Interactions of the developmental regulator ABI3 with proteins identified from developing Arabidopsis seeds


  • Smita Kurup,

    1. IACR-Long Ashton Research Station, Department of Agricultural Sciences, University of Bristol, Long Ashton, Bristol BS41 9AF, UK, and
    Search for more papers by this author
  • Huw D. Jones,

    1. Biochemistry and Physiology Department, IACR-Rothamsted, Harpenden, Herts AL5 2JQ, UK
    Search for more papers by this author
  • Michael John Holdsworth

    Corresponding author
    1. IACR-Long Ashton Research Station, Department of Agricultural Sciences, University of Bristol, Long Ashton, Bristol BS41 9AF, UK, and
      *For correspondence (fax +44 1275 394281; e-mail
    Search for more papers by this author

*For correspondence (fax +44 1275 394281; e-mail


The ABI3 locus is a major regulator of embryo development in Arabidopsis and is essential for the simultaneous activation of the maturation pathway, as well as repression of germination and seedling development. We used a two-hybrid screen in yeast in order to identify proteins that interact with ABI3. Four ABI3-interacting proteins (AIPs) were identified which showed specific in vivo and in vitro interactions with the C-terminal region of ABI3 that contains the B2 and B3 domains, previously shown to have DNA binding activity. The expression characteristics of the genes encoding the AIPs have also been analysed in wild-type and abi3, lec1 and fus3 embryo mutants. This analysis demonstrated differential expression of these genes during normal embryo development and in the mutant lines. All the AIPs show homology to existing transcription factors and therefore they may function with ABI3 within the network of transcriptional regulators that control embryo development in Arabidopsis.


Embryo development in higher plants consists of several phases ( Goldberg et al. 1994 ; West & Harada 1993), including an initial phase of early morphological embryo pattern formation, followed by maturation and subsequently by desiccation and quiescence. The maturation phase is characterized by the accumulation of storage reserves including proteins, carbohydrates and lipids ( Goldberg et al. 1994 ).

The regulatory events that control plant embryo development are still largely unknown. However, genetic studies have revealed that, in Arabidopsis thaliana, the ABA-INSENSITIVE3 (ABI3) ( Koornneef et al. 1982 ; Koornneef et al. 1984 ), FUSCA3 (FUS3) ( Baumlein et al. 1994 ; Keith et al. 1994 ) and LEAFY COTYLEDON1 (LEC1) ( Meinke et al. 1994 ; Meinke 1992; West et al. 1994 ) loci play major roles in regulating maturation. All three promote embryo-specific processes and simultaneously repress germination.

Severe mutations at the ABI3 locus result in the production of seeds that remain green, are insensitive to exogenous absisic acid (ABA), and fail to acquire desiccation tolerance ( Koornneef et al. 1982 ; Koornneef et al. 1984 ; Ooms et al. 1993 ). The embryos, however, are morphologically normal, and rescued immature seeds give rise to plants that are indistinguishable from the wild-type ( Nambara et al. 1992 ). Shoot apex development and xylem differentiation can be observed in abi3 seeds, demonstrating that these embryos exhibit characteristics of germinating seedlings ( Nambara et al. 1995 ). Mutant embryos are also impaired in the accumulation of many seed-specific transcripts ( Parcy et al. 1994 ), indicating that ABI3 activates the expression of a large number of genes during embryo maturation. The regulatory mechanism is complex, since ABI3 affects the accumulation of seed transcripts belonging to different temporal classes to varying extents. These studies show that ABI3 acts as a positive regulator of the late maturation programme in embryogenesis, and simultaneously as a negative regulator of germination programmes ( Parcy et al. 1994 ).

The ABI3 protein ( Giraudat et al. 1992 ) shows significant homology to VP1 (VIVIPAROUS1), a transcriptional activator from maize ( McCarty et al. 1991 ) that also regulates embryo maturation. The similarity between ABI3 and VP1 strongly suggests that ABI3 is also a transcription factor ( Giraudat et al. 1992 ).

ABI3/VP1 homologues have been identified in several plant species, including OsVP1 from Oryza sativa ( Hattori et al. 1994 ), PvAlf from Phaseolus vulgaris ( Bobb et al. 1995 ), AfVP1 from Avena fatua ( Jones et al. 1997 ), C-ABI3 from Daucus carota ( Shiota et al. 1998 ), CpVP1 from Creterostigma plantagineum ( Chandler & Bartels 1997) and PtABI3 from poplar (Populus trichocarpa) ( Rohde et al. 1998 ). The high level of protein sequence conservation in all these species indicates that this transcription factor class is fundamentally important in regulating embryo maturation in flowering plants.

Analysis of the predicted protein sequence of ABI3/VP1 reveals an N-terminal acidic region that has been demonstrated to be a transcriptional activator in VP1 ( McCarty et al. 1991 ); the homologous region in PvAlf has also been shown to activate transient gene expression in yeast and plant cells ( Bobb et al. 1995 ). Similar transient gene expression assays have demonstrated the ability of the rice homologue, OsVP1, to activate transcription in rice suspension-cultured cells ( Hattori et al. 1994 ). High sequence similarity between ABI3 and VP1 is also found in three basic domains, designated B1, B2 and B3. The B2 domain from VP1 has been shown to enhance the in vitro binding of transcription factors to their recognition sites ( Hill et al. 1996 ), and it has been suggested that this domain may interact with DNA in a non-sequence-specific manner. Such an interaction may alter the confirmation of the DNA to enhance the binding of sequence-specific transcription factors. The isolated B3 domain of VP1 has a highly co-operative, sequence-specific DNA binding activity in vitro, whereas the full-length protein shows no specific DNA binding activity ( Suzuki et al. 1997 ). Whether the requirement for the B3 domain to be isolated to observe DNA binding is due to steric hindrance, which might affect accessibility, or whether the N-terminal region of VP1 has a specific role in inhibition of the DNA binding activity associated with B3, is not known.

