The Duplicated B-Class Heterodimer Model: Whorl-Specific Effects and Complex Genetic Interactions in Petunia hybrida Flower Development

Phylogenetic Analysis

The phylogenetic relationship between the P. hybrida, Antirrhinum, and Arabidopsis B-class MADS box genes is represented by a neighbor-joining tree (Figure 1), which was constructed by comparing the MIK domain of these sequences, including a selection of sequences from other taxonomic groups. The FBP24 gene, a P. hybrida member of the B_sister subfamily (Becker et al., 2002), was used as an outgroup. For a more comprehensive overview of B-class MADS box genes, we refer to Kramer et al. (1998), Kramer and Irish (2000), and Vandenbussche et al. (2003a). The phylogenetic analysis indicates that in P. hybrida, gene duplications have occurred in both the GLO/PI and DEF/AP3 lineages. To facilitate mental comparisons between the different systems, we have decided to rename the GP, FBP1, and PMADS2 genes to PhDEF, PhGLO1, and PhGLO2, respectively, after the founding members of the different subfamilies; PhTM6 was previously named after TM6, the Lycopersicon esculentum (tomato) paleoAP3 representative (Kramer and Irish, 2000). PhGLO1 and PhGLO2 are more closely related to GLO than to PI, which is in accordance with the taxonomic distribution of the species involved; P. hybrida and A. majus belong to the Asteridae, whereas Arabidopsis belongs to the Rosidae. Likewise, PhDEF displays a closer phylogenetic relationship to DEF than to AP3, whereas the second P. hybrida DEF/AP3 member, PhTM6, belongs to the so-called TM6 lineage, which forms a distinct clade within the DEF/AP3 subfamily (Kramer and Irish, 2000).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Neighbor-Joining Tree of B-Class MADS Box Genes from P. hybrida, Arabidopsis, A. majus, and a Selection of Other Species.

Species names are indicated as follows: Am, A. majus; At, A. thaliana; Cm, Chrysanthemum x morifolium; Cs, Chloranthus spicatus; De, Dicentra eximia; Hp, Hieracium piloselloides; Hm, Hydrangea macrophylla; Md, Malus x domestica; Le, L. esculentum; Lr, Lilium regale; Os, O. sativa; Ph, P. hybrida (shaded); Pn, Papaver nudicaule; Sc, Sanguinaria canadensis; and Zm, Z. mays. We renamed three of the four P. hybrida putative B-function proteins; old names are shown in between brackets. The tree was rooted with FBP24, a P. hybrida member of the B_sister (B_s) MADS box subfamily (Becker et al., 2000). Altogether, 1000 bootstrap samples were generated to assess support for the inferred relationships. Local bootstrap probabilities (in percentages) are indicated near the branching points.

Analysis of phglo1 and phglo2 Single Mutants

To further characterize the B-function in P. hybrida, we have screened for transposon insertions into the PhGLO1, PhGLO2, and PhTM6 genes; phdef alleles were already available (de Vlaming et al., 1984; van der Krol et al., 1993). In this screening, we identified transposon insertions in two of our three primary targets, PhGLO1 and PhGLO2. The phglo1-1 and phglo1-2 alleles (Table 1) were isolated in a forward genetics manner (Van den Broeck et al., 1998; Vandenbussche et al., 2003b; M. Sauer, unpublished data) and contain a dTph1 insertion in the C-terminal domain and a dTph8 insertion in the K region, respectively. Flowers of plants homozygous for the phglo1-1 or phglo1-2 allele display an identical phenotype: the midveins of the petals become broader and greener, especially toward the edge of the corolla at the abaxial side of the petals (Figures 2A and 2B). Scanning electron microscopy of these regions revealed a conversion of the typical conical petal epidermal cells into sepal-like epidermal cells, the presence of stomata, and the development of trichomes, suggesting a shift of petal toward sepal identity in these regions (Figures 2C to 2E). In wild-type P. hybrida flowers, stamen filaments are fused partly with the corolla tube. By contrast, stamen filaments of phglo1 mutants are not fused (Figure 2F). Mutants for PhGLO1 thus exhibit a very partial B-mutant phenotype.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Phenotypes Observed in phglo1 Flowers and in the F2 Progenies of the phglo1 phglo2, phdef phglo1, and phdef phglo2 Crosses.

(A) to (F) Phenotypic analysis of phglo1-2 mutants.

(A) Abaxial side of wild-type (left) and phglo1-2 (right) petals.

(B) Close-up of a phglo1-2 petal showing green tissue and trichomes.

(C) to (E) Scanning electron microscopy images of the adaxial epidermis of wild-type petals (C) showing the characteristic conical petal cells; phglo1-2 petals (D) near a midvein at the edge of the corolla showing sepal-like epidermal cells, trichomes and stomata (arrows); and wild-type sepal (E) with typical sepal epidermis cells, trichomes, and stomata. Scale bars = 100 μm for (C) to (E).

(F) phglo1-2 flower (left) with freestanding stamen filaments and wild-type flower (right) with part of the stamen filaments fused to the corolla tube (indicated by a white brace); corolla and sepals have been removed partially to reveal the inner organs.

(G) Flower of a phglo1 phglo2 double mutant displaying full homeotic conversion of petals to sepals in the second whorl and of stamens to carpels in the third whorl, which are fused in a pentalocular gynoecium. Development of the central pistil is reduced.

(H) PhGLO1phglo1 phglo2 flower showing second-whorl organs with normal petal identity, whereas stamens are converted to multiple carpelloid organs. Development of the central pistil appears as wild type.

(I) and (J) phglo1 PhGLO2phglo2 flowers showing petals with a more pronounced petal to sepal conversion along the midveins compared with phglo1 mutants (I), stamens carrying stigmatic tissue on top of the anthers (J), and one stamen that has been fully converted to a wild type–appearing pistil (indicated with an arrow). Sepals and petals have been removed to reveal the inner organs.

(K) Flower of a phdef mutant displaying full conversion of petals to sepals, whereas the anthers remain unaffected (inset).

