Functional analysis of Arabidopsis NCED6 and NCED9 genes indicates that ABA synthesized in the endosperm is involved in the induction of seed dormancy


*(fax +33 1 30 83 30 99; e-mail


The cleavage of 9-cis-epoxycarotenoids to xanthoxin, catalyzed by 9-cis-epoxycarotenoid dioxygenases, is considered to be the key regulatory step of abscisic acid (ABA) biosynthesis. In Arabidopsis, genes for these enzymes form a multigene family with nine members, only five of which are thought to be involved in ABA production. In contrast to the prominent function of AtNCED3 in stress responses, the physiological and developmental role of the other 9-cis-epoxycarotenoid dioxygenases (NCEDs) remain unknown. Our functional and expression analyses have revealed that AtNCED6 and AtNCED9 are required for ABA biosynthesis during seed development. Reverse genetic analysis showed that ABA levels were reduced in Atnced6 and Atnced9 mutant seeds. In addition, transgenic plants overexpressing the AtNCED6 gene overproduced ABA. In accordance with mutant phenotypes, both AtNCED6 and AtNCED9 exhibited seed-specific expression. Detailed cytological studies were carried out, either by using transcriptional fusions of the promoter with GUS and GFP reporter genes, or by in situ hybridization. Expression of AtNCED6 was observed exclusively in the endosperm during seed development, that of AtNCED9 in both embryo and endosperm at mid-development. In addition to reduced ABA levels, Atnced6 and Atnced9 mutant seeds were also resistant to paclobutrazol, a gibberellin biosynthesis inhibitor. Although seeds of single mutants were still dormant, reduced dormancy was observed in the Atnced6 Atnced9 double-mutant seeds. These demonstrate that ABA synthesized in both the endosperm and the embryo participates in the hormonal balance that controls seed dormancy and germination.


The phytohormone abscisic acid (ABA) plays a major role in plant stress adaptation and in seed development and germination (Nambara and Marion-Poll, 2005). In accordance with this role, most ABA-deficient or -response mutants from various plant species show altered phenotypes in response to abiotic stress and in the expression of seed dormancy. Moreover, the regulation of these physiological processes depends largely on the fine-tuning of hormone levels. Indeed, ABA levels increase when plants are subjected to certain abiotic stresses, such as water deficit, and during seed development.

Abscisic acid is derived from carotenoid precursors. The cleavage of cis-epoxycarotenoids by 9-cis-epoxycarotenoid dioxygenase (NCED) is the first step specific to ABA biosynthesis. The NCED enzymes cleave the cis-isomers of violaxanthin and neoxanthin to a C15 product (xanthoxin) and a C25 by-product (Schwartz et al., 1997). The first NCED gene (VP14) was cloned from maize by insertional mutagenesis; NCED genes have subsequently been identified by sequence homology in several other plant species (Schwartz et al., 2003; Tan et al., 1997). In Arabidopsis nine NCED-related sequences have been identified, and phylogenetic analysis has indicated that five of these clustered with functionally characterized NCED proteins from other species (Schwartz et al., 2003). In agreement with this, recombinant proteins encoded by four of these genes exhibited cis-xanthophyll cleavage activity in vitro; but one of these, AtNCED5, remains to be characterized enzymatically (Iuchi et al., 2001).The remaining four Arabidopsis protein sequences, more distantly related to maize VP14, have been named AtCCD (carotenoid cleavage dioxygenase) and do not appear to be involved in ABA biosynthesis. The recombinant protein AtCCD1 has been shown to cleave symmetrically a variety of carotenoids at positions 9–10 and 9′–10′, whereas NCEDs cleaved carotenoids asymmetrically at positions 11–12 (Schwartz et al., 2001). The AtCCD7 and AtCCD8 recombinant proteins can cleave β-carotene sequentially, and their activities in plants produce a mobile signal that inhibits branching (Booker et al., 2004; Schwartz et al., 2004; Sorefan et al., 2003); as yet, no enzyme activity has been reported for AtCCD4.

Mutants affected in NCED genes have been identified from maize (vp14; Tan et al., 1997); tomato (notabilis; Burbidge et al., 1999; Thompson et al., 2004); and Arabidopsis (nced3; Iuchi et al., 2001; Ruggiero et al., 2004). Like vp14 and notabilis, Atnced3 mutant plants showed increased water loss and reduced ABA levels in vegetative tissues. In accordance with this, AtNCED3 transcript levels increased upon water deficit, whereas other AtNCED genes showed no or minor increases (Iuchi et al., 2001; Tan et al., 2003). Therefore AtNCED3 probably plays a major role in the regulation of ABA synthesis in response to stress. However, the phenotypes of mutants affected in other AtNCED genes remain to be analysed in order to evaluate the respective contributions of the five members of the gene family. AtNCED2, 3, 5 and 6 expression studies have been performed by GUS reporter gene analysis (Tan et al., 2003). Only AtNCED2 and 3 were found to be expressed in roots, whereas in aerial vegetative organs AtNCED2, 3 and 5 transcripts were detected. Their pattern of expression was tissue-specific, and restricted to particular cell types such as root tips, vascular tissue or guard cells. All four genes were expressed in reproductive organs and seeds, but again with variable patterns. Taken together, these results indicate that the tissue-specificity of ABA synthesis may be controlled largely by the differential expression of NCED genes. In addition, expression studies and phenotypic analysis of transgenic plants also indicate that carotenoid cleavage is the major regulatory step for ABA accumulation both upon stress and in seeds. Firstly, NCED gene and protein expression has been shown to be well correlated to stress-induced ABA accumulation in Phaseolus vulgaris (Qin and Zeevaart, 1999). Secondly, NCED transcript overexpression in transgenic plants from several species increased both water-stress tolerance and seed dormancy (Iuchi et al., 2001; Qin and Zeevaart, 2002; Thompson et al., 2000).

