During the initial stages of flower development, floral meristems increase in size without the formation of floral organs. When a critical meristem size is reached, the floral meristem begins to develop the floral organs. The first stages of flower development are characterized by the expression of genes such as APETALA 1 (AP1), CAULIFLOWER (CAL), AGAMOUS-LIKE 24 (AGL24) and SHORT VEGETATIVE PHASE (SVP). We have shown that AP1, AGL24 and SVP act redundantly to control the identity of the floral meristem and to repress expression of class B, C and E genes. Recently, it was shown that class E gene repression was direct and established by two independent pathways. We show here that repression of class B and C genes is also directly established by a co-repressor complex that comprises LEUNIG (LUG), SEUSS (SEU) and the MADS box dimers AP1–AGL24 and AP1–SVP. Furthermore, we show that the distantly related SUPPRESSOR OF OVEREXPRESSION OF CO 1 (SOC1) MADS box gene can complement for the loss of AGL24 and SVP activity; however, under normal conditions, this transcription factor does not play a role during the early stages of flower development.
In most angiosperm species, flowering is induced in response to environmental conditions, hormones and plant age. During the vegetative phase, the shoot meristem forms a series of leaf meristems on its flanks; after floral induction has occurred, the apical inflorescence meristem produces floral meristems instead of leaf meristems. The formation of flowers is controlled by a complex regulatory network in which MADS box transcription factors such as SHORT VEGETATIVE PHASE (SVP) and AGAMOUS LIKE24 (AGL24) have been shown to play important roles during various phases of reproductive development (Hartmann et al., 2000; Yu et al., 2002; Michaels et al., 2003; Gregis et al., 2008). These genes share substantial sequence homology but have opposite functions during the floral transition. SVP is a repressor of flowering, whereas AGL24 promotes floral transition. Recently, it has been shown that, during vegetative development, SVP acts as a central regulator of the floral transition by directly repressing the floral integrator SUPPRESSOR OF OVEREXPRESSION OF CO 1 (SOC1) in the shoot apex and the leaf (Li et al., 2008). Furthermore, SVP also regulates expression of the floral pathway integrator FLOWERING LOCUS T (FT) (Lee et al., 2007; Li et al., 2008). The repressive function of SVP is mediated by interaction with another potent repressor of flowering, FLOWERING LOCUS C (FLC) (Li et al., 2008). In contrast to the repressive function of SVP, it has been suggested that AGL24 and FT promote SOC1 expression during the floral transition (Michaels et al., 2003; Yoo et al., 2005; Searle et al., 2006; Liu et al., 2008).
We showed recently that, during early stages of flower development, AGL24 and SVP have redundant functions (Gregis et al., 2006, 2008). Combining the agl24-2 svp-41 double mutant with the strong apetala 1-10 (ap1-10) mutant resulted in loss of floral meristem identity, leading to the formation of a cauliflower-like curd. This phenotype is similar to the phenotype observed in the ap1 cauliflower (cal) double mutant, indicating that, like APETALA 1 (AP1) and CAULIFLOWER (CAL), AGL24 and SVP are also floral meristem identity genes (Gregis et al., 2008).
Furthermore, analysis of the agl24-2 svp-41 double mutant and the ap1-12 agl24-2 svp-41 triple mutant (ap1-12 is a weak allele) showed that, at stages 1 and 2 of flower development, AGL24 and SVP play a redundant role in repressing the homeotic class B genes PISTILLATA (PI) and APETALA 3 (AP3) and the class C gene AGAMOUS (AG) (Gregis et al., 2006). Mis-expression of these class B and C genes resulted in homeotic conversions in all four floral whorls. The repressive function of AP1, AGL24 and SVP is probably established by interaction with the LEUNIG (LUG)–SEUSS (SEU) co-repressor complex, as, in yeast assays, the AP1–SVP and AP1–AGL24 dimers are able to interact with the LUG–SEU dimer. By analogy with the Ssn6–Tup1 co-repressor complex in yeast (Conlan et al., 1999), SEU acts probably as an adapter protein between LUG and DNA binding proteins (Sridhar et al., 2004). Furthermore, Liu et al. (2009) showed that AGL24 and SVP repress SEPALLATA 3 (SEP3) via two distinct pathways. SVP interacts with TERMINAL FLOWER 2-LIKE/HETEROCHROMATIN PROTEIN 1 (TFL2/LHP1) to modulate trimethylation of histone H3 lysine 27 (H3K27me3), while AGL24 (like SOC1) interacts with SAP18, a member of the Sin3/histone deacetylase (HDAC) complex, to modulate histone H3 acetylation.
