The SOC1 MADS-box gene integrates vernalization and gibberellin signals for flowering in Arabidopsis

Authors


For correspondence (fax +82 2 872 1993; e-mail ilhalee@plaza.snu.ac.kr).
These authors made equal contributions to this study.

Summary

The floral transition in Arabidopsis is regulated by at least four flowering pathways: the long-day, autonomous, vernalization, and gibberellin (GA)-dependent pathways. Previously, we reported that the MADS-box transcription factor SUPPRESSOR OF OVEREXPRESSION OF CO 1 (SOC1) integrates the long-day and vernalization/autonomous pathways. Here, we present evidences that SOC1 also integrates signaling from the GA-dependent pathway, a major flowering pathway under non-inductive short days. Under short days, the flowering time of GA-biosynthetic and -signaling mutants was well correlated with the level of SOC1 expression; overexpression of SOC1 rescued the non-flowering phenotype of ga1-3, and the soc1 null mutant showed reduced sensitivity to GA for flowering. In addition, we show that vernalization-induced repression of FLOWERING LOCUS C (FLC), an upstream negative regulator of SOC1, is not sufficient to activate SOC1; positive factors are also required. Under short days, the GA pathway provides a positive factor for SOC1 activation. In contrast to SOC1, the GA pathway does not regulate expression of other flowering integrators FLC and FT. Our results explain why the GA pathway has a strong effect on flowering under short days and how vernalization and GA interact at the molecular level.

Introduction

Flowering, a transition from vegetative to reproductive growth, is regulated by both environmental and endogenous cues. Extensive genetic analyses to elucidate the molecular mechanism of flowering in Arabidopsis, a quantitative long-day plant, have revealed at least three flowering pathways as the long-day, autonomous, and vernalization pathways (reviewed in Koornneef et al., 1998a; Levy and Dean, 1998; Mouradov et al., 2002; Simpson and Dean, 2002). Mutations in the genes such as CONSTANS (CO), GIGANTEA (GI), and FT that are involved in the long-day pathway cause late flowering under long days but do not delay flowering under short days compared to the wild type. On the contrary, mutations in genes such as FCA, LUMINIDEPENDENS (LD), FVE, and FPA that are involved in the autonomous pathway cause late flowering under both long days and short days compared to the wild type. In addition, mutants of the genes involved in autonomous pathway show a strong response to vernalization, a prolonged cold treatment that accelerates flowering, suggesting that the vernalization and autonomous pathways merge at some point (Koornneef et al., 1991, 1998b).

The genetic analysis of natural variation of flowering among different geographical Arabidopsis ecotypes revealed two major genes, FRIGIDA (FRI) and FLOWERING LOCUS C (FLC), that confer the winter annual flowering habit (Burn et al., 1993; Clarke and Dean, 1994; Koornneef et al., 1994; Lee et al., 1993, 1994; Michaels and Amasino, 1999a; Napp-Zinn, 1985). FRI-FLC-containing plants show a very late flowering phenotype and a very strong response to vernalization. Four weeks of vernalization treatment offsets the effect of FRI FLC and thus fully suppresses the late-flowering phenotype. FRI encodes a novel protein with two coiled-coil domains, whereas FLC encodes a MADS-box transcription factor (Johanson et al., 2000; Michaels and Amasino, 1999a; Sheldon et al., 1999). FRI causes an increase of FLC expression level and thus acts as a positive regulator of FLC. FLC acts as a strong repressor of floral transition (Michaels and Amasino, 1999a; Sheldon et al., 1999, 2000). On the other hand, vernalization reduces FLC expression, thus causing early flowering (Michaels and Amasino, 1999a; Sheldon et al., 2000). Interestingly, the genes involved in the autonomous pathway act as negative regulators of FLC; if the genes are mutated, FLC expression is increased and the level of FLC is well correlated with the lateness in flowering (Michaels and Amasino, 2001; Sheldon et al., 2000). Therefore, the merging point of the autonomous and vernalization pathways is the FLC gene.

From activation-tagging mutagenesis to screening for FRI FLC suppressor mutants, we isolated an early flowering mutant that overexpresses AGAMOUS-LIKE 20 (AGL20), another MADS-box gene (Lee et al., 2000). Molecular and genetic analyses showed that AGL20 is negatively regulated by FLC and positively regulated by the genes involved in autonomous pathway through FLC. In addition, AGL20 is positively regulated by the long-day pathway. AGL20 was also isolated in a screen for suppressors of 35S::CO and was shown to be a direct target of CO (Onouchi et al., 2000; Samach et al., 2000). Thus, it is a synonym of SUPPRESSOR OFOVEREXPRESSION OFCO 1 (SOC1). Hereafter, we refer to AGL20 as SOC1 because the function of the gene was first reported with the designation of SOC1. Also, we suggest that the activation-tagged mutant agl20-101D be named soc1-101D and the T-DNA-tagged agl20 null mutant be named soc1-2 (Lee et al., 2000). As the expression of SOC1 is regulated by both FLC and CO, two central genes regulating the autonomous/vernalization pathways and the long-day pathway, respectively, it is proposed to act as an integrator of flowering pathways (Araki, 2001; Hepworth et al., 2002; Lee et al., 2000; Mouradov et al., 2002; Simpson and Dean, 2002).