(L) and (M) PhGLO1phglo1 phglo2 flowers showing enlargement (L) of the petal tube base (right) compared with the wild type (left), sepals have been removed partially; third-whorl organs (M) consisting of multiple carpelloid organs and the presence of placental tissue carrying ovules without being encapsulated in an ovary. Extra petal tissue (red) is emerging between second- and third-whorl organs. First- and second-whorl organs have been removed.

(N) phdef phglo1 flower with anthers carrying small stigmas on top (inset).

(O) Flower of a PhDEFphdef phglo1 mutant with pronounced sepal-like petal midveins and asymmetrically reduced development of petal tissue in between.

(P) phdef phglo2 flower displaying full conversion of petals to sepals and of stamens to carpels, which are fused in a central gynoecium. Two second-whorl sepals have been removed to show the central carpelloid structure densely covered with trichomes. Inset, top view of the central gynoecium.

(Q) Abaxial (top row) and side views (bottom row) of B-function gene mutant combinations, showing variations in corolla and tube structure. From left to right: PhDEFphdef phglo1, phglo1, phglo1 PhGLO2phglo2, and PhGLO1phglo1 phglo2.

The phglo2-1, phglo2-2, and phglo2-3 insertion alleles (Table 1) were identified in a reverse genetics screening (Vandenbussche et al., 2003b). Flowers of plants homozygous for any of these three insertion alleles appear morphologically as wild type (data not shown). However, we occasionally observed flowers of homozygous mutants with anthers that failed to dehisce properly. These anthers shrivel and become brownish at the time mature pollen grains are released from wild-type anthers. Because of the low penetration of the phenotype, we have not analyzed this phenomenon in detail.

To further elucidate the genetic interactions among these genes, we have selected all possible double mutant combinations using the available mutant alleles (Table 1).

phglo1 phglo2 Double Mutant Analysis

Besides plants displaying a wild-type or a phglo1 phenotype, three new phenotypic classes were encountered in the F2 progeny of the phglo1 × phglo2 cross. A first class consisted of plants that exhibit a complete conversion of petals to sepals and of stamens to carpels (Figure 2G). The five carpelloid organs in the third whorl are fused and form a central tubular structure. Development of the genuine bilocular gynoecium formed in the fourth whorl of the wild-type flower often is strongly reduced. Molecular analysis indicated these plants to be the expected homozygous double mutants. Because phglo1 and phglo2 single mutants both develop petals and stamens, this indicates that at least one of the two GLO/PI genes must be active to enable B-function activity in the flower. In a second class, an unexpected phenotype was encountered: plants exhibiting a qualitative change of anthers into carpels, whereas petal identity remained largely unaffected (Figure 2H). Moreover, the base of the petal tube in these flowers was considerably enlarged, and petal tube length and size of the corolla were reduced compared with wild-type and phglo1 single mutant flowers (Figures 2L and 2Q). Furthermore, there is a strong increase in third-whorl organ number, mainly consisting of numerous carpelloid structures, although rudimentary petaloid structures develop regularly in between these organs and the second-whorl petals. Often, ovules and placental tissue develop on and in between these third-whorl organs without being encapsulated in an ovary (Figure 2M). In contrast with phglo1 phglo2 double mutants, the wild-type bilocular gynoecium develops normally in plants of this class, although it can be partially fused with carpelloid tissue from the third whorl (Figure 2H). All analyzed plants belonging to this class turned out to be homozygous mutant for phglo2 and heterozygous for phglo1. A third class consisted of plants displaying a more pronounced phenotype than the subtle phglo1 single mutants (Figures 2I and 2Q). The partial conversion of petal to sepaloid tissue along the petal midveins is more outspoken, and in the majority of the flowers, stigmatic tissue develops on top of the anthers in the third whorl. Occasionally, a stamen was fully converted into a carpel (Figure 2J). Longitudinal growth of the petal tube and corolla is reduced compared with wild-type and single phglo1 mutants (Figure 2Q). Genotype analyses revealed that all tested plants of this class were homozygous mutant for phglo1 and heterozygous for phglo2. Remarkably, we thus found gene dosage effects for both the PhGLO1phglo1 phglo2 and the phglo1 PhGLO2phglo2 genotypes.

phdef phglo1 Double Mutant Analysis

Apart from plants displaying the phdef mutant phenotype (Figure 2K), the phglo1 mutant phenotype (Figure 2A), or a wild-type phenotype, two additional phenotypic classes were observed in the F2 progeny. First, we noticed plants that look very similar to phdef single mutants, except that they systematically develop stigmatoid tissue on top of the anthers (Figure 2N). Genotyping confirmed that these plants were phdef phglo1 double mutants. The second group consisted of plants with flowers displaying a strongly enhanced phglo1 phenotype, resulting in a particularly attractive new type of P. hybrida flower (Figures 2O and 2Q). The partial conversion of petals to sepals is more outspoken than in phglo1 single mutants, especially along the petal midveins. The corolla tissue in between the enlarged green midveins displays full petal identity, although the total corolla surface is considerably reduced compared with phglo1 single mutants. All of these plants appeared homozygous mutant for phglo1 while heterozygous for phdef, again indicating the occurrence of gene dosage effects.

phdef phglo2 Double Mutant Analysis

In the F2 progeny of this cross, two phenotypes were encountered, besides wild type–appearing plants and plants with the phdef phenotype. One class consisted of plants with flowers exhibiting a complete conversion of petals to sepals and of anthers to carpels (Figure 2P). Genotyping revealed that these plants were phdef phglo2 double mutants. The phenotype of these plants is strikingly similar to that of phglo1 phglo2 double mutants, although in general, the carpelloid tubes of the latter were more regularly organized than in phdef phglo2 mutants (cf. Figure 2G with 2P, inset). In both cases, the outer wall of the central carpelloid tube is densely covered with trichomes, which are never found in the inner two whorls of wild-type flowers, and development of the fourth-whorl pistil is often strongly reduced. These phenotypic effects are very similar to what has been observed in def and glo mutants in Antirrhinum (Sommer et al., 1990; Trobner et al., 1992). The phenotype of phdef phglo2 flowers further clearly demonstrates that PhDEF is involved in anther development, which has remained elusive so far because phdef single mutants exhibit normal anthers. Plants of the second class (data not shown) displayed a phenotype similar to the one shown in Figure 2N: flowers that look like phdef single mutants, except that they frequently develop stigmatic tissue on top of the anthers. Such plants proved to be homozygous for phdef and heterozygous for phglo2.

phglo1 bl, phglo2 bl, and phdef bl Double Mutant Analysis

To test whether the second-whorl defects observed in phglo1 and phdef flowers are independent from organ identity, we have crossed these mutants with the P. hybrida A-function mutant blind (bl) (Vallade et al., 1987). Because of ectopic expression of the C-function gene PMADS3 in whorls 1 and 2 (Tsuchimoto et al., 1993; Kater et al., 1998), bl flowers display homeotic conversion of the corolla limb into antheroid structures in the second whorl (Figure 3C) and under certain conditions of mainly the tips of the sepals into carpelloid tissue in the first whorl.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Flower Phenotypes of bl Single and phglo1 bl, phdef bl, and phglo2 bl Double Mutants.