Interestingly, several other proteins besides those belonging to the ABI3/VP1 family share domains with sequence similarity to the B3 domain. These include FUS3 ( Luerssen et al. 1998 ), MONOPTEROUS ( Hardtke & Berleth 1998) and ARF1 (AUXIN RESPONSE FACTOR1) ( Ulmasov et al. 1997 ). The precise role of this domain in these other proteins is as yet unknown.

The FUS3 transcript is expressed 2 days after pollination and expression levels increase during the first half of embryogenesis ( Luerssen et al. 1998 ). In comparison, ABI3 transcripts are induced around the same developmental stage as FUS3, but while FUS3 transcripts peak shortly after mid-embryogenesis, ABI3 expression spans mid-embryogenesis and is highest during late embryogenesis ( Luerssen et al. 1998 ; Parcy et al. 1994 ).

Genetic studies have shown that ABI3, FUS3 and LEC1 interact to regulate several processes during seed maturation, including accumulation of chlorophyll, sensitivity to ABA and expression of storage proteins ( Parcy et al. 1997 ), and that FUS3 and LEC1 (which has been shown to encode a homologue of the yeast transcription factor HAP3; Lotan et al. (1998) ) regulate the abundance of ABI3 protein. These data indicate that ABI3, LEC1 and FUS3 act in a synergistic and co-ordinated manner. Whether the existence of the B3 domain in FUS3 indicates some overlap or redundancy in function between ABI3 and FUS3 is unknown.

To gain insight into the regulatory processes that control embryo development, elucidation of the molecular functions of ABI3 is essential. While B2 and B3 domains show some DNA binding activity, it is likely that ABI3 interacts with other proteins, including other transcription factors, in order to correctly regulate the large number of processes that occur during embryo maturation. Identification of such proteins should provide new information on the molecular mechanisms controlling this process.

In this paper, we have used the yeast two-hybrid system to clone several ABI3-interacting proteins (AIPs) from Arabidopsis. We have analysed the interactions between these proteins and ABI3, and studied their expression properties.


Identification and analysis of ABI3-interacting proteins (AIPs)

Identification and analysis of proteins that interact with ABI3 will provide molecular information about the mechanism of ABI3 action in activating embryo maturation and repressing germination. In order to identify such proteins, we used the yeast two-hybrid interaction cloning protocol ( Fields & Song 1989). Two derivatives of ABI3 were initially tested for use in this system, a C-terminal region of ABI3 containing the B2 + B3 domains (pABI3B2 + B3), and a derivative also containing the B1 region (pABI3B1 + B2 + B3) ( Fig. 2a). In separate experiments, these were fused to the GAL4 binding domain and tested for self-activation in yeast. Both constructs were separately introduced into the yeast strain YRG-2 which was then tested for growth on media without histidine (–LTH media) and expression of the LacZ reporter gene. For both constructs, transformants did not grow on –LTH media and did not turn blue when grown on plates supplemented with leucine and histidine, indicating that the bait fusion protein alone could not activate transcription. Although introducing the pABI3B1 + B2 + B3 construct did not result in growth on –LTH media or reporter gene expression on its own, a small-scale test screen with the two-hybrid target cDNA library resulted in the growth of a very large number of colonies, indicating non-specific activation of the reporter genes (data not shown). Therefore, the pABI3B2 + B3 construct was used for further experiments. This plasmid was co-transformed into yeast with a mixed Arabidopsis silique and mature seed cDNA library, produced in the target plasmid pADGAL4, containing a fusion of the GAL4 activation domain with plant cDNAs. Transformants (1.4 × 105) were selected for growth on –LTH media.

Figure 2.

Interactions of AIP proteins with ABI3.

(a) Interaction of AIPs and ABI3 in yeast. (b) Interaction of the wild-type CO domain and mutant CO derivative domains with ABI3B2 + B3 in yeast. Diagrammatic representations of ABI3 and the ABI3 bait protein are shown. The activation (A) and basic (B1, B2 and B3) domains are shaded light and dark grey, respectively. Growth of yeast on –LTH media, and β-galactosidase activity (units ml−1 min−1, ± indicates standard deviation from the mean) from cultures of these colonies are shown. (c) Interaction of AIPs with ABI3B2 + B3 in vitro. Separate panels for each interaction are shown, in each case, lane 1 = purified AIP-his protein (load 1), lane 2 = E. coli soluble extracts containing recombinant ABI3B2 + B3 (load 2), lane 3 = specific eluate from columns following extensive washing to remove unbound proteins. Control refers to the protein encoded by the vector pET32(a)+, and contained a 20.3 kDa protein with the following structure: Trx.Tag-His.Tag-S.Tag (Novagen). The positions of the molecular weight markers are indicated.

Colonies that grew on –LTH media were tested for the expression of the second reporter gene, LacZ. Plasmids from colonies that showed β-galactosidase activity were recovered and the DNA sequence of cDNA inserts was determined. In this study, the identification and analysis of three cDNA clones identified from this screen (ABI3-interacting proteins, AIPs), and a fourth AIP identified as the Arabidopsis homologue of the A. fatua AfVIP3 protein ( Table 1) is reported.

Table 1. . ABI3-interacting proteins (AIPs) identified in the two-hybrid screen
Protein nameEMBL accession numberChromosome locationGene copy numberProtein homologyPredicted cellular localization
  1. The EMBL accession number for each cDNA sequence is shown along with the chromosome location and gene copy number. The homologies of the predicted proteins are shown, together with predicted cellular localisations. aPutterill et al. (1995) ; bBouchard & Cote (1993); cJones et al. (2000) ; dIljima et al. (1996) ; ePati & Weissman (1989); fWoychik et al. (1990) .

AIP1AJ25108651Arabidopsis CONSTANS aNuclear (82%)
AIP2AJ251087Unallocated?Drosophila GOLIATH b, mesoderm formation Nuclear (43%)
AIP3 cAJ25108811Human C1 d, cell cycle control Cytoplasmic 47%, nuclear 17%
RPB5AF01924831Human/yeast RNA pol II, RPB5 subunite, fNuclear

Full-length clones for these AIPs were obtained by screening a silique cDNA library with cDNA inserts derived from plasmids obtained from the two-hybrid screen. In addition, 5′ RACE ( Chenchik et al. 1995 ; Chenchik et al. 1996 ) was used to obtain the full-length sequence of AIP1.

