The MADS-box gene DEFH28 from Antirrhinum is involved in the regulation of floral meristem identity and fruit development


For correspondence (fax +49 221 5062113; e-mail


DEFH28 is a novel MADS-box gene from Antirrhinum majus. Phylogenetic reconstruction indicates that it belongs to the SQUA-subfamily of MADS-box genes. Based on its expression pattern and the phenotype of transgenic plants it is predicted that DEFH28 exerts a dual function during flower development, namely control of meristem identity and fruit development. Firstly, DEFH28 is expressed in the inflorescence apical meristem and might control, together with SQUAMOSA (SQUA), floral meristem identity in Antirrhinum. Also, DEFH28 is sufficient to switch inflorescence shoot meristem to a floral fate in transgenic Arabidopsis thaliana plants. Secondly, DEFH28 is expressed in carpel walls, where it may regulate carpel wall differentiation and fruit maturation. Support for this later role comes from overexpression of DEFH28 throughout the silique in transgenic Arabidopsis plants where it altered the identity of the replum and valve margin cells so that they adopted a valve cell identity. This late aspect of the DEFH28 function is identical to the FRUITFULL (FUL) function of Arabidopsis as demonstrated in gain-of-function plants. FUL, like DEFH28, belongs to the SQUA-subfamily of MADS-box genes. DEFH28 most likely represents the ortholog of FUL. Promoter analysis shows that the control mechanism conferring a carpel wall specific expression has been conserved between Antirrhinum and Arabidopsis during evolution. Although the overall flower development between Antirrhinum and Arabidopsis is very similar, their carpels mature into different types of fruits: capsules and siliques, respectively. Therefore, it is suggested that the role of DEFH28 in control of carpel wall differentiation reflects a conserved molecular mechanism integrated into two very different carpel developmental pathways.


The formation of reproductive plant organs requires a change in the life cycle of a plant, a transition from the vegetative to the reproductive growth phase. In Antirrhinum, SQUAMOSA (SQUA) is required to make flowers as it specifies the identity of meristems to become floral. In squa mutants, secondary inflorescences are formed instead of floral meristems (Huijser et al., 1992). However, squa inflorescences occasionally do produce abnormal flowers, indicating that in Antirrhinum other factors can probably substitute for the SQUA function to accomplish floral identity.

SQUA and its Arabidopsis ortholog APETALA1 (AP1;Mandel et al., 1992) are members of the MADS-box gene family. These genes exert pivotal roles throughout plant reproductive development, regulating floral meristem identity and floral organogenesis, as well as late processes such as tissue differentiation during fruit maturation. Plant MADS-box genes encode transcription factors which share an overall conserved structure composed of four domains (MIKC): a highly conserved DNA-binding MADS-domain (M) at the N-terminus is followed by an intervening (I) region of variable length and a K box (K) which is involved in protein–protein interactions (Theißen et al., 1996). The high variability of the C-terminal domain (C) contributes to the specific activity of each MADS-box protein (Egea-Cortines et al., 1999).

The Arabidopsis MADS-box genes CAULIFLOWER (CAL) and FRUITFULL (FUL) are closely related to AP1, but have no impact on meristem identity as single mutants (Gu et al., 1998; Kempin et al., 1995; Mandel and Yanofsky, 1995b). However, ap1/cal and most severely ap1/cal/ful mutants enhance the ap1 mutant phenotype, the latter resulting in a non-flowering phenotype (Ferrándiz et al., 2000a; Kempin et al., 1995). Thus, the developmental fate of floral meristems in both, Antirrhinum and Arabidopsis, is controlled by at least partly redundant activities. SQUA, AP1, CAL and FUL belong to the SQUA-subfamily of MADS-box genes, which was established by comparison of the amino acid composition in the MIK domain of different MADS-box proteins. As subfamily members often reveal similar expression patterns and overlapping functions, membership in a certain subfamily can be indicative for the putative function of a MADS-box gene (Theißen et al., 2000a). However, recent data show that besides the early, redundant function of FUL, shared with the other subfamily members, FUL has an additional function during fruit development, where it regulates valve differentiation and valve cell expansion after fertilization (Gu et al., 1998). Following fertilization, the Arabidopsis gynoecium expands strongly into an elongated silique. The silique is composed of two carpels, separated by a fused structure, called the septum. The carpel walls, known as valves, are joined by a structure of the outer septum, named the replum (Esau, 1977). The cells between the valves and the replum, at the valve margins, mediate dehiscence of the siliques. After silique maturation, the combination of cell differentiation patterns within the valves and desiccation of the fruit tissues forms tensions within the carpel walls that finally result in the separation of the valves from the replum (Spence et al., 1996; Figure 7b). In siliques of ful mutants, the valve cells fail to elongate and differentiate, which results in the formation of shortened siliques with overcrowded seeds. As no dehiscence along the replum occurs, seeds prematurely burst out of ful siliques (Gu et al., 1998). In angiosperms, a large number of divers fruit types exist, mediating seed dispersal by different mechanisms (Esau, 1977). Despite the increasing knowledge on the regulation of flowering, little is known about the molecular mechanisms controlling fruit morphogenesis. During fruit development in Antirrhinum, a capsule is formed by two fused locules that dry during maturation. In contrast to Arabidopsis, where seed dispersal is mediated by dehiscence along the septum, Antirrhinum displays a porose dehiscence mechanism. Three pores are formed at the tip of the capsule, initiated by a split parallel to the septum, which is followed by additional ruptures resulting in the final pore formation (Sutton, 1988; Figure 7a).

