ABSTRACT
Fertilisation in maize (Zea mays), in common with most angiosperms, involves two fusion events: one of the two sperm nuclei unites with the egg cell nucleus, while the other sperm nucleus fuses with the two central cell nuclei giving rise to the triploid endosperm. Since deviation from this nuclear ratio (2:1 maternal/paternal) in the endosperm can result in abortion, it has been suggested that the genomes of the sperm and/or central cell are differentially imprinted during sexual development. By crossing a normal diploid maize line as female with its autotetraploid counterpart, an unbalanced genomic ratio (2:2 maternal/paternal) is created in the endosperm which often results in the eventual abortion of the tissue. Detailed structural comparison of these aberrant endosperms with normal endosperms reveals that the formation of the transfer cell layer, a tissue formed some 8 days after pollination and responsible for the transport of nutrients into the endosperm, is almost completely suppressed under conditions of paternal genomic excess. The first structural analysis of the development of this tissue in normal and aberrant endosperms is reported, and the implications of regulating the formation of such a tissue by gametically imprinted genes are discussed in the light of current theories on the consequences of genomic imbalance on early embryonic development.
INTRODUCTION
In mammals it is now well established that the maternal and paternal genomes transmitted to the zygote during fertilisation differ in gene expression during embryonic development, the expression of a number of sequences being determined by their parental origin. This parent-specific pattern of gene expression is held to be established during gametogenesis by means of an epigenetic, reversible imprint (Reik et al., 1987; Sapienza et al., 1987). To be effective, this imprint must be capable of surviving mitosis, yet be removed during gametogenesis to permit reimprinting during passage through the germline of the opposite sex. How this gametic or genomic imprinting (Barlow, 1994) is achieved is not fully understood, but recent evidence from work on mice suggests that cytosine methylation may play a major role in this process (Reik et al., 1987; Sapienza et al., 1987; Stöger et al., 1993; Feil et al., 1994; for reviews see Solter, 1988; Razin and Cedar, 1991; Riggs and Pfeiffer, 1992; Surani, 1993), although other hypotheses have been proposed (Singh, 1994).
In plants, gametic imprinting has only convincingly been demonstrated to influence development of the endosperm, a tissue that nourishes the embryo (Kermicle, 1970; Lin, 1982, 1984; Kermicle and Alleman, 1990; Chaudhuri and Messing, 1994). Fertilisation in flowering plants involves the release of two sperm nuclei into the embryo sac; one unites with the egg to form the zygote, while the other fuses with the two haploid nuclei of the central cell to form the triploid endosperm. The endosperm thus possesses a maternal/paternal genomic ratio of 2:1 (2m:1p) and it has been demonstrated (Lin, 1984) that this ratio can be crucial for the successful development of this tissue, any divergence resulting in abortion. Using an indeterminate gametophyte (ig) mutant as the female parent, Lin (1984) was able to generate endosperms with ploidy levels ranging from diploid (2n) to nanoploid (9n) and, in all cases studied, abnormal endosperms resulted unless a maternal: paternal genome ratio of 2:1 was maintained. Further, Lin (1984) was able to demonstrate unambiguously that these effects resulted from interactions between nuclear genomes and not from the influence of organellar genomes derived from the sperm and/or central cell.
The molecular and cellular consequences of these genomic interactions are thus highly significant (for reviews see Birchler, 1993; Matzke and Matzke, 1993). Most importantly, molecular mechanisms must exist by which both the copy number and parental origin of certain genes are sensed and, in circumstances where the incorrect balance is detected, development of the early endosperm is aborted. The rapid degeneration of these unbalanced endosperms makes analysis of the molecular events involved in gametic imprinting, and its consequences, very difficult. However, Lin (1984) reported a cross leading to an abortive endosperm which involved fertilising a diploid female plant with pollen from its autotetraploid to produce tetraploid (2m:2p) endosperms. Interestingly, these endosperms appeared to undergo normal development until 10 to 12 days after pollination (DAP), providing an opportunity for the molecular and cellular analysis of the tissue prior to and during abortion. The general cytology of aberrant endosperms of this type has been investigated (Cooper, 1951), but we report here a more comprehensive structural analysis of early development of normal and aberrant endosperms. In particular, we reveal the primary effect of paternal excess to be on the early differentiation of the transfer tissue that links the base of the kernel with the embryo and endosperm. Molecular analysis of these events will form the subject of a later publication.
