The floral transition is the switch from vegetative development to flowering. Proper timing of the floral transition is regulated by different pathways and is critical for the reproductive success of plants. Some of the flowering pathways are controlled by environmental signals such as photoperiod and vernalization, others by autonomous signals such as the developmental state of the plant and hormones, particularly gibberellin. SUPPRESSOR OF OVEREXPRESSION OF CO 1 (SOC1) acts in Arabidopsis as an integrative centre of these pathways, promoting the floral transition. In this work, we show that AGAMOUS-LIKE 42 (AGL42), AGAMOUS-LIKE 71 (AGL71) and AGAMOUS-LIKE 72 (AGL72), which encode MADS-box transcription factors phylogenetically closely related to SOC1, are also involved in the floral transition. They promote flowering at the shoot apical and axillary meristems and seem to act through a gibberellin-dependent pathway. Furthermore SOC1 directly controls the expression of AGL42, AGL71 and AGL72 to balance the expression level of these SOC1-like genes. Our data reveal roles for AGL42, AGL71 and AGL72 in the floral transition, demonstrate their genetic interactions with SOC1 and suggest that their roles differ in the apical and axillary meristems.
During the floral transition in Arabidopsis, the shoot apical meristem (SAM) switches from the production of rosette leaves (vegetative phase) to the production of flowers (reproductive phase) (Hempel and Feldman, 1995; Suh et al., 2003). In Arabidopsis after the primary SAM undergoes the transition to flowering, axillary meristems (AMs) formed in the axils of cauline leaves (Hempel and Feldman, 1995) pass through a short vegetative phase before converting to reproductive development (Grbić and Bleecker, 1996). The floral transition is an irreversible process controlled by robust regulatory networks that integrate signals from different pathways to ensure the reproductive success of plants. Genetic and molecular analysis in Arabidopsis have revealed a set of flowering time genes that have been placed in six pathways: environmental response pathways regulated by photoperiod, vernalization and ambient temperature as well as the autonomous, age and gibberellin (GA)-dependent pathways that act independently of environmental cues (reviewed by Simpson et al., 1999 and Fornara et al., 2010). Genes that integrate the floral transition pathways have been named floral integrators (reviewed by Parcy, 2005). FLOWERING LOCUS T (FT) is such a floral integrator, as it integrates signals from the autonomous and vernalization pathways through its repression by the MADS-box factor FLOWERING LOCUS C (FLC) and its activation by the photoperiod pathway through the B-box transcriptional regulator CONSTANS (CO) (Samach et al., 2000; Wigge et al., 2005; Helliwell et al., 2006; Searle et al., 2006). Another floral integrator is the MADS-box gene SUPPRESSOR OF OVEREXPRESSION OF CO 1 (SOC1). Both FT and SOC1 are up-regulated in the flc mutant (Moon et al., 2003). FLC has been shown to bind to regulatory sequences of SOC1, and this binding is necessary for its repression (Hepworth et al., 2002). SOC1 is also activated by FT at the SAM (Hepworth et al., 2002; Searle et al., 2006).
Moon et al. (2003) showed that the GA pathway promotes SOC1 expression. In the GA-biosynthetic ga1-3 mutant, which fails to flower in short-day (SD) conditions, SOC1 expression levels remained basal whereas they increase upon GA treatment enabling the plant to flower. It has also been shown that SOC1 promotes the expression of LEAFY (LFY) (Lee et al., 2008). LFY is also considered to be a floral integrator since it integrates the photoperiod and the GA pathways through separate cis elements in its promoter (Blázquez and Weigel, 2000). After the floral transition LFY is the first meristem identity gene to be expressed and it promotes floral meristem development (Weigel et al., 1992; Ratcliffe et al., 1999). The other two genes involved in floral meristem identity determination, which play roles in the floral transition, are the MADS-box genes SHORT VEGETATIVE PHASE (SVP) and AGAMOUS-LIKE 24 (AGL24) (Hartmann et al., 2000; Yu et al., 2002; Gregis et al., 2008). Genetic and molecular evidence revealed that SVP binds to the same promoter regions of SOC1 and FT as FLC (Li et al., 2008; Gregis et al., 2009). It has recently been shown that the autonomous and GA pathways negatively regulate SVP. The expression of SVP was consistently higher in the ga1-3 mutant background and reduced in wild-type plants upon GA treatment (Li et al., 2008). On the other hand, AGL24 is positively controlled by the vernalization and photoperiod pathways (Yu et al., 2002; Michaels et al., 2003). Furthermore, it has also been shown that SOC1 is a direct target of AGL24 (Liu et al., 2008).