The AIP1 cDNA encodes a 618 amino acid protein ( Fig. 1a), and the ABI3-interacting region is located at the C-terminus, from residue 528–575. AIP1 has two direct repeats, residues 275–319 and residues 325–369. The N-terminal region (residues 141–273) of this protein shows some homology to two component response regulators from Arabidopsis ( Sakai et al. 1998 ). The predicted AIP1 protein shows high homology from residues 528–575 to the plant transcription factor CONSTANS ( Putterill et al. 1995 ). This 47-residue region shares 23 identical residues and eight similar residues with CONSTANS (CO); there is no other region of similarity to CONSTANS or COL (CO-like) proteins in the rest of AIP1 ( Fig. 1b). The protein localization predictor program PSORT indicated that AIP1 is located in the nucleus ( Table 1).

Figure 1.

Protein structures of AIP1, AIP2 and RPB5.

(a) (c) and (d) are schematic representations of the predicted AIP1, AIP2 and RPB5 proteins, respectively (numbers represent amino acid residues). The regions identified from the two-hybrid screen with pABI3B2 + B3 are depicted as dashed lines. The delimited ABI3-interacting regions are shown as solid lines. (a) AIP1; homology to CONSTANS is shaded light grey, the region showing homology to the two-component response regulators is shaded dark grey. The direct repeats are depicted by arrows. (b) Alignment of the predicted AIP1 protein with CONSTANS, CONSTANS-LIKE (COL1, COL2) proteins and homologues of CONSTANS from Brassica napus, Bn1CON10, Bn1CON19 and Bn9CON10. Amino acids that are identical in all proteins are shown in dark grey. Amino acids that are similar are shown in lighter grey. The substitutions in the mutant CO proteins are indicated by an asterisk, co5 (residue 566) and co7 (residue 567). (c) AIP2; the region showing homology to the Drosophila GOLIATH protein is shaded grey. (d) RPB5; the corresponding region of the human RPB5 that interacts with the human HBx protein is shown (dot-dashed line).

The AIP2 cDNA encodes a 310-residue protein ( Fig. 1c) that shows homology to the Drosophila transcription factor GOLIATH ( Bouchard & Cote 1993), which is involved in the regulation of gene expression during mesoderm formation and belongs to the C3HC4 family of zinc finger proteins ( Freemont et al. 1991 ). The ABI3 interacting region of AIP2 is located at the C-terminus (amino acids 173–310) ( Fig. 1c) and this protein has a predicted location in the nucleus ( Table 1).

We have identified the Arabidopsis RNA polymerase II subunit RPB5 as an AIP. The region of RPB5 that interacts with pABI3B2 + B3 is located towards the C-terminus (amino acids 162–205) ( Fig. 1d).

Analysis of interactions in yeast

The interaction of the B2 + B3 portion of ABI3 with AIP1, AIP2 and RPB5 was analysed in vivo using the yeast two-hybrid system ( Fig. 2a). Transformation of yeast with the pABI3B2 + B3 bait and each AIP target plasmid resulted in strong growth under histidine selection. Transformation of yeast with the pABI3B2 + B3 bait and target plasmid containing no cDNA insert resulted in no yeast growth (data not shown). The AIP targets were tested with the bait vector containing no cDNA insert, and no growth was observed on –LTH media (data not shown). Assays for β-galactosidase activity were carried out using protein extracts from yeast strains containing bait and target plasmids and showed a high level of expression of the LacZ reporter gene in all cases ( Fig. 2a). To delineate more closely the regions of interaction within the ABI3 B2 + B3 protein, AIPs were tested for interaction with a bait construct containing only the B3 domain (pABI3B3). For AIP1 only, co-transformation with this construct allowed growth on –LTH media. However, it was not possible to detect expression of the LacZ reporter gene in these colonies. Therefore, it is not clear whether growth was due to ‘leaky’ expression of the HIS3 gene or a very low level of LacZ reporter gene activation. For other AIPs, growth on –LTH plates was only detected following co-transformation with the pABI3B2 + B3 bait construct.

The domain identified in AIP1 with high sequence identity to CONSTANS and COL proteins was examined further to more closely define the interaction with ABI3. The full-length CONSTANS open reading frame, fused to the GAL4 activation domain, allowed growth of yeast on –LTH media and activated expression of the LacZ reporter gene when co-transformed with the pABI3B2 + B3 bait plasmid (data not shown). The 47-residue region from CONSTANS with high sequence identity to AIP1 (see Fig. 1b) was fused in-frame in the pADGAL4 target plasmid. This construct (pCOWT) was co-transformed with the ABI3 bait construct, pABI3B2 + B3, into yeast cells and tested for growth on –LTH media. His+ colonies were obtained and these showed high levels of β-galactosidase activity ( Fig. 2b), demonstrating that the COWT domain interacts in vivo with ABI3 B2 + B3. Two mutant alleles of CO, co5 and co7, have been shown to disrupt CO function ( Putterill et al. 1995 ; Robert et al. 1998 ). Both mutations are amino acid substitutions in the C-terminal domain of CONSTANS that is identical to AIP1 ( Fig. 1b). In co5, proline at residue 566 is substituted with leucine, whereas in co7, arginine at position 567 is substituted with glutamine. To study whether the substitution of these residues within this domain had any affect on interaction with ABI3, constructs were produced that incorporated these mutations separately into the 47-residue domain (constructs pco5 and pco7). These were cloned into the target plasmid pAD-GAL4 and tested in yeast with the ABI3B2 + B3 bait; His+ colonies were obtained in both cases. However, the pco7 colonies showed much slower growth in comparison with colonies containing the COWT and pco5 constructs. The His+ colonies were assayed for expression of the LacZ reporter gene. The level of β-alactosidase activity in extracts from both pco5 and pco7 colonies was reduced in comparison to the COWT extracts ( Fig. 2b). These results lead to the conclusion that the 47-residue region in AIP1 and CONSTANS is sufficient for interaction with ABI3. In addition, the substitution of residues that occur in co mutants resulted in a decreased level of reporter gene expression, indicating a role for these residues in this interaction.

