Dissection of sexual organ ontogenesis: a genetic analysis of ovule development in Arabidopsis thaliana

The detailed mechanisms governing organogenesis, particularly in plants, remain largely elusive. Flowers, for example, contain several types of floral organs, and a great deal has been learned about how the identity of the floral organs is established (Coen and Meyerowitz, 1991; Ma, 1994; Weigel and Meyerowitz, 1994). However, it is not understood how the three-dimensional architecture of an organ is set up and maintained once it is committed to a particular fate.

To address this problem we turned to the ovule of Arabidopsis thaliana as a model system. A variety of properties make this simply structured yet highly differentiated organ particularly useful to work with and its development has been well documented (Mansfield et al., 1991; Modrusan et al., 1994; Robinson-Beers et al., 1992; Schneitz et al., 1995). The ovules are major female reproductive structures with 40 to 50 ovules located within the gynoecium. The Arabidopsis ovule is of the common Polygonum-type (Bouman, 1984; Reiser and Fischer, 1993; Willemse and van Went, 1984). Each ovule contains a seven-celled embryo sac at maturity. One of these cells is the egg cell proper, which, upon fertilization, generates the zygote and the plant embryo. Two integuments envelop the embryo sac. They will eventually form the seed coat or testa. A stalk-like structure, the funiculus, connects the ovule to the placental region of the gynoecium (Fig. 1). Due to the alternation of generations typical of the plant life cycle, the mature ovule is composed of tissues differing in the ploidy level of their cells. While the embryo sac is haploid, representing the (female) gametophytic phase of the life cycle, the rest of the ovule is diploid forming the sporophytic part of the organ.

One way to approach organogenesis at the functional level is to genetically dissect the process and to study systematically a large number of mutants. In this respect only little is known about ovule development. In particular, simultaneous cosup-pression of two genes, FBP7 and FBP11, results in a conversion of ovule into carpel tissue in petunia (Angenent et al., 1995; Colombo et al., 1995), while in tobacco two mutants have been reported with a similar phenotype (Evans and Malmberg, 1989). In Arabidopsis, genes like APETALA2 (AP2) (Jofuku et al., 1994; Modrusan et al., 1994), BELL1 (BEL1) (Modrusan et al., 1994; Ray et al., 1994; Reiser et al., 1995; Robinson-Beers et al., 1992), AINTEGUMENTA (ANT) (Elliott et al., 1996; Klucher et al., 1996) and ABERRANT TESTA SHAPE (ATS) (Léon-Kloosterziel et al., 1994) seem important for early functions while the floral homeotic gene SUPERMAN (SUP) (Bowman et al., 1992; Sakai et al., 1995) acts at later stages (Gaiser et al., 1995) as does SHORT INTEGUMENTS (SIN1) (Lang et al., 1994; Robinson-Beers et al., 1992).

In this study we have attempted a systematic analysis of ovule development. To identify genes important for ovule development we performed a large-scale single-line mutagenesis. We screened through nuclear-recessive sterile mutants to identify candidates showing defects in ovule ontogenesis. We have isolated many novel mutants as well as additional alleles of most of the previously identified genes. Based on a systematic analysis of the phenotypes we define several classes of mutants and identify distinct steps of the process of ovule organogenesis. The data allow several general conclusions concerning ovule development in Arabidopsis.

