Interaction Between Maternal Effect and Zygotic Effect Mutations During Maize Seed Development
Matthew M. S. Evans, Jerry L. Kermicle


Double fertilization of the embryo sac by the two sperm cells of a pollen grain initiates seed development. Proper development of the seed depends not only on the action of genes from the resulting embryo and endosperm, but also on maternal genes acting at two stages. Mutations with both sporophytic maternal effects and gametophytic maternal effects have been identified. A new maternal effect mutation in maize, maternal effect lethal1 (mel1), causes the production of defective seed from mutant female gametophytes. It shows reduced pollen transmission, suggesting a requirement in the male gametophyte, but has no paternal effect on seed development. Interestingly, the defective kernel phenotype of mel1 is conditioned only in seeds that inherit mel1 maternally and are homozygous for the recessive allele (endogenous to the W22 inbred line) of either of two genes, sporophyte enhancer of mel1 (snm1) or snm2, suggesting redundancy between maternally and zygotically required genes. Both mel1 and snm1 map to the short arm of chromosome 2, and snm2 maps to the long arm of chromosome 10. The mode of action of mel1 and the relationship between mel1 and snm1 and snm2 are discussed.

THE plant life cycle alternates between a diploid sporophyte and a haploid gametophyte generation. In contrast to animals, the haploid products of meiosis do not differentiate into the gametes but undergo a few rounds of mitosis before producing gametes. Another distinguishing feature of plant reproduction is the dependence of the gametophytes on the expression of genes from their own haploid genomes as well as the diploid genome of the parent organism. This is obvious in lower plants with large free-living gametophytes but is also the case in angiosperms, the gametophytes of which are highly reduced in size. Evidence for this requirement is the lethality of large genomic deletions to the gametophyte (Patterson 1978; Viziret al. 1994; Vollbrecht and Hake 1995).

The angiosperm female gametophyte, called the embryo sac, consists of four cell types: synergid, antipodal, egg, and central cell (Reiser and Fischer 1993; Drewset al. 1998; Grossniklaus and Schneitz 1998; Yang and Sundaresan 2000). Double fertilization of the egg and central cell produce the embryo and endosperm, respectively. The embryo sac develops embedded within a specialized sporophytic organ, the ovule. Consequently, after fertilization the seed also develops within maternal sporophytic tissue. Recent genetic evidence distinguishes three classes of maternal effect genes in angiosperms: (1) genes required in the sporophyte for proper development of the embryo sac (Hulskampet al. 1995); (2) genes required in the maternal sporophyte for normal embryo development (Colomboet al. 1997; Rayet al. 1997); and (3) genes required in the female gametophyte for proper embryo development (Castleet al. 1993; Ohadet al. 1996; Chaudhuryet al. 1997; Grossniklauset al. 1998; Kiyosueet al. 1999; Luoet al. 1999; Springeret al. 2000).

This last class, gametophytic maternal effect genes, can have its effect through several distinct mechanisms. Mutations in genes required in two doses in the endosperm will show a maternal effect because the endosperm is triploid with two maternal genomes and one paternal genome (e.g., floury3 in maize; Ma and Nelson 1975). In this first mechanism, the gene is not strictly required in the embryo sac; the effect is an indirect result of the specialization of the endosperm. Mutations that cause abnormal embryo sac morphology can have maternal effects on seed development (e.g., indeterminate gametophyte1 in maize; Kermicle 1971; Lin 1978, 1981). Mutations in genes that are expressed in the embryo sac but whose gene products are stored cytoplasmically and function after fertilization show a maternal effect [proposed for PROLIFERA (PRL) in Arabidopsis; Springer et al. (2000)]. Mutations in genes that are imprinted such that the maternal copy is active and the paternal copy is inactive also show a maternal effect [e.g., MEDEA in Arabidopsis; Kinoshita et al. (1999); Vielle-Calzada et al. (1999)]. It can be difficult to distinguish these last two mechanisms from one another. They are clearly distinguishable only if transcripts are present only prior to fertilization but not after (perdurance) or if transcripts are present only after fertilization but not before (imprinting). In plants many genes may actually have gametophytic maternal effects through a combination of the last two mechanisms. In animals the last two classes can be distinguished on the basis of inheritance, but this is not the case in normal diploid plants because the maternal gametophyte has only one allele of each gene to pass on to the embryo.

Gametophytic maternal effect mutants in Arabidopsis, fertilization independent endosperm (fie), fertilization independent seed2 (fis2), and medea (mea), all have similar effects. They are poorly transmitted through the female but are transmitted normally through the male. In addition to being required for normal development of the embryo, they are also required to suppress fertilization-independent seed development. These genes have all recently been cloned. FIE and MEA encode WD domain and SET domain proteins, respectively (Grossniklauset al. 1998; Ohadet al. 1999) and have been shown to interact directly (Luoet al. 2000; Spillaneet al. 2000; Yadegariet al. 2000). Interestingly, MEA undergoes imprinting, which could explain its maternal effect (Kinoshitaet al. 1999; Vielle-Calzadaet al. 1999). A relationship between DNA methylation and maternal effects is suggested by the suppression of the mea phenotype by the decrease in dna methylation1 (ddm1) mutation and by the phenocopying of parent-of-origin effects on seed size using METHYLTRANSFERASE1 antisense transgenes (Vielle-Calzadaet al. 1999; Adamset al. 2000).