These data fit well with a hypothesis that we recently formulated (Gregis et al., 2008), i.e. for formation and growth of the floral meristem (stages 1 and 2 of flower development), AP1 interacts with SVP or AGL24, and these dimers are important to establish floral meristem identity and at the same time repress genes that control formation of the floral organs. Subsequently, when SVP and AGL24 are repressed at late stage 2, AP1 interacts with SEP proteins (whose expression starts from late stage 2), and development of the floral organs starts.
To unravel the regulatory mechanisms that control formation of the floral meristem, which is a crucial developmental phase for correct formation of flowers, we investigated this model further and present a detailed analysis of the regulation of AG, AP3, PI and SEP3 by AGL24, SVP and AP1 during early stages of flower development. We provide evidence that AP1, AGL24 and SVP directly bind to regulatory regions of the class B and C floral homeotic genes, confirming their fundamental role during floral meristem formation. We also present new genetic data showing that the AP1–SVP and AP1–AGL24 dimers repress their target genes through interaction with the SEU–LUG co-repressor complex. Moreover, in situ expression analysis of SOC1 in the agl24-2 svp-41 double mutant showed that this gene is ectopically expressed at stages 1 and 2 of flower development. This led to the discovery that SOC1 is able to rescue the loss of AGL24 and SVP during early stages of flower development, probably by interacting with AP1 and repressing directly class B and C genes.
In flowers, AGL24 and SVP are restricted to developmental stages 1 and 2
Previously published AGL24 mRNA expression profiles during flower development have been rather variable, probably due to differences in protocols and probes (Hartmann et al., 2000; Yu et al., 2002, 2004; Michaels et al., 2003; Gregis et al., 2008). To analyse the expression of AGL24 and SVP at the protein level, we introduced fluorescently tagged versions of these two MADS box transcription factors into Arabidopsis.
The full genomic regions of AGL24 and SVP, including 3 kb upstream of the start codon, were cloned as C-terminal fusions with mRFP1 (monomeric red fluorescent protein; Campbell et al., 2002; Shaner et al., 2004) and GFP (green fluorescent protein; Chalfie et al., 1994; Chiu et al., 1996), respectively. The AGL24–RFP and SVP–GFP fusion constructs were introduced into the agl24-2 and svp-41 single mutant backgrounds, respectively. Homozygous lines for these constructs were selected in T2 and T3 generations. To investigate whether these fusion proteins are biologically active, we measured the flowering time of these complementation lines, which showed a time to flowering comparable to that of wild-type plants (Figure 1c,d and Table S1). This shows that the fluorescent fusion proteins can complement the null alleles and retain their wild-type function in the svp-41 and agl24-2 mutant backgrounds.
In order to study AGL24 and SVP expression dynamics and protein localization in the reproductive tissues of Arabidopsis, we studied their protein expression profiles by confocal microscopy using the pSVP:SVP-GFP svp-41 and pAGL24:AGL24-RFP agl24-2 lines (Figure 1a,b). This analysis showed that both proteins are present in the nuclei throughout the floral meristem during stages 1 and 2 of flower development. These stages are characterized by the undifferentiated state of the floral meristem. As soon as the sepal primordia developed (stage 3 of flower development), neither AGL24–RFP and SVP–GFP were detectable any longer. Furthermore, in contrast to SVP, AGL24 was also detected in the inflorescence meristem. These expression profiles are in agreement with the in situ mRNA hybridization data obtained by Michaels et al. (2003) and Gregis et al. (2008). All published in situ data indicate the presence of significantly reduced levels of AGL24 mRNA during later stages of development; these reductions were also detected at the protein level (Figure 1a).
Silencing of LUG enhances the agl24-2 svp-41 double mutant phenotype
We previously showed that, at 22°C, the agl24-2 svp-41 double mutant has only a very mild floral phenotype in the first three flowers. When we grew this double mutant at high temperatures (>30°C) or combined the double mutant with the weak Arabidopsis ap1-12 allele, all flowers displayed homeotic conversions in the four floral whorls due to ectopic expression of the class B and C floral organ identity genes (Gregis et al., 2006). These floral phenotypes are very similar to those reported for the lug and seu single and double mutants, in which class B and C genes were also shown to be de-regulated (Liu and Meyerowitz, 1995; Franks et al., 2002). Furthermore, yeast two-, three- and four-hybrid protein interaction studies showed that a dimer comprising AP1 and SVP or AP1 and AGL24 can bind the LUG–SEU co-repressor (Gregis et al., 2006).