Here we report the functional analysis of three Arabidopsis MADS-box genes, named AGL42, AGL71 and AGL72, that are closely related to SOC1 (Parenicováet al., 2003), but were not previously implicated in the floral transition. We examine their roles during the floral transition in the SAM where they seem to act redundantly in the GA pathway, although the soc1 mutation is clearly epistatic to agl42, agl71 and agl72. We also uncovered a new role for the SOC1-like genes in the floral transition occurring in AMs. Moreover, we show that these three SOC1-like genes are directly regulated by SOC1.
Expression analysis of AGL42, AGL71 and AGL72
SOC1 is a key regulator of the floral transition since it is an integrator of the vernalization, photoperiod and GA pathways (Borner et al., 2000; Lee et al., 2000; Samach et al., 2000). Detailed phylogenetic analysis of the MADS-box transcription factor family indicates that SOC1 is closely related to AGL42, AGL71 and AGL72 (Parenicováet al., 2003). According to the publicly available microarray data AGL42, AGL71 and AGL72 are all expressed in roots, rosette and cauline leaves and inflorescences. However, expression of AGL71 and especially AGL72 is significantly lower than that of SOC1 and AGL42 (Schmid et al., 2005). We performed detailed in situ analysis of the Arabidopsis SAM to clarify the expression profiles during the floral transition (Figure 1). This analysis revealed that AGL42, AGL71 and AGL72 are all expressed in the SAM during the vegetative phase (Figure 1a,f,k) and during the floral transition (Figure 1b,g,l).
As mentioned above, SOC1 serves as an integrator of signals coming from different flowering pathways, including those from the photoperiod pathway (Borner et al., 2000; Lee et al., 2000; Samach et al., 2000). To evaluate if the SOC1-like genes have a role in the photoperiod pathway, we examined the effect of day length on their expression. We harvested plant material according to a collecting-sample model at three different time-points (Wigge et al., 2005; Schmid et al., 2003; Figure S1 in Supporting Information) to evaluate the expression of these genes upon long-day (LD) induction. In situ hybridization analysis using these samples showed that before the LD induction the expression of AGL42, AGL71 and AGL72 is localized in leaf primordia and in the SAM (Figure 1a,f,k). After the LD induction (T =3) expression was observed in the inflorescence meristem (IM) (Figure 1b,g,l). When the floral transition had occurred (T =5) AGL42, AGL71 and AGL72 expression was observed in the AM, IM and floral primordia (Figure 1c,h,m). While SOC1 expression is up-regulated after the LD induction (Figure 1p–r), the expression of AGL42, AGL71 and AGL72 does not seem to be changed. This result suggests that expression of AGL42, AGL71 and AGL72 is independent of day length.
We also analysed the expression profile of SOC1-like genes in reproductive tissues (Figures 1d,i,n and S2). This analysis revealed the expression of these SOC1-like genes in the IM, in floral meristems, in developing organ primordia and in ovules of mature carpels, indicating an overlapping expression profile.
Generation of soc1, agl42, agl71 and agl72 mutant combinations
For our analysis we used the agl42-2 T-DNA insertion mutant (Nawy et al., 2005), the agl71 allele that contains a T-DNA insertion located within the first exon and the agl72 allele that contains a T-DNA insertion in the second intron (Figure 2a). The positions of the T-DNA elements were confirmed by PCR and sequencing analysis. Using in situ analysis we were unable to detect AGL42, AGL71 and AGL72 mRNAs in longitudinal sections of agl42, agl71 and agl72 mutant inflorescence tissues, indicating that these are complete knockout mutants (Figure 1e,j,o).
Analysis of the agl42, agl71 and agl72 single mutants revealed no flowering phenotype when compared with wild-type plants. To investigate whether there is functional redundancy between these genes, we created various mutant combinations. Since AGL71 and AGL72 are closely linked on chromosome 5 (Figure 2a), an artificial micro-RNA (amiRNA) silencing strategy (Ossowski et al., 2008) was used to specifically knock down both genes. The similarity in nucleotide sequence between AGL71 and AGL72 allowed their silencing using a single amiRNA that recognises both of them. The amiRNA was expressed under the control of the CaMV 35S promoter. In order to obtain various mutant combinations we first generated the soc1-2 agl42 double mutant using the well-characterised soc1-2 mutant (Lee et al., 2000) and transformed this double mutant with the ami::agl71-72 construct to obtain soc1 agl42 ami::agl71-72 plants. We analysed the expression of AGL71 and AGL72 in 10 transgenic lines by semi-quantitative RT-PCR analysis (data not shown). Six lines showed aerial rosette leaves and almost complete silencing of AGL71 and AGL72. Four of these lines were crossed with the wild type to generate the ami::agl71-72, agl42 ami::agl71-72 and soc1 ami::agl71-72 combinations. We used one of these four lines (line 3) and its segregants for further detailed phenotypic characterization. Semi-quantitative RT-PCR was also performed in the ami::71-72 line that we obtained from the backcross with the wild type showing that the silencing of both genes was maintained in the next generation (Figure 2b). We measured the level of down-regulation of AGL71 in both apices (SAM enriched fraction) and leaves of the soc1 agl42 ami::agl71-72 plants by means of quantitative real-time PCR (qPCR). Plant material of mutant and wild-type plants was harvested at comparable developmental stages (respectively, 27 days and 15 days after the shift from SD to LD) based on a morphological analysis during the floral transition of the soc1 agl42 ami::agl71-72 mutant and wild-type plants (Figure S3 and Appendix S3). The qPCR analysis showed that AGL71 was reduced to <7% of wild-type expression levels (Figure 2d,f). Unfortunately for AGL72 it was not possible to obtain reliable expression data by quantitative real time PCR even if we tried different PCR programs and primer sets (see Table S1). Therefore, expression of this gene was analysed by semi-quantitative RT-PCR which showed a reduction of AGL72 expression in both apices (SAM enriched fraction) and leaves of the soc1 agl42 ami::agl71-72 plant (Figure 2g–j).
