Embryo and Endosperm Development Is Disrupted in the Female Gametophytic capulet Mutants of Arabidopsis

Corresponding author: Martin Hülskamp, University of Cologne, Gyrhofstr. 15, D-50931 Cologne, Germany., martin.huelskamp{at}uni-koeln.de (E-mail)

Communicating editor: C. S. GASSER

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The female gametophyte of higher plants gives rise, by double fertilization, to the diploid embryo and triploid endosperm, which develop in concert to produce the mature seed. What roles gametophytic maternal factors play in this process is not clear. The female-gametophytic effects on embryo and endosperm development in the Arabidopsis mea, fis, and fie mutants appear to be due to gametic imprinting that can be suppressed by METHYL TRANSFERASE1 antisense (MET1 a/s) transgene expression or by mutation of the DECREASE IN DNA METHYLATION1 (DDM1) gene. Here we describe two novel gametophytic maternal-effect mutants, capulet1 (cap1) and capulet2 (cap2). In the cap1 mutant, both embryo and endosperm development are arrested at early stages. In the cap2 mutant, endosperm development is blocked at very early stages, whereas embryos can develop to the early heart stage. The cap mutant phenotypes were not rescued by wild-type pollen nor by pollen from tetraploid plants. Furthermore, removal of silencing barriers from the paternal genome by MET1 a/s transgene expression or by the ddm1 mutation also failed to restore seed development in the cap mutants. Neither cap1 nor cap2 displayed autonomous seed development, in contrast to mea, fis, and fie mutants. In addition, cap2 was epistatic to fis1 in both autonomous endosperm and sexual development. Finally, both cap1 and cap2 mutant endosperms, like wild-type endosperms, expressed the paternally inactive endosperm-specific FIS2 promoter GUS fusion transgene only when the transgene was introduced via the embryo sac, indicating that imprinting was not affected. Our results suggest that the CAP genes represent novel maternal functions supplied by the female gametophyte that are required for embryo and endosperm development.

MATERNAL effects are fairly common in genetically tractable animals, such as Drosophila melanogaster and Caenorhabditis elegans, and maternal-effect mutations have allowed the identification of genes whose products play important roles in setting the stage for embryo development (JOHNSTON and NUSSLEIN-VOLHARD 1992 ). By contrast, the evidence for maternal effects in higher plants is rather scant, due to the plant-specific alternation of generations, with a haploid gametophyte generation occurring between two successive diploid sporophytic generations. In Arabidopsis, one of the female meiotic products, the megaspore, undergoes three rounds of nuclear divisions followed by cellularization, which results in a seven-celled female gametophyte, or embryo sac (MANSFIELD et al. 1991 ; WEBB and GUNNING 1994 ; SCHNEITZ et al. 1995 ; CHRISTENSEN et al. 1997 ). At the micropylar end where the pollen tube delivers the two sperm cells, the egg cell is flanked by two synergids that assist in fertilization. A large central cell that is diploid due to the fusion of two haploid nuclei separates the egg cell from the three antipodal cells of unknown function that occupy the chalazal end and degenerate before fertilization. Double fertilization of egg cell and central cell initiates development of the diploid embryo and the triploid endosperm, respectively. These two organisms develop in parallel in a coordinated interplay between sporophytic and gametophytic tissues to produce the mature seed.

Embryo and endosperm development are fundamentally different. In Arabidopsis, early embryo development is characterized by an invariant pattern of cell divisions and differentiation (GOLDBERG et al. 1994 ; LAUX and JURGENS 1997 ; HARADA 1999 ; JURGENS 2001 ). The endosperm initially undergoes synchronous nuclear divisions, but by the time of cellularization, different domains have been established by morphological criteria, mitotic activity, and reporter gene expression patterns (WEBB and GUNNING 1991 ; BERGER 1999 ; BROWN et al. 1999 ; BOISNARD-LORIG et al. 2001 ). In Arabidopsis and most other angiosperms, the endosperm is largely consumed during embryogenesis, suggesting a nutritive function (LOPES and LARKINS 1993 ). However, the endosperm has also been suggested to have roles in the regulation of embryo size and fruit development and may function as a check point for the initiation of sexual reproduction (LOPES and LARKINS 1993 ; HONG et al. 1996 ; OHAD et al. 1996 ).

To what extent embryo and endosperm development require maternal cues is a matter of speculation since both somatic embryogenesis and endosperm development can occur in vitro in the absence of maternal tissue (ZIMMERMAN 1993 ; KRANZ et al. 1998 ). However, there is genetic evidence for both sporophytic and gametophytic maternal effects. The Arabidopsis SHORT INTEGUMENT1 (SIN1) gene is required maternally in the ovule for proper embryo development, regardless of embryo or endosperm genotype (A. RAY et al. 1996 ; S. RAY et al. 1996 ). In barley, the maternal-effect mutations shrunken endosperm affect the endosperm in a similar manner (FELKER et al. 1985 ). Sporophytic maternal effects on the endosperm have also been shown by ovule-specific downregulation of the petunia MADS box genes FLORAL BINDING PROTEIN 7 (FBP7) and FBP11 (COLOMBO et al. 1997 ).

The inaccessibility of the female gametophyte has been a hindrance for large-scale genetic screens and, until recently, only a few female gametophytic mutations have been described (REDEI 1965 ; KERMICLE 1971 ; PATTERSON 1994 ; VIZIR et al. 1994 ; VOLLBRECHT and HAKE 1995 ). However, more recent screens have resulted in a growing mutant collection that yields new insights into female-gametophyte development and function (FELDMANN et al. 1997 ; CHRISTENSEN et al. 1998 ; HOWDEN et al. 1998 ; GRINI et al. 1999 ; VIELLE-CALZADA et al. 2000 ; for reviews, see DREWS et al. 1998 ; GROSSNIKLAUS and SCHNEITZ 1998 ). Some genes have been shown to be expressed in the female gametophyte before fertilization (SPRINGER et al. 1995 ; KRANZ and DRESSELHAUS 1996 ; NADEAU et al. 1996 ; PERRY et al. 1996 ; VIELLE-CALZADA et al. 2000 ; CORDTS et al. 2001 ). The PROLIFERA (PRL) gene encodes an MCM family protein that regulates replication during G₁ phase of the cell cycle. ß-Glucuronidase (GUS) activity from a PRL::GUS transgene accumulates in the central cell nucleus before fertilization, and a fraction of mutant prl embryo sacs arrest at G₁ checkpoints during syncytial endosperm development (SPRINGER et al. 2000 ).

