Control of embryonic meristem initiation in Arabidopsis by PHD-finger protein complexes

Apical meristems, located at the growing tips, are indispensable for plant development because these produce all plant organs post-embryonically (Weigel and Jürgens, 2002). The meristems and the stem cells contained within these are formed during embryogenesis. The first manifestation of embryonic root meristem initiation is marked by the specification of an initially extra-embryonic suspensor cell as hypophysis. The hypophysis divides asymmetrically and its small descendant cell will become the quiescent center (QC), which maintains stem cell identity in adjoining cells of the root meristem (reviewed by Möller and Weijers, 2009). Root meristem initiation has been studied mostly using genetic approaches. Few mutations that specifically affect embryonic root initiation have been identified (Mayer et al., 1991), and most of those that have been described converge on the activity of the auxin-dependent transcription factor MONOPTEROS (MP)/AUXIN RESPONSE FACTOR 5 (Hardtke and Berleth, 1998; Weijers et al., 2006) (reviewed by Möller and Weijers, 2009). MP is inhibited by the interacting BODENLOS (BDL)/IAA12 protein. The plant hormone auxin promotes degradation of BDL, thereby releasing MP from inhibition (Hamann et al., 2002). Knowledge about the network operating downstream of MP in root initiation, and the mechanisms of gene regulation by MP beyond inhibition by BDL is fragmented. Recently, a first set of MP targets was identified. Among these, TARGET OF MP 5 (TMO5) and TMO7 genes are directly activated by binding of MP to their promoters. In turn, TMO7 is required for MP-dependent embryonic root meristem initiation (Schlereth et al., 2010). MP also promotes the transport of auxin through controlling PIN1 activity, resulting in auxin accumulation in the future hypophysis that is a fundamental event for MP-dependent embryonic root meristem initiation (Friml et al., 2003; Weijers et al., 2006). In addition to MP and its direct target TMO7, several other factors have been shown to contribute to embryonic root formation. PLETHORA (PLT) proteins, which belong to the AP2-type transcription factor family, are essential for the specification and maintenance of the stem cells (Aida et al., 2004; Galinha et al., 2007), while GRAS family transcription factors SCARECROW (SCR) and SHORT-ROOT (SHR) are important for controlling the radial tissue organization of the root (Di Laurenzio et al., 1996; Helariutta et al., 2000; Sabatini et al., 2003). Although all these transcription factors have been shown to be involved in embryonic root meristem formation, their activity appears to be required after initial MP-dependent initiation. Key unanswered questions are what the connections between these components are, and what mechanisms ensure strict spatial control of these genes.

OBERON1 (OBE1) and OBE2 genes encode plant homeodomain (PHD)-finger proteins and these genes act redundantly in MP-dependent embryonic root initiation (Saiga et al., 2008; Thomas et al., 2009). The PHD-finger domain is found in a wide variety of proteins involved in the regulation of chromatin structure (Taverna et al., 2007). PHD-finger domain is constituted of a conserved Cys4-His-Cys3 zinc-finger domain (Aasland et al., 1995). Recent studies demonstrated that the PHD-finger domain specifically binds to histone H3 trimethylated at lysine 4 (Li et al., 2006; Peña et al., 2006; Shi et al., 2006; Wysocka et al., 2006; Lee et al., 2009), which is associated with nucleosomes near the promoters and 5′ ends of highly transcribed genes (Zhang et al., 2009), and recruit transcription factors and nucleosome-associated protein complexes to chromatin (Saksouk et al., 2009). Interestingly, although PLT1, PLT2, SCR and WOX5 are not expressed in obe1 obe2 double-mutant embryos, MP is normally expressed. As expression of PLT1, PLT2 and WOX5 depends on MP (Aida et al., 2004; Sarker et al., 2007), one possibility is that OBE1 and OBE2 act to control embryonic root meristem formation downstream or at the level of MP (Saiga et al., 2008). However, the function of OBE proteins in the MP pathway is not known.