Genetic interaction among SOC1-like genes in the regulation of flowering time
Since all four genes under study show expression in the SAM during the floral transition, we analysed all mutant combinations for changes in flowering time under SD and LD conditions (Figure 3a,b). For each mutant and knockdown combination as well as wild-type control we analysed at least 18 plants, and this experiment was repeated twice. Our analysis confirms that loss of SOC1 activity significantly delays the time to flowering. The flowering time of agl42, agl71, agl72 single mutants and the ami::agl71/72 line was similar to that of wild-type plants. Nevertheless the agl42ami::agl71/72 plants are late flowering, indicating that these genes have a redundant function in controlling flowering time in Arabidopsis. Moreover all mutant combinations that include the soc1 allele flower at the same time as the soc1 single mutant, suggesting that soc1 is epistatic to agl42, agl71 and agl72. As mentioned before, SOC1 is an integrator of the photoperiod, vernalization and GA pathways (Borner et al., 2000; Lee et al., 2000; Samach et al., 2000). Our results suggest that AGL42, AGL71 and AGL72 are not involved in the photoperiod pathway. In fact agl42ami::agl71/72 plants flowered late under both LD and SD conditions, whereas mutants in the photoperiod pathway are late flowering mainly under LD conditions (Abe et al., 2005). Moreover, the expression of AGL42, AGL71 and AGL72 doesn’t seem to be influenced by day length (Figure 1). Concerning the vernalization pathway, we screened the publicly available microarray data using the Genevestigator database (Hruz et al., 2008) to analyse the expression of the SOC1-like genes in vernalization mutants. The microarray data published by Lempe et al. (2005) and those deposited in the NASC arrays database by Edwards et al. (2006), show that while SOC1 is up-regulated in the flc mutant background, AGL42, AGL71 and AGL72 are all expressed at the same level as the controls, suggesting that these genes don’t appear to be influenced by FLC and are therefore unlikely to act in the vernalization pathway.
AGL42, AGL71 and AGL72 are involved in the GA flowering pathway
SOC1 is known to be involved in the GA pathway (Moon et al., 2003). Since AGL42, AGL71 and AGL72 have a role in the control of flowering time, but are probably not involved in the photoperiodic and FLC-dependent pathways we wondered whether these genes also act through the GA pathway. To investigate the putative role of SOC1-like genes in the GA pathway, the expression of AGL42, AGL71 and SOC1 was analysed by qPCR in seedlings of the highly GA-deficient ga1-3 mutant (Figure 3c–e). In this mutant, which fails to flower under SD and shows a slight delay in flowering under LD (Wilson et al., 1992), SOC1, AGL42 and AGL71 expression was lower than in Ler wild-type plants. This suggests that SOC1-like gene expression is modulated by GA.