Analysis of interaction of AIP2, AIP3 and RPB5 with ABI3 in vitro

To obtain additional evidence for the specificity of AIP interactions with ABI3, the interactions of these proteins with ABI3 were tested using an in vitro assay. These proteins and derivatives of ABI3 were produced in E. coli, using an expression system that allows purification of recombinant proteins through a histidine (his) tag.

The accompanying paper ( Jones et al. 2000 ) describes the cloning of AfVIP3, a protein from A. fatua that interacts with AfVP1 ( Jones et al. 1997 ) and shows strong homology to the human transcription factor C1, which has a presumed role at the G1–S phase transition of the cell cycle. AfVIP3 also showed very high homology (smallest sum probability measured using the BLAST algorithm 5.9e−46) to a presumptive Arabidopsis gene located on chromosome 1 ( Table 1). Expression of the Arabidopsis gene, called AIP3, was analysed (see next section). The open reading frame from this presumptive gene was subcloned into the expression vector pET29a and used to produce recombinant his-tagged protein that was analysed for interaction with ABI3B2 + B3.

For ABI3B2 + B3-his, AIP2-his and AIP3-his, it was possible to purify soluble protein using the pET29a vector system ( Fig. 2c). Purified his-tagged proteins were bound to Ni-NTA-agarose columns and challenged with the untagged ABI3B2 + B3 protein. Soluble bacterial extracts containing ABI3B2 + B3 were passed down columns containing his-tagged AIPs, and, following extensive washing, proteins bound to the columns were eluted. PAGE analysis of final wash fractions showed that all non-specific proteins had been removed (data not shown). Analysis of the eluate fractions from each separate experiment showed that only AIP2/AIP3 and ABI3B2 + B3 were retained on columns ( Fig. 2c). The observation that the untagged protein was ABI3B2 + B3 was confirmed by Western analysis using an antibody raised to the B2 domain (data not shown). Control experiments showed that ABI3B2 + B3 on its own did not bind the Ni-NTA-agarose columns (data not shown), and that when using an unrelated his-tagged protein, ABI3B2 + B3 was not retained ( Fig. 2c). The observation that AIP3 (originally identified by homology) binds ABI3B2 + B3 demonstrates that it is an ABI3-interacting protein. Soluble recombinant Arabidopsis RPB5 could not be produced in E. coli. Purified wheatgerm RNA polymerase II holoenzyme was used to perform affinity chromatography on Ni-NTA-agarose columns and the interaction tested using a his-tagged ABI3B2 + B3 derivative protein (containing the same residues of ABI3 as ABI3B2 + B3). Wheatgerm polymerase was retained on a Ni-NTA-agarose column containing the his-tagged ABI3 derivative, but only at a very low level ( Fig. 2c).

Expression analyses of AIPs

Northern analysis using mature seed poly(A)+ RNA was carried out to analyse the size of AIP transcripts ( Fig. 3a). For AIP1, AIP2, AIP3 and RPB5, transcripts of the expected length were observed in seed poly(A)+ RNA.

Figure 3.

Expression analyses of AIP transcripts.

(a) Northern blot analyses of AIPs in mature Arabidopsis seed. Each lane contains 2 μg poly(A)+ RNA. The size of the RNA transcripts is indicated. (b) RT–PCR analyses of AIP1, AIP2, AIP3, RPB5, CO and ABI3 gene expression patterns in Arabidopsis wild-type, mutant abi3-4, lec1-2 and fus3-3 total RNA extracted from leaf (L), seed and silique (S) and flower (F) tissue. Amounts of RNA used: AIP1 50 ng; AIP2, RPB5 10 ng; AIP3, ABI3 100 ng; CO 5 ng. The negative control contained no RNA, the positive control contained 10 ng of the corresponding cDNA for amplification.

In order to analyse expression patterns of AIP transcripts during development, RT–PCR analysis ( Fig. 3b) was carried out on total RNA from leaf, silique and flower tissue from wild-type and abi3-4 ( Koornneef et al. 1984 ), lec1-2 ( Meinke et al. 1994 ) and fus3-3 mutants ( Keith et al. 1994 ). Expression of the AIP transcripts was observed in all cases throughout development. However, specific differences in tissue and mutant backgrounds were observed. No differences in the expression levels of AIP1 and RPB5 were observed between mutant and wild-type silique tissue. The level of AIP2 expression was reduced in lec1-2 mutant silique tissue, whereas abi3-4 and fus3-3 siliques showed increased abundance of transcripts in comparison with the wild-type. Most noticeably, the abundance of AIP3 transcripts was very low in wild-type silique tissue, but much higher in abi3-4, lec1-2 and fus3-3 silique tissue, all mutants which show premature activation of germination. RT–PCR was carried out to study the expression of ABI3 transcripts in mutant siliques in comparison with wild-type siliques. Very low transcript levels were observed in wild-type tissue, whereas increased levels were observed in lec1-2 and fus3-3 and abi3-4 mutant siliques ( Fig. 3b). It is possible to analyse ABI3 RNA expression in the abi3-4 mutant because the mutation in this allele results in a truncated protein and does not affect mRNA abundance ( Giraudat et al. 1992 ). Expression of the CONSTANS transcript was detected in silique tissue using RT–PCR. The level of expression was low in wild-type silique tissue in comparison with abi3-4 and fus3-3 silique tissue; however, lec1-2 mutant siliques showed very low levels of CONSTANS RNA accumulation ( Fig. 3b).