To provide genetic evidence that confirm these yeast interaction data and further support the view that SEU and LUG, together with AGL24 and SVP, regulate class B and C expression, a LUG RNAi silencing approach in the agl24-2 svp-41 background was used. A specific portion of the LUG cDNA was cloned in antisense and sense orientations in an RNAi expression cassette, under the control of the CaMV 35S promoter. We transformed svp/+ agl24/+ plants and screened the T1 generation for those plants showing mild lug phenotypes. Subsequently, we analysed these plants by PCR for the presence of the agl24-2 and svp-41 mutant alleles. We chose two lines (Las1 and Las2) that were heterozygous for the agl24-2 and svp-41 mutant alleles. In the T2 generations of Las1 and Las2 lines, we verified the reduction of LUG expression by RT-PCR analysis using RNA extracted from cauline leaves. We used the segregants Las1-3 (svp/+ agl24), Las1-7 (svp agl24), Las2-3 (svp agl24/+) and Las2-14 (svp agl24) for further analysis. This revealed that the LUG expression levels in knockdown plants were significantly reduced compared to wild-type, confirming the efficiency of this RNAi approach for LUG silencing (Figure 2a). 35S:LUG-RNAi plants with at least one wild-type allele for SVP or AGL24 showed weak lug floral phenotypes, with short and narrow sepals, a reduced number of petals and stamens, and unfused carpels (Figure 2b,c). The phenotypes were always less severe than those of 35S:LUG-RNAi agl24-2 svp-41 double mutant plants (Figure 2d–f). These plants (Las1-7 and Las2-14) showed severe floral defects similar to those observed in the agl24-2 svp-41 double mutant at 30°C or the ap1-12 agl24-2 svp-41 triple mutant, as well as to the phenotype observed in strong lug mutants (Figure 2d–h) (Liu and Meyerowitz, 1995; Gregis et al., 2006). Morphological analysis of Las1-7 and Las2-14 plants revealed that they developed flowers with a reduced number of organs in the three outer whorls, and organs developing in the outer whorl showed homeotic transformations. They always displayed carpelloid and stamenoid features, suggesting class B and C gene de-regulation, which is typical for strong lug alleles. In fact, the narrow and elongated tips of first-whorl carpels resemble the horn-like protrusions of lug carpels (Figure 2d,f,h) (Liu and Meyerowitz, 1995). Scanning electron microscopy analysis confirmed the formation of chimeric organs in the first whorl that develop as sepal/carpel mosaics, with stigmatic tissues developing along the margins of the mosaic organs (Figure 2, compare g,h with i,j). The second-whorl organs were often lost, and stamens were strongly reduced in number or completely missing. In almost every flower, the carpels failed to fuse properly (Figure 2f).
The observed phenotypes show that LUG down-regulation in the agl24-2 svp-41 double mutant significantly enhances the floral defects of the double mutant, and confirm that LUG, together with SVP and AGL24, regulates class B and C genes.
AGL24, AP1 and SVP directly and redundantly regulate class B, C and E floral homeotic genes
To investigate whether AP1, AGL24 and SVP directly bind in vivo to MADS box protein-binding sites (CArG boxes) to regulate class B (AP3, PI), C (AG) and E (SEP3) floral organ identity genes, chromatin immunoprecipitation (ChIP) assays were performed using transgenic plants expressing GFP- and RFP-tagged versions of these proteins. Inflorescence tissue of homozygous pSVP:SVP-GFP svp-41, pAGL24:AGL24-RFP agl24-2 and pAP1:AP1-GFP ap1-15 (Wu et al., 2003) was used for chromatin extractions. Wild-type plants served as a negative control in these experiments. We analysed the genomic regions of AG, AP3, PI and SEP3 for the presence of CArG boxes [CC(A/T)6GG] (Riechmann and Meyerowitz, 1997) with a maximum of two mis-matches and A/T-rich regions that may be involved in MADS domain protein binding (Hill et al., 1998). The genomic regions that we analysed included 3 kb upstream of the ATG start codon, all exons and introns, and 1 kb downstream of the termination codon.
To analyse which CArG boxes are involved in binding the AGL24–AP1 and SVP–AP1 MADS box protein dimers, pilot ChIP experiments were performed using chromatin extracted from pSVP:SVP-GFP svp-41 plants and antiserum against GFP. Enrichment of CArG box-containing fragments was analysed by semi-quantitative PCR experiments. Those binding site fragments that were significantly enriched compared to the controls in three independent experiments were further analysed by real-time PCR (a complete set of pilot ChIP experiments is shown in Figure S1).
