Conserved intragenic elements were critical for the evolution of the floral C-function


(fax +44 113 343 3144; e-mail


The floral C-function, which specifies stamen and carpel development, played a pivotal role in the evolution of flowers. An important aspect of this was the establishment of mechanisms regulating the temporal and spatial expression domain of the C-function genes. Transcription of the Arabidopsis C-function gene AGAMOUS (AG) is tightly controlled by factors that interact with cis-elements within its large second intron. Little is known about the regulatory role of intragenic elements in C-function genes from species other than Arabidopsis. We show that a binding site for the LEAFY (LFY) transcription factor, present in the AG intron, is conserved in the introns of diverse C-function genes and is positioned close to other conserved motifs. Using an in planta mutagenesis approach, we targeted evolutionarily conserved sequences in the intron of the Antirrhinum PLENA (PLE) gene to establish whether they regulate PLE expression. Small sequence deletions resulted in a novel class of heterochronic C-function mutants with delayed onset of PLE expression and loss of stamen identity. These phenotypes differ significantly from weak C-function mutant alleles in Antirrhinum and Arabidopsis. Our findings demonstrate that the PLE intron contains regulatory cis-elements, including a LFY-binding site, critical for establishing the correct C-function expression domain. We show that the LFY site, and other conserved intron elements, pre-date the divergence of the monocot and dicot lineages, suggesting that they were a determinant in the evolution of the C-function, and propose a threshold model to explain phenotypic divergence observed between C-function mutants.


The central tenet of the long-established ABC model of flower development proposes that three homeotic functions (A, B and C), each operating in two adjacent whorls, specify the identity of the four floral organ types (Coen and Meyerowitz, 1991). The A-function acts alone in the outermost whorl (whorl 1) to specify sepal identity, and together with the B-function to specify petal identity in whorl 2. The specification of the reproductive organs requires the activity of the C-function. In whorl 3, the B- and C-functions together specify stamens. At the centre of the flower, in whorl 4, the C-function acts alone to initiate carpel development and to terminate further development of the floral meristem. The ABC model also proposes that activity of the C-function is restricted to the third and fourth whorls by the A-function and vice versa.

Complete loss of the C-function results in the conversion of third-whorl stamens to petals, and fourth-whorl gynoecium to a new indeterminate mutant flower consisting of many whorls of sepals and petals. Weak C-function mutant alleles, which are typified by loss of determinacy and female organ identity, but only partial loss of male identity, reveal that the roles of the C-function genes in stamen and carpel development and floral determinacy are separable and require different threshold levels of C-function activity (Mizukami and Ma, 1995; Sieburth et al., 1995; Roe et al., 1997; Davies et al., 1999; Durfee et al., 2003). Current models suggest that the C-function expression domain and C-function gene product thresholds are established through a dynamic relationship between C-function gene activators and repressors (Sridhar et al., 2006; Cartolano et al., 2007). For example, the Arabidopsis C-function gene AGAMOUS (AG) is activated at the centre of the flower by a key regulator of floral identity LEAFY (LFY), in combination with the meristem maintenance gene WUSCHEL (WUS), and other enhancers (Lenhard et al., 2001; Lohmann et al., 2001), but is repressed by the A-function proteins APETALA1 and APETALA2 (AP2), and other factors including LEUNIG (LUG), SEUSS, BELLRINGER (BLR), BELL1, RABBIT EARS, AINTEGUMENTA, STERILE APETALA, AGL24 and SHORT VEGETATIVE PHASE (Gregis et al., 2006; Wellmer et al., 2006; Zahn et al., 2006). The regulation of C-function gene expression by similar factors in the model species Antirrhinum majus suggests that some of the mechanisms used to control the C-function domain are conserved (Hantke et al., 1995; Motte et al., 1998; Navarro et al., 2004).

Early examination of AG expression revealed that intragenic elements were required to establish the correct AG domain (Sieburth and Meyerowitz, 1997). Later it was shown that these elements were located in the large second intron of the AG gene, and subsequent work concentrated on dissecting the AG intron. It is now established that the AG intron responds to many of the activators and repressors of AG expression, including LFY, WUS, AP2, LUG and BLR (Sieburth and Meyerowitz, 1997; Bomblies et al., 1999; Busch et al., 1999; Deyholos and Sieburth, 2000; Lohmann et al., 2001; Bao et al., 2004). Furthermore, the AG intron contains separate domains that independently activate AG expression in whorls 3 and 4 (Deyholos and Sieburth, 2000).