Here, we demonstrate the role of PHD-finger proteins involved in MP-dependent embryonic root initiation in Arabidopsis. TITANIA1 (TTA1) and TTA2 genes, which are closest homologs of OBE1 and OBE2, are functionally redundant and required for MP-dependent embryonic root initiation. Our data show that OBE1 locally mediates the activation of TMO5 and TMO7 genes. Construction of triple and quadruple mutants among obe1, obe2, tta1 and tta2 showed that OBE1/2 and TTA1/2 also act redundantly in embryogenesis. Our findings suggest that activation of transcription factor genes during root initiation requires the activity of a PHD-finger protein complex.

The Arabidopsis thaliana Columbia (Col-0) ecotype was used as the wild type. The tta1-1 (SALK_042597), tta2-1 (SALK_082338) and tta2-2 (SALK_016218) mutants were obtained from the Arabidopsis Biological Resource Center (ABRC). obe1 and obe2 mutants, OBE1p::OBE1-GFP, TMO5p::3×nGFP and TMO7p::3×nGFP transgenic lines have been described previously (Saiga et al., 2008; Schlereth et al., 2010). Plants were grown on MS agar plates containing 1% sucrose or on rock-wool bricks surrounded by vermiculite under long-day conditions (16 hours light/8 hours dark) at 22°C.

For the TTA1p::TTA1-GFP and TTA2p::TTA2-GFP constructs, genomic regions corresponding to 4780 bp upstream from the TTA1 stop codon TGA and corresponding to 5333 bp upstream from the TTA2 stop codon TAA, respectively, were cloned into pGEM-T (Promega) then subcloned into pBI-GFP (Saiga et al., 2008).

For the MPp::OBE1 and ARF13p::OBE1 constructs, the coding region of OBE1 was cloned into pGreenII BAR (pGreen-OBE1). MP and ARF13 promoter fragments (Schlereth et al., 2010) were introduced into pGreen-OBE1.

All constructs were transformed into wild-type or obe1/+ obe2 plants by the floral dip method (Clough and Bent, 1998).

For observation of embryos, histological analysis and microscopy were performed as described previously (Saiga et al., 2008).

In situ hybridization was performed as described previously (Saiga et al., 2008). The MP, PLT1, SCR and WOX5 riboprobes were generated as described previously (Saiga et al., 2008).

ChIP experiments were performed according to Gendrel et al. (Gendrel et al., 2005) with minor modification. Globular stage embryos from silique of OBE1p::OBE1-GFP were used to precipitate OBE1-GFP-bound chromatin. For immunoprecipitation, a polyclonal anti-GFP antibody (ab290, Abcam) was used.

Yeast two-hybrid interactions were performed using the HybriZAP-2.1 Two-Hybrid Predigested vector kit (Stratagene). The open reading frames of OBE1, OBE2, TTA1 and TTA2 were amplified from wild-type cDNA using gene-specific primers. Amplified DNA fragments were subcloned into pGEM-T and subsequently cloned into pAD-GAL4-2.1 and pBD-GAL4 Cam. The bait and prey constructs were transformed into the yeast strain YRG-2. Mating and selection for interactions were performed according to the manufacture’s protocol (Stratagene). All experiments were repeated at least three times.

For immunoprecipitation, 1 g of OBE1p::OBE1-GFP and Col-0 siliques were ground in a mortar with liquid nitrogen. Protein extraction, immunoprecipitation and mass spectrometry were performed as reported by Zwiewka et al. (Zwiewka et al., 2011) with minor modifications. nLC-MS/MS analysis was carried out using a LTQ-Orbitrap. Data were analyzed using the Bioworks software package version 3.1.1 (Thermo Scientific).