We evaluated under both SD and LD conditions the effect of GA3 and MOCK control treatments on wild-type plants and on the mutant combinations that we generated (see Experimental procedures). This analysis showed that GA3 treatments consistently reduced the flowering time of wild-type, soc1 (as previously demonstrated by Moon et al., 2003), soc1 agl42, agl42 ami::agl71-72, soc1 ami::agl71-72,ami::agl71-72 and soc1 agl42 ami::agl71-72 plants (Figure 3a,b). In particular, soc1 agl42, soc1 agl42 agl71 and soc1 agl42 ami::agl71-72 treated with GA3 all flowered at a time similar to soc1, while agl42 ami::agl71-72 plants flowered like the wild type upon GA3 treatment both in SD and LD conditions. The GA signalling pathway includes biosynthesis, turnover and signal transduction. The rescue of the late-flowering phenotype of the agl42 ami::agl71-72 plants obtained through applying exogenous GA3, as also previously observed for the ga1-3 mutant (Moon et al., 2003), suggests that AGL42, AGL71 and AGL72 could be redundantly involved in the regulation of GA biosynthesis or turnover but probably not in the GA signal transduction pathway. Therefore we investigated the expression levels of genes that are involved in the conversion of inactive GA into the bioactive form, such as GA20ox, and those in the reverse process (bioactive to inactive form) such as the GA2ox genes, in wild-type, soc1 agl42 ami::agl71-72 and soc1 plants. As explained above, we harvested the material at the same developmental stage based on our morphological analysis (Figure S3). In particular, soc1 agl42 ami::agl71-72 plants produce the IM and AMs after 27 days in LD, whereas the wild type produce these after 15 days in LD. According to the flowering time analysis (Figure 3a,b) soc1 agl42 ami::agl71-72 and soc1 plants flower at the same time, therefore we collected soc1 SAM-enriched material after 27 days in LD. Quantitative PCR analysis showed that the biosynthetic gene GA20ox1 was down-regulated in both soc1 agl42 ami::agl71-72 and soc1 plants (Figure 3f). Interestingly, the biosynthetic genes GA20ox2 and the gibberellin inactivation gene GA2ox6, are both also down-regulated in soc1 agl42 ami::agl71-72 plants but not in the soc1 single mutant, suggesting a specific role for SOC1-like genes in modulating the expression of genes involved in the biosynthesis of bioactive GA (Figure 3g,h).
These results show that the abundance of the AGL42 and AGL71 mRNAs is dependent on GA levels and that SOC1-like genes are involved in the regulation of GA biosynthetic genes.
SOC1-like genes control the floral transition in the axillary meristem
An accurate phenotypic analysis of the soc1-2 single mutant revealed a mild aerial rosette phenotype under LD and SD conditions (Figure 4b,e). Aerial rosettes develop from the AM as described in detail for the Arabidopsis Sy-0 ecotype by Grbić and Bleecker (1996). We counted the number of leaves in these aerial rosettes until secondary inflorescence bolting. This analysis revealed a gradually increasing number of aerial rosette leaves in the soc1 agl42, soc1 agl42 agl71, soc1 ami::agl71-72 and soc1 agl42 ami::agl71-72 mutant lines compared to the soc1 single mutant (Figure 4b,c,e,f,h). Interestingly, the aerial rosette phenotype appears only in those combinations that include the soc1 mutation, suggesting a central role for SOC1 in AM development. However, SOC1-like genes probably play a redundant role in this process since the soc1 single mutant had little effect on the floral transition in axillary meristems whereas multiple mutant and knock-down combinations of SOC1-like genes showed an increased number of aerial rosette leaves compared with soc1 single mutant plants. Detailed analysis of the single leaves developing at nodes forming the aerial rosettes confirmed that the rosette leaves develop from AMs in the axils of cauline leaves (data not shown). This observation suggests that there is a delay in the floral transition that occurs in the AM. Moreover, these data indicate that SOC1, AGL42, AGL71 and AGL72 are involved in the floral transition that takes place in the AM. Furthermore, when the soc1, soc1 agl42, soc1 ami::agl71-72 (not shown) and soc1 agl42 ami::agl71-72 lines were treated with GA3 (Figure 4d,g), the aerial rosette leaf phenotype completely disappeared, confirming that at least SOC1 and maybe also AGL42, AGL71 and AGL72 are involved in the GA pathway that controls the floral transition in the AM.
SOC1 directly regulates SOC1-like gene expression
According to our data SOC1 seems to be a central player in controlling the floral transition in both the SAM and AM. In the SAM the soc1 mutation was epistatic to the other soc1-like mutant combinations (Figure 3a,b). However, in the AMs, mutations in the SOC1-like genes enhanced the aerial rosette phenotype of soc1, indicating an additive role for these genes in the regulation of the floral transition in AMs (Figure 4). Since SOC1-like genes are involved in the same processes, we wondered whether SOC1 is able to control the expression of the others to balance their activities. Based on the morphological analysis (Figure S3) we used samples collected at the same developmental stage and containing both SAMs and AMs, which are the structures in which SOC1-like genes are functional. We collected the aerial part of wild-type and soc1 plants and then we separated the leaves from the rest, obtaining two samples: one containing only leaves (differentiated tissue) and another enriched in SAMs and AMs (meristematic tissue).