Plants were either grown as described by Schneitz et al. (1995) or on autoclaved Florabella cactus soil (Klasmann-Deilmann, Geeste, Germany) beneath OSRAM 58W/72 Biolux fluorescent bulbs (OSRAM, Winterthur, Switzerland) in continuous light (about 100 μE/m²/second at pot height) in a constant temperature room at 22°C. Seeds were mutagenized with ethyl methanesulphonate (EMS) essentially as described by Mayer et al. (1991). The wild-type strain used was Arabidopsis thaliana (L.) Heynh. var. Landsberg (erecta mutant). The M₂ families of approximately 15000 single M₁ lines were screened. We screened through nuclear-recessive sterile mutants to identify those showing defects in ovule ontogenesis. Families that segregated sterile plants were rescreened at least once. The type of sterility was checked with reciprocal outcrosses using the male-sterile line TH154 (Schneitz et al., 1995) and Ler wild-type plants. We investigated ovules of femalesterile (fs) as well as male and female-sterile (ms/fs) lines. Fs lines were sorted by phenotype. Allelism tests were performed with mutants showing a similar phenotype. Tests with bel1 were done using a previously isolated allele (Robinson-Beers et al., 1992). Tests for allelism with ant, ino, lug and sup were performed with appropriate lines (Bowman et al., 1992; Elliott et al., 1996; Gaiser et al., 1995; Liu and Meyerowitz, 1995) and hll, ant, bel and ino alleles were also complementation tested against one another. Because of their similar phenotype the sin1 (Lang et al., 1994; Robinson-Beers et al., 1992) and lal mutants were tested for complementation and the result showed that they are not allelic. The HLL and SUB genes were mapped to the upper region of chromosome 1 (K. Schneitz, unpublished data). Because SIN1 maps to the top of chromosome 1 (Lang et al., 1994) cross-wise allelism tests have been performed showing that SIN1, SUB and HLL are different genes. The bag, mog and ucn mutants are classified on the basis of their distinct phenotypes. The ms/fs lines could be propagated through the heterozygous siblings of individual M2 families. No complementation tests were done with these lines. Ovule preparations were only performed with mutant lines which had been rescreened once or twice.

Whole-mount ovule preparations and staging have been described by Schneitz et al. (1995). Preparations were examined using a Zeiss Axioplan microscope (Carl Zeiss, Oberkochen, Germany) with differential interference contrast (DIC) optics and pictures were taken on Kodak Ektachrome 160T slide film (Eastman Kodak, Rochester, NY). Slides were scanned by a local graphics bureau and put on compact discs (CD) in Kodak Photo-CD format.

For scanning electron microscopy (SEM), freshly opened flower buds were immersed in Jauch’s fixative (70% acetone, 2% glutaraldehyde, in H₂O) and fixed for 6 to 12 hours at room temperature. The tissue was washed 10 times in 70% acetone, rehydrated through an acetone series in cacodylate buffer (50 mM sodium cacodylate pH 7.0), postfixed in 2% osmium tetroxide in cacodylate buffer for 2 hours at room temperature, washed twice in cacodylate buffer for 15 minutes and dehydrated through an acetone series. Subsequently, critical point drying was performed, the specimens were mounted on stubs and the ovules were dissected free. The tissue was sputtered with gold and examined with a Hitachi S4100 field emission scanning electron microscope (Nissei Sangyo Co., Ltd.). Pictures were taken on Kodak TMAX 100 Pro film. Negatives were scanned using a UMAX Power Look PS-2400X flatbed scanner (UMAX Data Systems, Inc.). Pictures were processed for publication on a Power Macintosh 8100/110 computer (Apple, Inc. Cupertino, CA) using Adobe Photoshop 3.0.5 (Adobe Systems, Inc. Mountain View, CA) and Macromedia Freehand 5.0.1 (Macromedia, Inc. San Francisco, CA) software. Printouts were generated on a FUJIX Pictrography 3000 color printer (Fujifilm Dielsdorf Ltd., Dielsdorf, Switzerland).

Arabidopsis ovule development has been well described (Mansfield et al., 1991; Modrusan et al., 1994; Robinson-Beers et al., 1992; Schneitz et al., 1995) and a staging system has been put forward (for details see Schneitz et al., 1995). Therefore we provide only a brief summary here (Fig. 1A-D). Visible morphological development begins with the formation of a small conical protuberance from the placenta (stages 1-I/II). During stage 2-I the megaspore mother cell appears subepidermally within the nucellus. Subsequently meiosis/cytokinesis leads to four haploid megaspores (stage 2-V) and an inner and an outer integument initiate from the chalazal region (stage 2-II/III). They are entirely of epidermal origin and will form two sheets, each primarily consisting of two cell layers, that grow around the nucellus. Stages 3-I to 3-VI relate to the period of embryo sac formation in the nucellus and the maturation of the integuments. Only the proximal-most megaspore survives and undergoes further development to form the seven-celled embryo sac thereby replacing most of the nucellar tissue. Asymmetric growth of the integuments contributes to the characteristic curvature of the ovule. By the end of stage 3 the vascular strand has appeared in the funiculus. Stages 4-I to 4-V cover fertilization and the initial steps of endosperm development and embryogenesis. By this time the final amphitropous shape of the ovule is achieved.

