ABSTRACT
The ABI3 gene product of Arabidopsis is essential for correct completion of seed maturation. A severe mutant allele at this locus results in seed that remain green, fail to establish desiccation tolerance, and that germinate at a developmental stage when wild-type seed will not. Moreover, the formation of leaf primordia and xylem differentiation, both characteristic of germinating wildtype seedlings, can be observed in embryos harvested 12 days after flowering. Thus, mature abi3 embryos reach a developmental state that more closely resembles the character of a developing seedling rather than that of a dormant embryo. Previous studies of this gene have resulted in the suggestion that ABI3 is a transducer of abscisic acid induced seed dormancy. Our results demonstrate that the ABI3 gene product can be most accurately described as one of the major regulators of the transition between embryo maturation and early seedling development, rather than simply a transducer of the abscisic acid signal.
INTRODUCTION
Embryo development in dicotyledonous plants can be roughly divided into three phases. Early embryo development, which includes the zygotic and developmental stages during which embryo pattern formation occurs. This is followed by midstage embryo development during which reserve materials including storage proteins, lipids and carbohydrates are synthesized. Finally, during late embryo development, the embryo matures, at which time it becomes dormant, and desiccates. In this article we denote the developmental stages after embryo pattern formation as the maturation stage.
In most higher plants, germination will not occur while the seed is attached to the mother plant. However if embryos are excised from seeds during late embryo development, they are capable of ‘precocious’ germination; that is, they can grow autotrophically and develop into normal seedlings (Quatrano, 1986; Crouch, 1987). The inability of late stage embryos to germinate while connected to the maternal plant suggests that specific developmental control mechanisms act to inhibit germination, permitting the embryo to complete late functions that are necessary for the correct establishment of desiccation tolerance and dormancy in the mature seed.
Genetic analysis has been useful in identifying some of the regulators of late embryo development in higher plants (Neill et al., 1987; Koornneef et al., 1989; Nambara et al., 1992; Keith et al., 1994). Mutations in Arabidopsis thaliana that reduce either the ability to synthesize the plant hormone abscisic acid (ABA) or response to this hormone, also reduce seed dormancy (Koornneef et al., 1982, 1984). In Zea mays similar mutations cause the embryo to germinate viviparously (Neill et al., 1987). Two of these mutations, vp1 in maize and abi3 in Arabidopsis, are of particular interest because unlike other ABA-related mutations, they result in underdeveloped seed that are desiccation intolerant (Neill et al., 1987; Nambara et al., 1992; Ooms et al., 1993). Although the maturation stage is clearly altered in these mutants, embryo pattern formation is normal and immature seed, when imbibed, germinate to produce normal plantlets. Recent cloning and molecular comparisons of the VP1 and ABI3 genes have shown that they share regions of high sequence identity and may encode a seed-specific transcriptional activator (McCarty et al., 1991; Giraudat et al., 1992).
Phenotypic and molecular characterization of vp1 and abi3 mutants has been focused primarily on late embryo functions such as desiccation tolerance and dormancy; yet, these studies have not addressed why these mutants are so non-dormant (McCarty et al., 1991; Nambara et al., 1992; Ooms et al., 1993). Interestingly, physiological studies using an Arabidopsis strain carrying an ABA biosynthetic mutation and a weak abi3 allele showed germination-related proteins were synthesized during the maturation stage in the double mutant (Meurs et al., 1992). Although no visible germination was apparent, this result suggests highly non-dormant phenotypes may result because germinative development is induced while the embryo is still maturing.
Here we present developmental studies of molecular, physiological and morphological aspects of a severe allele of abi3. The alterations we have observed in differentiation of the shoot apex and vascular tissue in a severe abi3 allele indicate that these mutants are in a germinative stage of development well before seed desiccation begins. Our analyses also suggest that the ABI3 gene product (ABI3) is not only essential for entry into the maturation stage of embryo development, but also that mutant embryos continue directly into a germinative program while in the seed.