We measured the fold change in expression of AGL42, AGL71 and AGL72 to uncover the differential regulation among meristematic and differentiated enriched tissues (Figure 2c–j). AGL42 was up-regulated in both tissues of soc1 (Figure 2c,e), whereas AGL71 and AGL72 were down-regulated (Figure 2d,f–j). These data suggest a role for SOC1 in the regulation of SOC1-like genes.
To understand whether this regulation is direct, chromatin immunoprecipitation (ChIP) assays were performed. The genomic regions of AGL42, AGL71 and AGL72, comprising 3000 bp upstream of the ATG, all exons and introns and 1000 bp downstream of the stop codon, were analysed by bioinformatics tools to identify CArG boxes [CC(A/T)6GG], which are putative MADS-box protein-binding sites (Riechmann and Meyerowitz, 1997) (Figure 5a). We mainly focused on CArG boxes that are located in the region upstream of the ATG and in the largest intron, since these regions have been shown to be important for the regulation of MADS-box gene expression (Sieburth and Meyerowitz, 1997; Sheldon et al., 2002; Kooiker et al., 2005; Schauer et al., 2009). However, for the completeness of this analysis, CArG boxes located downstream of the stop codon were also tested for binding.
Wild-type and soc1 mutant (used as negative control) 21-day-old seedlings grown under SD conditions for 2 weeks and under LD conditions for 1 week, were collected for ChIP experiments using an antibody against a synthetic peptide of SOC1 (Gregis et al., 2009). Enrichment in the selected regions was quantified by qPCR and normalized against ACTIN and against the negative control (Appendix S1). Enrichment of region I on the LFY promoter was used as a positive control in these experiments (Figure 5b; Lee et al., 2008).
Significant enrichment was observed in region I of AGL71 (Figure 5d), in regions V, VII and VIII of AGL72 (Figure 5d), and in regions II and V of AGL42 (Figure 5c). These results indicate that the SOC1 protein is able to bind in vivo to the regulatory regions of all three SOC1-like genes to directly regulate their expression.
Phylogenetic analysis of the SOC1 subfamily of MADS-box genes
Phylogenetic analysis of the complete Arabidopsis MADS-box gene family showed that AGL42, AGL71 and AGL72 form a group of genes that are related to SOC1 (Parenicováet al., 2003). To get more information about how closely related these genes are a phylogenetic analysis of the SOC1 family was done using protein sequences of monocot and eudicot species. The phylogenetic tree falls into two distinct branches, supported by two different methods (Figure 6 and see Experimental Procedures). One branch (1) contains the Arabidopsis SOC1, AGL14 and AGL19 proteins and the other branch (2) contains the AGL42, AGL71 and AGL72 proteins. The monocot proteins fall as a separate subgroup into branch 1 whereas in branch 2 there are no monocot proteins. Our analysis suggests that in the common ancestor of monocot and eudicot species gene duplication occurred that gave rise to branches 1 and 2. Subsequently monocots and eudicots diverged. However, since in the monocots there do not seem to be any members of branch 2, it is most likely that the ancestral gene that in eudicots gave rise to the AGL42/AGL71 and AGL72-like genes got lost in monocots or diverged to such an extent that it is no longer recognised as a SOC1-like gene. In the common ancestor of the eudicots another duplication occurred that gave rise to two distinct groups, the SOC1 and the AGL14/AGL19 lineages. The phylogenetic analysis using the complete MADS-box families of rice and Arabidopsis performed by Arora et al. (2007) confirms this scenario.
The MADS-box gene family is one of the most intensively studied families of transcription factors in plants. Due to a number of genome duplications in the genus Arabidopsis, gene redundancy is often observed (Arabidopsis Genome Initiative, 2000). Good examples of gene redundancy come from studies of MADS-box genes controlling flower development (Favaro et al., 2003; Pinyopich et al., 2003; Ditta et al., 2004). SOC1 has been shown to be a key player in the control of flowering time by integrating flowering pathways (Hepworth et al., 2002; Moon et al., 2003; Yoo et al., 2005), and several genes are closely related to SOC1. These genes include AGL42, AGL71 and AGL72 and in the study described here we have analysed whether these genes also have functions in the control of flowering time and if some functions of SOC1 are hidden due to redundancy with these MADS-box genes. The possibility of functional redundancy between these four genes was further strengthened by our expression analysis, which showed that their expression profiles overlap in the vegetative and reproductive parts of the plant (Figure 1). However, our functional analysis showed that there is no clear redundant relationship between SOC1 and the other genes. Nevertheless, AGL42, AGL71 and AGL72 were shown to be redundantly controlling flowering time in Arabidopsis. Furthermore, we observed interesting interactions between SOC1 and these three related genes.