Figure 7.

Comparison between fruit develop ment of Antirrhinum and Arabidopsis.

(a) Capsules from Antirrhinum form upon maturation three pores at the dry tip: one elongated pore in the adaxial and two pores in the abaxial locule.

(b) In Arabidopsis, dehiscence occurs by ruptures along the septum, separating the two valves from the replum and revealing the seeds attached to the septum.

(c,d) Schematic representation of a cross section through the middle of the fruits shown in (a) and (b) depicting the different fruit tissues in Antirrhinum and Arabidopsis.

(e,f) Comparison of the regulatory interactions involved in fruit development in Antirrhinum and Arabidopsis. Arrows symbolize activation and barred lines indicate antagonistic interactions. In Arabidopsis, AGAMOUS (AG) controls carpel organogenesis (Yanofsky et al., 1990) and its activity can therefore be considered a prerequisite for FUL and SHP1/2 function. In contrast, in Antirrhinum, carpel organogenesis is regulated by two MADS-box genes: PLENA (PLE) and FARINELLI (FAR;Bradley et al., 1993; Davies et al., 1999; Ferrándiz et al., 2000b). It will be intriguing to find out if genes exerting a similar function in fruit dehiscence as the Arabidopsis genes SHP1/2 exist in Antirrhinum and are negatively regulated by DEFH28 or if novel mechanisms evolved independently in this species (‘X’).

Here, we report the characterization of the DEFICIENS-Homolog 28 (DEFH28). Several experiments indicate that DEFH28, similar to FUL, exerts a dual function during flower development. Firstly, it seems to affect valve differentiation during fruit maturation. Although the fruit types from Antirrhinum and Arabidopsis capsules and siliques differ fundamentally from each other, carpel wall differentiation seems to be controlled in both species by a conserved molecular mechanism. Secondly, DEFH28 might also participate in the regulation of floral meristem identity.


Isolation and sequence analysis of DEFH28

The Antirrhinum MADS-box gene DEFH28 was isolated from a flower specific cDNA library by screening with a probe comprising a mixture of MADS-box fragments (see Experimental procedures). The DEFH28 protein exhibits the MIKC structure that is typical for MADS-domain proteins from higher eudicots (Theißen et al., 1996). Phylogenetic tree reconstruction groups MADS-box genes into subfamilies according to the degree of amino acid conservation. Analyses showed that MADS-box proteins of the same subfamily share similar expression patterns and are often involved in regulating similar developmental processes (Theißen et al., 2000a; Theißen et al., 1996). Therefore, subfamily membership can be indicative for the putative function of a gene. Phylogenetic tree analysis of the MIK domains from all MADS-box genes known to date revealed that DEFH28 belongs to the SQUA-subfamily of MADS-box genes. The position of DEFH28 within a subset of SQUA-subfamily members is shown in Figure 1(a). This subfamily comprises SQUA from Antirrhinum majus (Huijser et al., 1992) and AP1, CAL and FUL from Arabidopsis thaliana (Kempin et al., 1995; Mandel and Yanofsky, 1995b; Mandel et al., 1992). Mutant analyses showed that these genes are involved in establishing floral meristem identity. Overexpression and co-suppression studies revealed a similar function for other subfamily members like EAP1/EAP2 from Eucalyptus, PEAM4 from Pisum or PFG from Petunia (Kyozuka et al., 1997; Berbel et al., 2001; Immink et al., 1999).

Figure 1.

Sequence analysis of DEFH28.

(a) Phylogenetic tree construction was conducted with the neighbor-joining algorithm using the MIK domain amino acid sequences (Winter et al., 1999) of DEFH28 and a subset of MADS-box genes. Numbers above nodes give bootstrap percentages. In order to root the SQUA-subfamily, SEPALLATA1 (SEP1), AGAMOUS (AG), DEFICIENS (DEF) and GLOBOSA (GLO) were used as representatives of the respective subfamilies. Information on the different genes as well as their accession numbers are provided within the worldwide web at the ‘MADS homepage’ ( The DEFH28 cDNA sequence is deposited in GenBank under accession number: AY040247.

