The globby1-1 (glo1-1) mutation disrupts nuclear and cell division in the developing maize seed causing alterations in endosperm cell fate and tissue differentiation

Control of cell division and expansion is crucial for developmental patterning and is likely to be mediated by factors operating at different organisational levels (Irish and Jenik,2001). Although some mutations causing altered division rates or aberrant cytokinesis affect overall plant morphogenesis(Liu et al., 1997), others do not (Blilou et al., 2002; Cleary and Smith, 1998). However, the increasing number of mutations reported to generate cytokinetic defects that lead to alterations in cell fate(Assaad et al., 1996; Chen and McCormick, 1996; Kessler et al., 2002) further demonstrate the importance of the precise timing and proper execution of cell division in plant organogenesis.

The maize seed is composed of an embryo and a persistent endosperm, and is proving an effective model system for the study of basic mechanisms of plant development. The endosperm is a simple structure that originates as a triploid central cell, the result of the fusion of the diploid central cell nucleus with one of the two sperm cell nuclei(Kiesselbach, 1949). A succession of free-nuclear divisions results in the formation of the single-celled syncytium, which bears strong similarities to the Drosophila blastoderm. Cellularisation of the syncytium is initiated by the formation of radial microtubule systems (RMS), which emanate from the nuclear envelopes. Adventitious phragmoplasts are deposited at the points of intersection, forming open-ended tube-like alveolar structures with the open end facing the interior of the endosperm(Brown et al., 1994; Olsen et al., 1995). Coordinated mitotic divisions of nuclei within alveoli are followed by cytokinesis, thus giving rise to a single peripheral layer of cells and a new layer of open-ended alveoli (Olsen,2001). Cellularisation proceeds in a centripetal manner, as this cycle is repeated, until the endosperm becomes fully cellular.

The molecular systems regulating early endosperm development, including the transition from syncytial to cellular stages, remain unclear. Studies in Arabidopsis have shown that the process of endosperm cellularisation shares multiple components with cytokinesis(Sørensen et al.,2002). However, the isolation of a small but significant number of mutations which specifically affect either cytokinesis only in the embryo(Assaad et al., 1996) or cellularisation of the endosperm (Liu and Meinke, 1998; Sørensen et al., 2002), point to the existence of additional mechanisms operating in these two structures(Dickinson, 2003).

Unlike Arabidopsis, maize endosperm develops four structurally and functionally distinct tissues – aleurone, basal endosperm transfer layer(BETL), embryo-surrounding region (ESR) and starchy endosperm (SE) –which are believed to be specified during the free-nuclear to cellularisation stages of development (Becraft,2001; Olsen,2001). A significant number of mutants with defective aleurone development have thus far provided valuable insight into aleurone cell fate and differentiation (Becraft and Asuncion-Crabb, 2000; Becraft et al., 1996; Lid et al.,2002; Shen et al.,2003). By contrast, no mutants with structural defects in either the BETL or the ESR domains alone have been identified, and as a result,little is known of the mechanisms regulating BETL and ESR formation(Becker et al., 1999; Becraft, 2001).

We describe here a recessive lethal mutant characterised by a distinctive globular embryo and endosperm morphology, termed globby1-1(glo1-1). A key feature of the glo1-1 phenotype is aberrant nuclear and cell proliferation in the early syncytial and cellular endosperm,respectively, which to our knowledge has not been previously reported for monocots. As a consequence of these early abnormalities, localised disruptions in cell organisation of the BETL arise, and some cells with aleurone characteristics form ectopically in the regions of SE and BETL. These findings are inconsistent with a previous model for aleurone development, where it was proposed that aleurone cell fate was acquired through signalling from the maternal tissue to the endosperm surface, mediated via the CRINKLY4 receptor kinase (Olsen, 1998). Instead, our data point to the presence of other unknown developmental cue(s) operating within the starchy endosperm that confer aleurone cell identity.

