Arabidopsis NF-YB subunits LEC1 and LEC1-LIKE activate transcription by interacting with seed-specific ABRE-binding factors

Authors


*(fax +81 52 789 5214; e-mail hattori@agr.nagoya-u.ac.jp).

Summary

LEAFY COTYLEDON 1 (LEC1) plays vital roles in the regulation of seed maturation in Arabidopsis. LEC1 encodes a homolog of yeast HAP3 or mammalian NF-YB/CBF-A subunit of trimeric CCAAT binding factor (CBF). Among the nine paralogs of NF-YB in Arabidopsis, LEC1-LIKE (L1L) is most closely related to LEC1, and can complement the lec1 mutation when expressed under the control of the LEC1 promoter. Although the nature of the B3-type seed maturation regulators as transcription factors have been investigated, knowledge of the molecular action of LEC1 is limited. When co-expressed with NF-YC2 in the presence of ABA, we found that LEC1 or L1L, but not other NF-YBs, activated the promoter of CRUCIFERIN C (CRC), which encodes a seed storage protein. However, additional expression of an NF-YA subunit interfered with the activation. The LEC1/L1L-[NF-YC2] activation depended on ABA-response elements present in the promoter, which led to the finding that LEC1/L1L-[NF-YC2] can strongly activate the CRC promoter in the absence of ABA when co-expressed with a seed-specific ABA-response element (ABRE)-binding factor, bZIP67. Functional coupling of LEC1/L1L-[AtNF-YC2] and bZIP67 was also observed in the regulation of sucrose synthase 2 (SUS2). Immunoprecipitation experiments revealed that L1L and bZIP67 formed a protein complex in vivo. These results demonstrate a novel plant-specific mechanism for NF-Y subunit function that enables LEC1 and L1L to regulate a defined developmental network.

Introduction

Embryogenesis in higher plants is a developmental process that not only establishes the basic body plan, but also produces seed, a strategy that secures progeny by maintaining a quiescent state under unfavorable conditions, and resuming growth under favorable conditions. Seed characteristics, such as storage of proteins and other nutrients, desiccation tolerance and dormancy, are conferred during the maturation phase, which corresponds to the latter half of embryogenesis after the completion of morphogenesis. In Arabidopsis, four genes, LEAFY COTYLEDON 1 (LEC1), LEC2, FUSCA 3 (FUS3) and ABSCISIC ACID INSENSITIVE 3 (ABI3), have been found to play central roles in the regulation of embryo maturation (Giraudat et al., 1992; Keith et al., 1994; Lotan et al., 1998; Luerssen et al., 1998; Meinke, 1992; Meinke et al., 1994; Stone et al., 2001; West et al., 1994). Mutation in any of these genes leads to defects in broad aspects of the maturation process, and thus they are regarded as master regulators of seed maturation; notably, however, these genes do not function exclusively during maturation (Vicente-Carbajosa and Carbonero, 2005). The lec1, lec2 and fus3 mutants, which are grouped as lec-type mutants, differ morphologically from the abi3 mutant in that they exhibit leafy traits in the cotyledons, which are typically recognized by the presence of trichomes on the cotyledons (Baumlein et al., 1994; Keith et al., 1994; Meinke, 1992; Meinke et al., 1994; West et al., 1994). abi3 mutant seeds also display unique phenotypes, not observed for the lec-type mutants, such as failure in chlorophyll degradation and ABA-insensitivity upon germination (Finkelstein and Somerville, 1990; Koornneef et al., 1984; Nambara et al., 1995, 1992; Parcy et al., 1997, 1994).

FUS3, LEC2 and ABI3 encode transcription factors containing a B3 DNA binding domain, which was originally found in VP1, the maize ortholog of ABI3 (Giraudat et al., 1992; Lotan et al., 1998; Luerssen et al., 1998; McCarty et al., 1991; Stone et al., 2001). The B3 domain is a plant-specific DNA-binding domain present in highly diverged forms, such as those in auxin response element binding factors (ARFs), VERNALIZATION RESPONSE 1 (VRN1), RAV family proteins and VAL/HIS2 family proteins, in addition to FUS3, LEC2 and ABI3 (Kagaya et al., 1999; Levy et al., 2002; Suzuki et al., 2007; Tsukagoshi et al., 2007; Ulmasov et al., 1997). However, the B3 domains of these three master regulators of seed maturation are highly conserved, and are suggested to bind a common DNA sequence element, referred to as both RY or Sph, where they have been proposed to play partially redundant roles in seed maturation (Kroj et al., 2003; Mönke et al., 2004; Reidt et al., 2000; Suzuki et al., 1997). Nevertheless, they also play distinct roles, as revealed by differences in the aforementioned morphological phenotypes of the mutants, presumably through the functions of domains other than B3, and via distinct spatial and temporal patterns of expression. The expression of these B3-containing factors is under complex mutual as well as self regulation, which occurs locally (To et al., 2006), and thus elucidating the overlapping and unique functions of these proteins appears challenging. The transcript levels of 12S globulin (Cruciferin A–C, CRA–CRC) and 2S (At2S1-4) seed storage protein genes are reduced in mutants of these genes (Baumlein et al., 1994; Kroj et al., 2003; Lotan et al., 1998; Meinke et al., 1994; Nambara et al., 1992; Parcy et al., 1997, 1994). The regulation is most likely mediated by direct interaction between these B3 factors and RY elements in the promoters of storage protein genes (Kroj et al., 2003; Mönke et al., 2004; Reidt et al., 2000). However, indirect regulation, via intermediate transcription factors under the control of FUS3 and/or ABI3, has also been suggested (Kagaya et al., 2005a).

