Synergistic Activation of Seed Storage Protein Gene Expression in Arabidopsis by ABI3 and Two bZIPs Related to OPAQUE2*

The expression of many seed storage protein genes in cereals relies on transcription factors of the bZIP class, belonging to the maize OPAQUE2 family. Here, we de-scribe a survey of such factors in the genome of Arabidopsis thaliana, and the characterization of two of them, AtbZIP10 and AtbZIP25. Expression analysis by i n situ hybridization shows that the occurrence of their mRNAs in the seed starts from early stages of development, peaks at maturation, and declines later in seed development, matching temporally and spatially those of the seed storage protein genes encoding 2S albumins and cruciferins. Gel mobility shift assays showed that AtbZIP10 and AtbZIP25 bind the ACGT boxes present in At2S and CRU3 promoters. Moreover, using the yeast two-hybrid system we show that AtbZIP10 and AtbZIP25 can interact in vivo with ABI3, an important regulator of gene expression in the seed of Arabidopsis . Transient expression analyses of a reporter gene under the control of the At2S1 promoter in transgenic plants overexpressing ectopically AtbZIP10 , AtbZIP25 , and ABI3 reveal that none of these factors could activate significantly the reporter gene when expressed individually. However, co-expression of AtbZIP10/25 with ABI3 resulted in a remarkable increase in the activation capacity over the At2S1 suggesting that they are part of a regulatory complex involved in seed-specific expression. This shows a common mechanism of ABI3 in reg-ulating different seed-specific genes through combinatorial interactions with particular bZIP proteins and a conserved role of O2-like bZIPs in monocot and dicot species.

During seed development, storage reserves accumulate mostly in the form of carbohydrates and proteins, whose degradation upon germination will provide nutrients to the growing seedling before the photosynthetic capacity is fully acquired. Seed storage proteins (SSP) 1 are specifically synthesized in developing seeds, both in the endosperm and in the embryo. In the seed of monocotyledoneous species, the former is the predominant reserve tissue, whereas in dicotyledoneous plants the endosperm is commonly re-absorbed as maturation proceeds, and storage proteins are preferentially accumulated in the embryo. Because SSP-encoding genes are specifically induced and tightly regulated during seed development, they represent an interesting model system for studying the mechanisms of temporal and tissue-specific gene regulation.
In the developing seed, different programs of gene expression have been defined that comprise distinct classes of genes that are coordinately regulated (1,2). The MAT (maturation) class includes major SSP genes (like 2S albumins and 12S globulins) expressed at early and mid-maturation phases, whereas the LEA (late embryogenesis abundant) class includes primarily genes involved in the acquisition of desiccation tolerance expressed at later stages of maturation (3,4). Unraveling the molecular basis of seed-specific gene expression has been mainly focused on the identification of cis-acting promoter elements and associated transcription factors (TFs). Both aspects have been considerably studied in monocots due to the importance of cereal seeds as crops, but in dicot species the molecular aspects of this regulation are still less understood. In cereals, the bipartite endosperm box (5,6) and the ACAA motif (7,8) are important regulatory elements, repeatedly found in the promoters of SSP genes. These cis-elements are bound by TFs of the bZIP, DNA binding with one finger (DOF), and MYB classes (8, 9 -11). One of the first-described transcription factors in plants and probably one of the best studied, involved in this regulation is the bZIP protein from maize OPAQUE2 (O2). Mutations in either the O2 gene or in its binding sites within the endosperm box of zein promoters, result in a dramatic reduction in the level of zein transcription (12). The central role of O2-like factors in seed-specific expression has been well documented in other cereal species, such as wheat (SPA), barley (BLZ1 and BLZ2), rice (RISBZ1 and RITA1), Sorghum (SBO2), and Coix (CLJO2) (5,(13)(14)(15)(16)(17)(18).
In dicot plants, the study of SSP gene promoters by deletion analysis, site-specific mutagenesis or gain-of-function approaches has revealed the importance of bZIP target sites to provide full activity and in conferring tissue specificity (19 -22). It has been also demonstrated that these sites are bound in vitro by maize O2 or O2-like factors from cereals, resulting in efficient activation in planta when they are co-transformed into the dicot plant tissues (19,23). Moreover, the fact that target sites of O2-like bZIPs in the promoters of monocot SSP genes are necessary for seed expression in transgenic dicot plants (24,25) suggests that similar, yet uncharacterized, factors are im-portant in the regulation of SSP genes in the dicot embryos.