Genetic analysis of interaction between ABI3 and CONSTANS

Double mutant plants were generated that contained either abi3-4 co5 or abi3-4 co7 combinations, to analyse whether genetic interactions occur between these loci during Arabidopsis development. Analysis of interactions in yeast ( Fig. 2) demonstrated that the B2 + B3 region of ABI3 was required for interaction with the COWT domain. The abi3-4 allele (which converts Gln417 to a premature stop codon; Giraudat et al. (1992) ) results in a protein truncated just before the B2 domain, and may not be expected to interact with the COWT domain in vivo based on experiments in yeast. Flowering time was observed both temporally (days to flowering) and developmentally (numbers of leaves before flowering). This analysis showed that abi3-4 flowered earlier under short- and long-day conditions than wild-type, co5 and co7 ( Fig. 4). The flowering time of abi3-4 co5 and abi3-4 co7 double mutants under short days was indistinguishable from abi3-4, indicating that for both these mutant alleles of co, the genetic function of abi3 is epistatic to that of co. Under long-day conditions, double mutants flowered at an intermediate time between abi3-4 and co5/co7 single mutants, indicating partial epistasis under these growth conditions. Under long days, flowering times of abi3-4 co5/+ or abi3-4 co7/+ heterozygous plants were intermediate between double mutant and abi3-4 phenotypes, consistent with the co-dominant nature of these co alleles. This co-dominance was not observed under short days, where abi3-4 showed complete epistasis.

Figure 4.

Genetic analysis of flowering time.

(a) Days to flowering under long days or short days for WT, mutant and double mutant combinations. (b) Number of leaves at flowering under long days or short days for WT, mutant and double mutant combinations. Standard error shown for each data set (n = 7–30).

A genetic function for CO in seed development has not previously been reported( Putterill et al. 1995 ), and analysis of seeds showed that double mutants were phenotypically indistinguishable from abi3-4 in both colour (seeds were green) and germination potential (fresh seeds showed no dormancy) (data not shown).


The ABI3 protein is an important regulator of Arabidopsis embryo development ( Koornneef et al. 1982 ; Koornneef et al. 1984 ; Nambara et al. 1992 ; Nambara et al. 1995 ), acting as an activator of the late maturation programme in embryogenesis, and, simultaneously, as a repressor of germination ( Nambara et al. 1995 ; Parcy et al. 1994 ). In order to regulate embryo maturation, it is likely that ABI3 interacts with other proteins, including other transcription factors. The yeast two-hybrid screen described in the present work has allowed the identification of several cDNAs that encode proteins that interact with ABI3.

Two derivatives of ABI3 were tested for use as baits in the two-hybrid system. The longer derivative (containing the B1 + B2 + B3 domains) could not be used in the screen because it was found to interact non-specifically with target cDNAs in yeast. The bait construct ABI3B2 + B3 did not show any reporter gene activation or growth on –LTH media and therefore was considered suitable as a bait for carrying out a two-hybrid screen ( Fig. 2a). For AIP2 and AIP3, it was not possible to delimit the ABI3 interaction region to less than a region of ABI3 containing both B2 + B3 domains. It is possible that the structural conformation of ABI3 (i.e. the inclusion of both B2 and B3 domains) is important for protein–protein interactions in these cases. It has previously been shown that the DNA binding activity of the B3 domain of VP1 is only observed when this region is isolated from the rest of the protein ( Suzuki et al. 1997 ), demonstrating that the conformation of this protein is important for specific functions. AIP1 interacted with the isolated pABI3B3 bait, but we could not detect LacZ reporter gene activity; thus, it is not possible to determine from these results whether this was a specific interaction.

The ABI3-interacting proteins identified in the screen described herein showed sequence homology to known transcription factors or contained known DNA binding domains ( Table 1). Since ABI3 is involved in complex co-ordination of embryo maturation events, along with other transcription factors such as LEC1 and FUS3 ( Parcy et al. 1997 ), it is not unexpected that we identified several different classes of transcription factors. The AIPs described here were expressed throughout development, as opposed to the embryo-specific expression of ABI3 ( Fig. 3b). These proteins may therefore play a more general role in transcription, with ABI3 providing the developmental specificity for activation during embryo maturation. We observed a lower level of expression of ABI3 transcripts in wild-type siliques in comparison with that in lec1-2 and fus3-3 silique tissue ( Fig. 3b). Previous studies have shown that ABI3 protein levels are reduced in silique extracts from lec1 and fus3 mutants; our results therefore extend this observation and agree with the previous suggestion that LEC1 and FUS3 regulate the level of ABI3 expression ( Parcy et al. 1997 ).

The RPB5 subunit of RNA polymerase II, which is highly conserved in eukaryotes, was identified as an AIP. Recently, evidence has been found to indicate that this subunit plays a specific role in transcriptional activation in humans and yeast. It has been reported that yeast RPB5 affects transcriptional activation at specific promoters ( Miyao & Woychik 1998). Human RPB5 has been shown to bind to the transcriptional activator, TAFII ( Bertolotti et al. 1998 ), and to interact with the hepatitis B virus X activator protein HBx ( Cheong et al. 1995 ). Both HBx and RPB5 specifically bind to the general transcription factor TFIIB and there is evidence of trimeric complex formation for HBx transactivation ( Lin et al. 1997 ). HBx interacts with the mid-region of RPB5 ( Fig. 1d) ( Cheong et al. 1995 ), whereas ABI3 interacts with the C-terminal region of the Arabidopsis RPB5 subunit ( Fig. 1d). HBx cannot bind DNA directly, hence, protein–protein interactions are vital for HBx transactivation ( Siddiqui et al. 1987 ). This observation is analogous to that of ABI3/VP1, where protein–protein interactions have been suggested to be an important mode of VP1 action ( Busk & Pages 1997). Although only a relatively weak interaction between RPB5 and ABI3 was obtained by affinity chromatography using the wheatgerm polymerase II and his-tagged ABI3B2 + B3, the in vivo and in vitro results together suggest that RPB5 may have a role in transcriptional activation by ABI3.