Female-gametophytic maternal effects on embryo and endosperm development have been demonstrated by mutations in the Arabidopsis MEDEA (MEA), FERTILIZATION INDEPENDENT SEED (FIS), and FERTILIZATION INDEPENDENT ENDOSPERM (FIE) genes (OHAD et al. 1996 ; CHAUDHURY et al. 1997 ; GROSSNIKLAUS et al. 1998 ). The mea, fis, and fie mutants show autonomous seed development, suggesting that the affected genes normally repress embryo and endosperm development in the unfertilized ovule. In addition, these genes are subject to imprinting such that only their maternal alleles are expressed whereas their paternal alleles are inactive during early seed development. The developmental consequence of imprinting is suppressed by activating additional factors in the paternal genome (KINOSHITA et al. 1999 ; KIYOSUE et al. 1999 ; LUO et al. 1999 , LUO et al. 2000 ; VIELLE-CALZADA et al. 1999 ; VINKENOOG et al. 2000 ; YADEGARI et al. 2000 ). Although the MEA, FIS, and FIE gene products are present in the egg and central cell before fertilization, the time of their requirement for embryo and endosperm development is not known.

Two separate approaches have been employed to isolate mutants affected in gametophyte development or function. Direct screens are based on seed-abortion phenotypes or transgene reporter gene expression in the female gametophyte (SPRINGER et al. 1995 ; CHRISTENSEN et al. 1998 ; VIELLE-CALZADA et al. 2000 ; SORENSEN et al. 2001 ). By contrast, linkage-based screens detect gametophytic mutations by the altered segregation of linked morphological markers or transgene-borne resistance markers (FELDMANN et al. 1997 ; BONHOMME et al. 1998 ; HOWDEN et al. 1998 ; GRINI et al. 1999 ). The linkage-based strategy enables direct mapping of identified mutants as well as the ability to distinguish between gametophytic and dominant maternal sporophytic effects.

To analyze the role of the female gametophyte in fertilization, we performed a linkage-based screen for mutants that were functionally impaired but did not display an easily scorable embryo sac phenotype. Our procedure was based on the assumption that recessive marker mutations closely linked in trans to a newly induced female gametophytic mutation would give up to 50% marker progeny rather than the Mendelian 25%. We used the multiply marked mm1 line that carries five visible mutations at 20-cM intervals on chromosome 1 (GRINI et al. 1999 ). M2 families of mutagenized heterozygous mm1 seeds were examined for increased frequencies of mm1 markers. Among the lines segregating >40% marker frequencies we identified two mutants that had gametophytic maternal effects on embryo and endosperm development.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Plant strains and growth conditions:
Arabidopsis thaliana (L.) Heynh. var. Landsberg erecta was used as wild type (WT) unless indicated otherwise. The mm1 marker line is homozygous for angustifolia (an), distorted1 (dis1), eceriferum5 (cer5), apetala1 (ap1), and glabra2 (gl2; GRINI et al. 1999 ). The tetraploid U408 line was in the Landsberg erecta background (M. HÜLSKAMP, unpublished results). The FIS2::GUS and the MET1 a/s transgene constructs were in the C24 ecotype background. The ddm1-2 ecotype was Columbia and it had been backcrossed to Columbia for eight successive generations. Progeny of selfed ddm1-2/DDM1-2 plants were genotyped by an allele-specific PCR test using dCAPS primers DDM1f (5'-GAGATCTCTACCCTCCTGT-3') and ddm1-2dRsa (5'-TGAGCTACG–AGCCATGGGTTTGTGAAACGTA-3'), as described by YADEGARI et al. 2000 . Digestion of the PCR fragments with RsaI restriction endonuclease and separation on a 4% agarose gel yields an 130-bp band for the ddm1-2 allele.

All seeds were germinated on a mixture of soil and sand after 4 days of vernalization at 4° and grown under long-day conditions (18 hr light/6 hr dark) at 20°. fis1 homozygous seed and progeny of crosses involving maternal fis1 were germinated on MS medium (MURASHIGE and SKOOG 1962 ). FIS2:: GUS transgene plants were selected on MS medium containing 50 µg/ml kanamycin. Seedlings were transferred to soil after 1 week and kept at high humidity for one additional week.

Mutagenesis screen and genetic characterization of mutants:
Heterozygous seed from the mm1 marker line were mutagenized with ethyl methanesulfonate as described previously (MAYER et al. 1991 ; GRINI et al. 1999 ). M2 plants were screened for distorted segregation of the mm1 morphological markers (for details, see GRINI et al. 1999 ). Lines segregating >40% of two adjacent mm1 markers were rescreened and checked for aborted seed development. For the genetic analysis of mutants, split mm1 marker lines were used. These lines were homozygous for an dis1 or for ap1 gl2. cap1 and cap2 were backcrossed to Ler three times and reintroduced to the split mm1 marker lines. For the genetic analyses, reciprocal cross data from independent lines were pooled only when both outcrosses and selfing were done. Plants used for crosses or phenotypic analysis were rescreened for increased marker frequencies.

Histology:
For whole-mount preparations of fertilized or autonomous ovules, siliques were dissected with hypodermic needles and carpel walls were removed so that all ovules remained connected to the placenta. This dissection technique enabled the position of each ovule in the silique to be scored. Dissected siliques were fixed on ice in FAA [10:7:2:1 EtOH:distilled water:acetic acid:formaldehyde(37%)] for 30 min, hydrated in a graded EtOH series to 50 mM NaPO₄ buffer, pH 7.2, and mounted on microscope slides in a clearing solution of 8:2:1 chloral hydrate:water:glycerol (ClH). The specimens were allowed to clear for 1 hr at 4° before inspection. Embryo sac phenotypes were inspected in methyl benzoate-cleared whole-mount ovule preparations. Ovules were stained with Mayer's Hemalaun and processed as described by SCHNEITZ et al. 1995 . Scanning electron microscopy studies were performed as described previously (HULSKAMP et al. 1995A ). Aniline blue visualization of pollen tubes was performed as described (HULSKAMP et al. 1995B ). The GUS assay was performed after a modified protocol from SCHOOF et al. 2000 . Plant material was prefixed in ice-cold 90% acetone for 10 min, rinsed for 10 min in staining buffer (50 mM NaPO₄, pH 7.2; 2 mM potassium-ferrocyanide; 2 mM potassium-ferricyanide; 0.1% Triton X-100; 2 mM X-Gluc) with no substrate and incubated in staining buffer at 37° for 3–5 hr. Following a graded EtOH dehydration series to 50% EtOH, the material was postfixed in FAA on ice for 30 min and hydrated in an EtOH series to 50 mM NaPO₄ buffer and mounted on microscope slides in ClH. Confocal scanning laser microscopy (CLSM) and Feulgen staining were performed as described by BRASELTON et al. 1996 .