Here we report on the initial characterization of 458 independent mutant lines. On a morphological basis one can discriminate three major phenotypic classes (Table 1). The sporophytic and megagametogene sis-defective (smd) class consisting of 28 mutants which represent 11 different loci (see Materials and methods). Mutants of this class usually show defects in the diploid and haploid tissues of the ovule. The megasporogenesis-defective (msd) group encompassing 270 mutants with defects apparently confined to spore development. The embryo sac-defective (emd) class consists of 160 mutants which show defects during gametophyte development. Mutants belonging to the smd class are female sterile (usually bearing relatively normal pollen) with one exception (see below). Mutants belonging to the msd and emd classes, however, produce either no pollen, a greatly reduced amount of pollen, or degenerate pollen and are therefore male and female sterile. In the following description of the different mutant phenotypes we focus on the defects detected within the ovule.

Mutations in five ‘early-acting’ genes either result in the absence of, or, in one case, in supernumerary ovule structures.The defects usually become visible by mid-stage 2. Mutations in six ‘late-acting’ loci, however, lead to an altered morphogenesis of the ovule and the alterations generally appear during early-to-mid-stage 3.

We isolated two alleles of HLL. The phenotype of the stronger allele 79F2 is depicted in Fig. 2A,B. Apparent defects are restricted to the ovules. At maturity hll ovules are small and lack an embryo sac, both integuments and a mature vascular strand. The defects first become visible around stage 2-II/III. The inner integument does not form beyond the first epidermal cell division. Occasionally even this first cell divsion is absent. With respect to outer integument development, cell enlargement often occurs but epidermal cell divisions are usually not detected. Gametogenesis seemingly stops after tetrad formation. Therefore it looks like the development of the ovule ceases at or shortly after stage 2-V. The tissue may start to degenerate, the nucellar region usually being affected first.

Several mutations in ANT have recently been described (Elliott et al., 1996; Klucher et al., 1996). We isolated four new alleles and the phenotype of a strong allele (72F5) is shown in Fig. 2C,D. Until stage 3-I development within the nucellus proceeds normally (mononuclear embryo sac), however, subsequent development ceases in this element. Within the central region in many cases there is no sign of integument initiation by this stage. Occasionally some development in the proximal half of the domain is apparent. The subepidermal cell divisions have occurred. In addition, there is some enlargement of epidermal cells indicating that outer integument development occurs to a certain extent (data not shown). At stage 4-V practically all ovules show enlarged epidermal cells at the outer integumentary position. This suggests that some outer integument development occurs quite regularly albeit somewhat delayed. However, there is still no morphological sign of inner integument development. The nucellus appears elongated (compare Fig. 2C with 2A) and lacks an embryo sac. Within the funiculus no mature vascular strand is observed.

Three bel1 alleles of similar strength were isolated and the phenotype of one (1460) is summarized in Fig. 2E,F. Its description does not deviate significantly from reports on other bel1 alleles (Modrusan et al., 1994; Ray et al., 1994; Robinson-Beers et al., 1992). Early ovule development in this mutant is normal. Development within the nucellus starts to deviate after stage 3-I (similar to the situation in ant-72F5) when embryo sac ontogenesis ceases. Much earlier (by stage 2-II) a defect is seen in the central domain. Instead of integument initiation, large bulges appear at the position where normally the outer integument is initiated. These bulges consist of epidermal and subepidermal cells and they continue to grow until they extend just above the nucellus. No inner integument development is observed. The stage 4-V ovules show an orthotropous stature (erect) rather than the amphitropous pattern and a broader than normal funiculus. No embryo sac is observed in the nucellus. At very late stages the bulges may develop into carpelloid structures (not shown).