(A) and (B) Young and full-grown phglo1 bl flowers, respectively, displaying second-whorl organs consisting of a tube carrying antheroid structures with stigmas on top (inset).

(C) Full-grown bl flower with second-whorl organs terminating with antheroid structures.

(D) and (E) Flowers of young and full-grown phdef bl double mutants, respectively, showing the presence of second-whorl organs consisting of antheroid structures with short style–stigma structures on top (top inset). Third-whorl anthers appear as wild type. Two sepals in (D) and all sepals in (E) have been removed. Bottom inset, full-grown phdef bl flower with a short greenish tube.

(F) phglo2 bl flower, phenotypically identical to bl single mutant flowers. The arrow indicates one of the sepal tips that is fully converted to a style–stigma structure. This aspect of the phenotype of bl mutant flowers is only rarely observed.

We have shown that the phglo1 mutation affects petal development only locally in the petal midveins at the edge of the corolla. A very similar local homeotic change occurs in phglo1 bl flowers: the second-whorl antheroid organs carry small style–stigma structures on top (Figures 2B and 3A), indicating an absence of B-function activity only in the distal ends of these organs. Therefore, we conclude that the subtle second-whorl phenotype of phglo1 is independent from organ identity. As expected, phglo2 bl flowers are indistinguishable from those of bl single mutants (cf. Figures 3C and 3F) because phglo2 single mutants do not exhibit homeotic conversions. By contrast, although phdef flowers display a full homeotic conversion of petals to sepals, the effect of the phdef mutation on the development of the second-whorl antheroid organs in bl flowers is much less severe (Figures 3D and 3E). Instead, these second-whorl organs resemble those of phglo1 bl (Figures 3A and 3B), exhibiting antheroid organs carrying small style–stigma structures on top. Furthermore, the tube often is replaced by five separate filamentous structures, whereas if present, the tube usually is more greenish, and its longitudinal outgrowth remains rather restricted (Figure 3E, inset). The same phenotype was obtained using another phdef allele (Table 1, gp [PLV]) crossed with the bl mutant (Tsuchimoto et al., 2000). These results are quite surprising because phdef single mutants have second-whorl organs that are fully converted to sepals, suggesting a complete absence of B-function activity in the second whorl. Therefore, because C-function genes are ectopically expressed in the second whorl of bl flowers, one would expect that the second-whorl organs of phdef bl double mutants would entirely consist of carpelloid organs rather than the observed antheroid structures carrying small style–stigma structures on top. This phenotype suggests that P. hybrida contains a B-class gene, presumably a DEF/AP3 homolog, which can complement the phdef mutation regarding antheroid formation in the second whorl of phdef bl double mutants but that is unable to complement petal development in the second whorl of phdef single mutants. Most likely, this postulated B-class gene represents the same gene that is responsible for the rescue of stamen development in the third whorl of phdef single mutants. PhTM6, a P. hybrida DEF/AP3 homolog, is the most logical candidate to represent this gene (Figure 1), although we cannot at this point formally exclude the possibility that another yet unknown candidate gene might be involved. To further investigate this, we have included the bl single and phdef bl double mutants in the expression analysis of the B-class genes described below.

RT-PCR Expression Analysis in Wild-Type and Mutant Floral Organs

To further elucidate regulatory interactions between the four P. hybrida B-class MADS box genes and to correlate these with the observed phenotypes, we have monitored their expression levels in the floral whorls of the wild type and a selection of informative mutant combinations by reverse transcription (RT)–PCR.

In wild-type P. hybrida flowers (Figure 4A), the expression domain of the two PI homologs PhGLO1 and PhGLO2 and the euAP3 homolog PhDEF is mainly confined to the second and third whorls, as described previously (Angenent et al., 1992; van der Krol et al., 1993), and signals are slightly stronger in 0.5-cm buds compared with the later stage. Low levels of PhDEF expression were also detected in the first and fourth whorls, confirming earlier observations (Tsuchimoto et al., 2000), which has also been reported for DEF in A. majus (Schwarz-Sommer et al., 1992). The expression patterns of PhGLO1, PhGLO2, and PhDEF are thus very similar to their counterparts in Arabidopsis and A. majus. By contrast, the expression of PhTM6 (a paleoAP3-type gene) differs drastically in this respect. Highest signals for PhTM6 transcripts in 0.5-cm buds are detected in carpels and stamens, whereas expression levels observed in petals and sepals are much lower. In 4- to 5-cm buds, PhTM6 levels remain high in the fourth whorl while declining in the stamens. Notably, L. esculentum TM6 (another paleoAP3-type gene) transcripts also accumulate to high levels in the center of the flower (Pnueli et al., 1991).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Expression Analysis of the Four P. hybrida B-Class MADS Box Genes in Floral Organs of Wild-Type and Various Mutant P. hybrida Flowers as Determined by RT-PCR.

(A) Expression in wild-type flowers.

(B) Expression in flowers of a selection of informative B-class mutant combinations.

(C) B-class gene expression in flowers of an A-function mutant (bl) and an AB-function double mutant (phdef bl).