Microscopy and processing of images:
Light microscope preparations were examined using a Zeiss Axiophot microscope with differential interference contrast optics and epifluorescence. Photographs were taken on Kodak Ektachrome 64T or PROVIA 400 color films or with a Nikon Coolpix 990 digital camera. Scanning electron microscopy was performed with a HITACHI S 800 microscope. CSLM was performed with a Leica microscope equipped with UV light and Kr/Ar laser. Microscopic images were processed using Adobe Photoshop 6.0 and Adobe Illustrator 9.0 software.

Mapping with flanking markers:
Genetic distances between flanking markers and cap1 or cap2 were calculated using the formula p = 1 - 2M, where M is the frequency of the flanking markers. The recombination frequencies were normalized relative to the size of the interval (GRINI et al. 1999 ). To determine the penetrance of the mutant alleles, cap1/CAP1 and cap2/CAP2 were crossed with wild-type pollen donors and their F₁ progeny were checked for the cap ovule phenotype (see Fig 1A). More than 100 F₁ plants from each cross were inspected and none showed the cap phenotype, suggesting that penetrance was nearly complete.

View larger version (181K): In this window In a new window Download PPT slide   Figure 1. Embryo sac and seed development in cap mutants. (A) Two classes of developing ovules in a cap1/CAP1 silique at 3 DAP as shown by scanning electron micrograph. Bar, 100 µm. (B and C) Mature embryo sacs from CAP2 control (B) and cap2/CAP2 (C) siliques. Optical sections were obtained from whole-mount preparations of ovules. Asterisk, egg cell nucleus; arrowhead, central cell nucleus; v, vacuole. Bars, 20 µm.

Molecular mapping:
cap1 and cap2 mapping populations were made by outcrosses with the Niederzenz (Nd-0) and Columbia (Col-0) ecotypes, respectively. Heterozygous mutants were crossed as male partners to wild-type plants. F₁ plants phenotyped for cap1 or cap2 were paternally re-outcrossed with Nd-0 or Col-0. The resulting F₂ mapping populations were grown on soil and phenotypes were determined for all plants. In this crossing scheme, the mapping population consisted only of male meiotic events and the maternal lethality of the mutations did not affect the population. Genetic distance (p) was the same as recombination frequency (Rf, p = Rf). DNA was isolated by a modified cetyltrimethylammonium bromide miniprep protocol as described by STEWART and VIA 1993 . Basic molecular biology techniques were performed according to SAMBROOK and RUSSELL 2001 .

Mapping was performed with cleaved amplified polymorphic sequences (CAPS) or simple sequence length polymorphisms as described previously (HAUGE et al. 1993 ; KONIECZNY and AUSUBEL 1993 ; BELL and ECKER 1994 ) or by the Arabidopsis Information Resource (http://www.Arabidopsis.org). cap1 was mapped in a population of 108 chromosomes. On the telomere side, 7 recombinants were found for marker nga59 [on bacterial artificial chromosome (BAC) T25K19], 3 recombinants for marker T2,5 (on BAC T7I23; FOLKERS et al. 2002 ), 1 recombinant for marker O846a (on BAC F19P19), and no recombinants for marker m488 (on BAC T25N20; LUKOWITZ et al. 1996 ). On the centromere side, 17 recombinants were found for marker m59 (on BAC F20D23), 5 recombinants for marker G5957 (on BAC T27G7; LUKOWITZ et al. 1996 ) and no recombinants for m488. The O846a–G5957 interval spans a region of 1.5 Mb corresponding to 5–7 cM. The genetic distance between cap1 and the PCR markers was 6 cM [(5 + 1)/108]. The cap2 population consisted of 102 meiotic events. On the centromere side, 14 recombinants were found for marker nF5I14 (on BAC F5I14), 5 recombinants for marker nga111 (on BAC F28P22), 2 recombinants for marker ADH (on BAC T14N5), and no recombinants for dSNP142 (on BAC F18B13), a CAPS marker we made from the single nucleotide polymorphism (SNP) 142 (SNP142). This marker detected 2 independent recombinants on the telomere side. dSNP142 primers were dSNP142F 5'-CGGGGACATCTTGACGGCTT-3' and dSNP142R 5'-TGCTCCGATACTGAACTCGTGGC-3'. Digestion of the 933-bp PCR fragment with SspI restriction endonuclease and separation on a 2% agarose gel yields 524- and 409-bp bands for the Col-0 ecotype. In the Ler ecotype, the 409-bp band is cut into 238 and 171 bp. The ADH-SNP142 interval spanned 1.2 Mb corresponding to 3–5 cM, in good accordance with the genetic distance found between the PCR markers and cap2 (2 x 2/102 = 4 cM). The flanking marker GL2 is located on BAC F19K16, adjacent to F18B13, which supports the genetic mapping data for cap2 within the ap-gl2 interval.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Isolation and genetic characterization of the two novel female gametophytic mutants capulet1 and capulet2:
In a linkage-based screen for increased transmission of recessive morphological markers on chromosome 1, we isolated two novel mutants termed capulet after Shakespeare's Romeo and Juliet (see MATERIALS AND METHODS for screen details). As shown in Table 1, selfing of capulet1 (cap1) and capulet2 (cap2) gave increased frequencies of the flanking marker pairs an dis1 and ap1 gl2, respectively (Table 1), suggesting that their wild-type alleles linked to the cap mutations were not transmitted through the female or through the male gametophyte. Reciprocal backcrosses of cap/CAP heterozygous plants with mm1 marker plants revealed that the cap1 and cap2 mutant alleles were specifically blocked in their transmission through the female gametophyte (Table 1). The map positions of cap1 and cap2 were calculated from flanking marker segregation data (see MATERIALS AND METHODS). cap1 mapped within the an-dis1 interval 8 cM from an and 13 cM from dis1 whereas cap2 mapped near the bottom end of chromosome 1, 18 cM south of ap1 and 2 cM north of gl2. These results were confirmed by mapping against molecular markers. cap1 was placed between PCR marker O846a on BAC F19P19 and marker G5957 on BAC T27G7 (see MATERIALS AND METHODS for details) and was thus separated from the closely linked maternal-effect mutant medea by the marker O846a (GROSSNIKLAUS et al. 1998 ). cap2 was mapped within an interval defined by the PCR marker ADH on BAC T14N5 and a CAPS marker made from SNP142 on BAC F18B13 (see MATERIALS AND METHODS for details). In summary, the female-gametophyte defects of cap1 and cap2 appear to result from single-locus genetic lesions.

cap mutant embryo sacs are morphologically normal but do not support embryo and endosperm development:
cap/CAP heterozygous plants displayed a reduced seed set of 50% (Fig 1A). To examine whether the CAP genes are required for the development of the female gametophyte, we inspected whole-mount preparations of ovules from unpollinated mature siliques. In Arabidopsis, the mature embryo sac consists of an egg cell, a central cell, two synergid cells, and three degenerated antipodal cells (WEBB and GUNNING 1990 , WEBB and GUNNING 1994 ; MANSFIELD et al. 1991 ; SCHNEITZ et al. 1995 ; CHRISTENSEN et al. 1997 ). By morphological criteria, embryo sac development in cap/CAP ovules was indistinguishable from wild type (Fig 1B and Fig C), and in both, <1% of embryo sacs were degenerated (n = 345 and 251, respectively). These results suggest that CAP1 and CAP2 genes are required for female-gametophyte function in pollen tube guidance, fertilization, or postfertilization processes.