These real-time PCR assays confirmed that the CArG box-containing fragments AG.V, AG.VI, AP3.III, AP3.IV, PI.I and SEP3.II were significantly enriched (Figures 3a,b and S1). Subsequently, we tested using pAP1:AP1-GFP ap1-15 plants and antiserum against GFP whether AP1 interacts with the same CArG boxes as SVP. These ChIP experiments confirmed that AP1 is associated in vivo with the same fragments as SVP (Figure 3c), which further strengthens our hypothesis that the AP1–SVP dimer recruits the LUG–SEU repressor to repress floral organ identity gene expression at stages 1 and 2 of flower development.
As genetic experiments (Gregis et al., 2006) clearly showed that AGL24 and SVP have redundant functions during early stages of flower development, we also tested whether AGL24 binds to the same CArG boxes as AP1 and SVP. These ChIP assays using inflorescence tissue from pAGL24:AGL24-RFP agl24-2 plants and RFP antiserum did not result in any enrichment of the CArG box-containing fragments to which AP1 and SVP bound (Figure 3d). This surprising result might be explained by assuming that both AGL24–RFP and SVP–GFP are able to interact with AP1 and bind to the same sequences; however, SVP–GFP may have higher affinity for these interactions. To test this, we transformed the agl24-2 svp-41 double mutant with the pAGL24:AGL24-RFP construct. The AGL24–RFP fusion protein was able to complement the agl24-2 allele in this double mutant background as flower development was normal at 30°C. However, these transgenic plants flowered early due to the loss of SVP activity (data not shown).
Using inflorescences of these plants for ChIP assays, we were able to show (Figure 3d) that, in the absence of SVP, the AGL24–RFP protein associates with the same regions that were shown to interact with SVP–GFP and AP1–GFP, suggesting that AGL24 can take over its function when SVP is not expressed.
Furthermore, these results suggest that AP1, AGL24 and SVP are important in repressing directly the genes that encode components of the MADS box protein complexes that control floral organ identity determination (Pelaz et al., 2000; Honma and Goto, 2001).
Liu et al. (2009) showed that SEP3 is up-regulated in vegetative tissues of the svp-41 mutant, and this was further enhanced by combining this mutant with the soc1-2 mutant. Furthermore, reporter gene studies in which the two CArG motifs in the SEP3 regulatory region were mutated expression of the reporter gene in the whole plant, including vegetative tissues. This indicates that SEP3 is directly repressed by SVP and SOC1. Analysis of class B and C floral homeotic genes showed that these were not de-regulated in the svp-41 mutant. To analyse whether the regulation of AG is based on direct repression by SVP in vegetative tissues, we performed ChIP assays using chromatin extracted from 14-day-old seedlings grown under short-day conditions (8 h light, 16 h dark). This showed that fragments containing the CArG boxes in the AG promoter were not enriched and thus were not associated with SVP, suggesting that a different repressive mechanism acts on this gene in vegetative tissue (Figure 3e).
Expression of class B and C genes in 35S::SVP and 35S::AGL24 backgrounds
Plants that constitutively express AGL24 and SVP from the CaMV 35S promoter show various phenotypic defects. This includes the formation of new flowers from the axils of first-whorl floral organs, alteration of floral organ number and the appearance of chlorophyll-containing petals (Michaels et al., 2003; Masiero et al., 2004; Yu et al., 2004). In these plants, the first- and second-whorl organs show clear vegetative characteristics, indicating partial loss of sepal and petal identity. As AP1 is expressed in the outer two whorls, the ectopic presence of AGL24 or SVP might allow formation of AGL24–AP1 or SVP–AP1 dimers, which can form a repressive complex with SEU and LUG to down-regulate class B gene activity. To analyse this option we performed in situ hybridization experiments with AP3 and PI probes using inflorescences of the 35S::AGL24 and 35S::SVP lines. Expression of these genes was not significantly altered in these over-expression lines (Figure 4), indicating that the observed phenotypes are probably not due to repression of class B genes. This also suggests that the repression mechanism is complex and may involve many more factors.
