Redefining C and D in the Petunia ABC

The petunia AGAMOUS subfamily of MADS box transcription factors contains two C- and two D-lineage genes. This work shows that the two C-genes redundantly specify stamen and carpel identity, whereas all four genes participate in ovule identity specification and floral meristem termination. These results illustrate that subfunctionalization among homeotic genes can vary considerably between species.

Abstract

According to the ABC(DE) model for flower development, C-genes are required for stamen and carpel development and floral determinacy, and D-genes were proposed to play a unique role in ovule development. Both C- and D-genes belong to the AGAMOUS (AG) subfamily of MADS box transcription factors. We show that the petunia (Petunia hybrida) C-clade genes PETUNIA MADS BOX GENE3 and FLORAL BINDING PROTEIN6 (FBP6) largely overlap in function, both in floral organ identity specification and floral determinacy, unlike the pronounced subfunctionalization observed in Arabidopsis thaliana and snapdragon (Antirrhinum majus). Some specialization has also evolved, since FBP6 plays a unique role in the development of the style and stigma. Furthermore, we show that the D-genes FBP7 and FBP11 are not essential to confer ovule identity. Instead, this function is redundantly shared among all AG members. In turn, the D-genes also participate in floral determinacy. Gain-of-function analyses suggest the presence of a posttranscriptional C-repression mechanism in petunia, most likely not existing in Arabidopsis. Finally, we show that expression maintenance of the paleoAPETALA3-type B-gene TOMATO MADS BOX GENE6 depends on the activity of C-genes. Taken together, this demonstrates considerable variation in the molecular control of floral development between eudicot species.

INTRODUCTION

While there is enormous variation in flower size, composition, and shape across the plant kingdom, one general molecular framework seems to lie at the heart of flower development. This framework is described by the ABC(DE) model, which proposes that distinct genetic functions control floral determinacy and specify the development of the floral organs in a combinatorial manner (Coen and Meyerowitz, 1991; Krizek and Fletcher, 2005).

We are systematically investigating the applicability of this model in petunia (Petunia hybrida) by reverse and forward genetics approaches. Arabidopsis thaliana and petunia belong to the rosids and asteroids, respectively, which are thought to have diverged ∼100 million years ago (Moore et al., 2010); therefore, comparison of the molecular mechanisms controlling their flower development helps to estimate the antiquity and divergence of this floral network. Previous work has shown that despite similarities in the architecture of petunia and Arabidopsis flowers, the molecular mechanisms that underlie organ identity specification in both species are not necessarily identical. For example, the B-function in petunia is more complex and includes petunia TOMATO MADS BOX GENE6 (TM6), which does not seem to fit in the basic rules of the ABC model (see below). Furthermore, to date, no A-function could be attributed to petunia orthologs of the Arabidopsis A-function genes. Instead, we found that in both petunia and snapdragon one aspect of the A-function, restriction of C-class activity to stamens and carpels, is encoded by a microRNA, named BLIND in petunia and FISTULATA in snapdragon (Antirrhinum majus; Cartolano et al., 2007).

Finally, major differences in floral architecture do exist between Arabidopsis and petunia, for example, during development of the ovary: In Arabidopsis, both the placenta and the ovules start to differentiate from the inner carpel wall, when the floral meristem has already developed completely. In petunia, as in rice (Oryza sativa), the floral meristem remains active for a longer period after emergence of the carpel primordia compared with Arabidopsis, since placenta and ovules arise directly from the inner part of the floral meristem (Colombo et al., 2008).

In this study, we focus on the petunia C- and D-lineage genes, which are involved in the development of reproductive organs, including the ovules, and in determinacy control. The C- and D-lineage genes together compose the AGAMOUS (AG) subfamily within the MADS box transcription factor family. This subfamily can be divided into two clades: the C-clade (which is further divided into two subclades; the euAG- and the PLENA-subclade) and the D-clade (Colombo et al., 1995; Kramer et al., 2004; Zahn et al., 2006).

In Arabidopsis, the C-function, required for reproductive organ development and floral determinacy, is essentially represented by one euAG-subclade gene, AG (Bowman et al., 1989; Yanofsky et al., 1990). In addition to AG, Arabidopsis contains two other C-clade genes: SHATTERPROOF1 (SHP1) and SHP2. SHP1 and SHP2 belong to the PLENA subclade of C-clade genes and function redundantly in seedpod shattering (Liljegren et al., 2000).

Petunia harbors two C-clade genes: the euAG-subclade gene PETUNIA MADS BOX GENE3 (PMADS3) and the PLENA-subclade gene FLORAL BINDING PROTEIN6 (FBP6) (Angenent et al., 1993; Tsuchimoto et al., 1993). During early development, these genes are expressed in proliferating stamen and carpel primordia, though later, expression patterns become more differentiated. PMADS3 is specifically expressed in the vascular bundles of the stamens and the pistil, in developing ovules, the epidermis of the pistil, nonsporogenic anther tissue, and the nectaries. FBP6 is more highly expressed in the transmitting tissue of the style and in the stigma (Kater et al., 1998; Kapoor et al., 2002). Overexpression and silencing experiments have already shed some light on the function of these genes (Tsuchimoto et al., 1993; Kater et al., 1998; Kapoor et al., 2002). Silencing of PMADS3 leads to a partial loss of stamen identity and slightly altered carpel morphology. In the third whorl, secondary inflorescences arise, bearing flowers with a phenotype similar to the primary flower. Ectopic expression of PMADS3 produces antheroid petals and rarely sepals with some weak carpelloid features. Loss-of-function lines for FBP6 have not been described. Overexpression of FBP6 was reported to lead only to a reduction in the size of petal limbs (Kater et al., 1998), suggesting that FBP6 does not play an important role in reproductive organ identity determination.

D-clade genes have first been described in petunia and were functionally characterized via cosuppression experiments. Cosuppression lines in which both D-clade genes FBP7 and FBP11 are knocked down show a transformation of ovules into carpel-like organs (Angenent et al., 1995), leading to the addition of a D-function to the ABC model, for ovule identity specification. Later analyses suggested that the role of FBP7 and FBP11 may not be restricted to ovule identity specification; they also play a role in ovule differentiation and seed development (Colombo et al., 1997). Arabidopsis has one D-function gene, SEEDSTICK (STK), which plays a role in seed abscission and funiculus development (Pinyopich et al., 2003), while all AG subfamily members from Arabidopsis have (partially) redundant functions in ovule development.