The glo1-1 and glo1-2 mutants were isolated from active Activator (Ac) transposon containing lines maintained in a standard W22 inbred. B-A translocation stocks and the following genetic stocks located on chromosome 1L were used for allelism tests with heterozygous glo1-1/+ plants: cp3-N888A, cp*-N918A, cp*-N948A, dek22-1113A, dek2-1315A, dek*MS6214, gm*-N1303 and ptd1, and were provided by the Maize Stock Center (Urbana, IL, USA). The embryo and aleurone-specific pVP1::GUS seeds, and BETL-specific pBET1-GUS(Hueros et al., 1999) seeds were kindly provided by Drs. P. Perez (Biogemma, France) and R. Thompson(Max-Planck-Institut, Germany), respectively.

For genetic analysis and mapping, plants were grown in fields at Cornell University (Aurora, NY). Glasshouse grown plants were used for all other analyses and either grown at Oxford or Jealott's Hill, under the following regime: 16 hours day length (supplemented with metal halide lamps at 250 mmols, when required) at 22-28°C during the day, and at 16-20°C at night. Humidity levels were set at ∼40-50% daytime and ∼60-70% at night.

Plants heterozygous for the glo1-1 allele were self-pollinated to recover mutants, which segregated as a single recessive Mendelian trait (data not shown). For analysis of early endosperm developmental stages, kernels were harvested from the mid portion of the ear and the remaining kernels were left to mature to score the mutant phenotype. Fresh hand sections of mature kernels were obtained after imbibing seeds on wet tissue for 2 days, and cutting through the central longitudinal axis with a sharp blade. For phenotypic descriptions of developing embryos, we followed nomenclature according to Abbe and Stein (Abbe and Stein,1954).

For epifluorescence microscopy, kernels were trimmed along the mediolateral axis and fixed with FAA (5% formaldehyde, 5% acetic acid, 45% ethanol) for 15 minutes under a gentle vacuum. After infiltration, samples were left overnight at 4°C in fresh fixative, dehydrated through a graded ethanol series,cleared with Histoclear (National Diagnostics, Hull, UK) and wax embedded in Paraplast Plus (Sigma, St Louis, MO). Sections were cut at 8-10 μm, mounted on Superfrost Plus slides (BDH) and stained with DAPI(Ruzin, 1999) at a final concentration of 1-2 mg/ml in Vectashield (Vector laboratories, Peterborough,UK). Slides were examined with a Zeiss Axiophot microscope using a 50 W mercury lamp and the following filter set: 365 nm excitation, 395 nm dichroic and 420 nm long-pass emission.

For light microscopy and transmission electron microscopy (TEM), tissue was fixed overnight in 100 mM phosphate buffer (pH 7.5), with 4% paraformaldehyde and 1% gluteraldehyde at room temperature. The material was then washed in four 15 minutes changes of phosphate buffer and treated in 1% aqueous osmium tetraoxide buffer, followed by dehydration in a graded acetone series and embedding in epoxy resin. Thick sections were cut at 2 μm with a tungsten-coated glass knife, affixed to pre-treated glass slides, stained with Toluidine Blue O (Feder and O'Brien,1968) and viewed under bright field optics using a Zeiss AxioPhot microscope. Ultra-thin sections were cut using a diamond knife at 80 nm,stained with uranyl acetate and lead citrate on a 2168 Ultrostainer (Carlsberg System), and collected on 2 mm copper grids coated with Butvar B98 support film. Grids were examined with a JEOL 2000EX TEM operating at 80 kV accelerating voltage and negatives were digitally imaged after conventional development.

In situ hybridisation was performed on developing kernels at 7 and 10 days after pollination (dap) according to Jackson(Jackson, 1991), with minor modifications. Briefly, kernels were trimmed along the medial-lateral axis and immediately fixed in ice-cold FAA, dehydrated in an ethanol series, and embedded in wax. Sections were cut at 10-12 μm and affixed onto pre-treated Superfrost Plus slides (BDH). Riboprobes were labelled using the DIG RNA labeling mix (Boehringer Mannheim, catalogue number 1175025) according to manufacturer's instructions, and slides were hybridised overnight at 50°C. Slides were viewed with a Zeiss AxioPhot microscope under DIC3-5 optics and images were digitally recorded.