Gibberellin (GA) has been known to induce flowering in many plant species (Bernier, 1988). Flowering of Arabidopsis is also promoted by GA. For example, exogenous treatment of GA accelerates flowering of Arabidopsis particularly under short days (Chandler and Dean, 1994; Langridge, 1957). In addition, the mutations that disrupt either GA biosynthesis or signaling show alterations in flowering time (Jacobsen and Olszewski, 1993; Wilson et al., 1992). The Arabidopsis mutant ga1-3 has a deletion in the gene encoding ent-kaurene synthetase A, which catalyzes the first committed step in GA biosynthesis (Sun and Kamiya, 1994). The mutant ga1-3 fails to flower under short days and shows a slight delay in flowering under long days (Wilson et al., 1992). Thus, GA is absolutely required for flowering under short days in Arabidopsis. The dominant gibberellic acid insensitive-1 (gai-1) mutant flowers extremely late under short days, and the flowering phenotype is not rescued by the exogenous treatment of GA (Wilson et al., 1992). On the other hand, the mutant spindly (spy), which causes constitutively active GA signaling, flowers early under both long days and short days (Jacobsen and Olszewski, 1993). Therefore, a GA-dependent flowering pathway has been proposed.

Genetic interactions between the GA pathway and other flowering pathways have been studied. The GA pathway was proposed to act independently of the long-day pathway because the flowering defect of ga1-3 is relatively minor under long days but a double mutant of ga1-3 with co often fails to flower under long days (Putterill et al., 1995; Reeves and Coupland, 2001). In addition, the comparison of double and triple mutants using co, ga1, and fca (a mutation in an autonomous pathway) showed that the GA pathway has the strongest effect on flowering under short days (Reeves and Coupland, 2001). It was proposed that GA may have a role in vernalization because a non-flowering phenotype of ga1-3 under short days could not be overcome by vernalization (Wilson et al., 1992). However, when a ga1-3 mutation was introduced into vernalization-sensitive late-flowering backgrounds, such as ga1-3 FRI FLC and ga1-3 fca-1, it showed a complete vernalization response under long days, suggesting that GA is not required for a vernalization response under long days (Chandler et al., 2000; Michaels and Amasino, 1999b). So far, it is not known why the GA pathway for flowering acts primarily under short days and how vernalization and the GA pathways interact at the molecular level.

Recently, models for the integration of the genetic networks for flowering have been proposed (reviewed in Araki, 2001; Muradov et al., 2002; Simpson and Dean, 2002). In addition to SOC1, FT and LEAFY (LFY) act as flowering pathway integrators. FT is an immediate target of CO, and the expression of FT is negatively regulated by FLC, suggesting that FT integrates the long-day and autonomous pathways (Samach et al., 2000). LFY is a gene regulating floral meristem identity, which, when overexpressed, causes premature flowering (Blázquez and Weigel, 2000; Blázquez et al., 1998; Weigel and Nilsson, 1995). LFY expression is decreased by mutations in genes of both the long-day and the autonomous pathways (Nilsson et al., 1998). LFY is regulated by the long-day and GA pathways through separate cis elements on the LFY promoter, suggesting that multiple flowering pathways are integrated at the LFY promoter (Blázquez and Weigel, 2000).

We further characterized the integrative role of SOC1, in particular, the integration of the GA-dependent flowering pathway. Our results showed that SOC1 integrates the GA pathway, and that such integration is necessary for flowering under short days. In contrast, the expression of other flowering integrators, FLC and FT, is not regulated by GA. Furthermore, we show that the repression of FLC is not sufficient to activate SOC1, and that positive factors are also required. Under long days, CO acts as a major positive factor, whereas under short days, the GA pathway provides a positive factor. Our results explain why the ga1-3 mutant is insensitive to vernalization for flowering under short days.

Results

GA positively regulates SOC1 expression under short days

To address the role of SOC1 in the GA-dependent flowering pathway, the SOC1 expression level was analyzed in the wild type Landsberg erecta (Ler) and GA-biosynthetic or -signaling mutants grown under short days with or without exogenous GA treatment (Figures 1 and 2). The Ler wild type showed acceleration of flowering by GA treatment, and this phenotype was correlated with an increase in SOC1 expression (Figure 2). In the ga1-3 mutant, which fails to flower under short days, SOC1 expression remained at a basal level. However, exogenous GA treatment led ga1-3 to produce flowers, and this rescue of the non-flowering phenotype was accompanied by an increase in SOC1 expression (Figure 2). The GA-insensitive mutant, gai-1, showed extremely late flowering under short days. In contrast to ga1-3, this mutant phenotype could not be rescued by exogenous GA treatment and the expression of SOC1 remained at a basal level irrespective of GA treatment (Figure 2). The mutant, spy-5, showing a constitutive GA response, flowered earlier than the wild type. Exogenous GA treatment slightly enhanced the early-flowering phenotype of this mutant. In correlation with the early-flowering phenotype, SOC1 was highly expressed even without GA treatment, and the expression level was similar to that of the GA-treated wild type (Figure 2). This correlation between the flowering time and the SOC1 expression level in GA mutants strongly suggests that SOC1 is a target of the GA signals for flowering.