Despite the highly similar intron/exon structure of C-function genes (Zhang et al., 2004), and conservation of C-function gene regulators, it is not clear whether the second intron plays an important regulatory role in other species. A combination of sequence comparisons and biological studies has identified a number of functionally important cis-elements conserved in the second introns of C-function genes from diverse species. One of the first to be identified was a 70-bp element characterised by a direct repeat of the sequence CCAATCA, which was later shown to be required for late-stage AG expression in Arabidopsis (Davies et al., 1999; Hong et al., 2003). Another conserved element, the aAGAAT box, for which no function has yet been established, was identified in the second introns of dicot C-function genes, with the exception of the Antirrhinum PLENA (PLE) gene (Hong et al., 2003). Furthermore, binding sites for LFY and other transcription factors have also been identified. While the functions of some of these elements are well characterised in Arabidopsis (Busch et al., 1999; Lohmann et al., 2001; Hong et al., 2003), their roles in other species have yet to be determined.

Our aim was to study the function of putative cis-elements in the Antirrhinum PLE gene. Sequence comparisons between introns from species spanning large evolutionary distances revealed that a LFY-binding site previously identified in AG (LBS2; Hong et al., 2003), located immediately upstream of the 70-bp element, was conserved. Using an in vivo mutagenesis approach we disrupted the PLE intron in the region containing these conserved elements to determine whether they are important for PLE expression, reproductive organ development and floral meristem determinacy. We found that deletion of a region of the intron containing the conserved LFY-binding site significantly delayed PLE expression, resulting in the loss of stamen identity in the third whorl of mutant flowers, while carpel development and determinacy were unaffected. This novel phenotype differs from those observed in weak C-function mutant alleles in all species. Our results suggest that the LFY-binding site, which was probably present in the ancestor of monocots and dicots, determines the onset of C-function gene expression, thus affecting its spatial expression domain and may therefore have been important in the evolution of the C-function.


cis-elements in the large intron of C-function genes are conserved in moncots and dicots

Previous studies identified two relatively small regions of conservation between the second introns of dicot C-function genes. The first of these was a 70-bp region, characterised by a repeat of the sequence CCAATCA (Davies et al., 1999), which plays a role in the maintenance of AG expression in Arabidopsis flowers (Hong et al., 2003). The second motif (the aAGAAT box) was found upstream of the 70-bp element using pairwise BLAST alignments of C-function gene introns from the Brassicaceae and other dicot species. Curiously, this motif did not appear to be present in the intron of the Antirrhinum PLE gene (Hong et al., 2003).

We expanded these studies to determine how far back in evolutionary time these elements might be conserved. We compared the second intron of the rice (Oryza sativa) C-function gene OsMADS3, a representative of the monocots which diverged from the dicots at least 140 million years ago (Chaw et al., 2004), with that of dicot introns, including sequences unavailable in previous studies. We found that the aAGAAT box is present in the second introns of all dicot C-function genes (including PLE), and that a variant of this sequence is also present in the second intron of OsMADS3 (Figure 1a,b). Similarly, the 70-bp region was also conserved in monocot and dicot introns (Figure 1a,c). Further studies revealed that the position of the aAGAAT box relative to the 70-bp element did not vary greatly between species (Figure 1a and Figure S1 in Supporting Information). Together the data suggests that the aAGAAT box and 70-bp element were present in the common ancestor of the monocots and dicots, and that the region including these elements may be important in the regulation of C-function gene expression. To investigate this further, we examined the sequence located between these two motifs in an attempt to identify additional regulatory elements. While these comparisons showed that this region had diverged significantly over time (Figure S2), we found that a LFY-binding site previously identified in the AG intron (LBS2; Hong et al., 2003) was conserved. Moreover, we found that the LFY site was positioned at a similar distance upstream of the 70-bp element in all C-function gene second introns (Figures 1a,c and S1).

Figure 1.

 Conserved elements in the second introns of C-function genes.
(a) Evolutionary relationships of C-function genes from diverse species belonging to the dicot AG (blue) and PLE (red) lineages, or to the monocots, are shown in the phylogram on the left. On the right, the relative positions of the aAGAAT box (white box), the conserved LFY site (grey box), and the 70-bp element (represented by two black boxes indicating the CCAATCA sequences) are shown for each intron (approximately to scale).
(b) ClustalW alignment of the aAGAAT box. Conserved nucleotides are shaded. The consensus sequence is presented below the main alignment.
(c) ClustalW alignment of the intron regions that include the conserved LFY-binding site (boxed, with matches to the LFY consensus binding site highlighted in red), and the two CCAATCA boxes (underlined in black), which are characteristic of the 70-bp element. Conserved nucleotides are shaded. C-function intron 2 sequences were taken from Antirrhinum majus (PLE and FAR; AY935269 and AJ239057, respectively), Arabidopsis thaliana (AG; At4g18960), Ipomoea nil (DP; AB281192), tomato (TAG1; AY254705), poplar (PTAG1; AF052570), Petunia (pMADS3; AB076051), cucumber (CUM1; AY254704) and rice (OsMADS3; nucleotide 5567050-5561700 of AP008207.1).