LEC1 encodes a homolog of yeast HAP3 and the mammalian NF-YB/CBF-A subunit, which together with HAP2 and HAP5, or NF-YA/CBF-B and NF-YC/CBF-C, respectively, constitutes the trimeric CCAAT binding factor (CBF) (Li et al., 1992; Lotan et al., 1998). LEC1 regulates the expression of FUS3 and ABI3 (and LEC2, although somewhat differently), based on the observed reduction in their expression in lec1 mutant seeds and their induction upon ectopic LEC1 expression (Kagaya et al., 2005b). Therefore, the LEC1 regulation of the expression of seed storage protein genes and other downstream target genes is considered to be, at least in part, mediated by the B3-containing regulators. However, knowledge about the molecular mechanism of transcriptional regulation by LEC1 is relatively limited compared with what is known for the B3 regulators.

Unlike fungi and mammals, the homologs for each subunit of the CBF complex are encoded by gene families in higher plants. The Arabidopsis genome encodes 10 each of NF-YA and NF-YB, and nine NF-YC paralogs (Edwards et al., 1998; Gusmaroli et al., 2002). An examination of amino acid substitution rates in plants has revealed evidence for the asymmetric evolution of certain duplicates of NF-YB and NF-YC, which appears to be coupled to asymmetric divergence in gene function (Yang et al., 2005). Besides LEC1, several other NF-Y subunits have been implicated in specific developmental processes or responses to developmental cues, including nodule formation, flowering time regulation, ABA/blue light response and chloroplast development (Ben-Naim et al., 2006; Cai et al., 2007; Combier et al., 2006; Kusnetsov et al., 1999; Miyoshi et al., 2003; Nelson et al., 2007; Warpeha et al., 2007; Wenkel et al., 2006). These observations indicate that distinctions exist in molecular functions among the NF-Y subunit family members, which contrasts with the paradigms in fungi and mammals. One likely mechanism that distinguishes the functions of the NF-Y subunit members is through specific interacting partners. Interactions between the NF-Y complex and a second transcription factor(s) have been reported. A rice MADS-box protein, OsMADS18, either with or without a heterodimerizing MADS-box partner, interacts with OsNF-YB1 (Masiero et al., 2002). Arabidopsis CONSTANS and CONSTANS-LIKE proteins, through their plant-specific CCT domain, interact with a complex of NF-YB and NF-YC (Wenkel et al., 2006). Similarly, in tomato, NF-YC interacts with the CCT domain of group-1A COL proteins (Ben-Naim et al., 2006). However, the specificities or preferences among the different members of the NF-YB/C subunit families in these interactions are not evident, or have not been examined. On the other hand, a specific amino acid residue has been identified in LEC1 and LEC1-LIKE (L1L) that differentiates their functions from those of other NF-YB members (Kwong et al., 2003; Lee et al., 2003). Arabidopsis L1L is the most closely related NF-YB member to LEC1, and is expressed preferentially during seed development (Kwong et al., 2003). As L1L can complement the lec1 mutation when it is expressed under the control of the LEC1 promoter, the molecular function of the two NF-YBs, LEC1 and L1L, is considered to be redundant (Kwong et al., 2003).

In this study, we show that LEC1 and L1L can activate the CRC and SUCROSE SYNTHASE 2 (SUS2) promoters in combination with an NF-YC subunit, through the interaction with a seed-specific ABA-response element (ABRE) binding bZIP factor, bZIP67. This activation was specific to LEC1 and L1L, and was inhibited by the expression of NF-YA subunits. These results shed further light on the plant-specific molecular mechanisms through which different members of the NF-Y subunit families play distinct roles in plant development.

Results

The CRC promoter is activated by LEC1 and L1L, in combination with an NF-YC subunit, in an ABA-dependent manner

Although the roles of LEC1 in embryo development have been investigated in many studies by genetic approaches, little is known about the nature of LEC1 and the molecularly redundant NF-YB subunit L1L as transcription factors. To gain insight into the molecular mechanisms of the two NF-YB subunits implicated in seed development, we examined the promoter of CRC, a seed maturation-specific gene, which encodes a seed storage protein (12S globulin) for its regulation by LEC1 and L1L.

A promoter-GUS reporter gene, CRC(-730)-GUS, which contains the GUS coding region and the CRC promoter from −730 to +81 bp, was tested for its regulation by LEC1 and L1L with transient co-expression experiments using Arabidopsis T87 cultured cell protoplasts. LEC1 or L1L alone had no effect on the CRC promoter, but when co-expressed with an NF-YC subunit, NF-YC2, they activated the promoter in the presence, but not in the absence, of ABA (Figure 1a). The very similar responses of the reporter to LEC1 and L1L confirmed the redundancy of their molecular function. The observed activities of LEC1 and L1L appeared to be specific to these two NF-YB members, because no CRC promoter activation was evident for a distinct NF-YB, NF-YB2 (Figure 1a).

Figure 1.

 L1L and LEC1, but not NF-YB2, can activate the CRC promoter in combination with NF-YC.
The specificities of NF-YB subunits (a) and NF-YC subunits (b), or the requirements of NF-YA subunits (c), were tested for the activation function of the CRC promoter.
T87 cultured cell protoplasts were transfected with the CRC promoter-GUS reporter plasmid [CRC(-730)-GUS] and the effector plasmids for the indicated transcription factors. Transfected cells were cultured for 16 h without or with 10 mm ABA, and were assayed for GUS and luciferase activities. The averages of GUS activities normalized by LUC activity obtained from triplicate transfections are indicated with SD. Treatments with the same protoplast preparation are given in each panel. ‘Control’ indicates transfection with the reporter plasmid and an empty expression vector alone, instead of an effecter plasmid. The empty vectors were also included in other transfections to keep the total level of transfected DNA constant.