MATERIALS AND METHODS
Plant material
Two lines of maize (Zea mays), a diploid cv. Wisconsin 23 (W23), kindly supplied by Dr Jane Langdale, and its autotetraploid cv. N107B, a gift from the Maize Seed Stock Centre (University of Illinois, Urbana-Champaign, USA), were grown at Zeneca Seeds (Jealott’s Hill, UK) under glass with supplementary lighting and heating. Two crosses were performed: W23 × W23 and W23 × N107B. The former provided control kernels while the latter resulted in a tetraploid endosperm formed by the fusion of the two haploid central cell nuclei with a diploid sperm nucleus from the tetraploid N107B. Prior to silk emergence, ears were protected from uncontrolled pollinations with shoot bags. As silks emerged, they were trimmed back to the tip of the ear until sufficient silks were exposed to ensure one application of pollen would produce a high level of fertilisation. Once pollinated, the ears were protected with tassel bags until harvested. All pollinations were performed in the early morning.
Preparation of material for light and electron microscopy
Dissection
Material from both crosses was collected for study at 0, 2, 3, 4, 6, 8 and 10 DAP. Following removal of the glumes from the kernel, embryo sacs or endosperm/embryo tissue at 0 and 2 DAP were prepared for fixation by either repeatedly piercing the kernel pericarp with a dissecting needle or removing a small section of the upper half of the kernel epidermis. Endosperm/embryos from 3 and 4 DAP were prepared by removing a half to two thirds of the nucellar tissue with a transverse cut being made close to the kernel base. Endosperm/embryo tissue at 6, 8 and 10 DAP was either completely excised from the nucellus or left seated in a cup of nucellar tissue to maintain the integrity of structures linking the base of the endosperm/embryo assembly to the nucellus.
Fixation and embedding
Tissue fixation was carried out for 4 hours at room temperature in Karnovsky’s fixative (4% paraformaldehyde, 3% glutaraldehyde buffered in 0.03 M phosphate, pH 7.0). After rinsing in buffer, the tissues were postfixed in 1% osmium tetroxide for 1–2 hours at room temperature before dehydration through a graded series of acetone:water mixes (30%, 50%, 75%, 90%, 95%, 100%; each for 1 hour) and embedding in medium grade Epon (TAAB) catalysed with 1% accelerator (TAAB). Prior to embedding in Epon, the tissue was supported in 1% low melting point agarose (Flowgen Instruments Ltd) buffered with 0.03 M phosphate, for protection against mechanical damage during resin impregnation.
Sectioning
Semithin sections (7 μm) were cut on a 5000 MT Sorvall ultramicrotome, stained with 1% Toluidine blue in 1% borax and/or Periodic acid/Schiff’s reagent (PAS) (Feder and O’Brien, 1968) and photographed with a Zeiss Axiophot light microscope on Pan F film. Following light photography, the semithin sections were re-embedded for ultrathin sectioning at 90 nm. Ultrathin sections were stained with uranyl acetate and lead citrate and examined using a JEOL 2000EX transmission electron microscope operating at 80 kV.
RESULTS
Pollinations with pollen from tetraploid parents
Fertilisation with pollen from tetraploid parents proved poor when compared with pollen from diploid parents. While the latter regularly produced fully fertilised ears, it was not uncommon to find ears of the former completely devoid of developing kernels or containing as few as 5–10 kernels. Only occasionally would more than 50% of ovules be successfully fertilised.
Pollinations that were successful gave rise to two phenotypes: abundant but shrivelled kernels, or a low number of small but well-formed, ‘normal’ kernels (Fig. 1). Although an insufficient number of ears were allowed to develop to maturity to permit a statistical analysis of these phenotypes, Randolf’s earlier work (1935) reported that less than 0.5% of the successful 2n × 4n pollinations gave rise to ‘normal’ kernels. The earlier work by Randolf also reported that approximately 10% of the well-formed ‘normal’ kernels and very few of the aborted kernels produced viable seedlings.
Kernels of Zea mays following pollination by diploid and tetraploid plants. Top row: normal kernels resulting from diploid × diploid pollination. Middle row: smaller kernels of near-normal appearance resulting from diploid × tetraploid pollination. Bottom row: shrunken kernels resulting from diploid × tetraploid pollination.
