• In orthodox seeds, the transcriptional activator ABI3 regulates two major stages in embryo maturation: a mid-maturation (MAT) stage leading to accumulation of storage compounds, and a late maturation (LEA) stage leading to quiescence and desiccation tolerance. Our aim was to elucidate mechanisms for transcriptional shutdown of MAT genes during late maturation, to better understand phase transition between MAT and LEA stages.
• Using transgenic and transient approaches in Nicotiana, we examined activities of two ABI3-dependent reporter genes driven by multimeric RY and abscisic acid response elements (ABREs) from a Brassica napus napin gene, termed RY and ABRE, where the RY reporter requires ABI3 DNA binding.
• Expression of RY peaks during mid-maturation and drops during late maturation, mimicking the MAT gene program, and in Arabidopsis thaliana RY elements are over-represented in MAT, but not in LEA, genes. The ABI3 transactivation of RY is inhibited by staurosporine, by a PP2C phosphatase, and by a repressor of maturation genes, VAL1/HSI2.
• The RY element mediates repression of MAT genes, and we propose that transcriptional shutdown of the MAT program during late maturation involves inhibition of ABI3 DNA binding by dephosphorylation. Later, during seedling growth, VAL1/HSI2 family repressors silence MAT genes by binding RY elements.
The process of embryo maturation shapes important seed traits such as composition of storage compounds and seed viability, and has therefore been subject to intense research (Holdsworth et al., 2008; Santos-Mendoza et al., 2008). In orthodox seeds, embryo maturation is regulated by abscisic acid (ABA) and proceeds mainly in two phases: an initial phase where storage compounds are accumulated, termed MAT (maturation), and a final phase where the embryo becomes dormant and desiccation tolerant, termed LEA (late embryo-abundant; Vicente-Carbajosa & Carbonero, 2005). In cotton and Arabidopsis, MAT and LEA stages temporally coincide with the accumulation of two major classes of coordinately expressed mRNAs, where MAT genes typically code for storage proteins, and LEA genes code for a variety of protective proteins, termed Lea proteins (Hughes & Galau, 1989; Parcy et al., 1994). MAT gene expression peaks during mid-maturation and declines during late embryo maturation, whereas expression of LEA genes peaks during late maturation; it has been postulated that MAT and LEA gene expression programs are regulated by distinct program-specific regulators (Hughes & Galau, 1989). The transcriptional activator ABI3 is a crucial regulator of both MAT and LEA gene programs in Arabidopsis (Parcy et al., 1994), mediating ABA-responsive transactivation of genes from both programs (Giraudat et al., 1994). The bZIP transcription factor ABI5 is coregulated with LEA genes and activates their ABA-dependent gene expression (Brocard-Gifford et al., 2003). Bensmihen and coauthors looked for possible coactivators that might recruit ABI3 to MAT gene promoters, and identified several bZIP proteins that are temporally coregulated with MAT and LEA programs (Bensmihen et al., 2002), but none of them were proved to be the regulators postulated by Hughes & Galau (1989).
ABI3 belongs to the B3 domain protein family, characterized by the B3 domain, a plant-specific DNA-binding domain. Within the B3 domain family, ABI3 branches together with FUS3 and LEC2, forming a cluster of evolutionary and functionally conserved transcription factors, termed ABI3/VP1-like (Riechmann et al., 2000). The ABI3/VP1-like B3-domain proteins regulate different aspects of embryo maturation and bind the sequence CATGCA, called the RY, or Sph, element (Holdsworth et al., 2008).
The RY/Sph element (hereafter termed RY) is crucial for transactivation through ABI3/VP1-like B3-domain proteins (Ezcurra et al., 2000; Reidt et al., 2000), and is found predominantly in seed-specific promoters (Lelievre et al., 1992) and in promoters of genes upregulated by ectopic ABI3 expression (Suzuki et al., 2003). The recombinant B3 domain of ABI3 and its orthologue in maize, VP1, bind the RY element in vitro (Suzuki et al., 1997; Mönke et al., 2004), though ABI3 does not bind DNA in a yeast one-hybrid (Y1H) assay, and it was suggested that, in vivo, ABI3 interacts with the RY element indirectly by protein–protein interaction with FUS3 and/or LEC2 (Kroj et al., 2003). The ABA response element (ABRE) mediates ABI3 ABA signaling and is bound by bZIP proteins that interact with ABI3, such as ABI5 (Nakamura et al., 2001), and bioinformatic analysis showed that RYs and ABREs are physically clustered in upstream regions of genes activated by VP1 and ABA (Suzuki et al., 2005).
Recently, new members of the VP1/ABI3-like B3-domain proteins were characterized and named as VAL or HSI2 proteins (hereafter termed VAL) (Suzuki et al., 2007; Tsukagoshi et al., 2007). The VAL proteins contain an EAR-domain that mediates active transcriptional repression (Ohta et al., 2001), and it was shown that they are repressors of maturation genes during early seedling growth. Because RY elements are frequent in genes downregulated by VAL, it was speculated that VAL proteins bind RY elements (Suzuki et al., 2007).