Figure 1.

Flowering phenotype of lines described in this study.

All plants were grown under short days (8-h light/16-h dark). The plants (e) to (h) were treated with 100 µm GA3 once a week, and all other plants were not treated with GA3. Scale bars are 1 cm.

(a) and (e) Ler wild type.

(b) and (f) ga1-3.

(c) and (g) gai-1.

(d) and (h) spy-5.

(i) soc1-101D.

(j) ga1-3 soc1-101D.

(k) ga1-3 35S::CO.

(l) ga1-3 35S::FT.

Figure 2.

Effect of GA on SOC1 expression and flowering time under short days.

(a) Flowering time of Ler, ga1-3, gai-1, and spy-5 plants grown under short days with (white bar) or without (black bar) 100 µm GA3 treatment. Flowering time was measured as the number of total leaves produced before flowering. Values represent the mean ± SD of at least 12 plants. (*) indicates that plants did not flower during our experimental period.

(b) RNA gel blot analysis of SOC1 transcript in Ler, ga1-3, gai-1, and spy-5 plants grown with (+) or without (–) 100 µm GA3 treatment. Total RNA was isolated from 6-week-old plants grown under short days. A TUBULIN 2 (TUB2) probe was used as a loading control.

Expressions of FLC and FT are not affected by GA

As SOC1 functions together with other genes in the flowering pathways, we compared the temporal expression pattern of SOC1 with those of two other flowering time genes, FLC and FT, in short-day-grown ga1-3 mutants with and without GA treatment (Figure 3). SOC1 expression remained at a basal level throughout the development of ga1-3 grown under short days if GA was not treated. However, GA-treated ga1-3 mutants showed a significantly increased SOC1 transcript level after 6 weeks. Consistently GA-treated ga1-3 started to express AP1 after 6 weeks, which indicates that flowering had occurred (Hempel et al., 1997, 1998; Kardailsky et al., 1999). In contrast to SOC1, the expression patterns of FLC and FT were not changed by GA treatment. FLC showed uniform expression, and FT showed a gradual increase of expression during the time course irrespective of GA treatment (Figure 3).

Figure 3.

Effect of GA on expression of SOC1, FLC, and FT in ga1-3 mutant.

Semiquantitative RT-PCR analysis was carried out for each gene. RNA was isolated from the plants harvested every week up to 7 weeks. Plants were harvested at 6 h after dawn all times. Numbers indicate weeks after germination. AP1 was used as a marker for floral transition, and TUB2 was used as a quantitative control.

(a) Expression of SOC1, FLC, and FT in ga1-3 grown under short days. AP1 transcript was not detected until 7 weeks, indicating that floral transition did not occur.

(b) Expression of SOC1, FLC, and FT in ga1-3 grown under short days with 100 µm GA3 treatment. AP1 transcript was detected from 6 weeks, indicating that floral transition occurred after 6 weeks.

Overexpression of SOC1 rescues the block to flowering in ga1-3

The non-flowering phenotype of ga1-3 under short days was correlated with the lack of increase in SOC1 expression (Figures 2 and 3). If the minimal SOC1 expression level is the main cause of the block to flowering in ga1-3, overexpression of SOC1 would be expected to induce flowering irrespective of the endogenous GA level. To test this hypothesis, we treated soc1-101D, a mutant constitutively overexpressing SOC1, with GA and paclobutrazol (PAC), an inhibitor of GA biosynthesis, and checked their flowering times (Figure 4). The wild type Ler showed an acceleration of flowering and a concomitant increase in SOC1 levels in response to exogenous GA treatments. When treated with PAC, Ler failed to produce flowers that phenocopied the ga1-3 mutant (Figure 4a). SOC1 expression was also decreased to a basal level in these plants (Figure 4b). In contrast to the counteracting effects of GA and PAC on Ler, soc1-101D did not show significant changes in flowering time in response to GA and PAC (Figure 4a). Such insensitivity of soc1-101D supports our hypothesis that saturation of SOC1 expression can sufficiently overcome the effects of changes in endogenous GA signaling.

Figure 4.

Effects of GA and PAC on flowering time and SOC1 expression in the wild type and soc1-101D grown under short days.

(a) Effect of GA and PAC on flowering time. Non-treated control (black bar), GA treatment (gray bar), and PAC treatment (white bar). Flowering time was measured as the number of total leaves produced before flowering. Values represent the mean ± SD of at least 20 plants.

(b) RNA gel blot analysis of SOC1 transcript in Ler grown with (+GA3) or without (–GA3) 100 µm GA3 treatment or with 37 mg l−1 concentrated PAC treatment (+PAC).