We also examined the large intron of a second rice C-function gene, OsMADS58, and found that although the LFY site and 70-bp element were conserved as they are in other species we could not identify an upstream aAGAAT box (Figure S1). Interestingly, the two rice C-function genes have different spatial expression patterns, and while OsMADS3 has a predominant role in stamen development, OsMADS58 is more involved in carpel development and floral determinacy (Yamaguchi et al., 2006). The structure of the intragenic regulatory unit in these genes may account for the functional differences.

Our data suggest that the DNA sequence, relative position and spacing of these elements have been maintained over long evolutionary distances. They are found in the second introns of C-function genes belonging to the AG (which includes the Antirrhinum FAR gene) and PLE lineages (Kramer et al., 2004; Causier et al., 2005) from diverse dicot species, and were probably present in the C-function genes of the common ancestor of the monocots and dicots. Since these regions have been conserved in otherwise divergent intron sequences, it is likely that they have a key regulatory function and were important in the evolution of the C-function.

The conserved LFY-binding site in the PLE intron is bound by FLO

The putative conserved LFY-binding site identified in PLE from Antirrhinum has one mismatch to the consensus Arabidopsis LFY-binding site (Figure 1c). Our first step in examining the functional importance of this LFY-binding site was to confirm that it was capable of being bound by the FLO protein (the Antirrhinum orthologue of LFY; Coen et al., 1990; Weigel et al., 1992). Using a region of the PLE intron containing the LFY site (the ‘onset box’, see below) as bait in a yeast one-hybrid assay we observed that FLO showed a clear interaction with the ‘onset box’, but was unable to interact with a negative control element (p53) (Figure 2).

Figure 2.

 The Antirrhinum FLO protein binds to a region of the PLE intron containing the conserved LFY site.
Yeast one-hybrid analysis was used to test binding of FLO and ROA proteins to the PLE‘onset box’. Binding of the PLE, ROA, FLO and murine p53 prey proteins (shown across the top) was tested against the following bait target sites (listed to the left): the ‘onset box’ (pHis2-ob), and the binding site for the p53 protein (p53His2) to assess specificity. The presence of yeast growth on media lacking histidine (–His) indicates an interaction between the prey protein and the target site.

Generating targeted small deletions in planta

The contribution of cis-elements to gene expression and function is usually determined by transgenic approaches, using reporter gene analyses. These approaches are subject to variability due to integration position effects and do not investigate gene expression in the endogenous chromosomal environment. To examine the function of the conserved intron elements we used an in planta mutagenesis to generate small targeted deletions within the second intron of PLE. The classic Antirrhinum ple-625 allele (Bradley et al., 1993) is caused by insertion of a transposable element (Tam3) within the second intron of the gene, between the aAGAAT box and the 70-bp element, close to the conserved LFY-binding site (Figure 3a). No PLE transcript is detectable in the ple-625 mutant allele (Bradley et al., 1993). ple-625 is somatically unstable and the transposon is readily mobilised by placing growing plants at 15°C. Imperfect excisions result in alterations to the sequence at the insertion point. A large-scale temperature shift experiment to induce excision of the transposable element led to the isolation of two novel recessive phenotypes, which we will henceforth refer to as plepetaloid double (plepd) and pleno anthers (plena).

Figure 3.

PLE intron 2 deletions in plepd and plena.
(a) The PLENA gene. The boxes represent the PLE exons that encode the PLE MADS, I, K and C domains (underlined in grey beneath the gene). The PLE loss-of-function allele ple-625 is caused by insertion of the Tam3 transposon into the second intron of PLE (position shown), which lies close to the conserved LFY site (grey box). The approximate positions of the aAGAAT box (white box) and the 70-bp element (black boxes) are also indicated. The size of the intron is shown (4087 bp), as are the positions of the full PLE gene (26447–36217) and intron 2 (30566–34646) within the published PLE BAC sequence (accession number AY935269).
(b) A PCR spanning the deletion sites in the PLE intron of plepd and plena. The PCR was carried out using the oligonucleotide primers shown schematically as grey arrows in (c).
(c) Position of the intron 2 deletions in plepd and plena. The size and position of the deletion in each allele is indicated (473 bp in the case of plepd, and 187 bp for plena). The first nucleotide position for the aAGAAT box (white box), the LFY site (grey box) and the first CCAATCA sequence of the 70-bp element (black boxes) are shown.
(d) Sequence of the ‘onset box’. The numbers represent the nucleotide position for the beginning and the end of the ‘onset box’ within the second intron of PLE. Predicted transcription factor binding sites, and their orientations, are shown. GAMYB is a class of MYB-related transcription factors (TFs), DOF proteins are Zn-finger TFs, and NF-YA represents a member of the CCAAT box binding proteins. The consensus binding sites for the LFY and WUS proteins are indicated. Vertical lines show homologies between the consensus sites and those predicted within the ‘onset box’ sequence.