(b) Alignment of the predicted MADS-domain amino acid sequences from DEFH28, BPMADS4, FUL, SQUA, AP1 and CAL. Identical amino acids are shown in dark shading, similar ones in bright shading.

(c) Schematic diagram of the exons (boxes) and introns (lines) of the DEFH28 gene.

Beyond this early function in floral development some genes like AP1 exert additional functions regulating organ identity in the first and second whorl (Irish and Sussex, 1990) or the recently reported control of valve differentiation by FUL (Gu et al., 1998). A comparison of the MADS-domain protein sequences of a subset from the SQUA-subfamily members shows the high degree of amino acid identity within the conserved MADS consensus sequence (Figure 1b).

In order to analyze the genomic organization of the DEFH28 locus, the genomic clone comprising 3.1 kb of the promoter and 7.3 kb of the transcription unit was isolated. Comparison with the cDNA sequence showed that the transcription unit consists of eight exons separated by seven introns (Figure 1c). The MADS-domain is encoded by the first exon, the K-domain by the third, fourth and fifth exon, with the remaining exons coding for the C-terminus of the DEFH28 protein. Database searches showed that the exon/intron number is shared with other SQUA-subfamily members like SQUA, AP1 and FUL. In contrast, genomic loci from members of other subfamilies, such as DEFICIENS and GLOBOSA, comprise six exons or, as for DEFH125, only five exons (Schwarz-Sommer et al., 1992; Tröbner et al., 1992; Zachgo et al., 1997).

DEFH28 expression during flower development

Northern blot experiments showed that the DEFH28 expression is flower specific (Figure 2). Using a DEFH28 probe without the MADS-box to avoid cross-hybridization, no signal could be detected in vegetative tissues. In contrast, high levels of a 0.85 kb long DEFH28 transcript were observed in inflorescences and buds. To determine the DEFH28 expression during flower development more precisely, RNA in situ hybridization experiments with digoxigenin-labeled DEFH28 antisense probes were conducted on longitudinal sections of Antirrhinum inflorescences (Figure 3).

Figure 2.

Northern blot analysis of the DEFH28 expression pattern.

Two µg of mRNA, isolated from wild-type roots, seedlings, inflorescences, buds and leaves were probed with a radioactively labeled DEFH28 probe devoid of the MADS-box. Equal loading was tested by hybridization with an actin control (not shown).

Figure 3.

In situ expression analysis of the DEFH28 mRNA during wild-type floral development.

Longitudinal sections hybridized with digoxigenin-labeled antisense DEFH28 mRNA revealed a biphasic DEFH28 expression pattern. The earliest detectable signal is in the inflorescence meristem (a). During young floral primordium stages DEFH28 signal decreases (a, arrowhead). The second expression phase is mostly confined to the fourth whorl, starting as a bipartite signal underlying the area where the future carpel wall primordia will be initiated (b,c). DEFH28 expression is restricted during further flower development (d) to the basal parts of the carpel walls and was not detected during older stages in ovule primordia, marked with arrows (e). Control experiments were conducted using a DEFH28 sense probe (not shown). 1, 2, 3 and 4 designate the four floral whorls in the flower buds. Bar = 100 µm.

DEFH28 reveals a biphasic expression pattern during floral development. The early expression phase starts after the transition from a vegetative to an inflorescence meristem has taken place. DEFH28 mRNA is detected in the apical inflorescence meristem and a weak DEFH28 signal is detectable in young floral primordia (Figure 3a) which resolves before sepal primordia are initiated. Furthermore, expression in young bracts is visible (Figure 3a). During the second phase, DEFH28 expression is mostly confined to the fourth whorl. Antirrhinum exhibits an axile placentation type where the carpel walls of the two locules fold inwards and thereby merging their margins and bringing the placentae together (Esau, 1977; Puri, 1952). After initiation of sepal primordia strong DEFH28 expression is detected as a bipartite signal flanking the center of the primordium where later the two carpel wall primordia will be formed (Figure 3b). During further development of the ovary the DEFH28 expression remains restricted to the lower parts of the carpel walls (Figure 3c,d) where it is maintained throughout later stages after ovule primordia have been initiated (Figure 3e). Occasionally, a weaker staining in the abaxial, basal parts of older petal and stamen primordia could be detected (not shown). The DEFH28 expression dynamic shows high similarity to the biphasic FUL expression (Mandel and Yanofsky, 1995b), which indicates that DEFH28, like FUL, might exert a dual function during flower development.