We have previously demonstrated that the PHD-finger proteins OBE1 and OBE2 are indispensable for the establishment and maintenance of the both shoot and root apical meristem (SAM and RAM, respectively) (Saiga et al., 2008). In Arabidopsis, there are two close homologs of OBE1 and OBE2, and we named these TITANIA1 (TTA1) and TTA2 (Fig. 1A). TTA1 (At1g14740) and TTA2 (At3g63500) proteins share 55% amino acid similarity, suggesting that TTA1 and TTA2 function redundantly, as is the case of OBE1 and OBE2. To test this possibility, we analyzed loss-of-function mutants of TTA1 (tta1-1) and TTA2 (tta2-1 and tta2-2) (supplementary material Fig. S1A). As none of single mutants exhibited obvious phenotypes (data not shown), we generated double mutant combinations of these mutants. All tta1 tta2 double mutants showed seedling lethality (supplementary material Fig. S1B,C) and these phenotypes were completely rescued by introducing TTA1p::TTA1-GFP or TTA2p::TTA2-GFP (supplementary material Fig. S1D; data not shown). These observations indicate that TTA1 and TTA2 indeed function redundantly. We used tta2-1 as the tta2 mutant for all further analyses.

tta1 tta2 double mutants exhibited a rootless phenotype and this defect is probably derived from disruption of normal pattern formation during embryogenesis. To determine how embryonic pattern formation is perturbed in tta1 tta2, we examined embryos from self-fertilized plants heterozygous for tta1 and homozygous for tta2 as double homozygous plants died before flowering. It is expected that ∼25% of embryos from these plants might segregate as tta1 tta2 double homozygous.

We found additional and abnormal cell divisions at the embryo proper from the two-cell to 16-cell stage (Fig. 1B,F; data not shown). Although it was observed in only a fraction of embryos from tta1/+ tta2 plants (Table 1), both TTA1p::TTA1-GFP and TTA2p::TTA2-GFP completely rescued those defects (data not shown), indicating that deprivation of both TTA1 and TTA2 is responsible for those phenotypes. At the globular stage, during which the hypophysis divides into a smaller apical cell and larger basal cell in the wild-type embryo (size of apical cell, 5.7±0.5 μm; size of basal cell, 11.1±0.3 μm; n=20), the hypophysis of tta1 tta2 embryos divided abnormally (Fig. 1C,G), resulting in production of two equally size descendants (size of apical cell, 8.1±0.6 μm; size of basal cell, 9.2±0.5 μm; n=20) (Fig. 1D,H). Furthermore, cell division of endodermis/cortex cell files in tta1 tta2 is missing (Fig. 1E,I). These observations suggest that the rootless phenotype in the tta1 tta2 double mutant results from an early defect in embryonic root initiation. After germination, tta1 tta2 seedlings have a variable number of cotyledons (Fig. 1J,K; Table 2) in addition to the rootless phenotype and eventually die after forming the first pair of leaves (Fig. 1L,M).

To address whether only cell division is affected in tta1 tta2 embryos, or whether cell identities are incorrectly specified, we examined the expression of marker genes by in situ hybridization or fluorescence microscopy-based expression analysis. In this analysis, we used embryos obtained from self-fertilized plants heterozygous for tta1 and homozygous for tta2, in which it is expected ∼25% of embryos might segregate as tta1 tta2 double homozygous.