To further confirm this result, we introduced soc1-101D into the ga1-3 mutant background by a genetic cross. The ga1-3 soc1-101D double mutant successfully produced flowers under short days, although the double mutant produced more leaves than the soc1-101D single mutant (Figure 1i,j and Table 1). Therefore, this result suggests that the failure of ga1-3 to flower under short days is at least in part caused by reduced SOC1 activity. We compared the flowering phenotype of ga1-3 soc1-101D with ga1-3 35S::CO and ga1-3 35S::FT. In a soc1-101D background, ga1-3 still had a relatively strong effect in delaying flowering, whereas in 35S::CO and 35S::FT backgrounds, ga1-3 had little effect on flowering under short days (Table 1 and Figure 1j,k,l). As ga1-3 has little effect on flowering under long days, such suppression of the defect in ga1-3 by the overexpression of long-day-pathway genes was expected. On the contrary, the flowering phenotype of ga1-3 soc1-101D suggests that GA regulates additional factor(s) as well as SOC1 under short days.

Table 1.  Flowering times of transgenic and mutant plants grown under short days
GenotypeNumber of total leavesan
  • a

    Flowering time was measured as total leaves produced before flowering.

Ler43.2 ± 2.814
soc1-101D4.3 ± 0.512
35S::CO2.9 ± 0.712
35S::FT3.0 ± 0.812
ga1-3 soc1-101D13.9 ± 0.812
ga1-3 35S::CO4.8 ± 0.812
ga1-3 35S::FT4.6 ± 0.512

Together, our results suggest that GA signaling for flowering is targeted to SOC1 and sufficient levels of SOC1 are able to suppress the defects of ga1-3 mutants in flowering under short days.

The soc1 null mutant is less sensitive to GA

If SOC1 mediates the GA pathway for flowering, it is expected that the soc1 null mutant is insensitive to GA for a flowering response. To test this possibility, we checked the sensitivity of the soc1-2 null mutant to various concentrations of GA (Figure 5). The mutant soc1-2 showed an acceleration of flowering with increasing amounts of GA. However, it showed a weak response to GA, compared to the wild type. The half-maximal concentration of GA for flowering was 4 nm for the wild type and 30 nm for soc1-2. Such a partial sensitivity to GA in a soc1-2 null supports the hypothesis that SOC1 integrates the GA pathway for flowering. In addition, the result suggests the presence of additional factors regulated by a GA-dependent flowering pathway. Such an interpretation is also consistent with the fact that soc1-1 mutation has a little effect on flowering under short days compared to the ga1-3 or gai-1 (Onouchi et al., 2000). Another floral pathway integrator, LFY, is most likely one of the additional factors because the LFY promoter has a cis element mediating the GA pathway for flowering (Blázquez and Weigel, 2000).

Figure 5.

Dose–response curve of GA effect on flowering time of soc1-2.

The wild type Col (solid line) and soc1-2 (dashed line) were sprayed with different concentrations of GA3 once a week under short days. Arrows designate half-maximal response to GA3.

Vernalization activates SOC1 and FT expression irrespective of FLC

Vernalization promotes flowering of late-flowering ecotypes and autonomous pathway mutants by the repression of FLC expression (Michaels and Amasino, 1999a; Sheldon et al., 1999). Subsequently, SOC1, which acts downstream of FLC, is upregulated by this repression (Lee et al., 2000). Recently, the flc null mutant (flc-3) was reported to respond to vernalization under short days, suggesting that vernalization is able to promote flowering via FLC-dependent and FLC-independent mechanisms (Michaels and Amasino, 2001). This result prompted us to check if vernalization can upregulate SOC1 expression in an flc null background. As previously reported, the flc-3 mutant showed an acceleration of flowering by intensive vernalization under short days (Figure 6a). The vernalization effect was saturated by approximately 9 weeks of cold treatment. Consistently, SOC1 expression in flc-3 was increased by prolonged vernalization under short days and the maximal expression of SOC1 was reached after around 9 weeks of vernalization (Figure 6b). This result clearly showed that vernalization increases SOC1 expression even in the absence of FLC.

Figure 6.

Effects of vernalization on flowering time and expression of SOC1 and FT in flc null mutant.

(a) Effect of vernalization on flowering time of flc null. Flowering time was measured as the number of rosette leaves produced before flowering. Values represent the mean ± SD of at least 20 plants. Plants were grown under short days after vernalization treatment.

(b) Semiquantitative RT-PCR analysis of SOC1 and FT in flc null. Tissues were harvested for RNA extraction when plants produced 10 rosette leaves. TUB2 was used as a quantitative control.

As another flowering pathway integrator FT is also repressed by FLC (Samach et al., 2000), we checked if vernalization upregulates FT in the flc null background (Figure 6b). Similar to SOC1, FT also showed an increase in expression in flc-3 by more than 9 weeks of vernalization. Thus, both FLC-dependent and FLC-independent mechanisms of vernalization promote flowering through the regulation of flowering pathway integrators, FT and SOC1. This result may indicate the presence of additional common upstream repressors of FT and SOC1 for which expression is repressed by vernalization.