Plants were genotyped for β-glucuronidase (GUS) transcriptional fusions via PCR using GUS-specific oligonucleotides (data not shown), and backcrossed to glo1-1/+ plants for three successive generations. Kernels were cut longitudinally and GUS was detected histochemically according to a method previously described(Jefferson et al., 1987), with slight modifications. After staining at 37°C overnight, the material was fixed in phosphate buffer and wax embedded, as described above. Sections were cut at 12-15 μm and viewed under Nomarski optics.

The glo1-1 mutant and a second allele, glo1-2, were both isolated from an active Ac transposon population and segregated as single recessive seed lethal traits. Although the degree of phenotypic severity varies within individual segregating ears(Fig. 1A,B), both glo1-1 and glo1-2 mutant kernels are small in size and possess identical phenotypes (data not shown). For this reason, only the glo1-1 allele has been used in the phenotypic characterisation described below. An Ac element linked to the glo1-1 mutation was identified through Southern hybridisation and a region flanking the glo1-1-linked Ac element was cloned and mapped to position 1L, bin 1.07 between SSR loci umc1358 and umc1128 using the Maize Missouri IBM mapping population. Furthermore, the glo1-1 mutation was uncovered by crosses to the B-A translocation stock TB-1La(Beckett, 1978), thus confirming its position to the long arm of chromosome 1(Fig. 1C). Complementation analysis of glo1-1 with other known seed mutants located on chromosome 1L demonstrated that glo1-1 and gm*-N1303(Neuffer and Sheridan, 1980; Sheridan and Neuffer, 1980)were allelic (data not shown).

To ascertain whether the lethality of the glo1-1 embryo was a consequence of abnormalities observed in the endosperm, we attempted to create mosaics with a hyperploid endosperm (glo1-1/ glo1-1/+ /+)and hypoploid (glo1-1/–) embryo(Fig. 1C,D). If the embryo lethality was due to the absence of the GLO1 product in the endosperm, then we reasoned that we should be able to rescue the mutant phenotype in the hyperploid endosperm and in the hemizygous embryos. As shown in Fig. 1D, many of these kernels lacked a distinct embryo, suggesting that glo1 is necessary and expressed in cells of the developing embryo. Furthermore, progeny containing hypoploid (glo1-1/ glo1-1/ –) endosperms and hyperploid (glo1-1/+ /+) embryos(Fig. 1D) failed to germinate,indicating that expression of glo1 in embryo tissues is necessary,but not sufficient, for proper kernel development.

To determine when in endosperm development GLO1 function is required,serial sections of sibling kernels were compared from wild-type and self-pollinated glo1-1/+ heterozygous plants between 2 and 4 dap. We examined endosperms that ranged from syncytial, to cellularising to fully cellular stages of development. In wild-type syncytia, nuclei are suspended in a thin layer of cytoplasm surrounding the large central vacuole(Fig. 2A). Remarkably, subtle structural differences and anomalies in the distribution of nuclei were observed in ∼25% of syncytia examined from segregating, but not wild-type,ears (Fig. 2B). We interpret these anomalies as indicating the lack of GLO1 in these syncytia. At a point when the first cell layer forms in wild-type syncytia(Fig. 2C), we noticed an unusual overproliferation of cells confined to basal regions of endosperms interpreted as glo1-1 (Fig. 2D). By 4 dap, wild-type endosperms were fully cellular and had a rounded, conical shaped morphology (Fig. 2E), whereas putative glo1-1 endosperms often showed signs of constrictions around the mid-apical region(Fig. 2F) that became greatly accentuated during later stages of development. Our current data thus indicate that the glo1-1 mutation affects both nuclear divisions in the syncytium and endosperm cellularisation (summarised in Table 1), which dramatically alter subsequent development of the mutant endosperm.

The comparisons of embryos isolated from 8 dap wild-type and glo1-1 kernels indicated that glo1-1 mutants showed a greatly retarded growth, appearing as mid-transition phase embryos while their wild-type siblings had entered the early coleoptilar phase (data not shown). A comparative analysis of wild-type and glo1-1 mid-transition stage embryos revealed profound differences in cellular anatomy(Fig. 3A,B). In wild-type, the mid-transition stage embryo is located within the ESR domain of the endosperm and consists of the apical embryo proper and basal suspensor(Fig. 3A). The embryo proper is characterised by small cytoplasmically dense cells surrounded by a single peripheral layer of protodermal cells, while cells of the suspensor are larger and vacuolated (Fig. 3A). By contrast, an irregular protoderm was often observed in glo1-1embryos, accompanied by an occasional misplacement of densely cytoplasmic cells located in the abnormal suspensor(Fig. 3B). At later stages,mutant embryos either degenerated, some completely, or persisted as a mass of undifferentiated cells, lacking any visible signs of a shoot apical meristem(SAM), while only 16% formed rudimentary primary roots upon germination (data not shown).