A functional ovule and embryo sac are required for correct pollen tube guidance (HULSKAMP et al. 1995B ; RAY et al. 1997 ; SHIMIZU and OKADA 2000 ). We examined how many ovules attracted pollen tubes in heterozygous cap1/CAP1 and cap2/CAP2 plants that had been pollinated with wild-type pollen. Ovules were inspected 20–24 hr after pollination (HAP), using an aniline-blue squash technique (see MATERIALS AND METHODS). Pollen tubes were observed at similar frequencies as wild-type controls (88% for both mutants, N = 495 and 153, respectively, as compared to 90% in wild type, N = 240). Thus, the CAP genes are not required for female-gametophyte function before fertilization.

Whole-mount preparations of ovules 12–60 HAP revealed that >90% of the embryo sacs had initiated embryo and endosperm development in the cap mutants (Table 2). However, embryo and endosperm were subsequently aborted in 50% of the seeds (Table 2). Reciprocal testcrosses between cap1/CAP1 or cap2/CAP2 plants and wild-type plants revealed that abortion of embryo and endosperm at 1–5 days after pollination (DAP) occurred only when the maternal plant was cap/CAP heterozygous (Table 2) and the lethal phenotype was the same as from selfed cap/CAP plants (see below). Taken together, these data strongly suggest that the cap mutants are female-gametophyte mutants displaying maternal effects on embryo and endosperm development.

Early embryo and endosperm development of cap mutant embryo sacs:
Following double fertilization of wild-type embryo sacs, endosperm development is initiated in the central cell before the first embryo division. The endosperm undergoes three rounds of synchronized syncytial nuclear divisions (Fig 2A). Some nuclei migrate toward the chalazal pole where they form a common cytoplasmic pocket termed chalazal cyst or chalazal endosperm (CZE; MANSFIELD and BRIARTY 1990A ; BERGER 1999 ; BROWN et al. 1999 ; BOISNARD-LORIG et al. 2001 ). Initial endosperm development is accompanied by an elongation of the zygote and the migration of the zygote nucleus toward the apex. Concurrent with the fourth syncytial endosperm mitosis, the zygote divides asymmetrically to produce a suspensor precursor cell and the one-cell embryo proper. Following two successive longitudinal divisions (Fig 2C and Fig E), the four-cell (quadrant) embryo divides transversely to form the eight-cell octant embryo (Fig 3G; MANSFIELD and BRIARTY 1991 ; JURGENS 2001 ). At this stage, the endosperm has gone through seven rounds of free nuclear division and is composed of 100 nuclei (Fig 3G). The syncytial mitoses are no longer synchronous although nuclear divisions are coordinated locally within the central peripheral endosperm (PEN), the micropylar peripheral endosperm (MCE), and the CZE. One or two consecutive syncytial divisions commence in the MCE and PEN and as the embryo reaches the heart stage (Fig 3H) cellularization takes place in the endosperm, initiated from the MCE surrounding the embryo. The CZE remains syncytial, containing nuclei of different sizes (MANSFIELD and BRIARTY 1990B ; BROWN et al. 1999 ; BOISNARD-LORIG et al. 2001 ).

View larger version (92K): In this window In a new window Download PPT slide   Figure 2. cap1 embryo and endosperm phenotypes. Comparisons of wild-type (A, C, and E) and cap1 mutant ovules (B, D, and F) were made at the same time after fertilization; staging refers to wild type only. (A and B) Zygote stage. In B, only one enlarged endosperm nucleus is in arrested cap1 mutant (asterisk). (C and D) Two-cell stage with a few enlarged endosperm nuclei in cap1 arrested at elongated-zygote stage. (E and F) Four-cell stage with no cell wall formed between nuclei in apical cell of arrested cap1 mutant; wild-type embryo (E) is not in focal plane. Asterisk, endosperm nucleus; arrow, embryonic nucleus; arrowhead, cell wall from division of zygote. Optical sections were obtained from whole-mount preparations of ovules from cap1/CAP1 siliques crossed with wild-type pollen. Bars: A and B, 20 µm; C and D, 40 µm; E and F, 50 µm.

View larger version (406K): In this window In a new window Download PPT slide   Figure 3. cap2 embryo and endosperm phenotypes. (A–F) Developmental stages of cap2 mutant embryos (asterisks, endosperm nuclei): (A) one-cell stage; (B) four-cell stage; (C) dermatogen stage; (D) early globular stage; (E) late globular stage; (F) early heart stage. (G and H) Wild-type control embryos at octant (G) and early heart (H) stages. Note heterogenous sizes of endosperm nuclei in B and C. (I and J) Comparison of wild-type (I) and cap2 (J) endosperm. Note aggregated endosperm nuclei of different sizes (asterisk, arrow, and arrowheads in J) and uniform nuclear size in I. (K and L) Precocious "septation" (arrowheads) in cap2 endosperm; two different focal planes. Optical sections were obtained from whole-mount preparations of ovules from cap2/CAP2 siliques crossed with wild-type pollen. (G) CLSM image of Feulgen-stained ovule. Bars: A, B, G, and I–L, 20 µm; C–F, 40 µm; H, 80 µm.

