NOZZLE regulates proximal-distal pattern formation, cell proliferation and early sporogenesis during ovule development in Arabidopsis thaliana

The ovule of Arabidopsis thaliana has become a prominent model system for the genetic and molecular study of floral organogenesis (Gasser et al., 1998; Schneitz, 1999; Schneitz et al., 1998b). It is located within the gynoecium, often normally composed of two carpels, and represents the progenitor of the seed. As such the ovule is the major female reproductive organ in higher plants. Within the ovule the egg cell is formed, fertilization occurs and subsequently, during the ontogeny of the seed, the embryo develops.

Pattern formation represents an important aspect of early ovule development. For example, several distinct regions can be identified morphologically along a proximal-distal (PD), or chalazal-micropylar, axis. Distally, the nucellus is characterized by the presence of the megaspore mother cell and eventually the embryo sac, which harbors the egg cell. The nucellus thus corresponds to the megasporangium. Centrally, usually two integuments that initiate at its flanks identify the chalaza. Proximally, the stalk or funiculus is characterized by the development of the vascular strand. This PD arrangement of different regional identities is recognizable at a very early stage in Arabidopsis thaliana indicating that it becomes established during the early proliferative phase of ovule primordium formation (Schneitz et al., 1995). Furthermore, it is highly conserved during evolution and thus typical for a generalized seed plant (Esau, 1977).

How is this pattern set up? We have proposed that an early corresponding prepattern is established during early ovule development (Schneitz et al., 1995). The distal pattern element or domain is the progenitor of the nucellus, the central pattern element develops into the chalaza and the proximal domain originates the funiculus. Part of the genetic evidence is based on the observation that many mutations that cause early defects in ovule development lead to corresponding region-specific defects within the ovule, leaving early development in other regions largely intact (Gasser et al., 1998; Schneitz, 1999; Schneitz et al., 1998b). In addition, it was shown that the distal element is established in an independent fashion because a nucellus develops even in the absence of the subnucellar region (Schneitz et al., 1998a). Early molecular support has come from studies on the BELL (BEL1) gene (Reiser et al., 1995). The BEL1 expression pattern provides evidence for the central pattern element as its expression within the ovule primordium gets restricted to a central band, approximately corresponding to the central domain, prior to morphological signs of integument differentiation.

Intriguing questions arise from such considerations. For example, what genes are involved in primordium outgrowth and PD pattern formation and how do they interact to orchestrate these aspects of ovule development? Known genes that control primordium emergence include AINTEGUMENTA (ANT) and HUELLENLOS (HLL). In ant mutants only about half the regular number of primordia are present and the ovules are slightly shortened. In addition, integument development fails to proceed beyond the initial steps, indicating that ANT plays a role in integument development as well (Baker et al., 1997; Elliott et al., 1996; Klucher et al., 1996; Schneitz et al., 1998a; Schneitz et al., 1997). ANT regulates growth and cell proliferation in several types of plant organs (Baker et al., 1997; Elliott et al., 1996; Klucher et al., 1996; Krizek, 1999; Mizukami and Fischer, 2000) and it has been postulated that ANT maintains the meristematic competence of cells during organogenesis (Mizukami and Fischer, 2000). The ANT locus encodes a putative transcription factor of the AP2 class (Elliott et al., 1996; Klucher et al., 1996). Shorter ovules and missing integuments are also observed in hll mutants and HLL and ANT share redundant functions in ovule primordium outgrowth (Schneitz et al., 1998a). Mutations in INNER NO OUTER (INO) do not lead to an obvious defect in primordium outgrowth; however, they do cause an early block in outer integument development (Gaiser et al., 1995; Schneitz et al., 1997; Villanueva et al., 1999). Interestingly, INO encodes a putative transcription factor of the YABBY class (Villanueva et al., 1999) and members of this group of proteins are involved in mediating abaxial cell specification in Arabidopsis (Bowman, 2000; Bowman and Smyth, 1999; Chen et al., 1999; Sawa et al., 1999; Siegfried et al., 1999).