AIP1 shows strong homology to the Arabidopsis flowering time regulatory protein, CONSTANS (CO). This homology to CONSTANS is restricted to a 47 amino acid residue region at the C-terminus, with 23 identical residues and eight similar residues ( Fig. 1b). Along with homology to CONSTANS, AIP1 also shares homology in this same region with Arabidopsis CONSTANS-LIKE (COL) proteins and homologues of CONSTANS from Brassica napus. Specific interaction of ABI3B2 + B3 with the isolated 47 amino acid CO domain was demonstrated in yeast. Although the exact role of the C-terminal region has not yet been defined, the presence of positively charged amino acids may indicate a role as a nuclear localization signal ( Putterill et al. 1995 ; Robert et al. 1998 ). Mutants in the CONSTANS gene, co5 and co7, have single amino acid substitutions within this 47 amino acid C-terminal region, and interaction of mutated versions of this domain with ABI3 was reduced in comparison with the wild-type domain. This indicates the possible importance of these amino acid residues for this interaction with ABI3. Therefore, ABI3 binds to this domain in AIP1, which suggests that there may be an interaction in vivo between ABI3 and CO. Although expression of CO is highest in leaf tissue, RT–PCR revealed lower levels of CO transcripts in silique tissue ( Fig. 3b), which has not been reported previously.

Genetic interactions between ABI3 and CO were studied using the double mutant combinations of abi3-4 co5 and abi3-4 co7. Based on results in yeast, the abi3-4 mutant protein should not interact with CO, because it does not contain the B2 + B3 domains shown to interact with the COWT domain ( Fig. 2). Although no observable differences in seed colour and germination capacity were observed between double mutants and abi3-4, genetic analysis of ABI3 function in relation to flowering showed that for the abi3-4 allele, plants flowered earlier than wild-type under both long- and short-day growth conditions ( Fig. 4). Analysis of the phenotypes of abi3 co double mutants showed that abi3 genetic function is epistatic to co, suggesting that expression of the B2 + B3 region of ABI3 during seedling development is necessary to observe the co5 and co7 mutant phenotypes, and indicating an interaction between these regions of ABI3 and CO proteins in vivo. Therefore, based on these genetic studies, one function of CO appears to be the repression of ABI3 function. Interestingly, it has been shown recently that both ABI3 expression and genetic functions are not limited to the seed ( Rohde et al. 1999 ). These authors showed that ABI3 is expressed during vegetative quiescence processes in Arabidopsis, particularly in the apex, but this expression is lost in severe abi3 mutants. In addition, it has also been shown that there is a genetic interaction between ABI3 and DET1 in the control of flowering ( Rohde et al. 2000 ). The ABI3 locus has been associated with induction and maintenance of embryo dormancy and inhibition of apical meristem growth in the seed ( Nambara et al. 1995 ). Our results, in conjunction with those of Rohde et al. (1999, 2000) , suggest that ABI3 is also required for quiescence in the developing seedling, and that removal of ABI3 function leads to removal of quiescence–associated processes, one result being the rapid induction of flowering (in comparison to wild-type plants). In this context, it is interesting that one AIP encodes a putative transcription factor with function in the cell cycle (AIP3; see below), which may also play a part in quiescence-related processes.

The N-terminal region of the predicted AIP1 also showed strong homology to the phosphate receiver domain within the two-component response regulators, ARR1 and ARR2, from Arabidopsis ( Sakai et al. 1998 ). The Arabidopsis genes, ETR1 and CKI1, which have a role in ethylene and cytokinin responses, also encode proteins similar to bacterial two-component response regulators ( Chang et al. 1993 ; Kakimoto 1996).

AIP2 encodes a C3HC4 type of zinc finger protein ( Freemont et al. 1991 ) and shows homology to the Drosophila developmental protein, GOLIATH ( Bouchard & Cote 1993), that is a variant zinc finger transcription factor, which may be involved in mesoderm formation. AIP2 expression was reduced in the embryo-specific lec1-1 mutant in comparison to wild-type, whereas there was an increased level of expression in fus3-3 and abi3-4 mutant silique tissue in comparison with wild-type ( Fig. 3b). This suggests that AIP2 may be differentially regulated by ABI3, FUS3 and LEC1. Similar differential regulation has previously been observed for the MYB factors, AtMYB13 ( Kirik et al. 1998a ), ATMYBR1 and ATMYBR2 ( Kirik et al. 1998b ).

The Arabidopsis AIP3 protein was identified as a homologue of the A. fatua protein, AfVIP3, reported in the accompanying paper ( Jones et al. 2000 ), that interacts with the A. fatua homologue of VP1 (AfVP1) ( Jones et al. 1997 ). This protein shows homology to the human C1 gene that is thought to be involved in regulation of the G1–S phase transition during the cell cycle. The embryo cells of dormant A. fatua seeds are arrested at G1 and DNA replication does not occur until the onset of germination ( Elder & Osborne 1993). Whether an analogous situation exists in Arabidopsis is not known. Interestingly, the pattern of expression of AIP3 and the A. fatua homologue in dormant and non-dormant Arabidopsis/A. fatua seed is similar. The A. fatua homologue shows higher levels of expression in non-dormant embryos in comparison with dormant embryos ( Jones et al. 2000 ). Similarly, the expression of AIP3 is very low in wild-type Arabidopsis siliques but much higher in abi3-4, lec1-2 and fus3-3 mutant siliques. These mutants all show reduced dormancy and premature activation of the apical meristem and other germination-specific processes.

The ABI3 transcription factor has been shown to function as both an activator of embryo maturation and as a repressor of germination. The AIP proteins identified in this study may function in either or both processes. In order to determine the function of these proteins during embryo development, it will be necessary to produce and analyse plants containing mutations in the corresponding genes. Analysis of the phenotypes of these plants should allow an evaluation of the function of the AIPs in the activation of embryo maturation and repression of germination, and other aspects of development.

Experimental procedures

Plant material and growth conditions

Arabidopsis thaliana plants were grown at 18–23°C, 16 h light (long day) and 75–80% relative humidity in controlled-environment growth rooms. For short-day conditions, plants were grown in 10 h light under similar temperature and humidity conditions. Seeds were stored at 24°C in the dark until use.

Yeast strain and plasmids

Saccharomyces cerevisiae YRG-2 (Matαura3-52 his3-200 ade2-101 lys2-801 trp1-901 leu2-3112 gal4-542 gal80-538 LYS2:: UASGAL1-TATAGAL1-HIS3 URA3::UASGAL417 mes(×3)-TATACYC1-LacZ was used for the two-hybrid assay ( Callahan et al. 1995 ), together with the phagemid vectors pBD-GAL4, pBDGAL4Cam and pAD-GAL4 ( Mullinax & Sorge 1995).