We also tested whether other NF-YC subunits could function similarly to NF-YC2 in combination with LEC1/L1L to activate the CRC promoter. NF-YC1 and NF-YC6, both of which belong to distinct phylogenetic clades (Edwards et al., 1998; Yang et al., 2005), activated the CRC promoter when co-expressed with L1L, suggesting no strong preference for the NF-YC partner (Figure 1b).

Because the effect of LEC1/L1L on the CRC promoter was strictly dependent on the co-expression of an NF-YC subunit, we next examined the requirement for an NF-YA subunit. Surprisingly, but interestingly, co-expression of any of the four distinct NF-YA subunits strongly inhibited the L1L-[NF-YC2]-mediated activation of the promoter (Figure 1c), suggesting that the L1L-[NF-YC2] dimer functions by interacting with a protein factor(s) other than NF-YA, with which these co-expressed NF-YAs compete. Alternatively, a specific endogenous NF-YA subunit(s) other than those tested may form a functional trimer with L1L and NF-YC2.

Activation of the CRC promoter by LEC1/L1L and NF-YC depends on ABREs

As shown above, the activation of the CRC promoter required exogenously added ABA. Thus, we examined whether the ABA-dependent activation by LEC1/L1L-[NF-YC2] required an ABA-response element (ABRE), as in typical ABA-induced promoters (Figure 2). The −730-bp CRC promoter has three potential ABRE sequences, one non-ACGT type (ABRE3 at −575 bp), and two with the ACGT core (ABRE2 at −103 bp; ABRE1 at −58 bp). When base-change mutations were introduced in ABRE3, or ABRE2 and ABRE1 at the same time, the ABA-dependent activation by L1L-[NF-YC2] was greatly reduced. When all three ABREs were mutated, no activation was observed. These results indicate that the ABREs are essential for the LEC1/L1L-[NF-YC2]-mediated activation of the CRC promoter.

Figure 2.

 Activation of the CRC promoter by L1L and NF-YC2 depends on ABA-response elements (ABREs).
The CRC promoter-GUS reporters with base-change mutations at ABREs were tested for the activation by L1L and NF-YC2, as described in Figure 1. The wild-type and mutant forms of the −730-bp promoters in the reporter plasmids are schematically indicated on the left. Hatched and filled boxes indicate ABREs and the CCAAT box, respectively. Mutations at ABREs are indicated by X.

LEC1/L1L and NF-YC2 strongly activate the CRC promoter in the absence of ABA when combined with a seed-specific ABRE-binding factor

Based on the result that the ABA-dependent LEC1/L1L-[NF-YC2] activation of the CRC promoter required ABREs, the effects of ABRE-binding factors were investigated. We first tested bZIP67 in this regard because its expression is highly seed specific, and is under the control of FUS3 (Jakoby et al., 2002). Expression of bZIP67 alone did not activate the CRC promoter, either in the absence or presence of ABA. However, when it was co-expressed together with LEC1/L1L-[NF-YC2], a more than 20-fold activation was observed, even in the absence of ABA (Figure 3a). In this case again, LEC1 and L1L exhibited similar activities. These levels of activation were much greater than those achieved only by LEC1/L1L-[NF-YC2] in the presence of ABA. The high-level activation of the CRC promoter by L1L-[NF-YC2], in combination with bZIP67, was again strictly dependent on the ABREs (Figure 4a, b). These results indicate that LEC1/L1L-[NF-YC2] acts on the ABREs through the ABRE-binding factors bound to them. Combination of NF-YB2 and NF-YC2 activated the promoter in the presence of bZIP67 only to a level significantly lower than that observed for LEC1/L1L-[NF-YC2] (data not shown).

Figure 3.

 L1L/LEC1, NF-YC2 and ABA-response element (ABRE)-binding factor in combination can strongly activate the CRC promoter without ABA.
(a) The effects of an ABRE-binding factor bZIP67 on L1L/LEC1-[NF-YC2] activation of the CRC promoter.
(b) The rapid activation (at 3 or 6 h) of the CRC promoter by L1L/LEC1-[NF-YC2] and bZIP67.
(c) Comparison among the ABRE-binding factors for the CRC promoter activation.
The details of presentation and experiments are as described in Figure 1.

Figure 4.

 The strong activation of the CRC promoter by the combination of L1L, NF-YC2 and bZIP67 depends on ABA-response elements (ABREs).
The ABRE3 (a) or ABRE1/2/3 mutants (b) of the −730-bp CRC promoter shown in Figure 2 were tested for activation with the combination of L1L, NF-YC2 and bZIP67. The details of the presentation and experiments are as in Figures 1 and 2.

In standard protoplast assays, cells were cultured for 16 h after transfection. However, even when the culture period was shortened to 3 h, similarly high levels of activation were observed upon co-expression of L1L-[NF-YC2] and bZIP67 (Figure 3b). Such early effects of L1L-[NF-YC2] and bZIP67 strongly suggest direct action of all these transcription factors on the CRC promoter, rather than activation through the induction of secondary transcription factors.