Kernels of Zea mays following pollination by diploid and tetraploid plants. Top row: normal kernels resulting from diploid × diploid pollination. Middle row: smaller kernels of near-normal appearance resulting from diploid × tetraploid pollination. Bottom row: shrunken kernels resulting from diploid × tetraploid pollination.
Development of triploid and tetraploid endosperms 0-6 DAP
‘Fertilisation’ of the binucleate central cell by a haploid sperm nucleus initiates endosperm development following a pathway first described by Cooper (1951) (see also Fig. 2). By c.6 DAP, the tissue has become fully cellular and, through repeated divisions, has increased in volume approximately 3 fold (Kies-selbach, 1980) to assume a characteristic cone-shaped structure. Under the light microscope little difference can be observed at this stage between the triploid and tetraploid endosperms, apart from a tendency for the upper third of the tetraploid endosperm to narrow more sharply.
Summary of the development of Zea mays endosperms formed following pollination of a diploid seed parent by a diploid pollen source (2n × 2n), and by a tetraploid source (2n × 4n). Structural development of the two types of endosperm appears identical until c.5 DAP.
During the development of the outermost basal cells of the triploid endosperms, fine wall ingrowths appear (Fig. 3A,B), signalling the first stages in transfer cell development (Kies-selbach and Walker, 1952; Kiesselbach, 1980). The developing ingrowths, which are principally restricted to the outer and tangential cellular faces, are frequently observed in association with cytoplasmic inclusions containing an electron-lucent matrix similar in appearance to that in the developing wall (Fig. 3A,B). The cytoplasm of these cells appears very active, being particularly rich in small, spherical mitochondria. Importantly, little or no development of the ingrowths takes place in the outermost basal cells of the tetraploid endosperms. In circumstances where slight development does take place, apparently random and malformed aggregations of wall material are laid down (Fig. 4). Here, the wall matrix itself is different, being more electron-opaque (Fig. 4). The cytoplasm surrounding these early wall inclusions is disorganised with few mitochondria. Additionally, while triploid endosperms appear firmly anchored at their bases to the nucellar tissue, tetraploid endosperms are readily detached.
(A,B) Electron micrographs of grazing sections of the outer surface of the transfer layer, 6 DAP of a diploid plant by diploid pollen. Ingrowths (I) are beginning to develop, frequently associated with cytoplasmic vesicles (V). The cells are interconnected by plasmodesmata (arrows). Profiles suggestive of vesicular contribution to the ingrowths are shown in A. Scale bars, (A) 500 nm; (B) 1 μm.
(A,B) Electron micrographs of grazing sections of the outer surface of the transfer layer, 6 DAP of a diploid plant by diploid pollen. Ingrowths (I) are beginning to develop, frequently associated with cytoplasmic vesicles (V). The cells are interconnected by plasmodesmata (arrows). Profiles suggestive of vesicular contribution to the ingrowths are shown in A. Scale bars, (A) 500 nm; (B) 1 μm.
Electron micrograph of outer face of transfer cells in a tetraploid endosperm, 6 DAP. No organised ingrowths have formed. Instead the cytoplasm appears to be discharging irregular masses (M) into the ‘nucellar’ surface of the wall (W). Scale bar, 500 nm.
Development of triploid and tetraploid endosperms 6-10 DAP
By 8 DAP the outermost basal cells of the triploid endosperms have become elongated, and their wall ingrowths readily observable under the light microscope (Fig. 5). The invaginations appear tubular in section and are about 0.5-1.0 μm in diameter; there are c. 1-3 per 10 μm2 of the cell surface and they extend some 5-10 μm into the cytoplasm (Figs 6, 7). They continue to be associated with inclusions containing a wall-like matrix and, again, the surrounding cytoplasm contains numerous mitochondria. By contrast, corresponding cells in the tetraploid endosperms remain devoid of maturing wall ingrowths. The cytoplasm of these cells appears inactive and frequently becomes highly vacuolate; in later stages cytoplasmic bridges may be observed (Fig. 8A). In addition, the narrowing of the apical region of the tetraploid endosperms has, by 8 DAP, become more pronounced when compared with triploid endosperms. Importantly, the inner cells of the tetraploid endosperms begin to degenerate at this stage. Originating in a mass of cells at the centre of the young endosperm, changes normally associated with necrosis take place; these include disintegration of the tonoplast, intense vesiculation, the accumulation of electron-opaque inclusions and, eventually, the condensation of nuclear material. Despite this apparent cellular degeneration, no clear difference may be detected either in endosperm length (approximately 2.7 mm) or width (approximately 1.5 mm) between triploid and tetraploid endosperms at 8 DAP.