A brief description of an ino mutant has appeared in the literature (Gaiser et al., 1995). Figure 2G,H shows the phenotype of a strong allele (46E4), one of seven new alleles that were isolated. As in the previous examples, early development appears normal and there is a difference in timing of the appearance of defects in the nucellus and the integuments. While at stage 4-V the embryo sac is lacking, its development never stops before the formation of a mono-nuclear embryo sac and a four-nuclear embryo sac is often discernible. Morpho-genesis of the central region is characterized by the lack of the outer integument. While the epidermal cell enlargement typical of early outer integument development occurs, no oblique or periclinal cell division is apparent and development does not proceed further. However, the subepidermal cell proliferation in this region does take place indicating that these two processes are separable. While at maturity the ino ovules exhibit a bending, they do not show the typical amphitropous shape (see Discussion). Except for its shape, the inner integument develops normally. At stage 4-V the endothelium is visible and the third layer, originating from the endothelium at stage 3-VI (Schneitz et al., 1995), exists.

The single ucn mutant we have isolated (135D12) features an astonishing phenotype (Fig. 2I,J). At stage 4-V a horn-like excrescence from the outer integument is present at a distinct proximal-distal location. This bump may vary in size from a mere swelling of a few epidermal cells (2-4 cells in an optical cross section) to a considerable protrusion consisting of two or three cell layers. Often however the layering is not clear cut. The embryo sac is usually missing but occasionally is normal. The mutant is semi-sterile and seeds can be obtained. This extra excrescence appears after stage 2-III when both integuments are initiated and embryo sac morphogenesis may fail at various times after stage 3-I.

We have isolated four mutant alleles of SUB and they all display similar phenotypes. Fig. 3A-C shows aspects of the phenotype of the allele 88C12. At maturity sub ovules exhibit several aberrant features. Most prominently the outer integument often forms several protrusions at the distal tip. In some more extreme cases the outer integument does not encapsulate the inner integument and the nucellus but rather generates an open-folded sheet-like structure (Fig. 3C). Inner integument morphogenesis is relatively normal, developing characteristics like the endothelium, for example. The funiculus, however, is affected because very often no mature vascular strand is discernible. The embryo sac is variably influenced. Its development usually proceeds beyond the mono-nuclear embryo sac and often arrests only at the eight-nuclear stage before cellularization takes place. Some ovules look relatively normal and are sometimes fertile. Development of sub ovules starts to deviate from wild type only after stage 3-I or stage 3-II when the outer integument may not close at the distal (micropylar) tip and embryo sac development can become aberrant. In addition to the defects in the ovule and a slightly reduced amount of pollen sub plants exhibit wrinkled and twisted pistils and the plants look much shorter and stubbier than wild-type plants.

One bag mutant was identified (94G3; Fig. 3D,E,F). Stage 4-V ovules from bag mutants exhibit dramatic alterations. Except for the funiculus with the vascular strand no clear morphological features are recognizable in whole-mount preparations. Interestingly the ovules stick together tightly, a property that was never observed in wild type. At higher magnifications, cells are larger and roundish particular at the edges of the tissue. This can also be seen in scanning electron micrographs (Fig. 3E). An embryo sac is usually not visible. Young ovules (until about stage 2-III) appear normal except that the epidermal cells may look somewhat more roundish. Nevertheless, integument initiation and megaspore mother cells can be clearly observed. Shortly after stage 3-I one fails to detect clear morphological structures. Even though the ovules have already stuck to each other they continue to grow in size. Plants defective for BAG function are also smaller than wild-type, bear flowers that open prematurely and have sepals and petals of irregular shape that may appear glassy. The anthers have a somewhat reduced amount of pollen.

We identified one mutant in the MOG gene (179A10; Fig. 3G,H). While the overall shape of an ovule is maintained, stage 4-V mog ovules display large, balloon-shaped cells in both the inner and the outer integuments and the cell layers are often not regular in appearance. This may lead to an aberrant layer structure in the integuments. The effects on embryo sac development vary from an unrecognizable patch of tissue to an occasional eight-nuclear embryo sac. Similarly, the mature vascular strand in the funiculus may or may not be present. The mog ovules appear normal until about stage 3-II when the first weak signs of larger integument cells become detectable. Similar to sub and bag mutants, mog plants show a pleiotropic phenotype. They have shorter internodes and therefore exhibit a more ‘crowded’ inflorescence. The pistils are thicker than normal and the stigmas bear a reduced number of papillar cells of smaller size. The stamens have short filaments and the anthers bear no pollen.