In Arabidopsis and in closely related species from the genus Brassica, regulatory cis-acting elements required for specific expression of SSP genes have been determined (22,26,27). Both in 2S albumin and napin gene promoters these include the RY motif and the ACGT box, representing putative targets for B3 and bZIP transcription factors, respectively. Alternative strategies, like the analysis of genetic mutants with altered patterns of gene expression in the seed, have also led to the discovery of genes corresponding to TFs responsible for these variations. Remarkably, factors with a B3 domain such as ABI3, FUS3, or LEC2, potentially implicated in the regulation of SSP genes through the RY element, were found among them (28 -30).
Concerning the interaction with the ACGT boxes in these promoters, several bZIP proteins of the ABI5 group expressed in the seed have been identified as putative candidates to participate in this regulation (31,32). However, mutations in the ABI5 and EEL genes were reported not to affect the expression of early and mid-maturation SSPs (e.g. 2S albumin genes) but instead of those genes expressed at late embryogenesis (e.g. LEA protein genes), suggesting that different bZIP factors should be involved in their regulation (32).
As part of a project devoted to the study of seed-specific transcriptional control, we focused on the identification of Arabidopsis bZIP transcription factors related to OPAQUE2, as putative candidates for the regulation through ACGT boxes of early and mid-maturation SSP genes. Here, we present evidence that two bZIP factors of the OPAQUE2 family in Arabidopsis are involved in MAT SSP gene activation in a concerted manner with ABI3.

EXPERIMENTAL PROCEDURES
Plant Material-Arabidopsis thaliana seeds from Columbia (Col-0) and C24 were obtained from the Arabidopsis Biological Resource Center (Ohio). Seeds from A. thaliana C24 expressing ectopically ABI3 (2) were obtained from the authors. Plants were grown under standard conditions in growth chambers at 22°C with a 16-h photoperiod.
Isolation and Analyses of cDNAs and Genomic Clones of O2-like bZIPs-cDNAs encoding AtbZIP9, -10, -25, and -63 were obtained from public Expressed Sequence Tag (EST) collections or amplified by reverse transcriptase-PCR. The corresponding genomic clones were isolated from an A. thaliana library prepared in GEM11 (Promega), kindly donated by J. Paz-Ares. Screening of the Arabidopsis genomic library was performed following general procedures (33) with genespecific probes corresponding to cDNA fragments encoding the following protein regions: AtbZIP9, aa 1-100; AtbZIP10, aa 1-125; AtbZIP25, aa 300 -403; AtbZIP63, aa 220 -314 and previously shown not to crosshybridize. DNA sequences were obtained using the ABI PRISM 377 dye terminator sequencing system and the ABI PRISM 377 DNA sequence analyzer (PerkinElmer-Applied Biosystems). Analyses of DNA and deduced protein sequences were done with computer services under the European Bioinformatics Institute (www.ebi.ac.uk/). Sequence alignments and phylogenetic analyses were performed using ClustalW and Phyllip packages, respectively, at the server of the DNA Data Bank of Japan (www.ddbj.nig.ac.jp/). Detection of conserved domains in protein sequences was done with MAST/MEME tools (meme.sdsc.edu/ meme/website/meme.html).
Northern Blots and in Situ Hybridization Analyses-Northern blot analysis was performed using total RNA from different Arabidopsis tissues as described in Ref. 2. Gene-specific probes (see above) were labeled by random priming following standard procedures (33). In situ hybridization was performed according to Ref. 34 with digoxigeninlabeled antisense RNA probes generated according to the manufacturer's specifications (Roche Diagnostics). Gene-specific probes, described above, were cloned into pGEM-T easy (Promega), and labeled RNA was synthesized on linearized templates using T7 or SP6 polymerase depending on insert orientation. An ABI3-specific probe was obtained by PCR amplification of a cDNA fragment corresponding to the last 103 aa of the ABI3 protein. This fragment was cloned into pGEM-T easy and used to synthesize RNA probes following the same procedure.