Light micrograph of developing ingrowths of the outer walls of transfer tissue cells, 8 DAP by a diploid pollen source. PAS staining reveals the ingrowths to contain high levels of carbohydrate. Scale bar, 15 μm.
Electron micrograph of ingrowths, more highly developed than those shown in Fig. 3, in endosperm 8 DAP by a diploid plant. Note the high density of mitochondria (arrows) in the cytoplasm investing these structures. Scale bar,1 μm.
Electron micrograph of ingrowths, more highly developed than those shown in Fig. 3, in endosperm 8 DAP by a diploid plant. Note the high density of mitochondria (arrows) in the cytoplasm investing these structures. Scale bar,1 μm.
Electron micrograph of glancing section of the outer surface of the transfer tissue, as in Fig. 6, but from material 10 DAP. Extensive ingrowths are present in all cells, and in some regions the cytoplasm between the ingrowths (arrows) is highly reduced. Scale bar, 1 μm.
Electron micrograph of glancing section of the outer surface of the transfer tissue, as in Fig. 6, but from material 10 DAP. Extensive ingrowths are present in all cells, and in some regions the cytoplasm between the ingrowths (arrows) is highly reduced. Scale bar, 1 μm.
(A) Electron micrograph of outer face of transfer cells in a tetraploid endosperm, 10 DAP. No organised ingrowths have formed and there is little development of the outer (nucellar) wall (W). Cells appear to be joined via large cytoplasmic bridges (arrows). Scale bar, 5 μm. (B) As A but at a higher magnification. The outer (nucellar) wall appears to contain compacted masses (arrows) of electron-opaque material. Scale bar, 1 μm.
(A) Electron micrograph of outer face of transfer cells in a tetraploid endosperm, 10 DAP. No organised ingrowths have formed and there is little development of the outer (nucellar) wall (W). Cells appear to be joined via large cytoplasmic bridges (arrows). Scale bar, 5 μm. (B) As A but at a higher magnification. The outer (nucellar) wall appears to contain compacted masses (arrows) of electron-opaque material. Scale bar, 1 μm.
Differentiation of the aleurone layer in the triploid tissues is observed between 6 and 10 DAP, with the accumulation of spherosomes and protein bodies in the outer layers of the endosperm (Fig. 9). While these cell layers contain normal cytoplasm in tetraploid endosperms, protein bodies and spherosomes are absent (Fig. 10).
Electron micrograph of surface cells of a normal endosperm, 10 DAP, in the region subsequently occupied by the aleurone layer. Large numbers of spherosomes (S) are present. Scale bar, 2 μm.
As Fig. 9, but material from a tetraploid endosperm. No spherosomes are visible. Scale bar, 1 μm.
As Fig. 9, but material from a tetraploid endosperm. No spherosomes are visible. Scale bar, 1 μm.
Wall ingrowths of triploid endosperms cover a large pro-portion of the placentochalazal region by 10 DAP (Fig. 7), and extend to the junctions between cells of the endosperm up to three cell layers beneath the superficial cell layer. The matrix composing these wall ingrowths is continuous with the walls of the nucellar cells and thus presumably contributes to the higher mechanical strength of this region when compared with kernels containing tetraploid endosperms. No ingrowths are present at the nucellar face of tetraploid endosperms (Fig. 8B).
In tetraploid endosperms 10 DAP, the boundary of the region of cellular dissolution approaches the periphery of the tissue, with no further development of the transfer cell layer. By contrast, triploid endosperms at 10 DAP are plump, fully cellular tissues, and at this stage the first evidence of a size difference between the two types of endosperm can be distinguished. On average, 10 DAP triploid endosperms are twice as long and twice as wide as tetraploid endosperms. Further, tetraploid endosperms assume a deflated appearance with their surfaces becoming increasingly infolded.