In seeds, ABA is synthesized in tissues of different origin: the testa is of maternal origin; the endosperm and the embryo both result from fertilization. The site of ABA synthesis conditions its physiological action. In particular, maternal ABA promotes reserve accumulation, and embryonic ABA induces seed dormancy and desiccation tolerance (Finkelstein et al., 2002). As in vegetative tissues, it is expected that the spatio-temporal regulation of certain NCED genes may be a key element in the control of ABA levels in seeds, for the regulation of physiological processes such as seed development, maturation, desiccation and germination. In this study we have undertaken a detailed analysis of the function and expression of genes AtNCED6 and AtNCED9. AtNCED6, according to a previous study, is expressed only in reproductive organs, but no precise information is available concerning AtNCED9 temporal and spatial expression (Tan et al., 2003). By reporter gene analysis and/or in situ hybridization, AtNCED6 expression in seeds was found to be restricted exclusively to the endosperm, whereas AtNCED9 was expressed in both endosperm and embryo. These observations led us to investigate the role played by ABA produced in the endosperm in the regulation of seed germination.


The AtNCED6 gene is involved in ABA biosynthesis in vivo

The in vivo involvement of NCED proteins in ABA biosynthesis has only been proven for AtNCED3 by functional analysis of mutants and transgenic plants (Iuchi et al., 2001; Ruggiero et al., 2004). As AtNCED6 is, among the five putative AtNCED enzymes, the most distantly related to other family members and, in particular, to AtNCED3 (Iuchi et al., 2001; Schwartz et al., 2003; Tan et al., 2003), we aimed to verify the involvement of AtNCED6 in ABA biosynthesis in vivo by expressing this protein ectopically in transgenic plants. The overexpression of ABA biosynthetic genes, such as LeNCED1 in tomato, PvNCED1 in Nicotiana plumbaginifolia and AtNCED3 in Arabidopsis, led to increased ABA levels in leaves and seeds (Iuchi et al., 2001; Qin and Zeevaart, 2002; Thompson et al., 2000). It would therefore be predicted that the overexpression of the AtNCED6 gene under the control of the cauliflower mosaic virus (CaMV) 35S promoter would result in a similar effect.

Primary transformants, carrying a 35S::AtNCED6 construct, were selected for kanamycin resistance. T2 progeny seeds of several lines showed reduced levels of germination. The number of germinated seeds could, however, be increased by addition of the ABA biosynthesis inhibitor norflurazon, or by stratification (data not shown), indicating that increased ABA synthesis might be responsible for increased seed dormancy. To facilitate subsequent analyses we selected three independent transgenic lines segregating with a 3:1 ratio for kanamycin resistance in T2 seeds, as predicted for a single T-DNA insertion which exhibited high levels of T2 seed germination after stratification. Expression of the AtNCED6 transgene was verified in the homozygous offspring of these three transgenic lines by Northern analysis. In contrast to wild-type (WT) leaves, where no AtNCED6 transcripts were detected, RNA was present in transgenic leaves (Figure 1a). Accumulation of ABA was then measured in non-stressed rosettes of the three lines, and an increase of up to fourfold was observed in comparison with WT levels (Figure 1b). Unexpectedly, when detached rosettes were dehydrated to reduce the relative water content (RWC) to 80% of that of non-stressed rosettes, the increase in ABA content was lower (1.5- to fourfold) for transgenic lines than for WT (about 10-fold). Therefore increased ABA synthesis apparently downregulated ABA accumulation upon stress, possibly through enhanced catabolism, as proposed for N. plumbaginifolia NCED overexpressers by Qin and Zeevaart (2002). Furthermore, higher ABA levels in turgid tissues increased their tolerance to subsequent desiccation. Firstly, the time required to reduce the water content of transgenic rosettes from 100% to 80% was longer than that for the WT (200 and 80 min respectively). In addition, analysis of the dehydration rate confirmed that transgenic plants were more resistant than WT to water loss (Figure 1c). After 4.5 h dehydration, wild-type detached rosettes had lost 40% of their water content, whereas transgenic lines had lost only 30%. These results prove that the AtNCED6 gene is implicated in ABA biosynthesis in vivo. In addition, the results demonstrate that ectopic expression of AtNCED6 in vegetative tissues results in increased tolerance to hydric stress, as shown previously for AtNCED3 (Iuchi et al., 2001). Furthermore, increased ABA levels in transgenic lines also induced alterations in their vegetative development. Rosette size was reduced and the floral spike appeared later than in WT (Figure 2). Therefore increased ABA synthesis in non-stressed plants reduced growth and delayed flowering time. Retardation of plant development might result from both the inhibition of growth by high levels of ABA, and the induction of stomatal closure which would reduce photosynthetic activity.