Ca, carpel; Pe, petal; Se, sepal; St, stamen. Homeotically converted organs are shown in bold and underlined. Whorl numbers are indicated below the gel images. Expression of glyceraldehyde-3-phosphate dehydrogenase was monitored as a positive control. a, size indications of the flower buds reflect petal length; b, sepals harvested from bl and phdef bl flowers have been enriched for having a strong penetration of the bl first-whorl phenotype, consisting of sepals carrying stigmatoid tissue on the tips; c, second-whorl tissue harvested from bl and phdef bl flowers consists of antheroid structures and antheroid structures topped with style–stigma structures, respectively; d, the gene-specific primer pair designed to monitor PhGLO1 expression flanks the insertion site of the phglo1-2 footprint allele. This results in the amplification of a slightly larger fragment than from samples containing the wild-type allele.

In the various B-mutant combinations (Figure 4B), the expression patterns of PhGLO1, PhGLO2, and PhDEF are highly consistent with the observed phenotypes; in all fully converted organs in the different B-class mutant combinations, the expression levels of these genes are either below detection limit or significantly reduced. By contrast, the expression level of PhTM6 in phdef, phglo1 phglo2, phdef phglo2, and PhGLO1phglo1 phglo2 does not differ from the wild type, indicating that its expression at the transcriptional level is not regulated by PhDEF, PhGLO1, or PhGLO2.

In flowers of the A-function bl mutant (Figure 4C), expression patterns of PhGLO1, PhGLO2, and PhDEF appear as in the wild type, mainly confined to second- and third-whorl organs, as described previously (Angenent et al., 1992, Tsuchimoto et al., 1993). By contrast, PhTM6 expression levels in a bl mutant background are clearly upregulated in second-whorl organs and to a lesser extent in first-whorl organs as compared with the wild type. This is quite surprising because in the original ABC model of flower development, high expression levels of B-class MADS box genes have been proposed to be confined to the second and third whorls only, independently from A- and C-function activity, as we observe for PhGLO1, PhGLO2, and PhDEF. In flowers of phdef bl double mutants, expression levels of PhGLO1 and PhGLO2 in the third whorl and in the second whorl are comparable to the wild type, confirming earlier observations (Tsuchimoto et al., 2000). This differs from what is observed in phdef flowers, where PhGLO1 and PhGLO2 expression levels are reduced in the second-whorl sepaloids (Figure 4B), indicating that PhDEF is required for the upregulation of PhGLO1 and PhGLO2 in the second whorl in wild-type flowers but not in the second whorl in bl flowers. Similar to what we observe in bl single mutants, PhTM6 is mainly expressed in whorls 2, 3, and 4 and at slightly lower levels in the first whorl, indicating that the absence or presence of PhDEF transcripts does not influence the upregulation of PhTM6 in the second and first whorls of bl flowers. Strikingly, the PhTM6 expression pattern in wild-type, bl, and phdef bl flowers offers a very logical explanation for the unexpected second-whorl phenotype of phdef bl mutants (see Discussion).

In Situ Hybridization

To further analyze the expression patterns of the P. hybrida B-class MADS box genes during the early stages of flower development, we have monitored the spatial and temporal expression patterns of PhTM6 and PhGLO2 by in situ hybridization (Figure 5). Expression patterns of PhGLO1 and PhDEF during early flower development have been documented previously (Angenent et al., 1995).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

In Situ Localization of PhGLO2 and PhTM6 Transcripts in P. hybrida Flower Buds.

Sections were hybridized with digoxigenin-labeled antisense RNA probes of PhGLO2 ([A] to [C]) and PhTM6 ([D] to [I]). Cross-sections of large flower buds are shown in (C), (F), and (I). All others are longitudinal sections. The first two columns show sections of wild-type P. hybrida flower buds; the last column consists of sections of bl mutant flower buds. First- to fourth-whorl organs are indicated by the according numbers. The red color reflects the presence of transcripts.

(A) Young wild-type flower bud with developing sepals and petals, stamen, and carpel primordia. PhGLO2 expression is detected in petal and stamen primordi, but not in sepals and in the center of the flower.

(B) and (C) PhGLO2 expression remains localized in petals and stamens during further development.

(D) Young wild-type floral bud at a comparable stage as in (A). In contrast with PhGLO2, highest levels of PhTM6 mRNA are present in stamen primordia and in the center of the flower, whereas low levels are detected in petal primordia.

(E) and (F) PhTM6 expression during stages as in (B) and (C). Highest expression levels are found in the developing pistil, especially in the tissue that will give rise to the placenta and ovules and in the stamens. Very low PhTM6 levels can be observed in the developing petals.

(G) Young bl mutant flower bud at a similar developmental stage as in (A) and (D). The three inner whorls show comparable PHTM6 expression levels, in contrast with PhTM6 expression in the wild type.

(H) PhTM6 expression in a bl floral bud at a similar developmental stage as in (B) and (E). The upregulation of PhTM6 expression in second-whorl organs is very pronounced compared with the wild type. Note that second-whorl organs start to enlarge according to the adaxial-abaxial axis, which finally will result in the formation of a tube terminating in five antheroid structures.

(I) bl floral bud at a slightly later developmental stage compared with (C) and (F), which allows capturing sections through the anthers and the well-developed second-whorl antheroid organs in the same plane. PhTM6 mRNA accumulates in the developing ovules and placental tissue, in the third-whorl anthers, and in the second-whorl antheroid structures.

Scale bars = 100 μm in (A), (D), and (G); 200 μm in (B), (C), (E), (F), (H), and (I).

PhGLO2 transcripts are first detected in the very early stamen and petal primordia, where transcripts are initially uniformly distributed (Figure 5A) and remain localized in these organs during further development (Figures 5B and 5C). The PhGLO2 expression pattern during early development is thus very similar to PhGLO1 (Angenent et al., 1995). To illustrate PhTM6 expression in comparison with PhGLO2, we have selected similar developmental stages of wild-type and bl flower buds. In early stage wild-type buds with emerging stamen and petal primordia (Figure 5D), highest levels of PhTM6 expression are detected in stamen primordia and in the center of the flower bud. Signal is also present in the emerging petal primordia but at a lower level compared with third and fourth whorls. During later stages of development (Figures 5E and 5F), expression in the petals drops further in comparison with stamens and pistil, where high expression levels are maintained. During development of the pistil, PhTM6 expression is mainly confined to the cells that give rise to the placental tissue and ovules. In bl flowers, PhTM6 expression levels and localization in third- and fourth-whorl organs are comparable to wild-type flowers throughout development (Figures 5G to 5I). By contrast, expression in the developing second-whorl organs is clearly upregulated and during later stages accumulates to levels comparable to third-whorl expression, even before the antheroid nature of the bl second-whorl organs becomes apparent (Figures 5G and 5H). Note that a weak signal can also be seen in the left sepal tip of the bud in Figure 5G. At a slightly later stage, the distal ends of the bl second-whorl organs adapt an antheroid morphology, which is clearly visible in the cross-section shown in Figure 5I. At this stage, PhTM6 expression in the second-whorl organs is mainly localized in the developing antheroid tissues. These results are in full agreement with the RT-PCR data.