At 8 dap, mutant endosperms were reduced in size relative to their wild-type siblings and glo1-1 endosperm tissues showed developmental abnormalities. In many glo1-1 mutant kernels the BETL region, which is located at the endosperm-maternal tissue interface(Fig. 3C), was disrupted. The severity of this disruption varied from a small number of comparatively undifferentiated cells (Fig. 3D) to a total disorganisation of cells. Conventional microscopy of these maturing mutant kernels also revealed purple pigmentation in a significant number of cells resembling anthocyanin pigment seen in the aleurone (data not shown). In contrast to the uniform patterning of cells in the wild-type SE (Fig. 3E),striking abnormalities occurred in the glo1-1 SE(Fig. 3F), including the clustering of small cells, to an apparent loss of cell adhesion and unusual cell wall thickenings. The monolayer of aleurone cells that regularly invests the larger starchy cells in wild-type endosperms(Fig. 3G) was also disrupted in glo1-1, principally on the abgerminal face of mutant endosperms (data not shown). Other abnormalities included areas containing multiple aleurone layers (Fig. 3H), irregular proliferation of aleurone at the endosperm apex(Fig. 3I), and regions devoid of aleurone altogether (Fig. 3J). To summarise, pattern formation is severely disrupted in glo1-1 embryos, whereas endosperm patterning appears to be affected to a lesser extent as the ESR, BETL, SE and aleurone domains are more or less established.

Interestingly, cell wall projections or `stubs'(Fig. 3F,I,J) and multinucleate cells were occasionally observed in both embryo (data not shown) and endosperm cells (Fig. 3D,I). Using TEM to investigate further these possible cytokinesis-associated defects, we examined wild-type and mutant kernels in the sub-aleurone region, as it is the most actively dividing area in the 8 dap endosperm(Olsen et al., 1999). In wild-type cells, the protoplast surface was lined by regular aggregations of ER (Fig. 4A), a feature absent in the mutant (Fig. 4B,E). Irregular deposits of cell wall material(Fig. 4B) and the appearance of striking cell wall `stubs' in the mutant, comprising all elements of the normal cell wall (Fig. 4B,C),were also observed in addition to multinucleate cells(Fig. 4D). Furthermore,cell-cell adhesion appeared to be poor in glo1-1, with interfaces between cells frequently featuring gaps and lacunae(Fig. 4E,F). Interestingly,multivesicular bodies (MVBs) were commonly associated with irregular cell wall thickenings, where they appeared to secrete their contents through the plasma membrane (Fig. 4E). Strikingly,large intercellular lacunae that characterise the SE region in glo1-1endosperms were bordered by isodiametric cells with dense cytoplasms, which were microscopically indistinguishable from aleurone cells(Fig. 4F).

mRNA in situ hybridisation on wild-type and glo1-1 endosperms was performed to assess the early expression patterns of genes normally restricted to the ESR and to the BETL. The ZmESR2 probe (P. Rogowsky, Lyon), was used to detect all three ESR-specific transcripts (ZmESR1, ZmESR2 and ZmESR3)(Opsahl-Ferstad et al., 1997). In 7 dap wild-type kernels, ZmESR transcripts were highly abundant in the ESR(Fig. 5A), whereas in 7 and 10 dap glo1-1 endosperms, ZmESR transcripts were restricted to the basal extremity of the suspensor, in close contact with the pericarp(Fig. 5B,C).