MATERIALS AND METHODS
Plant materials and growth conditions
The mutant abi3 allele, abi3-3, used in this study is one of the most severe alleles of this locus (Nambara et al., 1992, 1994). Plants were grown on rockwool moistened with a nutrient medium described previously (Fujiwara et al., 1992) under continuous fluorescent light at 22°C.
A transgenic strain, designated cab3-GUS, carrying the β-glucuronidase (GUS) gene under the control of the promoter of a gene encoding a chlorophyll a/b binding protein (Cab) has been described previously (Chory and Peto, 1990). The cab3-GUS strain was crossed to the abi3 mutant and a line homozygous for both abi3 and the transgene was established from the F2 generation.
Physiological analysis
Siliques were staged for phenotypic analysis of whole embryos and for dark germination experiments by tying colored thread around the peduncle on the day of flower opening and intact siliques were harvested on the indicated number of days after flowering (DAF). Seed sterilization for dark germination experiments was carried out either under a green safe light (Nambara et al., 1991) or within 3 minutes under normal lighting conditions. Dissected seeds were placed on nutrient agar plates containing 1% sucrose. Both mutant and wild-type seeds were scored for growth after 7 days of incubation in darkness at 22°C.
The water content of the seeds was determined as follows. Siliques were harvested 6-18 DAF and approximately 100 seeds were excised from siliques with a needle. These were weighed both within 8 minutes of harvesting and after being allowed to desiccate at room temperature for 3 days. The decrease in seed weight due to desiccation was not significant for at least 10 minutes after dissection for seeds older than 8 DAF. Exact numbers of seeds were counted after desiccation.
Northern hybridization
Siliques were staged as described for physiological analysis, and northern analysis was performed as previously described (Nambara et al., 1992; Naito et al., 1994). RNA equivalent to that obtained from 1.5 siliques was loaded into each lane. The Arabidopsis 12S SSP genes (CRA, CRB, and CRC; Pang et al., 1988) and the 2S SSP gene (AT2S1; Guerche et al., 1990) were used to probe seed reserve protein expression. The LEA76 gene probe encodes one of the late embryogenesis abundant protein (Lea) genes of Brassica napus (Harada et al., 1989). 32P-labeled DNA probes were prepared by random primer labeling (Feinberg and Vogelstein, 1984). We obtained similar results in triplicate experiments conducted with each probe. A representative result is presented for each probe.
Whole-mount preparations
Wild-type and abi3 seeds harvested 12 DAF were incubated for 1 hour or less on agar to soften seed coats and facilitate dissection of embryos from the seed. Germinating wild-type seedlings were imbibed on agar-solidified nutrient medium for the number of hours indicated then prepared for whole-mount observation. Dissected embryos or germinating seedlings were fixed in 70% ethanol for 2 days, then dehydrated in 1 change of 95% ethanol, followed by 2 changes of 100% ethanol. Each dehydration step lasted 2-4 hours. Dehydrated seedlings were cleared overnight in a saturated solution of chloral hydrate (60 g chloral hydrate in 15 ml H2O) then mounted on slides in a modified Hoyer’s solution (9 g gum arabic, 60 g chloral hydrate, 6 ml glycerol, in 15 ml H2O; Cunningham, 1972), and viewed under a dissecting microscope. A minimum of 15 embryos or seedlings at each stage of development were observed and representative embryos were photographed using a Reichert Polyvar compound photomicroscope with Nomarski (differential interference contrast) optics and Kodak TMAX 100 black-and-white film.
GUS assays
Immature seeds were dissected from siliques with a needle. Fifteen immature seeds or 30 seedlings that were grown under 18 hours light/6 hours dark condition were used for GUS assays as described using 4-methylumbelliferyl-β-D-glucuronide as a substrate (Jefferson et al., 1987).
RESULTS
Seed development in abi3 mutant
The early wild-type embryo passes through a number of morphological stages during which it is essentially colorless (Fig. 1A). At approximately 6 DAF it begins to green and remains green until 12-13 DAF (Fig. 1C), after which the embryo loses color and the seed coat begins to turn brown (Fig. 1E). The degreening of the embryo is also marked by a rapid decrease in water content (Fig. 2), which by 16 DAF yields a mature, white, quiescent embryo in a brown seed coat (Fig. 1G).