Analysis of the DEFH28 promoter

In situ experiments showed that the strongest expression of DEFH28 is confined to the carpel walls, similar to the FUL expression in Arabidopsis (Mandel and Yanofsky, 1995b). During carpel development of Antirrhinum and Arabidopsis, carpel walls participate in the formation of two completely different fruit types: capsules and siliques, respectively. To investigate if these differences in fruit development are reflected by a temporally or spatially altered DEFH28 and FUL expression, transgenic Arabidopsis plants were generated expressing the GUS reporter-gene under the control of the presumptive DEFH28 promoter region. Figure 4 shows the GUS expression pattern that was observed in 15 independent T1 plants.

Figure 4.

Expression of the GUS reporter-gene driven by the DEFH28 promoter in transgenic Arabidopsis plants.

T1 plants containing the DEFH28::GUS construct where stained to reveal the GUS expression. Whole plants just after the transition from vegetative to reproductive growth showed only staining in the stipules, which is most likely unspecific (a). Slightly later, after initiation of the first flower buds on the main shoot, staining in cauline leaves and whole buds was detected (b) , being similar to staining of coflorescences (d). During later stages of flower development strongest staining was observed in carpels (c,e). During further silique development the GUS signal disappears and is finally absent in mature siliques (f).

No GUS activity was detected in plants before the transition to flowering occurred, except for the artifact of stipules staining (Figure 4a), which was also observed in non-transgenic plants. Early GUS activity was observed in young inflorescences where it was detected in young floral buds and young cauline leaves (Figure 4b) and in young co-florescences (Figure 4d). During later stages of floral development, GUS staining was confined to the carpels, omitting the stigmatic papillae (Figure 4c). After fertilization, during maturation of the silique, the GUS signal started to vanish from the silique (Figure 4c,e,f). Besides strong carpel and young inflorescence specific GUS expression, a weak and transient DEFH28 expression was occasionally detected in other organs like pedicels (Figure 4c), petals, anthers and stems just below the inflorescence. However, GUS staining was never detected in vegetative plant parts like roots and leaves.

Overexpression of DEFH28 in Arabidopsis thaliana affects early and late stages of flower development

To gain further information about the function of the DEFH28 gene, the DEFH28 coding region was placed under the control of the CaMV 35S promoter and the vector was transformed into Arabidopsis thaliana plants. Overexpression of the 35S::DEFH28 transgene showed two striking effects, acting both early and late during flower development.

52 plants from 135 transgenic T1 plants differed markedly from wild-type plants being affected in meristem identity and flowering time (Figure 5, Table 1). The strength of the phenotype generally correlated with the expression level of DEFH28 in the transgenic plants, as determined by Northern blot experiments (data not shown). Phenotypic evaluation was carried out only on T1 plants displaying strong deviations from wild-type development due to high DEFH28 expression levels.

Figure 5.

Phenotype of 35S::DEFH28 plants.

35S::DEFH28 plants flower earlier and with a reduced number of rosette leaves (b) compared to wild-type Arabidopsis plants (a). In addition to the early flowering phenotype, an altered inflorescence architecture is observed. Floral development is terminated with a final compound flower that contains an increased number of floral organs (e). The 35S::DEFH28 plants strongly resemble tfl1–11 mutant plants with respect to plant habit (c) and flower formation (f). Plants in a, b and c are of the same age and were grown under identical LD photoperiod conditions.

Table 1.  Effect of ectopic DEFH28 expression on Arabidopsis phenotypes under long day conditions
  • a

    values expressed as mean ± standard deviation

  • *

    *days after sowing until anthesis

  • b

    b number of plants scored

  • c

    number of siliques scored (2–3 siliques/plant, measured in mm)

  • nd

    nd, not determined

total rosette leavesa
10.2 ± 0.9
4.4 ± 0.8
7.8 ± 0.9
flowering time*a
33.5 ± 2.9
22.2 ± 2.4
26.3 ± 3.3
length of siliquesa
13.3 ± 1.3
9.1 ± 1.5

Comparison of 35S::DEFH28 T1 plants with wild-type plants showed a more than two-fold reduction in the total number of rosette leaves (Figure 5b,a; Table 1). Concomitantly, under long day conditions flowering occurred more than 10 days earlier in transgenic plants than in wild-type plants (Table 1). Wild-type plants developed an indeterminate shoot producing only lateral flowers. Flowers comprised four concentric whorls formed by four sepals, four petals, six stamens and two carpels fused to a central gynoecium (Figure 5d). Growth of the shoot apex in 35S::DEFH28 plants terminated prematurely in an abnormal floral structure (Figure 5e). These fertile terminal flowers developed two to four carpels that were surrounded by a variable number of sepals, petals and stamens. Occasionally, mosaic organs, such as stamens with petaloid structures were observed (not shown).