Three other ABRE-binding bZIP factors, bZIP12/EEL, ABI5 and AREB1, were also tested (Figures 3c and S1). bZIP12 is another seed-preferentially expressed ABRE-binding factor with similar expression kinetics to that of bZIP67, and is implicated in the fine-tuning of a late embryogenesis abundant protein gene, AtEm1 (Bensmihen et al., 2002). bZIP12 activated the promoter to a very small extent, if any, by itself, even in the presence of ABA, but when combined with L1L and NF-YC2, a very high level of activation was observed, irrespective of the presence or absence of ABA (Figure S1). Thus, bZIP12 acted on the CRC promoter in a very similar manner to bZIP67. ABI5, which functions mainly at late embryogenesis and seed germination, or early seedling stages (Finkelstein and Lynch, 2000; Lopez-Molina et al., 2001), alone activated the CRC promoter to some extent in an ABA-dependent manner, and to a greater extent when it was co-expressed with L1L and NF-YC2. The levels of the reporter gene expression in the presence of ABA were comparable with those observed with bZIP67 and LEC1/L1L-[NF-YC2], although more ABA dependence was associated with the activation by the combination of ABI5 and LEC1/L1L-[NF-YC2]. AREB1 is thought to function in ABA-induced transcription mainly in vegetative tissues (Fujita et al., 2005). In contrast to the three seed-preferentially expressed ABRE-binding factors, AREB1 alone activated the CRC promoter to a high level in an ABA-dependent manner. The co-expression of L1L-[NF-YC2] increased the AREB1 activation by twofold. Some activation of the CRC promoter by AREB1 was observed in the absence of ABA, which was enhanced by L1L-[NF-YC2], but not as prominently as in the case of the activation by bZIP67. Thus, any of the ABRE-binding factors tested functionally interacted with L1L-[NF-YC2]. However, bZIP67, an ABRE-binding factor that primarily functions in seed development, was distinct from the others: whereas other ABRE-binding factors could activate the promoter by themselves in the presence of ABA, bZIP67 essentially required L1L-[NF-YC2] to function. Such a strong dependence of bZIP67 on L1L-[NF-YC2] indicates the importance of the observed interaction between LEC1/L1L and the ABRE-binding factor for gene regulation in developing seeds.

Activation of the CRC promoter by L1L-[NF-YC2] does not require a CCAAT Box

It is well established in fungi and mammals that NF-Y trimer complexes bind to the CCAAT box. The −730-bp promoter contained three CCAAT sequences. The most upstream one lies between −730 and −724 bp, but this CCAAT sequence did not function in the activation by L1L-[NF-YC2], because the deletion to −724 bp did not affect the activation (Figure 5a). The other two CCAAT boxes were then mutated in the −724-bp context to test whether they were required for the activation by L1L-[NF-YC2] and bZIP67. However, a high level of activation in terms of –fold activation was retained, although the basal promoter activity diminished (Figure 5b). These results indicated that the CCAAT boxes do not participate in the LEC1/L1L-[NF-YC2]-mediated activation of CRC.

Figure 5.

 The activation of the CRC promoter by the combination of L1L, NF-YC2 and bZIP67 does not depend on CCAAT boxes.
(a) The effect of the deletion of the 5′-most CCAAT box on the L1L-[NF-Y2C] activation of the CRC promoter.
(b) The effect of base-change mutations in the CCAAT boxes on L1L-[NF-Y2C]-bZIP67-mediated activation.
The details of the presentation and experiments are as described in Figures 1 and 2.

Activation of the CRC promoter by L1L-[NF-YC2] through ABREs is context dependent

A GUS reporter construct, 4XABRE-GUS, which was directed by four repeated ABREs fused to the 35S minimal (−46-bp) promoter, was tested to determine whether the presence of ABREs was sufficient for the activation (Figure 6). bZIP67 alone activated this synthetic promoter, indicating that bZIP67 could act on the ABREs in this promoter. However, no further activation was observed by adding L1L-[NF-YC2]. Therefore, the functional interaction between the NF-Y subunits and the bZIP factor through ABREs was context dependent with regard to some other cis-element(s) or the position of ABREs relative to the core promoter.

Figure 6.

 L1L/LEC1 and NF-YC2 cannot activate a synthetic promoter essentially composed of ABA-response elements (ABREs).
The promoter portions of the reporter genes are schematically illustrated on the left. The ‘pro35S’ indicates the minimal promoter of cauliflower mosaic virus 35S promoter up to −46 bp, relative to the transcription start site. The ‘4xABREs’ indicates the synthetic sequence comprising four repeated ABREs. The details of the presentation and experiments are as described in Figures 1 and 2.

LEC1/L1L-[NF-YC2] functions with bZIP67 in a protein complex

To test whether the functional interaction between LEC1/L1L-[NF-YC2] and bZIP67 reflected a physical interaction, immunoprecipitation experiments were performed (Figure 7). Myc-epitope-tagged L1L was expressed in T87 protoplasts, either without or with HA-epitope-tagged bZIP67. The lysates of the protoplasts were subjected to immunoprecipitation with anti-HA antibody agarose beads, and were then analyzed by immunoblotting with anti-myc antibodies. Myc-tagged L1L was detected in the immunoprecipitates only when HA-tagged bZIP67 was co-expressed, suggesting that bZIP67 and L1L formed a complex in vivo. The interaction was observed in the absence of co-expressed NF-YC2, and the additional expression of NF-YC2 reduced the quantity of L1L recovered in the immunoprecipitates. Thus, NF-YC2 apparently competed with L1L for bZIP67. This observation was unexpected, because L1L, NF-YC2 and bZIP67 function in concert to activate the target promoter, from which a stable trimeric complex would have been anticipated. The observation may be explained if one assumes that such a complex formation is only possible or maintained stably in the presence of target DNA. The observed interaction between L1L and bZIP67 was likely to have occurred mostly free from DNA because they were overexpressed. Without target DNA, L1L and NF-YC2 may form a complex in which bZIP67 is less accessible, and is inclined to form a conventional trimeric complex with pre-existing NF-YA subunits. Despite such unsolved questions, the results clearly demonstrated the ability of L1L to interact physically with bZIP67 in plant cells.

Figure 7.