Seed development in wild-type (A,C,E and G) and abi3 mutants (B,D,F and H) of Arabidopsis. Embryos dissected from seeds of wildtype and abi3 harvested 4 (A,B), 8 (C,D), 12 (E,F) and 16 (G,H) DAF are shown. Embryos were dissected after imbibing seeds on agar plates for 20 minutes, then photographed immediately. The 16-day old embryos and seeds appear more hydrated in this photograph than seeds that haven’t been dissected or photographed. All photographs were taken at the same magnification. Scale bar, 300 μm.
Seed development in wild-type (A,C,E and G) and abi3 mutants (B,D,F and H) of Arabidopsis. Embryos dissected from seeds of wildtype and abi3 harvested 4 (A,B), 8 (C,D), 12 (E,F) and 16 (G,H) DAF are shown. Embryos were dissected after imbibing seeds on agar plates for 20 minutes, then photographed immediately. The 16-day old embryos and seeds appear more hydrated in this photograph than seeds that haven’t been dissected or photographed. All photographs were taken at the same magnification. Scale bar, 300 μm.
Water content of wild-type and abi3 seeds during maturation. Approximately 100 seeds from staged siliques of wild-type (◻) and abi3 (◼) were harvested, weighed, allowed to desiccate for 3 days, then weighed again. The water content was determined by dividing the weight loss during the 3 day desiccation period by the weight of the same seeds prior to desiccation. The average and standard deviation of 3 replicates are shown
Water content of wild-type and abi3 seeds during maturation. Approximately 100 seeds from staged siliques of wild-type (◻) and abi3 (◼) were harvested, weighed, allowed to desiccate for 3 days, then weighed again. The water content was determined by dividing the weight loss during the 3 day desiccation period by the weight of the same seeds prior to desiccation. The average and standard deviation of 3 replicates are shown
Mutant abi3 embryos develop normal green cotyledons in a manner similar to wild-type and appear morphologically normal by 8 DAF (Fig. 1C,D). After that point, however, they fail to degreen and remain fully hydrated at a stage when wildtype seeds are beginning to lose water (Figs 1F, 2). By 12-14 DAF, anthocyanin accumulation can be detected at the cotyledon margins, which has not been observed in wild-type embryos (Fig. 1E-H). Although mutant embryos remain green and maintain a higher water content than wild-type during most of embryo maturation, the embryo finally desiccates by 16-18 DAF (Fig. 2) suggesting that ovule abscission does occur in abi3 embryos yielding a shriveled seed containing a dark green embryo (Fig. 1H).
Expression of seed-specific genes
The temporally regulated expression of a number of gene families during seed development has been studied in an attempt to distinguish between specific stages of seed development at the molecular level (Goldberg et al., 1989; Hughes and Galau, 1991). In Arabidopsis, 12S cruciferin and 2S napin gene expression marks the end of embryo pattern formation and the beginning of the mid-stage of embryo development when storage reserves accumulate. Lea mRNAs, which are thought to play a role in protecting the embryo from desiccation, are usually expressed later than storage reserves and mark the developmental stage after ovule abscission (Dure, 1993).
Three 12S cruciferin genes, CRA, CRB, and CRC (Pang et al., 1988), and a 2S napin gene, AT2S1 (Guerche et al., 1990), were used to follow the SSP gene expression. The three 12S genes isolated in Columbia background do not cross hybridize to each other (Pang et al., 1988) while the AT2S1 gene detects all four 2S family genes (Guerche et al., 1990). In wild-type, these SSP gene mRNAs showed similar patterns of accumulation, reaching a maximal level 10-12 DAF and then decreasing (Fig. 3A-D). In contrast, in the abi3 mutant accumulation of SSP gene mRNAs was highly reduced (Fig. 3F-I) with the exception of the CRB gene, which accumulated to a nearnormal level (Fig. 3G). Lea mRNA accumulation which was not detected until 10 DAF in wild-type (Fig. 3E), could not be detected in the mutant under the conditions used in these experiments (Fig. 3J). The reduction of Lea mRNA accumulation in the severe abi3 mutant seed is similar in pattern to that reported using an Arabidopsis Lea gene probe in expression studies of a weak allele of abi3 (Finkelstein, 1993). Taken together with the morphological studies, the reduction in expression of the mRNAs associated with mid to late stages of embryo development suggest abi3 mutants complete the embryo pattern formation but are unable to successfully complete the maturation stage of embryo and seed development, which include the establishment of dormancy and desiccation tolerance.