To determine the maternal-effect phenotypes of cap1 embryo and endosperm, we analyzed whole-mount preparations of ovules from cap1/CAP1 siliques pollinated with wild-type pollen during development from the zygote to the early heart stage. Although phenotypes were variable, developmental arrest of embryo and endosperm was generally restricted to a relatively small period of development. cap1 embryo sacs were already abnormal at the zygote stage, which corresponds to the second or third syncytial endosperm mitosis in wild type (Fig 2A). The zygotes appeared less elongated and the endosperm contained a single enlarged nucleus (Fig 2B). Approximately 29% of mutant embryos were arrested as zygotes (N = 257, Table 3). Most mutant embryo sacs contained 1–2 or 4–8 endosperm nuclei (Table 3). About 41% of mutant zygotes were able to divide asymmetrically to produce one-cell proembryos (Table 3) although this division was delayed compared to wild-type embryos (Fig 2C and Fig E; Fig 3H). When arrested at the one-cell stage, mutant embryos were most frequently associated with endosperms containing 4–8 nuclei. Approximately 23% of mutant embryo sacs lacked endosperm nuclei, which were probably degraded (Table 3). In 13% of cap1 embryo sacs, one-cell proembryos underwent nuclear division but no cell wall was formed between the daughter nuclei (Table 3; compare Fig 2C and Fig F). No mutant embryos were detected beyond this stage. In general, there was a close correlation between developmental progress of the embryo and endosperm (Table 3). Most two-nucleate cap1 embryos were surrounded by endosperm with 4–24 nuclei (Table 3) although endosperm development was often delayed as compared to their embryo partners (Fig 2F). Moreover, within an embryo sac, endosperm nuclei were always of the same size, with no typical chalazal cyst formed in endosperms with three or more syncytial mitoses. Thus, by morphological criteria, both embryo and endosperm development were affected in cap1 embryo sacs.

In comparison to cap1, cap2 embryo sacs generally did not support endosperm development beyond two syncytial mitoses whereas embryo development was delayed as compared to CAP2 controls or arrested at various developmental stages. When nearly 80% of the CAP2 embryos were at the four- or eight-cell stage, 66% of the cap2 embryos had not reached the four-cell stage (Table 4A; compare Fig 2E and Fig 3G with Fig 3A and Fig B). As development progressed, older stages of cap2 embryos were observed (Table 4B; Fig 3, C–F). Although cap2 embryos were generally larger than CAP2 embryos of the same age, their patterns of cell divisions were fairly normal. However, as wild-type embryos reached the heart stage, the proportion of cap2 embryo sacs with degenerated embryos increased dramatically (Table 4B).

The development of cap2 mutant endosperm was abnormal from the zygote stage onward (Table 4C). The number of endosperm nuclei ranged from 1 to 12, indicating that only very few rounds of nuclear division occurred (compare Fig 3, A–F, with Fig 3G). As CAP2 development progressed, the proportion of degenerated cap2 endosperms increased to 80% (Table 4C). cap2 endosperms had nuclei of different sizes and shapes, in contrast to the homogenous population of nuclei in CAP2 and also in cap1 endosperms (Fig 3, B–D and J; compare with Fig 3I). A chalazal cyst was not observed in cap2 endosperms. A rare feature of cap2 endosperms was the occurrence of domains with multiple nuclei of different sizes and shapes ("multiple" in Table 4C; compare Fig 3I with Fig 3J). These assemblies were surrounded or encapsulated by "walls" or membrane-like structures. No cellularization of cap2 late stage endosperms was observed. However, premature cellularization of endosperms with 8 or fewer nuclei occurred at a low frequency (Fig 3K and Fig L). In these rare cases, all endosperm nuclei were in a different focal plane from that of the embryo, and interestingly no "cell compartments" were binuclear (Fig 3K and Fig L). In conclusion, CAP2 appears to be required in the endosperm from the first syncytial mitosis. The variable enlargement of nuclei as well as the premature cellularization phenotype may suggest interference with cell-cycle regulation in the syncytial endosperm (see DISCUSSION).

The cap mutant phenotype does not depend on CAP gene dosage in the endosperm and embryo:
To determine whether the number of CAP gene copies, rather than the parental origin of the cap allele, may be critical for embryo and endosperm development in cap mutants, we crossed cap/CAP diploid (2n) plants with tetraploid (4n) wild-type pollen donors. Fertilization of cap embryo sacs with CAP/CAP pollen gave cap/cap/CAP/CAP tetraploid endosperms and cap/CAP/CAP triploid embryos. Similar wild-type crosses have been reported to produce viable seeds that were enlarged due to an enlargement of both embryo and endosperm (SCOTT et al. 1998 ). Whereas triploid embryos develop at the same rate as diploid embryos, tetraploid endosperms display an increased rate of endosperm nuclei proliferation (SCOTT et al. 1998 ). We also observed enlargement of embryo and endosperm in ovules of wild-type, cap1/CAP1, and cap2/CAP2 plants fertilized with diploid pollen. In addition, 15–20% of the embryo sacs did not develop, regardless of the maternal genotype, presumably due to the genetic background of the tetraploid paternal line (Table 5). In both cap1 and cap2 embryo sacs, embryo and endosperm development could not be rescued by the presence of a supernumerary paternal CAP allele (Table 5). cap1/CAP1/CAP1 embryos were arrested as binucleate proembryos. cap1/cap1/CAP1/CAP1 endosperms had somewhat enlarged nuclei but otherwise were phenotypically identical to cap1/CAP1/CAP1 endosperms. The same result was obtained for cap2 embryo sacs, although embryo and endosperm carrying a supernumerary CAP2 allele seemed to degenerate prematurely. In summary, these data suggest that the mutant phenotypes of endosperm and embryo do not depend on their own CAP gene dosage but rather on the parental origin of the cap mutant allele.

No fertilization-independent seed development in cap mutant embryo sacs:
In the maternal-effect mutants mea (also called fis1), fis, and fie, endosperm and, except in fie, embryo develop in the absence of fertilization (OHAD et al. 1996 ; CHAUDHURY et al. 1997 ; GROSSNIKLAUS et al. 1998 ). In our hands, 38% of fis1 embryo sacs developed autonomously (see Table 6). To test whether cap1 and cap2 mutants also had this capability, we emasculated cap/CAP plants and analyzed ovules 5–10 days later. All embryo sacs were undeveloped in both mutant lines (Table 2), suggesting that, unlike MEA and related genes, CAP1 and CAP2 are not required for prefertilization repression of embryo and endosperm development.

cap2 disrupts sexual and autonomous development of fis1 mutant embryo sacs:
Disruption of MEA, FIS, and FIE genes leads to autonomous development of embryo and endosperm in unfertilized embryo sacs (OHAD et al. 1996 ; CHAUDHURY et al. 1997 ; GROSSNIKLAUS et al. 1998 ). To examine whether CAP2 is required for development in mea mutant embryo sacs, we crossed plants homozygous for the MEA allele fis1 (LUO et al. 1999 ) with cap2/CAP2 pollen donors and analyzed their F₁ progeny for seed development (Table 6 and Table 7). Four of nine F₁ plants were fis1/FIS1 and exhibited normal seed set upon selfing, although embryo and endosperm development were arrested at heart stage and at endosperm cellularization, respectively, in approximately one-half of the seeds (Table 7). Five F₁ plants were classified as fis1/FIS1;cap2/CAP2 because 50% of the seed showed the cap2 phenotype. Furthermore, only one out of four seeds contained normal embryos upon selfing, as expected from genetic recombination between the MEA and CAP2 loci at opposite ends of chromosome 1. Thus, cap2 was epistatic to fis1 in sexual seed development. This result was confirmed in crosses of fis1/FIS1;cap2/CAP2 plants with wild-type pollen donors (Table 7).