P. hybrida DEF/GLO Heterodimer Formation: Promiscuity and Monogamy

It was demonstrated previously that PHDEF is able to interact with both PHGLO1 and PHGLO2 (Immink et al., 2003), indicating that heterodimer formation between DEF/AP3 and GLO/PI proteins is conserved in P. hybrida. We have further investigated protein–protein interactions among the four P. hybrida B-class proteins using the yeast GAL4 two-hybrid system. Growth of yeast colonies on selective media was scored after 7 d of incubation at 20°C (Table 2). Interactions could be demonstrated between PHDEF and both PHGLO1 and PHGLO2, confirming previous results. In contrast with PHDEF, our data indicate that PHTM6 is only capable of interacting with PHGLO2 in yeast. However, after a prolonged incubation of 10 d, limited yeast growth could also be detected in the PHTM6/PHGLO1 combination (but not in the PHTM6/PHDEF combination and in the negative control), suggesting that PHTM6 and PHGLO1 might also be capable of interacting, although very weakly. Nevertheless, these results suggest that PHGLO2 might be the preferred interaction partner of PHTM6, although we cannot exclude the possibility that in vivo additional factors are present, which may stabilize PHTM6/PHGLO1 heterodimers.

Duplicated B-Class Genes May Diverge in Function

Whereas in both A. majus and Arabidopsis a single pair of genes interacts to define the developmental fate of the meristems in whorls 2 and 3, the situation in P. hybrida is more complicated: two GLO/PI as well as two DEF/AP3 subfamily members have been identified, the latter consisting of a euAP3-type (PhDEF) and a paleoAP3-type (PhTM6) gene. Theoretically, when both genes encoding the two partners in a DEF/GLO (or AP3/PI)-type heterodimer undergo duplication, four heterodimers can be expected: DEF1/GLO1, DEF1/GLO2, DEF2/GLO1, and DEF2/GLO2. Eventually, this redundancy might lead to the loss of one or both of the duplicated copies, or alternatively, subfunctionalization and/or neofunctionalization might lead to retention of the extra copies. According to this duplicated B-class heterodimer model and assuming that eudicot B-class MADS box proteins can only act as heterodimers consisting of a DEF/AP3 and a GLO/PI member, def1 def2 and glo1 glo2 double mutants should result in full homeotic conversions of both the second and the third whorls, whereas the phenotype of the remaining double mutant combinations depends on the extent to which the duplicated genes have functionally diverged.

The Largely Redundant PhGLO1 and PhGLO2 Genes Are Required for B-Function Activity in P. hybrida

To elucidate the biological role of the two GLO/PI homologs in P. hybrida, we have applied an insertion mutagenesis approach to obtain loss-of-function alleles. phglo1 flowers display only a very local change from petal to sepal identity in the second whorl, whereas the stamen filaments and tube remain unfused. This indicates that PhGLO1 controls the formation of the petal midvein and growth under the zone of petal and stamen initiation, which causes the corolla tube and stamen filaments to emerge as a congenitally fused structure. Interestingly, nonfusion was also observed in phdef flowers and in mild PhDEF cosuppression lines, which still developed petals (van der Krol et al., 1993), suggesting that PHDEF and PHGLO1 jointly direct this process. phglo2 flowers on the other hand exhibit a wild-type architecture, although pollen maturation might be affected. Thus, for both PhGLO1 and PhGLO2 loss-of-function mutations, the drastic homeotic conversions as seen in Arabidopsis pi and Antirrhinum glo mutants were not observed. This, together with the fact that PhGLO1 (Angenent et al., 1995) and PhGLO2 (Figures 4 and 5) exhibit similar expression patterns in developing flowers, suggested that PhGLO1 and PhGLO2 act largely redundantly in petal and stamen development. This is confirmed by the phenotype of phglo1 phglo2 flowers, which display a complete homeotic change from petals to sepals and stamens to carpels. Moreover, this phenotype indicates a complete absence of B-function activity, implying that the two P. hybrida DEF/AP3 homologs PhDEF and PhTM6 by themselves or together do not exert B-function activity. This is in full accordance with the predictions based on a duplicated B-class heterodimer model. The function of PhGLO1 has been analyzed previously using a cosuppression approach (Angenent et al., 1993). The transgenic lines with the most severe alterations exhibited flowers with a phenotype identical to phglo1 phglo2 double mutants, indicating that cosuppression in these lines did not occur in a gene-specific way (cf. Vandenbussche et al., 2003b).

DEF/GLO Heterodimer Formation in Relation to Petal and Stamen Development

The results of the yeast two-hybrid analyses indicate that like in Antirrhinum and Arabidopsis, P. hybrida B-class MADS box proteins form heterodimers consisting of a DEF/AP3-like protein combined with a GLO/PI-like protein. Here, we integrate these data with the results obtained from the expression and mutant analyses into a model describing petal and stamen development in P. hybrida (Figure 6A).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Summarizing Model Describing Unique and Redundant Functions of the Proposed P. hybrida B-Class Heterodimers and the Regulatory Interactions among These Genes Based on Mutant Analyses, Two-Hybrid Interactions, and Expression Studies.

(A) Functions of the proposed heterodimers in wild-type flowers. The two-hybrid interaction data are represented by solid lines between the involved B-class proteins.

(B) Regulatory interactions among the P. hybrida B-class MADS box genes.

The proposed function of the PHTM6/PHGLO2 heterodimer and the regulatory interactions involving PhTM6 need to be confirmed.