Expression of SSP and LEA76 genes in wild-type (A-E) and abi3 (F-J) during seed development. Northern blots containing total RNA extracted from 1.5 siliques at the indicated number of days after flowering (DAF) were probed with 12S cruciferin (CRA; A and F; CRB, B and G; CRC, C and H), 2S napin (AT2S1; D and I) and the B. napus LEA76 genes (E and J).
Expression of SSP and LEA76 genes in wild-type (A-E) and abi3 (F-J) during seed development. Northern blots containing total RNA extracted from 1.5 siliques at the indicated number of days after flowering (DAF) were probed with 12S cruciferin (CRA; A and F; CRB, B and G; CRC, C and H), 2S napin (AT2S1; D and I) and the B. napus LEA76 genes (E and J).
Germination in darkness
Wild-type Arabidopsis requires light for efficient germination. The light requirement for germination can be circumvented by the addition of exogenous GA, suggesting light either directly or indirectly activates the synthesis of GA or increases the sensitivity of the seed to this hormone (Karssen and Lacka, 1985). The severe abi3 alleles used in this study were originally identified due to the mutant’s ability to germinate in the absence of GA synthesis in the light (Nambara et al., 1992, 1994), suggesting such mutants might germinate well in darkness.
We tested this possibility by imbibing abi3 and wild-type seeds in darkness for seven days. Mutant abi3 seeds harvested at 14 DAF showed 100% germination in the dark and 85-90% of these seedlings were similar in morphology to GA-induced dark germinated wild-type seeds (Fig. 4A). The remaining seedlings differed markedly from their siblings and although variable in phenotype all had fully expanded cotyledons attached to elongated cotyledonary petioles, and fully formed, unexpanded leaves bearing trichomes (Fig. 4B,C).
Dark germination of seedlings. Mature wild-type seeds and 14 DAF abi3 seeds were used to analyze dark germination profiles. A shows wild-type seedlings germinated in darkness for 7 days on agar containing 10−5 M GA3. B and C show abi3 seedlings germinated in darkness for 7 days. D shows a wild-type seedling imbibed in the light for 3 days then exposed to darkness for 7 days. Scale bars, 1 mm.
Dark germination of seedlings. Mature wild-type seeds and 14 DAF abi3 seeds were used to analyze dark germination profiles. A shows wild-type seedlings germinated in darkness for 7 days on agar containing 10−5 M GA3. B and C show abi3 seedlings germinated in darkness for 7 days. D shows a wild-type seedling imbibed in the light for 3 days then exposed to darkness for 7 days. Scale bars, 1 mm.
The unexpected ability of abi3 seedlings to produce expanded cotyledons and leaves in darkness could be accounted for if abi3 embryos have entered germinative development while still in seeds in siliques. If true, abi3 seeds imbibed in darkness are functionally equivalent to wild-type seeds that have been imbibed and germinated in light before being placed in darkness. To test this hypothesis, mature wildtype seeds were imbibed for varying lengths of time in the light and then moved to darkness for 7 days. Seeds imbibed in the light for 72 hours and then placed in darkness resulted in 100% germination with approximately half of the seedlings having expanded cotyledons and the first pair of foliage leaves (Fig. 4D), which is very similar to that observed in dark-germinated abi3 seedlings (Fig. 4B,C). Our result with wild-type seeds supports the hypothesis that late stage abi3 embryos are in a germinative state of development while in the seed.