Figure 1.

Phenotypes of transgenic lines overexpressing a P35S::AtNCED6 construct. AtNCED6 transcript levels (a); ABA content (b); and water loss (c) of detached rosettes of transgenic lines overexpressing a P35S::AtNCED6 construct.
(a) Northern blot analysis of AtNCED6 transcript abundance in well watered plants. Hybridization was performed using an AtNCED6-specific probe. RNA loading was visualized by ethidium bromide staining.
(b) The ABA content was measured at relative water content of either (bsl00000) 100 or (bsl00001) 80%. Bars, SE (n = 3).
(c) Water loss of (bsl00063) WT; (bsl00066) Ataba3-1; (bsl00000) P35S::AtNCED6-1; (bsl00043) P35S::AtNCED6-2; (bsl00080) P35S::AtNCED6-3 is expressed as percentage of the initial fresh weight of detached rosettes. Bars, SD.

Figure 2.

Vegetative growth of wild type and two P35S::AtNCED6 transgenic lines exhibiting increased abscisic acid content.
When the floral spike measured 1 cm, two criteria were measured: (a) number of days after transplantation; (b) rosette diameter. Bars, SE (n = 23).

Spatio-temporal regulation of AtNCED6 and AtNCED9 transcripts in developing seeds

Previous studies indicated that transcripts of AtNCED3, AtNCED5, AtNCED6 and AtNCED9 could be detected by RT-PCR in developing siliques, and that AtNCED3, AtNCED5 and AtNCED6 expression could be detected in developing seeds by reporter gene analysis (Tan et al., 2003). To determine the precise localization of AtNCED gene expression in siliques and seeds in comparison with other known ABA biosynthesis genes, RT-PCR analysis was performed on dissected siliques and seeds at 10 days after pollination (DAP; Figure 3). AtZEP, AtNCED3, AtABA2, AtABA3 and AtAAO3 were expressed in both empty silique envelopes and seeds. In contrast, expression of AtNCED5, AtNCED6 and AtNCED9 was detected only in seed tissues. AtNCED2 expression was very low in all tissues at this stage. AtNCED6 and AtNCED9 transcripts were, among the five NCEDs, the most abundant in seeds at mid-maturation. Furthermore they exhibited different tissue-specific expression, AtNCED6 transcript being detected almost exclusively in the tissue fraction containing testa and endosperm, and AtNCED9 in both embryo and testa plus endosperm fractions.

Figure 3.

Expression of abscisic acid biosynthesis genes in different seed organs at 10 DAP. RT-PCR analysis was carried out on dissected embryos (bsl00000); testa plus endosperm (bsl00001); and empty silique envelope (bsl00023) at 10 DAP. Transcripts were normalized using 18S rRNA as an internal control. Duplicate experiments gave similar results.

As the ABA accumulated during seed development has been shown to derive from both maternal and embryonic tissues (Groot and Karssen, 1992), it was possible that tissue-specific ABA synthesis might be regulated by the differential expression of the AtNCED family members. As shown here (Figure 3), and also according to previous data, several AtNCED genes showed distinct expression patterns in seeds (Tan et al., 2003); however, the precise site of AtNCED6 and AtNCED9 expression was not reported. Therefore we investigated the tissue-specific localization of these transcripts by reporter gene analysis and/or in situ hybridization.

AtNCED6 is expressed specifically in the endosperm

To determine AtNCED6 spatial and temporal expression patterns more precisely, the gene promoter was fused to two different reporter genes, GUS and GFP. The results presented here were obtained with four independent transgenic lines for each construct, all of which presented the same expression profiles. Before fertilization AtNCED6 transcripts were present in pollen and at the micropylar pole of ovules, as revealed by GUS activity (Figure S1). After fertilization, AtNCED6 expression was detected specifically by GFP analysis of intact seeds (Figure 4) and by GUS staining of seed cross-sections (Figure 5). By these techniques, we observed that AtNCED6 transcripts were present from fertilization up to seed maturity, but no reporter gene expression was detected at later maturation stages, or in dry or imbibed seeds. At early stages after fertilization (until the early globular embryo stage) GFP expression was observed uniformly throughout the embryo sac (Figure 4a,b). Fertilization of WT ovules with transgenic pollen showed that reporter gene expression was already detected 24 h after pollination, indicating that AtNCED6 was expressed very early in fertilized cells and did not result simply from residual expression in synergid or antipodal cells which had persisted in the embryo sac (Figure S1). During the early stages, nuclei of the syncytial endosperm are first equally spaced in the embryo sac; then a central vacuole is formed and the syncytium fills the chalazal and micropylar chambers, but is restricted to a single layer of nuclei in the central chamber surrounding the central vacuole (Brown et al., 1999). At the late globular stage, fluorescence remained high in the micropylar endosperm localized around the embryo, and also at the chalazal pole, but became weaker in the central endosperm layer (Figure 4c). The lower GFP signal in this layer might result from localization of the GFP protein around the widely spaced nuclei that are surrounded by lateral vacuoles, as described by Brown et al. (1999). At the heart stage, cellularization of the endosperm begins in the micropylar chamber and then proceeds through the central chamber to the chalazal chamber (Berger, 2003). At this point, from heart to curled cotyledon-stage embryo, the GFP signal became lower around the embryo, but remained strong at the chalazal pole. A high fluorescence intensity in the chalazal chamber might correlate with the persistence of a large coenocytic cyst of multinucleate cytoplasm (Olsen, 2004). At all these stages (Figure 4d–f), no fluorescence was detected in the embryo or seed testa. Therefore AtNCED6 expression was apparently limited to the endosperm, and its expression pattern correlated with endosperm development.