Ectopic SOC1 expression in the floral meristem of the agl24-2 svp-41 double mutant
Previously it has been shown that SVP acts as a repressor of the floral transition by down-regulation of the floral pathway integrator SOC1 (Li et al., 2008). In contrast, AGL24 has been shown to promote the floral transition and positively controls SOC1 expression (Liu et al., 2008). Interestingly, during flower development, AGL24 and SVP have redundant functions in development of the floral meristem at stages 1 and 2 of flower development. As SOC1 is not expressed during these early stages of flower development, and its expression only re-appears in flowers at stage 3 when floral organs start to develop (Borner et al., 2000), we were interested in whether AGL24 and SVP repress SOC1 at these early stages of flower development. Therefore, we analysed SOC1 expression in developing flowers of wild-type and agl24-2 svp-41 double mutant flowers by in situ hybridization (Figure 5a,b). In wild-type inflorescences, we confirmed the previously reported expression profile (Borner et al., 2000). SOC1 expression was observed in the inflorescence meristem, and its expression re-appeared in stage 3 flowers, in which SOC1 was expressed in the central dome that later will develop into stamens and carpels (Figure 5a) (Borner et al., 2000). Interestingly, in situ analysis using inflorescences of the agl24-2 svp-41 double mutant showed that SOC1 is ectopically expressed in flowers of developmental stages 1 and 2 (Figure 5b). These data suggest that both AGL24 and SVP act as repressors of SOC1 during the early stages of flower development, although we cannot exclude the possibility that only SVP or AGL24 represses SOC1 in these young flowers. To test whether AGL24 and SVP were directly associated with SOC1 in inflorescence tissues, ChIP assays were performed which showed that these two MADS box factors were indeed associated with the CArG boxes that were previously reported to bind AP1 in inflorescences tissue (Figure S2) (Liu et al., 2007).
The ectopic SOC1 expression at stages 1 and 2 of flower development appears to compensate for the loss of AGL24 and SVP activity, as, when we combined the soc1-2 mutant with the agl24-2 svp-41 double mutant, phenotypes were observed that were similar to those of the agl24-2 svp-41 double mutant at 30°C and the ap1-12 agl24-2 svp-41 triple mutant (Gregis et al., 2006), as recently also described by Liu et al. (2009). In these mutant flowers, development is affected due to de-regulation of class B and C gene expression (Figure 5d,h).
SOC1 is able to bind directly to regulatory regions of class B and C homeotic genes
To determine whether SOC1 is a direct repressor of AG, AP3 and PI, we performed ChIP assays. In particular, the fragments (AG.V, AG.VI, AP3.III, AP3.IV and PI.I) that were shown to bind AP1, AGL24 and SVP (Figure 3) were analysed. For these ChIP experiments, we immunoprecipitated chromatin extracted from wild-type and agl24-2 svp-41 double mutant inflorescence tissues using a SOC1-specific antibody. As a negative control, we used inflorescence tissue of the soc1-2 mutant. These experiments showed that the SOC1 protein bound to various fragments (AG.V, AG.VI, AP3.IV and PI.I) in two of four experiments. However, this was only observed in the agl24-2 svp-41 mutant background (Figure 3f); no association with these fragments was observed using wild-type plant tissues. These in vivo data confirm that SOC1 is involved in the direct regulation of AG, AP3 and PI, but only if AGL24 and SVP are not present.
Direct repression of class B, C and E MADS box floral organ identity genes by AP1, AGL24 and SVP
Recently, we proposed the hypothesis that the interaction between AP1 and SVP or AGL24 is essential for floral meristem identity, and that the floral meristem starts to generate floral organs from stage 3, which is when AP1 binds to SEP proteins. The AGL24 and SVP protein expression data presented here fit well with this model, as they show that their expression at stages 1 and 2 of flower development is mutually exclusive with that of SEP3, whose expression starts at stage 3. The protein expression data also show that the presence of AP1, AGL24 and SVP at stages 1 and 2 of flower development, when the floral meristem enlarges but floral organ primordia do not yet start to differentiate, are essential for flower formation, as their absence affects all later stages of flower development.
The experiments to partially silence LUG by RNA interference resulted in mild lug phenotypes. However, when this partial silencing of LUG was combined with the agl24-2 svp-41 double mutant, significantly enhanced floral phenotypes were observed, probably due to ectopic class B and C expression. In combination with previously reported yeast two- and four-hybrid assays, which showed that AP1 can directly interact with SEU and that the AP1–SVP and AP1–AGL24 dimers are able to recruit the LUG–SEU co-repressor complex (Gregis et al., 2006; Sridhar et al., 2004, 2006), these results reveal that such interactions are necessary to control the repression of class B and C genes. Furthermore, our ChIP data show that AP1 binds to the same CArG boxes as AGL24 and SVP, suggesting that AP1–SVP and AP1–AGL24 dimers bind to multiple sites of class B (AP3 and PI), class C (AG) and class E (SEP3) genes in flower meristems. This supports our hypothesis that the repression of class B, C and E genes by AP1, AGL24 and SVP is direct.