Interaction of GA and vernalization for the activation of SOC1 under short days

The role of GA in the vernalization response remains controversial because ga1-3 mutants grown under short days are insensitive to vernalization but ga1-3 FRI FLC and ga1-3 fca grown under long days show a strong acceleration of flowering by vernalization (Chandler et al., 2000; Michaels and Amasino, 1999b; Wilson et al., 1992). It is possible that GA mediates the vernalization response only under short days. Alternatively, vernalization may cause only the de-repression of flower-promoting genes like SOC1, but for the activation of such genes, GA may be absolutely required under short days. As our results showed that both vernalization and GA activate the same target gene, SOC1, we addressed this issue by determining SOC1 expression in FRI FLC and ga1-3 FRI FLC grown under different environmental conditions.

FRI FLC and ga1-3 FRI FLC were subjected to 0, 2, 4, and 6 weeks of vernalization and divided into two groups that were grown under long days and short days until plants produced four rosette leaves under long days and 10 rosette leaves under short days, and then the FLC and SOC1 expression levels were determined by RNA gel blot analysis (Figure 7). As previously reported, ga1-3 FRI FLC responded normally to vernalization under long days (Michaels and Amasino, 1999b). The flowering time of these plants was accelerated as the FRI FLC wild types after 6 weeks of vernalization if grown under long days (Figure 7c). In correlation, the expression of FLC was decreased to an undetectable level after 6 weeks of cold treatment, while the SOC1 level reached the maximum (Figure 7a). Therefore, under long days, ga1-3 FRI FLC showed a response to vernalization very similar to that of the FRI FLC wild type, both physiologically as an acceleration of flowering time and molecularly as a decrease in FLC and an increase in SOC1 expression.

Figure 7.

Effect of vernalization on expression of FLC and SOC1 in plants containing ga1-3.

(a) RNA gel blot analysis of FLC and SOC1 transcript in FRI FLC and ga1-3 FRI FLC grown under long days after 0, 2, 4, and 6 weeks of vernalization treatment. Plants were harvested for total RNA extraction when four rosette leaves were produced.

(b) RNA gel blot analysis of FLC and SOC1 transcripts in FRI FLC and ga1-3 FRI FLC grown under short days after 0, 2, 4, and 6 weeks of vernalization treatment. Plants were harvested for total RNA extraction when 10 rosette leaves were produced. TUB2 probe was used as a loading control.

(c) RNA gel blot analysis of SOC1 transcript in Ler and ga1-3 plants grown under short days after 0, 3, 6, and 9 weeks of vernalization treatment. Total RNA was isolated when 10 rosette leaves were produced.

(d) Effect of vernalization on the flowering time of FRI FLC and ga1-3 FRI FLC grown under long (left) and short (right) days. Gray bars represent the plants without vernalization treatment, and white bars represent the plants with vernalization. Flowering time was measured as the number of total leaves produced before flowering. Values represent the mean ± SD of at least 12 plants. (*) indicates that plants were unable to flower.

(e) Flowering time of Ler (black bar) and ga1-3 (white bar) plants grown under short days after 0, 3, 6, and 9 weeks of vernalization treatment.

However, under short-day conditions, ga1-3 FRI FLC lacked their response to vernalization and failed to produce flowers while the FRI FLC wild type showed normal acceleration of flowering by vernalization (Figure 7c). Consistently, the FRI FLC wild type showed normal repression of FLC expression and a concomitant increase in SOC1 expression by vernalization (Figure 7b). In contrast, vernalization of ga1-3 FRI FLC failed to activate SOC1 expression, although vernalization reduced FLC transcript levels sufficiently (Figure 7b). After 6 weeks of vernalization, SOC1 expression remained at a basal level in ga1-3 FRI FLC even though vernalization completely repressed FLC expression. This result clearly shows that vernalization represses FLC expression irrespective of GA under both long days and short days. However, it shows that SOC1 activation is dependent on the presence of GA under short days.

The ga1-3 mutant, which has low FLC expression and fails to respond to vernalization under short days, did not show any increase in SOC1 expression even after 9 weeks of vernalization, although the wild type Ler showed a strong increase in SOC1 expression after 6 weeks of vernalization under short days (Figure 7d,e). Together, our results suggest that the repression of FLC expression is not sufficient for activation of SOC1, but that positive factors are also required for the activation of SOC1 expression. The GA pathway provides such a positive factor under short days as the long-day pathway does under long days. The possibility that GA mediates the vernalization response under short days was excluded because vernalization activates both FT and SOC1 but GA activates only SOC1 under short days (Figure 6). This suggests that the vernalization and GA pathways are independent.