In the original ABC model, maintenance of B-gene expression was thought to be independent of the C-function. More recently, it was shown that in Arabidopsis, AG is required for maintenance of AP3 expression during stamen organogenesis, while in developing petals this role is fulfilled by APETALA1 (AP1) (Gómez-Mena et al., 2005). We were particularly interested to further investigate this aspect of C-dependent B-gene expression maintenance in petunia to better understand the evolutionary divergence between the two B-class genes DEFICIENS (DEF), the petunia AP3 ortholog (van der Krol et al., 1993; Vandenbussche et al., 2004), and TM6 (Rijpkema et al., 2006), a member of a more ancient B-class lineage gene that has been lost in Arabidopsis but is present in most of the eudicot species investigated so far (Kramer et al., 1998). TM6 is a paleoAP3 gene expressed, atypically for a B-function gene, in the third and fourth whorls. This gene redundantly specifies stamen identity, together with DEF, a euAP3 gene required for petal development (Rijpkema et al., 2006).

In the petunia bl mutant, the C-clade genes PMADS3 and FBP6 are ectopically expressed in the outer floral whorls, resulting in the formation of antheroid petals and, in the first whorl, sepals with rarely some weak carpelloid features (Tsuchimoto et al., 1993; Cartolano et al., 2007). In addition, these mutants show a strong increase in expression of TM6 in the perianth whorls, while DEF remains expressed at wild-type levels (Vandenbussche et al., 2004; Rijpkema et al., 2006). This suggests a more direct regulatory link between the petunia C-clade genes and TM6 than between AP3 and AG in Arabidopsis.

We specified unique and redundant functions of the C- and D-clade genes and the regulatory interactions between C- and B-function genes in petunia by analyzing a set of transposon insertion mutants, transgenic lines, and combinations thereof. As a result, we propose that FBP6 and PMADS3 function largely redundantly in both reproductive organ development and determinacy control. Both C-function genes contain cis-regulatory elements and show expression patterns consistent with these functions. Some specialization can also be observed; FBP6 has a unique function in style and stigma development. Furthermore, we find that all AG subfamily genes are involved in ovule identity specification and play a role in floral determinacy. Finally, we demonstrate that maintenance of TM6 expression is dependent upon the C-function. These data further expand our knowledge of the molecular mechanisms controlling flower development and the variations within these mechanisms.

RESULTS

Quantitative RT-PCR Shows Differential Expression of the AG Subfamily Genes in Petunia

To characterize the petunia AG genes in more detail, a quantitative expression analysis of C- and D-clade genes was performed (Figure 1A). Expression of all four AG genes is mostly restricted to the inner two whorls of developing flowers, as described previously (Tsuchimoto et al., 1993; Kater et al., 1998; Kapoor et al., 2002) The individual expression patterns show some differences, which correlate to a large extent with their functions (as presented below). While PMADS3 is expressed at similar levels within the stamens, the pistil, and ovules, the expression of FBP6 is markedly lower in the stamens and ovary compared with the pistil, in which the levels of PMADS3 and FBP6 are relatively equal. As expected for putative D-class homeotic genes, FBP7 and FBP11 are most strongly expressed in the placenta and ovules, with the expression levels of these two genes being very similar.

Figure 1.

The Petunia AG Subfamily Genes.

(A) Expression patterns of the petunia AG subfamily genes. mRNA levels in different floral organs were determined by qRT-PCR using two biological and two technical duplicates. Standard curves were created to calculate absolute numbers of mRNA molecules in each cDNA sample. These numbers were then normalized against the reference genes. Results are plotted as the ratio to the lowest expression level. a.u. absolute units. Bars indicate se.

(B) and (C) Putative LEAFY binding sites and CCAAT-boxes present in the second intron of FBP6 (B) and PMADS3 (C), as identified with the use of a position-specific scoring matrix using a cutoff value of −18. Stars indicate the position of the CCAAT-boxes.

(D) CCAAT-boxes in the second introns of different C-function genes. Asterisks indicate nucleotides conserved in all five genes. The conserved CCAAT-boxes are lined in red. TAG1, TOMATO AGAMOUS 1; TAGL1, TOMATO AGAMOUS-LIKE1.

(E) to (H) Schematic representations of the gene structures and insertion alleles of the petunia AG subfamily genes: FBP6 (E), PMADS3 (F), FBP7 (G), and FBP11 (H). Black boxes represent exons, and lines represent introns. Triangles represent dTph1 transposon insertion alleles. All gene models start at the start codon and end at the stop codon. Promoter and 5′- and 3′-untranslated regions are not shown, except in the case of PMADS3, where the two parallel lines indicate a part of the 3′ untranslated region, which was used for construction of the RNAi construct as indicated by the double arrow. Bars = 500 bp.

(I) Coding sequence and amino acid sequence of the FBP11 wild-type allele and the fbp11-1 allele, in which a 7-bp footprint creates an out-of-frame start codon, resulting in a different amino acid sequence (asterisk) and a premature stop codon (arrowhead).