SOC1-like genes are flowering promoters involved in the GA pathway
Since our expression analysis shows that AGL42, AGL71 and AGL72 have overlapping expression profiles in both the SAM and IM, we analysed the role of these SOC1-like genes during the floral transition, evaluating the flowering time of several mutant combinations. Interestingly only the agl42 ami::agl71-72 plants revealed a late flowering phenotype (Figure 3) indicating that AGL42, AGL71 and AGL72 play a redundant role as flowering promoters. When we combined soc1 with agl42 ami::agl71-72 flowering time was the same as in the soc1 single mutant, suggesting that soc1 is epistatic to agl42, agl71 and agl72.
Evidence that AGL42, AGL71 and AGL72 act in the GA-dependent flowering pathway come from the experiment showing that expression of AGL42 and AGL71 was down-regulated in the GA-biosynthetic ga1-3 mutant that fails to flower under SD conditions (Figure 3). The same behaviour was shown for SOC1 expression (Borner et al., 2000; Lee et al., 2000; Samach et al., 2000). The GA pathway includes biosynthesis, turnover and signal transduction. The flowering time phenotype of the multiple mutant combinations that we describe here is fully rescued by treating plants with GA, suggesting that SOC1-like genes are not involved in the GA signal transduction pathway since the mutants still respond to exogenous GA. It has already been shown that some MADS-box transcription factors are involved in the regulation of GA biosynthesis and turnover. In fact, AGAMOUS (AG) positively regulates a GA3-oxidase (AtGA3ox1) by directly binding to its promoter (Gomez-Mena et al., 2005). Moreover, AtGA2ox6, which is involved in deactivation of GA, has been identified as a direct positive target of AGL15 (Wang et al., 2004). We analysed the expression of GA20ox1 and GA20ox2, which are reported to be the most highly expressed gibberellin biosynthetic genes during vegetative and early reproductive development (Rieu et al., 2008a). We found that both these genes were down-regulated in meristematic enriched tissues of soc1 agl42 ami::agl71-72 plants. GA20ox1 is also down-regulated in the soc1 mutant, whereas expression of GA20ox2 decreased only in soc1 agl42 ami::agl71-72 plants, suggesting a specific role for SOC1-like genes in the control of its expression (Figure 3). Moreover we investigated the expression of GA2ox6, which is the most highly expressed gene in all developmental stages (Rieu et al., 2008b). GA2ox6 is involved in the inactivation of bioactive GA. Again a down-regulation of GA2ox6 was only detectable in soc1 agl42 ami::agl71-72 plants. These results are in agreement with the model of GA homeostasis; in fact the down-regulation of GA biosynthetic genes causes a decrease in the amount of GA that in turn releases the feed-forward regulation on GA-inactivation by GA2ox genes (Rieu et al., 2008b). Taken together, these data support the existence of a positive regulatory loop: GA promotes the expression of AGL42, AGL71 and AGL72, which in turn modulate the expression of GA biosynthetic genes.
In the SAM, SOC1 seems to act upstream of AGL42, AGL71 and AGL72 (Figure 2) since in the soc1 mutant all three genes were deregulated. Furthermore, ChIP analysis showed that the regulation of AGL42, AGL71 and AGL72 by SOC1 is direct (Figure 5). While in the soc1 mutant background AGL71 and AGL72 were down-regulated, AGL42 was up-regulated. Since we clearly showed that SOC1-like genes act redundantly as flowering promoters, their regulation via SOC1 points to a complex regulatory loop, which is probably a mechanism to balance the expression of these three downstream genes. This also suggests that the epistatic relation between soc1 and agl42, agl71 and agl72 in the SAM can probably not be explained by the regulation of these downstream genes by SOC1. It is more probable that SOC1 forms protein complexes with AGL42, AGL71 and AGL72 in which the presence of SOC1 is essential and the others are redundant. This is further supported by the observation that SOC1 directly interacts with AGL42 and AGL71 (de Folter et al., 2005). For AGL72 a direct interaction with SOC1 was not observed even if this protein is very similar to AGL71. This might be a false negative in this large-scale yeast two-hybrid assay but it might also be that the interactions are bridged by other MADS-domain proteins like for instance AGL74 (Figure S4).