Yeast Strains and Methods-Strains HF7c and SFY526 of Saccharomyces cerevisiae carrying LacZ (␤-galactosidase) and HIS3 (imidazole glycerol phosphate dehydratase) reporter genes under the control of a truncated Gal1 promoter that contains Gal4-responsive elements (Gal1UAS) were obtained from Clontech. The effector plasmids pGBT9 and pGAD424 (Clontech), which contain the alcohol dehydrogenase I (AdhI) promoter fused to the Gal4 DNA binding domain (Gal4DBD; pGBT9 vector) or to the Gal4 DNA activation domain (Gal4AD, pGAD424 vector), respectively, were used to generate translational fusions with AtbZIP10, AtbZIP25, or ABI3 or with particular fragments derived from them (ABI3 N: aa 1-525, ABI3 C: aa 400 -720; AtbZIP10 C: aa 184 -410; and AtbZIP25 C: aa 200 -403). All constructs were generated by PCR amplification with selected primers and verified for in-frame fusions and sequence fidelity. Yeast transformation, activity assays for ␤-galactosidase in liquid cultures and colony growth in Hisdepleted medium, were done at 30 or 23°C, according to the manufacturer's instructions with minor modifications (13,14) Particle Bombardment in Arabidopsis Leaves-Particle coating and bombardment were carried out basically according to Ref. 13 with a biolistic Helium gun device (DuPont PDS-1000, BioRad), using rupture disks of 1,100 psi and a 6-cm distance between macrocarrier and target samples. Leaves from Arabidopsis plants grown under sterile conditions were isolated from a rosette stage, placed onto a 5-cm dish agar-MS/2 medium, and shot in sets of six. After bombardment leaves were incubated in the same plate for 24 h, and GUS activity was quantified by chemiluminescence (GUS light kit, Tropix) and expressed as relative luminescence units (RLU) per g of protein.
AtS1::GUS reporter vector is a pUC19-derived plasmid containing the ␤-glucuronidase reporter gene under the control of the At2S1 promoter and fused to the 3Ј-terminator of the nopaline synthase gene (3Ј-nos). Effector plasmids for the expression of AtbZIP10, AtbZIP25, and ABI3 were generated by cloning the corresponding cDNAs under the control of the CaMV35S promoter and a 3Ј-nos terminator in pBSII (Stratagene). A. thaliana transgenic plants overexpressing AtbZIP10 and AtbZIP25, were generated by Agrobacterium transformation following procedures referred to in Ref. 34 and using the vector pROKII, in which the expression of both cDNAs is driven by the CaMV35S promoter. Plants overexpressing ABI3, previously described (2), were a kind gift from the authors.

The Arabidopsis Genome Contains Four Genes Encoding
OPAQUE2-related bZIPs-The completion of the A. thaliana sequencing project (35) facilitates the identification of genes encoding defined protein families within the genome. O2-related bZIPs have several features that make them distinguishable from other members of the bZIP group of TFs (Fig. 1A). The bZIP region has an extended zipper comprising up to nine leucine heptad repeats with a high degree of sequence conservation; outside this domain, other sequence signatures are also shared specifically within the O2 family. Among 75 bZIP-encoding genes annotated in the Arabidopsis genome (36, 37), comprehensive sequence searches (see "Experimental Procedures") using both the conserved bZIP region and the different signatures (1-6) schematically represented in Fig. 1A, allowed the identification of four genes, AtbZIP10, AtbZIP25, At-bZIP63, and AtbZIP9 as the only members of the O2 family in Arabidopsis. Naming of these genes followed the nomenclature proposed by the EU-Regulatory Gene Initiative in the Arabidopsis project (37). Amino acid sequence alignments and comparisons with other O2 members in different plant species, led to the recognition of similar structural signatures common to these TFs. The results of such analyses are summarized in Fig.  1A, where the occurrence of domains 1-6, alongside the conserved bZIP region, is indicated for each individual member of the family. Among the four Arabidopsis O2-like bZIP proteins, AtbZIP10 shares the maximum number of conserved domains with members of the family in other species, and in particular with maize O2 and O2-like cereal proteins. The phylogenetic tree in Fig. 1B, based on sequence similarities of the bZIP region, shows that AtbZIP25 is very closely related to At-bZIP10, suggesting that both genes might represent redundant functions. AtbZIP63 forms a clade together with parsley CPFR2 and soybean G/HBF, two dicot bZIPs putatively involved in responses to pathogen and environmental stress (38,39). Finally, AtbZIP9 is related to RITA1 and other rice bZIPs of yet unknown functions (15).