Development of the first leaf pair and vascular differentiation in abi3 embryos
To clarify whether or not the leaf development observed in the dark-germinated abi3 seedlings occurred before or after imbibition, embryos dissected from seeds harvested 12 DAF, as well as germinating seedlings were prepared for whole-mount observation and photographed. The shoot apical meristems of all wild-type embryos observed were essentially flat (Fig. 5A,D) and correspond well to those described by Irish and Sussex (1992). Wild-type seedlings imbibed for 48 hours were essentially identical in appearance to wild-type embryos dissected from seeds harvested 12 DAF, except that the imbibed cotyledons were slightly larger and the root tips were more elongated. In contrast, two distinct populations of apical meristems were observed in abi3 embryos fixed 12 DAF. Approximately 90% of these embryos contained apical meristems at the leaf buttress stage of development. The height of the meristem (measured from the level of cotyledon attachment) was twice that of wild-type embryos and precursors of the first two leaves were clearly present as leaf buttresses (Fig. 5H). This stage of meristem development corresponds to that observed in wild-type seedlings 48 hours after initiation of imbibition (HAI, Fig. 5B,E). The remaining (approximately 10%) abi3 embryos contained meristems with well developed leaf primordia (Fig. 5G,I), corresponding to a stage of apical meristem development similar to that observed in germinating wild-type seedlings 72 HAI (Fig. 5C,F).
Shoot apical meristem formation and vascular differentiation in embryos and germinating seedlings. (A-F) Wild-type embryo and seedlings. (A,D) Wild-type embryo dissected from a seed harvested 12 DAF. The shoot apical meristem is flat. (B,E) A wild-type seedling imbibed for 48 hours. Meristem development has progressed such that buttresses of the first foliage leaf primordia are apparent. The diameter of the procambium is indicated by a double headed-arrow. (C,F) A wild-type seedling imbibed for 72 hours. Primordial leaves are apparent in the meristem and protoxylem differentiation is continuous throughout the seedling. (G-I) abi3 embryos dissected from seeds harvested 12 DAF. (G) Protoxylem tracheal elements have differentiated within the procambium of the stem-root axis and cotyledons of the embryo, but this has not occurred in wild-type embryos of the same age (A). (H,I) abi3 embryonic shoot apical meristems exhibit leaf primordia characteristic of wild-type seedlings imbibed for 48 and 72 hours (E,F). The degree of protoxylem differentiation in abi3 embryos at 12 DAF (G-I) is only comparable with that in wild-type seedlings imbibed for 72 hours (C,F). bp, leaf buttress primordia; lp, true leaf primordia; m, shoot apical meristem; pc, procambium; px, protoxylem. All scale bars, 100 μm.
Shoot apical meristem formation and vascular differentiation in embryos and germinating seedlings. (A-F) Wild-type embryo and seedlings. (A,D) Wild-type embryo dissected from a seed harvested 12 DAF. The shoot apical meristem is flat. (B,E) A wild-type seedling imbibed for 48 hours. Meristem development has progressed such that buttresses of the first foliage leaf primordia are apparent. The diameter of the procambium is indicated by a double headed-arrow. (C,F) A wild-type seedling imbibed for 72 hours. Primordial leaves are apparent in the meristem and protoxylem differentiation is continuous throughout the seedling. (G-I) abi3 embryos dissected from seeds harvested 12 DAF. (G) Protoxylem tracheal elements have differentiated within the procambium of the stem-root axis and cotyledons of the embryo, but this has not occurred in wild-type embryos of the same age (A). (H,I) abi3 embryonic shoot apical meristems exhibit leaf primordia characteristic of wild-type seedlings imbibed for 48 and 72 hours (E,F). The degree of protoxylem differentiation in abi3 embryos at 12 DAF (G-I) is only comparable with that in wild-type seedlings imbibed for 72 hours (C,F). bp, leaf buttress primordia; lp, true leaf primordia; m, shoot apical meristem; pc, procambium; px, protoxylem. All scale bars, 100 μm.