Concerning the involvement of two DEF/AP3-like proteins in petal development, we found that in contrast with PhDEF, PhTM6 is expressed at very low levels in the developing petals, indicating that PhTM6 might not be involved in petal development. Therefore, according to the duplicated B-class heterodimer model, phdef mutations should result in full homeotic conversion of petals to sepals because in this whorl, the only putatively available DEF/AP3 partner, PHDEF, is mutated. This is indeed supported by the phenotype of phdef flowers (Figure 2K). Second, we and others (Immink et al., 2003) found that PHDEF is capable of interacting with PHGLO1 and PHGLO2, both of which are well expressed in developing petals. This indicates that PHGLO1 and PHGLO2 might exhibit functional redundancy as common interacting partners of PHDEF. This is confirmed by the phenotype of phglo1 phglo2 double mutants (Figure 2G). Therefore, we conclude that petal development is controlled by the largely redundant PHDEF/PHGLO1 and PHDEF/PHGLO2 heterodimers, whereas PHTM6 most likely is not involved in this process.

The expression analysis further shows that PhTM6 is well expressed in developing stamens, as has been found for PhDEF, PhGLO1, and PhGLO2, indicating that all four identified B-class MADS box genes might be involved in stamen development. The phenotypes of phglo1 phglo2 and phdef phglo2 mutants (Figures 2G and 2P) indeed directly prove the involvement of PhDEF, PhGLO1, and PhGLO2 in stamen development, and the two-hybrid analysis suggests that these genes act through the formation of PHDEF/PHGLO1 and PHDEF/PHGLO2 heterodimers. The two-hybrid analysis further indicated that PHTM6 interacts specifically with PHGLO2 and to a much lesser degree with PHGLO1, suggesting that the PHTM6/PHGLO2 heterodimer (and presumably not PHTM6/PHGLO1) might also be involved in stamen development. Integrating this information in the duplicated B-class heterodimer model and assuming that the capacity to form a heterodimer is a prerequisite for B-function activity, than the following predictions can be made: In phdef phglo2 double mutants, the only available heterodimer is PHTM6/PHGLO1. However, the two-hybrid analysis revealed only a very weak interaction, if any at all, for these proteins. Therefore, in phdef phglo2 double mutants, B-function activity should be severely reduced or completely absent. On the other hand, in phdef phglo1 mutants, the only heterodimer theoretically available is PHTM6/PHGLO2, and the strong two-hybrid interaction between these proteins indicates that this heterodimer might be functional in vivo. The phenotypes obtained in the double mutants are in full agreement with these interpretations. In phdef phglo2 flowers (Figure 2P), stamens are fully replaced by carpels, indicating that the PHTM6/PHGLO1 heterodimer is not formed or not sufficient to confer stamen identity. On the other hand, phdef phglo1 flowers still develop stamens (Figure 2N), suggesting that the PHTM6/PHGLO2 heterodimer might be sufficient to induce stamen development. The partner preference of PHTM6 toward PHGLO2 as suggested by the yeast two-hybrid analysis is thus strongly supported by the phenotypes of these mutants.

Functional Divergence within the DEF/AP3 Lineage during Evolution: Defining Functions for PhDEF and PhTM6

Within the DEF/AP3 subfamily, a major duplication event was identified that coincides with the base of the higher eudicot radiation, resulting in two types of DEF/AP3-like proteins, which can easily be distinguished on the basis of their completely divergent paleoAP3 and euAP3 C-terminal motifs (Kramer et al., 1998). Whereas euAP3 genes appear to control stamen and petal identity in higher eudicot species, paleoAP3 genes until now have only been characterized in the monocot species maize (Zea mays) and rice (Oryza sativa). Both the SILKY1 and OsMADS16 (SPW1) genes (Figure 1) have been shown to control stamen and lodicule identity (Ambrose et al., 2000; Nagasawa et al., 2003). Recently, data involving C-terminal motif swapping experiments were published indicating that paleoAP3 and euAP3 motifs within the eudicot lineage encode divergent functions (Lamb and Irish, 2003); a chimeric construct containing the Arabidopsis AP3 gene, of which the C-terminal euAP3 motif was replaced by a paleoAP3 motif, was able to partially rescue stamen development in an ap3-3 mutant background, whereas second-whorl organs remained fully sepaloid. These results indicate that the C-terminal motif of paleoAP3 proteins promotes stamen but not petal development in higher eudicots, supporting the hypothesis that euAP3 genes may have obtained a novel function, leading to the development of petals.

In P. hybrida, all currently available data fit this hypothesis. Null mutations in the P. hybrida euAP3-like PhDEF gene cause homeotic alterations in the second whorl of the flower, whereas stamen development remains largely unaffected. Further, we and others (Tsuchimoto et al., 2000) have shown that phdef bl flowers develop antheroid organs in the second whorl, although the full conversion of petals to sepals in phdef flowers implies a complete absence of B-function activity in the second whorl in a phdef genetic background. According to the duplicated B-class heterodimer model, this suggests that the paleoAP3-like PhTM6 is able to complement the phdef mutation in stamen development but not in petal development. The expression pattern of PhTM6 in wild-type, phdef, bl, and phdef bl flowers indeed fully supports this hypothesis: In wild-type and phdef flowers, PhTM6 is highly expressed in the developing stamens, but only at low levels in the second-whorl petals. On the other hand, in the second-whorl antheroid organs of bl and phdef bl flowers, PhTM6 expression is strongly upregulated to levels comparable to third-whorl expression levels in wild-type developing stamens. Finally, the phenotype of phglo1 phglo2 and phdef phglo2 flowers together with the two-hybrid analysis indicate that PHTM6 requires PHGLO2 for B-function activity, suggesting that PHTM6 B-function activity can be abolished through mutagenesis of its interaction partner, PHGLO2. Recently, we obtained phdef phglo2 bl triple mutants, and in contrast with phdef bl flowers, these flowers do exhibit a full homeotic conversion to carpels in the second whorl (M. Vandenbussche, S. Royaert, and T. Gerats, unpublished data). Although a full functional analysis of PhTM6 will be required to unequivocally prove it, this suggests that PhTM6 is responsible for the rescue of stamen development in the third whorl of phdef flowers and in the second and third whorls of phdef bl flowers. PhTM6 is thus a very likely candidate to represent gene X as postulated by Tsuchimoto et al. (2000). It may be noted that PhTM6, by sequence a B-function gene, behaves as a C-function gene both in respect to its wild-type expression pattern as well as its changed pattern in an A-function mutant background.