Similarly, expression of an early BETL-specific gene, ZmMEG1-1(J.F.G.-M., unpublished) was dramatically altered in glo1-1endosperms. At 7 dap, ZmMEG1-1 transcripts were abundant throughout the wild-type BETL (Fig. 5D), while in glo1-1, transcripts were present at lower levels and were often restricted to small fields within the transfer tissue(Fig. 5E). To investigate further anomalies in the glo1-1 BETL at later stages of development(14 dap), plants carrying a BET1-specific promoter-GUS fusion(Hueros et al., 1999) were introgressed in the glo1-1 background(Fig. 5F). In all circumstances, GUS staining was only ever detected in the BETL region,although expression of the pBET1-GUS reporter was frequently observed to be`patchy' (Fig. 5G) in the basal region of glo1-1 endosperms. Thus, it appears that the glo1-1 mutation results in structural irregularities and altered patterns of gene expression in basal endosperm (ESR and BETL) domains.

To ascertain the positioning of aleurone cells in glo1-1endosperms, we introgressed pVP1-GUS plants (J.F.G.-M., unpublished data) in the glo1-1 background, thus enabling GUS detection of aleurone cells. We examined 10 dap wild-type kernels and found GUS staining patterns confined to aleurone cells (Fig. 6A)However, in glo1-1 endosperms, non-uniform expression in the aleurone was occasionally observed (data not shown) as was the unexpected presence of GUS in few cells of the central SE region(Fig. 6B). More frequently, GUS staining was observed in the BETL of severe glo1-1 endosperms(Fig. 6C). An additional unique feature of the glo1-1 phenotype was the appearance of discrete cysts encapsulated by extremely thick cell walls proximally located to peripheral regions of the endosperm. By using the pVP1-GUS reporter, we found that the layer of cells investing these ectopic structures also showed GUS staining(Fig. 6D), while cells of the interior of these structures contained well-defined starch grains and did not stain for GUS (Fig. 6E). Thus,our current data indicate that cells with many features of the aleurone are ectopically formed throughout the glo1-1 endosperm.

A large number of maize seed mutants have been isolated(Neuffer and Sheridan, 1980; Scanlon et al., 1994; Sheridan and Neuffer, 1980),although only few have been comprehensively characterised to date(Becraft et al., 2002; Consonni et al., 2003; Fu et al., 2002; Lid et al., 2002; Shen et al., 2003). We describe here the phenotypic characterisation of the glo1-1 mutation,which arose in an active transposon population. The distinctive mutant phenotype not only points to the nature of the gene affected, but when considered in the perspective of current knowledge, also provides further insight into the factors governing endosperm cell fate and tissue differentiation.

Using a range of microscopic techniques, in situ hybridisation and reporter gene assays, we have shown that early cellularisation events in the glo1-1 endosperm are abnormal, and, as a result, lead to alterations in cell fate and subsequent differentiation of endosperm tissues. Examination of mutant embryos revealed severe cellular disorganisation with few recognisable tissues formed. Taken together, these findings suggest that the GLO1 product is essential for the regulation of cellular morphogenesis in maize embryo and endosperm tissues.

It is possible that the variation observed in endosperm and embryo phenotypes among homozygous glo1-1 mutants results from varying gene dosage. Through the use of a B-A translocation affecting the long arm of chromosome 1, additional copies of the glo1-1 wild-type allele were introduced in either the embryo or endosperm independently, but this failed to rescue the mutant phenotype, suggesting that variations observed in glo1-1 mutant phenotypes are not attributable to differences in gene dosage. These studies also demonstrated that the glo1-1 mutation affects development of the embryo and endosperm independently. Moreover, we showed that wild-type (hyperploid) embryos were unable to develop in the presence of defective (hypoploid) endosperms, suggesting that under the glo1-1 mutant condition, development of the embryo is strongly influenced by the endosperm.

Interestingly, we found glo1-1 to be allelic to the EMS-induced mutant gm*-N1303 (data not shown), originally described as a germless mutant by Neuffer and Sheridan(Neuffer and Sheridan, 1980). The variation in seed phenotype observed between the two alleles (data not shown) may be attributable to differences in genetic background (e.g. Gethi et al., 2002; Vollbrecht et al., 2000): gm*-N1303 was described in a vigorous hybrid background, whereas the glo1-1 allele is in a standard inbred line.