 L1L forms a complex with bZIP67 in plant cells.
T87 cultured cell protoplasts were transfected with the expression plasmids for the indicated combination of the transcription factors. The expressed proteins were immunoprecipitated with anti-hemagglutinin (HA) epitope conjugated Sepharose beads, and then analyzed by immunoblotting with anti-myc-epitope. LEC1, L1L, NF-YB2, NF-YC2 and bZIP67 had myc and HA epitope tags, respectively, at the C terminus. Lanes under ‘Input (1/00)’ were loaded with 1% of the total protein subjected to immunoprecipitation. The origins of the doublet bands of L1L-myc and NF-YC2-myc are not known. *Immunoglobulin light chains that escaped from the antibody beads.

The expression of L1L during seed development, and the effects of T-DNA insertion mutation at L1L

Although the expression of CRC is known to be significantly decreased in the lec1 mutant seeds, the apparent downregulation of CRC in the mutant does not necessarily indicate a direct role for LEC1 in the regulation of CRC. The overall effect of the lec1 mutation on the expression of CRC is more likely to be conveyed through the LEC1 regulation of FUS3 or the B3 networks (Kagaya et al., 2005b; To et al., 2006). It could also result from developmental disturbances of the mutant. Nevertheless, the high levels of the CRC promoter activation, as well as other data shown above, strongly suggest direct involvement of LEC1 and L1L in the regulation of CRC during seed development.

We then asked whether CRC expression was affected in an l1l T-DNA insertion line, l1l-1 (SALK_118236; Figure 8a). In l1l-1, the L1L transcript probed with a sequence upstream of the T-DNA insertion was almost undetectable (Figure 8b). In addition to the structural disruption of the gene, the extreme reduction of the transcript indicated that l1l-1 is a null allele. Unlike the reported embryo-lethal phenotype of L1L RNAi lines (Kwong et al., 2003), l1l-1, as well as another T-DNA insertion allele, l1l-2, did not have apparent altered phenotypes during seed development. Interestingly, the expression of L1L was found to be greatly reduced in lec1 mutant seeds (Figure 8b). This reduction of L1L mRNA may be, at least in part, the result of the regulation of FUS3 by LEC1 (Figure S2; Kagaya et al., 2005b). The lack of remarkable phenotypes in the l1l mutant may be explained by the presence of the normal level of molecularly redundant LEC1. In contrast, the obvious defects in embryo development in the lec1 mutant could result from the greatly reduced level of the molecularly redundant NF-YB subunit.

Figure 8.

 Effect of l1l and lec1 mutations on the expression of L1L and SUS2.
(a) T-DNA insertion alleles of L1L, designated l1l-1 and l1l-2, are schematically illustrated. Boxes indicate exons, and the included black portions indicate protein coding regions.
(b–e) Relative L1L (b, d) and SUS2 (c, e, f) transcript levels in developing seeds at the indicated days after flowering (DAF) from wild-type (Col-0) plants or plants homozygous for the indicated mutant alleles. The relative transcript levels were determined by real-time RT-PCR. Data with error bars are means ± SD of triplicate seed samples.

The expression of CRC tended to be decreased in l1l-1 compared with the wild type at 8 days after flowering (DAF): in the lec1-1 l1l-1 double mutant, a further reduction of CRC mRNA was observed (Figure S3). However, the time-course experiment between 7 and 10 DAF failed to show a statistically significant decrease in the CRC mRNA levels because of the large variations among repetitive samples, probably resulting from the steep rise in the expression levels at the early phase of CRC induction (Figure S3). Although the genetic demonstration of the participation of L1L in the CRC regulation in planta was not conclusive, it was still likely, considering the result with the double mutant.

To verify the functional coupling of LEC1/L1L and bZIP67 in transcriptional regulation of other genes, which could be confirmed more convincingly to be under the control of LEC1/L1L in developing seeds, we searched for potential L1L target genes for which expression was affected in l1l mutant seeds. As the expression of L1L mRNA was found to be highest at 8 DAF, and declined thereafter (Figure 8d), the effect of the l1l mutation on gene expression was expected to be most pronounced around this stage. We first searched for seed-specific genes in which the temporal pattern of expression was similar to that of L1L, using public array databases with the aid of the ATTED II web site (http://atted.jp (Obayashi et al., 2007). We also performed a preliminary microarray analysis without biological replication. Among several candidate genes picked up by these analyses, the expression level of a sucrose synthase gene, SUS2, was significantly reduced (30–50%) in l1l-1 developing seeds compared with the wild type from 8 to 10 DAF (Figure 8e). A further striking reduction of SUS2 expression was observed in the l1l-1 lec1-1 double mutant (Figure 8b). The reduction of the expression of SUS2 at 8 DAF was also confirmed with the l1l-2 mutant allele (Figure 8f). These results strongly suggest that SUS2 is regulated additively by the molecularly redundant NF-Y2B subunits in developing seeds.

SUS2 is regulated by L1L via a similar mechanism to that for CRC

Having confirmed the in planta role of L1L in the regulation of SUS2, we asked whether the SUS2 promoter was regulated by L1L and bZIP67. The expression of the SUS2 promoter-GUS reporter [SUS2(-620)-GUS] was not activated by L1L, NF-YC2 or bZIP67 alone, whereas the simultaneous expression of L1L and NF-YC2 activated the SUS2 promoter (Figure 9). The co-expression of bZIP67 with L1L-[NF-YC2] boosted the activation dramatically (Figure 9). Therefore, the promoter of SUS2, the expression of which was reduced in the l1l-1 developing seeds, is regulated by L1L-[NF-YC2] and bZIP67 in a manner very similar to that of CRC. The only qualitative difference was that the activation by L1L-[NF-YC2] occurred without ABA. This might be partly because the basal activity of the SUS2 promoter without effectors was repressed by ABA.

Figure 9.