Very few known loci have been implicated to function in PD pattern formation. In fact, the BEL1 gene has been the only good candidate for quite some time. BEL1 encodes a homeodomain putative transcription factor (Reiser et al., 1995). In bel1 mutants the chalazal domain undergoes altered development and irregular outgrowths of unknown identity form instead of integuments (Modrusan et al., 1994; Robinson-Beers et al., 1992; Schneitz et al., 1997). The biological function of BEL1 remains enigmatic as the data are compatible with both a direct role of BEL1 in establishing the chalazal identity, and with a role in chalaza-specific growth control in response to PD pattern formation (Reiser et al., 1995). More recently the NOZZLE (NZZ) gene has been considered as a second prominent candidate (Schiefthaler et al., 1999). A preliminary analysis suggested an early role for NZZ in sporangium formation and sporogenesis. In nzz mutants the nucellus and the pollen sacs are strongly reduced and the founder cells of the sporogeneous lineage fail to develop. The NZZ locus encodes a putative protein of unknown biochemical function. However, computer-based sequence analysis raised the possibility that NZZ represents a nuclear protein and possibly a transcription factor (Schiefthaler et al., 1999). Recently, a mutant termed sporocyteless (spl) was identified and found to be impaired in sporocyte differentiation (Yang et al., 1999). Sequence analysis of the SPL gene showed it to be identical to NZZ and further experiments indicated that the SPL protein is localised in the nucleus.

Here we present a detailed characterisation of NZZ largely focusing on its function in ovule development.

Plants were grown as described by Schneitz et al. (1997) and Arabidopsis thaliana (L.) Heynh. var. Landsberg (erecta mutant) was used as the wild-type strain. The isolation and molecular characterization of the mutant nzz alleles has also been described (Schiefthaler et al., 1999).

For the double-mutant analysis the following alleles were used: nzz-2, nzz-1, bel1-1460 and ant-72F5 (Schneitz et al., 1997), ino-2 (Schneitz et al., 1997; Villanueva et al., 1999) (previously known as ino-46E4), hll-2 (Schneitz et al., 1998a), bel1-3 (Modrusan et al., 1994; Reiser et al., 1995) and aberrant testa shape (ats; Léon-Kloosterziel et al., 1994). Double mutants were identified by their novel phenotypes segregating in the expected frequencies and because they exhibit a nzz-like anther phenotype in addition to the ovule phenotype. Between 14 and 20 individual double mutants were obtained, depending on the combination, except for nzz-2 ant-72F5 (6/1573; double mutants/scored plants) double mutants. The ANT gene (Elliott et al., 1996; Klucher et al., 1996) maps closely to NZZ rendering it difficult to isolate the corresponding double mutant plants.

In situ hybridization experiments, including the generation of NZZ, BEL1 and ANT antisense transcripts, were carried out as described previously (Schiefthaler et al., 1999; Schneitz et al., 1998a). In situ hybridization using the INO probe was performed using a different protocol (Vielle-Calzada et al., 1999) with minor modifications. Near full-length INO antisense and sense transcripts were generated from plasmid pJMV50 (Villanueva et al., 1999) after linearising the template with PstI and XhoI and performing the in vitro transcription reaction with T7 and T3 RNA polymerase, respectively. This antisense probe yielded identical results compared to previously obtained data using a probe lacking the information for the YABBY domain (pJMV86; Villanueva et al., 1999). Light microscopy, scanning electron microscopy, staging, whole-mount ovule preparations and artwork have also been described earlier (Schiefthaler et al., 1999; Schneitz et al., 1998a, 1997, 1995).

Wild-type ovule development in Arabidopsis has been well described previously (Christensen et al., 1997; Mansfield et al., 1991; Modrusan et al., 1994; Robinson-Beers et al., 1992; Schneitz et al., 1995; Webb and Gunning, 1990; Fig. 1A-D,I-L). During stage 1 the primordium forms (stages according to Schneitz et al., 1995). During stage 2 the megaspore mother cell differentiates within the nucellus and enters meiosis followed by cytokinesis. Concomitantly to meiosis the two integuments initiate from the chalaza, each originally consisting of two cell layers, and begin to grow around the nucellus in an asymmetric fashion. This uneven growth will eventually lead to the characteristic kinked (anatropous) shape of the mature ovule. During stage 3 the functional megaspore develops into the embryo sac, integument ontogenesis proceeds and the vascular strand becomes visible within the funiculus. At the end of stage 3 prefertilization development has ended and the mature ovule is ready for postfertilization development which begins with stage 4. For the most part the ovule consists of L1 and L2-derived tissue (Jenik and Irish, 2000).