AtbZIP10 and AtbZIP25 Are Expressed in the Developing Seed-For a transcription factor to participate in the regulation of seed storage protein genes, a necessary requisite is to be expressed in the developing seed. Consequently, the expression patterns of the four O2-like bZIPs from Arabidopsis were investigated by Northern blot analysis. Total RNA samples isolated from different tissues, including siliques at two developmental stages, were hybridized after electrophoretic separation to gene-specific probes corresponding to AtbZIP10, AtbZIP25, AtbZIP63, and AtbZIP9 genes. As shown in Fig. 2 the four genes are expressed in roots (R), shoots (S), rosette leaves (L), and flowers (F), but only hybridization to the AtbZIP10 and AtbZIP25 probes presented intense signals in lanes S1 and S2, containing RNAs from siliques.
If AtbZIP10 and AtbZIP25 were regulators of SSP gene expression, an overlapping pattern of expression of these genes should be anticipated. To explore this possible relationship, the same blot was subsequently hybridized with a probe specific for a 2S albumin gene. The peak for the TF genes was at the S1 stage whereas that for the 2S gene was at the later S2 stage of silique development. The patterns of AtbZIP10/25 and 2S mRNA accumulation are therefore consistent with the possibility of the former being regulators of this SSP gene.
To clarify whether the signals detected in Northern blots derived specifically from the seed or from other silique tissues, a more precise expression analysis, namely in situ mRNA hybridization, was done for these genes (Fig. 3). AtbZIP63 and AtbZIP9 mRNAs could not be detected in developing seeds, AtbZIP9 expression being restricted to the valve vascular veins in the siliques (data not shown). In contrast, expression in the seed could be clearly detected for AtbZIP10 and AtbZIP25, whose patterns of expression were found to be identical throughout seed development. AtbZIP10/25 transcripts were present from early stages of development both in the developing embryo and, at lower levels, in the endosperm. Expression of AtbZIP10/25 decreased as development proceeded. At later stages, they were expressed at very low levels in the mature embryo and the remaining endosperm, but were strongly upregulated in the inner layer of the seed coat. tion with an ABI3 probe was also performed. ABI3 RNA was detected in the embryos, endosperm, and outer teguments of the seed throughout seed development, with a pattern similar to that previously shown for AtbZIP10/25. However, in contrast to AtbZIP10/25 expression, which is detected preferentially in the embryo, ABI3 RNA levels were high, both in the endosperm and in the embryo, even at later stages of seed development. In mature seeds, ABI3 expression was also strong in the outer layers of the seed coat AtbZIP10 and AtbZIP25 Bind to ACGT Core Sequences in Seed Storage Protein Gene Promoter; Effect of ABI3 on Binding Affinities-Arabidopsis major seed storage proteins are the arabidins (2S albumins) and cruciferins (12S globulins). While 2S albumins map to a single locus, as a four tandem repeat, with minimal sequence divergence, cruciferins comprise a gene family with at least three different members (CRU1, CRU2, and CRU3). All these genes are expressed in the embryo in a coordinated fashion during the early and midstages of seed maturation. Sequence comparisons of these Arabidopsis SSP gene promoters revealed the presence of conserved ACGT boxes. Similar boxes in Brassica napin promoters have been shown to be critical for their expression in seeds (22). The finding that the Arabidopsis AtbZIP10 and AtbZIP25 spatial and temporal patterns of expression in developing embryos overlap with those of At2S and At12S genes, prompted us to investigate whether these factors could recognize in vitro the conserved ACGT boxes present in these SSP gene promoters. To this end, AtbZIP10 and AtbZIP25 proteins expressed in Escherichia coli or produced in vitro, were tested for their ability to bind different oligonucleotide sequences containing ACGT boxes derived from At2S1 and CRU3 promoters. These boxes present adjacent RY regulatory elements (CATGCATG) in these promoters (schematically represented in Fig. 4A). In Fig. 4B the results of EMSA demonstrate that the AtbZIP10 and AtbZIP25 binding to these sequences occurs specifically through the ACGT boxes, since mutations (AaGg) introduced at this core abolished the interaction in the context of both At21S and CRU3 promoters.