The effect on plant architecture, flower development and flowering time is strikingly similar to the Arabidopsis tfl1–11 mutant phenotype (Figure 5c,f) and to Arabidopsis plants ectopically expressing the MADS-box gene AP1 (Shannon and Meeks-Wagner, 1991; Alvarez et al., 1992; Mandel and Yanofsky, 1995a). A similar phenotype was mentioned for 35S::FUL plants (Ferrándiz et al., 2000b) but has not yet been documented in detail. Comparison between wild-type, 35S::DEFH28 T1 plants and tfl1–11 mutants revealed that rosette leaf numbers and flowering time were most severely affected in plants overexpressing DEFH28 (Table 1).

A second strong effect of DEFH28 overexpression was detected after fertilization during silique development. Siliques of T1 plants failed to dehisce and to release their seeds upon fruit ripening (Figure 6a). In wild-type Arabidopsis plants, siliques are composed of two valves, which are separated by the replum. Dehiscence occurs at the valve margins, the junction between the valve and the replum (Figure 6c; for schematic representation see Figure 7d). Cells adjacent to the dehiscence zone and cells in the inner valve layer, named the endocarp, are lignified in wild-type plants (Figure 6e). During fruit ripening, the combination of drying tensions and the valve-specific lignification pattern results in a breakdown between middle lamella of the cells in the dehiscence zone and thereby mediates valve separation and release of the seeds (Ferrándiz et al., 1999; Spence et al., 1996).

Figure 6.

Overexpression of DEFH28 throughout Arabidopsis siliques changes the identity of valve margin and replum cells.

During fruit dehiscence of mature wild-type siliques the two valves separate from the replum, starting from the base of the silique (a , lower part). Dehiscence is mediated by drying tensions and middle lamella breakdown in lignified cells at the valve/replum border, the valve margins. Siliques from 35S::DEFH28 plants fail to dehisce (a, upper part) and seeds can only be harvested after manual silique opening (b). In transgenic siliques a normal septum is formed separating the two locules and fertile seeds are produced (b). Comparison of wild-type (c) and siliques from DEFH28 overexpressing plants (d) by scanning electron microscopy (SEM) reveals that the cells at the valve margins and the outer replum adopt a valve cell identity that results in a failure to mediate dehiscence. No style is formed, but stigmatic papillae develop directly at the tip of the valves. SEM analysis was carried out at stage 18, when the siliques turn yellow (for stage definition see Ferrándiz et al., 1999). Comparison of the lignification pattern on cross sections through siliques showed that the stretch of lignified cells mediating dehiscence in wild-type siliques (e) is not formed at the valve margins in the transgenic siliques (f). Lignification of the endocarp layer and vascular bundles is not affected by DEFH28 overexpression (f) stg, stigmatic papillae; sty, style; vm, valve margin; v, valve; me, valve mesophyll layers; len, lignified endocarp layer; vb, vascular bundles. Bar = 100 µm.

In contrast to wild-type siliques, no distinction could be made between valve and replum cells in the undehisced siliques of 35S::DEFH28 plants. SEM analysis showed that the short style was absent. No dehiscence zone was formed, instead cells of the valve margins and replum adopted a valve-like identity (Figure 6d). This observation was supported by histological analysis of the lignification pattern of 35S::DEFH28 siliques, as no lignification could be observed in the small stretch of cells at the valve margins (Figure 6f). Endocarp cells and vascular bundles were lignified as in wild-type siliques (Figure 6e,f), but obviously that was not sufficient to mediate dehiscence in the absence of lignified cells at the valve margins. A similar silique phenotype was observed in 35S::FUL plants, where also no lignified cells at the valve margins were formed (Ferrándiz et al., 2000b). The similarities between the DEFH28 and FUL overexpression silique phenotypes show that DEFH28 can substitute the FUL function in promoting ectopic valve differentiation.

Manual opening of the siliques revealed that a normal septum was formed separating the two locules (Figure 6b). However, the length of the siliques in the transgenic plants was reduced by about 30% compared with wild-type plants (Table 1). The seeds had a normal appearance and were fully viable. Analysis of the progeny from selfed T1 plants showed that the observed phenotypes were heritable and displayed by the T2 generation (data not shown).


DEFH28 is a novel member of the SQUA MADS-box gene subfamily

The DEFH28 gene encodes a novel Antirrhinum MADS-domain protein comprising the MIKC domain structure, which is characteristic of plant MADS-box transcription factors. DEFH28 groups into the SQUA-subfamily, named after the SQUA gene from Antirrhinum, which determines floral meristem identity. As flowers are occasionally formed in squa mutants, floral meristem initiation seems to be redundantly controlled (Huijser et al., 1992). In Arabidopsis, the process of floral meristem initiation is also redundantly controlled by a group of three MADS-box genes, all members of the SQUA-subfamily, namely AP1, CAL and FUL (Ferrándiz et al., 2000a). From all so far functionally characterized SQUA-subfamily members, DEFH28 shows the highest degree of homology to FUL, indicating that DEFH28 might exert a function similar to FUL.