Discussion

Previous studies on flowering time in Arabidopsis have revealed many genes that can be placed in four genetic pathways: the autonomous, vernalization, long-day, and gibberellin pathways (Koornneef et al., 1998a; Simpson et al., 1999). Recently, three genes, FT, SOC1, and LFY, were shown to be regulated by multiple flowering pathways and were termed floral integrators (Araki, 2001; Blázquez and Weigel, 2000; Kardailsky et al., 1999; Kobayashi et al., 1999; Lee et al., 2000; Onouchi et al., 2000; Samach et al., 2000; Simpson and Dean, 2002). Particularly, SOC1 is antagonistically regulated by a B-box zinc finger protein CO, encoded by a gene involved in the long-day pathway, and a MADS-box protein FLC, which itself is an integrator of the autonomous/vernalization pathway (Hepworth et al., 2002; Lee et al., 2000; Samach et al., 2000).

In this study, we focused on the role of SOC1 in relation to the GA-dependent flowering pathway. Our studies showed that the flowering time of GA-biosynthetic and -signaling mutants are well correlated with the SOC1 expression level, and that the introduction of SOC1 overexpression into ga1-3 rescues the non-flowering phenotype under short days, demonstrating that SOC1 also integrates a GA-dependent flowering pathway. In addition, we showed that the repression of FLC by vernalization is not sufficient for SOC1 activation, but the GA pathway is also required under short days.

SOC1 integrates a GA-dependent flowering pathway

Previously, Borner et al. (2000) showed that exogenous GA treatment increased the SOC1 expression level in short-day-grown Arabidopsis plants, which were already at the reproductive stage. It was also shown that GA treatment activated the expression of the SOC1 ortholog SaMADSA in the shoot apex of Sinapsis alba plants grown under non-inductive photoperiods (Bonhomme et al., 2000). However, the correlation between flowering time and the increase in SOC1 expression by GA has not been firmly addressed yet. Here, we intensively studied how the SOC1 expression level correlated with flowering time by using short-day-grown GA-biosynthetic and -signaling mutants with or without GA treatment. The GA-biosynthetic mutant ga1-3 failed to flower under short days, and it was correlated with the minimal expression of SOC1 (Figure 2). On the contrary, exogenous GA treatment caused ga1-3 to flower at a time similar to that for a GA-treated wild type, and caused the SOC1 expression in ga1-3 to increase to a level similar to that in a GA-treated wild type. Therefore, similar levels of SOC1 expression seem to reflect similar flowering times. Such an activation of SOC1 expression by GA is most likely to be mediated by the GA-signaling pathway because the GA-signaling mutants gai and spy also showed this correlation between flowering time and SOC1 expression level. GAI encodes a GRAS (for GAI, RGA, SCARECROW) family regulatory protein with a DELLA domain at the N-terminus (Peng et al., 1997). The function of GAI is to inhibit GA responses in the absence of active GA, and such inhibition is suppressed by GA. The suppression of GAI activity is likely to be mediated through the DELLA domain because the DELLA-domain-deleted mutant gai-1 shows the dominant gain-of-function phenotype, thus showing the GA-signaling defect even with exogenous GA treatment (Peng et al., 1997). The flowering of gai-1 is extremely delayed irrespective of GA treatment, and the plant shows a minimal level of SOC1 expression, showing the correlation between flowering time and a low SOC1 expression level as a result of the GA-signaling defect (Figure 2). The SPY gene, encoding a tetratricopeptide repeat protein, is another negative regulator of GA response, and SPY has been proposed to act upstream of GAI (Jacobsen et al., 1996; Silverstone et al., 1998). The recessive mutant, spy-5, shows a constitutive GA response and earlier flowering than the wild type. The earlier flowering phenotype in spy-5 was again correlated with higher expression of SOC1 than that in the wild type (Figure 2). In addition to the correlation between flowering time and SOC1 expression level, the introduction of SOC1 overexpression into a ga1-3 mutant rescued the non-flowering phenotype under short days. Similarly, PAC treatment could not block the flowering of 35S::SOC1 under short days (Blázquez et al., 2002). These results suggest that the failure of flowering by GA deficiency under short days is caused by the lack of SOC1 activation. Therefore, all our results strongly suggest that SOC1 integrates the GA-dependent flowering pathway.

GA pathway activates additional factors

Although the expression level of SOC1 is regulated by GA and is correlated with the flowering time, SOC1 is not the only flowering time determinant regulated by GA. The double mutant ga1-3 soc1-101D, which overexpresses SOC1, flowered later than soc1-101D single mutant (Table 1). Such a delaying effect of ga1-3 on the flowering time of soc1-101D suggests the presence of additional factor(s) regulated by GA. Consistent with this, the soc1 null mutant showed partial sensitivity to various concentrations of applied GA for a flowering response (Figure 5). Compared to the ga1-3 or gai-1, the soc1-1 mutant shows only a slight delay in flowering under short days (Onouchi et al., 2000). It also supports the presence of partially redundant factor(s) regulated by GA. A flower meristem identity gene, LFY, may be one of the additional factors regulated by GA (Blázquez et al., 1998). The expression of LFY is also regulated by GA; LFY expression remains at a minimal level in ga1-3 mutants throughout development under short days, whereas stronger expression is detected in spy-5 mutants than in the wild type. Similar to soc1-101D, 35S::LFY rescued the flowering defect of ga1-3 under short days but ga1-3 35S::LFY flowered later than 35S::LFY, indicating that LFY is also not sufficient to mediate the GA pathway (Blázquez et al., 1998). SOC1 was proposed to act partially upstream of LFY because overexpression of SOC1 in the absence of FRI caused the production of ectopic flowers subtended by cauline leaves, which are usually observed in 35S::LFY plants (Lee et al., 2000). Thus, it is possible that GA-promoted LFY expression is mediated through SOC1. However, the cis element on the LFY promoter, which mediates the GA-signaling pathway for flowering, is known to be bound to a GAMYB-like protein, AtMYB33, which shows an increase in the expression at the shoot apex during floral transition (Blázquez and Weigel, 2000; Gocal et al., 2001). Therefore, it is more likely that GA regulates the two flowering pathway integrators SOC1 and LFY independently. In addition, the presence of additional factor(s), other than SOC1 and LFY, that regulate flowering in response to GA cannot be excluded.