ABI3 has been previously demonstrated to be an important regulator of 2S albumin and cruciferin genes, since abi3 mutants display a significant, although not complete, reduction in the expression levels of SSP genes (2). The ABI3 protein has a modular structure with high sequence similarity to the maize Viviparous1 (Vp1) gene product (40). Similar to VP1, ABI3 has a B3 domain, but its ability to bind DNA through this domain has not been described. In addition, both VP1 and ABI3 contain a B2 domain, reported in VP1 to increase the DNA binding affinity of different proteins, including O2 (41). We have detected that AtbZIP10 and AtbZIP25, as well as ABI3 are coexpressed in the same tissues during seed development, and that AtbZIP10 and AtbZIP25 bind to ACGT core sequences in the promoters of SSP genes expressed herein too. It was pertinent, therefore, to check whether possible changes in their binding affinities were produced in the presence of ABI3. In vitro, EMSA experiments shown in Fig. 4C, display the binding behavior of different amounts of AtbZIP10 or O2 proteins incubated with target DNA sequences. A clear increase in the binding of both proteins to their target sites was obtained in the presence of ABI3 (see lanes under ϩABI3 in Fig. 4C) as compared with controls lacking the ABI3 protein.
In analogous experiments performed with AtbZIP25, we were unable to see an increase of the binding affinity in the presence of ABI3 (data not shown).
In Vivo Interactions between AtbZIP10/25 and ABI3 in the Yeast Two-hybrid System-AtbZIP10 and AtbZIP25 display high sequence similarity between them and are closely related to other bZIPs of the O2 family over particular protein regions. Two of the shared signatures (1 and 2 in Fig. 1A) span the activation domain AD-I. To test their potential as transcriptional activators, as well as their putative interactions with ABI3, the yeast one-and two-hybrid systems were used (Fig. 5). Yeast plasmid vectors were generated to express proteins containing the GAL4 DNA-binding domain as N-terminal fusions to the complete AtbZIP10 and AtbZIP25 factors or their Cterminal regions. When these vectors were transformed into appropriate yeast strains, activation of the reporter activity was detected in both cases (see Fig. 5D), AtbZIP10 being a much stronger activator than AtbZIP25 in this system as shown by the higher concentration of 3AT required to counteract its growth in His-depleted medium. As in other O2-like factors, we could map the activation capacity in the N-terminal region of both proteins, since GAL4BD-fusions to the C-terminal fragments fail to induce reporter activities compared with the complete proteins (e.g. see Fig. 5C for AtbZIP25).
Yeast two-hybrid experiments were performed at two different temperatures, 30°C (standard conditions) and 23°C (commonly used to disclose weaker interactions). The results of a systematic analysis to check for possible interactions between FIG. 3. AtbZIP10 and AtbZIP25 expression in developing seeds. A-C, in situ mRNA hybridization of AtbZIP10 at different stages of seed development. AtbZIP10 expression was detected from early stages of seed development both in the developing embryo and, at lower levels, in the endosperm. AtbZIP10 level of expression decreased as development proceeded, at later stages the expression was low in the mature embryo and the remaining endosperm, but increased in the inner layer of the seed. Patterns of expression of AtbZIP25 were identical to AtbZIP10. D-F, in situ hybridization with an ABI3-specific probe was also performed to test for possible overlapping expression patterns with AtbZIP10/25. ABI3 RNA was detected in the embryos, endosperm, and outer teguments of the seed throughout development, as previously shown for AtbZIP10/25. However, contrary to AtbZIP10/ 25, whose expression is detected preferentially in the embryo, ABI3 RNA levels were high both in the endosperm and in the embryo, and remained elevated even at later stages of seed development. In mature seeds, ABI3 expression was also strong in the outer layers of the seed coat. A-D, seeds ϳ3 days after fertilization; B-E, seeds ϳ6 days after fertilization; C-F, seeds ϳ12 days after fertilization.
AtbZIP10 and AtbZIP25 with ABI3 are also reflected in Fig. 5. Since ABI3 is very potent activator in yeast (our data not shown) constructs of ABI3 were prepared in the plasmid containing the GAL4 activation domain. Two different constructs were generated encompassing the N-(residues 1-525) or Cterminal (residues 400 -720) regions of the ABI3 protein fused to GAL4-AD (AD-ABI3N and AD-ABI3C, respectively) These constructs and an empty vector containing the AD domain alone, were transformed into yeast strains expressing fusions of the GAL4 DNA-binding domain (BD) to the complete or C-terminal regions of AtbZIP10 and AtbZIP25 proteins. Liquid medium assays of ␤-galactosidase activity were performed for each combination of AD and DB vectors and represented as percentages relative to basal activities obtained for BD constructs cotransformed with empty AD vectors (100% values). The results of such analyses are displayed in Fig. 5B, showing that AtbZIP10 and ABI3 are able to interact through the Nterminal regions of both proteins. Likewise, an interaction between the ABI3 N and AtbZIP25 C-terminal region was also detected. Interactions between AtbZIP25 and ABI3 could not be reliably reproduced at 30°C, but lowering incubation temperatures to 23°C resulted in consistent data in all replicated experiments.