Besides this meristematic function that seems to be shared between the SQUA-subfamily members and was probably established over 200 million years ago (Theißen et al., 2000a), some of the subfamily members are involved in additional regulatory processes. Interestingly, FUL also has a late function during flower development, regulating valve cell differentiation. Loss of the FUL function blocks elongation of the silique after fertilization and inhibits silique dehiscence (Gu et al., 1998). The high degree of homology between DEFH28 and FUL raised the question if DEFH28 also exerts a dual function during flower development. Expression pattern analyses and overexpression studies in Arabidopsis addressed this assumption.

The DEFH28 expression analyses suggest a dual function of DEFH28 during flowering

Northern analysis showed that DEFH28 is expressed in inflorescences and older floral buds but is not detectable in vegetative organs. In situ experiments revealed the onset of DEFH28 expression in the apical inflorescence meristem. Once flower primordia were initiated the DEFH28 expression in the primordia decreased, coinciding with the onset of SQUA expression in these structures (Huijser et al., 1992). Similarly, early FUL expression was detected in the shoot apical meristem immediately after the transition from the vegetative to the reproductive growth phase and disappears from floral primordia due to negative regulation by AP1 (Mandel and Yanofsky, 1995b). Whether this antagonism is realized by a direct or indirect mechanism remains to be answered.

The late DEFH28 expression phase starts after initiation of sepal primordia. DEFH28 expression is predominantly confined to the fourth whorl, where it is localized in the carpel primordia and, during further carpel development, in the basal part of the carpel walls. Antirrhinum exhibits an axile placentation, explaining the observed central basal DEFH28 expression in the carpel, which is originating from expression in the folded and joined carpel walls. Arabidopsis exhibits a parietal placentation type where carpels are marginally joined. FUL transcript is detected exclusively in carpel primordia and valve cells and is absent in the septum, as this is a partition formed by an outgrowth of the placentae at the carpel margins (Esau, 1977; Mandel and Yanofsky, 1995b; Puri, 1952). Although Antirrhinum and Arabidopsis have evolved different placentation types and fruit types, carpel wall differentiation seems to be regulated by a conserved mechanism, controlled by DEFH28 and FUL, respectively. These observations are further supported by DEFH28 promoter studies in Arabidopsis. Arabidopsis plants expressing the GUS gene under the control of a 2.8-kb DEFH28 promoter fragment showed a strong GUS expression in the siliques. This indicates that not only do the DEFH28 and FUL proteins exert similar functions during carpel wall development but also the mechanism controlling their valve specific expression seems to be conserved between the two species.

DEFH28 overexpression in Arabidopsis affects meristem identity and flowering time

To investigate the function of DEFH28, Arabidopsis plants were analyzed expressing DEFH28 under the control of the CaMV 35S promoter. Two striking effects on early and late flower development, correlating well with the biphasic expression pattern of DEFH28, were observed.

Arabidopsis plants overexpressing DEFH28 showed a drastically reduced flowering time. Flowering occurred after 22 days compared with wild-type plants, which flowered after about 33 days under long day conditions. This was also reflected by a strongly reduced number of rosette leaves. Instead of an indeterminate, a determinate inflorescence with an abnormal terminal flower was formed at the tip. This altered phenotype is very similar to the Arabidopsis tfl1 mutant. TERMINAL FLOWER1 (TFL1) delays the transition of the shoot apex from vegetative to reproductive growth and prevents conversion of an indeterminate to a determinate inflorescence meristem (Alvarez et al., 1992; Shannon and Meeks-Wagner, 1991). The meristem identity genes AP1 and CAL act antagonistically to TFL1 and promote establishment of floral meristems, which is also reflected by their complementary expression patterns (Bradley et al., 1997; Kempin et al., 1995; Mandel et al., 1992). Plants ectopically expressing AP1 reveal the same phenotype as tfl1 mutants, which is caused by an opposing effect of high AP1 levels on the TFL1 function by inhibiting its transcription (Mandel and Yanofsky, 1995a; Ratcliffe et al., 1999).

For 35S::FUL plants, subtle to severe effects on flowering time, as well as formation of a terminal flower, are mentioned by Gu et al. (1998) and Ferrándiz et al. (2000b), but have not been documented in detail. However, functionality of the 35S::FUL transgene was shown by its ability to partially rescue the valve defects of ful mutants (Gu et al., 1998).