We have isolated three mutant alleles of NZZ (Schiefthaler et al., 1999). All three mutant nzz alleles behave as recessive hypomorphs and exhibit a Mendelian segregation pattern (not shown). The major defects appear to be restricted to ovule and anther development, thus the mutant plants exhibit male and female sterility (Schiefthaler et al., 1999) (Fig. 1). The inflorescence often appears more compact as well. The phenotypes of the three mutants are quite similar, however, the alterations in nzz-2 are somewhat stronger than the defects in nzz-1 and nzz-3. Therefore, we focus on a description of the ovule aspects of the nzz-2 phenotype. The nzz-2 mutant carries a point mutation in exon 1 of the NZZ gene resulting in a stop codon. The putative wild-type protein has a length of 314 residues. In contrast the mutant protein is shortened to 142 residues. This shorter protein lacks a putative nuclear localisation signal and the D1/D2 region (Schiefthaler et al., 1999).

An obvious defect in nzz-2 mutants is the apparent absence of the prominent megaspore mother cell (MMC) at stage 2-I (Fig. 1M) and a strongly diminished nucellar tissue, which is best seen at stages 2-II/III (Fig. 1F,N). At these stages a nucellus is readily apparent in wild type because of the presence of the MMC (Fig. 1I), and because the nucellus forms a distinct dome atop the developing integuments (Fig. 1B). We have scored ovules of wild type and nzz-2 mutants for the presence of MMCs (at stage 2-III) and embryo sacs (at stages 4-V) (Table 1). All ovules from wild-type plants showed a MMC or embryo sac. In nzz-2 mutants only 5% of ovules at stage 2-III showed a MMC in a still somewhat reduced nucellus (Table 1). Because we cannot detect a MMC it is not clear whether the much reduced distal part in nzz-2 mutants actually represents a nucellus, and thus we refer to it as the distal tip region. The development of the nucellus is not simply delayed in nzz-2 mutants, because the ovule continues development, and, while in about 10% of cases a tiny distal tip region can be seen, a regular-size nucellus is usually not observed even at stage 4-V (Fig. 1P; Table 1). In this context it is noteworthy that two large groups of mutants with a defect in megasporogenesis and/or embryo sac development exist and that in those mutants a nucellus can be observed (Schneitz et al., 1997). We have quantified the reduction by counting the epidermal cells in the distal part of the primordium in a mid-sagittal plane of a single profile flanked by the inner integument at stage 2-IV/V. 17.6±2.2 (mean±s.d.) cells were present in nucelli of wild-type plants (n=17), in contrast to 7.7±2.1 cells in the distal tip region of nzz-2 mutants (n=20). Thus, while being somewhat variable (Fig. 1F) the extension of the distal tip region is routinely only about 40% compared to the nucellar extension in wild type. The entire PD extension of stage 1-II/2-III primordia does not appear to be notably reduced in nzz-2 mutants when compared to wild type (Fig. 1A,B,E,F). As the MMC is usually not detected, the defect is likely to occur prior to or around stage 2-I (see also below). However, we cannot exclude the possibility that nzz-2 mutants are defective in nucellar growth during early to mid-stage 2.

In addition to the defect in the formation of the nucellus the temporal arrangement of integument initiation is altered in nzz-2 mutants. In wild type, the inner integument often initiates slightly earlier (stage 2-II) than the outer integument (stage 2-III) (Robinson-Beers et al., 1992; Schneitz et al., 1995) (Fig. 1B). In nzz-2 mutants the outer integument is usually seen first (Fig. 1F,N). While regular integuments may develop (Fig. 1P) both integuments often appear reduced at later stages (Fig. 1G,O). In rare instances the integuments are even absent, resulting in ovules consisting only of a funiculus (Fig. 1G). Furthermore, the length of the funiculus is also affected. The funiculus is extended due to a larger number of cells (Fig. 1F-H; Table 1). This is particularly visible during mid to late ovule development.