The role of SOC1, AGL42, AGL71 and AGL72 in the floral transition of the AM
In Arabidopsis AMs are mainly formed in the axils of leaf primordia when the SAM goes through the floral transition (Figure S3; Hempel and Feldman, 1995). Once they develop they pass through a very short vegetative phase before entering reproductive development (Grbić and Bleecker, 1996). If the vegetative phase in these AMs is prolonged, due to a significant reduction in environmental signals promoting flowering or due to the loss of gene activity promoting flowering, aerial rosettes develop in the axils of cauline leaves. For instance when wild-type Arabidopsis plants are grown for a long period under non-inductive SD conditions a mild form of aerial rosette formation can be observed. In the Arabidopsis Sy-0 ecotype that makes a large number of aerial rosettes, it was shown that the delay in the floral transition at the SAM and the AM was due to the synergistic up-regulation of the floral repressor FLC by FRI and ART1 (Poduska et al., 2003). Also in the soc1 mutant, where the floral transition is significantly delayed, aerial rosettes are formed under LD conditions, indicating that there is also a delay in the floral transition at the AM in this mutant. Interestingly, the number of leaves in the aerial rosettes was significantly enhanced when the soc1 mutant was combined with agl42, agl71 and agl72 mutant alleles (Figure 4), indicating that the loss of the activity of these SOC1-like genes further delays the flowering time in the AMs. Furthermore, the whole developmental program of the soc1 agl42 ami::agl71-72 quadruple mutant plant seemed to be further delayed (Figure S3). This all suggests that the mechanism controlling the floral transition in the SAM seems to be different from that in the AMs, since in the SAM of soc1 mutant plants there was no additive effect of the loss of AGL42, AGL71 and AGL72 activity on the time to flowering and the soc1 mutation was epistatic to the agl42, agl71 and agl72 mutations (Figure 3). Furthermore, in the SAM AGL42, AGL71 and AGL72 seem to act redundantly whereas in the AM there is a clear additive effect. Nevertheless, of the four genes also present in the AM, SOC1 is the key controlling gene since aerial rosette development was only observed when there is no SOC1 activity, whereas in the agl42 ami::agl71-72 triple mutant no aerial rosettes were observed.
The AMs develop when the plant goes through the floral transition. Furthermore, it is likely that the signals that promote the floral transition at the SAM immediately act on the AM so that these, when they are formed, transit instantly to the reproductive phase keeping the vegetative phase very short. Since the floral repressor FLC also seems to act in the AM (Poduska et al., 2003), and since it is an upstream regulator of SOC1, it might well be that SOC1 is also integrating signals from the different flowering pathways in the AMs. In this respect the GA-dependent flowering pathway is certainly important, since when we treated the soc1 agl42 ami::agl71-72 quadruple mutant with GA the aerial rosette phenotype completely disappeared, indicating that in the AMs these genes also play a role in the GA-dependent flowering pathway as explained before with regard to the SAM transition.
The evolution of the SOC1 gene family
The functional analysis of the SOC1-like genes AGL42, AGL71 and AGL72 made clear that, like SOC1, they play a part in the floral transition. Here we mainly focused on their role in the GA pathway. Their function seems not to be redundant with SOC1, although they seem to be directly regulated by SOC1. These SOC1-like genes have probably diversified their function during evolution. A phylogenetic analysis to understand the evolutionary relationship between SOC1 and AGL42, AGL71 and AGL72 showed that the three SOC1-like genes are closely related (Figure 6) and it is therefore not surprising that they are functionally redundant. However, our analysis interestingly showed that there was a gene duplication that gave rise to the SOC1 gene and the AGL42/AGL71/AGL72 genes that occurred before the separation of monocot and eudicot plants. This means that this gene duplication is very ancient and might explain the functional diversity. Interestingly, a function in flowering time is conserved between these genes. Another interesting fact is that we did not find any AGL42/AGL71/AGL72-related gene in monocot species, despite the gene duplication that led to the formation of this group of genes having occurred in the ancestor of monocot and eudicot species. The ancestral gene was probably lost early in monocot plant evolution, although more monocot species will have to be analysed to understand if this was a very early event.
The monocot genes that group into the SOC1 lineage all group together separate from the eudicot genes. However, they seem to be active as flowering time regulators. At least for the rice (Oryza sativa) genes OsMADS40 and OsMADS46 a function in the control of flowering time has been observed (Ryu et al., 2009). Of the genes that belong to branch 1, only VvMADS8 has been characterized. Interestingly the VvMADS8 gene, that is expressed in IM in Vitis vinifera, cause early flowering when constitutively expressed in Arabidopsis suggesting a role as flowering promoter (Sreekantan and Thornas, 2006).
In conclusion, our data show that the SOC1-like genes are involved in controlling the floral transition in the SAM and AM and that they seem to act through the GA-dependent flowering pathway. Our data also suggest that these genes, in contrast to SOC1, do not seem to play a role in the photoperiod- and FLC-dependent pathways. However, we cannot exclude that they might play roles in other flowering pathways, like the endogenous or ambient temperature-sensitive pathways. This remains the domain of future studies.