The pABI3B1 + B2 + B3 bait plasmid was constructed by digesting the ABI3 cDNA with PvuII and HincII, and the resulting 2120 bp fragment (containing amino acids 115–720) was cloned into the SmaI site of pBDGAL4 (Stratagene, La Jolla, CA, USA). To obtain the pABI3B2 + B3 bait plasmid, the ABI3 cDNA was digested with PstI and NsiI, and the resulting 900 bp fragment (containing amino acids 439–720) was cloned into the PstI site of pBDGAL4. The pABI3B3 bait was obtained by PCR amplification using Pwo polymerase (Hybaid, Middlesex, UK) according to the manufacturer’s recommendations, using BR3-5′ (5′-CTACCATGGAC- CGAGCTGGCTC-3′) and BR3-3′ (5′-CTGTTTTTCATTTAACAGT- TTG-3′). This DNA fragment (amino acids 538–720) was cloned into the SmaI site of pBDGAL4Cam.

DNA corresponding to the wild type CO domain (COWT) was amplified from pGAD424/CO (a generous gift of Dr George Coupland, John Innes Centre, Norwich, UK) using CO-5′ (5′-AACTCAGTCCAATGGACAGAG-3′) and CO-3′ (5′-TCTCTTTGCGA- ACCGGCC-3′). The co5 allele was amplified using CO-5′ and co5-3′ (5′-TCTCTTTGCGAACCGGCCATTGACCCGCAGTCT-3′ mutated nucleotide in bold). The co7 allele was amplified using CO-5′ and co7-3′ (5′-TCTCTTTGCGAACCGGCCATTGACCTGCGGTCT-3′ mutated nucleotide in bold). The resulting 145 bp fragment in each case was cloned into the SmaI site of pADGAL4Cam.

Yeast two-hybrid selection

YRG-2 competent cells (0.1 ml) were used for co-transformation of the pABI3B2 + B3 bait plasmid with 20 μg of the Arabidopsis silique and seed cDNA library in the pADGAL4 target vector. Transformants were plated onto media lacking leucine, tryptophan and histidine (–LTH) according to the Stratagene YRG-2 Competent Yeast Cell Kit Manual.

An overlay assay to analyse β-galactosidase activity in yeast cells growing on solid media was carried out as previously described ( Duttweiler 1996). Plasmid DNA was isolated from yeast ( Hoffman & Winston 1987) and transformed into E. coli XL-1 Blue MRF′ (Stratagene). The cDNA inserts from several clones for each AIP were sequenced using the Amersham Life Science Thermo Sequenase Dye Terminator Cycle Sequencing Pre-mix Kit/Perkin-Elmer ABI Prism dRhodamine Terminator Cycle Sequencing Ready Reaction Kit, and reaction products were assayed using an ABI Prism 377 DNA Sequencer (PE Applied Biosystems, Foster City, CA, USA). DNA sequence analysis was carried out using the MACVECTORTM program (Oxford Molecular Group plc, UK). Sequence data was used to perform BLAST searches of the GenBank database ( The protein localization predictor program PSORT was from

Construction and analysis of cDNA libraries

The target silique and seed cDNA library was constructed in the λ HybriZAP vector (Stratagene) using poly(A)+ RNA derived from mixed silique and mature seed material. The primary library consisted of 0.5 × 106 plaque-forming units. To obtain full-length cDNA clones, a mixed-stage Arabidopsis silique and seed cDNA library (Meinke and Castle, obtained from the Arabidopsis Biological Research Centre, Columbus, OH, USA) was used.

5′ RACE (rapid amplification of cDNA ends) was carried out using the Marathon cDNA amplification kit (Clonetech Laboratories Inc., Palo Alto, CA, USA). RACE-PCR was primed with an oligonucleotide corresponding to position 2149–2174 of the full-length AIP1 cDNA.

Cloning of the AIP3 open reading frame

DNA encoding the AIP3 protein was obtained by ligating together the open reading frames from the two exons of the gene. Using sequence information from BAC F22013, this gene was identified as a homologue of the human C1 transcription factor, position 108964–108833, exon 1; position 108108–107848, exon 2. These were obtained by PCR from wild-type Arabidopsis Landsberg erecta genomic DNA, prepared using the Phytopure plant DNA extraction kit (Nucleon Biosciences, Glasgow, UK). The PCR reactions for each exon were carried out using Pwo polymerase under standard conditions, for exon 1 using primers AIP3-5′1 (5′-ATGATGATGTTGCAGGGGAGT-3′, initiating methionine in bold) and AIP3-5′2 (5′-CTTGGCGGATTTGATATCGTC-3′), and using primers AIP3-3′1 (5′-GAAAAGTGTGAAAATCTAGAG-3′) and AIP3-3′2 (5′-GTCTTCTTCAAGATTGATAGA-3′) for PCR of full-length exon 2. For his-tagging experiments, the primers AIP3-3′1 and AIP3-3′3 (5′-TCTTCTTCAAGATTGATAGA-3′) were used, resulting in a DNA fragment that did not include the final base of the last codon. The exon 1 and 2 DNA fragments, resulting from PCR, were ligated together and cloned into either the SmaI site of pUBS3, for general use, or the EcoRV site of pET29a, for his-tagging experiments.

Quantitative β-galactosidase assay

This assay was carried out as described by Miller (1972).