In summary, we find that in nzz-2 mutants the distal region is strongly reduced or missing, a prominent MMC is usually absent, the integuments can be variably shortened, and the funicular PD extension increases due to extra cell proliferation.

In order to investigate the genetic interactions of NZZ and other known genes with a function during early ovule development we analysed a series of corresponding double mutants (see Materials and Methods).

ABERRANT TESTA SHAPE (ATS) is important for integument development since in the ats mutant integument development is impaired leading to a single ‘fused’ integument (Léon-Kloosterziel et al., 1994; Fig. 2A,B). Except for the failure of the ovule to reach an anatropous shape (resembling ovules from superman (sup) mutants; Gaiser et al., 1995; Schneitz et al., 1997) the nzz ats double mutants exhibit a largely additive phenotype (Fig. 2C,D). HUELLENLOS (HLL) is repeatedly required for growth control during ovule development (Schneitz et al., 1998a, 1997). In hll mutants the ovule primordia are shorter, the integuments fail to develop beyond the initial steps and many cells die (Fig. 2E,F). The nzz hll double mutants exhibit an additive phenotype as well (Fig. 2G,H). The results indicate that NZZ functions in a different pathway than ATS or HLL during ovule development.

Ovules of bel1 mutants lack the inner and outer integument and exhibit multicellular protuberances which emerge from the central region (Modrusan et al., 1994; Robinson-Beers et al., 1992; Schneitz et al., 1997). The nzz-2 bel1-1460 double mutants show a synergistic phenotype that is restricted to the ovules (Figs 3E-H, 4AC,D). Such plants originate conspicuous elongated structures instead of true ovules. As judged by the morphology of the epidermal cells the majority of such a structure resembles, an albeit much too long, funiculus. In nzz-2 bel1-3 double mutants this effect is more pronounced than in nzz-2 bel1-1460 mutants (Table 1). However, with a few exceptions, a vascular strand is not routinely detected (data not shown). The distal tip region consists of small uniform cells and is often found bi-or trifurcated at late stages. It is difficult to assess the extent of the distal tissue at early stages. MMCs were observed in this tip region at a frequency comparable to that of nzz-2 single mutants (Table 1). Similar results were obtained with nzz-1 bel-1460 and nzz-2 bel1-3 double mutants (data not shown and Table 1). This indicates that NZZ and BEL1 are still active in bel1 and nzz mutants, respectively, and that the two genes have overlapping functions in the specification of the chalaza.

INNER NO OUTER (INO) has a prominent role in outer integument development (Baker et al., 1997; Schneitz et al., 1997; Villanueva et al., 1999). In ino mutants the outer integument development does not proceed beyond the initial epidermal cell enlargement (Figs 3I-L, 4E,F). Megasporogenesis takes place but embryo sac development ceases around stage 3-IV or later.

In nzz-2 ino-2 double mutants (Figs 3M-P, 4G,H), and in contrast to nzz-2 mutants, the distal region appears to be present at full size. However, only 20% of young ovules at stage 2-III show a MMC (Table 1). At later stages the distal tissue continues to grow and often protrudes from a micropyle which is formed by a somewhat shorter inner integument. The outer integument fails to develop as in ino mutants and the funiculus is of about normal length (Table 1). Comparable results were obtained with nzz-1 ino-2 double mutants (data not shown).

It appears that loss of INO function partially restores nucellar development and suppresses the extra growth of the funiculus in the absence of normal NZZ function.

Besides other functions, AINTEGUMENTA (ANT) has an important role during ovule primordium outgrowth and integument development (Baker et al., 1997; Elliott et al., 1996; Klucher et al., 1996; Schneitz et al., 1998a, 1997). In ant mutants sporogenesis occurs (Fig. 4I; Table 1) but subsequent embryo sac development is aberrant. In addition, the integument development is blocked at an early stage and rarely proceeds beyond the epidermal cell enlargement of the outer integument (Figs 3R-T, 4I,J). Finally, the funiculus is reduced (Schneitz et al., 1998a; Table 1).