Mature ovules of the lal mutant resemble to some degree ovules of plants defective for SIN1 (Lang et al., 1994; Robinson-Beers et al., 1992). However, as opposed to the sin1 mutant the defects in lal plants are restricted to the ovule and the two mutants are not allelic (data not shown). We have isolated one lal mutant (196F6; Fig. 3I,J). Stage 4-V ovules of lal plants lack an embryo sac and show a much reduced outer integument with the nucellus and the inner integument protruding. The inner integument appears wild type. Early ovule development in the lal mutant appears normal until early stage 3-I. Within the nucellus embryo sac development proceeds until the four-nuclear stage. Older embryo sacs were not observed. By stage 3-III lal mutant plants bear ovules which show the mutant features. The phenotype suggests that either cell expansion or the anticlinal cell divisions within the outer integument is affected.

The LUG gene was shown to exert a cadastral role during early flower development (Liu and Meyerowitz, 1995). Here we show that LUG has a previously unreported function during ovule development. We have isolated two new lug alleles and the phenotype of one (44D2) is presented in Fig. 3K,L. The phenotype of mature ovules defective in LUG function varies considerably. They often lack an embryo sac and it appears as if development simply stopped prematurely. Occasionally they show an enlarged nucellus and inner integument that extend distally beyond the tip of the outer integument. The defect first becomes evident after stage 3-I. The embryo sac rarely develops beyond the mono-nuclear stage and at stage 3-III the outer integument often remains open distally.

The floral homeotic gene SUP (Bowman et al., 1992; Sakai et al., 1995) was recently shown to have a function during ovule development (Gaiser et al., 1995). We have identified two new sup alleles and the analysis of the ovule phenotype leads to similar conclusions. Ovules of allele 198G3 are depicted in Fig. 3M,N. Most of the stage 4-V sup ovules show a relatively well developed embryo sac. However, instead of showing the amphitropous shape of wild-type ovules they appear straight from about midway through the proximal-distal extension of the integuments. This is due to excess growth of the outer integument in the anterior (adaxial) domain while the inner integument appears comparatively unaffected. At a more proximal position within the chalaza curvature of the ovule is observed and the funiculus also exhibits its typical S-shape. The defect becomes evident by stage 3-II when the outer integument elongates and demonstrates abnormal straight growth.

These two groups are characterized by mutants exhibiting defects restricted to gametogenesis. In all these mutants we cannot detect any obvious defects in the sporophytic tissue of the ovule.

We have isolated 270 mutants which fall into the msd class. At maturity they all show a similar phenotype of displaying no sign of embryo sac development. As an exemplary case we describe the mutant 170F5 (Fig. 4A,B) in more detail. At stage 4-V, either nothing or only a patch of tissue is visible instead of an embryo sac. This patch probably represents a degenerated tetrad because the tetrad usually forms in this mutant although subsequent embryo sac development fails. In 170F5, early megaspore development appears normal and meiosis proceeds until the formation of the tetrad. Embryo sac development is commonly absent. In only two out of over 70 ovules examined did we observe a two nuclear embryo sac with a vacuole. We therefore conclude that the defect in 170F5 occurs prior to embryo sac formation some time during megaspore formation.

As in the msd class, the defects of the 160 mutants of the emd group seem restricted to gametogenesis. However, the defects become apparent during embryo sac development. Characteristically one observes a broad spectrum of defects even within a given mutant line, often ranging from an msd-type defect to a normal-looking embryo sac. Here we show two examples with a relatively narrow range of phenotypes. Fig. 4C depicts a stage 4-V ovule of the mutant 211E6. In nearly all 211E6 ovules, embryo sac development seems to be arrested at the four-nuclear (stage 3-IV) stage with an irregular spatial arrangement of the nuclei. Some ovules show a mature embryo sac. Fig. 4D shows a mature ovule of the mutant 54E9. Generally the embryo sac in this mutant appears normal but shows no sign that fertilization occurred. A few examples exhibit a large scrap of unidentifiable tissue instead of an embryo sac, which suggests that sometimes the embryo sac degenerates.