To better understand the role of embryo sac gene expression in seed development we have isolated a mutant in maize, maternal effect lethal 1 (mel1), with effects on postfertilization development similar to those of mea in Arabidopsis. Progeny of mel1 mutant embryo sacs have defective endosperm and embryo development. Unlike mea, mel1 is poorly transmitted through the pollen, although the progeny derived from mel1 pollen grains are normal. mel1 also interacts with either of two genes, sporophyte enhancer of mel1 (snm1) or snm2, that act after fertilization to condition the defective phenotype. Our results suggest redundancy of genes with different patterns of inheritance.


Genetic stocks and growth conditions: The mel1 mutation arose spontaneously in a cross between two W22 lines. One of several R1-r:standard (R1-r:st) Leaf color1 (Lc1) females pollinated with golden1 (g1) R1-self color (R1-sc) males produced progeny segregating defective kernels. This mutant was given the provisional designation dex-1146 (Kermicle 1978). The mel1 mutation was subsequently maintained by self-crossing or backcrossing to a r1-g:stadler (r1-g) W22 line, generating families segregating mel1 and homozygous for r1-g or R1-sc or segregating both. Stocks of visible genetic markers used for mapping had previously been introgressed into the W22 inbred line by multiple rounds of backcrossing and selection. For mapping experiments, the mel1 stock, which carries the b1-bar allele of the booster1 (b1) gene as well as a recessive allele of the purple plant1 (pl1) gene conferring a sun-red phenotype, was crossed by a W22 stock carrying dominant B1 and Pl1-Rhoades (Pl1-Rh) alleles of these genes that confer intense plant color or by a W22 line carrying the B1-Peru allele of b1, which conditions aleurone color. The F1 plants then were crossed with a standard b1-bar; pl1 W22 tester. Plants of all four combinations, b1-bar pl1, B1 pl1, b1-bar Pl1-Rh, and B1 Pl1-Rh, can be distinguished from one another on the basis of plant color, and seeds carrying b1-bar and B1-Peru can be distinguished on the basis of aleurone color.

Defective seeds from mel1 crosses were routinely germinated on filter paper to identify both root and shoot abnormalities. Seedlings that produced a viable shoot were then transplanted into soil and grown to maturity to use for crosses. Plants were grown to maturity under summer field conditions or in green-houses during long days. In experiments to determine male transmission and mel1 penetrance normal kernels were also germinated on filter paper to ascertain whether there were any seedling defects and to ensure that all planted seeds were included in the final total.

Pollen tube growth: Plants were grown in the greenhouse to flowering. The day before pollen collection all old anthers were removed from the tassels. Pollen was collected in the morning from newly extruded anthers that had not yet dehisced. A single anther was used for each plant. The anther was opened, and the pollen was dusted onto the surface of pollen germination media (Walden 1994). A reference point was established on the agar of the pollen germination plates to ensure photography of the same pollen grains at each time point. Pollen tubes were photographed at 35, 55, 120, 180, and 240 min on a Leica Wild M10 dissecting microscope using darkfield illumination on Kodak Gold 100 ASA print film. Prints were scanned as TIFF files in Adobe Photoshop and imported into NIH image (available by anonymous FTP from Pollen tubes were measured by tracing along the length of the tube and measuring the resulting line in NIH image. Pollen tubes were numbered to compare growth rates of individual tubes between different time points. For wild type a total of 242 pollen tubes from three plants were measured, and for mel1 heterozygotes a total of 269 pollen tubes from four plants were measured. The frequency of pollen tubes of different lengths was then determined for each genotype after measurement.

Histology: For embryo sac phenotypes ovules were fixed, stained, and analyzed according to the method of Vollbrecht and Hake (1995). For mature pollen phenotypes mature tassel spikelets were collected 1 or 2 days prior to anther dehiscence on the basis of the dehiscence of anthers from other spikelets located on the same tassel branch. Spikelets were fixed overnight in FAA (50% ethanol, 10% formalin, 5% acetic acid) at 4° and rinsed twice in 95% ethanol and once in 70% ethanol prior to storage in 70% ethanol at 4°. Anthers were dissected out of these spikelets, rehydrated through an ethanol series, placed in a drop of 2.5 mg/ml KI and 0.45 mg/ml I2, and opened to release pollen grains. Seven hundred to 1000 pollen grains were examined for each plant using a Nikon Optiphot microscope.

For embryo development, whole kernels were removed from ears of mel1 heterozygotes that had been pollinated with wild-type W22 males. A single ear was used for each time point after pollination. Normal and mutant siblings could be distinguished from one another on the basis of the smaller size and/or paler yellow color of mutant kernels at all time points. Kernels were fixed overnight at 4° in FAA and rinsed twice in 95% ethanol and once in 70% ethanol prior to storage in 70% ethanol at 4°. Embryos were dissected from fixed kernels under 70% ethanol to prevent dessication. Embryos were dehydrated through an ethanol series, critical point dried, sputter coated, and visualized on a Hitachi S-570 scanning electron microscope.

Figure 1.

Seed phenotypes of mel1 progeny. (A) Ear of a mel1 heterozygote pollinated with a wild-type male. mel1 embryo sacs produce miniature, etched, and defective kernels. (B) Ear of a homozygous wild type pollinated with a mel1 heterozygous male. All kernels are normal. (C) Germinal face of normal seeds showing large, bilaterally symmetrical embryos. (D) Germinal face of mel1 seeds with twin embryos (far left), small embryos (middle two), and no visible embryo (far right).