As the hypophysis division defect in tta1 tta2 embryos strongly resembles the mp mutant, we first addressed whether MP expression is lost in tta1 tta2 embryos. Ninety-six percent (23 out of 24) of early globular stage embryos showed wild-type MP expression (data not shown). At the heart stage, MP was still expressed in the tta1 tta2 embryos (Fig. 2A,E), suggesting that the phenotype is not due to a loss of MP expression. We next investigated the expression of the PLT1 and SCR genes. PLT1 is required for the QC specification and is expressed in the basal region of embryo proper at the globular stage in wild type (Fig. 2B) (Aida et al., 2004). However, no PLT1 expression was detectable in 25% (11/44) of embryos (Fig. 2F). SCR is also required for the QC specification, in which it acts in parallel with PLT genes (Sabatini et al., 2003; Aida et al., 2004). SCR is initially expressed in the hypophysis at the early globular stage and is subsequently activated in the ground tissue (Fig. 2C) (Wysocka-Diller et al., 2000). By contrast, of the globular stage embryos, tta1 tta2 embryos (14/57) failed to express SCR (Fig. 2G). These data indicate that specification of the QC is defective in tta1 tta2 embryos. Interestingly, SCR expression was also lost from ground tissue cells (Fig. 2G), which is consistent with the failure of these cells to divide in the double mutant (Fig. 1I). Consistent with a loss of QC identity, WOX5 expression, which initiates in the hypophysis at the globular stage and subsequently becomes restricted in the lens-shaped cell and its derivatives in the wild-type embryo (Fig. 2D) (Haecker et al., 2004), was completely lost in tta1 tta2 embryos (8/30) (Fig. 2H). In summary, MP was still expressed in tta1 tta2 embryos, but expression of PLT1, SCR and WOX5 was lost, suggesting that cell identity specification in this mutant is compromised downstream of MP activity.

As TTA1 and TTA2 are redundantly required for specification of the hypophysis and establishment of the embryonic root, we predicted that TTA1 and TTA2 proteins are expressed in the basal region of embryo proper and/or upper-most suspensor cells at the early globular stage, when the hypophysis is specified (Weijers et al., 2006). To investigate the expression pattern of TTA1 and TTA2 proteins, we generated TTA1p::TTA1-GFP and TTA2p::TTA2-GFP transgenic lines, and analyzed the expression pattern of these throughout embryonic development. Both constructs complemented the double mutant phenotype (supplementary material Fig. S1D; data not shown), indicating that the fusions encode functional proteins. GFP fluorescence was first detected in the two-cell stage embryos (Fig. 3A). At the early globular stage, TTA1-GFP was found both in the basal region of embryo proper and suspensor cells (Fig. 3B), which is consistent with the finding that TTA1 has a role for hypophysis specification. During embryonic development, TTA1 was expressed not only in the basal region but also in the apical region of the embryo proper (Fig. 3C,D). TTA2 displayed the same expression pattern at all stages examined (Fig. 3E-H). Interestingly, despite the ubiquitous expression of both genes, the phenotype resulting from the loss of both genes is remarkably specific to the hypophysis.

Given the finding that tta1 tta2 exhibited similar defects as observed in obe1 obe2 embryos (compare with Saiga et al., 2008), we generated multiple mutant combinations among these mutants. We found that obe1 tta1, obe1 tta2, obe2 tta1 and obe2 tta2 double mutants exhibited no obvious phenotypes. This result suggests that the OBE1/2 and TTA1/2 proteins are not simply redundant, but rather that one protein from each pair is required for normal development. To determine the consequences of progressively eliminating the entire OBE1/2 TTA1/2 clade, we next analyzed embryos of obe1 obe2 tta2 triple mutants from an obe1/+ obe2 tta2 mother plant. Among the progeny of such plants, 20% of embryos exhibited embryonic lethality (Table 3). obe1 obe2 tta2 did not show novel phenotypes in the basal region where the formation of the embryonic root meristem is already disrupted in obe1 obe2 and tta1 tta2 embryos. However, development of the apical region where cotyledon primordia and shoot apical meristem are produced was disturbed (Fig. 4A-C,E-G). During transition from triangular to heart stage in wild-type siblings in the same silique, cotyledon primordia had correctly emerged (Fig. 4A,B); however, emergence of cotyledon primordia was not observed in obe1 obe2 tta2 embryos (Fig. 4E,F). In addition, the apical region of obe1 obe2 tta2 embryos was abnormally expanded compared with wild-type siblings (Fig. 4B,F). obe1 obe2 tta2 triple mutant embryos arrested at the triangular stage (Fig. 4C,G). We further investigated other triple mutant combinations and found that all of them showed same phenotypes (data not shown). Finally, we investigated the phenotypes of obe1 obe2 tta1 tta2 quadruple mutants. We found that ∼5% of embryos from obe1/+ obe2 tta1/+ tta2 mother plants were swollen when wild-type siblings were at the bent-cotyledon stage (Fig. 4D,H; Table 3). These data indicate that although the OBE1/2 and TTA1/2 pairs are not redundant in root formation, all four proteins function redundantly in development of the apical pole, as well as in progression beyond the triangular stage of embryogenesis.