The PCR analysis revealed that, as expected, excision had generated targeted deletions in the second intron of the PLE gene in both plepd and plena mutants (Figure 3b). DNA sequencing confirmed that 473 bp of intron 2 was deleted in plepd, and 187 bp was deleted in the plena mutant (Figures 3c and S3). It is interesting to note that although the plena phenotype is more severe than that of plepd (see below), it results from the smallest deletion. The plepd deletion encompasses most of the plena deletion, but there is a region of 49 bp that is deleted only in the more severe plena (Figure 3c,d). We have named this region the ‘onset box’. To confirm that the plepd and plena phenotypes were due to the deletions caused by mobilising the Tam3 element, and not due to its reinsertion elsewhere in the PLE gene, we carried out PCRs spanning the entire PLE gene. These analyses confirmed that there were no transposon insertions anywhere within the PLE gene, in either mutant (data not shown).

An examination of the ‘onset box’ sequence for potential cis-elements revealed that this region includes the conserved LFY-binding site (Figure 3d), suggesting that the more severe plena phenotype may correlate with the loss of this element. In addition, a number of other potentially interesting sites were revealed within the ‘onset box’ (Figure 3d). In the AG intron, the conserved LFY-binding site (LBS2; Hong et al., 2003) is closely associated with a binding site for another activator of AG, WUS (Lohmann et al., 2001). Similarly, the LFY site in the PLE intron is also adjacent to a WUS-like binding site (Figure 3d). However, we were unable to detect binding of the ROA protein (the Antirrhinum orthologue of WUS; Kieffer et al., 2006) to the ‘onset box’ in our assay system (Figure 2).

Other potentially interesting transcription factor consensus binding sites found within the ‘onset box’ include those for factors such as GAMYB, DOF and NF-YA (Figure 3d). GAMYB and DOF proteins play a role in gibberellin (GA) signalling (Diaz et al., 2002), and floral homeotic genes, including AG and LFY are targets of GA signalling pathways (Gocal et al., 2001; Yu et al., 2004). NF-YA proteins, which bind the sequence CCAAT, have been implicated in the regulation of C-function genes (Hong et al., 2003; Cartolano et al., 2007). However, these elements do not appear to be conserved in this region in other species, and further study will be required to establish the importance of these sites in PLE regulation.

Deletion of the conserved LFY site from the PLE intron affects stamen development

Wild-type Antirrhinum flowers normally consist of an outer whorl of five sepals, a second whorl of five lobed petals fused at the base to form the corolla tube, a third whorl of four free-standing stamens, and a central gynoecium (Figure 4a–d). In contrast, the loss-of-function ple phenotype (Carpenter and Coen, 1990; Schwarz-Sommer et al., 1990; Bradley et al., 1993) is characterised by the conversion of the reproductive organs to perianth organs. An indeterminate number of petals form in the second and third whorls, with partial fusions between whorls. In the fourth whorl, a new indeterminate mutant flower that consists solely of whorls of perianth organs replaces the carpel.

Figure 4.

 Floral phenotypes of Antirrhinum wild-type, plepd and plena plants.
(a) Graphical representation of the structure of a wild-type flower in cross section. The positions of the dorsal, lateral and ventral petals of the flower are indicated.
(b) Side view of an entire wild-type flower.
(c) Side view of a longitudinal section of a wild-type flower showing stamens (st) and central gynoecium (ca).
(d) Longitudinal section of an early wild-type floral bud. Floral whorls are numbered. The structure of wild-type stamens in whorl 3, and carpels in whorl 4 can be seen.
(e) Graphical representation of the structure of a plepd flower in cross section, highlighting fusion of stamens to the lateral/ventral petals in the second whorl (arrowed) and petaloid stamens (p/st) in whorl 3.
(f) Side view of an entire plepd flower.
(g) Side view of a longitudinal section of a plepd flower showing petaloid stamens (p/st) and central gynoecium (ca).
(h) Longitudinal section of an early plepd floral bud. Floral whorls are numbered. The structure of petaloid stamens (p/st) in whorl 3 can be seen.
(i) Graphical representation of the structure of a plena flower, grown under standard conditions, in cross section.
(j) Side view of an entire plena flower.
(k) Side view of a longitudinal section of a plena flower showing conversion of stamens to petals (arrowed) and central gynoecium (ca).
(l) Structure of an entire, and partially dissected gynoecium from a plena flower grown at 30°C. The carpel walls appear to consist of several whorls of petal-like tissue (centre image) that enclose developing ovules (right image).
(m) Longitudinal section of an early plena floral bud. Floral whorls are numbered. The conversion of third-whorl stamens to petals can be seen.