Mapping with simple sequence repeat markers: DNA was extracted from samples using the protocol of Dellaporta (1994). PCR reactions were performed on a PTC-200 thermal cycler (MJ Research, Watertown, MA). The amplification conditions were the same as the “touchdown” profile of Senior et al. (1998) except that the last cycle was repeated 30 times instead of 20 times prior to terminating with a continuous 4° cycle. The 15-μl reaction mix consisted of 3 pmol of each primer, 2.5 mm MgCl2, 100 μm each dNTP, 10 mm Tris pH 9, 50 mm KCl, 0.1% Triton X-100, 1 mg/ml purified BSA (New England Biolabs, Beverly, MA), 0.6 units Taq DNA polymerase (Promega, Madison, WI), and ~30 ng template DNA. After amplification, 3 μl of loading dye (30% glycerol, 0.25% bromophenol blue, 0.25% xylene cyanol) was added to each sample, and 6 μl of each mix was electrophoresed on 4% Metaphor (FMC Bioproducts, Vallensbaek, Denmark) agarose gels in 1× TBE (Sambrooket al. 1989). After electrophoresis, gels were stained in 0.5 μg/ml ethidium bromide and visualized on a UV transilluminator.

Allelic constitution was first determined for the W22 and W64A parental lines. Simple sequence sepeat (SSR) markers were chosen on the basis of their bin location listed in MaizeDB ( Attempts were made to choose markers from every other bin and from the terminal bins of each chromosome to ensure complete genomic coverage. Approximately 250 SSR markers were tested on the W22 and W64A parental lines to ascertain which had detectable polymorphisms. Polymorphic markers were then tested on 45 mel1 progeny from the cross of a mel1/+ W22/W64A hybrid female by a wild-type W22 inbred male to ascertain which markers cosegregated with the mutant phenotype. Those markers for which fewer than 50% of the mel1 individuals were heterozygous were then tested on an additional 27 mel1 individuals and on 45 normal individuals to show that the distorted ratio was a result of association with the phenotype and not a consequence of a general reduction in transmission of the W64A allele. All DNA samples were extracted from seedling tissue, so only those mel1 individuals that produced a root or shoot were used.


Isolation and inheritance of mel1: The mel1 mutation arose spontaneously in a standard W22 inbred stock. It was originally isolated as a single plant that had 40% (121/303) defective kernels in a cross between two separately maintained W22 lines. A few of these defective kernels were viable, allowing recovery of the mutation and confirmation of the dependence of the phenotype on a maternal effect. These mel1-affected seeds are generally smaller than wild type; their endosperms are frequently etched, particularly at the crown; and the embryos of most of these seeds are abnormally shaped or absent (Figure 1). Embryo defects may be more common than endosperm defects because many mutant individuals have abnormally shaped embryos but endosperms that are normal in appearance.

Pollination of mel1 heterozygotes with wild-type W22 males produces defective seeds (defective embryo or endosperm or both), the percentage of which varies from 9 to 52% (Table 1). Three modes of action can explain the inheritance of mel1: a dominant defective kernel mutation, a dominant sporophytic maternal effect with variable penetrance, and a gametophytic maternal effect with variable penetrance or transmission. The different outcomes of these three possibilities are described in Table 2.

To distinguish these possibilities reciprocal crosses between mel1 heterozygotes and Mel1+ W22 homozygotes were made. Defective kernels were produced only when the mel1 heterozygotes were used as the female parent and not as the male parent, suggesting either a maternal effect or a dominant mutation with zero transmission through the pollen. To confirm that paternal inheritance of mel1 had no effect on kernel development, seeds produced from the pollen outcross of mel1 heterozygotes were tested for the inheritance of mel1 by growing them to maturity and crossing them as females. mel1 was transmitted through the pollen but at a much reduced frequency relative to wild type (7–23% of the progeny carry mel1 instead of 50%). To test pollen effects more stringently some of the rare abnormal kernels in these crosses were tested for the presence of mel1. These defective kernels were no more frequent than in crosses between homozygous wild-type plants and were not found to consistently carry mel1, demonstrating that mel1 was not responsible for their abnormal development. Therefore, paternal inheritance of mel1 has no effect on kernel development. However, the mel1 mutation is detrimental to pollen development or function.

View this table:

Frequency of defective seeds caused by mel1

View this table:

Models for action of mel1

To distinguish the stage of action of the mel1 maternal effect, the defective kernels were tested for inheritance of the mel1 mutation. If mel1 were a dominant mutation acting in the maternal sporophyte then it would not matter whether the mutant or wild-type allele were present in the embryo sac allowing half of the mutant progeny to carry mel1 and half to be homozygous wild type. The vast majority of individuals tested inherited mel1 (117/126), demonstrating that the genotype of the maternal gametophyte, rather than the genotype of the maternal sporophyte, is critical. The rare defective individuals that do not carry mel1 could be the result of abnormal development of wild-type seeds, as is seen in homozygous wild-type W22 individuals or by reversion of the mel1 mutation to wild type in these individuals.

Within the W22 inbred background, the percentage of defective kernels in crosses of mel1 heterozygotes by wild type is frequently <50%. This could be explained either by incomplete penetrance of mel1 or by failure of some mel1 embryo sacs to be fertilized to produce seed. The normal kernels from crosses of mel1 heterozygous females by wild-type pollen were grown to maturity and pollinated to determine how many, if any, carried mel1. The percentage of defective kernels and the percentage of normal kernels carrying mel1 were inversely correlated (Table 3). The two classes together comprise ~50% of the total in all crosses regardless of the percentage of defective kernels (i.e., half of the offspring were mel1 heterozygotes and half were homozygous wild type). Therefore, the reduced percentage of defective kernels in some crosses reflects incomplete and variable penetrance of the mel1 phenotype rather than embryo sac inviability or other prefertilization effects.