Plant materials and growth conditions
The plants were grown al 22°C under SD (8-h light/16-h dark) or LD (16-h light/8-h dark). The soc1-2, agl42, agl71 and agl72 mutant plants are all of the same Columbia background. soc1-2 and agl42-1 are T-DNA insertional mutants that were described previously (Lee et al., 2000; Nawy et al., 2005). Seeds from the agl71 and alg72 mutants in Columbia were obtained from the Nottingham Arabidopsis Stock Centre (http://nasc.life.nott.ac.uk/). The agl71 and agl72 alleles contain a T-DNA insertion (GABI-Kat line 249D03 and GABI-Kat line 799A05, respectively). Genotyping primers are listed in the Table S1. The Arabidopsis Genome Initiative number for AGL71 is At5G51870.1 and AGL72 is At5G51860.1. The GA mutant ga1-3 was in Arabidopsis thaliana Ler ecotype.
Treatment of plants with GA was started with seedlings at 1 week after germination, and weekly application of 100 μm GA3 was performed as published (Moon et al., 2003).
Flowering time was determined by counting the total number of rosette leaves.
In situ hybridization analysis
Arabidopsis seedlings and flowers were fixed and embedded in paraffin as described previously (Huijser et al., 1992). Sections of plant tissue were probed with digoxigenin-labelled SOC1, AGL42, AGL71 and AGL72 antisense RNA. Sequences of primers used for in situ hybridization are listed in Table S1. Hybridization and immunological detection were performed as described by Coen et al. (1990). Slides were observed using a Zeiss Axiophot D1 microscope equipped with differential interference contrast (DIC) optics. Images were captured on an AxiocamMRc5 camera (Zeiss, http://www.zeiss.com/) using the Axiovision program (version 4.1).
RNA extraction and reverse transcription
For RT-PCR, roots, leaves, inflorescences and siliques of the wild type (grown in LD conditions), were collected for RNA extraction and reverse transcription as previously described in Gregis et al. (2008). The samples were treated with DNase (TURBO DNA-free®; Ambion, http://www.ambion.com/) and reverse transcribed according to the ImProm-II® Reverse Transcription System (Promega, http://www.promega.com/) instructions. Sequence primers for RT-PCR amplification are listed in the Table S1.
For semi-quantitative PCR, roots and seedling tissues from the wild type, soc1 and soc1 agl42 amiR-agl71-72 (grown in LD conditions) were collected for RNA extraction and reverse transcription as explain above. Amplification was performed in 20, 25, 30 and 35 cycles. For details see Appendix S2.
Artificial micro RNA silencing technology
The amiRNA sequence against AGL71 and AGL72 5′-ACCGACCTATTTATCGGATTG-3′ was selected among the proposed sequences and designed as previously described in http://wmd.weigelworld.org. The Primers Ve-49 and Ve-50 was designed to introduced the sequences for the TOPO vector (Invitrogen, http://www.invitrogen.com/). The 700 nucleotides obtained by using PCR amplification were introduced in the pB2GW7 binary vector downstream of the CaMV35S promoter. Cloning was done using the Gateway® system (Invitrogen). The insert was extracted from the vector and purified using the Purification Kit (GE Healthcare, http://www.gehealthcare.com/) and cloned in Agrobacterium tumefaciens to subsequently infect the plants according to the floral dip method (Clough and Bent, 1998). Sequences of primers for amiRNA silencing constructs against AGL71-72 are listed in the Table S1. Information about vector pB2GW7 is available at http://www.psb.ugent.be/gateway.
Chromatin immunoprecipitation assays
For ChIP experiments, an antibody against a synthetic peptide (SVKCIRARKTQVFK) was used. Chromatin was prepared from 21-day-old seedlings grown under SD conditions for 2 weeks and under LD conditions for 1 week. soc1 single mutant plants were used as negative controls. Positive binding site fragments were considered only if they were significantly enriched compared with the controls in at least three independent experiments.
The complete primer sets are shown in Table S1. The ChIP assays were performed as described in detail in Appendix S1.
Quantitative real-time RT-PCR
Enrichment folds were detected using a SYBR Green assay (Bio-Rad, http://www.bio-rad.com/). The real-time PCR assay was performed in triplicate using a Bio-Rad C1000 Thermal Cycler optical system. For expression analyses normalized expression (ΔΔC(t)) was calculated using Bio-Rad CFX manager software. For ChIP experiments, relative enrichment was calculated as described in Appendix S1. For the expression analysis ubiquitin was used as reference gene. Relative enrichments were calculated as previously described (Gregis et al., 2008). Primers used for Real-Time PCR can be found in Table S1.
We thank S. Masiero, L. Dreni, F. Fornara and D. Horner (University of Milan) for technical support, P. N. Benfey for providing the agl42-1 mutant and H. Yu for the ga1-3 mutant. The post-doctoral and PhD fellowships for VeG and VaG, respectively, were funded by the Università degli Studi di Milano. CD-F was supported by Marie Curie EU project Transistor (512285).