Regulatory Interactions

In both Arabidopsis and A. majus, it has been shown that the expression of either one of the B-function genes is initiated independently but that the maintenance of high levels of AP3 and PI depends upon the presence of the heterodimeric protein complex itself. In A. majus, this interdependence occurs in both whorls 2 and 3, where DEF and GLO gene products are required to regulate each others' expression positively at the level of transcription. In Arabidopsis, such an interdependent relationship exists on the transcriptional level between AP3 and PI in whorl 3, but AP3 continues to be transcribed in the second-whorl sepals in a PI mutant background. AP3 protein is however not detected in second-whorl organs of pi mutants (Jack et al., 1994). The results from our RT-PCR analysis (Figure 4) indicate that for PhDEF on the one hand and for PhGLO1 or PhGLO2 on the other hand, a similar interdependent relation at the transcriptional level exists, as has been found in Antirrhinum. In phglo1 phglo2 flowers, PhDEF expression is strongly reduced in whorls 2 and 3, indicating that PhDEF requires the presence of at least one of the two GLO/PI homologs to maintain its own expression. Similarly, in the second whorl of phdef mutants, the expression levels of both PhGLO1 and PhGLO2 are reduced, suggesting that these genes require PhDEF to maintain high levels of expression in the second whorl (Van der Krol et al., 1993). In the third whorl of phdef flowers, PhGLO1 and PhGLO2 are still expressed at high levels, indicating that another gene, most likely PhTM6, acts redundantly with PhDEF to maintain expression of the P. hybrida GLO genes (see below). Nevertheless, in phdef phglo2 flowers, expression levels of phdef, PhGLO1, and phglo2 are severely downregulated in both the second and third whorls. Other evidence that PhDEF affects the expression of PhGLO1 and PhGLO2 came from transgenic lines overexpressing PhDEF (Halfter et al., 1994). In these lines, sepals are converted to petals, and PhGLO1 and PhGLO2 are ectopically expressed in these converted tissues.

By contrast, the expression pattern of PhTM6 remained virtually unchanged in all tested B-function mutants, indicating that maintenance of PhTM6 expression at the transcriptional level does not depend on the activity of PhDEF, PhGLO1, and PhGLO2. A schematic representation of all of these regulatory interactions in wild-type P. hybrida flowers is shown in Figure 6B.

Further, we found that PhTM6 is upregulated in the second whorl of bl flowers (and to a lesser degree in the first whorl), indicating that PhTM6 expression either is repressed by the BL gene product or, alternatively, that PhTM6 requires C-function activity to maintain high expression levels. The latter hypothesis is consistent with the idea that B- and C/D-function genes originally might have been more tightly associated with each other in reproductive organ development, before specialized sterile perianth organs appeared during flower evolution (Theissen et al., 2000, and references therein). Unfortunately, we cannot yet verify the hypothesis that PhTM6 requires C-function activity because C-function mutants are not yet available in P. hybrida. Finally, our data suggest that both PhGLO1 and PhGLO2 expression in whorl 3 of phdef mutants and in whorls 2 and 3 of phdef bl flowers is possibly maintained by PhTM6, despite the fact that at least in in vitro experiments, PhTM6 appears to interact only with PhGLO2 on the protein level. A plausible explanation for this would be that although PHTM6 is not capable of heterodimerizing with PHGLO1, the PHTM6/PHGLO2 heterodimer might be capable of activating PhGLO1 expression.

B-Class Gene Dosage Modulates the Developmental Fate of Second- and Third-Whorl Meristems in a Quantitative and Qualitative Manner

Normally, B-class gene dosage is not directly associated with phenotypic changes in flower morphology. Like the known B-class mutant alleles from Arabidopsis and Antirrhinum, all available phglo1, phglo2, and phdef mutations are fully recessive, implying that a reduction of 50% in gene dosage because of the presence of a mutant allele in a heterozygous state does not induce any phenotypic effect. However, when a function that is normally encoded by one gene is duplicated, gene dosage can be further reduced to a level of 25% in plants homozygous mutant for gene copy A and heterozygous for gene copy B. This readily explains our observation of gene dosage effects in specific progeny classes, as, for example, in the second whorl of PhDEFphdef phglo1 flowers and in the second and third whorls of phglo1 PhGLO2phglo2 flowers (Figures 2I, 2J, 2O, and 2Q). Further, in PhGLO1phglo1 phglo2 flowers, second-whorl organs fully retain their petal identity, whereas stamens are replaced by carpelloid structures (Figures 2H, 2L, 2M, and 2Q). This indicates that the sensitivity of petal and stamen development toward a reduction in B-class gene dosage differs and that the third whorl apparently requires a higher dose of B-function activity to maintain normal identity. Remarkably, a similar phenomenon was observed in Antirrhinum in heterozygous plants carrying a def null allele in combination with the temperature sensitive def-101 allele (Schwarz-Sommer et al., 1992; Zachgo et al., 1995). Gene dosage effects have also been described for specific sepallata mutant combinations (Favaro et al., 2003).

Evolution of P. hybrida B-Function Regulation Compared with Antirrhinum and Arabidopsis

The data presented here demonstrate that gene duplications in the DEF/AP3 and GLO/PI lineages in P. hybrida have led to a functional diversification of their respective members, which is reflected by partner specificity and whorl-specific functions among these proteins. Nevertheless, when the individual actions of PhDEF, PhGLO1, and PhGLO2 are considered together, they appear basically to function in a similar way as DEF/GLO in Antirrhinum and AP3/PI in Arabidopsis. Analyzed in more detail, the P. hybrida B-class loss-of-function phenotypes and the regulatory network among these genes resemble more the situation in Antirrhinum than in Arabidopsis, which is logical given the taxonomic distribution of the three species. In addition, we identified a novel function fulfilled by a B-class heterodimer: PHDEF/PHGLO1 seems to be controlling the fusion of the stamen filaments with the petal tube. In Antirrhinum and Arabidopsis, this function apparently is not present because in these flowers, wild-type stamens emerge as freestanding structures. This might be an example of a subtle difference in function that accounts for species-specific differences in floral architecture.