Recently, it was suggested that SVP and AGL24 are only directly associated with the class E gene SEP3, and that the ectopic expression of class B and C genes is due to ectopic SEP3 activity (Liu et al., 2009). However, activation by SEP3 is probably not the only regulatory mechanism, as active repression of class B and C genes at early stages of flower development is needed. Indications for this derive from the in situ analysis by Liu et al. (2009), which showed that expression class B (AP3, PI) and class C (AG) genes is not completely abolished in the agl24 svp soc1 sep3 quadruple mutant. Furthermore, previous studies have shown that regulation of AG is established by activation and repression (Sieburth and Meyerowitz, 1997; Bomblies et al., 1999; Busch et al., 1999; Deyholos and Sieburth, 2000; Lohmann et al., 2001; Bao et al., 2004). Repression by LUG was shown to act via the second intron of AG (Sieburth and Meyerowitz, 1997; Deyholos and Sieburth, 2000), which fits with our observation that AP1–SVP and AP1–AGL24 dimers bind CArG boxes in the second AG intron and interact with LUG and SEU. Furthermore, mutations in these CArG boxes resulted in ectopic AG expression during early stages of flower development (Hong et al., 2003).
Interestingly, in floral meristems, SVP appears to be the main player in repression of the floral organ identity genes, as no binding of AGL24 to any of the class B, C and E genes was observed in the wild-type background. This is probably not due to the fact that we used fusions with fluorescent proteins for our ChIP analysis, as Liu et al. (2009) made a similar observation and used a peptide antibody against the native AGL24 protein for their ChIP assays. AGL24–RFP binding was only observed in the svp-41 agl24-2 mutant background. This could indicate a higher affinity of the AP1–SVP dimer for CArG boxes present in the class B, C and E genes, but could also indicate that the interaction between AP1 and SVP is more stable than the one between AP1 and AGL24. Our yeast two-hybrid assays (Gregis et al., 2006) suggest that the interaction between AP1 and AGL24 is stronger than the one between AP1 and SVP. This result was obtained in yeast, and does not take into account other stabilizing factors that could be involved. Nevertheless, it might suggest that the higher affinity of the AP1–SVP dimer for the CArG box is the reason for this difference.
The analyses that we present here do not exclude the possibility that SEP3 might be regulated by the SEU–LUG co-repressor, as AP1–SVP and AP1–AGL24 dimers bind to CArG boxes that have been shown to be important for SEP3 regulation (Liu et al., 2009). Nevertheless, it is clear that SEP3 gene expression is also regulated via two other independent pathways (Liu et al., 2009). AGL24 interacts with SAP18, a member of the Sin3/HDAC complex (Silverstein and Ekwall, 2005; Hill et al., 2008), whereas SVP interacts with TFL2, a protein that has been suggested to be involved in gene repression by recognizing H3K27me3 (Zhang et al., 2007). These repressive pathways appear to act upon SEP3 in both vegetative and floral tissues. In the svp mutant, SEP3 was up-regulated in vegetative tissues, whereas class B and C genes were not ectopically expressed. This seems to indicate that the silencing mechanism based on SVP and TFL2 does not act on class B and C genes in vegetative tissue. This was further supported by our ChIP assays, which showed that, in vegetative tissues, SVP is not associated with the CArG boxes located in the 2nd intron of AG that normally bind SVP in the floral meristem. It is known that, in vegetative tissue, AG (and also AP3) is repressed by CURLY LEAF (CLF), an Arabidopsis homolog of ‘enhancer of zeste’ (EZ) (Goodrich et al., 1997). It is therefore likely that multiple mechanisms act on the class B, C and E genes, and that these differ between vegetative and reproductive tissues.
SOC1 complements the agl24-2 svp-41 double mutant but plays no role during early stages of flower development
Combining the soc1-2 mutant with the agl24-2 svp-41 mutant showed a significant enhancement of the double mutant phenotype, as recently also reported by Liu et al. (2009). This observation cannot be explained by the SOC1 expression profile, as it is not expressed during stages 1 and 2 of flower development in wild-type plants (Borner et al., 2000). An explanation for the observed phenotypes comes from in situ analysis of SOC1 expression in the agl24-2 svp-41 double mutant, which showed that SOC1 is expressed at stages 1 and 2 of flower development in the double mutant. This indicates that AGL24 and SVP repress SOC1 expression in the floral meristem. This repression is probably direct, as our ChIP assays showed that both SVP and AGL24 are associated with CArG box-containing fragments in the SOC1 promoter.