GA pathway is targeted downstream of FLC

The SOC1 promoter contains a MADS-domain protein binding element (CArG) box that mediates the repression by FLC (Hepworth et al., 2002). Although the GA pathway activates the expression of SOC1, GA does not influence the expression of FLC. FLC showed uniform expression throughout development in the ga1-3 mutants grown under short days, and the expression level was not changed by GA treatment (Figure 3). On the contrary, SOC1 expression was gradually increased after GA treatment, suggesting that the GA pathway targets downstream of FLC. Consistent with our result, Sheldon et al. (1999) showed that the flowering of plants overexpressing FLC was accelerated by GA treatment, and that the FLC level in wild-type plants was not changed by the GA application. Similarly, the expression of another flowering pathway integrator, FT, which is also regulated by FLC, is not affected by the GA pathway (Figure 3). Therefore, we propose that the GA pathway is directly integrated via SOC1.

It is noteworthy that there is some specificity in the integration of flowering pathways among pathway integrators. FT integrates the long-day and autonomous/vernalization pathways, but not the GA pathway. SOC1 integrates all pathways: the long-day, autonomous/vernalization, and the GA pathways. Finally, LFY integrates the autonomous/vernalization and the GA pathways but is not an immediate target of CO, a central regulator of the long-day pathway. Such integration specificity may further elaborate the fine tuning of environmental and endogenous cues for flowering.

Interaction of GA, photoperiod, and vernalization

Previously, it was proposed that SOC1 is directly repressed by FLC through a CArG box on the SOC1 promoter and is activated by CO under long days through a separate cis element (Hepworth et al., 2002). In this study, we showed that SOC1 is also positively regulated by the GA pathway mainly under short days. We propose that the GA pathway is the only flower-promoting pathway under short days, and that removal of FLC repression is a prerequisite but is not sufficient for flowering. Six weeks of vernalization reduced the expression of FLC in ga1-3 FRI FLC to an undetectable level, but SOC1 expression remained at a basal level under short days (Figure 7). This result shows that the repression of FLC expression is not sufficient to activate SOC1 expression but positive factors are also required. Our results also show not only that CO acts as a positive factor but also that the GA pathway provides a positive factor for SOC1 activation and concomitant flowering. Under long days, CO and GA redundantly activate SOC1 expression. Although a ga1-3 or co single mutation causes only a slight decrease in SOC1 expression, blocking the GA pathway in a co mutant causes strong reduction in SOC1 expression under long days (data not shown). On the contrary, under short days, only the GA pathway activates SOC1 expression. The mutant ga1-3 fails to activate SOC1 expression under short days even with prolonged vernalization treatment (Figure 7). Our results suggest that the GA pathway is a constitutive flower-promoting pathway regardless of the photoperiod, but the long-day pathway is a physiologically conditional flower-promoting pathway. In contrast, vernalization causes a de-repressed state of flower-promoting genes such as SOC1. Such a model explains how vernalization interacts with the GA pathway at a molecular level. The vernalization and GA pathways eventually affect the same target molecules, but vernalization acts through de-repression while the GA pathway acts through activation of flower-promoting genes.

Our hypothesis explains well the phenotype of ga1-3 FRI FLC plants. Without vernalization, the repression of SOC1 by FLC is too strong to be overcome by CO activation under long days; thus, ga1-3 FRI FLC shows a minimal expression of SOC1 and a failure to flower (Michaels and Amasino, 1999b). However, if vernalized, the repression by FLC is relieved and the positive factor CO activates SOC1 expression under long days. On the other hand, under short days, vernalization of ga1-3 FRI FLC fails to activate SOC1 because the positive factor provided by GA as well as CO is absent in short-day-grown ga1-3 FRI FLC, although FLC expression is reduced completely as under long days by vernalization (Figure 7). Consistent with this hypothesis, overexpression of CO in ga1-3 almost completely rescues the flowering defect of ga1-3 under short days (Table 1). Alternative to this hypothesis, vernalization may have promotive effect on flowering and GA-dependent activation of SOC1 is enhanced by vernalization.