Starch development in triploid and tetraploid endosperms 0-10 DAP
Apart from differences in the development of the aleurone layer described above, surprising differences also occur in the accumulation of starch in the interior of the endosperms. By contrast with events at the tissue surface, PAS staining reveals that tetraploid endosperms accumulate starch at an earlier stage than triploid tissues, small starch grains becoming visible in tetraploid cells as early as 8 DAP, whereas the first appearance of starch occurs in triploid tissues at 10 DAP (data not shown).
DISCUSSION
By crossing a female diploid with a male autotetraploid, we have generated tetraploid endosperms with a genomic ratio of 2:2 and like Lin (1984) found abortion to occur during development. However, by contrast with reports of complete abortion of tetraploid endosperms by maturity (Lin, 1984), a small number of the endosperms produced from our 2n × 4n crosses were normal, plump tissues, albeit slightly smaller than the wild type. Similar observations have been previously reported by Randolf (1935) and Cooper (1951), but not in the context of genomic balance. Such ‘normal’ kernels are unlikely to be the result of contamination by pollen from a diploid plant or a reversion of tetraploid to diploid pollen as the triploid status of plants derived from these apparently normal kernels studied by Randolf (1935) was confirmed cytologically. It is thus important to determine whether the dramatic developmental changes reported relate to the near-normal or to the very small, shrunken kernels. However, over the course of this study, a sufficiently large number of kernels from each developmental stage was examined to indicate strongly that the absence of transfer tissue is associated with the development of the very small, shrunken kernels. We believe the larger kernels do possess transfer cells, although this has not been checked during our studies.
Following fusion of the two central cell nuclei with a sperm nucleus approximately 24 hours post pollination, both triploid and tetraploid endosperms commence parallel developmental pathways until approximately 6 DAP when, in the triploid endosperm, early signs of transfer cell development are first observed. In the tetraploid endosperms, this development fails to be fully initiated. The wall ingrowths, which first identify the transfer cells, are contiguous with the walls of cells forming the placentochalazal region and, early in their differentiation, appear to assume an anchorage role, holding the developing endosperm to the base of the nucellus. Although the first signs of transfer cell differentiation may sometimes occur in the tetraploid endosperms, development clearly does not progress to the stage where anchorage occurs; for this reason these endosperms are easily detachable by microdissection. By 8 DAP in triploid endosperms, the wall ingrowths of the transfer cells are readily observed under the light microscope and by 10 DAP the majority of the endosperm cells adjacent to the placentochalazal region possess these ingrowths, as do cells up to three layers beneath the principal transfer cell layer.
Since Harz first reported this presumably nutrient-transporting tissue for maize in 1885 (cited by Kiesselbach, 1980), this layer has been described in some detail by Harrington and Crocker (1923) in Johnson grass (Holcus halepensis L.), in Saccharum by Artschwager and co-workers (1929), in Coix by Weatherwax (1930), and in maize by Kiesselbach and Walker (1952), Kiesselbach (1980) and Felker and Shannon (1980), who first recognised the presence of transfer cells in this tissue, and by Schief and co-workers (1984). Transfer cells occur in all major taxa, and although normally associated with the transfer of nutrients from sporophyte to gametophyte, are found in many other situations (Gunning and Pate, 1969), occurring in a diversity of tissues and being involved in the transfer of a wide range of solutes. Even within the reproductive structures of the Gramineae, this tissue is not restricted to the endosperm, for studies of the wheat caryopsis (Smart and O’Brien, 1983) revealed the development, at 15 DAP, of a layer of cells with wall ingrowths within the nucellar epidermal cells at the base of the kernel. This interesting observation suggests that transfer cells in this region are capable of performing both export and import functions.