Early endosperm development in many angiosperms is characterised by nuclear migration and suppression of phragmoplast formation(Brown et al., 1999; Brown et al., 1994; Olsen, 2001), requiring tight coordination between the cell cycle and cytoskeletal organisation. Phenotypic analysis of the glo1-1 mutant suggests that glo1-1 may affect nuclear division, perhaps through a deregulation, or partial loss of control of the cell cycle. Aberrant cellular development is often localised to small patches within glo1-1 endosperms and may occur at different developmental stages, leading to the observed range of glo1-1 mutant phenotypes. Thus, early proliferation of nuclei in the syncytium would induce a severe kernel phenotype, whereas random proliferation in the cellular endosperm would result in minimal disruption to tissue organisation. Cell wall aberrations and other defects typically associated with cytokinesis (see Söllner et al., 2002)were apparent in glo1-1 kernels. These aberrations have been reported in several seed mutants, many of which have led to early developmental arrest of embryos (Consonni et al.,2003; Söllner et al.,2002). This link between defective cell division and suppression of embryo morphogenesis may account for the lack of pattern formation in glo1-1 embryos.

The GLO1 gene product therefore seems necessary for both nuclear division and cytokinesis in the developing seed. Evidence from Arabidopsis is accumulating that the processes of mitosis and cytokinesis are intimately interrelated (Dickinson, 2003; Mayer et al., 1999; Sørensen et al., 2002),and it remains possible that GLO1 is involved only in an aspect of nuclear division which, when disrupted, causes secondary or downstream effects on cytokinesis. This is also true for TITAN3, which encodes an SMC2 condensin (Liu et al., 2002). Mutant endosperms lacking the functional TITAN3 protein develop giant nuclei in the syncytial cytoplasm, which consequently results in defective cellularisation (Liu and Meinke,1998). Clearly, the role of GLO1 remains to be determined and efforts are now under way to clone the glo1 gene using a closely linked Ac element in regional mutagenesis(Brutnell, 2002; Singh et al., 2003).

Disruptions to the basal endosperm region are witnessed from an early stage in glo1-1 mutants, i.e. during syncytial development, when the ESR and BETL are thought to be specified(Becraft, 2001; Olsen, 2001).

Our data showed reduced levels of ZmESR gene expression in glo1-1 kernels, often restricted to small regions of the ESR. This may be a consequence of either aberrant development of other endosperm tissues, impacting indirectly on the ESR, or defective development of the embryo caused by the glo1-1 mutation. It has been proposed that the ZmESR gene products are involved in cross-talk between embryo and endosperm (Opsahl-Ferstad et al.,1997), as ZmESR expression is absent in spontaneously occurring embryo-less mutants(Opsahl-Ferstad et al., 1997). Our observations are consistent with the interpretation that correct establishment of the ESR relies upon successful development of the embryo– which is of course highly defective in glo1-1 kernels.

In glo1-1 endosperms, the BETL tissue was often disrupted to various degrees, and reduced levels of two BETL-specific transcripts were recorded – defects that are likely to result from abnormal syncytial development caused by the glo1-1 mutation. As BETL identity is held to be acquired via maternally derived diffusible signal(s) or morphogen gradient(s) present in the syncytium(Becker et al., 1999), it follows that the irregularities in the syncytium induced by the glo1-1 mutation may affect the ability of specific basal nuclei to perceive or act upon these positional cues. Our data therefore suggest that BETL specification is a fixed process, which occurs only during a short period in syncytial development. Nuclei that fail to perceive this information(either through the action of glo1-1 or otherwise) are thus unable to differentiate as BETL cell types. Therefore, if the acquisition of BETL cell fate were both irreversible and to occur within a narrow developmental window,then `BETL identity' must be passed on in a lineage-dependent fashion (see Fig. 7 for proposed model),which, although is rare for plants(Scheres, 2001), is common in animal systems such as Drosophila – where positional cues in the blastoderm lead to lineage-dependent differentiation(Lawrence and Struhl, 1996). In support of this hypothesis, we found that the BETL was frequently patchy in glo1-1 endosperms. Additionally, it has been reported that following interploidy crosses, endosperms (with a paternal excess genomic contribution)often contain tiny interspersed clusters of BETL cells(Gutierrez-Marcos et al.,2003). As BETL cells are known to undergo a finite number of divisions (Slocombe et al.,1999), these clusters may have originated from a single nucleus that had initially perceived the developmental cues to assume BETL cell fate. The fact that cells occupying basal positioning in the glo1-1endosperm are not able to adopt BETL cell identity, together with evidence that BETL cells are observed in apical regions of endosperms following interploidy crosses (Gutierrez-Marcos et al., 2003), would imply that basal positioning is not strictly necessary for BETL cell fate. This contradicts previous models for BETL cell development, where, according to Becker et al.(Becker et al., 1999),induction of transfer cells is thought to occur either via a signalling mechanism or through mRNA transport present in the basal endosperm region.