Figure 4.

Expression of an AtNCED6 promoter::GFP construct in the endosperm of developing seeds. Fluorescence was observed in the embryo sac of a fertilized ovule (a); the embryo sac of an early globular stage embryo (b); at the chalazal pole and surrounding the embryo in a seed with late globular stage embryo (c); heart-stage embryo (d); torpedo-stage embryo (e); and curled cotyledon-stage embryo (f). Scale bar = 40 μm (a); 50 μm (b–d); 150 μm (e, e′, f). Es, embryo sac; ch, chalazal pole; mi, micropylar pole.

Figure 5.

Expression of AtNCED6 and AtNCED9 in developing seeds. (a) Expression of AtNCED6 promoter::GUS constructs in seed sections 14 days after flowering. Double-headed arrow in a1 indicates position of cross-section in a2.
(b) In situ hybridization of wild-type developing seeds using an AtNCED6-specific probe 14 days after flowering. Arrows indicate location of a strong hybridization signal with the antisense probe in the chalazal endosperm (b1) which is not detected with the sense probe (b2).
(c) In situ hybridization of wild-type developing seeds using an AtNCED9-specific probe 10 days after flowering. Arrows indicate location of a strong hybridization signal with the antisense probe in the periphery of the embryo and in the endosperm (c1–c3), which is not detected with the sense probe (c4). Double-headed arrow in c1 indicates position of cross-section in c3. Scale bar = 100 μm.

Localization of AtNCED6 gene expression in the endosperm was then confirmed by the observation of seed sections after GUS staining or in situ hybridization. In 14-day-old seeds, GUS staining was clearly visible in the endosperm at the chalazal and micropylar poles, but also in the cellularized cell layer surrounding the embryo (Figure 5a). In situ hybridization using seeds at the same stage also showed that transcripts were detected in these tissues (Figure 5b). Together, these observations confirmed that AtNCED6 expression was tightly restricted to the endosperm in accordance with the data obtained by RT-PCR (Figure 3).

AtNCED9 is expressed in embryo and endosperm

A 3-kb AtNCED9 gene promoter was fused transcriptionally to the two reporter genes in order to determine the tissue specificity of AtNCED9 expression. However, neither GUS nor GFP could be detected in transgenic plants. In a previous study, Tan et al. (2003) were unable to detect reporter gene expression using a 1.5-kb promoter fragment fused to the GUS gene. We have shown here that extending the promoter fragment at the 5′ end did not restore promoter functionality. Therefore we can hypothesize that an important regulatory sequence is missing, which could reside after the start codon. We also performed in situ hybridization, using an AtNCED9-specific probe, of 10-day-old seed cross-sections. As presented in Figure 5(c), AtNCED9 is expressed in the peripheral cell layers of the embryo, as well as in the endosperm. These data confirmed those obtained by RT-PCR (Figure 3).

Atnced6 and Atnced9 mutants did not exhibit ABA-deficient phenotypes in vegetative tissues

In order to investigate the contribution of both AtNCED6 and AtNCED9 genes to ABA biosynthesis during seed development, the effects of mutations in both genes were analysed. Two independent insertion mutants in AtNCED6 were identified in the databases of mutant collections used for systematic sequencing of insertion-flanking sequences. The first, Atnced6-1, was in the Sainsbury Laboratory Arabidopsis thaliana Transposants collection generated by dSpm element insertions in the Col-0 accession (Tissier et al., 1999). The second, Atnced6-2, was a Versailles collection T-DNA mutant in the Ws accession. Genomic DNA was extracted either from individual plants from a seed pool containing Atnced6-1, or from Atnced6-2 T3 plants, and mutants were identified by PCR amplification from the dSpm 5′ or left border extremity, respectively. The location of the insertions was then characterized on homozygous mutants (Figure 6a). Similarly, two independent insertion mutants, Atnced9-1 and Atnced9-2, in Col-0 accession were found among the salk T-DNA insertion lines (Alonso et al., 2003) and obtained from the Nottingham Arabidopsis Stock Centre (NASC) seed bank. These two mutants were characterized as described for Atnced6 mutants (Figure 6a).

Figure 6.