Similarly to nzz ino double mutants, nucellar development appears to be partially restored in plants simultaneously defective for NZZ and ANT function (Figs 3U-X, 4K,L). At stage 2-III the distal tip region seems fully grown. A MMC is observed in 60% of these mutants (Table 1) but it never develops into an embryo sac. At late stages the outer integument is slightly more advanced than in ant mutants, while the funiculus appears to be of similar size to that of ant single mutants (Figs 3X, 4L; Table 1).

These data indicate that the establishment of the nzz mutant phenotype requires ANT function as well.

We tested whether the observed interactions between NZZ and BEL1, INO and ANT are reflected at the transcriptional level with a set of in situ hybridization experiments (Figs 5, 6).

In wild type, strong expression of NZZ transcripts can be detected throughout the ovule primordium at stage 1 (Schiefthaler et al., 1999; Fig. 5A-D). During stage 2 NZZ expression is still detected everywhere but at a lower level. Stronger epidermal expression can be seen in the chalaza. It starts prior to visible signs of integument initiation and continues to be present in the developing integuments (Fig. 5B-D). We first tested NZZ expression in nzz-2 mutants and found no deviation from the normal pattern (not shown). In bel1 mutants early expression of NZZ appears normal (Fig. 5E,F) as is the weaker overall expression at slightly later stages. However, strong staining is detected in the developing irregular outgrowths (Fig. 5G,H).

The overall early strong and later weaker expression of NZZ is also observed in ovules of ant and ino mutants (Fig. 5I-P). Often it appears as if the staining in the MMC is somewhat more pronounced in ant and ino (and bel1) mutants compared to wild type. In ant mutants the generally weaker staining is also observed in the chalazal epidermis at stages 2-II/III when in wild type the initiating integuments exhibit the strong labeling (compare Fig. 5J,K with 5B,C). In ino mutants, while NZZ is strongly expressed in the developing inner integument, only weak labeling is detected in the chalazal epidermis at the position where the outer integument normally develops (Fig. 5N-P). These results indicate that ANT and INO act as direct or indirect upstream transcriptional regulators of NZZ during early integument development.

The BEL1 pattern appears normal at early stage 1 (Fig. 6A,E). Around late stage 1-II in wild type, BEL1 expression is excluded from the distal and proximal regions and thus the pattern resembles a central stripe approximately marking the prospective chalaza (Reiser et al., 1995; Fig. 6B,C). Slightly later, particularly the outgrowing integuments exhibit strong BEL1 expression but the rest of the chalaza and a distal part of the funiculus stain as well (Fig. 6D). In nzz-2 mutants a central stripe is generally not observed and the BEL1 pattern occupies the distal-half of the primordium (Fig. 6F,Gb,Hb). While the pattern is clearly widened distally it is likely to be broadened proximally as well since at around stage 2-IV/V a considerable part of the funiculus, compared to wild type, is labeled by the BEL1 probe (Fig. 6Hb). In some cases (less than 5%) a stripe can be detected (Fig. 6Ga,Ha). Those examples probably correspond to the few specimens that regularly develop more distal tissue and eventually a MMC (compare Table 1). The result indicates that in nzz-2 mutants BEL1 mRNA is ectopically expressed in the distal region of the ovule primordium around late stage 1/early stage 2.

In wild type the ANT gene is expressed ubiquitously in the ovule primordium at early stage 1 (Elliott et al., 1996; Fig. 6I). By late stage 1/early stage 2 ANT expression can be found in a central domain (Fig. 6Ja,Jb) and during later stage 2 particularly in the developing integuments and in most of the developing funiculus (Elliott et al., 1996; Fig. 6Jb,K,L). In nzz-2 mutants ANT expression, during early stage 1, is again normal. At late stage 1/early stage 2 the ANT and BEL1 patterns are in part comparable since ANT RNA is also detected in the distal-half of the primordium (Fig. 6N,O). However, some stronger staining is regularly detected in a few cells of the epidermis at an abaxial (posterior) location (arrows). The cells at this position are likely to be part of the developing outer integument which initiates in a similarly asymmetric fashion at the abaxial side of the primordium (Robinson-Beers et al., 1992) and which becomes visible prior to the inner integument in nzz mutants (Fig. 1F,N). Later, ANT expression is found throughout both developing integuments and most of the funiculus (Fig. 6P). As with the BEL1 probe, in a few specimens a less-stained region can be seen at the distal tip (Fig. Pa). In summary, we observe ectopic expression of ANT mRNA in the distal region of the ovule primordium of nzz-2 mutants around late stage 1/early stage 2.