The plane of cell division is strictly controlled in aleurone cells by preprophase bands (PPB), defining the anticlinal plane of division, thus permitting lateral expansion of the aleurone layer (reviewed by Olsen, 2001). Data presented here revealed perturbations in the organisation of aleurone cells, which were mainly confined to the abgerminal face of glo1-1 kernels – a pattern prevalent in many other mutants affecting early kernel development(Becraft and Asuncion-Crabb,2000). These defects included the absence of aleurone cells and/or the presence of additional aleurone cell layers in regions of the glo1-1 endosperm. Similarly, the extra cell layers1(xcl1) mutation causes aberrant periclinal divisions in the maize kernel, which results in the formation of extra aleurone layers(Kessler et al., 2002).

Little information exists of how aleurone fate is determined; however, the simplest model holds that aleurone cell fate is acquired through a steep ligand gradient away from the maternal cell wall(Olsen et al., 1998), which is perceived at the endosperm surface by the CRINKLY4 receptor kinase(Becraft et al., 1996). Based on microscopic observations and GUS marker gene expression analysis, our data unexpectedly revealed that aleurone-like cells have the ability to form from within the glo1-1 endosperm. Thus, aleurone formation cannot exclusively rely upon maternally derived signals. Instead, we suggest that, at least in glo1-1 endosperms, aleurone identity might be acquired internally, perhaps as a result of signals from within the SE itself. Certainly, studies have demonstrated that the aleurone and SE cells can develop from common progenitors and that neither cell type fates are terminally determined, as aleurone formation is dependent on the constant input of positional cues (Becraft and Asuncion-Crabb, 2000; Becraft et al., 2002; Lid et al.,2002). This developmental plasticity is further supported by our observations: anthocyanin-like pigment (data not shown) and patchy reporter-GUS expression patterns observed in glo1-1 basal endosperms suggest that the undifferentiated cells often observed within the BETL during early development are capable of assuming aleurone identity at later stages.

glo1-1 endosperms thus contain two classes of aleurone-like cells,the first are peripherally located and bounded on their `outward face' by a thickened wall, as in wild-type kernels (data not shown). The second class of cells, despite their location within the SE, are also bounded on one or more faces by thickened cell walls. Assuming differences in permeability between these thickened walls and others of the endosperm, it is possible that short-range cues that trigger aleurone formation may accumulate in these cells. This interpretation is also supported by emerging evidence that cell fate in plants is specified by short-range cell-cell interactions, mediated through signal transduction cascades, as occurs in Drosophila and C. elegans (Irish and Jenik,2001; Marx,1996).

We are indebted to Ian Hempshall and co-workers at Syngenta (Jealott's Hill, UK) for invaluable support and plant growth facilities. Cledwyn Merriman(Wellcome Trust Institute for Human Research, UK) and Judith Sheldon (Oxford Brookes University, UK) are kindly acknowledged for their help with resin sectioning and electron microscopy, as well as John Baker for image processing. We thank the Maize Genetics Cooperation Stock Center (Urbana, IL,USA) for providing the valuable seed stocks and the Maize Mapping Project Team(University of Missouri Columbia, MO, USA) for mapping information. We also thank Dr P. Becraft (Iowa State University, USA) for supplying the promoter-VP1-GUS construct and Dr P. Perez (Biogemma, France) for generation of transgenic plants. The pBET1-GUS seeds were a kind gift from Dr R. Thompson(Max-Planck-Institut für Züchtungsforschung, Germany) and the ZmESR2 plasmid was generously provided by Dr P. Rogowsky (RCAP ENS, Lyon, France). This work was funded by a BBSRC-CASE award and Syngenta.