Analysis of AtNCED6 and AtNCED9 mutants. (a) Schematic representation of AtNCED6 and AtNCED9 genes indicating sites of insertion in different mutant alleles.
(b) Levels of ABA in dry seeds of wild-type, Atnced6 and Atnced9 single mutants and the Atnced6-1 Atnced9-1 double mutant. Bars, SD. Question marks indicate that inserted T-DNA might either be truncated or have a complex structure, as no PCR amplification was obtained using a primer in the right border sequence and gene-specific primers, although the AtNCED gene sequence next to the insertion was unaltered.

Abscisic acid plays a major role in drought-stress responses and is involved in stomatal closure. However, AtNCED6 and AtNCED9 exhibited respectively no or very low expression levels in vegetative tissues (Tan et al., 2003), therefore they are not expected to play a major role in water-stress tolerance. Nevertheless, to verify gene function, Atnced6 and Atnced9 mutants were tested for their resistance to rapid dehydration. We observed that dehydration rates, and the increase in ABA levels observed in detached rosettes upon desiccation stress, were similar to those of their respective WT (Figure S2). The absence of vegetative phenotypes in the Atnced6 mutants correlates well with the observed tissue-specific expression of AtNCED6 transcripts. Although detailed information is not available for AtNCED9 expression, very low transcript levels were detected in vegetative tissues by both RT-PCR and Northern analysis, and these increased slightly upon water stress (Iuchi et al., 2001; Tan et al., 2003). Nonetheless, the absence of vegetative phenotypes in the mutants confirmed that the AtNCED9 gene does not play a major role in ABA synthesis and stress tolerance upon rapid dehydration.

Atnced6 and Atnced9 mutants exhibited ABA-deficient phenotypes in seeds

In contrast to vegetative tissues, AtNCED6 and AtNCED9 transcripts were abundant in seed tissues (Figure 3). Levels of ABA were measured in dry mutant seeds. The ABA content was shown to be reduced to about half that of WT in Atnced6-1 and Atnced6-2, indicating that ABA synthesis in the endosperm makes a significant contribution to the amount of ABA in dry seeds (Figure 6b). For Atnced9-1 and Atnced9-2 mutants, ABA levels were 25% and 40% lower, respectively, than in WT seeds.

Most ABA-deficient mutants exhibit reduced seed dormancy, and genetic studies have demonstrated that embryonic ABA, not maternal ABA, is responsible for induction of dormancy (Karssen et al., 1983). The availability of ABA-deficient mutants that were specifically affected in ABA synthesis in the different seed tissues allowed the respective roles of ABA synthesized in the endosperm and in the embryo to be evaluated. Freshly harvested seeds of the Atnced6 and Atnced9 mutants germinated at a rate similar to WT seeds, indicating that dormancy was correctly established, in contrast to a classic ABA-deficient mutant such as Ataba3-1 (Figure 7a). Lack of seed dormancy in ABA-deficient mutants has also been described as being associated with increased resistance to paclobutrazol, an inhibitor of gibberellin (GA) synthesis (Karssen and Laçka, 1985). When sown on a medium supplemented with paclobutrazol, Atnced6 and Atnced9 seeds were clearly more resistant than WT (Figure 7b). This indicated that fewer GAs were required for Atnced6 and Atnced9 germination, as has been observed previously for ABA-deficient mutants affected in the step encoded by a single gene (Leon-Kloosterziel et al., 1996). Interestingly, the Atnced9 mutant was more resistant to paclobutrazol compared with the Atnced6 mutant. These observations suggest that the reduction of ABA levels in either Atnced6 or Atnced9 seeds is not sufficient to reduce seed dormancy, and that both endosperm-and embryo-derived ABA affect the level of GA required for germination.

Figure 7.

Seed phenotypes associated with AtNCED6 and/or AtNCED9 deficiency.
Germination of freshly harvested seeds (a) and paclobutrazol resistance of germinating seeds (b) of mutant Atnced6-1 (bsl00001); mutant Atnced9-1 (bsl00063); double-mutant Atnced6-1 Atnced9-1 (bsl00066); mutant Ataba3-1 (•); and WT (Col-0; bsl00043). Bars, SD. Data show Atnced6-1 and Atnced9-1 mutants, compared with Ataba3-1 and WT (Col-0) controls. Similar results were obtained for Atnced9-2 and Atnced6-2.

Cross-fertilization between Atnced6-1 and Atnced9-1 mutants was performed, giving rise to the Atnced6 Atnced9 double mutant. Plants homozygous for the two insertions were selected by PCR amplification, and the progeny of two double-mutant lines, issued from fertilization of two independent plants for each genotype, were studied further and gave identical results. Paclobutrazol resistance of double-mutant seeds was enhanced compared with Atnced6 and Atnced9 single mutants, as seeds exhibited a level of resistance similar to that of Ataba3-1 (Figure 7). Moreover, dormancy was notably reduced in the double-mutant seeds, although germination rates of double-mutant seeds were intermediate between those of WT and Ataba3-1. The observed heterogeneity of germination for double-mutant seeds may be due to their reduced level of dormancy being affected to different degrees by variations in physiological and environmental factors during seed development on the mother plant. In accordance with the reduction in seed dormancy and increased paclobutrazol resistance, ABA levels in double-mutant dry seeds were lower, although only about half that of WT in three independent experiments, and the maximal reduction compared with Atnced6-1 seeds was only 18%. Therefore other AtNCED genes probably contribute to ABA synthesis for the induction of seed dormancy.