To confirm the interactions observed, a strain with HIS reporter activity, driven through GAL4 binding sites in the context of a different promoter, was used. In Fig. 5D, basal levels of HIS activity, derived from intrinsic activation of AtbZIP10 or AtbZIP25, could be eliminated at precise concentrations of 3-AT. Nevertheless, these concentrations resulted inefficient when cells were cotransformed with AD vectors containing the interacting protein regions previously described.
Seed-specific Gene Activation by Ectopic Expression of At-bZIP10 and AtbZIP25 in Conjunction with ABI3-In the absence of mutants, to explore the possible involvement of At-bZIP10 and AtbZIP25 in the regulation of SSP genes, we decided to study the effects of ectopic overexpression of these factors on the activation of a reporter gene under control of a 2S albumin promoter. Transgenic plants expressing AtbZIP10 or AtbZIP25 under the CaMV35S promoter were generated and checked for overexpression of these genes by Northern-blot analysis of leaf tissue. Leaves of these plants were used in experiments of transient expression analysis after particle bombardment. As expected for a seed-specific gene, activity of the 2S promoter::GUS reporter was minimal in leaf tissue of wild-type plants. When this construct was bombarded into leaves overexpressing either AtbZIP10 or AtbZIP25 (Fig. 6A,  lanes 3 and 5), a significant although moderate increase of reporter activity was observed over controls. The interactions detected between these bZIP factors and ABI3 in the yeast two-hybrid system (Fig. 5) led us to explore their possible functional relevance in planta. Using wild-type plants we tested the activation effect of ABI3 on the 2S-driven reporter activity and determined that it was of a low range, similar to that observed for AtbZIP10 or AtbZIP25 alone. In contrast, transformation of ABI3 in transgenic tissue overexpressing AtbZIP10 or AtbZIP25 resulted in a significant increment of reporter activity (Fig. 6A, lanes 4 and 6). To validate the significance of these observations, the reciprocal experiment was performed using stably transformed transgenic plants overexpressing ABI3 ( 35 S::ABI3). In this background, the same reporter construct, 2S::GUS, was transiently expressed either absence (Ϫ) or presence (ϩ) of ABI3 in the binding reaction. An increase in the intensity of the bands is observed when ABI3 is added to the reaction. The lower panel shows a similar effect exerted by ABI3 on the maize protein OPAQUE2.  (Fig. 6B, lanes 1 and 4) or in combination with AtbZIP10 or AtbZIP25 as effectors (Fig. 6B, lanes 2, 5, and 3, 6, respectively). In Fig. 6B, the synergistic activation effect resulting from the combination of any of the Arabidopsis bZIP factors (AtbZIP10 or AtbZIP25) with ABI3, on 2S::GUS reporter activity, fully confirmed the previous observations (Fig. 6A). DISCUSSION In Arabidopsis, major seed storage protein genes are expressed in the embryo at early and midstages of maturation under a tight temporal and tissue-specific regulation. Here, we present several lines of evidence on support that the TFs, AtbZIP10 and AtbZIP25, characterized in this study, participate in this regulation. First, they are structurally very closely related to OPAQUE2-like factors involved in the regulation of seed storage protein genes in cereal species. Likewise, they are transcriptional activators, expressed in the seed at the maturation phase and bind to regulatory sequences in SSP gene promoters. Finally, overexpression of these factors, in conjunction with ABI3, induces the ectopic expression of SSP genes in Arabidopsis leaves.