Loss-of-Function Analysis of the Two Petunia C-Clade Genes FBP6 and PMADS3

By screening petunia W138 dTph1 transposon insertion libraries, we identified two unrelated individuals, each containing a dTph1 element inserted in the coding region of FBP6, at positions 222 bp (fbp6-1 allele) and 153 bp (fbp6-2 allele) downstream of the translational start, respectively (Figure 1E). Given the localization of both insertions in the N-terminal part of FBP6 and the fact that the dTph1 sequence encodes a stop codon in all reading frames and in both orientations, we consider these alleles to represent null alleles. Both homozygous insertion mutants exhibit identical phenotypes; further detailed analyses and crosses were performed using the fbp6-1 allele. fbp6 flowers are for the greater part wild-type in appearance, showing only mild developmental defects of the third- and fourth-whorl organs: Anthers can be partially transformed into petal limb-like tissue, usually arising from the connective tissue of the anther (Figure 2B), and the stigma frequently is split (Figure 2G). Upon closer examination, the pistil of fbp6 plants was shown to exhibit several developmental defects. Most commonly, the style and stigma show incomplete fusion of the inner tissues and a partial loss of transmitting tissue (Figures 2V and 2W). In addition, the stigma is partially converted into sepal- or leaf-like tissue with stomata and trichomes (Figure 2X). The severity of these aberrations is quite variable; only the apical part of the pistil may be incompletely fused, or occasionally style and stigma are completely transformed into sepal-like organs, maintaining only a wild-type ovary in the fourth whorl. Notably, the phenotype is more severe in early flowers than in later flowers and is consistently stronger when plants were growing under lower temperatures. The abnormalities of the pistil seem to severely reduce the female fertility of fbp6 mutants. Pollinations on fbp6 mutant flowers sometimes lead to seed set, but usually only small seedpods can be obtained. Male fertility of these mutants seems not to be impaired. Quantitative RT-PCR (qRT-PCR) analysis shows that the mRNA level of FBP6 is highly reduced (∼20-fold) in developing fbp6 flowers, while all other AG subfamily genes are still expressed at wild-type levels (Figure 3A).

Figure 2.

C-Clade Gene Loss-of-Function Analysis.

(A) to (E) Third-whorl organs of wild-type and C-clade gene loss-of-function lines.

(F) to (J) Fourth-whorl organs of wild-type and C-clade gene loss-of-function lines.

(K) to (O) Longitudinal sections through ovaries of wild-type and loss-of-function lines.

(A), (F), and (K) The wild type.

(B), (G), and (L) fbp6-1.

(C), (H), and (M) PMADS3-RNAi.

(D), (I), and (N) fbp6-1/⁺ PMADS3-RNAi.

(E), (J), and (O) fbp6-1 PMADS3-RNAi.

(P) Wild-type flower.

(Q) fbp6-1/⁺ PMADS3-RNAi flower.

(R) fbp6-1 PMADS3-RNAi flower.

(S), (T), (V), and (W) Longitudinal sections through of the stigma and apical part of the style ([S] and [V]) or the ovary and basal part of the style ([T] and [W]).

(S) and (W) The wild type.

(V) and (W) fbp6-1.

(U) and (X) Scanning electron micrographs of the stigmatic surface of wild-type (U) and fbp6-1 (X) flowers. Arrows in (U) and (X) indicate stigmatic papillae and trichomes, respectively.

Bar in (A) = 1 cm for (A) to (E); bar in (F) = 1 cm for (F) to (J); bar in (K) = 0.5 cm for (K) to (N); bar in (O) = 0.5 cm; bars in (U) and (X) = 100 µm.

Figure 3.

qRT-PCR Expression Analyses of the Different Petunia AG Loss-of-Function Lines.

(A) Relative expression levels of the petunia AG subfamily genes in flower buds of C-clade loss-of-function lines. WT, the wild type.

(B) Relative expression levels of the petunia AG subfamily genes in flower buds of D-clade gene mutants and of combined C⁺ D-clade loss-of-function lines.

Results are plotted as the mean value of three biological replicates ± se. In each case, the wild-type expression level is set at 100%, and other samples are normalized to this value. Asterisks indicate a significant (P < 0.05) difference between the wild-type and loss-of-function line.

To analyze the function of PMADS3, plants carrying a 35S:PMADS3-RNAi (for RNA interference) construct were generated. We identified five independent lines for which PMADS3 mRNA levels were drastically reduced compared with wild-type plants (Figure 3A). These lines displayed identical phenotypes and were retained for further morphological and molecular characterization. Plants in which PMADS3 is silenced show mild aberrations in stamen and pistil development. A deep furrow is formed in the stigma and anthers often show petal limb-like tissue, usually originating from the connective tissue, similarly as found in fbp6 mutants, but occurring more frequently and in a more pronounced fashion (Figures 2C and 2H). In addition, trichomes are sometimes growing on the carpel wall and/or epidermis of the style and ovary wall (data not shown). The ovary develops normally, and in contrast with fbp6 mutants, neither female nor male fertility is markedly affected, although the amount of pollen was frequently less than in wild-type flowers, most likely as a result of the partial anther-to-petal conversion. In PMADS3 silenced lines, we found FBP6 levels to be increased around twofold, while FBP7 levels are somewhat decreased (60%), with FBP11 expression remaining at the wild-type level (Figure 3A).

To investigate the effect of the combined loss of function of FBP6 and PMADS3, the 35S:PMADS3-RNAi construct was introduced in fbp6 mutants. Plants homozygous for fbp6 carrying the PMADS3-RNAi construct show a near-complete to complete loss of both stamen and pistil identity (Figures 2E, 2J, 2O, and 2R). In the most severely affected flowers, the anthers and stamen filaments are completely converted into petal tissue (Figure 2E). In the fourth whorl, unfused or partly fused sepals arise instead of carpels (Figures 2J and 2O). These sepals form the first whorl of a new flower, phenotypically similar to the primary flower. This process may reiterate a few more times. As expected, the expression of both C-clade genes in these flowers is severely reduced compared with the wild type. In addition, the expression level of both D-clade genes decreases drastically in these complete C loss-of-function lines (Figure 3A).

fbp6/⁺ PMADS3-RNAi plants display homeotic transformations intermediate between the single loss-of-function lines and the fbp6 PMADS3-RNAi lines. Compared with fbp6 and PMADS3-RNAi flowers, petal tissue proliferates more extensively on the stamens and the style of these flowers is shorter, greener, and thicker (Figures 2D, 2I, and 2Q) and densely covered by trichomes (data not shown). The lower part of the ovary is occupied by a wild-type placenta and normal ovules, but the upper part is frequently transformed into new floral organs (Figure 2N).