To determine the effect of the cap2 mutation on fertilization-independent seed development, we analyzed ovules of emasculated putative fis1/ FIS1;cap2/CAP2 and fis1/FIS1;CAP2/CAP2 plants 5–10 days after anther removal (Table 6). For both genotypes, 19% of the embryo sacs displayed autonomously developing endosperm (Table 6). In contrast to the CAP2 control, 50% of the ovules from fis1/FIS1;cap2/CAP2 contained endosperms with enlarged nuclei and were arrested after 0–2 syncytial mitoses (Table 6, Fig 4D). Some of the arrested endosperms were mononucleate or trinucleate, and, in rare cases, displayed mitotic spindles in metaphase (Fig 4C). These data indicate that cap2 is epistatic to fis1, implying that CAP2 is also required in autonomous seed development.

View larger version (151K): In this window In a new window Download PPT slide   Figure 4. cap2 fis1 autonomous endosperm development. (A and B) fis1 control. Several endosperm nuclei (asterisks) are shown. (C and D) fis1 cap2: arrested metaphase (C, arrow) and arrested enlarged endosperm nuclei (D, arrowheads) are shown. Optical sections were obtained from whole-mount preparations of unfertilized ovules 5–10 days after anther removal. Bars: A and B, 20 µm; C and D, 10 µm.

No rescue of the cap mutant phenotypes by altering epigenetic gene regulation:
The maternal-effect genes MEA, FIS2, and FIE are expressed in an imprinted manner in early seed development (KINOSHITA et al. 1999 ; VIELLE-CALZADA et al. 1999 ; LUO et al. 2000 ; YADEGARI et al. 2000 ). In addition, their mutant phenotypes can be rescued by pollen genotypes, such as METHYL TRANSFERASE1 antisense (MET1 a/s) transgene or mutations in DECREASE IN DNA METHYLATION1 (DDM1), that are thought to alter epigenetic gene regulation (KINOSHITA et al. 1999 ; VIELLE-CALZADA et al. 1999 ; LUO et al. 2000 ; YADEGARI et al. 2000 ). We examined whether these pollen genotypes also rescued the maternal defects of cap1 and cap2 mutants. As a control, fis1 mutant embryo sacs were fertilized with pollen carrying the MET1 a/s transgene (Table 5). Ovules from the pollinated siliques were examined from 1 to 8 DAP. The control fis1 embryo sacs produced 84 and 47% viable seeds in the presence of MET1 a/s and ddm1-2, respectively (Table 5). By contrast, cap1 and cap2 embryo sacs displayed their characteristic developmental defects, regardless of the presence or absence of MET1 a/s and ddm1-2 (Table 5). The developing endosperms from CAP1 and CAP2 sister embryo sacs carrying the MET1 a/s transgene had larger but fewer endosperm nuclei than those usually found after outcrosses to wild type, which is consistent with the findings of ADAMS et al. 2000 .

FIS2::GUS expression in cap mutant endosperm:
To further investigate the roles of the CAP genes in endosperm development, we analyzed the expression of a FIS2::GUS transgene in the cap mutants (LUO et al. 2000 ). FIS2 encodes a C₂H₂ zinc-finger transcription factor proposed to indirectly regulate the maternal-effect genes FIE and MEA (LUO et al. 1999 , LUO et al. 2000 ). Due to imprinting of FIS2, the FIS2::GUS transgene is not expressed in the endosperm when introduced via pollen (LUO et al. 2000 ). When wild-type, cap1/CAP1, or cap2/CAP2 plants were crossed with FIS2::GUS transgenic donors, no GUS expression was observed during endosperm development, indicating that cap1 and cap2 mutations did not relieve imprinting-mediated expression barriers.

To study FIS2::GUS expression in cap1 and cap2 mutant embryo sacs, FIS2::GUS transgenic plants were crossed with cap1/CAP1, cap2/CAP2, and wild-type control pollen donors. F₁ plants were emasculated and pollinated with wild-type pollen, and ovules were stained for GUS activity. The wild-type control yielded the expected temporal and spatial expression pattern in the developing endosperm (Fig 5A&NDASH;F; see also LUO et al. 2000 ). GUS expression was initially observed in the polar nuclei before and after fusion (Fig 5A and Fig B). After fertilization, GUS expression was strictly limited to endosperm nuclei during the first five to six syncytial mitoses (Fig 5, C–E) before GUS expression became restricted to the nuclear cyst of the CZE (Fig 5F).

View larger version (75K): In this window In a new window Download PPT slide   Figure 5. FIS2::GUS expression in cap1 and cap2 endosperm. (A–F) Wild-type control. GUS expression starts in central cell nuclei before fusion (A) and continues after fusion (B) and in syncytial endosperm nuclei (C–E) and is later restricted to the chalazal cyst (F, asterisk). (G) cap1: faint GUS staining in endosperm nuclei (asterisks). (H–J) cap2: strong GUS staining in enlarged endosperm nuclei (asterisks). Note lack of GUS staining in some endosperm nuclei (J, arrowheads). Optical sections were obtained from whole-mount preparations of GUS-stained ovules. Arrows, embryos. Bars: A–E, 10 µm; F–J, 20 µm.

Ovules from cap1/CAP1;FIS2::GUS/- plants gave the same overall staining frequency as CAP1 controls, and 45% (N = 121) of the cap1 mutant embryo sacs expressed the FIS2::GUS transgene in the endosperm. The GUS signal was present in all endosperm nuclei, although it was much weaker than that in the CAP1 control (compare Fig 5G to Fig 5D), decreasing with each syncytial mitosis. The weaker staining may reflect lower expression levels of FIS2::GUS in the cap1 background or, alternatively, expression of the transgene may have ceased early, with the remaining protein being partitioned during each free nuclear division.

cap2 mutant endosperms displayed high levels of FIS2::GUS expression in their enlarged nuclei at 3 DAP (Fig 5H and Fig I). At 5–7 DAP, CAP2 endosperms expressed the transgene exclusively in the chalazal cyst (Fig 5F) that is missing in cap2 endosperms (see above). Instead, nuclear FIS2::GUS expression persisted in the peripheral and micropylar regions of cap2 endosperms (Fig 5, H–J) whereas other nuclei lacked detectable GUS activity (Fig 5J, arrowheads). Thus, cap1 and cap2 mutant embryo sacs were able to activate the FIS2::GUS transgene.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

We have isolated two gametophytic mutants, cap1 and cap2, in which both embryo and endosperm are developmentally arrested only if the female gametophyte carries the mutant alleles. Our linkage-based screen ruled out both incompletely penetrant dominant maternal effects of the sporophyte and incompletely penetrant dominant zygotic mutations as the mutant phenotype that is always segregated with the genotype of the female gametophyte during meiotic recombination. We will first discuss whether this apparent gametophytic maternal effect is caused by the mutant alleles in the female gametophyte itself or instead by the inactivity of the pollen-derived wild-type alleles during embryo and endosperm development. Subsequently, we will address possible roles of the CAP gene functions in embryo and endosperm development.