 The L1L, NF-YC2 and ABA-response element (ABRE)-binding factor in combination can strongly activate the SUS2 promoter.
The details of the presentation and experiments are as described in Figures 1 and 2. The promoter portion of the SUS2(−620)-GUS reporter gene is schematically illustrated on the left. ABRE1 (GCCAAGTGTC) and ABRE2 (AACGTGGC) in this promoter are non-ACGT and ACGT-type ABREs, respectively.

Discussion

As mentioned above, in mammals and fungi, each subunit of the trimeric CCAAT binding factor/NF-Y is encoded by a single gene, but they are encoded by gene families in higher plants. The Arabidopsis genome encodes 10 NF-YA, 10 NF-YB and nine NF-YC paralogs (Edwards et al., 1998; Gusmaroli et al., 2002). At least some plant NF-YA and NF-YB subunits have been shown to complement yeast mutants lacking the corresponding NF-Y subunit, and to be able to form a trimeric complex with mammalian cognate subunits in vitro (Ben-Naim et al., 2006; Edwards et al., 1998; Masiero et al., 2002). These observations indicate that plant NF-Y subunits can constitute a trimeric complex that can function as a CCAAT binding factor, and may act as a general transcription factor, as in mammals. However, some plant NF-Y subunits function in specific developmental processes, or in response to environmental stimuli (Ben-Naim et al., 2006; Cai et al., 2007; Combier et al., 2006; Kusnetsov et al., 1999; Lotan et al., 1998; Miyoshi et al., 2003; Nelson et al., 2007; Warpeha et al., 2007; Wenkel et al., 2006). A well-established example is LEC1, which specifically controls embryo development (especially maturation). LEC1 plays specialized roles not only through its developmentally regulated expression but also by its distinct molecular activity, because the in vivo function of LEC1 cannot be replaced by other NF-YB subunits except for the most closely related L1L (Kwong et al., 2003; Lee et al., 2003). Our results provide a deeper insight into how such molecular and biological distinctions among the NF-YB paralogs are achieved.

In this study, we demonstrated that a plant seed-specific NF-YB subunit, either LEC1 or L1L, but not others, was able to function as a potent activator of the promoter of the CRC seed storage protein gene, as well as SUS2, when it was co-expressed with NF-YC2 or other NF-YCs. The ABA-dependent activation of the CRC promoter by LEC1/L1L-[NF-YC2] led us to the identification of ABRE as a necessary cis-element. Furthermore, we showed that a seed-specific ABRE-binding factor, bZIP67, functionally interacted with LEC1/L1L-[NF-YC2] to strongly amplify the activation, whereas bZIP67 itself was unable to activate the promoter. Additionally, the results of the co-immunoprecipitation experiment strongly suggest that this functional interaction is based on a physical interaction between the bZIP factor and the LEC1/L1L-[NF-YC2] complex. Another seed-specific ABRE-binding factor, bZIP12, also behaved very similarly to bZIP67 in combination with LEC1/L1L-[NF-YC2] in the activation of the CRC promoter. Although detailed analyses were only performed with bZIP67, the following discussions referring to bZIP67 may also apply to bZIP12.

In the presence of ABA, LEC1/L1L-[NF-YC2] was able to activate the CRC promoter by several fold without bZIP67. An ABRE-binding factor(s) endogenous to the protoplasts probably participated in the ABA-dependent activation in the absence of bZIP67. The activity of the endogenous ABRE-binding factor(s) may be regulated by ABA signaling, most probably via phosphorylation (Kagaya et al., 2002; Kobayashi et al., 2005). The apparent ability of LEC1/L1L-[NF-YC2] to confer ABA responsiveness to the promoter may be related to the ABA-insensitive germination phenotype of lec1 seeds (West et al., 1994). However, considering the developmental alteration and the reduced expression of ABI3 in the mutant (Kagaya et al., 2005b), this possibility is an open question. More importantly, in the absence of ABA, bZIP67 enabled LEC1/L1L-[NF-YC2] to activate the CRC promoter through the ABREs to a high level – nearly comparable with that in the presence of ABA. This indicates that an ABRE can function as a cis-element that can mediate developmental rather than hormonal signals in certain promoter contexts. The tight dependence of bZIP67 on LEC1/L1L-[NF-YC2] also strongly suggests that bZIP67 is a critical component for the function of LEC1/L1L-[NF-YC2].

The co-expression of several of the NF-YA subunits we tested strongly interfered with the CRC promoter activation by LEC1/L1L-[NF-YC2]. Although the possibility that LEC1/L1L forms a trimeric complex with an NF-YA and NF-YC subunit, as occurs in mammals and fungi, cannot be excluded, this result rather suggests that LEC1/L1L and NF-YC2 form a functional complex without an NF-YA subunit, which may compete for one of the components in the complex, which is most likely to be an ABRE-binding factor. An interesting future problem will be to address whether the observed competitive activity of NF-YA on LEC1/L1L-[NF-YC2] function implies a regulatory role for NF-YA in this regard. An example for the function of NF-YB and NF-YC in a complex without an NF-YA subunit has recently been reported in Arabidopsis. Several NF-YB and NF-YC subunits have been demonstrated to interact with the CCT domain-containing zinc finger proteins, CONSTANS (CO) and CONSTANS-LIKE 15 (COL15) (Wenkel et al., 2006). This interaction has been proposed to be based on the structural similarity between NF-YA and the CCT domain. CO or COL15 is considered to form a complex with NF-YB and NF-YC by competing with NF-YA through the CCT domain. In fact, the overexpression of an NF-YA subunit represses the CO function, resulting in late flowering (Wenkel et al., 2006). The inhibitory activity of NF-YA on LEC1/L1L-[NF-YC2] function may be an indication of a similar manner of interaction, although not through the CCT domain. Alternatively, the ABRE-binding factor may interact with a CCT-containing protein in a complex comprising LEC1/L1L and NF-YC2. However, the functional specificity of LEC/L1L among the NF-YB subunits contrasts with the case in the interaction with CCT-containing proteins.