Overall, these data indicate that although some subfunctionalization has occurred between the two petunia C-clade genes, FBP6 and PMADS3 function largely redundant in specifying the identity of stamens and carpels and in flower termination. Recently, a new model has been developed to predict LEAFY binding sites based on a position-specific scoring matrix (Moyroud et al., 2011). The presence or absence of LEAFY binding sites identified by this approach appears to be helpful in predicting to what extent a C-clade gene acts as a true C-function gene. The PMADS3 second intron has already been shown to harbor LEAFY binding sites (Moyroud et al., 2011), in line with its role as C-function gene. We resequenced the PMADS3 second intron and newly sequenced the second intron of FBP6 and could identify putative LEAFY binding sites in both genes (Figures 1B and 1C), which, in the case of PMADS3, corresponds to the initial analysis (Moyroud et al., 2011). In addition, we identified conserved CCAAT-boxes (Figure 1D), which are a hallmark of AG genes playing a role in early flower development and are required for maintenance of C-gene expression in reproductive organ primordia, at least in Arabidopsis (Hong et al., 2003; Moyroud et al., 2011). As shown, the resequenced second intron of PMADS3, like FBP6, contains two fully intact CCAAT-boxes (Figure 1D), in contrast with previous results (Hong et al., 2003).

FBP6 and PMADS3 Gain-of-Function Analysis

It has been reported that ectopic expression of FBP6 only leads to smaller petal limbs (Kater et al., 1998), seemingly in contradiction with our loss-of-function analysis, which identifies FBP6 as a classical C-function gene together with PMADS3. To resolve this issue, we have regenerated FBP6 overexpression lines in petunia and Arabidopsis.

In the second whorl of 35S:FBP6 petunia flowers, the corolla is replaced by antheroid tissue, while sepals are more narrow and rigid and curl up adaxially (Figure 4C), as observed also in leaves and bracts of these plants (data not shown). Third- and fourth-whorl organs are wild-type in appearance, although smaller. Thus, overexpression of FBP6 in the flower is sufficient to phenocopy the bl mutant. Introduction of the same 35S:FBP6 construct in Arabidopsis leads to very early flowering (flowers appear after just a few leaves in the strongest lines) and results in the homeotic conversion of petals into stamens and of sepals into carpels (Figure 4E). These first-whorl organs develop stigmatic tissue and carry ovules (Figure 4F), as also observed in 35S:AG lines (Mizukami and Ma, 1992).

Figure 4.

Gain-of-Function Analysis of the Petunia C-Clade Genes in Petunia and Arabidopsis.

(A) to (C) Gain-of-function phenotypes in petunia. Floral buds of petunia V26, the wild type (A), 35S:PMADS3 (B), and 35S:FBP6 (C).

(D) to (F) Gain-of-function phenotypes in Arabidopsis. Mature flowers of Arabidopsis Columbia-0, the wild type (D), and 35S:FBP6 (E). Cryoscanning electron micrograph of Arabidopsis 35S:FBP6 (F), the arrow indicates a first whorl ovule. Bars = 1 mm in (D) and (E) and 100 µm in (F).

35S:PMADS3 plants were earlier reported to phenocopy the bl flower phenotype (Tsuchimoto et al., 1993). We obtained very similar results with our own 35S:PMADS3 lines, with flowers highly similar to those of 35S:FBP6 plants, but with the conversion of petals to anthers being slightly more pronounced (Figure 4B). Together, these observations are in agreement with the loss-of-function analysis, which identifies both FBP6 and PMADS3 as canonical C-function genes.

We also obtained weak overexpression lines for FBP6 or PMADS3 resulting only in a reduction in petal size, suggesting that in an earlier study (Kater et al., 1998), no strong overexpressors for FBP6 were obtained. Furthermore, it was reported in the same study that the transgenic traits of both FBP6 and PMADS3 overexpression lines could not be transmitted to the progeny through either the male or female parent. By contrast, we found both 35S:PMADS3 and 35S:FBP6 constructs and phenotypes to inherit in a normal Mendelian way in self-pollinations and crosses.

Regulation of AP3-Clade Genes by the C-Clade Genes

To investigate possible regulatory interactions between the petunia C-genes and the B-genes DEF and TM6, we monitored the expression of DEF and TM6 in the different C-class loss- and gain-of-function lines by qRT-PCR. Our results (Figure 5) show that DEF expression is not influenced by ectopic expression of either PMADS3 or FBP6, in line with its unmodified expression pattern in a bl mutant background (Rijpkema et al., 2006), in which both PMADS3 and FBP6 are ectopically expressed. In the single and full loss-of-C-function mutants, DEF also remains expressed at wild-type levels. By contrast, we found that TM6 is strongly upregulated in the perianth of 35S:FBP6 flowers and of 35S:PMADS3 flowers and drops drastically in the fbp6 PMADS3-RNAi lines, while wild-type levels are maintained in single loss-of-C-function lines.

Figure 5.

Expression Analysis of the Petunia AP3 Genes in C-Clade Gain- and Loss-of-Function Lines via qRT-PCR.

Results are plotted as the mean value of three biological replicates ± se. In each case, the wild-type (WT) expression level is set at 100%, and other samples are normalized to this value. Asterisks indicate a significant (P < 0.05) difference between the wild-type and loss-of-function line.

The Petunia D-Class Genes FBP7 and FBP11 Are Not Essential to Confer Ovule Identity

Multiple transposon insertion alleles have been identified for both of the D-clade genes in petunia: FBP7 and FBP11. The fbp7-1, fbp7-2, fbp11-1, and fbp11-2 alleles all disrupt the essential MADS box domain located at the N terminus of the protein and cause a premature stop codon (Figures 1G to 1I). Therefore, all alleles are likely to represent null alleles, but none of the single insertion lines displayed a mutant phenotype when homozygous. In addition, seed set and seed viability in homozygous single mutants were not different compared to wild-type plants. For detailed analysis and further crosses, the fbp7-2 and fbp11-1 alleles were used. The absence of a phenotype in the single insertion mutants suggests that FBP7 and FBP11 are completely redundant, in line with their high sequence conservation (94% overall protein similarity) and their identical expression levels and patterns (Figure 1A). We subsequently analyzed fbp7 fbp11 double mutants, in which the expression level of both mutant genes was found to be significantly lower than the wild-type expression level (Figure 3B). Surprisingly, in most fbp7 fbp11 ovaries, ovule identity is not affected. Occasionally in some ovaries, a few ovules are transformed into carpel-like structures, which were consistently situated at the top of the ovary (Figure 6D). This demonstrates that the D-lineage genes FBP7 and FBP11 are not required to confer ovule identity, in contrast with what was previously proposed based on the drastic phenotype of FBP7/11 cosuppression lines (Angenent et al., 1995). To rule out differences in genetic background between the W138 transposon line and the variety that was used for the cosuppression experiment (petunia Mitchell), we crossed both insertion alleles into the Mitchell background and reselected homozygous double mutants. Again, the majority of the ovaries contained wild-type ovules, with occasionally a few ovules that were transformed into carpel-like structures. Nevertheless, we found that female fertility was dramatically decreased in these lines. However, repeated pollinations occasionally do lead to seed set, but the seeds that developed are fewer and larger and always have a shriveled/wrinkled seed coat (Figure 6B), as observed earlier in milder FBP7/FBP11 cosuppression lines (Colombo et al., 1997). This indicates that the FBP7/11 gene pair does have a unique function in seed development that cannot be taken over by the other AG clade members.