Of the mutants isolated in the excision experiment, plepd has the milder phenotype and is characterised by a partial conversion of third-whorl stamens to petals (Figure 4e–h). The second-whorl petals of plepd flowers tend to be split, especially between the two dorsal petals, to the base of the corolla tube, and there is often a deep notch between the dorsal and lateral petals. The third-whorl organs are petaloid and are often fused to the inner face of the corolla tube in whorl 2, particularly where the lateral and ventral petals of the flower fuse. Freestanding stamens are found in the third whorl, but these are often fused to ectopic strips of petal-like tissue, producing hybrid organs in these positions (see Figure 4h). Strips of freestanding petal-like tissue are also seen in whorl 3 of plepd flowers. The carpels appear as wild type, and the flowers are male and female fertile.

The more severe mutant, plena, is characterised by the complete conversion of third-whorl stamens to petals (Figure 4i–m). The second and third whorls of plena flowers resemble those of the most severe loss-of-function ple mutants. However, unlike ple-625, the fourth whorl of plena terminates with the production of a fertile gynoecium (viable seed is produced following cross-fertilization with pollen from wild-type flowers). Interestingly, though, at elevated temperatures (in excess of 30°C) there is a loss of floral determinacy and the carpel becomes petaloid, with several extra whorls of petals fusing to form the carpel wall, enclosing the ovules that form at the centre of this hybrid structure (Figure 4l).

Although plepd and plena can be considered weak ple alleles, they differ significantly in phenotype from the weak C-function mutants (ple-1, ag-4 and ag-5) and weak AG RNAi lines (Mizukami and Ma, 1995; Sieburth et al., 1995; Roe et al., 1997; Davies et al., 1999; Durfee et al., 2003). In plepd and plena flowers, stamen identity is compromised while carpel identity and determinacy are unaffected. In contrast, in ple-1, ag-4, ag-5 and the weak AG RNAi lines determinacy and gynoecium development are disrupted to a much greater extent than stamen identity.

Early PLE expression is affected in plepd and plena

The expression of the PLE gene was examined in each mutant by in situ hybridization to determine what effect the intron deletions had on PLE expression at different floral stages. In wild-type Antirrhinum flowers, PLE expression is detectable from floral stage 4 (floral stages defined according to Carpenter et al., 1995; Zachgo et al., 1995; Vincent and Coen, 2004) at the centre of the floral meristem (Figure 5a). As the flower matures PLE expression is maintained at the centre of the flower in developing stamens and carpels (Figure 5b–d). In contrast, early PLE expression in the plepd and plena mutants was disrupted.

Figure 5.

In situ hybridization analysis of PLE expression in plepd and plena flowers.
(a) PLE expression in an early wild-type flower (approximately stage 4).
(b) PLE expression in a stage 5–6 wild-type flower.
(c), (d) PLE expression in later stage wild-type flowers.
(e) PLE expression in a stage 4 plepd flower.
(f) PLE expression in a stage 5-6 plepd flower.
(g), (h) PLE expression in later stage plepd flowers.
(i) PLE expression in a stage 4 plena flower.
(j) PLE expression in a stage 5–6 plena flower.
(k), (l) PLE expression in later stage plena flowers.
Each column of panels shows floral sections at approximately comparable developmental stages (indicated above), and are at the same scale with black bars representing 250 μm. Floral whorls are numbered and sepals (Se) indicated. Signal is seen as purple/blue staining on the sections.

In plepd, PLE expression was detectable from floral stage 4, but was weaker than that observed in wild-type flowers (Figure 5e). However, as third- and fourth-whorl organs developed, PLE levels were similar to that observed in the wild type (compare Figure 5b–d with 5f–h, respectively). In plena, PLE expression was not detected at floral stage 4 (Figure 5i), and was not observed until stage 5–6, albeit weakly (Figure 5j), significantly later than in both wild-type and plepd flowers, and was restricted to the fourth whorl (Figure 5j). Later-stage plena flowers showed wild-type PLE expression in the developing carpels but a complete absence of PLE transcript in third-whorl organs (Figure 5k,l). The RT-PCR analysis confirmed that the fourth-whorl transcript was of the expected size, ruling out possible splicing defects (data not shown).

Together the data suggest that the PLE intron deletions have generated heterochronic mutants. The onset of PLE expression, particularly in the plena mutant, is delayed, leading to third-whorl organ identity defects. PLE levels in the fourth whorl of both mutants are similar to those observed in wild-type flowers, and carpels develop normally under our standard glasshouse conditions.


An evolutionarily conserved intragenic regulatory region

The evolution of the C-function, a single genetic function that specifies reproductive organ identity, and the mechanisms that control its expression domain were key factors in the emergence of flowering plants (Meyerowitz, 1994). While some mechanisms for modulating C-function gene activity appear to vary in different species (e.g. Cartolano et al., 2007), others have been conserved over large evolutionary distances. Arabidopsis and Antirrhinum diverged approximately 125 million years ago (Wikström et al., 2001), yet the homologous proteins LUG and STYLOSA, respectively, negatively regulate C-function gene expression in these species (Liu and Meyerowitz, 1995; Motte et al., 1998; Conner and Liu, 2000; Navarro et al., 2004). Similarly, both the Antirrhinum FLO gene and its Arabidopsis counterpart LFY are involved in the early activation of C-function gene expression (Weigel and Meyerowitz, 1993; Hantke et al., 1995). LFY activates AG expression by binding directly to positions within the second intron of AG (Busch et al., 1999). To date, there is little information regarding the regulatory importance of the second intron of C-function genes in other species, or whether the intragenic mechanism used by LFY to regulate the Arabidopsis C-function gene is conserved.