The disruption of all of the CCAAT boxes in the CRC promoter did not cause a loss of L1L-[NF-YC2] activation, irrespective of the presence of bZIP67. This is consistent with the idea that this combination of NF-YB and NF-YC subunits in the activation of the CRC promoter is unlikely to constitute a conventional trimeric complex of a CCAAT binding factor. One possible mechanism for L1L/LEC1-[NF-YC2] function is that the combination of the NF-YB and NF-YC subunit acts as a co-activator of an ABRE-binding factor. The ability of the mammalian NF-Y complex to recruit TFIID is in line with this mechanism (Bellorini et al., 1997; Frontini et al., 2002). However, the action of LEC1/L1L-[NF-YC2] depended not only on ABREs but also on the context of the promoter, because a synthetic promoter essentially composed of ABREs and a minimal promoter failed to be activated by L1L/LEC1-[NF-YC2], either in the presence or absence of bZIP67. Therefore, an interaction with a specific DNA sequence may be important. There could be specific cis-elements that are bound by the L1L/LEC1-[NF-YC2] dimer. Even if this is the case, the specificity of the DNA recognition could be rather loose, considering that NF-Y subunits have a histone-like structure. Another possibility is that the positions of ABREs relative to the core promoter might be critical for LEC1/L1L-[NF-YC2] to function with the CRC promoter.

We previously suggested that LEC1 controls the expression of CRC, as well as other seed storage protein genes, through FUS3 and ABI3 (Kagaya et al., 2005b). This proposition was based on the result that the expression of FUS3 and ABI3 during seed development was greatly reduced in the lec1 mutant, and that the induction of CRC by the ectopic expression of LEC1 in young seedlings required functional FUS3 and ABI3. However, this proposition does not necessarily exclude the direct action of LEC1 on the CRC promoter. The lec1 fus3 double mutant caused a greater reduction of seed storage protein accumulation compared with that in either single mutant. In addition, FUS3 expression was not completely eliminated in the lec1 mutant. Together with these observations, our transient expression results that the CRC promoter was strongly activated by the combination of LEC1/L1L, NF-YC2 and bZIP67 – and that this activation was observed even at a very early time (3 h) after transfection – support the direct participation of LEC1/L1L in the transcriptional activation of CRC in developing seeds. On the other hand, temporal patterns of LEC1 and L1L expression did not coincide with those of the seed storage protein genes. The transcript levels of LEC1 are substantially decreased at 12 DAF, when those of seed storage protein genes peak. The decrease in the L1L transcript level is delayed from that of LEC1, but still precedes the steep rise in the expression of the seed storage protein gene transcripts. Therefore, the direct action of LEC1 and L1L on the seed storage protein gene promoters might only contribute to the initial phase of seed storage protein gene activation, and their roles might be relayed to other transcription factors. Highly programmed temporal and quantitative regulation of expression for each gene during seed development may rely on such relay mechanisms.

We confirmed the functional interaction of LEC1/L1L and bZIP67 in the regulation of another gene, SUS2. Although we were unable to clearly demonstrate that L1L participates in the regulation of CRC in developing seeds, the observed reduction in SUS2 expression in l1l-1 mutant seeds and the further decrease in the l1l lec1 double mutant make a more convincing argument for the in planta roles of the coupled function of LEC1/L1L and bZIP67.

Although the molecular mechanism for the B3 master regulators of seed development as transcription factors have been studied in various aspects, only limited information is available as to the molecular mechanisms of transcriptional regulation by the NF-YB master regulators, LEC1 and L1L. The present study provides the functional characterization of LEC1 and L1L as transcription activators, and revealed the molecular basis for how a specific NF-YB subunit member(s) can function in a specific biological process. Our results also add further complexity to transcriptional networks that govern seed development, typically represented by the local and redundant regulations among the B3 regulators. A simplified view of transcriptional cascade with LEC1 could be that LEC1 regulates the B3 networks, which in turn regulate various seed-specific genes. However, our results indicate that LEC1 can also directly regulate the terminal seed-specific genes together with transcription factors under the control of the B3 networks. The relationship between the molecularly redundant LEC1 and L1L provides additional complexity, because LEC1 probably controls L1L through the regulation of FUS3. The information about the molecular mechanisms for LEC1 and L1L function revealed in this study should provide further opportunities for resolving the transcriptional networks in seed development.

Experimental procedures

Growth of plants and cell culture

Plants [Arabidopsis thaliana, ecotype Columbia (Col-0) and Wassilewskija (WS) and mutants derived from these ecotypes] were grown under long-day conditions (16-h light/8-h dark) under white fluorescent light (40 μmol m−2 sec−1) on Rock Fiber (Nittobo, http://www.nittobo.co.jp), with 1/5th-strength MS salt mixture supplemented with B5 vitamins (pH 5.7), after establishing seedlings on agar media [half-strength MS salt mixture, pH 5.7, B5 vitamins, 1% (w/v) sucrose and 0.8% (w/v) agar]. The mutant lines used in this study were fus3-3 (Col-0 background; Keith et al., 1994), lec1-1 (WS background; Meinke, 1992) and lec2-1 (WS background; Meinke et al., 1994).

The Arabidopsis cell line T87 was maintained in 100-ml of liquid medium [Gamborg’s B5 salt mixture, B5 vitamins, 0.05% (w/v) 2-(N-morpholine)-ethanesulphonic acid (MES), pH 5.7, 1.5% (w/v) sucrose and 1 μm 1-napthaleneacetic acid potassium salt] by subculturing once a week with 4% of the cell mass of the previous culture (Axelos et al., 1992).