To test whether differences in mel1 penetrance are heritable, the offspring of mel1 families with high penetrance (≥30% defective kernels) were compared to those of families with low penetrance (<30%). Both groups showed a full range of defective kernel frequencies, and, in fact, the frequency of low- and high-penetrance expression of mel1 among the two classes is very similar (data not shown). These results suggest that the difference in the level of expression of the mel1 phenotype does not reflect heritable changes in the state of the mel1 mutation itself but may reflect environmental differences during their development.

Self-crosses of mel1 heterozygotes in a W22 inbred line produce the same frequency of defective kernels as mutant heterozygotes crossed as females by wild type. To determine if there is an effect of the mel1 mutation on the sporophyte, the progeny of these self-crosses were tested for the presence of mel1 homozygotes. Defective kernels were selected to ensure female transmission of mel1, germinated, grown to maturity, and pollinated to test for the presence and percentage of defective kernels; most of these were also tested for the presence of mel1 in the next generation. If there were no effect of mel1 on the sporophyte, then the frequency of homozygotes would equal the male transmission rate of mel1 (from 7 to 23%). None of the 55 individuals tested were homozygous for mel1, suggesting that mel1 homozygotes fail to reach maturity and set seed (P = 0.018 for an expected frequency of 7%; P = 5.7 × 10−7 for an expected frequency of 23%). One possible explanation is that mel1 homozygotes in a W22 background fail to germinate—a common phenotype of mel1 seeds in both backcrosses and self-crosses.

Chromosomal location of mel1: To map mel1, mutant plants were crossed with W22 stocks carrying visible genetic markers. The only one among those tested that showed linkage to mel1 was the b1 gene on chromosome 2. The recombination frequency between mel1 and B1 (which confers plant color) is 34% (129/370; χ2 = 33.9, P < 0.0001 such that the two genes assort independently). This linkage was confirmed with a second allele of the b1 gene, B1-Peru, which confers seed color rather than plant color. In these crosses the recombination frequency between mel1 and B1-Peru was 28% (88/310). The B1 W22 line used for mapping also differed from the mel1 W22 stock at the linked SSR marker bnlg1537. This marker was tested on a subset of this population to verify linkage and to determine on which side of the b1 gene mel1 is located. The distance from mel1 to bnlg1537 is 22% (16/72) and from bnlg1537 to b1 is 11% (8/72). These data place mel1 on the short arm of chromosome 2 distal to bnlg1537.

View this table:

Penetrance of mel1

Testcrosses indicate that there is no linkage between mel1 and pl1 (or p1, bz2, a1, c2, a2, pr1, y1, c1, or r1). Testcrosses between mel1/+; R1-sc/r1-g; W22 females and r1-g W22 males indicate no linkage between r1 and mel1. However, self-crosses of mel1/+; R1-sc/r1-g W22 individuals show an association between the R1-sc allele and expression of the mel1 phenotype (Table 4). The defective kernels from these crosses are less likely to be homozygous for the r1-g allele than their normal siblings. These data suggest that an enhancer of the mel1 phenotype is linked to R1-sc and was crossed into the W22 inbred stock along with the R1-sc allele.

Effect of mel1 on the gametophytes: Embryo sacs and mature pollen grains of mutant heterozygotes were compared to those of homozygous wild type to determine whether mel1 causes morphological defects in the gametophytes. Embryo sacs were examined with Nomarski and confocal laser scanning microscopy and displayed no gross abnormalities.

Male transmission of b1 and bnlg1537 from mel1 heterozygotes was tested to determine whether the reduced frequency of mel1 recovery reflects reduced transmission of the mel1-carrying chromosome or alterations of the mel1 mutant allele itself (e.g., through imprinting). In crosses of mel1 bnlg1537w22 b1-bar/+ bnlg1537B B1; pl1/Pl1-Rh males onto standard W22 females, b1-bar was transmitted at a rate of only 42% (181/432; χ2 = 11.3, P < 0.001) and bnlg1537w22 was transmitted at a rate of only 27% (41/150; χ2 = 30.8, P < 0.0001). In contrast, pl1 was transmitted at a rate of 46% (198/432; χ2 = 3.0, P > 0.05). The transmission of b1-bar was also followed in crosses of mel1 b1-bar/+ B1-Peru males onto standard W22 females. In these crosses the transmission of the b1-bar allele was variable with a range of 27% (66/242) to 39% (146/371), likely reflecting variable transmission of mel1 as described above. The reduction in transmission of the mel1 chromosome region indicates it is detrimental to some aspect of pollen viability or function.

Mature pollen grains from mel1 heterozygotes were examined for mutant phenotypes. Pollen was stained for starch, and the frequency of pollen grains that lacked starch, were incompletely filled with starch, or were smaller than normal, was measured. Of the pollen from mel1 heterozygotes, 4–22% was abnormal, a frequency similar to that of wild-type W22. No other abnormal phenotypes were observed.

