Floral induction is controlled by a plethora of genes acting in different pathways that either repress or promote floral transition at the shoot apical meristem (SAM). During vegetative development high levels of floral repressors maintain the Arabidopsis SAM as incompetent to respond to promoting factors. Among these repressors, FLOWERING LOCUS C (FLC) is the most prominent. The processes underlying downregulation of FLC in response to environmental and developmental signals have been elucidated in considerable detail. However, the basal induction of FLC and its upregulation by FRIGIDA (FRI) are still poorly understood. Here we report the functional characterization of the ARABIDOPSIS THALIANA HOMEOBOX 1 (ATH1) gene. A function of ATH1 in floral repression is suggested by a gradual downregulation of ATH1 in the SAM prior to floral transition. Further evidence for such a function of ATH1 is provided by the vernalization-sensitive late flowering of plants that constitutively express ATH1. Analysis of lines that differ in FRI and/or FLC allele strength show that this late flowering is caused by upregulation of FLC as a result of synergism between ATH1 overexpression and FRI. Lack of ATH1, however, results in attenuated FLC levels independently of FRI, suggesting that ATH1 acts as a general activator of FLC expression. This is further corroborated by a reduction of FLC-mediated late flowering in fca-1 and fve-1 autonomous pathway backgrounds when combined with ath1. Since other floral repressors of the FLC clade are not significantly affected by ATH1, we conclude that ATH1 controls floral competency as a specific activator of FLC expression.
The transition from vegetative to reproductive growth is a highly plastic developmental process that requires continuous monitoring of environmental cues and endogenous signals. Antagonistic action of promoting and repressing pathways prevents floral transition until the plant has reached a certain age or size and growth conditions are favourable for sexual reproduction and seed maturation. At a certain time, the balance of promoting and repressing factors is such that the by then competent vegetative shoot apical meristem (SAM) is ‘evoked’ in a florally determined state by the activation of a set of so-called floral pathway integrators, including FLOWERING LOCUS T (FT) and SUPPRESSOR OF OVEREXPRESSION OF CONSTANS1 (SOC1)/AGAMOUS-LIKE 20 (AGL20; reviewed by Bernier, 1988; Boss et al., 2004; Mouradov et al., 2002).
Given its major contribution to the reproductive switch, precise temporal expression of FLC is of utmost importance. The two major pathways that are responsible for repression of FLC are the vernalization and autonomous pathways. Vernalization, the response to winter temperatures, promotes flowering by lowering FLC levels (Michaels and Amasino, 1999; Sheldon et al., 1999). Initial downregulation of FLC is mediated by the PHD domain protein VERNALIZATION INSENSITIVE 3 (VIN3; Sung and Amasino, 2004). After prolonged cold exposure FLC is epigenetically maintained at a low level by the combined action of the chromatin-modifying proteins VERNALIZATION1 (VRN1), VRN2 and LIKE HETEROCHROMATIN PROTEIN 1 (LHP1), rendering plants responsive to the perceived photoperiod in the season that is optimal to commit to flowering (Bastow et al., 2004; Gendall et al., 2001; Levy et al., 2002; Mylne et al., 2006; Sung et al., 2006).
Chromatin-associated proteins also make up part of the second FLC repression pathway, the autonomous pathway. One member, FVE, is homologous to histone deacetylase complex subunits and has been shown to repress FLC by means of histone deacetylation (Ausin et al., 2004; Kim et al., 2004). Initially, a second autonomous pathway component, FLD, was also described being involved in histone deacetylation of the FLC locus (He et al., 2003). However, the human FLD orthologue LSD1 was recently shown to function as a histone demethylase (Shi et al., 2004, 2005; Metzger et al., 2005). Therefore, more likely, lack of FLD leads to disturbance of a chromatin remodelling cascade that acts on both methylation and acetylation of target chromatin. A second group of autonomous pathway components, including FCA, FY, FPA and FLOWERING LATE WITH KH MOTIFS (FLK), are potential RNA-binding or RNA-processing factors (Lim et al., 2004; Macknight et al., 1997; Mockler et al., 2004; Schomburg et al., 2001; Simpson et al., 2003). A seventh member, LUMINIDEPENDENS (LD), encodes a homeodomain protein with unknown function (Lee et al., 1994). A mutation in any of the autonomous pathway components results in increased expression of FLC and late flowering under all photoperiods. Since loss-of-function flc mutations completely suppress the effects of the autonomous pathway mutants, the autonomous pathway components most likely affect flowering solely by the downregulation of FLC (Michaels and Amasino, 2001).
Although the molecular mechanism underlying the activation of FLC by FRI has remained elusive, several essential co-factors in this process have been isolated recently. The emerging picture is that FRI probably acts in a complex containing either one of its homologues FRIGIDA-LIKE1 (FRL1; Michaels et al., 2004) or FRL2 (Michaels et al., 2004; Schläppi, 2006) and SUPPRESSOR OF FRI4 (SUF4) (Kim and Michaels, 2006; Kim et al., 2006). SUPPRESSOR OF FRI4, a putative zinc-finger transcription factor, is most likely accountable for DNA-binding activity of such an FRI-containing complex, as SUF4 has been found to physically interact with both FRI and FRL1 and is capable of binding to FLC chromatin (Kim et al., 2006).