The genetic interactions seen among OBE1/2 and TTA1/2 proteins are consistent with joint requirement of multiple proteins for biological function, for example, in a protein complex. To determine whether protein complexes can be formed among these proteins, we initially tested all possible pairwise combinations between OBE and TTA proteins, including their homodimers, using a yeast two-hybrid assay. In this assay, all tested interactions were positive (Fig. 5A), suggesting extensive interaction potential among all proteins.

To determine whether complexes involving multiple OBE/TTA proteins are found in vivo, we used a translational fusion construct for OBE1-GFP that was previously shown to be functional (Saiga et al., 2008). We isolated the OBE1-GFP protein complex using immunoprecipitation on silique tissue. Next, associated proteins were identified by tandem mass spectrometry. Strikingly, in addition to OBE1, peptides uniquely representing OBE2, TTA1 and TTA2 were identified in pull-down experiments with OBE1-GFP siliques, but not with wild-type siliques (Fig. 5B; Table 4). Given the finding that TTA and OBE proteins act downstream or at the level of MP, we analyzed the mass spectrometry results for MP peptides, but did not find any. Hence, there is no evidence for a direct association between OBE1 and MP. These results demonstrate that OBE1 is found in complex with other OBE/TTA proteins, although this experiment does not resolve the size or topology of such complexes.

We have shown previously that obe1 obe2 double mutants are defective in embryonic root initiation, resulting in rootless phenotype similar to mp mutants (Saiga et al., 2008). In obe1 obe2 embryos, the expression of PLT1 is lost; however, MP is still expressed, suggesting that OBE1 and OBE2 act downstream or at the level of MP.

Recently, TMO5 and TMO7 genes, both of which encode bHLH type transcription factors, have been identified as direct targets of MP that mediate MP-dependent embryonic root initiation (Schlereth et al., 2010). To determine whether OBE1 is involved in the regulation of these direct MP target genes, we examined TMO5 and TMO7 expression in obe1 obe2 embryos. TMO5 is expressed in cells adjacent to the hypophysis and in cotyledon primordia in wild type (Fig. 6A). Although the expression in the lower domain of the embryo is abolished in obe1 obe2 embryos, the apical expression is maintained (Fig. 6C). The expression of TMO7, which is expressed in the hypophysis-adjacent cells in wild type (Fig. 6B), was completely abolished in obe1 obe2 embryos (Fig. 6D). These findings indicate that OBE1 and OBE2 are required for the expression of TMO5 and TMO7 genes in cells adjacent to the hypophysis. In addition, the observation that only basal expression of TMO5 was abolished in obe1 obe2 embryos suggests different requirements for OBE/TTA gene activity in root and cotyledon patterning, as was also suggested by the genetic analysis.

As MP is still expressed, but not TMO7 (the direct target of MP) in obe1 obe2 embryos, we hypothesized that OBE1 and OBE2 mediate the TMO7 expression through modification of, or binding to, chromatin at the TMO7 locus. To confirm the association of OBE1 with TMO7 promoter region, we performed chromatin immunoprecipitation (ChIP) analysis with the OBE1p::OBE1-GFP transgenic line that could rescue the defects of obe1 obe2. Three DNA fragments in the TMO7 promoter region were enriched using a GFP antibody (Fig. 6E,F; supplementary material Fig. S2), demonstrating in vivo binding. Interestingly, the binding profile of OBE1 along the tiles chosen for the TMO7 locus closely resembled that of MP binding, as previously demonstrated (Schlereth et al., 2010), suggesting that a functional interaction may exist.