Figure 6.

D-Clade Gene Loss-of-Function Phenotypes and Combined D- and C-Clade Loss-of-Function Phenotypes.

(A) Wild-type petunia V26 seed.

(B) fbp7-2 fbp11-1 seeds.

(C) to (F) Petunia ovaries after removal of the carpel wall.

(C) The wild type.

(D) fbp7-2 fbp11-1.

(E) fbp6-1 fbp7-2 fbp11-1.

(F) PMADS3-RNAi fbp7-2 fbp11-1.

(G) Longitudinal section of a PMADS3-RNAi fbp7-2 fbp11-1 ovary.

(H) to (K) Scanning electron micrographs of ovules and/or carpelloid organs.

(H) Wild-type ovules.

(I) fbp6-1 fbp7-2 fbp11-1 ovary showing both ovules and carpelloid organs.

(J) PMADS3-RNAi fbp7-2 fbp11-1 flower showing carpelloid organs.

(K) Detail of a carpelloid organ of a PMADS3-RNAi fbp7-2 fbp11-1 flower.

(L) to (O) Latero-longitudinal sections through ovaries at different developmental stages. Numbers 1 to 8 correspond to the following developmental stages: stage 1, ovary of a 1.2-cm bud; stage 2, 3.5-cm bud; stage 3, flower at the end of anthesis; stage 4, 3 d after pollination (DAP) or 3 d after the end of anthesis (DAA) in case of the female sterile lines or the unpollinated wild-type; stage 5, 4 DAP/DAA; stage 6, 5 DAP/DAA; stage 7, 6 DAP/DAA; stage 8, 14 DAP/DAA.

(L1) to (L8) Ovary development in unpollinated wild-type flowers.

(M1) to (M8) Ovary development in female sterile fbp6-1 fbp7-2 fbp11-1 flowers.

(N1) to (N8) Ovary development in female sterile PMADS3-RNAi fbp7-2 fbp11-1 flowers.

(O1) to (O8) Ovary development in self-pollinated wild-type flowers.

Bars = 0.1 cm in (A) and (B), 0.5 cm in (C) to (G), 100 µm in (H) and (J), 1 mm in (I), 10 µm in (K), 0.4 cm for (L) series, and 1 cm for (M) to (O) series.

Petunia C-Clade Genes FBP6 and PMADS3 Also Play Important Roles in Ovule Identity Specification

Since ovule development is not severely impaired by the simultaneous loss of the D-clade genes FBP7 and FBP11, other factors must be able to confer ovule identity in these mutants. Because of the homology and partially overlapping expression patterns within the AG subfamily (all four genes are expressed in ovules), PMADS3 and FBP6 may also function redundantly with FBP7 and FBP11 in ovule development. To test this, we separately introduced the fbp6 mutation and PMADS3-RNAi construct in the fbp7 fbp11 background. fbp6 fbp7 fbp11 plants show a variable phenotype, where usually part of the ovules (often those at the top of the ovary) is transformed into carpel-like structures, while other ovules may remain unaffected (Figures 6E and 6I).

In PMADS3-RNAi fbp7 fbp11 ovaries, we frequently observed a full transformation of all ovules to carpelloid organs (Figures 6F, 6G, and 6J). This phenotype was also observed in PMADS3-RNAi fbp7 fbp11/⁺ and PMADS3-RNAi fbp7/⁺ fbp11 plants, indicating gene dosage effects also play a role. The carpel-like structures observed in all these lines (including fbp6 fbp7 fbp11) had a broadened tip that carried stigmatic papillae (Figure 6K), as observed on wild-type carpels.

In fbp6 fbp7 fbp11 ovaries, PMADS3 remained expressed at wild-type levels (Figure 3B), while FBP6 was modestly upregulated in PMADS3-RNAi fbp7 fbp11 ovaries (Figure 3B), as also observed in single PMADS3-RNAi lines. These data explain why in both triple mutants a pistil and carpelloid organs still can be formed, since one C-clade gene is sufficient for pistil development (see above).

FBP6, PMADS3, FBP7, and FBP11 Redundantly Terminate Growth of the Ovary in the Absence of Fertilization

Throughout early development, the ovaries of PMADS3-RNAi fbp7 fbp11 plants were larger than wild-type ovaries, which can be attributed to overproliferation of placental tissue, abnormal elongation of the gynophore, and the growth of carpelloid organs instead of ovules (Figures 6N1 to 6N8), while fbp6 fbp7 fbp11 (Figures 6M1 to 6M8) ovaries were only somewhat larger compared with wild-type ovaries (Figures 6O1 to 6O8) at the end of anthesis. Both fbp6 fbp7 fbp11 and PMADS3-RNAi fbp7 fbp11 ovaries continue to grow long after anthesis ends, which normally only occurs in wild-type flowers after successful fertilization (cf. Figures 6L1 to 6L8 to Figures 6O1 to 6O8). This phenotype is not observed in the single C-gene loss-of-function lines nor in the full D-function mutant (Figures 2L, 2M, and 6D). However, we did find a similar although less drastic increase in ovary size in fbp6/⁺ PMADS3-RNAi plants (Figure 2N). There, the enlargement can also be partially attributed to an increase in placental tissue. This indicates that all four AG subfamily genes contribute to floral termination.