The shoot apical meristems of 12-day old abi3 embryos clearly had entered a developmental program corresponding to that observed in germinating wild-type seedlings. In addition to this, all the abi3 embryos we observed exhibited a marked degree of vascular differentiation (Fig. 5G-I). The procambium, which eventually differentiates to form the entire vascular system, was clear and could be traced in several focal planes throughout all embryos and seedlings observed (indicated by a double headed arrow in Fig. 5E). Several cell files of the procambium in abi3 embryos (Fig. 5G-I) and in wild-type seedlings observed 48-72 HAI (Fig. 5E,F) had helical secondary cell wall thickenings, indicating that an early stage of vascular differentiation, i.e. differentiation of protoxylem tracheary elements, had occurred. Protoxylem differentiation within the stem-root axis and cotyledons can be very clearly observed throughout abi3 embryos at lower levels of magnification (Fig. 5G). By observing the procambium using all available planes of focus, we confirmed that at least one continuous strand of protoxylem extended throughout the stem-root axis and each cotyledon of wild-type seedlings observed 72 HAI (Fig. 5C,F). Wild-type seedlings observed 48 HAI clearly exhibited some degree of protoxylem differentiation (Fig. 5E), however continuous strands of protoxylem were not found in these seedlings. There were regions of the cotyledons and just below the shoot apical meristem in which fully differentiated tracheal elements could not be observed (Fig. 5B,E).
Expression of Cab promoter during abi3 embryo maturation
Physiological and morphological observations of late embryo-genesis in abi3 mutants suggest genes induced after wild-type seed germination should be activated during late embryogenesis in abi3 mutants. We tested this hypothesis by analyzing the promoter activity of a gene encoding a chlorophyll a/b binding protein (Cab), a gene that has been shown to be induced by light after germination (Chory and Peto, 1990; Kubasek et al., 1992). A fusion gene containing the Cab promoter and the coding region of the E. coli uidA gene was crossed from a wildtype line into an abi3 mutant line, and GUS activity resulting from Cab promoter expression was measured in developing wild-type and abi3 seeds. We observed much higher levels of GUS activity in abi3 seeds compared to wild-type from 8-12 DAF (Fig. 6A). In abi3 mutant seeds the GUS activity reached a plateau about 10 DAF, and this level was similar to that observed in wild-type seedlings that had been imbibed for 3 days (Fig. 6B). These results are consistent with the notion that abi3 embryos precociously enter a germinative state.
Cab promoter expression during seed maturation and germination. GUS activity driven by a Cab promoter in wild-type (○) and abi3 mutant (•) seeds are shown. A shows the temporal change in GUS activity during wild-type and abi3 seed maturation and B shows the induction of GUS activity during wild-type germination. DAI, days after imbibition.
Cab promoter expression during seed maturation and germination. GUS activity driven by a Cab promoter in wild-type (○) and abi3 mutant (•) seeds are shown. A shows the temporal change in GUS activity during wild-type and abi3 seed maturation and B shows the induction of GUS activity during wild-type germination. DAI, days after imbibition.
DISCUSSION
Inactivation of ABI3 changes the developmental state of the embryo
The events that occur after embryo pattern formation in Arabidopsis are similar to those reported for other dicotyledonous plants, but they generally occur on a much shorter time scale. In null abi3 lines these events, associated with embryo maturation, are either absent or delayed in comparison to wild-type (Nambara et al., 1992, 1994). Although abi3 embryos cannot complete the maturation stage, these embryos are capable of initiating shoot apex development and protoxylem differentiation, so they are not developmentally arrested at the green cotyledon stage. In fact, late stage abi3 embryos appear morphologically very similar to germinating wild-type seedlings.
Premature seedling development accompanies loss of late embryo development
The unusual pattern of seedling development we observed after germinating abi3 seeds in the dark also occurred in wild-type seeds when germinated in the appropriate light/dark regime. Thus, abi3 embryos not only appear morphologically similar to germinating wild-type seedlings, but abi3 seedlings develop in the same way as 3-day old wild-type seedlings in response to both light and darkness. The difference between abi3 and wild-type seedlings lies in differences between the late stage embryos: abi3 embryos form aerial structure primordia at a stage when wild-type embryos do not.