A screen for early flowering mutants in an FRI-containing background further identified the CCCH zinc finger-encoding gene FRIGIDA ESSENTIAL 1 (FES1) (Schmitz et al., 2005). Like SUF4, FRL1 and FRL2, FES1 is necessary for the promotion of FLC expression in an FRI-dependent manner. In addition, FES1 requires SUF4 in order to upregulate FLC and delay flowering (Kim and Michaels, 2006). Epistasis analyses further indicate that FES1 acts cooperatively with FRI and FRL1 in this process rather than that these proteins function in a linear pathway (Schmitz et al., 2005). Whereas SUF4 activity is, at least partially, also responsible for the late-flowering phenotype of ld, fve and fca autonomous pathway mutants, in the case of FES1, and also FRL1, elevated FLC expression in autonomous pathway mutants does not depend on these factors. This indicates that FES1 and FRL1 probably act specifically with FRI to promote FLC expression.
In all situations elevated FLC expression coincides with increased levels of histone H3 lysine 4 (H3-K4) tri-methylation at its locus (He et al., 2004). This type of modification at the 5′-end of genes is generally associated with an active chromatin state, and, in the case of FLC, requires the Arabidopsis POLYMERASE II-ASSOCIATED FACTOR 1 (PAF1) transcriptional activator complex and the EARLY FLOWERING IN SHORT DAYS (EFS)/SET DOMAIN GROUP 8 (SDG8) methyltransferase (He et al., 2004; Kim et al., 2005; Oh et al., 2004; Soppe et al., 1999; Zhang and van Nocker, 2002; Zhao et al., 2005). The EFS/SDG8 is further required for di-methylation of histone H3-K36, a second epigenetic mark associated with active FLC, in regions that contain essential cis-elements for FLC transcription (Zhao et al., 2005). Mutations in members of the PAF1 complex, as well as in EFS/SDG8, also affect the expression of other members of the FLC clade that function in the regulation of flowering time, such as FLOWERING LOCUS M (FLM) and MADS AFFECTING FLOWERING 2 (MAF2) (Ratcliffe et al., 2001, 2003; Scortecci et al., 2001, 2003), suggesting that they play roles in multiple flowering pathways. Similarly, mutations in PHOTOPERIOD-INDEPENDENT EARLY FLOWERING1 (PIE1) suppress both FRI- and autonomous pathway mutation-dependent, FLC-mediated late flowering, but also cause early flowering in non-inductive photoperiods independently of FLC (Noh and Amasino, 2003). Recently, three groups independently reported that mutations in ACTIN-RELATED PROTEIN6 (ARP6)/SUPPRESSOR OF FRIGIDA3 (SUF3)/EARLY IN SHORT DAYS 1 (ESD1) result in phenotypes strikingly similar to pie1 loss-of-function phenotypes (Choi et al., 2005; Deal et al., 2005; Martin-Trillo et al., 2006). It was found that ARP6/SUF3/ESD1 is required for expression of both FLC and the FLC clade members FLM, MAF4 and MAF5. In the case of FLC this involves both H3 acetylation and H3-K4 tri-methylation at its chromatin (Deal et al., 2005; Martin-Trillo et al., 2006). Both PIE1 and ARP6/SUF3/ESD1 show clear similarities with subunits of the ATP-dependent SWR1 and SRCAP chromatin-remodelling complexes identified in yeast and humans, respectively, suggesting that these proteins might be part of a similar protein complex in Arabidopsis (Choi et al., 2005; Deal et al., 2005, 2007; He and Amasino, 2005; Martin-Trillo et al., 2006). In line with this, ARP6/SUF3/ESD1 and PIE1 were found to physically interact both with each other and with a third Arabidopsis homologue of an SWR1/SRCAP-complex component, SERRATED LEAVES AND EARLY FLOWERING (SEF; March-Diaz et al., 2007). Similar to pie1 and arp6/suf3/esd1 loss-of-function mutants, sef plants flower early in both short-day (SD) and long-day (LD) photoperiods and display a strong reduction of FLC and MAF4 mRNA levels (March-Diaz et al., 2007). In yeast and animals, SWR1/SRCAP complexes are recruited by tri-methylated H3-K4 loci and locally adjust the histone composition by substituting histone H2A with the active locus-associated histone variant H2A.Z (Kobor et al., 2004; Krogan et al., 2003; Mizuguchi et al., 2004; Ruhl et al., 2006). Consistently, pie1, arp6/suf3/esd1 and sef mutants show alterations in accumulation of H2A.Z in FLC as well as several other FLC-clade member loci, while other H2 variants remain unchanged (Deal et al., 2007; March-Diaz et al., 2007).
Whereas PAF1 complex, PIE1, ARP6/SUF3/ESD1 and SEF promote FLC transcription, most likely through chromatin remodelling, other factors required for accumulation of FLC mRNA may act post-transcriptionally. These include ABA HYPERSENSITIVE 1 (ABH1), which encodes the large subunit of the eukaryotic nuclear mRNA cap-binding complex, the zinc-finger protein SERRATE that can act in a miRNA gene-silencing pathway and the HUA2 protein that is required for the efficient processing of AGAMOUS (AG) pre-mRNA (Bezerra et al., 2004; Doyle et al., 2005; Grigg et al., 2005). Taken together, most of the loci identified as being necessary for proper accumulation of FLC encode proteins that play broader roles in plant development and that are generally necessary for gene expression. Most of these loci act independently of FRI, and therefore detailed knowledge about the initial induction of FLC expression and its upregulation by FRI is lacking.