In this paper, we show that the RY element mediates transcriptional repression from late maturation to early seedling growth in transgenic tobacco (Nicotiana tabaccum). The RY elements are over-represented in upstream and noncoding regions of Arabidopsis MAT genes, suggesting that transcriptional repression of MAT genes after mid-maturation is mediated by the RY element. In a transient assay using Nicotiana benthamiana, staurosporine as well as a seed-specific protein phosphatase 2C (PP2C) inhibit ABI3 transactivation of the RY reporter, suggesting that repression of the RY element requires dephosphorylation. VAL repressors inhibit ABI3 transactivation of the RY reporter, and their B3 domains bind the RY element in yeast. We propose a model that accounts for mechanisms underlying phase transition between MAT and LEA gene programs.
Materials and Methods
Y1H effectors The ABI3, B1B2B3, B2B3 and B3 effectors were kindly donated by R. Finkelstein and are described in Nakamura et al. (2001). The domain-swap effector FUS/ABI3sw was produced by gene synthesis (Sloning, Puchheim, Germany), and codes for the FUS3 protein with the B3 domain of ABI3 (amino acids Q560 to Q675) inserted between amino acids A79 and E195. The effectors FUS/VAL1sw and FUS/VAL2sw were produced by inserting a PCR fragment of the B3 domain of VAL1 (amino acids G285 to N397) and VAL2 (amino acids P275 to N388), obtained by PCR of an Arabidopsis cDNA library from mixed tissues, kindly donated by Dr Magnus Eklund (SLU, Uppsala, Sweden) into FUS3 as above. Primers were as indicated in the Supporting Information, Table S2 (and include UidSau-fwd, UidSau-rev, Moon & Callahan, 2004; and CB9, CB10, Burger et al., 2003). The wild-type FUS3 effector was obtained in a Y1H screen of a Brassica napus cDNA library from immature seeds using the RY trimer as bait (unpublished). All Y1H effectors were fused to the GAL4 activation domain and had a pGAD424 vector backbone.
Y1H reporters A synthetic 135 bp oligonucleotide representing a trimer of the Brassica napus napA composite RY fragment, CGTGCATGCATTATTACACGTGATCGCCATGCAAA (RY and G-box underlined), was inserted into vectors pHisi-1 and pLacZi; the resulting constructs were named pRY-HIS3 and pRY-Lac2. Contructs containing RY or G-box elements mutated as previously (Ezcurra et al., 2000) were made by the same procedure in vector pLacZi and called pMutRY-Lac2 and pMutG-Lac2. Vectors were provided with the Matchmaker one-hybrid kit (Clontech, Mountain View, CA, USA).
Binary constructs The P-309 and ABRE reporters, as well as the ABI3 reporter were described previously (Ezcurra et al., 1999, 2000). The RY reporter was produced by inserting the synthetic RY trimer described earlier upstream of the minimal 35S promoter (−46 to +8) in the min35S::GUS binary reporter (Ezcurra et al., 1999). The 35S::AHG1 and 35S::VAL1 constructs were produced by inserting the corresponding coding regions, obtained by PCR of the Arabidopsis cDNA library mentioned earlier between the 35S promoter and the NOS terminator of an in-house produced binary vector (PCR primers, Table S2). All binary vectors have a pGA581 vector backbone. All constructs were sequenced.
All Y1H reporters were stably integrated in the YM4271 yeast (Saccharomyces cerevisiae) strain at the HIS3 and/or URA3 loci following the manufacturer's instructions (Clontech). The ABI3-derived effectors were transformed into yeast strains RY-His3/RY-LacZi, mutRY-LacZi and mutG-LacZi, whereas FUS3 domain-swap effectors were transformed into strain RY-His3/RY-LacZi. Yeast colonies growing on synthetic dropout (SD) –Leu medium were tested by β-galactosidase colony-lift filter assay (Clontech protocol) and by assessing growth of restreaked cells on SD –Leu –His medium supplemented with 45 mm 3-amino-1,2,4-triazole (3-AT; Sigma).
Plant material and crossings
Transgenic N. tabaccum lines harboring construct 35S::ABI3 and different napA-derived GUS reporters (T0 generation), as well as their crossings to obtain the different napA::GUSx35S::ABI3 genotypes (T1 generation) have been described elsewhere (Ezcurra et al., 1999, 2000; see the Supporting Information, Methods S1). Growth and ABA treatment of T1 seedlings and plantlets were as described previously (Ezcurra et al., 1999, 2000).