The site of ABA production is critical for its regulatory action in seeds (Finkelstein et al., 2002; Nambara and Marion-Poll, 2005). As they belong to a gene family, AtNCED genes might contribute to the fine-tuning of tissue-specific synthesis of the hormonal signal in seeds by exhibiting restricted expression patterns. In contrast, other ABA biosynthesis genes (AtZEP, AtABA2, AtABA3 and AAO3) are unique, therefore their expression should be more ubiquitous. Expression of AtZEP, AtABA2, AtABA3 and AAO3 was detectable by RT-PCR in silique envelope, embryo and testa plus endosperm extracts (Figure 3). In accordance with this, AtZEP gene expression has been detected previously by in situ hybridization in the embryo, endosperm and testa throughout seed development (Audran et al., 2001). Expression of genes for steps further downstream, AtABA2 and AAO3, has been reported to be very low in seeds (Cheng et al., 2002; Seo et al., 2004), and effectively their expression levels were lower at 10 DAP than those of AtZEP (Figure 3). As AAO3 expression in the embryo at 10 DAP appeared to be very low, further studies will be necessary to investigate the spatio-temporal expression of this gene in seeds, which might constitute a limiting step for ABA accumulation in the embryo. In contrast to AtZEP, AtABA2, AtABA3 and AAO3, AtNCED genes effectively exhibited more restricted expression patterns, as reported previously (Tan et al., 2003). AtNCED6 was expressed before fertilization in male and female gametophytes, and then immediately after pollination in seed endosperm (Figures S1 and S2). This suggests that ABA synthesis might take place during the fertilization process, although its regulatory function remains to be discovered. During seed development, AtNCED6 expression was detected reproducibly and exclusively in seed endosperm (Figures 4 and 5). As AtNCED6 was highly expressed in seeds, compared with other family members, a large part of seed ABA might be synthesized in the endosperm. Indeed, ABA content was notably reduced in Atnced6 mutant seeds, indicating that a significant amount of the ABA present in dry seeds is produced in endosperm tissues (Figure 6b). AtNCED9 was, with AtNCED6, the most expressed gene in developing seeds on RT-PCR analysis at 10 DAP (Figure 3). However in contrast to AtNCED6, AtNCED9 was shown by in situ hybridization to be expressed in both endosperm and embryo, (Figures 3 and 5c). High levels of expression of AtNCED6 and AtNCED9 at mid-maturation were consistent with data reported by in silico Northern analysis throughout silique or seed development ( AtNCED6 and AtNCED9 transcripts were detected at early and mid-maturation stages, whereas AtNCED5 expression was observed only at later stages, and very low levels of AtNCED2 and AtNCED3 transcripts were detected by this technique. In accordance with the concomitant expression of NCED genes in seeds, the germination phenotypes of Atnced6 and Atnced9 mutants were milder than ABA-deficient mutants affected in unique genes (Figure 7), indicating that gene redundancy occurs. Genetic studies based on reciprocal crosses proved that embryonic ABA, not maternal ABA, is responsible for the induction of seed dormancy (Groot and Karssen, 1992; Karssen et al., 1983). However, it remains unclear whether this active ABA is synthesized in the embryo and/or the endosperm, as both are embryonic tissues produced from the double-fertilization process (Berger, 2003). The observation that the Atnced6 Atnced9 double mutant shows reduced dormancy clearly indicates that ABA synthesis in the endosperm does contribute, probably together with the ABA synthesized in the embryo, to the induction of seed dormancy (Figure 7). Nevertheless, as residual dormancy and ABA levels were detected in Atnced6 Atnced9 double mutants, it is highly likely that other AtNCED genes also participate in the synthesis of seed ABA responsible for dormancy induction.

Atnced6 and Atnced9 single mutants exhibited a typical characteristic of ABA-deficient mutants, as mutant seeds were able to germinate on paclobutrazol concentrations that inhibited WT germination (Figure 7). This has been observed in most ABA-deficient mutants in association with a reduction in seed dormancy, and interpreted as a lower requirement for gibberellins for germination when ABA levels are reduced (Leon-Kloosterziel et al., 1996). The mild reduction in ABA levels found in single mutants was sufficient to affect paclobutrazol resistance, whereas dormancy was maintained (Figure 7). Although ABA levels in double-mutant dry seeds were not reduced significantly compared with single mutants, ABA synthesis ability in double mutants was probably decreased further as dormancy was reduced. It can be hypothesized that a threshold level of ABA is necessary during seed development to induce dormancy, and that Atnced6 and Atnced9 single mutants would synthesize enough ABA, whereas this threshold would not be attained in double mutants. Therefore the phenotype of double mutants would be similar to that of some ABA-deficient mutants of N. plumbaginifolia and Arabidopsis, in which residual ABA levels were found to be insufficient to induce seed dormancy (Frey et al., 2004; Koornneef et al., 1989). In contrast, the GA requirement for germination would be affected by a smaller reduction in ABA synthesis ability, confirming that germination is indeed regulated by the GA:ABA hormonal balance. The ABA synthesized in the embryo, as well as in the endosperm, during seed development must contribute to the hormonal balance controlling germination, as both Atnced6 and Atnced9 seeds exhibited paclobutrazol resistance. During seed germination several events occur concomitantly, notably embryo growth is possible due to reserve mobilization and hydrolysis of the external barriers, endosperm and testa (Bentsink and Koornneef, 2002). These physiological processes are positively regulated by GAs and inhibited by ABA. It can be hypothesized that reduced synthesis, in particular of seed tissues, might be sufficient to reduce GA requirement for activation of only a subset of physiological processes, but insufficient for other germination processes that take place in other seed compartments or do not exhibit the same level of sensitivity to the hormone. For instance, in Atnced6 mutants, lower ABA levels in the endosperm might affect some of these processes, such as hydrolysis of endosperm and its reserves, but not other mechanisms, such as embryo growth. This could also explain how the GA requirement for germination would be decreased, despite dormancy being maintained in single mutants.