Here we report the characterization of the BEL1-like (BELL) family homeobox gene ARABIDOPSIS THALIANA HOMEOBOX 1 (ATH1). This gene was originally isolated in a screen for light-regulated transcription factors (Quaedvlieg et al., 1995). In Lolium perenne, ectopic expression of Arabidopsis ATH1 cDNA results in significantly later heading, but also in an increased leaf biomass due to both an extended vegetative growth phase and increased tillering (van der Valk et al., 2004). The putative role of ATH1 in the flowering process was investigated using expression analysis and a reverse genetics approach. Our results show that ATH1 functions as a floral repressor by specifically affecting FLC expression levels.
Developmental regulation of ATH1 promoter activity
In a previous study, it was found that ATH1 mRNA is most abundant in light-grown seedlings (Quaedvlieg et al., 1995). Here we studied the spatio-temporal localization of ATH1 expression during Arabidopsis development in more detail using a ProATH1:GUS reporter gene construct. To test whether GUS gene expression driven by a 2.6-kb genomic ATH1 promoter-containing fragment corresponds to endogenous ATH1 expression, the abundance of GUS mRNA was measured during dark adaptation of seedlings and consequent de-etiolation. The same responses as seen for endogenous ATH1 expression were observed (Figure 1f; compare with Figure 7 in Quaedvlieg et al., 1995), demonstrating the presence of the main regulatory elements in the ProATH1:GUS construct. Consistent with previous data on ATH1 expression (Quaedvlieg et al., 1995), ATH1 promoter activity was first detectable in 2-day-old light-grown seedlings (data not shown). High levels of GUS activity were present in the SAM, emerging leaf primordia and the vasculature of the cotyledons (Figure 1a,c). No GUS expression was found in the roots. When true leaves developed, strong ATH1 promoter activity was seen in these leaves (Figure 1c,d). In young leaves GUS expression was evenly distributed over the tissue, whereas it became confined to the vascular tissue as the leaves matured (Figure 1d).
We previously observed that in seedlings ATH1 mRNA peaks 2–3 days after germination and then gradually declines when plants grow older (Quaedvlieg et al., 1995). Such a decline was also found for GUS activity in developing ProATH1:GUS plants. The same GUS staining pattern as seen in 4-day-old light-grown seedlings was found up to 7 days after germination (Figure 1a). However, visible GUS activity was no longer detectable in the SAM of older plants (Figure 1b). The expanded and dome-shaped appearance indicates that the SAM has developed from a juvenile into an adult form (Figure 1b). This developmental stage marks the final phase of vegetative development and ends with the transition to an inflorescence meristem (Medford et al., 1992, 1994). Intriguingly, downregulation of ATH1 in the SAM coincides with the commitment time of inflorescence development (Bradley et al., 1997; Mockler et al., 1999). The commitment time is referred to as the developmental stage after which plants are committed to floral initiation and the quality and/or quantity of light has little effect on the flowering time. Analysis of expression data collected from Genevestigator (Zimmermann et al., 2004) showed that ATH1 is repressed in the SAM upon floral induction in both Col and Ler accessions (Figure 1e, data not shown). These microarray experiments were done with 30-day-old SD-grown plants in which flowering was induced by a photoperiod shift to LD (Schmid et al., 2005). This indicates that ATH1 downregulation in the SAM is not just age dependent but coincides with floral induction. Taken together, these results indicate that ATH1 might function in the floral transition process.
Altered expression of ATH1 affects flowering time in an FRI-dependent manner
The proposed function of ATH1 in floral transition was tested by constitutively expressing ATH1 cDNA in transgenic C24 Arabidopsis plants in either sense (Pro35S:ATH1) or antisense (Pro35S:asATH1) orientation. Interestingly, plants with reduced ATH1 mRNA levels (Pro35S:asATH1 no. 7) showed a pronounced acceleration of flowering under both LD and SD conditions (Figure 2a,b).
Ectopic overexpression of ATH1 had the opposite effect (Figure 2a,b). Lateness in these plants coincided with the formation of aerial rosettes (Figure 2e). These rosettes were established following floral transition at the first three to four nodes of the primary stem, instead of cauline leaves bearing secondary inflorescences. The aerial rosettes formed up to seven leaves and were similar in structure to the original basal rosette. Later in development the aerial rosettes gave rise to inflorescences with cauline leaves and flowers. Additional rosettes also developed in axils of basal rosette leaves. However, as a result of their extreme lateness, Pro35S:ATH1 plants formed a very compact, dome-shaped rosette of leaves, which made it very difficult to recognize these additional rosettes (Figure 2a).
Increased levels of ATH1 severely delay flowering in the C24 accession. Most remarkably, ectopically expressed ATH1 hardly had an effect on flowering time in the Col-0 and Ler accessions. Long-day-grown Col-0 Pro35S:ATH1 plants flowered 22 ± 2 days after germination, compared with 21 ± 4 days for control plants. Similar results were obtained when using the Ler accession to express the haemagglutinin (HA)-tagged ATH1 gene (Figure 3a). This suggests that ATH1 alone is not sufficient to delay the floral transition and it depends on one or more loci that naturally vary among these accessions.