Differential pollen tube growth could also cause mel1 pollen to be selected against in favor of wild-type pollen. Pollen from mel1 heterozygotes and homozygous wild-type plants were germinated in vitro, and tube length was measured. At 35 min after placing pollen on germination media, there were differences between mutant and wild-type individuals. While the tube length for the wild-type individuals had a normal distribution with a sharp peak in the 250–300 μm range, the length distribution for mutant heterozygotes had a broader range, suggesting a bimodal distribution (Figure 2). If the wild-type fraction is subtracted from the curve for mel1 heterozygotes to reveal the mutant class, the mel1 pollen tubes appear to have a peak in the 150–200 μm range. Pollen tube lengths from homozygous wild-type plants became more variable at later time points, making it difficult to determine whether mel1 continues to affect pollen tube growth after germination (data not shown).

View this table:

Association of the mel1 phenotype with r1

Figure 2.

Effects of mel1 on pollen tube growth. Pollen tube length at 35 min. Pollen was germinated in vitro. Pollen from mel1 heterozygotes has a broader distribution with a larger percentage of shorter tubes than wild type. Tube length for mel1 pollen grains was derived by assuming that half of the pollen tubes from heterozygotes were wild type. The wild-type half of the distribution was removed by subtracting half of the values for the wild-type curve from the mel1 heterozygote curve and then multiplying the difference by 2 to standardize the total back to 100%. Open triangles, wild-type length frequencies (W22). Half-solid squares, mel1 heterozygote length frequencies (W22). Solid circles, derived mel1 length frequencies (W22).

Effect of mel1 on seed development: mel1 progeny have a variety of defects. Endosperms are generally smaller than wild type and are etched, particularly at the crown or adgerminal to the silk scar (Figure 1A). Most of these seeds also have abnormally shaped embryos (Figure 1D). Embryos can be shorter along the apical-basal axis or not visible at all in mature kernels. Most of these seeds fail to germinate, and those that do usually produce abnormal seedlings (Table 5 and Figure 3). The shoot is more frequently affected than the root. Seedlings can have a root and no shoot and that root can be normal or fasciated. Some seedlings produce abnormal shoots that produce only the coleoptile or only the coleoptile and the first leaf. Some seedlings have twin shoots, each of which can be normal or abnormal. mel1 progeny that survive beyond the seedling stage have normal subsequent development.

View this table:

Phenotype of seedlings from mel1 affected kernels

Embryo development of mel1 progeny deviates from wild type as early as 8 days after pollination (DAP; Figure 4). Early in development mel1 seems primarily to retard embryogenesis. At 8 DAP the mel1 embryos were smaller than wild type and extremely difficult to remove from the developing kernel undamaged but were shaped normally (Figure 4B). At 11 DAP mel1 progeny are very similar to wild-type transition stage embryos at 8 DAP in both size and morphology (Figure 4D). Embryos at this stage are club shaped, and the apical half of the embryo is just beginning to enlarge to form the scutellum (cotyledon). Later in development mel1 progeny become more variable. Some continue to develop like wild type except more slowly, while others are shaped abnormally in addition to being reduced in size. Frequently, the cotyledon is abnormally shaped, most commonly with reduced apical-basal expansion. Twinning is also common and may occur by two different mechanisms. Some individuals have twin shoots that develop adjacent to one another on the face of the scutellum (Figure 4P). In other individuals a second scutellum and embryo axis appear to have developed from the base of the embryo (reminiscent of the suspensor mutants of Arabidopsis; Schwartzet al. 1994) (Figure 4, K and O).

Figure 3.

Seedling phenotypes of mel1 progeny. (A) Normal seedling. (B) Seedling with no shoot but a normal root. (C) Seedling with no shoot and a fasciated root. (D) Seedling with no primary root and an abnormal shoot—only a coleoptile was produced. (E) Seedling with twin, normal shoots. (F) Seedling with twin abnormal shoots—each shoot only produced a coleoptile.

Figure 4.

Effect of mel1 on embryogenesis. (A and B) 8 days after pollination (dap). (C and D) 11 dap. (E–G) 13 dap. (H–L) 17 dap. (M–Q) 26 dap. (A, C, E, H, and M) Wild type. (B, D, F, G, I–L, and N–Q) mel1. (A, D, and F) Transition-stage embryo. (B) Early transition-stage embryo. (C) First-leaf-stage embryo. (G) Coleoptilar-stage embryo. (I and N) Embryos with large but abnormally shaped scutella and no shoot axis. (J and K) Embryos with a small scutellum and a second axis of outgrowth (arrow). (O) Embryo with two scutella, each with an embryonic axis. (P) Embryo with a single embryonic axis but two shoots. (L and Q) Fairly normal mutant embryos with slightly smaller shoot axes than wild type. Bar, 1 mm.

In addition to causing abnormal morphology and retarded development mel1 evidently is lethal to some embryos. In early stages embryos are present in every seed examined. However, at later stages many of the seeds lack an identifiable embryo, suggesting that they have degenerated.

Because the phenotype of mel1 progeny is similar to that of mea in Arabidopsis, unpollinated ears of mel1 heterozygotes and wild type were examined for fertilization-independent seed development, which occurs in mea mutants (Kiyosueet al. 1999; Luoet al. 1999; Grossniklaus and Vielle-Calzada 1998). In both mel1 heterozygotes and homozygous wild type almost all unpollinated ears showed no sign of seed development. A small number of individuals in both classes produced fewer than five seeds. Consequently, in contrast to mea, mel1 has no affect on suppression of seed development prior to pollination.