DISCUSSION

The C-Function in Petunia Is Redundantly Encoded by PMADS3 and FBP6

A duplication event in the AG lineage at the base of the angiosperm radiation has led to the presence of two or more AG clade genes in most higher eudicot species (Kramer et al., 2004; Zahn et al., 2006). Here, we show that the petunia C-function is encoded by two genes, PMADS3 and FBP6, in a largely redundant fashion: We observed only mild phenotypes in the single loss-of-function lines, but fbp6 PMADS3-RNAi plants display a complete loss of reproductive organ identity and of determinacy. In line with this, we identified high affinity binding sites for LEAFY in the second intron of both FBP6 and PMADS3. LEAFY is a known upregulator of AG genes, and the presence of these binding sites corresponds to a function in early flower development (Moyroud et al., 2011). Redundancy between these two genes is also reflected in their interaction partners, since yeast two-hybrid screens have shown both PMADS3 and FBP6 are able to interact with the E-function proteins FBP2, FBP5, FBP9, and FBP23 (Immink et al., 2003). However, unlike PMADS3, FBP6 is also able to interact with FBP26 (Immink et al., 2003), perhaps indicative of the unique function of FBP6 in style and stigma development or in other processes yet to be uncovered. Recently, it was shown that the rice C-function is also redundantly encoded by its two C-clade genes (Dreni et al., 2011). This contrasts with the situation in Arabidopsis and snapdragon, harboring three and two C-clade genes, respectively, where in both cases a single gene has taken up the role of the classical C-function (Bowman et al., 1989; Yanofsky et al., 1990; Bradley et al., 1993).

The fbp6 single mutant phenotype nevertheless indicates a unique role for FBP6 during style and stigma development. These observations correspond well with the expression pattern of FBP6, since previous studies have shown this gene to be relatively highly expressed in transmitting and stigmatic tissue, while PMADS3 is not (Kater et al., 1998). Recently, it has been shown that SHP1 and SHP2 in Arabidopsis play a redundant role in style and stigma development, notably in the fusion of the carpels (Colombo et al., 2010). It seems then that this function is conserved between Arabidopsis and petunia, since FBP6 is the ortholog of SHP1 and SHP2.

PMADS3-RNAi lines show generally milder phenotypes compared with fbp6 mutants, although the connective tissue of the anthers is more frequently converted into petaloid tissue. This is in agreement with the fact that PMADS3 is expressed at higher levels in the stamens than FBP6 (this work). The fact that this transformation is observed in individuals losing either of the two C-function genes could reflect a necessity for a high C-gene dosage to maintain the correct identity of this part of the stamen. In PMADS3-RNAi plants, we observed a twofold increase in FBP6 expression, which might compensate, at least partially, for the loss of PMADS3 function. A similar observation was made in snapdragon, where PLE expression is enhanced in the far mutant (Davies et al., 1999). The PMADS3-RNAi phenotype reported here is similar but not identical to the PMADS3 loss-of-function lines that have been described earlier (Kapoor et al., 2002). Most notably, we never observed secondary inflorescences arising from within the third whorl of PMADS3-RNAi flowers. The difference between these lines might be due to the fact that our lines still produce some PMADS3 transcripts, while no transcripts are detected in the lines described by Kapoor et al. Alternatively, since Kapoor et al. created these lines by introducing a transgene carrying the MADS box region of PMADS3, other MADS box genes may have been (post)transcriptionally repressed in these lines, Finally, it should be noted that different petunia lines were used in the two studies, which could also explain the differences in phenotype.

The Phenotype of 35S:PMADS3, 35S:FBP6, and bl Mutants Suggests the Presence of an Unknown C-Repressor in Petunia Flowers

According to the ABC model, ectopic C expression in the flower results in the conversion of petals to stamens and of sepals to carpels, as observed in Arabidopsis 35S:AG lines (Mizukami and Ma, 1992). In line with this, the second whorl of 35S:FBP6, 35S:PMADS3, and bl flowers becomes antheroid. However, although FBP6 and PMADS3 are expressed in the first floral whorl of our transgenic lines, as in bl mutants, sepals are not converted into carpels. Only rarely is some carpelloid tissue formed at the distal end of the sepals, as in bl flowers. This could indicate that other factors (such as protein interaction partners of the C-genes) necessary for true carpel formation are absent in the first whorl. Remarkably, heterologous ectopic expression of the cucumber (Cucumis sativus) C-class gene CUM1 in petunia and of petunia PMADS3 in tobacco (Nicotiana tabacum) lead to a complete sepal to carpel conversion (Tsuchimoto et al., 1993; Kater et al., 1998), making this hypothesis quite unlikely. Instead, we favor a model in which C-function activity is repressed posttranscriptionally in the sepals of these lines by a hitherto unknown mechanism. The complete sepal-to-carpel conversions in a heterologous context could indicate that the heterologous AG transcripts or proteins are not recognized by this mechanism. Recently, we recovered in a secondary mutagenesis screen for bl enhancers a recessive mutant called blind enhancer1, which in combination with bl leads to the full conversion of sepals to carpels, including the formation of a closed ovary containing ovules (K. Heijmans and M. Vandenbussche, unpublished data), further supporting the existence of such a repressor. Intriguingly, such a mechanism seems not to exist in Arabidopsis, since AG is perfectly capable to convert sepals in carpels when ectopically expressed, as also observed in the Arabidopsis 35S:FBP6 flowers we present here.

Maintenance of Ph-TM6 Expression Requires C-Function Activity

To investigate the involvement of C-function genes in the maintenance of B-function gene expression, we analyzed the expression patterns of DEF and TM6 in C loss- and gain-of-function contexts. We found that the expression level of DEF is not significantly altered in any of the C-function mutants or transgenics, similarly to what is observed for its ortholog AP3 in Arabidopsis (Jack et al., 1992). By contrast, we found that TM6 expression is strongly upregulated in the perianth of FBP6 and PMADS3 overexpression lines, as in bl mutants, and is strongly downregulated in fbp6 PMADS3-RNAi lines, but not in the single loss-of-function lines. This demonstrates that the activity of minimally one of two C-genes is both required and sufficient to maintain wild-type levels of TM6. Furthermore, this dependence on the C-function explains the unusual B-class expression pattern of TM6 in wild-type, globosa1 globosa2 (glo 1 glo2), and bl mutant flowers (Vandenbussche et al., 2004). Together, these data indicate that regulation of TM6 and DEF expression is fundamentally different. Gómez-Mena et al. (2005) showed earlier that maintaining AP3 expression in Arabidopsis requires, besides the presence of a GLO/PISTILLATA protein, the presence of either AG (in stamens) or AP1 (in petals). Because of the antiquity of the stamen identity program, the regulatory loop involving AG most likely represents an ancestral regulatory function of B-class MADS box genes. This implies that the extant euAP3 B-class MADS box proteins must at some moment during evolution have acquired the capacity to maintain their expression independently from the C-function genes, as one of the crucial steps needed for the origin of petals. The incapacity of the ancestral TM6 gene to maintain high expression levels outside the C-domain further supports this hypothesis.