Ovule development in Arabidopsis can be formally divided into several distinct processes: initiation, specification, pattern formation and morphogenesis (Schneitz et al., 1995). As a first step towards the identification of key factors involved in these processes we isolated 458 sterile mutants which show a defect during ovule development. The morphological analysis of the mutants suggests a hierarchy of genes responsible for initiating and directing morphogenesis in response to specification and pattern formation (Fig. 5).

Even though a large mutant screen was performed several types of mutants are conspicuously absent from this collection. We did not isolate mutants with a defect in specification of ovule identity or pattern formation. We can only speculate about the reasons but one likely cause is genetic redundancy. For example, in petunia two MADS-box genes, FBP7 and FBP11 were identified that specify ovule identity (Angenent et al., 1995; Colombo et al., 1995). Cosuppression experiments with transgenic plants yielded conversions from ovules to carpelloid tissue. However, in these plants expression of both genes was absent. There are many MADS-box genes that are expressed in the young Arabidopsis ovule (Ma et al., 1991; Rounsley et al., 1995; Savidge et al., 1995) which may have such redundant functions. In addition, the analysis of a particular mutant allele of AP2 implicated a function for this gene in ovule specification (Modrusan et al., 1994).

Earlier we put forward a hypothesis in which pattern formation in the ovule primordium results in an arrangement of three main pattern elements along a proximal-distal (P-D) axis (Schneitz et al., 1995). The distal element (D) gives rise to the nucellus, the central element (C) originates the chalaza, and the proximal element (P) forms the funiculus. Morphogenesis then produces element-specific differentiation. While not affecting the proposed proximal-distal patterning process per se (see below), some early-acting SMD-class mutants provide supporting evidence for the existance of such a process. Defects can be region-specific indicating that distinct regions (i.e., pattern elements) are set up and continue to develop independently from each other to a large extent. For example, in bel1 and ino mutants a central (chalazal) domain is severely affected very early on, but development within distal and proximal domains (nucellus and funiculus) proceeds relatively normally. Gametogenesis in the nucellus continues until after early embryo sac formation in both mutants, while funiculus development is usually quite normal. Similar arguments can be made with respect to hll and ant. Additional support stems from the investigation of the BEL1 mRNA expression pattern (Reiser et al., 1995). At early stage 2, preceding the initiation of the integuments, BEL1 transcripts accumulate in the central region of the ovule providing molecular evidence for a central pattern element.

While nucellar development proceeds largely independently of the other elements, the data also indicate that at very late stages (after stage 3-I) its independent mode of development is lost. This implies the importance of other ovular tissue, most likely the integuments, for embryo sac ontogenesis (see also Reiser and Fischer, 1993). The expression patterns of both BEL1 and ANT support this notion. In both instances, the transcripts are not detected in the megaspore mother cell or the developing embryo sac (Elliott et al., 1996; Klucher et al., 1996; Reiser et al., 1995). However, integument and funiculus development is independent of the presence of an embryo sac as indicated by the msd-type mutants.

We have identified mutations in five loci that seem to act at an intermediate level in the sequence of events outlined in Fig. 5. They probably function in response to pattern formation because in all mutants the defects become detectable during ovule mid-stage 2 after the P-D regionalization has become apparent. The earliest steps in the formation of a tissue, particularly the central-region-derived integuments, are impaired. While the typical initial cell shape changes occur, subsequent cell divisions do not, or only irregularly, take place and development ceases. This leads to the absence of entire structures. The mutants seem to be defective in what we conveniently call initiation of morphogenesis, as opposed to defects in cellular aspects of morphogenesis. The latter type of defect leads to a structure being present but having an altered appearance.