RNA preparation and real-time RT-PCR

Flowers were marked by loosely tying them with thread upon flowering. Seeds were collected from 7–12-DAF siliques, and were immediately frozen in liquid nitrogen. From the frozen materials, RNA was prepared using the PureLinkTM Plant RNA Reagent (Invitrogen, http://www.invitrogen.com). The RNA was further purified by an RNeasy mini-kit column (Qiagen, http://www.qiagen.com), and then reverse transcribed with a Omniscript RT kit (Qiagen) using oligo-dT primer after DNase I (TAKARA BIO INC., http://www.takara-bio.com) treatment. Real-time PCR analysis was performed by the Syber Green method using a Power SYBR® Green PCR Master Mix kit and Genetic Analyzer PRISM 7000 (Applied Biosystems, http://www.appliedbiosystems.com). The primer sets used for the PCR are listed in Table S1. Based on the results of a preliminary microarray analysis using the same RNA set, transcript levels of the myosin-like protein VIIIA gene (AT1G50360) were used to normalize the expression levels.

Preparation of protoplast and transfection

Protoplasts were prepared from T87 cultured cells on the third day of subculture, as described by Axelos et al. (1992), and were transfected essentially as described by Kovtun et al. (2000). A 150-μl aliquot of protoplasts (5 × 106 cells ml−1) in Ma-Mg solution (0.4 m mannitol, 15 mm MgCl2 and 5 mm MES-KOH, pH 5.6) was mixed with 50 μl of DNA solution in sterile water containing 15 μg of effecter, 2 μg of GUS reporter and 2 μg of 35S-LUC (a luciferase internal standard) plasmid (Tsukagoshi et al., 2005), and kept on ice while working on other samples. Then, 65 μl of PEG-CMS solution [0.4 m mannitol, 0.1 m Ca(NO3)2·4H2O, 40% (w/v) PEG 4000] was added, and kept on ice for 20 min. The PEG-treated protoplast suspension was sequentially diluted with 0.4, 0.8, 1.6 and 3.2 ml of dilution solution (0.4 m mannitol, 125 mm CaCl2, 5 mm KCl, 50 mm glucose and 1.5 mm MES-KOH, pH 5.6) at intervals of 20 mins. After the final dilution, protoplasts were recovered by centrifugation at 60 g for 5 mins at 4°C, and suspended with 2 ml B5-mannitol medium [Gamborg’s B5 medium salt mixture, 1 ×  B5 vitamin, 2.56 mm MES-KOH, pH 5.7, 1.5% (w/v) sucrose, 1 μmα-naphthaleneacetic acid and 0.37 m mannitol]. The protoplast suspension was divided into two 1-ml aliquots, one of which was supplemented with ABA at a final concentration of 10 μm, and cultured for 16 h (unless noted) at 22°C under continuous light. Transfection with each combination of plasmids was performed in triplicate.

GUS and luciferase activities were determined as described by Hobo et al. (1999). Promoter activity was calculated as GUS activity divided by luciferase activity, expressed in arbitrary units.

Plasmid construction

Details of plasmid construction are available in the online Supporting Information.

Immunoblotting and immunoprecipitation

Immunoblot analysis was performed essentially as described by Kobayashi et al. (2004), except that polyvinylidene fluoride (PVDF) membranes (Hybond-P; GE Healthcare, http://www.gehealthcare.com) were used instead of nitrocellulose membranes, and the blocking solution contained 10% skimmed milk and 2% BSA. Monoclonal anti-myc (Covance, http://www.covance.com) was used as the primary antibody, and horseradish peroxidase (HRP)-conjugated anti-mouse IgG (GE Healthcare) was used as the secondary antibody.

To perform immunoprecipitation, transfected protoplasts (corresponding to 3.3–7.5 × 106 cells used for transfection) were ground in liquid N2 with a mortar and pestle, suspended in 0.5 ml of IP buffer [50 mm Tris-HCl, pH 7.6, 150 mm NaCl, 10% glycerol, 1 mm EDTA, 1 mm DTT, 1% Igepal CA-630, 0.5 mm PMSF, and Complete EDTA-free (proteinase inhibitor cocktail tablet; Roche, http://www.roche.com)]. The cell lysates were cleared by centrifugation at 18 600 g for 30 mins, and incubated with a 25-μl bed volume of monoclonal anti-hemagglutinin (HA)-conjugated agarose beads (Sigma-Aldrich, http://www.sigmaaldrich.com) at 4°C overnight. The beads were washed five times with 1 ml of IPW buffer (20 mm Tris-HCl, pH 7.5, 150 mm NaCl, 0.1% Igepal CA-630, 0.5 mm PMSF and Complete EDTA-free). The bound protein was eluted with twofold-concentrated Laemmli SDS-PAGE sample buffer, without β-mercaptoethanol, and used for immunoblot analysis.

Acknowledgements

The authors thank Ms Ayumi Fujie, Ms Eriko Kawana, Ms Yuka Ozaki and Dr Hirokazu Kato for excellent technical assistance, and ABRC (http://www.biosci.ohio-state.edu/~plantbio/Facilities/abrc/abrchome.htm) and the NASC (http://arabidopsis.info) for providing the seed stocks of T-DNA insertion and other mutant lines. This work was funded in part by Grants-in-Aid for Scientific Research on Priority Areas (grants 12138204 and 08075007) and the Center of Excellence from the Japanese Ministry of Education, Culture, Sports, Science and Technology, and the Japan Society for the Promotion of Sciences (Research for the Future Grant JSPS-00L1603).

Ancillary