Interestingly, Lyznik and co-workers (1982) have shown that translocation of amino acids into the endosperm, presumably via the transfer cells, takes place against a concentration gradient, created predominantly by high levels of alanine present in the tissue. The energy requirements for the transport of amino acids and other nutrients, combined with the absence of an apoplastic pathway between the sporophytic nucellus and gametophytic endosperm via plasmodesmata, may well explain the very large numbers of mitochondria in the transfer tissue. Other physiological studies (Shannon, 1972; Shannon and Dougherty, 1972; Felker and Shannon, 1980) have shown that during nutrient uptake by the endosperm, sucrose is unloaded from the pedicel parenchyma into the apoplast of the placenta-chalazal tissue. It is then either hydrolysed by acid invertase(s) into fructose and glucose, prior to uptake by the transfer cell layer into the endosperm symplast (Felker and Shannon, 1980; Doehlert and Felker, 1987), or transferred to the basal cells of the endosperm as unhydrolysed sucrose (Porter et al., 1985, 1987) to be hydrolysed by invertases in the endosperm.
The apparent lack of a nutrient transport tissue in tetraploid endosperms is consistent with the observations by Cooper (1951) of ‘undifferentiated basal cells’ in tetraploid and pentaploid endosperm resulting from 2n × 4n and 4n × 2n crosses. Although Cooper (1951) states that the ‘basal layer of cells fails to differentiate in a normal manner’, it is never explicitly stated that an absorbing tissue fails to form. Further, in his summary, failure of the basal layer is described only for the tetraploid endosperms, with no mention of the pentaploid endosperms from the 4n × 2n crosses. Since, by 16 DAP, the pentaploid endosperms are packed with starch (growth ceasing only at later stages) aberrant development is unlikely to result from the lack of transfer tissue.
After uptake by the transfer layer cells, nutrients are translocated into the rapidly expanding endosperm by a core of elongated cells extending from the transfer layer through the central region of the endosperm and into the apex (Weatherwax, 1930; Lampe, 1931; Brink and Cooper, 1947; Kiesselbach, 1980). No evidence of this conducting tissue is seen in the tetraploid endosperms at 10 DAP. Rather than containing specialised conducting tissue, this region is occupied by apparently necrotic cellular debris.
Our observations suggest that the departure from the normal 2m:1p genomic ratio in the maize endosperm results in the almost complete suppression of the development of the transfer cell layer. Other abnormalities found, such as cellular necrosis and retarded aleurone development, may merely be a consequence of nutrient starvation resulting from the absence of the transfer tissue. The precocious appearance of starch in the tetraploid endosperms is, however, less easily explained.
It is possible that suppression of transfer cell development results indirectly from changes in maternal tissue, which could restrict the uptake of nutrients essential for this complex cellular differentiation. Such an effect has been reported by Miller and Chourey (1992). Using the miniature 1 mutant first described by Lowe and Nelson (1946), they showed that, in its homozygous condition (mnmn), the mutation produces an invertase deficiency in the endosperm resulting in the degeneration of the placentochalazal cells and their eventual separation from the endosperm basal cells. The degeneration of these cells in turn causes a loss of invertase in this region. A combination of the structural separation, and a lack of invertase in both the basal cells and the placento-chalazal region, effectively prevents the uptake and processing of nutrients by the endosperm, causing a shrunken phenotype. There is, however, very little evidence that similar events occur in the tetraploid endosperms for, while we did not study the maternal tissue of the tetraploid-containing kernels beyond 6 DAP, neither Cooper’s (1951) study of tetraploid endosperms, nor Brink and Cooper’s (1947) report of the de17 mutant – which produces a phenotype similar to that of the tetraploid endosperms – described any disruption of maternal tissue.
Animal studies reveal that androgenetic mouse embryos (0m:2p) develop poorly, yet have well developed extraembryonic membranes (Barton et al., 1984); paternal excess in humans (1m:2p) leads similarly to under-developed foetuses with large placentas (Hall, 1990). Gynogenetic and parthenogenetic mouse embryos (2m:0p), and maternal excess in human foetuses, result in a reverse situation - with smaller placentas than normal. These studies have led Haig and Westoby (1989) to propose an elegant theory of parental ‘tug-of-war’ to explain the relationship between gametic imprinting and embryo development. The theory holds that selection favours the imprinting of certain genes from a male so that they ensure maximum investment from the female towards his offspring during development, while the female genes are imprinted to limit her investment in particular offspring. Although the transfer tissue of maize may reasonably be considered analogous to the placenta of mammals, our observations suggest that this theory fails to hold for endosperm imprinting in plants for, if Cooper’s (1951) results are taken at face value, both maternal and paternal excess give rise to a deficient ‘placenta’. This would indicate that it is not paternal excess per se which causes this deficiency but rather any deviation from the 2m:1p ratio. On the other hand, if penta-ploid endosperms (Cooper, 1951) do produce a transfer layer, as some evidence suggests, the situation conflicts directly with Haig and Westoby’s (1989) theory - for seemingly seed development is favoured despite maternal excess.