Mutations in HLL and ANT lead to a similar but not identical phenotype in the ovule. A major difference between the two mutants lies in the fact that plants defective for ANT show additional defects in other floral organs and bear a reduced number of floral organs, including the ovules (Elliott et al., 1996; Klucher et al., 1996). This suggests that ANT is repeatedly required during floral development and that it plays a role in floral organ and tissue initiation in general. This hypothesis is supported by the ANT mRNA expression pattern and the fact that the gene encodes a putative transcription factor with homology to the floral homeotic gene AP2 (Elliott et al., 1996; Klucher et al., 1996; Weigel, 1995). The apparent defects of mutations in HLL are restricted to the ovule. No significant reduction of the number of floral organs has been observed in hll mutants (K. Schneitz, unpublished data). Experiments are currently underway that address the relationship between HLL and ANT.

The phenotypes of ovules from bel1 and ino mutants differ from the previous two cases, particularly in the central and proximal regions. In bel1 mutants the entire central region is affected. The inner integument seemingly fails to initiate and instead of an outer integument a structure arises which was previously characterized as integument-like (Modrusan et al., 1994; Robinson-Beers et al., 1992). As judged by the mode of initiation (epidermal and subepidermal tissues contribute to the bulges), the difference in cell characteristics (lack of large trapezoidal epidermal cells with prominent vacuoles at the outside), and the subsequent growth pattern, this structure shows no similarity to an outer integument and its identity is not known. The BEL1 gene encodes a homeodomain-containing putative transcription factor (Reiser et al., 1995). Such a molecular function is compatible with the idea of BEL1 being a regulatory gene which initiates morphogenesis in the central domain. The ino mutant specifically lacks an outer integument, while the inner integument seems to undergo normal development until stage 4-V. Both mutants indicate that development of the inner and outer integument is differentially controlled, corroborating ideas about a separate evolutionary origin of the two integuments (Stebbins, 1974; Stewart and Rothwell, 1993). The fertile seed-shape mutant aberrant testa shape (ats) also falls into this early-acting class (Léon-Kloosterziel et al., 1994). At the time of integument initiation the two integuments are not distinguishable as separate structures. The boundary between the inner and outer integument is missing, resulting in a single ‘fused’ integument.

The defect in ucn leads to a supernumerary structure. In this mutant a protrusion forms at a specific location on the circumference of the ovule reminiscent of the initiation of the integuments (Robinson-Beers et al., 1992). The protrusion is generated very early in development, proximal to the outer integument and is often formed by enlarged epidermal cells. It is therefore a possibility that this protrusion represents a small ectopic integument which failed to undergo further development. If this interpretation is correct then UCN is a suppressor of integument initiation.

Following patterning and initiation of morphogenesis the next hierarchical step involves the direction of morphogenesis proper. We identified six different mutants that show impaired morphogenesis. Typically the mutant phenotypes become apparent during ovule stage 3 and occasionally by the end of stage 2. Therefore, they probably act after morphogenesis has been initiated by the genes discussed in the previous section. It also indicates that the morphology of the cells of the ovule primordium is controlled by a different set of genes because the primordia appear normal in the mutants. The late-acting smd class genes appear to control general growth functions, in particular the regulation of cell shape and/or cell proliferation. In the case of the bag mutant the ovules stick together tightly, indicating that cell surface properties are also altered. With one exception (lal) the mutant defects are not restricted to the ovule. They either affect the entire plant like sub, bag or mog mutants or early flower development, like lug and sup (Bowman et al., 1992; Liu and Meyerowitz, 1995; Sakai et al., 1995). In the latter two cases the defects have also been interpreted as disruptions in growth processes (Gaiser et al., 1995; Liu and Meyerowitz, 1995; Sakai et al., 1995). However, the defects in the ovules of sup mutants are much more specific. Mutations with such pleiotropic effects as observed in the former three mutants may be expected in genes that have a function in cellular differentiation. The observed effects on gametogenesis usually become visible after embryo sac onto-genesis is initiated. Either these effects are of secondary nature due to, for example, abnormal integument development or a given gene may also have a specific function during embryo sac development.

There is one more known mutant that belongs in this class. A mutation in SHORT INTEGUMENTS (SIN1), previously identified in our lab, causes a missing embryo sac and shorter integuments in a Landsberg erecta background (Robinson-Beers et al., 1992). The cell number is about normal but cell enlargement seems affected. However, when the sin1 mutation is introduced into a Columbia background the inner integument is extroverted and often expanded (Lang et al., 1994). We have isolated two more putative sin1 alleles but both were unfortunately lost.