W22 carries an enhancer required for mel1 expression: During attempts to map mel1 genetically, mel1 heterozygotes in a W22 background were crossed as females by pollen from six other inbred lines. Surprisingly, the mel1 phenotype was completely suppressed in all of these crosses (see Table 6 for W64A; data for A158, B73, M14, Mo17, and W23 not shown). Apparently, the effect of mel1 depends on a factor(s) present in W22 and absent from other inbred lines. Because the mel1 phenotype is suppressed in F1 kernels, this factor(s) does not act maternally. It could act in either the zygote or the pollen grain.

View this table:

Suppression of mel1 by W64A

To study their inheritance, two sets of crosses were performed to generate families segregating mel1 and any potential modifiers (Figure 5A). In cross I, mutant heterozygotes in a W22 inbred line were crossed as females by W22/W64A hybrids. The frequency of abnormal kernels in these crosses had a maximum of 21% instead of 50%, consistent with the segregation of one recessive modifier in the hybrid male (Figure 5B). In cross II mel1/+ W22/W64A hybrid females were pollinated by wild-type W22 males. The frequency of defective kernels in these crosses had a maximum of 45%. The results of cross I suggest that two major factors are segregating 1:1 in these crosses, mel1 and a single modifier, snm1. The higher percentage of defective seeds in cross II than cross I suggests either that snm1 is tightly linked to mel1 or that paternal inheritance of snm1, rather than homozygosity, is required for the mutant phenotype.

These crosses provided families segregating mel1 and snm1 to map both factors. In cross I, molecular markers linked to snm1 in the male cosegregate with the mutant phenotype, and in cross II markers linked to snm1 and mel1 in the female cosegregate with the mutant phenotype. The predicted genotypes of the defective kernels in the two types of crosses used are shown in Figure 5. Markers were first tested to cover the whole genome for linkage with the mutant phenotype in cross II, which should detect linkage to mel1 and any zygotic modifiers required for the mutant phenotype. The majority of the markers tested segregated independently from the defective kernel phenotype (Figure 6). Only markers on the short arm of chromosome 2 showed linkage to the mutant phenotype. Markers on chromosome 6 showed preferential transmission of W22 alleles to both mutant and wild-type progeny independently of phenotype. All other markers segregated 1:1 in mutant individuals.

Markers on chromosome 2 showed close linkage to the mel1 phenotype in cross II (Figure 6). These data put mel1 11% (8/72) from bnlg1017 and 10% (7/72) from bnlg125. There were no double crossover or noncrossover chromosomes carrying W64A alleles for bnlg-1017 and bnlg125 in this population. Based on linkage to b1 and bnlg1537 within the W22 inbred mapping population described above, this position is in good agreement with the map position for mel1.

Figure 5.

(A) Crosses used to map mel1 and snm. Note that for the defective kernels the mel1 region may be heterozygous W22/W64A in cross I and but not cross II, while the snm region is homozygous W22 for both crosses if the suppression is zygotic, but may be heterozygous in cross II if the suppression is paternal. These differences are also reflected in the alleles of tightly linked markers. (B) Effect of snm on the penetrance of mel1. Cross I families with a snm/+ heterozygous male parent have a maximum frequency of defective kernels that is half that of the W22 control. However, cross II families with females heterozygous for mel1 and snm1 give the same distribution as females heterozygous for mel1 but homozygous for snm1.

Figure 6.

Mapping of mel1 and snm1. All individuals tested for these markers were defective kernels (mel1 kernels) from cross II. An association with the phenotype is indicated by rare heterozygosity for a particular marker. SSR markers are shown to the left of the chromosome. The frequency of heterozygotes is shown to the right of each marker. Recombination frequency between markers for the test population is shown in boldface type between markers. Asterisks indicate approximate positions of core bin markers. The double dagger on chromosome 6 indicates the region that is preferentially homozygous W22 in both mutant and wild-type kernels. Total chromosome length, position of SSR markers used, and position of core markers are taken from the ACEMAZ database (

View this table:

Segregation of snm1 and snm2 in different crosses involving mel1

The results from cross I differed. Two classes of families were observed (Table 7). In the first, close linkage was observed between the defective kernel phenotype and bnlg1017 and bnlg125. However, in the second class there was weak or no association between this region and the defective kernel phenotype. In class 1 families, 13% of the mel1 individuals were heterozygous for bnlg1017 (11/87) and 14% were heterozygous for bnlg125 (12/85), demonstrating linkage of mel1 to the recessive enhancer snm1. However, snm1 is not absolutely required for the mutant phenotype. In class 2 families of cross I, only a weak association was seen between these markers and the mel1 phenotype. Forty-four percent (32/73) of the defective kernels were heterozygous for bnlg1017, and 35% (23/66) were heterozygous for bnlg125. These data taken together suggest that the class 2 families carry an additional recessive modifier(s), allowing expression of the mel1 phenotype independently of snm1.

Because of the association of the mel1 phenotype with r1 in self-crosses in a W22 inbred line and the fact that W22 lines with different r1 alleles had been crossed to mel1, we tested chromosome 10 in the r1 region, using bnlg1028, for association with the mutant phenotype in cross I. Class 1 families, which showed tight linkage of the mutant phenotype to chromosome 2, showed no association of the phenotype with chromosome 10 markers (42/87 individuals were heterozygous for bnlg1028). In contrast, class 2 families showed linkage to chromosome 10 in the region of r1 (12/64 mel1 individuals were heterozygous for bnlg1028), demonstrating the presence of a second modifier locus, snm2. It should be stressed that the class 1 and class 2 individuals are the mel1 heterozygotes used as females. The polymorphism for the SSR markers lies in the male parents, which are the same for both class 1 and class 2 crosses. This suggests that the male parent used is heterozygous for snm2 (over the wild-type W64A allele) and that while class 2 families carry snm2, class 1 families do not. The mel1 individuals used for cross II were related by descent to the mel1 individuals in class 1 rather than class 2, suggesting that the cross II families lacked snm2 and so showed linkage only to chromosome 2 markers.