In wild type, INO RNA can first be seen around stage 2-II in about 2 epidermal cells (Fig. 6R). This pattern quickly broadens to encompass about 4-5 epidermal cells. These cells will be part of the abaxial (outer) cell layer of the developing outer integument around stage 2-III-IV (Fig. 6S,T). Interestingly, the tip cell, in a cross section, does not show INO label (Fig. 6S,T). This indicates that INO may have a function in establishing the adaxial-abaxial polarity within the outer integument that is separate from its proposed function in setting up adaxial-abaxial polarity within the chalaza (Villanueva et al., 1999). In nzz-2 mutants the pattern resembles wild type (Fig. 6U-X). However, it is likely to be displaced distally by a few cells since the outer integument initiates more distally in nzz-2 (Fig. 1F,N) (see Discussion). The displacement would be quite difficult to detect in in situ hybridization experiments using sectioned material.

Our genetic and molecular analysis indicates that NZZ is a repressor and plays a central role in anther and several aspects of ovule development (Fig. 7). The phenotypes of our nzz mutants deviate from the reported phenotype of the spl mutant (Yang et al., 1999). In ovules of the single spl mutant that was isolated the alterations in the integuments and the funiculus were not observed. In addition, some late aberrant growth of epidermal cells of the nucellus occurs in spl mutants that we did not detect in our nzz mutants. We assume that the observed discrepancies are essentially due to allele-specific effects or the genetic background. All of the three different nzz mutants isolated in our lab show, to various degrees, the phenotypes described in this paper. In addition, the observed phenotypes of the nzz mutants correlate well with the expression pattern of NZZ (Schiefthaler et al., 1999).

The reduction of the nucellus and the failure to develop a MMC indicates that NZZ has an early function in the formation of the nucellus and/or the MMC. The mutant phenotype and the broad expression pattern raised the possibility that NZZ is required for the early formation of the nucellus and that the absence of a MMC is a secondary defect (Schiefthaler et al., 1999). As an alternative explanation for the reduction of the nucellus it had been suggested that NZZ is primarily required for early megasporogenesis and that nucellar ontogenesis depends on correct MMC formation (Schiefthaler et al., 1999; Yang et al., 1999). The data presented above shed some more light on NZZ function during nucellus development. They provide supporting evidence for the former model without excluding an additional role for NZZ in MMC development.

The double mutant analysis indicates that a main function of NZZ in the formation of the nucellus consists of antagonizing BEL1 and ANT activities in the distal region at late stage 1 prior to the onset of megasporogenesis. Ectopic BEL1 and ANT activity is detrimental to nucellus development with ectopic ANT function leading to more severe defects (see below). Interestingly, the transcription of both genes ceases in the nucellus at stage 1-II in wild type (Elliott et al., 1996; Reiser et al., 1995; Fig. 6B,C,J). Could NZZ antagonize ANT and BEL1 by being responsible for their transcriptional repression in the developing nucellus? Within its limits of resolution our analysis suggests that ovule primordia from wild type and nzz-2 mutants are of comparable PD extension at the end of stage 1 (Fig. 1A,B,E,F). In addition, the expression domains of BEL1 and ANT are broader and extend to the distal edge of the primordium at stage 1-II/2-I. Thus, these findings raise the possibility that NZZ is a negative regulator of BEL1 and ANT transcription in the incipient nucellus at stage 1-II/2-I. At present it is unclear how direct this regulation is. Given that NZZ is expressed in the distal region around late stage 1/early stage 2 and that the putative NZZ protein shows features of a transcription factor (Schiefthaler et al., 1999; Yang et al., 1999) and localizes to the nucleus (Yang et al., 1999) we would like to propose that NZZ acts as a transcriptional repressor of ANT and BEL1. Of course, further experiments are needed to clarify this issue. Taking into consideration the early ubiquitous wild-type NZZ expression, the data also indicate that additional factors must be involved in controlling ANT and BEL1 expression distally. Furthermore, in contrast to the situation with BEL1 the interactions of ANT and NZZ are more complex. Besides the ectopic distal expression at stage 1-II, there is a focus of stronger epidermal ANT expression at an abaxial position, close to where the outer integument develops, which is not found in wild type (Fig. 6N,O). We believe this pattern is reflecting a function of NZZ during integument development (see below).