It has been demonstrated that while the TMO7 transcript is expressed in the cells adjacent to the hypophysis, the TMO7 protein moves to the hypophysis where it acts to mediate root formation (Schlereth et al., 2010). If OBE1 mediates root formation in part by controlling TMO7, one would predict a requirement for OBE1 in the cells adjacent to the hypophysis but not in the hypophysis itself. As OBE1 is ubiquitously expressed at this stage (Saiga et al., 2008), it cannot be deduced where its activity is required. To determine the domain of OBE1 activity in root formation, we misexpressed OBE1 in obe1 obe2 mutants from two different promoters, and observed embryonic root initiation of those plants. OBE1 driven by MP promoter, which is expressed in the cells adjacent to the hypophysis but not in the hypophysis itself (Schlereth et al., 2010), could rescue the embryonic root initiation defects in obe1 obe2 (Fig. 6G). By contrast, OBE1 expression driven by the suspensor-specific ARF13 promoter (Schlereth et al., 2010) in ARF13p::OBE1 lines did not rescue the defects in obe1 obe2 roots (data not shown). These data indicate that OBE1 in the cells to the adjacent to the hypophysis is crucial for the embryonic root initiation, and OBE1 is important for TMO7 expression but not for its protein function.

Our results indicate that TTA1 and TTA2 are redundantly required for embryonic root initiation in Arabidopsis. The observations that cell divisions of the hypophysis of tta1 tta2 are defective, and that MP is expressed but PLT1, SCR and WOX5 are absent in tta1 tta2 embryos suggest that the rootless phenotype observed in tta1 tta2 is mainly derived from disruption of the hypophysis specification.

TTA1 and TTA2 seem to function in the same pathway in which OBE1 and OBE2 act because: (1) phenotypes of tta1 tta2 double mutants are similar to those of obe1 obe2; (2) the expression patterns of cell identity marker genes are identical to those of obe1 obe2; (3) expression patterns of all four proteins completely overlap; and (4) OBE1/2 and TTA1/2 proteins could interact with each other in vivo.

Because MP is present in the adjacent cells to the future hypophysis but not in the hypophysis itself, it follows that MP promotes the hypophysis specification in a non-cell-autonomous manner (Hardtke and Berleth, 1998). TMO7 expression is activated by MP in the adjacent cells to the hypophysis and TMO7 protein moves to the hypophysis. Our findings indicate that OBE1 mediates the MP-dependent TMO7 expression because: (1) the expression of TMO7 but not MP is lost in obe1 obe2 embryos; (2) OBE1 associates with the TMO7 promoter region; and (3) OBE1 function in the adjacent cells to the hypophysis but not in the hypophysis is required for embryonic root initiation (Fig. 7). OBE1, OBE2, TTA1 and TTA2 expression seem not to be regulated by MP, and protein complex identification with either OBE1 or MP failed to detect interactions between the proteins (B.M. and D.W., unpublished), suggesting that MP does not interact with OBE proteins. Although these findings implicate the function of chromatin regulators in the MP pathway, a key issue is through what molecular mechanisms these PHD finger proteins control TMO5/7 expression. One possibility is that OBE proteins are contained in a chromatin remodeling complex such as histone acetyltransferase (HAT), and mediate the transcriptional activation of target genes. Decondensation of nucleosome structure mediated by histone acetylation allows transcription factors to access target genes (Jenuwein and Allis, 2001). In Saccharomyces cerevisiae, the PHD-finger protein Yng1, which interacts with H3K4me3, is contained in NuA3 HAT complex. Yng1 mediates NuA3-dependent H3K14 acetylation through a specific interaction between the PHD-finger domain and H3K4me3, and promotes gene activation (Taverna et al., 2006). As similar mode of action could underlie OBE/TTA function, and this hypothesis awaits the identification of the chromatin mark that is recognized by OBE-TTA complexes. It can of course not be excluded that OBE and TTA do not act through recognition of histone modifications. We have recently identified non-PHD-finger OBE1-interacting proteins by mass spectrometry analysis of the OBE1 protein complex (S.S., B.M. and D.W., unpublished). Future analysis of these proteins could reveal how OBE proteins control the gene expression downstream of the MP.