Other aspects of the mel1 phenotype are also suppressed in the W22/W64A hybrid background. Male transmission was measured and found to be normal by two criteria. The linked marker bnlg125 was examined in the progeny of mel1/+ W22/W64A individuals crossed as males onto W22 inbred females. The W22 allele of bnlg125 linked to mel1 was transmitted to ~50% of the progeny (32/59). The rate of mel1 transmission in this same cross was measured using the percentage of individuals producing defective kernels in the next generation as an indicator of the presence of mel1. These plants were also present in ~50% of the progeny (82/156).

The presumed lethality of mel1 homozygotes is also suppressed in this hybrid background. The frequency of mel1 homozygotes in self-crosses of mel1/+ W22/W64A plants was measured using the flanking markers bnlg1017 and bnlg125, which are 21 map units apart. Twelve of 81 individuals tested were homozygous for the W22 alleles of both markers linked to mel1, so the frequency of mel1 homozygotes is at least 15%. If the recombination distance between these markers and mel1 is taken into account, then the frequency of mel1 homozygotes in these self-crosses likely approaches 25%. These homozygotes were indistinguishable from their wild-type siblings. They are distinguished only by the presence of a few miniature kernels segregating in the self-progeny of the mel1 homozygotes (data not shown).


Angiosperm reproductive biology allows for maternal contribution to the development of the next generation at several points in the life cycle. The female gametophyte develops to maturity within sporophyte tissue, and the angiosperm seed develops within the maternal sporophyte after double fertilization of the gametophyte. Recent genetic evidence supports a role for the maternal sporophyte and maternal gametophyte in development of the angiosperm seed. Crosses in maize and Arabidopsis between diploid and polyploid individuals cause seeds to develop differently, depending upon whether the polyploid is used as the male or female parent, suggesting that the maternal and paternal genomes are not equivalent (Lin 1984; Scottet al. 1998). This difference could reflect gene expression before or after fertilization, but the ability to phenocopy this effect using lines with reduced DNA methylation suggests an epigenetic component such as imprinting (Adamset al. 2000).

Here we define gametophytic maternal effects broadly, on the basis of inheritance rather than mechanism. The timing of gene expression is different for different mechanisms. First, maternal effects can be caused by mutations in genes expressed during the gametophyte phase of the life cycle. These effects could be caused by a requirement for the gene product for normal embryo sac development or by a requirement for perdurance of the gene product after fertilization. Second, maternal effects can be caused by mutations in genes whose paternally contributed alleles are silenced. Third, maternal effects can be caused by mutations in genes required in two doses in the endosperm. In the last two classes gene expression occurs after fertilization. Unlike the first two classes, maternal inheritance of dose-sensitive genes is not strictly required, because the presence of an additional paternal copy of the wild-type allele promotes normal development.

Other recent results suggest that very early in seed development the entire paternal genome is silenced (Vielle-Calzadaet al. 2000). However, very few mutations have noticeable maternal effects. Either very few genes are required during this period of silencing or embryos can recover from early defects to produce normal seeds. The latter seems to be the case for the EMB30/GNOME gene (Vielle-Calzadaet al. 2000). Imprinted genes with maternal effects must be distinct in some way from the majority. Either embryos cannot recover from an early absence of the gene product (which would be expressed from the maternal allele or present in the cytoplasm of the egg or central cell in wild type) or silencing of the paternal allele is maintained longer for these genes than for others.

Nature of the mel1 maternal effect: The mel1 mutant resembles the mea mutant of Arabidopsis. Maternal inheritance of mel1 is associated with abnormal endosperm and embryo development, while paternal inheritance has no effect on seed development. Although a wild-type paternal allele of mel1 is not required in the embryo, the mel1 mutation is detrimental to pollen function. This is in contrast to mea, fie, and fis2, which have no effect on the male gametophyte.

One explanation of the mel1 mutation is a gametophytic maternal effect. However, an alternative explanation is that mel1 is a dominant mutant that acts during seed development. A separate mechanism is necessary to explain the lack of a paternal effect. In this scenario, paternal transmission would silence the dominant allele, which would only rarely be reactivated upon subsequent female transmission. In other words, the mutant allele is transmitted at a rate equal to the wild-type allele but is modified by this transmission. Silencing in this case would be a special property of the mel1 mutant allele and not a normal condition of the wild-type mel1 gene. This would explain both the lack of a paternal effect and the deficit of mutant heterozygotes after male transmission. However, transmission of flanking markers demonstrates that the mel1 mutation is detrimental to the male gametophyte, and reduced male transmission of the mel1 chromosomal region is sufficient to explain the infrequency of mel1 recovery in these crosses.

In many respects inheritance of mel1 is similar to that of the Defective aleurone pigmentation (Dap) mutants of maize (Gavazziet al. 1997). These mutations cause abnormal endosperm development when maternally inherited and show reduced male transmission. They cause abnormal development of the outer layers of the endosperm but have no effect on embryo development, in contrast to mel1. The Dap mutants have been interpreted as dominant mutants that have a maternal effect either because they are silenced upon male transmission or because two mutant doses are required for the mutant phenotype in the endosperm. The reduced male transmission of some of these is attributed to an independent but linked deletion. An alternative explanation is that like mel1 the Dap mutations disrupt genes that are required in the male gametophyte and in the maternal gametophyte. The mel1 mutation, at least, is unlikely to be associated with a large deletion or other rearrangement, because the reduced male transmission is suppressed in other genetic backgrounds. It is unlikely that reduced transmission of a large deletion (i.e., involving many essential genes) would be dependent on genetic background. The simpler explanation is that a single gene is responsible for both effects.