How does INO fit into the picture? The phenotype of nzz ino double mutants suggests that NZZ is also a repressor of INO in the distal region and that ectopic INO activity leads to a failure in nucellus formation as well. We believe that the simultaneous misexpression of BEL1 and ANT probably leads to the ectopic activity of INO, since rendering ANT expression alone independent of NZZ activity, by putting ANT under the control of the ubiquitously active 35S CaMV promoter, does not result in a reduction of the nucellus (Krizek, 1999; Mizukami and Fischer, 2000). Furthermore, we have found that at early stage 2, INO RNA is absent in ant and bel1 mutants (S. B. and K. S., unpublished data). Still, INO continues to be expressed exclusively in a few abaxial epidermal cells. The homeobox gene ATML1 could be a possible candidate for providing epidermis-specific activity (Lu et al., 1996) in this process. How does INO control nucellus development in nzz mutants? We hypothesize that INO acts in a non cell-autonomous fashion and influences the development of the subepidermal region adjacent to the early INO expression domain at a stage before the outer integument is discernible as an outgrowth. This could also explain the funicular phenotype of nzz ino double mutants (see below). Non cell-autonomous INO function could be restricted to nzz mutants but we think it likely to occur also in wild type since it could well explain the early formation of an adaxial bulge, located within the chalaza, in ino mutants (Baker et al., 1997; Schneitz et al., 1997; Villanueva et al., 1999; Fig. 3K).

In summary we outline one possible model that rationalizes our findings (Fig. 7). A key aspect is that NZZ functions as a negative regulator of ANT and BEL1 in the distal region of the ovule primordium. If NZZ function is absent around stages 1-II/2-I co-misexpression of BEL1 and ANT in the distal domain, in concert with additional factors, leads to the ectopic activation of INO at a slightly more distal position within the epidermis. INO indirectly influences the development of the tissue neighboring its expression domain, leading to a chalazal rather than a nucellar specification, thus resulting in a ‘distal shift’ of the chalaza at the expense of the nucellus. In nzz ino double mutants ANT and BEL1 would still be ectopically expressed in the distal tip; however, this expression, due to the absence of INO activity, does not result in the chalaza ‘shifting distally’ and the replacement of nucellar tissue, rather, it interferes with later aspects of nucellar differentiation. Compared to BEL1, ANT would be the factor with a higher impact in negatively influencing further nucellar differentiation since plants defective for NZZ and ANT function have a much larger proportion of nucelli with a MMC than nzz bel1 double mutants. Alternatively, it is possible that NZZ has a function during MMC differentiation or meiosis as well or that additional, as yet unknown factors, come into play.

In nzz bel1 double mutants the chalaza seems at least partially substituted by a funiculus, as indicated by the epidermal morphology at the corresponding PD position. The molecular details of the interaction between NZZ and BEL1 in the central domain are presently unclear. However, the defect can be described as a homeotic phenotype with the chalaza replaced by a funiculus (or a pattern duplication of the proximal domain) indicating that NZZ and BEL1 are required for the specification of the chalaza and thus represent chalaza identity genes. Since the nzz and bel1 alleles used are recessive loss-of-function mutations it appears that in wild type, NZZ and BEL1 somehow direct a developmental pathway towards chalazal identity, which otherwise, perhaps by default, would lead to funicular identity.