Auxin is another signal involved in this signaling but its accumulation alone is not sufficient to promote the hypophysis specification (Weijers et al., 2006). Whereas auxin response is activated in extra-embryonic cells below the future hypophysis, TMO7 protein exists only in upper-most extra-embryonic cell, suggesting that accumulation of both auxin and TMO7 is required for the hypophyisis specification (Schlereth et al., 2010). TMO7 expression is absent in obe1 obe2 embryos, whereas the establishment of auxin response maxima in obe1 obe2 embryos is largely similar to the wild-type pattern (Thomas et al., 2009) (S.S., M.A., D.W. and Y.K., unpublished), suggesting that OBE1 mainly controls the expression of the TMO7 rather than establishment of the auxin maxima in the hypophysis specification.

The recent identification of TMO genes as MP targets provides entry points to connect the upstream regulator MP with its several downstream pathways. A key question is how region-specific MP activity is controlled. The analysis of TMO5 and TMO7 expression in obe1 obe2 mutants provides insight into this problem. Although TMO7 is eliminated entirely, TMO5 expression is lost only in the basal embryo domain. This suggests that the requirements for gene activation by MP in the basal and apical embryo domains differ. The precise molecular mechanisms for this regional MP activity remain to be determined, but the OBE proteins should allow dissecting these.

All triple mutant combinations among obe1, obe2, tta1 and tta2 exhibited no additional phenotypes in the formation of embryonic root meristem that are already disrupted in obe1 obe2 and tta1 tta2 double mutants, whereas development of apical region in triple mutants displayed more severe phenotypes than those of double mutants, indicating that OBE1/2 and TTA1/2 function in development of embryonic shoot meristem and cotyledons synergistically. Previously. we have demonstrated that the embryonic shoot meristem of obe1 obe2 might be formed initially but is not maintained because the expression of shoot meristem marker genes WUSCHEL and CLAVATA3 in obe1 obe2 embryos is initiated but is not maintained. By contrast, the embryonic root meristem of obe1 obe2 was not formed, as judged by the expression patterns of root meristem marker genes (Saiga et al., 2008). These suggest that a more complex mechanism operates in the embryonic shoot meristem and cotyledon development, as was also suggested by the differential effect of obe1 obe2 mutations on TMO5 expression in the two embryo poles.

The observation that all triple mutant combinations exhibited more severe phenotypes than those of double mutants is curious because both TTA1 and TTA2 and OBE1 and OBE2 are functionally redundant. One possible explanation is that there are differences in functionality among dimers containing OBE1/2 or TTA1/2. During embryogenesis, hetero-dimer formation might be important. For example, the obe1 obe2 tta1 triple mutant should only have TTA2 homo-dimer, and this results in embryo lethality. However, the obe1 obe2 double mutant, in which TTA1-TTA2 hetero-dimer can exist, can form cotyledons and germinate. However, the obe1 obe2 and tta1 tta2 double mutants have no OBE-TTA hetero-dimers and functional embryonic apical meristems were not established. Taken together, our results indicate that OBE-TTA dimer formation might be most important for Arabidopsis embryogenesis. More detailed analysis should elucidate how the OBE-TTA protein complex acts in the apical region during embryonic development. Finally, this work opens up avenues for studying the regulation of developmentally important genes through transcription factors and chromatin proteins.

We thank Sjef Boeren (Biqualis, Wageningen) for help with mass spectrometry, and Akihiko Nakano and Takashi Ueda for helping us with confocal laser scanning microscopy. We also thank Ayako Yamaguchi, Chihiro Furumizu and Kurataka Otsuka for discussions and comments on the manuscript.