Total RNA was isolated from seeds, reverse transcribed, and the cDNA obtained was used as a template for real-time reverse transcription polymerase chain reaction (RT-PCR) analysis according to Kuusk et al. (2006). Real-time PCR reactions were performed in triplicate using β-glucuronidase (GUS)-specific and actin-specific primers (Table S2).
Nicotiana benthamiana plants were grown under a 16-h photoperiod at 23°C. Agroinfiltration was carried out on 4-wk-old plants as described by Voinnet et al. (2003). After 5 d, 0.5-cm leaf discs were cut and assayed for GUS activity and/or RT-PCR. To study the effects of staurosporine, leaf discs from agroinfiltrated tobacco leaves were incubated for 5 h in Murashige and Skoog (MS) medium supplemented with the kinase inhibitor at a final concentration of 5 µm. For the ABRE reporter/VAL1 effector study, leaf discs were cut 4 d after agroinfiltration and incubated in 50 µm ABA 48 h. Where indicated, RT-PCR of agroinfiltrated leaves was performed on reverse-transcribed total RNA using gene-specific primers (see Table S2), and gene expression was quantified by densitometry of bands using NIH imagej software (NIH, http://rsb.info.nih.gov/ij/).
Our study uses transient and stable tobacco-based expression systems that we previously showed reflect well napA regulation in Brassicaceae (Stålberg et al., 1996; Ezcurra et al., 1999, 2000). Our choice is further supported by the high conservation of embryo maturation in angiosperms, which consistently involves ABI3/VP1 action through ABRE and RY elements, as well as dual MAT and LEA gene programs. Consistent with this, maize VP1 can complement the Arabidopsis abi3 mutant (Suzuki et al., 2001).
Regulation of RY-mediated gene expression
We previously showed that P-309, a napin promoter from the Brassica napus napA gene, requires two composite elements, termed ABRE and RY, for seed-specific expression and ABA-responsive transactivation by ectopic ABI3 in tobacco, and that ABI3 transactivates GUS reporters driven by multimers of RY and ABRE (Ezcurra et al., 1999, 2000). The RY reporter proved to be a marker of ABI3 DNA binding in vivo, because its transactivation requires the ABI3 protein with an intact DNA-binding B3 domain. Further, the ABA-responsive activity of the ABRE reporter requires the ABI3 protein with an intact conserved basic B2 domain. The two functions of ABI3, DNA binding and relaying of the ABA signal, are independent of each other because the B3-deletion mutant of ABI3 can still effectively relay ABA signaling and, conversely, the B2-deletion mutant can still bind the RY element. To establish whether DNA binding and ABA signaling in ABI3 are coregulated during embryo maturation, we monitored gene expression driven by the RY and ABRE reporters (shown in Fig. 1a) in developing seeds of transgenic tobacco. Figure 1b shows that the two constructs follow strikingly different time-curves: the ABRE reporter is inactive during early maturation at 7–14 d after pollination (dap), and rises steadily at 21–28 dap, whereas the RY reporter's activity starts rising at 14 dap, peaks at 21 dap and thereafter drops during late maturation. Because the GUS protein is very stable in planta (Jefferson, 1987), we addressed whether the drop in activity of the RY reporter at 28 dap is more pronounced at the transcriptional level by performing quantitative real-time PCR of the GUS transcript. Figure 1c shows that the ABRE reporter has high activity at 28 dap, showing that ABI3 is stably expressed and transcriptionally active, because the ABRE reporter is strictly seed-specific and ABI3-dependent (Ezcurra et al., 2000). Interestingly, the RY reporter has no detectable activity at 28 dap, indicating that ABI3 transactivation of the RY element is abolished during late embryo maturation.
Expression of ABI3 is mainly seed-specific, but can be reinduced by ABA during germination (Lopez-Molina et al., 2002) and early seedling growth both in Arabidopsis and rice (Teaster et al., 2007; Shobbar et al., 2008). To assess whether downregulation of RY-mediated transactivation is maintained in developmental stages following embryo maturation, we measured GUS activity in 8 d seedlings of transgenic tobacco harboring a 35S::ABI3 effector construct combined with three napA reporters, P-309, ABRE and RY (shown in Fig. 1a). We show that at 8 d after imbibition (dai) ectopic ABI3 is unable to mediate neither ABA-responsive transactivation of P-309, nor transactivation of the RY reporter (Fig. 2, top section). Notably, ABI3 is still able to activate ABA-responsive transcription through the ABRE, showing that it is stably expressed and transcriptionally active. This suggests that RY repression underlies repression of P-309, because P-309 is RY-dependent (Ezcurra et al., 2000). Moreover, repression of the RY element is developmentally regulated, because both RY and P-309 are transactivated by ectopic ABI3 after true leaves are formed (Fig. 2, bottom section; Ezcurra et al., 2000). Accordingly, inhibition of RY-mediated transactivation observed during late embryo maturation is maintained throughout germination and early seedling growth, but ceases after emergence of true leaves, suggesting that the ABI3 transactivation through the RY element is inhibited during a developmental window that encompasses late maturation, germination and early seedling growth.