Comparison of second intron sequences from diverse species has revealed a number of conserved elements that are likely to have been present in the common ancestor of the monocots and dicots (Davies et al., 1999; Hong et al., 2003; this study). The aAGAAT box was originally identified in AG, the C-function genes of other Brassicaceae and some other dicot species, although it was not found in the Antirrhinum PLE gene (Hong et al., 2003). We have expanded this study to show that the motif is present in the intron of all dicot C-function genes analysed (including PLE), and is also present in the intron of the rice C-function gene OsMADS3 (Figure 1), a representative of the monocot lineage. Similarly, a 70-bp motif, situated downstream of the aAGAAT box, is also conserved in the second intron of C-function genes from both monocots and dicots (Davies et al., 1999; Hong et al., 2003; this study). While no function has yet been elucidated for the aAGAAT box, the 70-bp element is required for late stage AG activity in Arabidopsis (Hong et al., 2003). Interestingly, these two elements are located similar distances apart in the second introns of all C-function genes analysed (Figures 1 and S1), suggesting that the region encompassing these motifs may be functionally important. Although DNA sequence comparisons in this region show considerable sequence divergence, the LFY-binding site located immediately upstream of the 70-bp element in the AG intron (LBS2; Hong et al., 2003) is also present at a similar position relative to the 70-bp element in all other species examined (Figures 1 and S1). Binding of the FLO protein to a portion of the PLE intron containing this conserved LFY-binding site provides some evidence for its functional importance (Figure 2). The sequence and spatial conservation of these three elements (aAGAAT–LFY–70 bp) suggests that they form a regulatory unit that pre-dates the divergence of the monocot and dicot lineages. The mechanistic connection between the temporal and spatial expression patterns of genes expressed in the developing floral meristem (see below) therefore implies that this region is part of an ancient mechanism that ultimately led to the appropriate third- and fourth-whorl expression of C-function genes now seen in perfect flowers.

In planta mutagenesis reveals a role for the conserved region in the onset of C-function expression

To investigate the function of this region of the Antirrhinum C-function gene PLE without disrupting its endogenous chromosomal environment, we adopted an in vivo mutagenesis approach. Using induced imprecise transposon excision we generated two small deletions in the targeted region of the second intron, each of which had a novel phenotype. Unlike the parental ple-625 flowers, that fail to produce reproductive organs and are indeterminate, the new mutants are only defective in stamen development (Figure 4). Analysis of PLE expression revealed that the loss of stamen identity corresponded with a delay in the onset of PLE expression, particularly in the more severe mutant plena (Figure 5). This demonstrates that the second intron of PLE, like that of AG, plays an important role in regulating appropriate expression of the C-function gene. It also suggests that the complete loss of reproductive organ identity and floral determinacy in ple-625 is not due to the disruption of a single cis-element by the transposon, since deletion of the same region in plepd and plena results in weaker phenotypes.

It remains possible that the loss of PLE gene expression in ple-625 may be due to disruption of the spatial conservation between the three elements (aAGAAT–LFY–70 bp) caused by insertion of the 3-kb Tam3 transposon into this region of the intron (Figure 3). However, we cannot rule out the possibility that the ple-625 phenotype is caused by an alternative mechanism, such as a transcription terminator sequence within the transposon.

The loss of the conserved intron sequences, especially the ‘onset box’ which includes the conserved LFY site, correlates with significantly delayed onset of PLE expression, resulting in complete conversion of third-whorl stamens to petals in plena. This suggests that the conserved LFY site plays a significant role in early C-function gene expression and thereby the establishment of the correct expression domain. Several lines of evidence support a key role for this LFY-binding site. First, AG expression in the loss-of-function lfy-6 allele is similar to that of PLE in plena, showing delayed onset of expression but reaching a wild-type pattern at later stages (Weigel and Meyerowitz, 1993). Second, deletion or mutation of the conserved LFY site in the AG intron (LBS1 and LBS2 – two functionally redundant LFY sites found close to one another upstream of the 70-bp element) also results in reduced AG activity (although it is unclear whether timing of AG expression is altered) (Bomblies et al., 1999; Busch et al., 1999; Hong et al., 2003). Finally, although functional studies in Arabidopsis have revealed that the AG intron can be separated into distinct regions required for early and late AG activity, each containing LFY-binding sites, it is only the LFY-binding site between the aAGAAT and the 70-bp elements that is absolutely conserved in all species tested. For example, the second intron of the cucumber C-function gene, CUM1, contains only the conserved LFY site, yet exhibits the full pattern of early and late expression seen for AG (Kater et al., 1998, 2001; Hong et al., 2003). However, the fact that delayed PLE expression is also observed in plepd, which retains the LFY-binding site and the 70-bp element, shows that other elements, in addition to the LFY site, are required to confer appropriate early expression of PLE. The conservation of the LFY site within a conserved regulatory unit, and the functional similarity of the LFY site in distantly related species, points strongly to an ancient evolutionary role in regulation of the C-function.