The formation of an enlarged basal rosette and aerial rosettes similar to the ones observed in Pro35S:ATH1C24 plants has been reported to require specific modulation of FLC expression by the synergistic activity of either a dominant FRI allele or mutations that disrupt the FLC-repression pathway, and an active AERIAL ROSETTE 1 (ART1) locus (Poduska et al., 2003). The ATH1 spatial expression pattern largely overlaps that of FLC (Figure 1a–d; Michaels and Amasino, 2000; Sheldon et al., 2002). Moreover, very similar to ATH1, FLC is repressed in the SAM upon transfer from short to long days (Figure 1e). Therefore, it was tested whether ATH1 might function as a positive regulator of FLC. As expected, C24 Pro35S:ATH1 plants have increased FLC mRNA levels (Figure 4d). Furthermore, ATH1-induced late flowering in this accession can be eliminated by vernalization, a treatment which is known to stably repress FLC expression, irrespective of the presence of FRI or an autonomous pathway mutation (Michaels and Amasino, 1999; Sheldon et al., 1999; Figure 2c,d).
Taken together, the late-flowering effect of ATH1 in a C24 background is probably due to an ATH1-dependent increase of FLC expression. Ler, like C24, contains a weak but functional allele of FLC, whereas Col-0 contains a strong FLC allele. However, neither Ler nor Col-0 respond to ATH1 overexpression. Therefore, ATH1 by itself does not seem sufficient to induce FLC expression to high levels and probably requires the further presence of an active FRI allele.
ATH1 ectopic expression and FRI synergistically affect FLC expression levels
To address this ATH1 was ectopically expressed in backgrounds that differ in FRI and/or FLC allele strengths, and flowering times in LD photoperiods were determined. For this purpose Pro35S:HA-ATH1 lines were generated in Ler:FLCCol (strong Col-0 FLC allele, loss-of-function Ler fri allele) and in Ler:FRISF2 (weak Ler FLC allele, active SF-2 FRI allele), in addition to the Ler Pro35S:HA-ATH1 lines that combine a weak FLC allele with a loss-of-function fri allele, and that are flowering like Ler control plants (Figure 3a). Introduction of Pro35S:HA-ATH1 in a background carrying the strong Col-0 FLC allele did not result in further delay of flowering when compared with the parental line. In contrast, overexpression of ATH1 in combination with the active SF-2 FRI allele dramatically affected floral transition (Figure 3a). Some plants never flowered within the duration of the experiment, and all plants that did flower formed more than 60 leaves before the transition. Importantly, the delay could be suppressed by a 4-week vernalization treatment as in C24 Pro35S:ATH1 plants (Figure 3a).
In Ler Pro35S:HA-ATH1 plants FLC expression was increased about 20-fold compared with control plants (Figure 3b). However, this increase was clearly not sufficient to cause a significant delay in flowering. Similarly, the presence of an active FRI allele in a Ler background caused a 40-fold rise in FLC mRNA without dramatically affecting flowering time (Figure 3b), suggesting that there might be a threshold for FLC to display a substantial effect. In the case of Pro35S:HA-ATH1 Ler:FRISF2 plants this putative threshold is exceeded as a result of a synergistic action of ATH1 and FRI, causing a 350-fold increase in FLC levels accompanied by a severe reduction in the mRNA levels of its direct target, the floral promoter FT (Figure 3c). Expression of FLM and MAF2, the two other members of the FLC clade of MADS-box genes that have been reported to function in floral transition, was hardly, if at all, affected by ATH1 in any of the genetic backgrounds (Figure 3d).
Thus, ATH1 alone is not sufficient to delay the floral transition despite its capacity to induce FLC expression to some extent. ATH1-induced late flowering was only observed in a C24 background and in the Ler:FRISF2 introgression line, both of which contain an active FRI and an FLC allele with an attenuated response to this FRI allele.
As in FRI-containing plants, autonomous pathway mutants are also late flowering because of elevated levels of FLC expression (Michaels and Amasino, 2001). To determine whether ATH1 is also capable of enhancing FLC expression to high levels in such backgrounds, Ler Pro35S:HA-ATH1 plants were crossed reciprocally to fca and fve autonomous pathway mutants. However, despite repeated attempts, we were never able to obtain plants homozygous for either autonomous pathway mutation carrying at least one of the Pro35S:HA-ATH1 transgene alleles. In the F2 offspring of the crosses all plants hemi- or homozygous for the transgene carried either none or only one of the respective mutant autonomous pathway alleles. Conversely, none of the F2 plants homozygous for either one of the autonomous pathway mutations was ever found to carry a transgene allele (data not shown). Moreover, Pro35S:HA-ATH1 fca/FCA and Pro35S:HA-ATH1 fve/FVE plants never gave rise to ATH1-overexpressing homozygous fca or fve mutants in the next generation (data not shown). Taken together, this suggests that a combination of ectopic expression of ATH1 together with fca or fve loss-of-function results in lethality.
In conclusion, our data strongly suggest that ATH1 functions as a floral repressor by activation of FLC, at least partly in an FRI-mediated manner. However, interaction with autonomous pathway components in the control of FLC mRNA levels cannot be ruled out.