We have shown that expression driven by the trimeric RY promoter peaks at mid maturation and follows a MAT-type program of expression, suggesting that downregulation of MAT genes during late maturation might be brought about by inhibition of RY element-dependent transactivation, and that LEA gene expression might be RY-independent. To evaluate this hypothesis, we examined the frequency of RY elements in MAT versus LEA genes in Arabidopsis. We chose to examine a subset of nine MAT and nine LEA genes that have been reported in the literature as belonging to either gene program by experimental evidence (Parcy et al., 1994; Nambara et al., 1995; see Table S1 for a list of genes analysed), and we confirmed their gene expression program by analysis of microarray data from AtGenExpress (Schmid et al., 2005; Fig. S1). In this dataset, downregulation of MAT genes is not detected at stage 10 (green cotyledons), although it is consistently reported in at least two independent studies by Northern analysis at both 18 and 21 days after flowering (daf; Parcy et al., 1994; Nambara et al., 1995), suggesting that seeds used in the AtGenExpress analysis had not reached the very late stages of maturation. To look for RY elements, we examined a region encompassing 1500 bp upstream of translational start (ATG codon), plus all introns as well as 200 bp 3′ UTR. We found that MAT genes have, on average, three-fold higher amounts of RY elements than LEA genes, and this difference is highly significant (P = 0.00004; Fig. 3). By comparison, we found no significant difference between the average amounts of G-boxes in MAT versus LEA genes.
Analysis of ABI3 DNA binding by Y1H assay
Our results suggest a mechanism for regulation of MAT genes involving RY element-dependent transcription, where RY mediates both initial transactivation during early and mid-maturation, and subsequent repression during late maturation. Repression of RY-dependent genes could be brought about either by inhibition of DNA binding in ABI3 (passive repression), or by repression of the RY element through repressors binding the RY sequence (active repression), or by a combination of both mechanisms. Interestingly, ABI3 is unable to bind the RY element in vivo in a Y1H system (Kroj et al., 2003), consistent with ABI3 DNA binding being regulated. We reasoned that it might be possible to reconstitute ABI3 DNA binding by deletion of eventual inhibitory domains, because DNA binding in transcription factors may be regulated by allosteric autoinhibition (Graves et al., 1998). To test this hypothesis, we examined RY element binding in ABI3 and ABI3 deletions (shown in Fig. 4a; kindly provided by Ruth Finkelstein), using the Y1H system. Figure 4b shows that the full-length ABI3 protein is unable to bind the RY element in yeast, confirming results by others (Kroj et al., 2003). Interestingly, removal of the acidic activation domain uncovers ABI3 DNA binding, in agreement with the autoinhibition hypothesis. However, further removal of conserved B1 and B2 domains abolishes DNA binding (Fig. 4b), consistent with the results of Kroj et al. (2003), suggesting that these domains are necessary for DNA binding by B3, at least in the yeast system. We confirmed that the B1B2B3 deletion mutant binds the RY element, because mutation of the RY sequence in the yeast lacZ reporter construct abolishes its activity (Fig. 4c), whereas mutation of a G-box internal to this composite element does not. Moreover, the B1B2B3 deletion mutant does not bind the napA ABRE in yeast (not shown).
Inhibition of ABI3 RY-mediated transactivation by staurosporine and by a seed-specific PP2C
Our studies in yeasts suggest that ABI3 DNA binding is regulated by allosteric autoinhibition, a process commonly controlled by phosphorylation/dephosphorylation. To establish whether ABI3 transactivation of RY requires phosphorylation, we assayed whether RY-dependent transcription by Arabidopsis ABI3 is sensitive to staurosporine, a broad-spectrum kinase inhibitor. Using agroinfiltration of N. benthamiana leaves, we show that a 5-h treatment with staurosporine reduces RY-mediated transcription by ABI3 significantly (3.3-fold), but not transcription of the 18S housekeeping genes (Fig. 5). In this experiment, ABI3 expression is supported by the 35S CaMV promoter, and we show that this expression is only slightly reduced by staurosporine (1.3-fold), and could therefore not explain the obtained reduction of ABI3-dependent RY-mediated GUS expression. Our results showing that the ABI3 transactivation of the RY element is inhibited by the kinase inhibitor staurosporine suggest that ABI3 DNA binding requires phosphorylation. We hypothesized that protein phosphatases expressed during late embryo maturation might dephosphorylate ABI3 and inhibit RY transactivation. Using AtGenExpress developmental microarray data, we established that out of the 67 PP2C-type phosphatases that we found annotated in the Arabidopsis genome, two were seed-specific and displayed a LEA-type expression pattern. One of them is AHG1, a PP2C that negatively regulates the ABA response during embryo maturation (Nishimura et al., 2007). We noted that MAT genes are highly expressed in germinating seeds of ahg1 (Nishimura et al., 2007; Fig. S2), suggesting that AHG1 negatively regulates MAT genes. To establish whether AHG1 is able to inhibit RY transactivation by ABI3, we coexpressed ABI3 and AHG1 together with the RY reporter by agroinfiltration in N. benthamiana leaves. We found that AHG1 specifically inhibits ABI3 transactivation of the RY element (Fig. 6a) as RT-PCR analysis showed no effect in ABI3 expression (Fig. 6b,c).