Prenylation of the CaaX box located at the C-terminus of AP1 was shown to be a prerequisite for directing compound terminal flower development in plants ectopically expressing AP1 (Yalovsky et al., 2000). As this motif is not present in the DEFH28 protein and a cross between ap1 mutants and 35S::DEFH28 plants did not rescue the ap1 phenotype (data not shown), ectopic DEFH28 expression is probably not acting by directly replacing the AP1 function. More likely, it exerts its function on the redundant network of genes regulating flower meristem identity in a fashion similar to ectopically expressed FUL, which might be reflected by an increase in AP1 or decrease in TFL1 expression levels.

Overexpression of DEFH28 results in an altered replum cell identity causing failure of silique dehiscence

Analysis of siliques from 35S::DEFH28 transgenic Arabidopsis plants showed that DEFH28 expression throughout the whole carpel causes the replum and valve margin cells to adopt a valve cell identity, resulting in a failure to form a functional dehiscence zone. The same silique phenotype is observed in plants overexpressing FUL (Ferrándiz et al., 2000b). Similarly, double mutants of the two MADS-box genes SHATTERPROOF1 (SHP1) and SHATTERPROOF2 (SHP2) reveal an absence of the dehiscence zone, however, without affecting the replum development (Liljegren et al., 2000). FUL negatively regulates SHP1 and SHP2, which redundantly control differentiation of the dehiscence zone cells and lignification of neighboring cells (Figure 7f; Liljegren et al., 2000; Savidge et al., 1995). Ectopic expression of FUL in the dehiscence zone represses SHP1 and SHP2 expression resulting in ectopic valve cell formation and therefore omits the formation of lignified cells at the valve margins (Ferrándiz et al., 2000b). Loss of lignified cells at the valve margins also occurs in the siliques of 35S::DEFH28 plants. This indicates that ectopic DEFH28 activity might exert its function similarly to ectopic FUL activity by negatively regulating SHP1 and SHP2 expression. Dehiscence of Antirrhinum capsules differs fundamentally from the Arabidopsis silique as no split along a septum occurs. In contrast, Antirrhinum exhibits a porose dehiscence pattern, where three pores are formed at the tip of the capsule by rupture of the carpel wall (Sutton, 1988; Figure 7a). It will be interesting to find out if MADS-box genes similar to SHP1/2 exist in Antirrhinum and mediate this rupture (Figure 7e,f). Phylogenetic reconstruction revealed that SHP1 and SHP2 belong to the AGAMOUS-subfamily and most likely originated by a duplication event that occurred after the separation of the Arabidopsis and Antirrhinum lineages had taken place (Theißen, 2000b). This supports the idea that the regulation of dehiscence evolved independently in the two species after a conserved mechanism for control of valve differentiation had already been established.


Although carpel wall development in Arabidopsis and Antirrhinum results in two different fruit types, namely capsules and siliques, overexpression of DEFH28 in Arabidopsis causes the same shatterproof silique phenotype as overexpression of FUL, its likely Arabidopsis ortholog. These data show for the first time that carpel wall differentiation might be regulated by a conserved mechanism between different eudicot plant species. This observation is of agronomic importance, as the problem of pod shatter occurs in species with silique and capsule fruit types. Pod shattering causes up to 50% of the potential yield losses in oilseed crop plants such as canola (MacLeod, 1981) and can be addressed by genetically engineering plants ectopically expressing either FUL or DEFH28. Further analysis of MADS-box genes involved in dehiscence zone and valve differentiation will allow fruitful insights into the mechanisms by which different plant species evolved different fruit types.

Experimental procedures

Plant material and growth conditions

Antirrhinum majus plants were grown in the greenhouse at 18–22°C with additional light in the winter. Arabidopsis thaliana plants of ecotype Col were grown under long day photoperiod (16 h light, 8 h dark) at 22°C in the greenhouse. The strong mutant alleles of tfl1–11 (CS6235) and ap1–10 (CS6230), Columbia ecotype, were provided by the Arabidopsis Biological Resource Center (Ohio State University, OH, USA).

Isolation of the DEFH28 cDNA and genomic locus

The DEFH28 cDNA was isolated from a bacterial cDNA library constructed from mRNA from about 1 cm long Antirrhinum majus inflorescences. By applying the Clontech PCR-Select subtractive hybridization technique from Clontech, Palo Alto, transcripts from floral organs were enriched and cloned into the pT-Adv vector (Clontech, Palo Alto, USA). 576 PCR amplified DNA fragments from single bacteria were spotted on nylon membranes. The filters were screened at low stringency (50°C in 3xSSPE for 16 h) with a mixture of individual Antirrhinum MADS-box fragments, each about 250 bp long. Under these conditions one clone (#28) gave a strong signal. Sequencing and database searches revealed that it contained a 280-bp long DNA-fragment spanning a novel MADS-box. In order to isolate the full size of the novel MADS-box gene named DEFH28, a cDNA phage library from young wild-type inflorescences (Sommer et al., 1990) was screened using the DEFH28 fragment as a probe. One full size DEFH28 cDNA was isolated and sequenced.