SOC1 is a MIKC type MADS box factor that is not closely related to AGL24 and SVP (Parenicováet al., 2003). However, a matrix-based interaction study in yeast to investigate all possible MADS box protein interactions (de Folter et al., 2005) showed that SOC1 has a partially overlapping interaction profile with AGL24 and SVP, as, like AGL24 and SVP, SOC1 interacts with FRUITFULL (FUL), AP1 and CAL (see also Pelaz et al., 2001; Gregis et al., 2006). It is therefore possible that, in the absence of AGL24 and SVP, SOC1 interacts with AP1 to recruit the SEU–LUG co-repressor to silence class B and C genes. Therefore, we tested whether SOC1 was associated with class B and C genes in inflorescence tissues of the agl24-2 svp-41 double mutant. These ChIP assays did indeed show that SOC1 binds these floral homeotic genes, explaining how SOC1 can complement the loss of AGL24 and SVP activity and thereby explaining the mild phenotype observed in the agl24-2 svp-41 double mutant at normal growth temperatures. However, the fact that, at 30°C, the agl24-2 svp-41 double mutant (Gregis et al., 2006) has the same phenotype as the agl24-2 svp-41 soc1-2 triple mutant indicates that insertion of SOC1 in this repressive complex results in a complex that does not have exactly the same characteristics as those containing AGL24 or SVP.
A new model for early stages of flower development
Our data show that during stages 1 and 2 of flower development, when the floral meristem increases in size, AP1 forms dimers with SVP (mainly) and AGL24 to repress class B, C and E floral homeotic genes. Furthermore, we show that the repression of class B, C and E genes is direct, as these factors were found to bind to CArG boxes present in AP3, PI, AG and SEP3 (Figure 3). Moreover, protein interaction studies and genetic data show that SVP and AGL24 repress class B and C gene expression via the activity of LUG and SEU (Figure 6a).
Recently, Liu et al. (2009) showed that SEP3 is regulated by two independent pathways. SVP interacts with TFL2 to maintain SEP3 chromatin in a silenced state by modulating trimethylation of histone H3 at lysine 27 (3 at lyusine 27 (H3K27me3), whereas AGL24 interacts with SAP18 (a member of the HDAC complex) to modulate histone H3 acetylation. SEP3 expression is an important step in flower development, as when this MADS box factor is expressed, it will (probably via interaction with AP1) activate B and C genes, repress the expression of AGL24 and SVP, and subsequently floral organ formation will start (Gregis et al., 2008; Liu et al., 2009). This means that SEP3 has to be repressed during stages 1 and 2 of flower development. However, we found that the regulation of class B and C floral organ identity genes is not exclusively controlled by SEP3, as there is also a pathway in which complexes containing AP1, AGL24, SVP, SEU and LUG directly repress B and C class genes (Figure 6a).
An interesting observation is that the repressive activity of LUG is mediated via two independent pathways (Gonzalez et al., 2007). One pathway is based on the interaction of LUG with the histone deacetylase HDA19, and the other is based on interaction with the mediator components MED14 and CDK8. Although more research is needed, it is not possible to exclude at this point that these two pathways represent the two distinct repressive pathways in which AGL24 and SVP are involved. This would simplify our model, in which the repression of class B, C and E genes would be controlled by two independent pathways (Figure 6b).
It is important to note that this model explains the genetic and molecular interactions observed during stages 1 and 2 of flower development. In vegetative tissue, class B and C genes appear not to be regulated by this pathway. In contrast to the regulation in floral meristems, repression of SEP3 by the SAP18 and TFL2 pathways in vegetative tissue does not seem to be completely redundant, as loss of SVP activity results in SEP3 up-regulation. Furthermore, in vegetative tissue, the complex that controls SEP3 expression by SAP18 is mainly based on SOC1 activity and not that of AGL24 (Liu et al., 2009), whereas, in flower tissue, it is AGL24 that controls the repression via SAP18. These complexes are probably also different, as ectopic SOC1 in the floral meristems of the agl24-2 svp-41 double mutant can only complement the absence of AGL24 at normal growing temperatures, whereas at higher temperatures SOC1 is unable to repress class B and C gene expression.
Chromatin immunoprecipitation assays
For ChIP experiments, the commercial antibodies GFP:Living Colors® full-length A.v. polyclonal antibody and Living Colors® DsRed polyclonal antibody (Clontech, http://www.clontech.com/) were used. For the SOC1 ChIP assays, an antibody against a synthetic peptide (SVKCIRARKTQVFK) was used. Chromatin was prepared from inflorescences of pSVP:SVP-GFP svp-41, pAP1:AP1–GFP ap1-15 and pAGL24:AGL24-RFP agl24-2 (the latter also in the agl24-2 svp-41 mutant background) and from 14-day-old seedlings grown under short-day conditions (8 h light, 16 h dark). Wild-type plants were used as negative controls. The SOC1 experiments were also performed on the soc1 single mutant and the agl24-2 svp-41 double mutant. The complete primer sets are shown in Table S2. PCR assays were performed as previously described (Gregis et al., 2008). ChIP assays were performed as described in detail in Appendix S1.
