin and controlling access of transcription factors (12). As members of thePolycomb and trithorax group, SET domain proteins regulate homeotic gene expression (13). Many show parent-of-origin–specific effects and regulate cell proliferation (11, 14). For instance, the human homologsAll-1/Hrx and Enx-1 are involved in the control of lymphocyte proliferation (15), and theCaenorhabditis elegans homolog MES-2 supports survival of the germ line (16). The plant E(z)homolog CURLY LEAF (CLF) regulates expression of floral homeotic genes (17). MEA shares 43% identity with E(z) in the CXC domain, a cysteine-rich region NH2-terminal to the SET domain (Fig. 2B). The CXC domain and five additional highly conserved cysteine residues (Fig. 2B) are unique to E(z) and its vertebrate and plant homologs (17). Although the function of these regions is unknown, they are required for E(z) activity (18). The similarity to SET domain proteins suggests that MEA controls cell proliferation by regulating gene expression through modulation of higher-order chromatin structure.

Figure 2

Sequence analysis and expression profile of MEA. (A) Deduced amino acid sequence ofMEA (34). An acidic region similar to that in the trx protein is underlined. The five cysteines that are conserved in E(z) homologs are boldface, and the 18 cysteine residues of the CXC domain are boldface and underlined. Basic residues of a putative bipartite nuclear localization signal are indicated by asterisks above the symbol. The 115–amino acid SET domain is boxed. (B) Schematic alignment of MEA andE(z) shows the relative position and amino acid identity of the SET and CXC domains, the putative nuclear localization signals (N), the five conserved cysteines (C5), and the acidic domain (A). (C) DNA sequences flanking the mea-1 Dsinsertion and derivative alleles. The 8-bp target site duplication is indicated by boldface. All revertants analyzed showed the wild-type sequence. In the stable mea-2 excision allele, a 7-bp footprint remains; an additional base is altered (underlined). The mea-2 footprint introduces two stop codons (lower case). (D) RT-PCR analysis of MEAduring flower and seed development. The three panels show amplification of MEA, and actin-11 (ACT11) and the seed storage protein At2S2 as controls for cDNA synthesis (33). RNA was isolated from floral buds (B), unpollinated carpels (U), pollinated carpels (P), and siliques containing embryos at the preglobular (PG), globular (G), heart (H), torpedo to early cotyledonary (T/C), cotyledonary (C), early (EM), and late maturation (LM) stages. (M) indicates the marker lane and (G) genomic DNA as a control.

To confirm that the isolated gene corresponds to themea mutation, we sequenced the region spanning the insertion site in three independently recovered revertants (19). TheDs element in mea-1 inserted NH2-terminal to residues of the SET domain that are invariant among the E(z) homologs. Ds excisions usually create characteristic footprints, and phenotypic revertants should restore the open reading frame of the disrupted gene. In all three revertants, the sequence was that of the wild type (Fig. 2C), indicating a strong sequence constraint on this region of the SET domain. In the phenotypically mutant excision allele mea-2, a 7–base pair (bp) footprint remains, introducing two stop codons (Fig. 2C).

We analyzed the expression profile of the MEA transcript using reverse transcription–polymerase chain reaction (RT-PCR) (20) on floral tissues and developing siliques (Fig. 2D).MEA is not expressed at early stages of flower development during early megagametogenesis. MEA expression is first detectable in unpollinated siliques that contain maturing gametophytes indicating maternal expression. The transcript remains detectable throughout the morphogenetic phase of embryogenesis and starts to disappear during seed maturation. Thus, MEA is either an unusually stable mRNA that is maternally deposited in egg or central cell, or both, and persists for 2 weeks throughout seed development orMEA is expressed both maternally and zygotically.MEA expression after fertilization may be regulated by genomic imprinting, because paternally provided copies cannot rescue embryo lethality.

The regulation of cell proliferation and growth during seed development by MEA is under maternal control. Haig and Westoby proposed that parent-of-origin–specific effects evolved as a consequence of an intragenomic conflict over the allocation of nutrients from the mother to its offspring (21). Although their theory is usually discussed with respect to imprinting, it is equally applicable to other postmeiotically established differences such as a maternal effect of cytoplasmic nature. The intragenomic conflict theory predicts that paternally expressed genes would tend to promote the growth of the embryo and maternally expressed genes would tend to reduce it. Supporting evidence has been provided by studies on imprinted genes in mice and humans (22) and from the manipulation of entire genomes in flowering plants (23). Our observations on themea mutant phenotype suggest that similar molecular mechanisms operate in animals and plants to control cell proliferation and to mediate parent-of-origin–specific effects.

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

  • Present address: Graduate Program in Natural Resources, Ohio State University, Columbus, OH 43210–1085, USA.


Stay Connected to Science

Navigate This Article