A genomic library was screened with the DEFH28 cDNA as a probe as described by Zachgo et al., 1997. Under high stringency conditions, a clone with a 13.5 kb insertion comprising the DEFH28 transcription unit plus 3.1 kb of upstream sequences was identified.

DNA sequencing and computer analysis

Determination of DNA sequences was performed on Applied Biosystems (Weiterstadt, Germany) Abi Prism 377 and 3700 sequencers using Big Dye-terminator chemistry by the MPIZ DNA core facility. Pre-mixed reagents were supplied by Applied Biosystems. Oligonucleotides were purchased from GibcoBRL Life Technologies (Eggenstein, Germany). DNA sequence analysis was performed using the GCG program package (Genetics Computer Group Inc., Madison, WI, USA; version 9.0) and for compilation of multiple alignments the program PILEUP therein. Phylogenetic tree analysis was conducted with the neighbor-joining algorithm as described by Winter et al. (1999).

Expression analysis and Scanning Electron Microscopy

RNA blotting and hybridization was carried out as described by Zachgo et al. (1997). Tissue preparation and in situ hybridization was carried out as detailed by Perbal et al. (1996) using a digoxigenin-labeled antisense DEFH28 probe devoid of the MADS-box to avoid cross-hybridization. Digoxigenin-probe preparation was done by T7 transcription of the DEFH28 cDNA, subcloned in the BS KS ± vector (Stratagene), using the DIG labelling mix (Roche Molecular Biochemicals, Mannheim, Germany). For tissue visualization sections were stained with calcofluor (Sigma, Taufkirchen, Germany) and viewed under fluorescent light and under bright field in a Zeiss Axiophot microscope (Zeiss, Olerkochen, Germany).

For scanning electron microscopy (SEM), freshly harvested siliques were shock-frozen in liquid nitrogen, transferred to a cryo-chamber from Oxford instruments (Oxford, UK) and gold coated. Examination of the specimens was carried out with a Zeiss DSM 940 electron microscope at 5 kV. The images were processed and assembled with Adobe Photoshop (Adobe Systems Inc, Miami, USA), Canvas (Deneta Software, Edinburgh, UK) and Corel Draw programs (Corel Corporation, München, Germany).

Generation of A. thaliana plants harboring a DEFH28::GUS construct

A 2.8-kb fragment from the DEFH28 promoter was amplified by PCR with the primer pair BM8: 5′-ACGTCGACATGCCATCAGTCCAGCTAAAG-3′ and BM22: 5′-CAGTCTAGACGCCGACCTTCTTCACTATTTC-3′. The respective 5′-ends of the primers had SalI (BM8) and XbaI (BM22) restriction sites that allowed the integration of the amplified promoter fragment into the binary vector pGPTV-HPT (Becker et al., 1992). Transformation of binary vectors into the Agrobacterium tumefaciens strain GV3101 harboring plasmid pMP90RK (Koncz and Schell, 1986) was conducted by electroporation. A. thaliana plants of the Columbia ecotype were transformed with A. tumefaciens using the standard in planta vacuum infiltration method (Bechtold et al., 1993). Transgenic plants and their progeny were selected with hygromycin (15 µg ml-1) on solid MS plates. Histochemical GUS analysis of tissue from T1 and their respective T2 progeny plants was carried out overnight at 37°C in an assay buffer as described by Heidmann et al. (1998).

Ectopic expression of the DEFH28 cDNA in transgenic A. thaliana plants

For the functional analysis of the DEFH28 gene transgenic A. thaliana plants were generated in which the DEFH28 cDNA was fused to the Cauliflower Mosaic Virus 35S promoter. Amplification of the DEFH28 cDNA with the primer pair BM6: 5′-AATGCTAGCAATATGGGGAGAGGTAGGGTTC-3′ and BM7: 5′-ACGGCTAGCGCGTTATTCATCAACACAAG-3′ yielded DNA fragments with NheI restriction sites at their 5′-ends. These sites allowed the integration of the amplified DEFH28 cDNA into the XbaI restriction sites of the binary vector pBAR-35S (accession#: AJ251014), a derivative from the binary vector pGPTV-BAR (Becker et al., 1992), carrying a CaMV35S promoter/terminator cassette. Selection of transgenic plants carrying the sense construct was done by application of a 0.1% BASTA solution (AgrEvol) to young seedlings.


We would like to thank Thomas Münster for support in phylogenetic analyses and Ming Ai Li for support in Northern experiments. We thank Zsuzsanna Schwarz-Sommer for stimulating discussions and Desmond Bradley and Peter Huijser for critical reading of the manuscript. This work was supported in part by a Lise Meitner scholarship from Nordrhein-Westfalen (NRW) to S.Z.