Recently, the ‘Aphrodite’ floral mutant, which has a phenotype similar to plena, was described for lily, a non-grass moncot (Akita et al., 2008). The phenotype correlated with reduced expression of the C-function gene in the third whorl of ‘Aphrodite’, although the molecular cause of this was not determined. Given our findings, it would be interesting to examine the integrity of the intron of the lily C-function gene in the region of the conserved regulatory unit in this mutant.

In the AG intron, LFY sites, including LBS2, are generally closely associated with binding sites for the meristem maintenance factor WUS (Bomblies et al., 1999; Busch et al., 1999; Hong et al., 2003). Similarly, the conserved LFY site found in the PLE‘onset box’ is also adjacent to a WUS-like binding site. However, examination of sequences close to the conserved LFY site in other species do not reveal conserved WUS sites and we could not detect binding of ROA, the Antirrhinum orthologue of WUS (Kieffer et al., 2006), to the site. The mechanistic significance of the close association between LFY- and WUS-binding sites in the AG second intron remains to be confirmed in other species. Similarly, other potentially interesting cis-elements identified within the PLE‘onset box’ (Figure 3) do not seem to be conserved in other species.

Models for C-function regulation

Expression of the C-function is tightly regulated to ensure timely development of the reproductive organs at the correct positions within the flower. Current models suggest that C-function genes are repressed in all floral whorls and are only induced where activators antagonise the repressors at the centre of the flower (Sridhar et al., 2006; Cartolano et al., 2007). Appropriate spatial and temporal expression of the C-function therefore depends on the balance between activators and repressors through developmental time. Consistent with this, loss of a micro RNA that would normally reduce the expression of a C-function activator shifts the dynamic between regulators at the centre of the flower such that the C-function domain expands (Cartolano et al., 2007). In contrast, the failure of an activation mechanism would result in reduced C-function expression, loss of reproductive organ identity and a contraction of the C-function expression domain. Deletion of the conserved intragenic sequences in plepd and plena affects only the early regulation of PLE expression. Thus the phenotypes of these flowers provide insight into the effects of a subtle shift in the dynamic between C-function gene regulators in favour of the repressors.

Another emerging theme is that the C-function gene product must reach particular threshold levels in order to specify reproductive organ identity, and that development of stamens and carpels, as well as floral meristem termination, each require a different threshold level (Mizukami and Ma, 1995; Cartolano et al., 2007). Weak alleles of ple (ple-1) and ag (ag-4 and ag-5) all show floral determinacy defects, but less severe disruption to stamen development. Consistent with this, an allelic series of AG RNAi lines demonstrates that floral meristem termination requires a higher threshold of C-function gene product than is necessary for the specification of stamen identity (Mizukami and Ma, 1995). However, the phenotypes of plepd and plena flowers do not resemble these weak C-function alleles and AG RNAi lines. Instead, the fourth whorl shows no organ identity or determinacy defects, while stamen identity is disrupted. plepd and plena represent a novel class of heterochronic mutants, affecting the onset of PLE expression. The differences between the phenotypes observed in wild-type, weak C-function alleles and these heterochronic mutants can be explained by a threshold model (Figure 6). In plepd and plenaPLE expression is delayed to the extent that the PLE gene product does not reach sufficient levels to direct full stamen development, although levels sufficient to drive normal gynoecium development and floral determinacy are attained later. In contrast, the onset of C-function gene expression is unaffected in the weak ple, ag and AG RNAi lines, but the gene function is maintained at low levels. Such levels are sufficient to induce stamens, but not to correctly specify carpel development or to terminate the floral meristem (Figure 6). It is interesting to note that, despite its slow start, PLE expression reaches levels similar to those seen in wild-type flowers at later stages. This suggests that the balance of C-function activation and repression is subject to feedback regulation allowing the correct levels to be achieved even if the initial timing is disrupted.

Figure 6.