The ovules of nzz mutants often exhibit a reduced growth of both integuments indicating that NZZ has a function in integument development (Fig. 1G,H). This is supported by the fact that elevated levels of NZZ expression can be found in both integuments starting around the time of their inception (Schiefthaler et al., 1999; Fig. 5B-D). The genetic analysis suggests that ant is essentially epistatic to nzz as far as integument development is concerned (Fig. 7). With respect to outer integument ontogenesis this holds true for ino as well. Interestingly, in ant mutants the elevated NZZ expression in the entire chalazal epidermis is not observed, while in ino mutants specifically the elevated NZZ expression in the outer integument domain of the chalaza is not detected. Thus, it appears that during the initial phase of integument development NZZ functions downstream of ANT and INO. Furthermore, ANT and INO act as transcriptional regulators of NZZ in this process with ANT functioning during the development of both integuments and INO being specific for outer integument ontogenesis. Both genes are not required for basal level of NZZ transcription, however.

One of the functions of NZZ in integument development could again be in repressing ANT activity. This hypothesis could explain the focus of stronger epidermal ANT expression in young ovule primordia of nzz mutants. In addition, constitutive overexpression of ANT in the Ler background leads to a reduction of the integuments comparable to the reduction observed in nzz mutants (Krizek, 1999). If true it lends support to the idea that proper integument development depends on the precise level of ANT activity (Krizek, 1999).

How do ANT and INO relate to each other during early outer integument development? Genetic experiments indicated that ant is epistatic to ino (Baker et al., 1997) but it is presently unclear how this is reflected at the molecular level. In any case, both ANT and INO encode putative transcription factors (Elliott et al., 1996; Klucher et al., 1996; Villanueva et al., 1999) and it will be interesting to explore whether the two genes control NZZ transcription directly.

The funiculus of a nzz mutant ovules shows hyperplastic growth since the funiculus is elongated due to a larger cell number but the overall morphology seems not affected (Table 1; Fig. 1G,H). The hyperplastic growth in the funiculus is suppressed in nzz ant and nzz ino double mutants indicating that NZZ functions as a repressor of those two genes during funiculus development. This interpretation fits well with respect to ANT, since overexpressing ANT in the Ler background essentially leads to a phenocopy of the nzz phenotype in the funiculus (and the integuments, see above. Compare Fig. 1G,H with figure 3N-P in Krizek, 1999). Unfortunately, due to the absence of suitable markers, it remains an open question whether or not the funicular repression of ANT by NZZ occurs at the transcriptional level as well. It often appears as if the expression domain of ANT in the distal part of the funiculus at early stage 2 is somewhat broader in nzz mutants than in wild type. However, we found it difficult to compare the exact position of the proximal boundary (or the level) of the ANT expression in funiculi of wild type and nzz mutants at later stages. This was in part due to the fact that it was difficult to establish whether or not the very proximal base of the funiculus was stained.

The genetics suggest that NZZ is a repressor of INO during funiculus formation. How this is achieved remains at present difficult to assess. One possible explanation of the nzz ino funicular phenotype is that INO renders the neighboring developing funicular cells competent to respond to elevated ANT levels caused by the absence of wild-type NZZ activity. Experiments are underway that address this issue in more detail.

Since there is generally no cell movement in plants, organogenesis essentially results from the patterned control of cell division and cell shape changes (Meyerowitz, 1997). How is this achieved during ovule development? NZZ and BEL1 have properties of chalaza identity genes. In addition, NZZ negatively regulates BEL1, in the nucellus, and the cell number control gene, ANT, in the nucellus, the funiculus and probably the developing integuments. Thus, there are aspects of NZZ function related to conferring identity and to controlling cell number. It suggests that NZZ links PD patterning and the control of cell proliferation during ovule development. Further analysis of NZZ should reveal additional details about the underlying molecular mechanism.

We thank U. Jauch for help with the SEM and J.-J. Pittet for the artwork. We also thank David Chevalier, Ursula Schiefthaler and Patrick Sieber for stimulating discussions and Martin Hülskamp for comments on the manuscript. This work was supported by the Swiss National Science Foundation (grant 31-53032.97) and by the Kanton of Zürich.