AtbZIP10 and AtbZIP25 Encode O2-like Proteins Expressed in the Arabidopsis Seed-Developmental stages of embryogen-esis during seed formation in Arabidopsis are characterized by the accumulation of distinct sets of mRNAs, resulting from sequential gene expression programs, activated by different regulatory factors. MAT-specific or Class II genes (2) encode major seed storage proteins, the expression of which starts from 9 days after fertilization, peaks at 15, and declines about 20 days after fertilization. The specificity of this expression pattern has been unambiguously associated with the presence in their promoters of conserved regulatory elements. Among them, and in striking parallelism to what has been experimentally determined for SSP genes of monocot species, ACGT core sequences are of particular relevance in their regulation. With the hypothesis that similar factors might be involved the regulation of analogous processes in Arabidopsis and in cereals, we have surveyed the Arabidopsis bZIP proteins to search for O2-related members. Comparisons of O2-like proteins from diverse plant species allowed the recognition of conserved features within the bZIP domain and the detection, outside this region, of additional signatures characteristic of the O2-bZIP family. A search with these features in the Arabidopsis genome led to the identification of four O2-like encoded proteins of which AtBZIP10 and 25 share the maximum number of signa- were cotransformed with the indicated AD-constructs derived from ABI3 or an empty AD vector (AD-ø). Measurements of ␤-galactosidase activity in liquid medium assays were obtained from three independent replicates and represented as percentages referred to the basal activity (100%) of BD constructs. (S.E. not shown, were Ͻ15%). C, mapping interactions between ABI3 and AtbZIP25. The indicated AD and BD constructs were cotransformed and yeast cells grown at 23 or 30°C, before performing ␤-galactosidase filter assays. Lowering incubation temperatures (23°C) allowed disclosing an interaction between the AtbZIP25 and ABI3 C-terminal regions that was not detected at 30°C. D, yeast cell growth of S. cerevisiae HF7C strain containing the His3 reporter gene under control of the Gal1-truncated promoter. Cells were transformed with BD-AtbZIP10/25 constructs and AD-ABI3 N, AD-ABI3 C or an empty AD vector (AD-ø). Minimal His depleted medium with increasing concentrations (0 -60 mM) of 3-aminotriazole (3-AT) used to suppress the basal activation capacity displayed by AtbZIP10/25. Interactions between AtbZIP10-ABI3 N and AtbZIP25-ABI3 C are revealed by cell growth above 3-AT concentrations that suppress intrinsic AtbZIP10/25 activation. tures with O2-like members of cereal species, known to participate in the control of SSP gene expression. Important information, derived form the phylogenetic tree of O2-like factors from different species, is that AtbZIP10 and AtbZIP25 integrate a distinct branch that clusters individually, indicating that they are structurally closer between them than to any other protein of this group. In addition, the identical patterns of expression observed for both genes suggest that they might represent a case of gene duplication with functional redundancy in Arabidopsis.
We have determined that AtbZIP10 and AtbZIP25 are highly expressed in the developing seed from very early stages of development to mid-maturation, then decaying as desiccation proceeds. This pattern of expression is consistent with a putative participation in the regulation of SSP genes, such as 2S albumins and cruciferins. Since the expression of AtbZIP10 and AtbZIP25 is not restricted to the seed, an important matter of further study, outside the scope of this work, is the function of these factors in organs unique to the seed.
Activation of SSP Genes Results from the Combinatorial Effects of Different Regulatory Factors-Systematic analyses of different SSP gene promoters, both from monocot and dicot species, have led to the identification of multiple protein binding sites responsible for their regulation. Among them, ACGT core elements are critical in conferring seed-specific expression, as determined in functional assays of mutated promoters or in gain-of-function experiments (6,20,22). Remarkably, in monocot species, transcription factors responsible for SSP gene activation through ACGT elements have been clearly identified as members of the O2-like bZIP class; maize O2 being originally isolated from seed mutants defective in SSPs (9). In Arabidopsis and other dicot species, a number of mutants are known also with altered expression programs at early and midstages of maturation, affecting MAT genes like 2S albumins. Interestingly, none of the mutants characterized so far (e.g. abi3, fus3, or lec1 and 2) correspond to a TF interacting with ACGT boxes, despite unequivocal participation of this element in the regulation of these genes. It was shown, however, that coupling of additional elements to these boxes in the context of native promoters was essential for promoter hormonal modulation and resulted in synergistic activation effects (22), indicating that expression changes could arise from mutations in any of the participating components.
According to a previously proposed model of 2S regulation (27), full activation of 2S promoters results from the interaction of diverse factors acting on defined elements organized into two major regulatory domains. In particular, the RY-G domain consists of a G-box (the target of a yet uncharacterized protein), flanked by RY elements (putatively bound by B3 domain containing factors, like ABI3). The analysis of abi3 mutants revealed the participation of the ABI3 protein in the regulation of 2S promoters (2), which was presumed to occur through its direct interaction with the RY elements (27). In support of that, transient expression experiments with 2S promoters demonstrated that ABI3 activation capacity was lost when mutations in the RY elements were introduced, although direct DNA binding was never observed for the ABI3 protein. Interestingly, despite the integrity of RY elements, ABI3 could not activate when the G-box within the RY-G domain was mutated (27), indicating that other factors interacting with this element are crucial in the ABI3-mediated activation of 2S promoters, perhaps by tethering and stabilizing ABI3 into a protein complex acting on the RY-G sequences.