Michaels and Amasino (1999b) proposed that vernalization affects meristem competency for flowering whereas GA affects a flowering signal under short days. Our data provide some of the molecular details of the relationship between the vernalization and GA pathways. It is tempting to propose that the meristem competency is a reduced state of floral repressor genes like FLC, and that GA acts as a flowering signal at least under short days. In addition, our results may have an implication about the nature of qualitative photoperiod response. Although Arabidopsis is a quantitative long-day plant, the mutation in GA1 converts it into qualitative long-day plant as ga1-3 never flowers under short days. It is because the only flowering pathway under short days is blocked in ga1-3. This may be the evolutionary mechanism through which plants acquire a qualitative photoperiod response. If a constitutive pathway, the GA pathway in the case of Arabidopsis, is blocked by mutation, flowering would be absolutely dependent upon inductive photoperiods.

Although the promoter analysis of SOC1 gene revealed the presence of cis elements that mediate activation by CO and repression by FLC (Hepworth et al., 2002), there must be many more cis elements on the promoter that mediate integration of other flowering pathways. More detailed analysis of SOC1 promoter and comparison with the promoters of FT and LFY will lead to further understanding of the molecular mechanism of flowering pathway integration.

Experimental procedures

Plant materials

The GA mutants ga1-3, gai-1, and spy-5 were all in the Arabidopsis thaliana Ler ecotype. The seeds were obtained from the Arabidopsis Biological Resource Center (Ohio State University, Columbus, OH 43210, USA). The mutant soc1-2 is re-named from T-DNA-tagged null allele of agl20, and soc1-101D is re-named from activation-tagged allele of agl20-101D (Lee et al., 2000). The mutant flc-3 in Col background has been previously described by Michaels and Amasino (1999a).

To generate double mutants between ga1-3 and soc1-101D, soc1-101D in Ler was generated by backcrossing the soc1-101D in Col background to Ler three times (Lee et al., 2000). The double mutants of ga1-3 35S::CO were obtained by crossing ga1-3 and 35S::CO in a Ler background. The 35S::CO seeds were kindly provided by Dr George Coupland (Max-Planck-Institut für Züchtungsforschung), and the ga1-3 35S::FT seeds by Dr Miguel Blázquez (CSIC-UPV). The ga1-3 FRI FLC line provided by Dr Rick Amasino (University of Wisconsin-Madison) has been previously described by Michaels and Amasino (1999b).

Growth conditions

To break seed dormancy, seeds were stratified on 0.65% phytoagar containing 1.5% sucrose and half-strength MS (Gibco-BRL, Gaitherburg, MD, USA) plates for 2–3 days at 4°C. Afterwards, plants were transferred and grown at 23°C under long (16-h light/8-h dark) or short (8-h light/16-h dark) photoperiod conditions in cool white fluorescent lights (100 µmol m−2 sec−1). At least 20 plants were used to measure the flowering time of each genotype. The flowering time was measured as a mean of the total leaf number including rosette and cauline leaves.

To germinate ga1-3 mutants, seeds were soaked in 100 µm GA3 (Duchefa, Biochemie, the Netherlands) for 5 days under 4°C dark conditions, and then rinsed thoroughly with water before sowing on MS plates, which were then moved to the chambers. For vernalization treatment, the MS plates were incubated for several weeks at 4°C under short-day conditions. Exogenous application of GA3 was carried out by spraying the plants with 100 µm GA3 every week. Ler and soc1-101D were treated with PAC by watering the plants with 37 mg l−1 concentrated solution twice a week.

RNA analysis

Total RNA was extracted as described before by Puissant and Houdebine (1990). For RNA gel blot analysis, 20 µg of RNA was separated on 1% denaturing formaldehyde agarose gels and transferred to NYTRAN-PLUS membranes (Schleicher and Schuell, Keene, NH, USA). The SOC1 and FLC probe were cDNA fragments lacking MADS-domain sequences. Blots were probed with TUBULIN 2 (TUB2)-coding regions as a control for the quantity of RNA loaded.

The RT-PCR procedure and primers used for SOC1, FLC, AP1, and TUB2 were described previously by Lee et al. (2000). For FT, 5′-ATG TCT ATA AAT ATA AGA GAC C-3′ and 5′-CTA AAG TCT TCT TCC TCC GCA G-3′ were used as primers.

Acknowledgements

We dedicate this publication to Dr Young Myung Kwon who retired from SNU after three decades of devotion to the plant biology field. We thank ABRC for providing ga1-3, gai-1, and spy-5 seeds; G. Coupland for 35S::CO seeds; M. Blázquez for ga1-3 35S::FT seeds; R. Amasino for ga1-3 FRI FLC seeds and critical reading of the manuscript; and anonymous reviewers for their useful comments. This study was supported by a grant (code: PF003201-01) from the Plant Diversity Research Center of the 21st Century Frontier Research Program funded by the Ministry of Science and Technology of the Korean government to I.L. and by a Korea Research Foundation Grant (KRF-2000–015-DP0398) to C.B.H., N.C.P. and I.L. J.M., S.-S.S., H.L., and K.-R.C. were supported by the Brain Korea 21 program.

Ancillary