The Petunia C- and D-Clade Genes Redundantly Specify Ovule Identity and Floral Termination

The two petunia D-clade genes FBP7 and FBP11 have been studied via cosuppression and ectopic expression (Angenent et al., 1995; Colombo et al., 1995, 1997). Based on those studies, an extension of the ABC model with a D-function was proposed, in which the D-function genes are required for ovule identity specification. We found that ovule development was largely unaffected in fbp7 fbp11 mutants, showing that ovule identity is not solely specified by the D-class genes. In the cosuppression experiment mentioned above, C-class genes may also have been partly downregulated, since cosuppression can lead to repression of homologous genes (Stam et al., 1997). This process might take place on a posttranscriptional level (Stam et al., 1997) and therefore escape detection via mRNA expression analyses. We did obtain a strong loss of ovule identity when combining the fbp7 fbp11 double mutant with either the PMADS3-RNAi line or the fbp6 mutant. The more frequent ovule-to-carpel transformation observed in the PMADS3-RNAi fbp7 fbp11 line could be due to a higher expression of PMADS3 in the ovules (this work; Kater et al., 1998), which is therefore better suited to compensate for the loss of FBP6, FBP7, and FBP11 in the triple mutant. Overall, these results demonstrate that all petunia AG members redundantly specify ovule identity, as in Arabidopsis (Pinyopich et al., 2003). This overlap in function is corroborated by the overlap in interaction partners, since in yeast two-hybrid screens both FBP7 and FBP11, like PMADS3 and FBP6, have been shown to be able to interact with the E-function proteins FBP2 and FBP5 (Immink et al., 2003). Consistent with that, we showed earlier that FBP2 and FBP5 are required for ovule identity (Vandenbussche et al., 2003), providing genetic evidence for protein–protein interactions found in yeast.

The Arabidopsis D-clade gene STK does have a unique function in funiculus development and seed abscission (Pinyopich et al., 2003). Likewise, the seed development phenotype in our fbp7 fbp11 mutants and in the weaker cosuppression lines (Colombo et al., 1997) indicates a unique function for the petunia D-clade genes during seed development. Thus, MADS13, the rice ortholog of STK and FBP7/11, seems to be so far the only D-clade gene that is absolutely required for ovule identity (Dreni et al., 2011).

Classically, the floral termination function is attributed to the combined action of C- and E-class MADS box proteins. In petunia, we found that the D-clade genes FBP7 and FBP11 also participate in floral termination, together with the C-clade genes PMADS3 and FBP6. The continuing growth of the gynophore, placenta, and carpel-like organs in fbp6 fbp7 fbp11 and PMADS3-RNAi fbp7 fbp11 lines well after anthesis is very similar to the phenotype of a SEPALLATA loss-of-function line, the fbp2 fbp5 double mutant (Vandenbussche et al., 2003), except that in the latter the ovules are replaced by leaf-like organs rather than carpels. A role for the petunia D-clade genes in determinacy control was earlier suggested by an ectopic expression experiment (Ferrario et al., 2006), in which the simultaneous constitutive expression of FBP11 and the E-function gene FBP2 led to growth arrest of seedlings at the cotyledon stage (however, the C-clade gene FBP6 was also ectopically expressed in these lines, calling into question whether meristem arrest was actually caused by D-clade function). Interestingly, an overlapping role in floral termination between D- and C-clade AG members has recently also been found in rice (Dreni et al., 2011), while the D-clade gene STK in Arabidopsis does not seem to be involved at all in determinacy control (Pinyopich et al., 2003). This might be the consequence of the different placentation topologies in these species: Petunia and rice belong to the central placentation type, with the placenta and ovules arising directly from the inner part of the floral meristem, while in Arabidopsis, the parietal placenta and ovules differentiate from the inner ovary wall, after the developing carpels have used up the floral meristem (Colombo et al., 2008). It will be interesting to find out if a floral meristem termination function for D-clade genes correlates with central placentation topology.

Mechanisms of Subfunctionalization and Specialization Events in the AG Subfamily

Thorough research in several model species has now confronted us with different scenarios of function distribution within the C-clade genes. In snapdragon, the differential ability of FARINELLI (FAR) and PLENA (PLE) to promote male or female development is a clear example of subfunctionalization at the protein level (Airoldi et al., 2010). In Arabidopsis, the specialized function of the SHP genes is at least in part due to a divergent expression pattern, since SHP genes are capable of maintaining ectopic carpel formation in ap2 mutants in the absence of AG (Pinyopich et al., 2003). In petunia, we determined a unique function for FBP6 in style and stigma development. The divergent expression patterns of PMADS3 and FBP6 in these tissues (Kater et al., 1998) combined with the similarity in FBP6 and PMADS3 gain-of-function phenotypes suggest that this specialized function is mainly due to a divergence in the expression pattern of the two C-clade genes.

In conclusion, gene duplications within the AG subfamily of MADS box genes have given rise to different patterns of redundancy and specialization (and most notably subfunctionalization) in different species, the only apparent constraint being that all necessary functions (stamen and carpel development, determinacy control, and ovule development) are performed by this family as a whole, underlining both the constraints and freedom inherent in the course of evolution.