Gametophytic maternal effect or gamete-specific imprinting of CAP genes?
By genetic criteria, the CAP genotype of the haploid embryo sac determined whether or not its fertilization products, the diploid embryo and the triploid endosperm, developed normally. The CAP genotype of the pollen had no effect, as shown by reciprocal crosses between cap/CAP heterozygous plants and wild-type plants as well as by crosses between cap/CAP plants and tetraploid pollen donor plants. Furthermore, the CAP2 gene was also required in the autonomous development of fis1 diploid endosperm from unfertilized embryo sacs.

A 2:1 ratio of maternal-to-paternal genomes has long been recognized as crucial for proper endosperm development (for review, see SCOTT et al. 1998 ). Parents of the same ploidy produce viable seed, whereas reciprocal interploidy crosses lead to a reduction or increase in seed size or, in extreme cases, to seed abortion (LIN 1984 ; BIRCHLER 1993 ; SCOTT et al. 1998 ). Different mechanisms have been proposed to account for this effect. The parental conflict theory suggests that conflicting interests of the maternal and paternal genomes are balanced by genomic imprinting (LIN 1984 ; HAIG and WESTOBY 1989 ; SCOTT et al. 1998 ). Experimental evidence for imprinting comes from preferential expression of maternal zein alleles in maize (LUND et al. 1995 ) as well as from recent analyses of the Arabidopsis MEA, FIS, and FIE genes (KINOSHITA et al. 1999 ; VIELLE-CALZADA et al. 1999 , VIELLE-CALZADA et al. 2000 ; LUO et al. 2000 ; VINKENOOG et al. 2000 ; YADEGARI et al. 2000 ). The critical question is whether the cap mutant embryo and endosperm phenotypes are due to imprinting of the paternal alleles or due to a true maternal effect of the female gametophyte.

Mutants of the MEA, FIS, and FIE genes can be rescued by the ddm1 mutation or by MET1 a/s transgene expression, which supports the notion that these genes are imprinted in the paternal genome (KINOSHITA et al. 1999 ; VIELLE-CALZADA et al. 1999 ; LUO et al. 2000 ; VINKENOOG et al. 2000 ). Using the same approach, cap1 and cap2 mutants were not rescued, suggesting that the CAP genes are not imprinted in a similar way. However, DDM1 and MET1 may not be involved in all sorts of genomic imprinting. MET1 is the major maintenance cytosine methyltransferase in Arabidopsis, and a functional knockout by METI a/s transgene expression reduces methylation to 15% of the wild-type level (FINNEGAN and DENNIS 1993 ; FINNEGAN et al. 1996 ; GENGER et al. 1999 ). The DDM1 gene product is a member of the SWI2/SNF2 protein family of chromatin remodeling factors (JEDDELOH et al. 1999 ), and its primary effect in imprinting may involve changes in chromatin conformation (VONGS et al. 1993 ; KAKUTANI et al. 1995 , KAKUTANI et al. 1996 ).

The parent-specific effects in mea and fie mutants may involve different imprinting mechanisms, since ddm1 pollen rescues mea mutants but not fie mutants (YADEGARI et al. 2000 ). In addition, rescue by the MET1 a/s transgene also occurs when the pollen is mutant for mea and fis but not for fie, thus indicating that the phenotypical rescue is not mediated by the activation of paternal alleles and suggesting that different factors are involved in relieving transcriptional silencing (LUO et al. 2000 ; VINKENOOG et al. 2000 ). Candidates include genes such as HOG1, MOM1, SIL1, SIL2, and SOM/DDM, which modify transcriptional gene silencing in Arabidopsis (VONGS et al. 1993 ; FURNER et al. 1998 ; MITTELSTEN SCHEID et al. 1998 ; AMEDEO et al. 2000 ). Considering that mechanisms for imprinting and transcriptional gene silencing are not fully understood in Arabidopsis, we cannot rule out imprinting of CAP1 and CAP2 genes. However, so far no evidence supports imprinting of these genes and therefore we consider it more likely that cap1 and cap2 mutants represent true gametophytic maternal-effect mutants.

Possible roles of CAP gene functions in embryo and endosperm development:
The development of cap1 embryos was arrested very early, with no mutant embryos progressing beyond a binucleate proembryo stage. Furthermore, most embryos were arrested before this stage, suggesting that CAP1 gene function is required from fertilization on in the developing embryo itself. By contrast, the developmental arrest of cap2 embryos was delayed, and in rare cases, the embryo developed to the early heart stage. cap2 embryos displayed no major pattern defects although they appeared enlarged compared to equivalent wild-type stages. The later developmental arrest of cap2 embryos could result from a nursing defect of the mutant endosperm (see also review by LOPES and LARKINS 1993 ). In this interpretation, the primary target of CAP2 function would be the developing endosperm.

Both cap1 and cap2 affect endosperm development, although in different ways. In the cap1 endosperm, development is arrested after a few syncytial mitotic divisions, thus resembling the defect observed in the cap1 embryo. It is therefore likely that CAP1 gene function is required independently in both embryo and endosperm development, in contrast to CAP2. It is conceivable, for example, that CAP1 plays an activating role after fertilization and that, in its absence, embryo and endosperm development depend on maternal supplies. This idea is supported by the decreasing FIS2::GUS signal in cap1 endosperms, which may result from prefertilization expression in the central cell. The presence of maternal supplies in early embryo and endosperm development is also evidenced by one or a few mitotic divisions that occur after fertilization in the absence of zygotic factors required for the formation and/or maintenance of the microtubular cytoskeleton (MAYER et al. 1999 ). In contrast to cap1, cap2 endosperm nuclei are irregular in shape and size, possibly representing variable endoreduplication cycles. The FIS2::GUS expression data also suggest regional heterogeneity among endosperm nuclei. Whether this reflects interference with the cell division machinery or incorrect specification of endosperm domains remains to be determined.