Basal levels of FLC are attenuated by ath1 mutations
Most results obtained so far follow from experiments in which ATH1 was ectopically expressed. Thus the possibility remains that the biological function of ATH1 is somehow masked. Therefore two mutants, ath1-1 and ath1-3, were obtained from publicly available collections of Arabidopsis T-DNA insertion lines (Alonso et al., 2003; Li et al., 2003). ATH1 contains the SKY and BELL domains conserved in BELL proteins, in addition to a three-amino-acid loop extension (TALE) homeodomain (Bellaoui et al., 2001; Bürglin, 1997) in the C-terminal half of the protein. The T-DNA insertion in ath1-1 is located in the third exon, just before the part of the gene that encodes these three conserved domains (Figure 4a). The T-DNA insertion in ath1-3 is located in exon 4, immediately after the start of the homeobox. Reverse transcriptase-PCR using ATH1-specific primers showed that full-length ATH1 mRNA is not expressed in either of the mutants (Figure 4a). Therefore, most likely both T-DNA insertions create null mutants. However, preliminary phenotypic analyses indicate that the ath1-1 allele is somewhat stronger than ath1-3 (MP and BR, unpublished data). Therefore, ath1-1 was used in most of our experiments.
Under LD conditions, the flowering time of ath1-1 is indistinguishable from that of wild-type Col-8 (Figure 4b). In contrast, under SD conditions, ath1-1 plants flower with approximately 20% fewer leaves than wild-type controls (Figure 4b) similar to what has been observed before for flc loss-of-function mutants in a Col background (Michaels and Amasino, 2001). When compared with flc-7 plants under these conditions, ath1-1 mutants flower somewhat earlier. Introduction of the flc-7 mutation in the ath1-1 background, however, had no additive effect on flowering time (Figure 4b), suggesting that lack of FLC is the major contributor to the ath1-1 flowering phenotype. Accordingly, ath1-1 mutants showed a 95% reduction in FLC expression and a resulting increase of FT (Figure 4c). Like ATH1 ectopic expression, ath1 loss-of-function had no significant effect on the expression of FRI (Figure 4d).
Mutations that affect expression of both FLC and other members of the FLC clade of MADS-box genes (FLM, MAF2-5) flower significantly earlier than flc single mutants in SD and LD (Deal et al., 2005; He et al., 2004; Martin-Trillo et al., 2006; Oh et al., 2004). In accordance with the observed flowering time of ath1-1 mutants, expression of FLM and MAF2, the two other reported floral repressors in the FLC clade of MADS-box genes, was not affected (MAF2) or was only slightly affected (FLM) by the ath1-1 mutation in both LD and SD (Figure 4c; data not shown). Such a small reduction in FLM mRNA levels might explain the added earliness of ath1-1 plants compared with flc-7 mutants in SD.
In conclusion, our data suggest that ATH1 does not affect the entire FLC clade and functions as a repressor of floral transition almost entirely by affecting FLC expression levels.
Late flowering of both FRI-containing plants and autonomous pathway mutants depends on ATH1
The dependence on ATH1 for basal FLC expression and the synergism between FRI and ATH1 overexpression in delaying floral transition prompted us to test whether ATH1 is necessary for proper FRI function. The ath1-1 mutation caused a significant reduction in flowering time in a Col:FRISF2 introgression line with a corresponding reduction in FLC levels, but these plants were still considerably later than Col-8 plants that carry a non-functional FRI allele (Figure 5a,b).
Thus FRI is able to delay flowering independently of ATH1 activity. This may indicate that FRI controls flowering through both ATH1-dependent and ATH1-independent pathways. Alternatively, ath1 mutants may not represent a total loss in activity due to functional redundancy with other related BELL genes.
As is the case for FRI-containing lines, late flowering of autonomous pathway mutants is the result of increased FLC levels. To test whether ATH1 is also required for the late flowering of autonomous pathway mutants, the flowering time of ath1-1 fca-1 and ath1-1 fve-1 mutants was determined. In both mutant backgrounds, a significant though not complete reduction in flowering time was observed (Figure 5a) and also here a corresponding reduction in FLC transcript levels was found (Figure 5b). Thus, in addition to FRI function, autonomous pathway mutations are also affected by ath1-1 mutations. As a result, these data suggest that ATH1 functions as a general regulator of FLC expression.
In Arabidopsis, the antagonistic action of promotion and repression pathways prevents the switch from vegetative to reproductive growth until the developmental state of the plant and favourable environmental conditions allow floral transition. The different floral promotion pathways induce flowering by activating the expression of the floral pathway integrator genes. This activity is antagonized by floral repressors, of which FLC is the most prominent. High levels of the floral repressors prevent the SAM from responding to the promoting signals. Therefore, as reviewed by Boss et al. (2004), the factors that regulate the expression of these floral repressors eventually enable flowering. These factors can be regarded as regulators of meristem competence, since they control the susceptibility of the SAM to the floral-promoting factors.
Here, we propose that the BELL gene ATH1 is a specific, positive regulator of the floral repressor FLC and functions as a competency modulator of the vegetative SAM.