On this basis, we searched and identified O2-like factors from Arabidopsis and tested their putative contribution to the activation of 2S albumin promoters, particularly in relation to ABI3. Our in situ hybridization results show that AtbZIP10 and AtbZIP25 are highly expressed in developing seeds at a time when 2S albumin expression takes place and overlapping with that of the ABI3 gene. Gel retardation assays confirmed that AtbZIP10 and AtbZIP25 bind in a sequence-specific manner to the G-box within the RY-G regulatory domain, and most significantly, using the yeast two-hybrid system we demonstrated that AtbZ10/25 and ABI3 are able to physically interact. The use of experimental conditions with lower incubation temperatures (23°C), allowed us to detect consistently a weaker interaction between ABI3 and AtbZIP25 and to map the interacting domain of the later to the C-terminal region. All these findings are consistent with the established idea of a regulatory complex acting on 2S promoters and support that the O2-like proteins identified herein represent the ACGTinteracting factors implicated in 2S activation. If this was so, full activation of 2S promoters should be achieved in the presence of both ABI3 and AtbZIP10/25. Transient expression analyses with the 2S promoter driving GUS reporter activity corroborate these expectations. Activity levels obtained with this construct in leaves of transgenic plants overexpressing At-bZIP10, AtbZIP25, or ABI3 alone were higher than the basal ones obtained in wild-type backgrounds, although of limited range, as previously reported by Northern blot analysis in CaMV35S::ABI3 lines (2). However when AtbZIP10/25 were included as co-effectors in CaMV35S::ABI3 plants or ABI3 was a co-effector in CaMV35S::AtbZIP10/25 plants, a considerable synergistic effect on the activity of the 2S promoter was observed. These results demonstrate that neither AtbZIP10/25 nor ABI3 overexpression is sufficient to attain the full induction of SSP genes, and that a combined effect of these factors is required for complete activation. It is also worth mentioning that these observations reflect a parallelism in the mode of action of ABI3 in controlling both MAT and LEA genes through specific interactions with different bZIP proteins (42) and are in support of the importance of combinatorial control in the regulation of transcription in plant promoters (43).
Although the positive effect of ABI3 on 2S activation has been consistently demonstrated in the past, the mechanisms underlying this process were obscured by the fact that no DNA binding capacity could be demonstrated for ABI3. This capacity, presumed on the basis of structural similarities of this factor to maize VP1, is a matter of controversy even for the later, since only a truncated version of VP1 spanning the B3 domain exhibits in vitro cooperative interaction with DNA (44). It has been proposed that ABI3 interaction in vivo with RY motives occurs only when it is part of a larger regulatory complex (45), by analogy to other mammalian co-activators (46). In this respect, the interactions that we have detected in the yeast 2-hybrid system are of particular relevance, indicating that AtbZIP10 and AtbZIP25 represent the interface that links ABI3 to RY-G-mediated activation of 2S promoters.
In conclusion, this study allowed the identification in Arabidopsis of four bZIPs belonging to the O2 family. A detailed characterization of two of them, AtbZIP10 and AtbZIP25, showed that they are both structurally and functionally the most closely related to O2-like cereal TFs, and might represent their Arabidopsis counterparts involved in SSP gene regulation. In this respect, it is worth mentioning the conservation discovered among TFs controlling equivalent processes of seed development in monocot and dicot species. Examples of such a kind, are the cereal VP1-like factors and ABI3 in Arabidopsis, and their interacting bZIP proteins TRAB1 and ABI5 (of rice and Arabidopsis, respectively), which represent orthologous functions (47,48). Similarly, the recently described transparent testa (TT) factors in Arabidopsis, TT2 (49), and TT8 (50), which correspond to members of the maize Myb C1/PI (51) and bHLH R/B classes (52), involved in the control of flavonoid biosynthesis during seed formation.
All these findings reveal an evolutionary conservation, remarkably higher than initially anticipated (53), in the mechanisms controlling seed development in monocot and dicot species. In this framework, the identification of the O2-like factors AtbZIP10 and AtbZIP25 invites the exploration in Arabidopsis of the existence of additional factors known to participate in seed storage protein regulation in monocot species and to in-vestigate whether relevant interactions among them have also been maintained.