The cap mutants represent novel genes that are presumably expressed in the female gametophyte itself and whose products are required for embryo and endosperm development. This interpretation can be rigorously tested only by the molecular characterization of the CAP genes, which allows the molecular basis for their gametophytic maternal influence on seed development to be determined. The CAP genes may thus represent models for the analysis of maternal factors and mechanisms crucial for embryo and endosperm development.

	FOOTNOTES

¹ Present address: Division of Molecular Biology, Department of Biology, University of Oslo, P.O. Box 1031, Blindern, N-0315 Oslo, Norway.

	ACKNOWLEDGMENTS

We thank Abed Chaudhury and Ming Luo (CISRO, Canberra, Australia) for the FIS2::GUS transgenic and homozygous fis1 lines, Jane Finnegan (CISRO, Canberra, Australia) for the MET1 a/s transgenic line, and Eric Richards (Washington University, St. Louis) for heterozygous ddm1-2 seeds; Heinz Schwarz and Jürgen Berger (Max-Planck Institute for Developmental Biology, Tübingen, Germany) for their assistance with the scanning electron microscope; and our colleagues R. Gross-Hardt, M. Heese, and K. Schrick for critical reading of the manuscript. P.E.G. was a recipient of a graduate student fellowship (Promotionsstipendium des Landes Baden-Württemberg). This work was supported by a Leibniz Award from the Deutsche Forschungsgemeinschaft to G.J. and by a Junior Group Award from the Volkswagen Foundation to M.H.

Manuscript received April 16, 2002; Accepted for publication September 6, 2002.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ADAMS, S., R. VINKENOOG, M. SPIELMAN, H. G. DICKINSON, and R. J. SCOTT, 2000  Parent-of-origin effects on seed development in Arabidopsis thaliana require DNA methylation. Development 127:2493-2502.[Abstract]

AMEDEO, P., Y. HABU, K. AFSAR, O. M. SCHEID, and J. PASZKOWSKI, 2000  Disruption of the plant gene MOM releases transcriptional silencing of methylated genes. Nature 405:203-206.[Medline]

BELL, C. J. and J. R. ECKER, 1994  Assignment of 30 microsatellite loci to the linkage map of Arabidopsis.. Genomics 19:137-144.[Medline]

BERGER, F., 1999  Endosperm development. Curr. Opin. Plant Biol. 2:28-32.[Medline]

BIRCHLER, J. A., 1993  Dosage analysis of maize endosperm development. Annu. Rev. Genet. 27:181-204.[Medline]

BOISNARD-LORIG, C., A. COLON-CARMONA, M. BAUCH, S. HODGE, and P. DOERNER et al., 2001  Dynamic analyses of the expression of the histone::yfp fusion protein in Arabidopsis show that syncytial endosperm is divided in mitotic domains. Plant Cell 13:495-509.[Abstract/Free Full Text]

BONHOMME, S., C. HORLOW, D. VEZON, S. DE LAISSARDIERE, and A. GUYON et al., 1998  T-DNA mediated disruption of essential gametophytic genes in Arabidopsis is unexpectedly rare and cannot be inferred from segregation distortion alone. Mol. Gen. Genet. 260:444-452.[Medline]

BRASELTON, J. P., M. J. WILKINSON, and S. A. CLULOW, 1996  Feulgen staining of intact plant tissues for confocal microscopy. Biotech. Histochem. 71:84-87.[Medline]

BROWN, R. C., B. E. LEMMON, H. NGUYEN, and O. A. OLSEN, 1999  Development of endosperm in Arabidopsis thaliana.. Sex. Plant Reprod. 12:32-42.

CHAUDHURY, A. M., L. MING, C. MILLER, S. CRAIG, and E. S. DENNIS et al., 1997  Fertilization-independent seed development in Arabidopsis thaliana.. Proc. Natl. Acad. Sci. USA 94:4223-4228.[Abstract/Free Full Text]

CHRISTENSEN, C. A., E. J. KING, J. R. JORDAN, and G. N. DREWS, 1997  Megagametogenesis in Arabidopsis wild-type and the Gf mutant. Sex. Plant Reprod. 10:49-64.

CHRISTENSEN, C. A., S. SUBRAMANIAN, and G. N. DREWS, 1998  Identification of gametophytic mutations affecting female gametophyte development in Arabidopsis.. Dev. Biol. 202:136-151.[Medline]

COLOMBO, L., J. FRANKEN, A. R. VAN DER KROL, P. E. WITTICH, and H. J. DONS et al., 1997  Downregulation of ovule-specific MADS box genes from petunia results in maternally controlled defects in seed development. Plant Cell 9:703-715.[Abstract]

CORDTS, S., J. BANTIN, P. E. WITTICH, E. KRANZ, and H. LORZ et al., 2001  ZmES genes encode peptides with structural homology to defensins and are specifically expressed in the female gametophyte of maize. Plant J. 25:103-114.[Medline]

DREWS, G. N., D. LEE, and C. A. CHRISTENSEN, 1998  Genetic analysis of female gametophyte development and function. Plant Cell 10:5-17.[Abstract/Free Full Text]

FELDMANN, K. A., D. A. COURY, and M. L. CHRISTIANSON, 1997  Exceptional segregation of a selectable marker (KanR) in Arabidopsis identifies genes important for gametophytic growth and development. Genetics 147:1411-1422.[Abstract]

FELKER, F. C., D. M. PETERSON, and O. E. NELSON, 1985  Anatomy of immature grains of eight maternal effect shrunken endosperm barley mutants. Am. J. Bot. 72:248-256.

FINNEGAN, E. J. and E. S. DENNIS, 1993  Isolation and identification by sequence homology of a putative cytosine methyltransferase from Arabidopsis thaliana.. Nucleic Acids Res. 21:2383-2388.[Abstract/Free Full Text]

FINNEGAN, E. J., W. J. PEACOCK, and E. S. DENNIS, 1996  Reduced DNA methylation in Arabidopsis thaliana results in abnormal plant development. Proc. Natl. Acad. Sci. USA 93:8449-8454.[Abstract/Free Full Text]

FOLKERS, U., V. KIRIK, U. SCHÖBINGER, S. FALK, and S. KRISHNAKUMAR et al., 2002  The cell morphogenesis gene ANGUSTIFOLIA encodes a CtBP/BARS-like protein and is involved in the control of the microtubule cytoskeleton. EMBO J. 21:1280-1288.[Medline]

FURNER, I. J., M. A. SHEIKH, and C. E. COLLETT, 1998  Gene silencing and homology-depe