The observed high levels of ATH1 expression in the vegetative SAM and subsequent downregulation prior to floral transition in both LD and after SD to LD transfer is consistent with such a function. Gradual downregulation of ATH1 expression is accompanied by a similar decrease in FLC expression (Figure 1e) and as a result the inhibition will be progressively released. In this way downregulation of ATH1 allows the meristem to become more susceptible to floral stimuli. The suggestion that ATH1 functions in the regulation of meristem competency is further corroborated by the formation of aerial rosettes in C24 Pro35S:ATH1 plants. Aerial rosettes are also found in Sy-0, a late-flowering accession of Arabidopsis, and in transgenic Arabidopsis plants that ectopically express TERMINAL FLOWER 1 (TFL1;Grbic and Bleecker, 1996; Ratcliffe et al., 1998). In both genotypes this characteristic phenotype is proposed to be the result of a prolonged vegetative phase of axillary meristems. In Arabidopsis these meristems usually undergo an obligatory, but short-lived, vegetative phase before converting to reproductive development (Grbic and Bleecker, 1996). In Sy-0 plants an increased insensitivity to floral stimuli prolongs this period and results in the production of vegetative structures at positions where cauline leaves normally develop (Grbic and Bleecker, 1996; Poduska et al., 2003). This requires specific modulation of FLC expression by the cooperative action of a dominant FRI allele and an active ART1 locus (Poduska et al., 2003). Analogously, in C24 Pro35S:ATH1 plants a synergistic interaction of continuous CaMV 35S promoter-driven ATH1 expression in the axillary meristems and a dominant FRI most probably result in substantially elevated levels of FLC expression in these meristems. As a result, such meristems are less sensitive to floral stimuli and undergo a prolonged vegetative phase.
Recently, two other redundant Arabidopsis BELL genes, PENNYWISE (PNY; also known as BELLRINGER, REPLUMLESS, LARSON and VAAMANA) (Bao et al., 2004; Bhatt et al., 2004; Byrne et al., 2003; Roeder et al., 2003; Smith and Hake, 2003) and POUND-FOOLISH (PNF) (Smith et al., 2004) were also proposed to function as competency regulators of the SAM in the process of floral evocation. Unlike ATH1, PNY and PNF act as positive regulators of meristem competence. Moreover, pny pnf plants appear to respond normally to floral induction signals, since early floral markers such as SOC1 and FRUITFULL are expressed after floral induction. Nevertheless, the pny pnf SAM fails to complete floral evocation, as internode and floral patterning events do not take place (Smith et al., 2004), indicating that these two BELL genes function at a later stage in meristem competence than ATH1. Possibly, BELL genes play a broader role in meristem competence at various stages of development.
Like FRI-containing plants, autonomous pathway mutants are also late flowering because of elevated levels of FLC expression (Michaels and Amasino, 2001). Surprisingly, despite several attempts in independent experiments, we have never been able to retrieve ATH1 overexpressing plants in an fca or fve loss-of-function background, suggesting that such a combination causes embryo lethality. Autonomous pathway mutants have been previously associated with lethality. For example, fy null alleles are embryo lethal (Henderson et al., 2005). In addition, hypomorphic fy alleles and fpa cause weak gametophytic defects which, when combined, cause synergistic lethality (Henderson et al., 2005; Koornneef et al., 1998). These and other findings have led to the insight that the autonomous pathway is a series of semi-redundant sub-pathways which are unlikely to be floral specific (Marquardt et al., 2006). For instance, FCA has also been reported to function in lateral root formation (Macknight et al., 2002), and has recently been identified as an abscisic acid receptor (Razem et al., 2006), whereas FVE seems to have an additional role in cold stress signalling (Kim et al., 2004). In animals, the presumed orthologues of plant BELL proteins, PBC proteins, have been shown to directly bind to class I histone deacetylase (HDAC) proteins and to co-precipitate with components of class I HDAC co-repressor complexes (Saleh et al., 2000), including conserved WD40-repeat proteins that resemble FVE (Guenther et al., 2000; Yoon et al., 2005). We therefore suggest that although ath1 fca and ath1 fve double-mutant analyses did not reveal any crucial developmental defects besides the timing of the floral transition, ectopic ATH1 may interfere with as yet unknown processes that involve autonomous pathway components and render proper embryo development impossible in an already disturbed fca or fve background.
The observation that ath1 mutations suppress FLC-mediated late flowering of both an FRI-expressing line and that of fca-1 and fve-1 autonomous pathway mutants indicates that ATH1 functions as a general regulator of FLC. Except for the autonomous pathway and the FRI pathway that act specifically on FLC, most loci identified as being necessary for proper accumulation of FLC have broader roles in the control of flowering time. The Arabidopsis PAF1 and SWR1/SRCAP complexes, as well as EFS/SDG8, control floral transition by regulating the expression of both FLC and other members of the FLC clade (Choi et al., 2005; Deal et al., 2005; He and Amasino, 2005; He et al., 2004; Kim et al., 2005; March-Diaz et al., 2007; Martin-Trillo et al., 2006; Noh and Amasino, 2003; Oh et al., 2004; Soppe et al., 1999; Zhang and van Nocker, 2002; Zhao et al., 2005; March-Diaz et al., 2006). As a result, corresponding mutants flower considerably earlier than both flc null mutants in SD and fri plants in LD. In contrast, ath1 mutant flowering is indistinguishable from that of fri plants in LD, whereas ath1 flowering in SD is similar to that of flc. Moreover, ath1 mutations result in a significant reduction of FLC mRNA, but no substantial effects on the expression of the FLC-clade members FLM1 and MAF2 are observed. Therefore, ATH1 is unlikely to act in close conjunction with these above-mentioned more general floral regulators. Taken together, we propose that ATH1 acts as a floral repressor by specifically regulating FLC expression levels.