METHODS

Plant Material and Genotyping

Petunia (Petunia hybrida) plants were grown under standard greenhouse conditions (16 h day/8 h night artificial lighting, 20°C minimally) and further influenced by local seasonal changes (51°82′ north; 5°87′ east). The fbp7-1; fbp7-2 and fbp11-1 alleles were isolated in a general screen for insertions in members of the petunia MADS box gene family (Vandenbussche et al., 2003). Since the original fbp11-1 allele contains a dTph1 transposon 2 bp before the start codon and, therefore, in theory might not fully disrupt gene function, we used a line in which the transposon has excised, leaving behind a 7-bp footprint and resulting in a rearrangement disrupting the original reading frame (Figure 1I). In this study, fbp11-1 always refers to this footprint allele. The fbp6-1, fbp6-2, and fbp11-2 alleles were identified by homology search in our transposon flanking sequence collections (Vandenbussche et al., 2008). Exact insert positions were confirmed by sequencing of homozygous mutant insertion loci (Figures 1E to 1H). Plants were genotyped by PCR using primers flanking the insertion site (see Supplemental Table 1 online). Segregation of the transgenic constructs was followed via standard PCR reactions with primers targeting the kanamycin resistance gene and using primers targeting the LR recombination sites (attb1/attb2).

Transgene Construction

For the PMADS3-RNAi construct, a 313-bp nonconserved fragment of the 3′ untranslated region of PMADS3 was isolated, whereas the FBP6 and PMADS3 overexpression constructs contained the full coding sequence of both genes. After their initial isolation, these fragments were introduced into the entry vector pDONR221 (Invitrogen) via the BP clonase (Invitrogen) recombination reaction as described by the manufacturer. The PMADS3-RNAi vector was then created by transferring the RNAi fragment from pDONR221 to the pKGWIW2(I) binary RNAi vector (Karimi et al., 2002) via the LR clonase (Invitrogen) recombination reaction, again performed as described by the manufacturer. The PMADS3 and FBP6 overexpression constructs were created in a similar fashion but using the binary pK2GW7 overexpression vector (Karimi et al., 2002). Both vectors use the cauliflower mosaic virus 35S promoter. Agrobacterium tumefaciens strain EHA105 was transformed with these constructs using the freeze-thaw method (Chen et al., 1994). For the PMADS3-RNAi construct, leaf discs of petunia fbp6-1/⁺ individuals (resulting from a cross between the fbp6-1 mutant and the easily transformable Petunia Mitchell variety) were transformed as described (Horsch et al., 1985). The overexpression constructs were transformed to petunia V26 wild-type lines using the same protocol. The single PMADS3-RNAi lines were obtained by outcrossing of the PMADS3-RNAi fbp6-1/⁺ plants with wild-type mothers. These progenies were then selfed to obtain families segregating for the PMADS3-RNAi construct. The 35S:FBP6 construct was introduced into Arabidopsis thaliana Columbia-0 using the floral dip method (Clough and Bent, 1998).

Quantitative Real-Time PCR Analysis

For analysis of the expression of PMADS3, FBP6, FBP7, and FBP11 in different wild-type plant tissues, total RNA was isolated from sepals, petals, stamens, and carpels of flower buds (petal length of 7 mm) and from the placenta carrying the ovules isolated from flowers at the end of anthesis. Every organ type was sampled in duplicate. For the examination of the petunia DEF, TM6, PMADS3, FBP6, FBP7, and FBP11 mRNA levels in the different transposon insertion mutant and transgenic lines, typically RNA of small floral buds (∼5 mm long from receptacle to stigma tip, without sepals) from three different representative plants was isolated. In case of the overexpression lines, RNA was isolated from sepals and petals of similarly sized buds from three independent individuals, RNA was extracted using TRIzol reagent (Invitrogen) according to the manufacturer’s instructions, and cDNA synthesis was performed as described (Rijpkema et al., 2006) cDNA-specific primer pairs were designed with the help of Beacon Designer 4 software (PREMIER Biosoft International). All qRT-PCR reactions were performed in a Bio-Rad MyIQ iCycler, using the following two-step PCR profile: 95°C for 3 min, followed by 40 cycles of 95°C for 15 s and 58°C for 1 min. Afterwards, the reaction mixtures were subjected to the machine’s standard melt curve analysis. Expression levels were corrected for PCR efficiencies for the analysis of expression in wild-type tissue series only and normalized to ACTIN and GAPDH expression levels as described (Livak and Schmittgen, 2001). To be able to compare expression levels between different genes in different tissues from wild-type plants, each primer set was used for calibration using a known number of template molecules (via introduction of the cDNA in a plasmid) in serial dilution. Relative expression levels in wild-type tissues or the lowest value (as indicated in each figure) were set at 1 or 100%, and other values were related to this value. Error bars represent the se of three biological replicates. The Student’s t test was used to test for statistical significance. In each case, the three wild-type values were compared with the three experimental values in a two-tailed t test assuming equal variance. Asterisks indicate P values < 0.05.

Microscopy

Samples for cryoscanning electron microscopy were analyzed as described previously (Vandenbussche et al., 2009). Pistils and ovaries shown in Figures 2S, 2T, 2V, and 2W were fixed in Carnoy’s solution II 1:3:6 glacial acetic acid:chloroform:ethanol for 12 h at 4°C, dehydrated in ethanol, and embedded in Technovit 7100. Sections of 1 and 5 µm for Technovit embedded tissue were stained with 0.1% Toluidine blue in 1% borax. Sections were viewed and photographed with a Leitz Orthoplan microscope, equipped with a Leica DFC 420C camera.

Isolation of the FBP6 and PMADS3 Regulatory Intron and Scoring of LEAFY Binding Sites

After cloning and sequencing, the complete FBP6 and PMADS3 second intron sequences were searched for potential LEAFY binding sites as described at http://biodev.cea.fr/morpheus and using the previously constructed position-specific scoring matrix (Moyroud et al., 2011). A cutoff score of −18 was used to select only for high confidence LEAFY binding sites based on previous results and confirmed targets (Moyroud et al., 2011; Winter et al., 2011).

Oligonucleotide Primers

All primers used in this study are listed in Supplemental Table 1 online.

Accession Numbers

Sequence data from this article can be found in the GenBank/EMBL data libraries under the following accession numbers: FBP6 second intron (JQ073292) and PMADS3 second intron (JQ073293).

Supplemental Data

The following materials are available in the online version of this article.

• Supplemental Table 1. List of Primers Used in This Study.

Glossary

qRT-PCR

quantitative RT-PCR