Despite an apparent overlap in gene function, the analysis of AtNCED6 and AtNCED9 gene expression and mutant phenotypes has yielded insights and working hypotheses about the role of ABA during seed development and germination, and has highlighted the contribution of the ABA synthesized in the endosperm. Further studies on other members of the AtNCED family will be necessary to determine precisely whether other AtNCED genes are expressed in the endosperm and in the embryo, and which AtNCED genes are responsible for ABA synthesis in the testa.

Experimental procedures

Plant material and growth conditions

Arabidopsis WT and mutant plants were grown either in a glasshouse, with a minimum photoperiod of 13 h assured by supplementary lighting or in a growth chamber, 16 h light, 8 h dark photoperiod, 25°C, 70% RH, 250 μE m−2 sec−2. The Atnced6-1 mutant (Columbia-0 accession) was obtained from SINS (Sequenced Insertion Site; Tissier et al., 1999: The Atnced6-2 mutant (Wassilewskija accession) was obtained from the Versailles mutant collection FlagDB/FST (Balzergue et al., 2001: The Atnced9-1 and Atnced9-2 mutants (Columbia-0 accession) were identified among salk lines (Alonso et al., 2003) from the SiGnAL database ( and obtained from NASC ( The mutant Ataba3-1 (Columbia-0 accession) was kindly provided by Dr M. Koornneef.

Germination experiments and water-loss assays

For dormancy analysis, freshly harvested mature seeds were sown in triplicate in Petri dishes containing a 0.5% (w/v) solidified agarose solution and placed in the growth chamber. Germination was scored each day based on radicle protrusion. For paclobutrazol resistance tests, seeds were surface-sterilized and sown on agarose medium supplemented with paclobutrazol (SOPRA, Velizy, France). The dishes were placed in a cold room (4°C) for 3 days before transfer to the growth chamber for 4 days. Plantlets able to develop two green cotyledons were scored as paclobutrazol-resistant.

For rapid dehydration assays, rosettes from plants grown in a glasshouse for 2–3 weeks were cut from the root system and weighed. Four plants of each genotype were placed in a laminar flow hood, abaxial surface uppermost, for dehydration and weighed at regular intervals. Water loss was estimated as the percentage of fresh weight lost relative to the initial fresh weight.

ABA content

Rosettes or dry seeds were frozen in liquid nitrogen and freeze-dried. Extraction in non-oxidative methanol:water (80:20, v/v), pre-purification through SepPak C18 cartridges (Waters, Milford, MA, USA) and HPLC fractionation in a Nucleosil C18 column (Macherey-Nagel, Düren Germany) have been described previously (Kraepiel et al., 1994). Recovery of purified ABA was determined by means of 3H-ABA added to the extracts and scintillation counting of aliquots of the purified fractions. The ELISA procedure was based on competition for a limited amount of monoclonal anti-ABA antibody (LPDP 229, UPMC, France) between standard ABA–bovine serum albumin (BSA) conjugate adsorbed onto the wells of a microtitre plate, and free ABA extracted from the samples. Bound antibodies were labelled with a peroxidase-conjugated goat antibody to mouse immunoglobulins (Sigma, St Louis, MO, USA), and peroxidase activity was then measured. A standard curve was established on each microtitre plate. The ABA content was determined five times for each sample.

Quantitative RT-PCR

To obtain silique samples for RNA extraction, flowers were tagged on the day of flowering and sampled after 10 days. Siliques were dissected into three parts: embryo, testa/endosperm, and other parts of silique, under a stereoscopic microscope. Total RNA was isolated using RNAqueous column with Plant RNA isolation aid (Ambion, Austin, TX, USA) according to the manufacturer's protocol. Quantitative RT-PCR was performed as described by Kushiro et al. (2004) using primers described by Seo et al. (2004).