The precocious initiation of shoot apical meristem development observed in abi3 embryos is similar to that observed in the fusca3 (fus3) mutant of Arabidopsis, which also fails to complete embryo maturation successfully (Keith et al., 1994). fus3 cotyledons have characteristics of both cotyledons and leaves, so two developmental programs that ordinarily do not overlap do so in fus3 embryos. Mutations such as fus3, which shift the timing of expression of developmental programs have been described as heterochronic. The abi3 mutation can also be interpreted as causing a heterochronic shift in embryo development; however, it differs distinctly from fus3 in one important way. abi3 cotyledon morphology, biochemistry and cellular ultrastructure more closely resemble germinating cotyledons than leaves (K. Keith, unpublished data). The fus3 embryonic leaves appear to be incapable of becoming fully functional cotyledons, whereas abi3 embryos form functional cotyledons, but the late stage of abi3 embryo development is incomplete and the embryos enter germinative development without pausing to permit desiccation tolerance to develop or dormancy induction to occur. Reductions in seed storage protein and Lea mRNA accumulation observed in abi3 seeds in this study and others, and quantitative changes in another major storage reserve in Arabidopsis seeds, triacylglycerol (Finkelstein and Somerville, 1990; Finkelstein, 1993) may be due to the incomplete operation of the maturation program in abi3 mutants.
A role for ABI3 in establishing embryo maturation
The ability of abi3 and fus3 embryos to germinate within the first 10 DAF, a period when wild-type embryos will not (Nambara et al., 1992; Keith et al., 1994), provides further evidence that these embryos are in germinative states of development. In fact, the abi3 mutation allows a high level of expression of the Cab gene (Fig. 6), which normally occurs after germination (Chory and Peto, 1990; Kubasek et al., 1992). From the perspective of the usual sequence of embryo development the ability to germinate appears to be a consequence of functional loss of embryo maturation as evidenced by embryo morphology, seed water content, and seed gene expression (Figs 1-3). It is also possible that the germinative state is, in fact, a default developmental state and that entry into the desiccation tolerance and dormancy programs may require positive or activating regulation, which is, at least in part, controlled by ABI3 (Fig. 7A). An alternative model is that a functional ABI3 gene product is required to delay or inhibit entry of the embryo into the germinative program after cotyledon development is complete (Fig. 7B). If this delay does not occur, the model predicts that embryo maturation will not occur successfully. The fact that expression of one of the 12S SSP genes, CRB, was not greatly affected in abi3 mutants (Fig. 3B,G) may indicate the presence of other regulatory pathways that control seed gene expression. It is intriguing to note that CRB gene expression is highly reduced in fus3 mutant seeds (Bäumlein et al., 1994; E. Nambara, unpublished results) while CRA and CRC gene expression is relatively normal in the fus3 mutant (E. Nambara, unpublished results) suggesting that the FUS3 gene product has a role complementary to ABI3.
A working hypothesis for the role of the ABI3 gene product during embryogenesis. The model in A shows an inductive role for the ABI3 gene product and that in B shows a repressive role for the ABI3 gene product.
The sequence identity between ABI3 and the maize transcriptional activator, VP1 (Giraudat et al., 1992), suggests that ABI3 acts as an activator of gene expression. Therefore, ABI3 may activate genes required for embryo maturation, repress genes that are required for germination to progress, or perhaps do both. While ABI3 function may be required for ABA action during seed dormancy, it is clear that ABI3 plays a much more complex role than simple transduction of the ABA signal in the seed, as proposed previously. Further understanding of how ABI3 regulates development in Arabidopsis requires identification of the genes or gene products that interact with ABI3.
ACKNOWLEDGEMENTS
We thank Drs A. Watanabe and Y. Komeda for substantial and insightful discussion. We are indebted to Dr N. G. Dengler and C. M. Kampney for advice regarding morphological and anatomical analysis and interpretation. We thank Dr J. Chory for the cab3-GUS transgenic strain. We also thank Ms K. Fujiwara for her general assistance. This work was supported in part by a Grant-in-Aid from Ministry of Education, Science and Culture of Japan to S. N. and by a grant from the Natural Science and Engineering Council of Canada to P. M.