Introduction of additional doses of the paternal allele to determine whether mel1 has its effects through a dose-dependent mechanism has not yet been done. However, the embryo, in which the maternal-to-paternal gene ratio is one to one, is more commonly affected than the endosperm. We cannot rule out that the embryo defects are indirect even in those individuals with no visible endosperm defect. If abnormal embryo development is a secondary effect, then all maternal effects of the mel1 mutation could be an effect of dosage in the endosperm. Alternatively, if mel1 is not dose dependent in the endosperm, then it acts through either imprinting or perdurance of maternal gene product after fertilization, like mea or prl, respectively. Until the mel1 gene is cloned, it is very difficult to distinguish these last two possibilities from one another. Gene dosage effects in maize endosperm development have also been attributed to a potential interaction of prefertilization and postfertilization dose-sensitive gene products (Birchler 1993). mel1 may act through a similar combination of maternal gametophytic mechanisms rather than only through one.

Function and relationship of mel1 and snm: All of the evidence suggests that the mel1 mutation acts maternally and not as a dominant zygotic mutant. However, because the embryo sac is haploid, mel1 could be a recessive maternal effect mutant or a dominant maternal effect mutant. The only evidence that mel1 is recessive is the lethal phenotype in the sporophyte. If mel1 were dominant, then heterozygous sporophytes would be expected to show a mutant phenotype regardless of the parental origin of mel1. If heterozygotes were abnormal or occasionally inviable, then some of the heterozygotes produced by mel1 male transmission would show a mutant phenotype, but this has not been observed. From the time of germination until flowering, heterozygotes are indistinguishable from wild type. If mel1 is recessive, then the wild-type function of mel1 is to promote normal growth of the endosperm and the embryo. Mutant embryos grow more slowly, and the axis of growth is frequently abnormal. Mutant endosperms also grow more slowly than wild type. This growth-promoting function can be provided either maternally by mel1 or zygotically by snm. Only when both functions are absent is growth abnormal.

Although mel1 and snm1 or snm2 contribute redundant functions genetically, they need not share sequence homology. As unrelated genes they could be mutations in different biochemical pathways that produce the same necessary factor. Alternatively, they could be in homologous genes. For example, snm1 and mel1 could be the result of a tandem duplication. Interestingly, snm2 is located on the long arm of chromosome 10, the region of the maize genome that is the duplicated segment of the short arm of chromosome 2 (Helentjariset al. 1988; Gaut and Doebley 1997). Therefore, snm2 may be the duplicate of mel1 or snm1 in this region.

While mel1 is redundant with snm1 or snm2, snm1 and snm2 do not show redundancy with each other. Only one of the two needs to be mutant to condition the mel1 phenotype. The mode of inheritance is also potentially different in snm1 and snm2. snm2 likely acts in the zygote, because inheritance from both parents is required for a mutant phenotype. Inheritance through the male is demonstrated by flanking markers, and inheritance through the female is demonstrated by the difference between cross I, class 1 females that lack snm2 and cross I, class 2 females that carry it. For snm1, however, there is currently no evidence that it needs to be inherited through both male and female parents to cause a mutant phenotype. Only a requirement in the pollen grain is currently supported, so its time of action could be in the zygote or the pollen grain. The tight linkage of snm1 and mel1 make a test for a female effect of snm1 difficult.

If mel1 and snm are homologous genes, the differences in their inheritance could be explained by differences in expression. For example, mel1 may have acquired sequences that cause its paternal allele to be silenced, while both alleles of snm1 and snm2 are expressed. It is also possible that mel1 and snm1 are alleles of the same gene. In this scenario the endogenous allele of mel1/snm1 in W22 would have to be different from that of all other inbreds tested. The W22 allele could either be silenced paternally while other inbred “wild-type” alleles are not (possibly by insertion of sequences in or near mel1 in the W22 line that causes its silencing), or the W22 allele could be weakly expressed compared to other wild-type alleles, causing it to be required in at least two doses. Either of these mechanisms would cause a subsequent null allele to have a gametophytic maternal effect in conjunction with the W22 allele but not with other alleles.

Although it is possible that mel1 and snm1 are alleles of the same gene, snm2 maps to a different chromosome and therefore represents a different gene. This demonstrates redundancy between a gene with a maternal effect (mel1) and a gene acting in the zygote (snm2). Since cross-generation suppression occurs, it remains a feasible explanation for the action of snm1 instead of allelism with mel1. Such a system of redundancy allows the allocation of resources to the seed and the promotion of embryonic growth to be under both maternal and zygotic (i.e., biparental) control.


We thank Kathy Barton for helpful comments on the manuscript. We also thank Beverly Oashgar and David Heller for their expert help in growing plants. Scanning electron microscopy was done at the University of Wisconsin Microscopy Resource. This work was supported by a Plant Postdoctoral Fellowship to M.M.S.E. This is paper number 3581 of the Laboratory of Genetics.


  • Communicating editor: J. A. Birchler

  • Received March 27, 2001.
  • Accepted June 18, 2001.


View Abstract