Whether paternal excess alone or genomic imbalance results in the phenotype observed, at least one of the genes involved in transfer cell development must be differentially imprinted. Since genomic imbalance can suppress the development of the complete transfer tissue, the imprinted genes are likely to be regulatory in function and act at an early developmental stage. On the basis of these assumptions, a simple model (see Fig. 11) for their operation can be proposed on the basis of the central cell nuclei but not the sperm nuclei carrying imprinted loci, with imprinting suppressing gene expression in the endosperm. Thus, a regulatory protein (A), transcribed from the paternal allele (A) and not from the imprinted maternal allele, complexes with a second regulatory protein (B) transcribed from both the maternal and paternal loci (B). Depending upon the activities of A and B, a ratio of A and B would be generated which results in an active complex. Upsetting the genomic balance would disturb the ratio of A and B and thus affect the activity of the AB complex resulting, eventually, in the arrest of development. Such concentration-dependent activation and repression has recently been shown to exist in Drosophila by Sauer and Jäckle (1993) for, at a certain concentration, the zinc finger-type transcription factor Kruppel (Kr) functions as a transcriptional activator in its uncomplexed form; however, at higher concentrations, homodimerisation of the Kr protein transforms it into a transcriptional repressor. Interestingly, the de17 mutation first described by Brink and Cooper (1947) may involve just such a regulatory gene responsible for the initiation of transfer tissue development. The phenotype of this mutant includes shrunken kernels and lack of development of the basal endosperm cells. Further, the description of the undifferentiated basal cells of this mutant is remarkably similar to our observations for the corresponding cells of the tetraploid endosperm. While this model was proposed to explain specific observations in maize endosperms, it is also applicable to other circumstances where ‘imbalance’ in the number of endosperm nuclei affects development, such as in the Solanaceae (Hawkes and Jackson, 1992).
A simple model to explain the regulation of endosperm development through gametic imprinting. In the normal triploid endosperm, genes A and B are transcribed from the paternally donated genome to produce gene products A and B. The expression of A is suppressed in the maternally donated genomes through gametic imprinting, and thus only the maternal B gene product is synthesized. It is suggested that only a combination of A and B polypeptides, in the correct ratio, can activate development.
A simple model to explain the regulation of endosperm development through gametic imprinting. In the normal triploid endosperm, genes A and B are transcribed from the paternally donated genome to produce gene products A and B. The expression of A is suppressed in the maternally donated genomes through gametic imprinting, and thus only the maternal B gene product is synthesized. It is suggested that only a combination of A and B polypeptides, in the correct ratio, can activate development.
There are important reasons why mechanisms have evolved to preserve the genomic ratio in the endosperm. Clearly the system will prevent the spontaneous development of the endosperm in the absence of pollination, and equally any development resulting from the fusion of both sperm nuclei with the two polar nuclei. More importantly it must also restrict the success of agamospermous embryos - thus protecting sexual reproduction (Haig and Westoby, 1991). Interestingly, in outcrossing plants this mechanism would also act to suppress polyploidy, which is held to be a major source of genetic variation in the process of evolution (Stebbins 1950). However, these aims would also be achieved were the embryo itself to be imprinted, and the reason for the imprinting of the endosperm rather than the embryo probably lies in the reproductive strategies adopted by the angiosperms: while it is important to maintain a strong selection for sexual reproduction, it is advantageous for plants to retain the option of asexual reproduction should the need arise. By imprinting the endosperm, selection for sexual reproduction is maintained, and by keeping the embryo free from imprinting, the option still exists for asexual reproduction. Alternatively, it may simply be that the process of imprinting carries a cost in evolutionary terms; thus by imprinting the endosperm only, this cost is lessened yet sex still ensured.
ACKNOWLEDGEMENTS
This work was funded under the UK BBSRC Stem Cell Initiative. The authors wish to thank Kelly Pritchard, Stephen Moore and John Baker for technical assistance and Ann Rogers and Melissa Spielman for help in preparing the manuscript.