Repression of ABI3 RY-dependent transactivation by VAL1
Our results suggest that repression of the RY element is a passive process involving downregulation of ABI3 DNA binding by dephosphorylation. However, such a mechanism does not necessary preclude active repression by repressors binding the RY element. Here, VAL/HSI repressors are strong candidates for this role because homozygous val1/val2 double mutants express embryo maturation genes during early germination (Suzuki et al., 2007; Tsukagoshi et al., 2007). The VAL proteins could repress the MAT gene program by binding the RY element through their ABI3/VP1-type B3 domains, and effect gene silencing through their EAR domains. To assess whether targets of VAL repressors are mainly MAT genes, we examined the levels of relative expression of the MAT and LEA genes as previously defined (Table S1) in 5 dai seedlings of a double val1/val2 homozygous mutant using microarray data by Suzuki et al. (2007). We found that VAL repressors selectively inhibit MAT gene expression, but not LEA, because, in a homozygous double val1/val2 mutant, expression of MAT genes is induced 1800-fold, whereas LEA genes are only induced 20-fold (Fig. S3).
To establish whether VALs are able to silence ABI3 transactivation of the RY element, we coinfiltrated the RY reporter and the 35S::ABI3 and 35S::VAL1 effectors into leaves of N. benthamiana. Figure 7 shows that after 1 wk of ABI3 and VAL1 co-expression, ABI3 transactivation of the RY element is significantly downregulated by VAL1. Because VAL1 has a repression EAR domain, and VAL double mutants express MAT genes during seedling growth, we propose that VAL1 actively represses the RY reporter, rather than inhibiting ABI3 RY transactivation by passive squelching. Silencing of MAT genes during late embryo maturation could then be brought about by initial downregulation of DNA binding in ABI3, and subsequent active repression through VAL proteins binding the RY elements.
Analysis of DNA binding in B3 proteins by Y1H assay
Our data suggest a mechanism for repression of MAT genes through the RY element by initial passive repression through inhibition of ABI3 DNA binding by dephosphorylation, and subsequently active repression through VAL repressors binding the RY element. Central to this hypothesis is that VAL proteins bind the RY element, although there is no experimental evidence for this yet. In a Y1H screen using the RY element as bait, we never obtained any of the VAL proteins, (nor ABI3), though we did obtain FUS3 and LEC2 multiple times (unpublished), suggesting that the full-length VAL proteins fail to bind DNA in the Y1H assay (similar to ABI3). Alternatively, the EAR repression domain is able to cause active silencing of the reporter genes by recruiting the yeast histone deacetylase (HDAC) machinery, despite the protein being fused to the strong GAL4 activation domain. To assess whether the B3 domains of VAL proteins bind the RY element, and in order to avoid interference by eventual autoinhibition or repressor domains, we tested RY element binding by B3 domains of VAL1 (HSI2, AT2G30470) and VAL2 (HSL1, AT4G32010) in a Y1H assay, using domain swap chimeras of FUS3, as shown in Fig. 8a. The rationale of this experiment is that FUS3, being a B3 protein that is able to bind the RY sequence in yeast, may act as a protein scaffold for analysing RY binding of B3 domains from proteins that fail to bind DNA in the yeast system because of autoinhibition. Figure 8b shows that the B3 domain of VAL2, but not VAL1, binds the RY element when placed within the context of the FUS3 protein, confirming that VAL repressors can potentially bind the RY sequence in vivo. The FUS3/VAL1 chimera mediated only very weak growth on selective media, which was comparable with a FUS3 deletion lacking the B3 domain (not shown). The growth rate on selection media mediated by the FUS3/VAL2 chimera was comparable to the growth rates of the B1B2B3 deletion, and of a FUS3/ABI3 chimera, whereas the FUS3 protein mediated very strong growth on selection media.