The fact that both FRI and the fca-1 and fve-1 mutations are able to significantly delay flowering independently of ATH1 activity might indicate that ath1 mutants may not represent a total loss in activity due to functional redundancy with related BELL genes. The Arabidopsis genome contains a total of 13 BELL genes, and functional redundancy between paralogous BELL genes has been reported for PNY and PNF in the process of floral evocation and inflorescence development, and for SAWTOOTH 1 and -2 in leaf development (Kumar et al., 2006; Smith et al., 2004). Therefore, we are currently investigating whether ATH1 functions as a regulator of FLC expression together with other BELL genes.
At the moment, it is unclear how ATH1 increases the levels of FLC expression. Interestingly, the ATH1-related BELL protein PNY has also been identified as a transcriptional repressor of the floral homeotic gene AG in floral and inflorescence meristems, in addition to its function in meristem competence. It was proposed that PNY might serve as an initiating factor for the repression of AG by recruiting transcriptional co-repressors to the AG chromatin (Bao et al., 2004). Remarkably, striking parallels exist between the regulation of FLC expression and that of AG. FLC and AG both encode a type II MADS-box transcription factor, and both genes appear to be regulated both epigenetically and by proteins involved in RNA metabolism, e.g. HUA2 (reviewed by Simpson, 2004; Quesada et al., 2005). Moreover, in either case many of the cis-regulatory sequences that are important for transcriptional control are found in an unusually long intron that is characteristic of these two genes (3.5-kb intron 1 of FLC, 3.0-kb intron 2 of AG) (Hong et al., 2003; Sheldon et al., 2002). Knowing that PNY was found to act upon this long second intron of AG in vivo and to directly bind to it in vitro (Bao et al., 2004), it will be interesting to determine whether ATH1 directly binds to FLC chromatin, in particular to sequences of intron 1, and whether additional factors are associated with it.
Plant materials and growth conditions
The wild types used in this study were the Arabidopsis thaliana C24, Columbia-0 (Col-0), Col-8 (N60000), and Landsberg erecta (Ler) strains. The FRISF−2 and FLCCol introgression lines used in this study have been described previously: lines homozygous for FLCCol in Ler (Ler:FLCCol) (Koornneef et al., 1994), FRISF2 in Ler (Ler:FRISF2), and FRISF2 in Col (Col:FRISF2; Lee et al., 1994; Michaels and Amasino, 1999). The effect of ectopic expression of ATH1 in Ler introgression backgrounds was tested by introducing the Pro35S:HA-ATH1 transgene in the respective backgrounds.
The ath1-1 (GK line 114A12) and ath1-3 (SALK_113353) T-DNA insertion lines were obtained from GABI-Kat (Rosso et al., 2003) and the European Arabidopsis Stock Centre (NASC; University of Nottingham, UK) respectively. Both alleles were backcrossed two times into the Col-8 ecotype and lines were selected for single insertions. Allele ath1-1 was used in most of our phenotypic and molecular analyses. The flc-7 allele (SALK_092716) was obtained from the NASC. In SD flc-7 plants flowered like the flc-3 null mutant.
Plants mutant for both ath1-1 and fca-1 were made by crossing Col-8 ath1-1 with Ler fca-1. Ler-looking F2 plants were genotyped for both mutations and the Ler FLC allele. Selected plants were subsequently backcrossed twice to the fca-1 mutant. The flowering time of ath1-1/fca-1 plants was compared with fca-1 single mutants that segregated from the second backcross population. Two independent lines were used.
Plants were grown in long days (16 h of light/8 h of dark) or short days (8 h of light/16 h of dark) under cool white fluorescent lights (150 μmol m−2 sec−1) at 22°C with 70% relative humidity (RH). For vernalization, seeds were imbibed and allowed to germinate on full-strength Murashige and Skoog medium at 4°C in low-light short days for 4 weeks in the absence of sucrose. Flowering time was measured by counting the total leaf number.
For the ATH1 promoter–GUS fusion a 2.6-kb SpeI–NcoI genomic fragment of ATH1 (Quaedvlieg et al., 1995) was isolated, and after filling in the NcoI restriction site with Klenow polymerase, it was inserted into the unique SmaI/XbaI sites of pBI101.1 (Jefferson et al., 1987). This fragment contains approximately 1.3 kb of promoter sequence, the entire, 700 nucleotide (nt) long 5′-untranslated region (5′-UTR), and the 42 most N-terminal amino acids encoding part of ATH1. In this way a translational fusion between the ATH1 promoter and the GUS gene was created. Fifteen representative homozygous lines in the C24 background were chosen for further analysis. Histochemical assays for GUS activity at different stages of development showed similar results for all lines, with only the levels of GUS expression differing. Three lines (nos 1.3, 6.5 and 9.6) with intermediate levels of GUS expression were chosen for detailed analysis. GUS histochemistry was performed as described in Dockx et al. (1995).