In contrast with what we found for PhDEF, all current data indicate that the function and mode of action of the paleoAP3-type PHTM6 differs significantly from the known euAP3-type DEF/AP3-like proteins; PhTM6 is mainly expressed in the third and fourth whorls, its maintenance of expression does not require functional GLO/PI-like proteins, its expression pattern is altered in an A-function mutant background, and most likely PhTM6 is not involved in petal development. The latter is consistent with the hypothesis that the evolutionary origin of the higher eudicot petal structure coincided with the appearance of euAP3-type MADS box genes (Kramer et al., 1998; Lamb and Irish, 2003; Vandenbussche et al., 2003a). Future research focused on a functional analysis of PhTM6 may shed new light on the recruitment of B-class MADS box genes in petal and stamen development during evolution.

Phylogenetic Analysis

The neighbor-joining tree shown in Figure 1 was obtained according to the methodology described previously (Vandenbussche et al., 2003a).

Plant Material and Genotyping

For the crosses described in this article, we have used a phglo1-2 line in which the dTph8 transposon has excised, leaving behind a 5-bp footprint, which introduces a stop codon at amino acid position 143 of the PhGLO1 coding sequence. This phglo1-2–derived footprint allele induces an identical phenotype as the phglo1-1 and the original phglo1-2 alleles. From the phglo2 alleles (Table 1), we have selected phglo2-3 because in this line, the first exon encoding the DNA binding MADS domain is disrupted by the transposon insertion, most likely resulting in a null mutation. Homozygous phglo1-2, phglo2-3, and phdef-1 mutants were crossed with each other and to the bl mutant (Vallade et al., 1987), and the resulting F1 plants were self-fertilized to obtain F2 progenies. All the different phenotypic classes described segregated in agreement with the expected Mendelian ratios, and similar results were obtained when we analyzed additional F2 progenies obtained from different F1 individuals. Genotyping of the phglo1-2 and phglo2-3 insertion alleles was done by PCR using gene-specific forward (fw) and reverse (rv) primer pairs flanking the insertion sites (PhGLO1-fw, 5′-CTTGAAGGGTGAAGATATCACATC-3′; PhGLO1-rv, 5′-TTCCTCATCATCCTCAGAACCTC-3′; PhGLO2-fw, 5′-GAGAAGTGAGATATTAGGTATGG-3′; and PhGLO2-rv, 5′-GCTACAATATTCATGCATCTTGCCAGA-3′) (data not shown). Genotypes for phdef and bl mutant alleles were scored on a phenotypic basis and, in specific cases, confirmed by backcrossing with phdef and bl homozygous mutants, respectively.

Electron Microscopy

Samples for cryo-scanning electron microscopy were first frozen in slush, prepared in an Oxford Alto 2500 cryo-system (Catan, Oxford, UK), and then analyzed in a JEOL JSM-6330F field emission electron scanning microscope (JEOL, Tokyo, Japan).

RT-PCR Analysis

Total RNA from different tissues was isolated using Trizol reagent (Life Technologies, Cleveland, OH) according to the instructions of the manufacturer. First-strand cDNA synthesis was done by combining 2 μg of total RNA diluted in 20 μL with 1 μL of oligo(dT)₂₅ primers (700 ng/μL), 4 μL of water, 8 μL of first-strand buffer, 4 μL of 0.1 M DTT, 2 μL of 10 mM deoxynucleotide triphosphate, and 1 μL of Superscript II RT (200 units/μL; Gibco BRL, Cleveland, OH) in a total volume of 40 μL. After incubation for 2 h at 42°C, the mixture was diluted 10 times. Five microliters of this dilution was used as a template for PCR amplification. PCR products were visualized by radiolabeling one primer of each gene-specific primer pair and analyzed by PAGE as described previously (Vandenbussche et al., 2003b). Integrity of the RNA samples and cDNA synthesis was monitored by measuring the expression level of glyceraldehyde-3-phosphate dehydrogenase as a positive control (25 PCR cycles). Signals for the B-class MADS box genes as presented here were obtained after 33 PCR cycles. Expression in the wild type was analyzed at two developmental stages (in 0.5-cm and 4- to 5-cm buds); for the mutant samples, these two stages were pooled. All reactions have been done in duplicate starting from independent RNA samples, and all results were in good agreement (data not shown). Some of the expression patterns shown here have been analyzed previously by RNA gel blot analysis; we have obtained comparable results in all cases (Angenent et al., 1992, 1995; van der Krol et al., 1993; Tsuchimoto et al., 2000).

In Situ Hybridization

3′ gene-specific fragments of PhGLO2 and PhTM6 were generated by PCR using gene-specific primer pairs and subsequently cloned in pGEM-T (Promega, Madison, WI), containing T7 and SP6 transcription sites. Probe synthesis and in situ hybridizations were performed as described previously (Caňas et al., 1994). Images were recorded with an AxioCam digital camera (Zeiss, Jena, Germany).

Two-Hybrid Analysis

The pBD-GAL4 bait and pAD-GAL4 prey vectors containing PhDEF, PhGLO1, or PhGLO2 were provided by Richard Immink and described previously (Ferrario et al., 2003). A full-length cDNA copy of PhTM6 was generated by PCR and cloned into the pAD-GAL4 vector. The GAL4 yeast two-hybrid analyses were performed as described previously (Immink et al., 2003), using the yeast strain PJ69-4a (James et al., 1996). Selection for interaction was performed on selective medium lacking His supplemented with 10 mM 3-AT (3-amino-1,2,4-triazole; Sigma, St. Louis, MO) and confirmed on selective medium lacking His and Ade and on medium supplemented with X-α-Gal (CLONTECH, Palo Alto, CA). Growth of yeast on selective media was scored after 7 d of incubation at 20°C, whereas blue staining was scored after an overnight incubation at 20°C.

Sequence Deposition

The genomic structure of PhGLO1 (AY532265) and PhTM6 (AY532264) was determined by sequencing PCR-generated fragments amplified from genomic DNA, covering the full coding sequence of these genes.

Sequence data from this article have been deposited with the EMBL/GenBank data libraries under accession numbers AY532264 and AY532265.