The RY element mediates repression of MAT genes
Two major gene expression programs, called MAT and LEA, shape the mature angiosperm embryo. Despite numerous studies, the molecular mechanisms underlying phase transition between these two distinct gene expression waves remain unknown. Although it is well established that ABI3 regulates both programs, and that ABI5 is involved in transactivation of LEA-program genes by ABI3 and ABA (Brocard-Gifford et al., 2003), the mechanism of downregulation of MAT genes during late maturation is unknown. Our studies in tobacco show that the RY element mediates initial transcriptional activation during embryo mid maturation and subsequent repression during late maturation (Fig. 1) and early seedling growth (Fig. 2). We propose that the RY element mediates both initial activation and subsequent repression of the MAT gene program, and that transactivation of LEA genes is mediated mainly by the ABRE and does not require the RY element. Accordingly, MAT genes are ABA-responsive and RY-dependent, whereas LEA genes are ABA-responsive and RY-independent. Our hypothesis is in agreement with several independent lines of evidence. First, the RY element is not necessary for expression mediated by the wheat Em promoter in maize (Vasil et al., 1995), a typical LEA protein. Further, the maize vp1-McW mutant, which has a disrupted B3 domain in VP1, produces kernels that are colorless and quiescent (Carson et al., 1997 and references therein), in contrast to the colorless and viviparous kernels of vp1 mutants. The vp1-McW phenotype indicates that the B3-disrupted VP1 protein is unable to transactivate the RY-dependent C1 gene mediating anthocyanin production, but can still transactivate RY-independent LEA genes mediating dormancy, consistent with our hypothesis. Similarly, induction of Arabidopsis MAT genes CRC and At2S3 by ectopic ABI3 requires the B3 domain, whereas induction of LEA genes AtEm1 and AtEm6 is B3-independent (Kagaya et al., 2005). Finally, the RY element is over-represented in regulatory sequences from Arabidopsis MAT genes (Fig. 3).
ABI3 DNA binding may be developmentally regulated
Our study provides several lines of experimental evidence suggesting that ABI3 DNA binding is developmentally regulated. First, ABI3 is able to bind the RY sequence during mid maturation, but transactivation of the RY reporter is abolished during late maturation. The napA RY fragment used in our studies contains two RY elements flanking a G-box, and it is theoretically possible that the G-box contributes to repression of the RY reporter. However, G-boxes are equally frequent in MAT and LEA genes (Fig. 4), and therefore the RY element is more plausible as mediator of MAT repression. Here, it should be noted that the G-box in the RY fragment is not an ABRE, because it lacks the ABRE signature 5′-GC flanking sequence (Ezcurra et al., 1999; Choi et al., 2000) and has no ABRE activity (Ezcurra et al., 2000; Fig. 3). Although G-boxes are core elements in ABREs, not all G-boxes are ABREs, as G-boxes are ubiquitous elements mediating responses to diverse stimuli such as ABA, light and jasmonic acid (JA) (Kim et al., 1992; Terzaghi & Cashmore, 1995; Busk & Pagès, 1998; Menkens et al., 1995). Second, in a Y1H assay, ABI3 fails to bind the RY element, but this function is restored when the N-terminus is removed (Fig. 7), further supporting the notion that ABI3 DNA binding is regulated by a mechanism involving conformational autoinhibition. The full-length ABI3 construct mediates ABI3–ABI5 protein–protein interaction in a yeast two-hybrid assay (Nakamura et al., 2001), showing that ABI3 is stably accumulated in yeast strains harboring this construct. Further deletion of the B1 domain abolishes DNA binding, which is surprising, because the resulting protein fragment is structurally similar to FUS3. However, in transient assays, deletion of the B1 domain abolishes all ABI3 activities, including transactivation through the RY and ABRE elements (Mönke et al., 2004; our unpublished results), showing the crucial importance of B1 in ABI3 function and stability in vivo. The yeast data presented by us (Fig. 4) and by others (Kroj et al., 2003) are in stark contrast with in vitro results, where purified recombinant full-length ABI3, as well as the ABI3-B3 domain fragment (Suzuki et al., 1997; Mönke et al., 2004), bind the RY element. However, in vitro assay conditions may circumvent allosteric autoinhibition and promote DNA binding, as observed with recombinant mammalian proteins such as p53 and Elk-1, which bind DNA in in vitro assays but require phosphorylation in vivo (Takenaka et al., 1995; Yang et al., 1999). Using domain-swap chimeras, we also show that the ABI3 B3 domain binds the RY element when placed within the context of the FUS3 protein (Fig. 8), further supporting the notion that the ABI3 B3 domain is able to bind DNA in vivo when in a favorable protein conformation.