For antisense expression, the entire ATH1 cDNA sequence (Quaedvlieg et al., 1995) was fused in antisense orientation to the 35S cauliflower mosaic virus (CaMV) promoter (Odell et al., 1985). For ectopic overexpression of ATH1, a 1573-bp cDNA fragment encoding ATH1 was fused in sense orientation to the 35S CaMV promoter. The resulting chimeric genes (Pro35S:asATH1 and Pro35S:ATH1, respectively) were introduced into Arabidopsis C24 plants. Nineteen independent Pro35S:asATH1 and 10 independent Pro35S:ATH1 lines segregating for a single T-DNA insertion, as determined by kanamycin resistance, were selected and further analysed in the T3 generation. Pro35S:asATH1 lines nos 3 and 7, which have undetectable ATH1 levels, and Pro35S:ATH1 lines nos 2 and 11, which show significantly higher ATH1 expression than wild-type control plants, were selected for further phenotypic analysis. Results are shown for lines nos 7, and 2 and 11 respectively.
For ectopic expression of a HA-tagged version of ATH1 (Pro35S:HA-ATH1) ATH1 cDNA flanked by Gateway linkers (generated within the European Community REGIA project) was inserted in the pAlligator2 vector (Bensmihen et al., 2004) by Gateway recombination following Invitrogen recommendations (http://www.invitrogen.com/). Transgenic Ler seeds were selected using a Leica MZFLIII stereomicroscope (http://www.leica.com/) equipped with GFP3 (470 nm/525 ± 50 nm) and YFP (510 nm/640 ± 50 nm) filter sets.
Analysis of gene expression
RNeasy minikit columns (Qiagen, http://www.qiagen.com/) were used to isolate RNA for quantitative (Q)-PCR and RT-PCR analysis. Total RNA was treated with DNaseI (RNase-free; Fermentas Gmbh, http://www.fermentas.com/) to remove genomic DNA. Absence of DNA was analysed by performing a PCR reaction (40 cycles, similar to the real-time PCR programme) on the DNaseI-treated RNA using Taq-DNA polymerase. Quantitative-PCR analyses were performed as described before (van der Valk et al., 2004). Primers used for Q-PCR expression analysis were as described in Czechowski et al. (2004) (FLC, FLM, MAF2) and El-Din El-Assal et al. (2001) (FT). Primers used for RT-PCR analysis were as described in Kumaran et al. (2002) (ACT8) and Michaels et al. (2004) (FLC, FRI). For ATH1 RT-PCR the primers ATH1 RT FWD (5′-TCCTCCACTTCATCCTTTGG-3′) and ATH1 RT REV (5′-CGTTGGGTTGAATGTGACTG-3′) were used. For each mRNA under study the exponential range of amplification was determined. The GUS RNase protection assays were done as in Rook et al. (1998).
Genomic DNA was isolated using the quick-prep method (Cheung et al., 1993). Genotypes of ath1-1 and ath1-3 plants were determined using primers to the 35S CaMV promoter (35S-mini, 5′-CTGCAGCAAGACCCTTCCTCTAT-3′) and the left border (LBb1, 5′-GCGTGGACCGCTTGCTGCAACT-3′), respectively, and ATH1 gene-specific primers (ATH1TAG3.3 FWD, 5′-GCTCGGAGATAAGTCTTTGTGCAGCTA-3′, and SALK_113353 LP, 5′-TTTGTAGTTCAAGAGAAAAGCTTGA-3′, respectively). ATH1 wild-type alleles were identified using the primer combination SALK_113353 LP (5′-TTTGTAGTTCAAGAGAAAAGCTTGA-3′) and SALK_113353 RP (5′-GGCGGGTTTCGGATCTACATT-3′).
The flc-7 insertion was detected using the LBb1 primer in combination with SALK_092716 RP (5′-CTTCTGTCCCTTTTTCATGGG-3′). The presence of a wild-type FLC allele was determined using the primer combination SALK_092716 LP (5′-TCCCTTAACTCTAACCAGCCG-3′) and SALK_092716 RP (5′-CTTCTGTCCCTTTTTCATGGG-3′). FRI allelic variation (Col, Ler or SF-2) was scored as described by Johanson et al. (2000). FLC alleles (Col or Ler) were genotyped using the primers designed by Gazzani et al. (2003) to detect the Mutator-like transposon in FLC in Ler. The fca-1 and fve-1 mutations were genotyped using the respective primer combinations fca-1 FWD (5′-AAAAACCTCTTCACAGTCCACA-3′) and fca-1 REV (5′-AGTTAAAACAACACAATAGCAGCTGAA-3′), and fve-1 FWD (5′-GGATGTTGAAACCCAACCAA-3′) and fve-1 REV (5′-GCTGAATGCCACATCTTCAA-3′). The fca-1 mutation creates a Tru9I restriction polymorphism, whereas the fve-1 mutation creates an HpyF10 VI polymorphism.
The authors thank the Nottingham Arabidopsis Stock Center and the Arabidopsis Biological Resource Center for providing T-DNA insertion lines. FRISF−2 and FLCCol introgression lines and flc-3 seeds were a kind gift of Richard Amasino. The ath1-1 T-DNA mutant was generated in the context of the GABI-Kat programme and provided by Bernd Weisshaar (MPI for Plant Breeding Research, Cologne, Germany). We thank Evelien van Eck-Stouten and Dongping Bao for technical assistance. We are grateful to The Netherlands Organisation for Scientific Research (STW-NWO) for funding this research.