A large and distinct gametogenesis-defective class consists of two groups, msd and emd, specifically affecting megaspore/embryo sac development within the nucellus. The lesions in the msd class most likely occur prior to the start of embryo sac development. Development in msd-type mutants fails to proceed beyond the megaspore or mono-nuclear embryo sac stage. Hence they seem to affect sporogenesis. The range of defects observed in emd-class mutants, often even within a mutant line, is not unknown in mutants affecting the embryo sac (Benavante et al., 1989; Huang and Sheridan, 1996; Kennell and Horner, 1985; Kermicle, 1971; Lin, 1978). Currently however, it does not allow for an easy interpretation of their function. The commonality resides in the defects becoming visible during embryo sac development. They may show aberrant numbers of nuclei, generally distorted morphology or pass through development until maturity but fail to proceed further (embryo/endosperm development). Mutants of both classes may show no, few or abnormal pollen and hence are male and female sterile. This indicates that they also function during pollen development. Therefore they open up roads to the study of sporogenesis and gametogenesis in general. The nuclear-recessive nature of the emd class suggests that many of these mutations may simply be leaky msd-type lesions. However, some of them could be involved in embryo sac development indicating that megagametophyte development depends to some extent on the sporophytic genome.

A major aspect of ovule morphogenesis is the formation of the curved shape of the Arabidopsis ovule which is exemplified by the location of the distal tip (micropyle) next to the funiculus. This feature is shared by the majority of Angiosperm species (Bouman, 1984). Erect ovules are called orthotropous and curved ovules anatropous. Amphitropy refers to the curved nucellus and embryo sac that may appear on top of anatropy (Gifford and Foster, 1989).

Analysis of the sup mutant indicated that the anterior (adaxial) side of the outer integument is responsible for causing amphitropy (Gaiser et al., 1995) suggesting that a single gene, SUP, is largely responsible for the conversion of orthotropy to amphitropy. However, the sup mutant still exhibits an approximately right angle bend (Fig. 3M). This is probably due to the cell proliferation that still occurs in the posterior (abaxial) region of the chalaza beneath the outer integument. This differential cell proliferation is maintained in ino mutants (Fig. 2H) and they bear ovules with a bend similar to that seen in ovules defective in SUP function. Ovules of bel1 mutants, lacking this differential cell proliferation in the chalaza, are orthotropous. Taken together the observations imply that at least three processes contribute to the curvature of the ovule, and that only the proximal half of the chalaza and the outer integument are involved. First, extra subepidermal cell divisions in the posterior half of the chalaza cause an initial bending of the ovule (hemitropy). Secondly, differential growth in the outer integument leads to anatropy. Thirdly, enlargement of cells from the anterior side of the base of the outer integument leads to the kink in the nucellus and embryo sac resulting in the amphitropous shape. The amphitropy of mature ovules from msd-class mutants support the notion that the embryo sac is not actively involved in this process. In addition to the ovule-autonomous processes described above for Arabidopsis it is possible that in a number of species the carpel-structure may also dictate ovule shape (Endress, 1996).

With this work we contribute to an emerging framework for an understanding of ovule ontogenesis (Angenent and Colombo, 1996; Reiser and Fischer, 1993) that provides an extensive basis for future investigations of ovule development and game-togenesis at the molecular level.

We thank R. Dudler and P. Endress for critical reading of the manuscript and Antje Redweik for expert technical assistance. We also thank Charles Gasser and David Smyth for exchanging material prior to publication and Elliot Meyerowitz for sending the sup and lug seeds. K. S. is especially indebted to H. R. Hohl and P. Matile for the opportunity to continue this work in Zürich and to Ulrike Utans for all her support. K. S. was a recipient of postdoctoral fellowships from the Swiss National Science Foundation (NF) and from the Ciba-Geigy Jubiläumsstiftung. M. H. was supported by an EMBO long-term post-doctoral fellowship and S. D. K. received a postdoctoral fellowship from the United States Department of Agriculture (USDA). This work was funded by NSF grants DCB-9018889 and IBN-9405391 to R. E. P. and by NF grant 31-42175.94 to K. S. and by the Kanton of Zürich.