ABI3 DNA binding requires phosphorylation
In the transient expression system, staurosporine inhibits ABI3 transactivation through RY, indicating that ABI3 DNA binding requires phosphorylation. Consistent with these data, AHG1, a seed-specific LEA-program PP2C inhibits ABI3 transactivation of the RY reporter in N. benthamiana leaves. AHG1 is a seed-specific negative regulator of ABA response in Arabidopsis (Nishimura et al., 2007) and is functionally comparable to ABI1 and HAB1 (Schweighofer et al., 2004). AHG1 mutants have reduced dormancy and their germination is ABA hypersensitive, and seeds of ahg1-1 display enhanced expression of embryo maturation genes during germination in 0.5 µm ABA. Interestingly, MAT genes are more strongly upregulated by AHG1 disruption than LEA genes (Nishimura et al., 2007; Fig. S2). The increased upregulation of MAT genes in seeds of ahg1-1 cannot be explained by higher ABA response in these genes, because both MAT and LEA genes are equally sensitive to increased ABA in transient assays (Gampala et al., 2002). This unexpected phenotype suggests that AHG1 inhibits MAT genes during late maturation, when its expression peaks (Fig. S4). As shown in Fig. S2, enhanced upregulation of MAT genes is not observed in germinating seeds of ahg3, another PP2C mutant that has ABA-hypersensitive germination (Yoshida et al., 2006), indicating that enhanced upregulation of MAT genes is coupled to loss of AHG1 function and is not a general feature of the ABA-hypersensitive-germination phenotype. Another PP2C that negatively regulates the seed's ABA response, ABI1, is a negative regulator of seed dormancy but not of storage accumulation (Merlot & Giraudat, 1997). ABI1 exerts its action in the seed by negatively regulating transactivation of LEA-type genes by ABI5, as shown using a wheat Em reporter (Gampala et al., 2002), but is unable to inhibit ABI3 ABA-dependent transactivation of At2S3 (Parcy & Giraudat, 1997). Clearly, the different PP2Cs ABI1, AHG3 and AHG1 target different subsets of ABA-responsive genes, probably by targeting different transcription factors. However, the question whether ABI3 is a direct target of AHG1 awaits further studies.
The VAL repressors inhibit ABI3 transactivation by binding the RY element
Our results suggest that RY repression involves inhibition of ABI3 DNA binding through dephosphorylation by PP2Cs. However, VAL repressors could also mediate RY repression, because they have ABI3/VP1-like B3 domains, suggesting that they bind the RY sequence. Confirming this notion, VAL1 is able to repress ABI3 RY-mediated transactivation in N. benthamiana (Fig. 7), and the VAL2 B3 domain binds the RY element when placed within the context of FUS3 in a domain swap-Y1H setup (Fig. 8). The fact that the VAL1 B3 domain fails to bind DNA in this setup may be the result of an alteration of conformation caused by domain swapping, rather than inherent lack of DNA binding. Although domain swapping is a useful molecular biology tool, it is conceivable that individual domains may have co-evolved with their surrounding context and in some circumstances cannot fully replace each other. Alternatively, the sites of domain truncation may need optimization, because domain boundaries leading to stable protein fragments are difficult to predict, where even minor shifts of just a few amino acid residues may dramatically affect protein stability (Prodromou et al., 2007). Another possible explanation is that DNA binding by the VAL1 B3 domain requires conditions that are absent in yeast, such as post-transcriptional modifications or an interacting partner. Whatever the case, the fact that VAL1 represses ABI3 RY-mediated transactivation (Fig. 7), suggests that VAL1 binds the RY element in planta. Because VAL1 and VAL2 are functionally redundant and can replace each other (Suzuki et al., 2007; Tsukagoshi et al., 2007), our results showing that VAL2 binds the RY element and VAL1 mediates downregulation of ABI3 RY-dependent transactivation suggest that VAL proteins mediate repression of the MAT gene program during early seedling growth by binding the RY element. Consistent with this, in Arabidopsis protoplasts, VAL1 represses a 210-bp sugar-inducible sporamin promoter, which contains a prolonged RY element (TGCATGCATG; Tsukagoshi et al., 2005). Further, in a homozygous double val1/val2 mutant, expression of MAT genes is induced 1800-fold, whereas LEA genes are only induced 20-fold (Fig. S3; Suzuki et al., 2007), consistent with the RY element mediating repression by VALs mainly of MAT genes, but not of LEA.