In situ hybridization

A 465-bp fragment of the AtNCED6 genomic sequence and a 521-bp fragment of the AtNCED9 sequence were used as probes in subsequent in situ hybridization and Northern analysis. These fragments were amplified from Arabidopsis (Col-0) genomic DNA using the following primers: for AtNCED6, 5′-ATG CAA CAC TCT CTT CGT TCT-3′ and 5′-AAT CAT TCC GTC ACC GTC AAA-3′; and for AtNCED9, 5′-ATG ACG ATA ATA CCA TTA TTT CTG G-3′ and 5′-GGA TGG GGA TGA CGG CGG CGC-3′, then cloned into the pGEM-T vector (Promega, Madison, WI, USA) and sequenced. The specificity of this probe was tested by dot-blot analysis against the five AtNCED and two AtCCD (AtCCD1 and AtCCD4) genes after amplification of full-length genomic fragments (data not shown).

Single-stranded RNA probes were synthesized after linearization of plasmid DNA carrying the 465-bp AtNCED6 and 521-bp AtNCED9 fragments by digestion using NcoI and SalI restriction enzymes (Invitrogen, Carlsbad, CA, USA) for antisense and sense probes, respectively. Transcription was performed using either SP6 (antisense probe) or T7 (sense probe) as the origin of transcription using a kit (Riboprobe combination system SP6-T7 RNA polymerase, Promega). Transcripts obtained were treated with DNAse (10 U, Promega) and then precipitated.

Tissue sections of immature siliques (Col-0 accession) were fixed in 4% (p/v) formaldehyde and 0.1% (v/v) triton-X-100 under vacuum for 1 h, and left overnight at 4°C. After fixation samples were washed in water, dehydrated and treated in histoclear (Histo-clear II, National Diagnostics, Atlanta, GA, USA) before being embedded in paraffin (Paraplast plus, Sherwood Medical, St Louis, MO, USA) as described by Jackson (1991). Sections 9 μm thick were cut and placed on slides (Dako, Glostrup, Denmark). Prehybridization was carried out essentially as described by Jackson (1991) with the following modifications: after dehydration, slides were rinsed in water and incubated for 20 min in 2 × SSC. For protease treatment, slides were incubated for 30 min at 37°C in a solution (20 mm Tris–HCl pH 7.5, 2 mm CaCl2) containing 1 μg ml−1 proteinase K. Hybridization was performed using Dako hybridization buffer at 60°C overnight. Slides were then washed with 0.2 × SSC for 2 h at 55°C, followed by 5 min in phosphate buffered saline (Eurobio, Courtaboeuf, France) at room temperature. Prior to immunological detection, slides were incubated for 1 h in 0.5% blocking reagent (Roche, Mannheim, Germany), then in 1% (w/v) BSA, 0.1% (v/v) triton-X-100. Detection of hybridized transcripts was performed using anti-digoxigenin antibodies conjugated with alkaline phosphatase (Roche).

Northern blot analysis

Total RNA from rosettes was extracted as described by Audran et al. (2001). RNA (10 μg) was loaded onto a 1.2% (w/v) agarose gel (Sambrook et al., 1989) and electrophoresis was performed in 1 × MOPS buffer pH 7.0 (MOPS 20 mm, sodium acetate 5 mm, EDTA 1 mm) at 60 V for 2 h. Fractionated RNA was then transferred onto nylon membranes (Genescreen, NEN, Boston, MA, USA) according to the manufacturer's instructions. Hybridization was performed using the AtNCED6 radiolabelled probe.

Cloning, plant transformation and reporter gene analysis

For reporter gene analysis, a 1.8-kb fragment of the AtNCED6 promoter (−1820 to −4 bp upstream of the ATG codon) was amplified from WT Col-0 genomic DNA and cloned into either pBI101-2 (Clontech, Palo Alto, CA, USA) for GUS expression (pBI101-2::pAtNCED6), or pBINmGFP5-ER ( for GFP (pBINmGFP5-ER::pAtNCED6). To overexpress the AtNCED6 transcript in transgenic plants, the transformation vector pKYLX71-35S2 was used and the full-length AtNCED6 cDNA cloned in the sense orientation under control of the 35S promoter of cauliflower mosaic virus (Audran et al., 2001). Binary vectors were electroporated into the Agrobacterium tumefaciens C58C1pMP90 strain and used for agroinfiltration of WT (Col-0; Clough and Bent, 1998). Transformed plants were selected for kanamycin resistance.

GUS-staining assays were performed as described by Jefferson et al. (1987) on dissected Arabidopsis siliques, at different developmental stages, fixed with acetone 90% (v/v) for 30 min at −20°C. Potassium ferricyanide/potassium ferrocyanide were used at 5 mm. For sections, siliques were paraffin-fixed and embedded as described above for in situ hybridization. Siliques from plants transformed with pBINmGFP5-ER::pAtNCED6 were dissected, and seeds were placed between a slide and a cover slip with water. Samples were observed under UV with a microscope (Leica, Wetzlar, Germany).


We thank Jocelyne Kronenberger for advice and assistance with seed sections and in situ hybridizations, Michael Anjuere and Bruno Letarnec for technical assistance in plant culture, and Sachiyo Harada for her technical help in collecting embryo, testa/endosperm and silique samples. This work was supported by a doctoral fellowship from the Ministère de l'Enseignement et de la Recherche to V.L.