RNA extraction and expression analysis

Poly(A)+ RNA was isolated for the target cDNA expression library from Arabidopsis Landsberg erecta silique and seed tissue (including silique material from post-fertilization to dehisced, and mature dry seeds) as described previously ( Jones et al. 1997 ). Total RNA was isolated from silique, leaf and flower tissue using the RNeasy Plant mini kit (Qiagen GmbH, Hilden, Germany). Seed poly(A)+ RNA was extracted as described previously ( Rushton et al. 1995 ) but the homogenization buffer was modified to include 10 mM β-mercaptoethanol. RNA was size-fractionated on 1% agarose MOPS–formaldehyde gels ( Sambrook et al. 1989 ) and transferred onto nylon membrane (Magnacharge, Micron Separations Inc., Westborough, MA, USA). cDNA inserts labelled with α(32P)dCTP using the Prime-it II kit (Stratagene) were used to detect corresponding mRNAs. Hybridization conditions were 50% formamide, 6× SSPE, 5× Denhardt’s solution, 0.2% SDS, 100 μg ml−1 denatured herring sperm DNA, 10% dextran sulphate at 42°C for 16 h. Filters were washed once with 2× SSC, 0.5% SDS at room temperature, once for 20 min in 2× SSC, 0.5% SDS at 42°C, once for 20 min in 1× SSC, 0.5% SDS at 42°C, once for 30 min in 0.5× SSC, 0.5% SDS at 42°C (for AIP2 and AIP3), once for 15 min at 60°C in 0.1× SSC, 0.5% SDS (for AIP1 and RPB5). Hybridization was visualised by autoradiography.

RT–PCR was carried out with Ready To Go RT–PCR beads (Pharmacia Biotech, Uppsala, Sweden) using total RNA and following the manufacturer’s recommendations. The oligonucleotides used were: for AIP1 5′-ATGTGGAGAAGAAGACGCATG-3′ and 5′-GTTCCCAAAGCATCATCCTG-3′ for AIP2 5′-ATGGATGCA- TCGTCTTCACCG-3′ and 5′-ACGTACATATATTCACCTCC-3′ for AIP3 5′-GGATGATGATGTTGCAGGGGAGT-3′ and 5′-GTCTTCTTC- AAGATTGATAGA-3′ for RPB5 5′-ATGTTGACGGAAGAGGAGCTC-3′ and 5′-ACAACATAACGATAGGTAAC-3′ for ABI3 5′-GCAGGA- CAAATGAGAGATCAG-3′ and 5′-TCATTTAACAGTTTGAGAAG-3′ and for CO 5′-CGACCCTGTGACACATGCCGG-3′ and 5′-TCTCTT- TGCGAACCGGCC-3′.

One-tenth of each PCR reaction was analysed by Southern blotting following electrophoresis through 1% (w/v) agarose gels ( Sambrook et al. 1989 ). The amplified products were detected by hybridization as described above for Northern analyses. Initial RT–PCR experiments were performed using a range of RNA concentrations to ensure that the PCR reactions were carried out within a range that was quantitative (data not shown). For all AIPs, primers used in the RT–PCR experiments amplified regions spanning introns, and RNA samples were DNAase-treated prior to use.

In vitro analysis of protein interactions

Recombinant versions of ABI3, AIP2 and AIP3 proteins were produced using the pET29a vector system (Novagen, Madison, WI, USA). Full-length open reading frames were amplified from cDNA inserts by PCR with the Pwo proofreading polymerase (Hybaid) under standard conditions using oligonucleotides designed to the 5′ end (from the initiating methionine) and to the 3′ end (up to the penultimate nucleotide of the final codon). DNA products from PCR reactions were cloned into the EcoRV site of pET29a, to produce in-frame derivatives that encode an N–terminal S-tag and a C-terminal his-tag. The his-tagged derivative of ABI3 (ABI3B2 + B3-his) was produced by cloning a PCR product corresponding to nucleotides 1720–2564 of the ABI3 cDNA (up to the penultimate nucleotide of the final codon) into the EcoRV site of pET29a, to produce an in-frame fusion encoding an N-terminal S-tag and a C-terminal his-tag. The untagged ABI3 derivative ABI3B2 + B3 was produced by cloning a PCR product corresponding to nucleotides 1720–2568 (including the stop codon) of the ABI3 cDNA into the EcoRV site of pET29a, to produce an in-frame derivative incorporating an N-terminal S-tag only. Proteins were expressed in E. coli and extracted and purified according to the manufacturer’s recommendations (Novagen), by Ni-NTA-agarose affinity column chromatography.

For interaction studies, his-tagged AIPs and ABI3B2 + B3-his were purified on Ni-NTA-agarose columns, which were then challenged with E. coli soluble extracts containing recombinant ABI3B2 + B3 (or pure wheatgerm polymerase II for ABI3B2 + B3-his interaction experiments, a generous gift of Professor T. Guilfoyle, University of Columbia, MI, USA). Following extensive washing to remove unbound proteins (wash fractions were assayed until no proteins could be detected by polyacrylamide gel electrophoresis (PAGE); Sambrook et al. (1989) ), proteins specifically retained on the columns were eluted with 1 M imidazole and the fractions obtained analysed by PAGE.

Double mutant analysis

F2 seedlings homozygous for abi3-4 from reciprocal crosses between abi3-4 and co5 and co7 were grown under long- and short-day conditions. Genomic DNA from the seedlings was used in CAPS-PCR ( Konieczny & Ausubel 1993) using primers designed to amplify the COWT region (5′-GCCAAGGAG- GTTGCTTCGTG-3′ and 5′-CCCATTTGCACAACAGTAG-3′). The loss of a SacII restriction site within the amplified DNA fragment in co5 and co7 alleles was used as a marker to identify double mutants and heterozygotes. Flowering time was recorded as the number of days to the time the flower bud was visible and the number of leaves in the rosette excluding the cotyledons at that time.


IACR Long Ashton Research Station receives grant-aided support from the Biotechnological and Biological Sciences Research Council (BBSRC) of the UK. S.K. was supported by a BBSRC Cell Commitment and Determination Special Initiative grant, H.D.J. was supported by a BBRSC ROPA grant. We are grateful to Dr John Lenton for critical reading of the manuscript, Dr J. Giraudat for ABI3 cDNA, Dr G. Coupland for pGAD424/CO and co5 and co7 alleles, Professor T. Guilfoyle for pure wheatgerm RNA polymerase II and for hosting S.K. in his laboratory, Dr P. McCourt for fus3-3 seed, Dr M. Koornneef for abi3-4 seed and Dr J. Harada for lec1-2 seed.


  1. EMBL accession numbers AJ251086( AIP1), AJ251087( AIP2) and AJ251088( AIP3).