A model for repression of MAT genes
Based on this, and previous studies (Ezcurra et al., 2000; Ng et al., 2006, Suzuki et al., 2007; Tsukagoshi et al., 2007), we propose a model for the dual regulation of MAT gene expression by the RY element in angiosperm orthodox seeds (Figs 9, 10). The model proposes that initial activation of MAT genes during mid-maturation is mediated by ABI3 binding of the RY element, most probably involving cooperation with FUS3 and LEC2 (Kroj et al., 2003), as well as recruitment of histone acetyl transferase (HAT) activity (Ng et al., 2006). MAT activation also requires that ABI3 mediates the ABA response through interaction with bZIP-type ABRE-binding factors (ABFs) (Bensmihen et al., 2002). The expression of LEA genes requires late maturation-specific ABI5 (Brocard-Gifford et al., 2003) and LEA genes are therefore not expressed during mid-maturation. As maturation ensues, the RY element becomes repressed, causing MAT gene expression to drop, because their expression is RY dependent. ABI5 expression commences at this stage, causing onset of the LEA gene program (Fig. 9). We propose that initial repression of the RY element during late maturation is caused by inhibition of ABI3 DNA binding through a mechanism involving either direct, or indirect, dephosphorylation, rather than by repression through VAL proteins (Fig. 10). Here, it could be argued that VAL repressors are the sole mechanism of RY-dependent downregulation of MAT genes from late maturation to early seedling growth. However, in double homozygous val1/val2 mutants, expression of the MAT genes AtS3 and OleS3 starts only after 5 dai and peaks at 7 dai (Tsukagoshi et al., 2007), indicating that repression by VALs starts well after germination, and cannot explain the repression of the RY element that we detected during late maturation (Fig. 1). The peak of AtS3 and OleS3 expression at 7 dai in val1/val2 temporally coincides with the repression of the RY reporter that we observe in seedlings (Fig. 2), consistent with a role for VALs in mediating repression of MAT genes during postgermination growth. Several lines of experimental evidence support our contention that ABI3 DNA binding is regulated by phosphorylation. First, both staurosporine and AHG1, a seed PP2C, inhibit ABI3 transactivation of the RY element in a transient assay, and AHG1 expression overlaps with RY-dependent repression in the embryo. Second, germinating seeds of ahg1 express MAT genes to a higher extent than LEA genes, suggesting that AHG1 inhibits MAT genes by a mechanism other than negative modulation of ABA signaling. Finally, ABI3 DNA binding is inhibited in yeast by a mechanism suggesting allosteric autoinhibition, consistent with ABI3 DNA binding requiring post-translational modification. Our model suggesting control of ABI3 DNA binding by phosphorylation/dephosphorylation should, nevertheless, be regarded as tentative because inhibition by staurosporine and AHG1 could involve another protein necessary for ABI3 function. After germination, the naked RY element becomes bound and actively repressed by VALs (Fig. 10), most probably involving recruitment of histone deacetylase (HDAC) activity, as shown with Arabidopsis Ethylene Response Factors (ERF) repressors, which are also EAR domain proteins (Song & Galbraith, 2006). Thus, our model suggests a repression mechanism for MAT genes involving initial passive repression by inhibition of transcription factor DNA-binding, and subsequent active repression through DNA binding by repressors, where the same cis-element, the RY sequence, mediates both passive and active repression (in addition to initial gene activation). In eukaryotic promoters, regulatory cis-elements frequently display dual regulatory roles and mediate both activation and repression of gene expression. A close example in plants is provided by AuxRE, a cis-element mediating auxin response through B3-domain ARFs, where ARFs can be activators or repressors of transcription, depending on the nature of a central protein domain (Ulmasov et al., 1999; Tiwari et al., 2003). A mechanism involving both passive and active repression may appear redundant, but it is now well established that these two mechanisms have distinct functions and are not redundant (Thiel et al., 2004). Accordingly, initial passive repression of a given gene or gene program gives an organism flexibility to rapidly reactivate gene transcription, whereas active repression produces a more permanent silencing that can eventually spread to adjacent regions by heterochromatin formation. In mammals, several studies report repression mechanisms involving initial passive repression and subsequent active repression mediated by a single regulatory element (Mailly et al., 1996; Lee et al., 1999; Nair et al., 2007).
In the case of embryo maturation, a mechanism involving initial passive repression by inhibition of ABI3 DNA binding, rather than VAL repressors displacing ABI3 by competition for binding sites, makes sense because during late maturation the levels of ABI3 mRNA are significantly higher than those of VALs (Fig. S4), and since their DNA-binding activities appear to be similar in the yeast assay, it is hard to imagine how VALs could compete with ABI3 for RY occupancy.
Our data showing that RY-mediated repression of MAT genes extends beyond late maturation over germination and early seedling growth suggests a mechanism for seedling adaptation to ABA-mediated abiotic stress. Here, enhanced ABA signaling activates ABI3 expression during germination (Lopez-Molina et al., 2002) and early seedling growth (Teaster et al., 2007; Shobbar et al., 2008), leading to selective activation of LEA genes, and concomitant enhanced stress tolerance plus growth inhibition under adverse conditions. In this scenario, it is crucial for the arrested seedling that storage accumulation does not resume, as it would sequester available nutrients indispensable for rapidly resuming heterotrophic growth upon normal conditions. Interestingly, after the transition from heterotrophic to autotrophic metabolism, MAT silencing is ended, suggesting that storage synthesis is less detrimental in the autotrophic phase, or that plantlets lack ABI3 induction, making RY silencing redundant.
We thank Dr Ruth Finkelstein for generously providing the pGAD-ABI3 plasmids, and anonymous referees for constructive suggestions to improve the paper. This work was supported by grants from the Swedish Research Council Formas and The Nilsson-Ehle Fund.