 An explanation for the phenotypic differences observed between weak C-function gene mutant alleles and the plena mutant.
(a) A graph predicting the expression levels of PLE in wild-type (intact line), weak alleles (dotted line) and plena (dashed line) over developmental time. In the wild type, PLE reaches sufficient levels, at the appropriate time, to specify stamens, carpels and floral determinacy. In plena, PLE expression is delayed and does not reach sufficient levels within the appropriate time frame to specify stamen development in whorl 3. Despite the delay in PLE onset, PLE accumulates at similar rates to that in the wild type, and reaches the appropriate level at the centre of the flower to specify carpels and terminate floral meristem development. In contrast, PLE is maintained at low levels throughout flower development in weak ple alleles. While this is sufficient to specify stamens, levels do not reach that required to specify carpel development or floral determinacy.
(b) Comparison of the phenotypes of wild-type, weak ple alleles, and plena flowers.

Transgenic experiments utilising reporter gene constructs suggested that the AG intron had separate domains regulating stamen development and carpel development, respectively (Deyholos and Sieburth, 2000). Since plena flowers fail to produce stamens, and plepd flowers are male fertile, our data could also be interpreted to suggest that the PLE‘onset box’ contains elements critical for PLE expression and stamen development in the third whorl. However, since PLE expression is affected in both the third and fourth whorls at early stages and because plena flowers also show carpel identity and determinacy defects at elevated temperatures (Figure 4l), we suggest that the regulation of PLE by this region is more complex than the simple model of separable whorl-specific enhancer elements proposed for the AG intron. Indeed, the heterochronic shift in PLE expression observed in the novel floral mutants plepd and plena suggests that the correct establishment of the C-function domain, and subsequent development of the reproductive organs, is a product of C-function threshold levels, and precise timing of the onset of C-function gene expression.

Experimental procedures

Plant growth conditions and mutant isolation

Antirrhinum plants were grown at 22°C under long-day conditions (16-h light/8-h dark), unless stated otherwise.

Mutant lines were generated by placing a population of plants heterozygous for the ple-625 allele (Bradley et al., 1993) at 15°C during the flowering period to stimulate transposon excision. The progeny were examined for novel floral phenotypes.

Yeast one-hybrid experiments

Yeast one-hybrid experiments were conducted using the Clontech Matchmaker one-hybrid library construction and screening kit (; 630304). Two complementary oligonucleotides, corresponding to a tandem repeat of the PLE intron ‘onset box’ sequence, were synthesized, annealed and cloned into the pHIS2 reporter vector, to generate pHIS2-ob. Complementary DNAs encoding the Antirrhinum proteins PLE, ROA and FLO were PCR amplified and cloned into the prey vector pGADT7-Rec2. Yeast strain Y187 was transformed with either pHIS2-ob or the control vector p53HIS2 to generate the bait strains, which were tested for background HIS3 expression. Bait strains were individually transformed with each prey construct, including the control bait plasmid pGAD–Rec2–53, which encodes the murine p53 protein that binds to the cis-element carried by p53HIS2. Interactions between DNA and protein were tested between all combinations of bait and prey on synthetic defined (SD) growth medium lacking histidine and containing 50 mm 3-amino-triazole. The interaction between the ‘onset box’ sequence and the prey protein was considered specific if the prey did not interact with the control cis-element in p53HIS2. Yeast growth confirming interaction between the control p53 protein and the p53HIS2 cis-element demonstrated that the assay system was operating as expected.

Accession numbers, sequence alignments and scanning for predicted cis-elements

Accession numbers or gene assignments for the C-function genes or intron sequences used in this study are as follows: Antirrhinum majus PLE (AY935269), A. majus FAR (AJ239057), Arabidopsis thaliana AG (At4g18960), Ipomoea nil DP (AB281192), tomato TAG1 (AY254705), poplar PTAG1 (AF052570), petunia pMADS3 (AB076051), cucumber CUM1 (AY254704), rice OsMADS3 (AP008207.1; nucleotide 5567050-5561700).

DNA sequence alignments were performed using ClustalW (Thompson et al., 1994), using default settings.

A number of web-based algorithms were used to predict cis-elements within the ‘onset box’ sequence. These included PLACE (; Higo et al., 1999), PlantCARE (; Lescot et al., 2002), TFBIND (; Tsunoda and Takagi, 1999), TFSEARCH ( and TESS (; Schug, 2008). All analyses were performed using default settings. Binding sites for LFY- and WUS-related proteins were identified manually based on published consensus sequences (Hong et al., 2003).

In situ hybridization

In situ hybridization was performed essentially as described by Zachgo (2002). Anti-sense DIG-labelled probes were generated from PCR-amplified gene fragments by in vitro transcription using T7 RNA polymerase, as previously described (Causier et al., 2003).


The authors wish to thank Enrico Coen, Lucy Copsey and Rosemary Carpenter (John Innes Centre, Norwich, UK) for access to mutant lines and for helpful discussions, and Zsuzsanna Schwarz-Sommer and Maria Cartolano (Max Planck Institut für Züchtungsforschung, Köln, Germany) for critical reading of the manuscript. We acknowledge the BBSRC for funding through the ERA-PG programme as part of the CisCode project.