Quantitative real-time RT-PCR
Enrichments fold 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 iCycler iQ optical system (software version 3.0a). Relative enrichment was calculated as described in Appendix S1.
Plasmid construction and plant transformation
For construction of p35S:LUG-RNAi, we used a specific portion of the LUG coding region. The LUG sequence was amplified using Gateway primers 5′-GGGGACCACTTTGTACAAGAAAGCTGGGTATAGTTTTCACTTCCACAG-3′ and 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTGGTTATTCCCTTCGTAC-3′ cloned into Gateway vector pDONOR207 and then recombined into pFGC5941.
For construction of pAGL24:AGL24-RFP, we cloned the AGL24 genomic sequence, including 3 kb upstream of the start codon, in a C-terminal fusion with mRFP1 (Campbell et al., 2002). We modified the pB7RWG2 Gateway binary vector (Karimi et al., 2002) that contains the mRFP1 protein, by deleting the CaMV 35S promoter using restriction enzymes SpeI and HindIII and subsequent filling of cohesive ends and back-ligation.
The pAGL24:AGL24 sequence was amplified using Gateway primers 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTTCTCGTCTTTTTGTGTGTCTGG-3′ and 5′-GGGGACCACTTTGTACAAGAAAGCTGGGTTTTCCCAAGATGGAAGCC-3′, cloned into Gateway vector pDONOR207 and then recombined into the modified version of pB7RWG2.
For construction of pSVP:SVP-GFP, we cloned the SVP genomic sequence, including a fragment of 3 kb upstream of the start codon, in a C-terminal fusion with GFP. The pSVP:SVP sequence was amplified using Gateway primers 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTCCGTAAATATCGTCAGTCTCG-3′ and 5′-GGGGACCACTTTGTACAAGAAAGCTGGGTCCACCACCATACGGTAAGCTGCAC-3′, cloned into Gateway vector pDONOR207, and then recombined into pGW::GFP.
Confocal laser scanning microscopy
To observe the localization of GFP-tagged SVP and RFP-tagged AGL24 in living plant tissue, inflorescence material was dissected until the relevant meristems and flower buds became visible. The tissues were fixed in 4% paraformaldehyde solution (0.2 m phosphate buffer, pH 7.2), and washed in 0.2 m phosphate buffer, pH 7.2. Tissues were embedded into 5% agar (0.2 m phosphate buffer, pH 7.2). Sections of 100 μm thickness were obtained using a vibratome. Confocal laser scanning microscopy (CLSM) of the living plant tissue was performed using a Leica TCS SP2 AOB5 microscope (http://www.leica.com). GFP was excited with the 488 nm line of an Ar/Kr laser. The GFP emission was 500–550 nm. RFP was excited with the 594 nm line of a He/Ne laser. The RFP emission was 600–640 nm.
Arabidopsis flowers were fixed and embedded in paraffin as described previously (Huijser et al., 1992). Sections of plant tissue were probed with digoxigenin-labelled PI (primers 5′-TAATACGACTCACTATAGGGGTTTCACACAGATCAATTAGTTCG-3′ and 5′-AACAACAGGAGATGGCTATAGC-3′), AP3 and AG antisense RNA (Gregis et al., 2006). Hybridization and immunological detection were performed as described by Coen et al. (1990).
We thank Drs G. Coupland (Max Planck Institute for Plant Breeding Research, Cologne, Germany), D. Shubert (Institut für Genetik Heinrich-Heine Universität Düsseldorf, Germany) and R.M. Amasino (Department of Biochemistry, University of Wisconsin, WI) for providing the pGW::GFP plasmid, the ChIP’s protocol and the soc1-2 mutant, respectively. Thanks also to Simona Masiero, Lorena Simioni, Vittoria Brambilla and the Mantovani and De Biasi groups (University of Milan) for technical support. Scanning electron and confocal microscopy were performed at the Centro Interdipartimentale di Microscopia Avanzata (CIMA), Museo Civico di Storia Naturale (Milano) and Interdisciplinary Centre for Nanostructureo Materials and Interfaces (CIMAINA) with the help of Drs G. Melone, M. Zilioli, S. Rodighiero and U. Fascio. The post-doctoral and PhD fellowships for V.G. and A.S., respectively, are founded by the University of Milan. C.D.-F. was supported by Marie Curie EU project Transistor (512285).