The D-lineage MADS-box gene OsMADS13 controls ovule identity in rice


(fax +39 02 50315044; e-mail


Genes that control ovule identity were first identified in Petunia. Co-suppression of both FLORAL BINDING PROTEIN 7 (FBP7) and FBP11, two D-lineage genes, resulted in the homeotic transformation of ovules into carpelloid structures. Later in Arabidopsis it was shown that three genes, SHATTERPROOF1 (SHP1), SHP2, and SEEDSTICK (STK), redundantly control ovule identity, because in the stk shp1 shp2 triple mutant ovules lose identity and are transformed into carpel and leaf-like structures. Of these three Arabidopsis genes STK is the only D-lineage gene, and its expression, like FBP7 and FBP11, is restricted to ovules. OsMADS13 is the rice ortholog of STK, FBP7, and FBP11. Its amino acid sequence is similar to the Arabidopsis and Petunia proteins, and its expression is also restricted to ovules. We show that the osmads13 mutant is female sterile and that ovules are converted into carpelloid structures. Furthermore, making carpels inside carpels, the osmads13 flower is indeterminate, showing that OsMADS13 also has a function in floral meristem determinacy. OsMADS21 is most likely to be a paralog of OsMADS13, although its expression is not restricted to ovules. Interestingly, the osmads21 mutant did not show any obvious phenotype. Furthermore, combining the osmads13 and the osmads21 mutants did not result in any additive ovule defect, indicating that osmads21 does not control ovule identity. These results suggest that during evolution the D-lineage gene OsMADS21 has lost its ability to determine ovule identity.


In angiosperms ovules develop as part of the gynoecium, which consists of one or more carpels. Megasporogenesis and megagametogenesis take place in ovules, and after fertilization ovules develop into seeds. Ovule development has mostly been studied in Arabidopsis thaliana, where a large number of mutants have been described and genes controlling ovule development have been characterized (Skinner et al., 2004).

The first two genes that control ovule identity were isolated from Petunia hybrida, and have been named FLORAL BINDING PROTEIN 7 (FBP7) and FBP11 (Angenent et al., 1995; Colombo et al., 1995). They both encode MIKC-type MADS-box transcription factors, a class of transcriptional regulators that has been shown to play key roles in the determination of floral organ identity in many angiosperm species (Becker and Theissen, 2003; Gutierrez-Cortines and Davies, 2000; Kater et al., 2006; Parenicova et al., 2003). FBP7 and FBP11 share 90% identity and were shown to be specifically expressed in ovules. Their concomitant silencing by co-suppression resulted in the homeotic transformation of ovules into carpelloid structures, showing that they are necessary to determine ovule identity (Angenent et al., 1995). Furthermore, ectopic expression of FBP7 or FBP11 in Petunia resulted in the formation of ovules on sepals and petals, suggesting that these genes are sufficient to induce ovule development in flowers (Colombo et al., 1995).

SEEDSTICK (STK) from Arabidopsis is very similar to FBP7 and FBP11, and, like the Petunia genes, is also specifically expressed in ovules, indicating that these are orthologous genes (Rounsley et al., 1995). However, the stk mutant does not show loss of ovule identity. In the stk mutant seeds are irregularly spaced as a result of a drastic enlargement of the funiculus; furthermore, seed detachment is inhibited, because of the absence of proper abscission-zone formation (Pinyopich et al., 2003). Combining the stk mutant with the shatterproof1 (shp1) and shp2 mutants in a stk shp1 shp2 triple mutant showed that these three MADS-box genes have a redundant function in ovule identity determination. Ovule and seed development was almost completely disrupted in this triple mutant, with some of the ovules converted into leaf-like or carpel-like structures (Pinyopich et al., 2003).

STK, SHP1, and SHP2 belong to the phylogenetic AGAMOUS (AG) clade. The AG gene is necessary for plant sexual reproduction because stamens and carpels are absent from ag mutant flowers (Bowman et al., 1989; Yanofsky et al., 1990). AG is also expressed in developing ovules, and together with STK, SHP1, and SHP2 plays a role in ovule identity determination (Pinyopich et al., 2003; Western and Haughn, 1999). Evidence for this came from the analysis of ectopic ovules that develop on sepals in the apetala2 (ap2) mutant. Whereas in the ap2 mutant 40% of the ectopic ovules develop into carpelloid structures, in the ap2 ag double mutant this number is increased to 55%, thereby confirming the role of AG in ovule identity determination.

Recently, Favaro et al. (2003) showed that STK, SHP1, SHP2, and AG can make multimeric complexes in yeast; however, these interactions require SEPALLATA (SEP) MADS-box factors. Genetic evidence to sustain the role of SEP factors in the establishment of ovule identity complexes was obtained by showing that a reduction in SEP activity leads to a loss of normal ovule development similar to what occurs in the stk shp1 shp2 triple mutant.

Phylogenetic analysis of rice (Oryza sativa L. ssp. japonica) MADS-box genes showed that the AG clade contains four members, which are OsMADS3 (Kang et al., 1995), OsMADS58 (Yamaguchi et al., 2006), OsMADS13 (Lopez-Dee et al., 1999), and OsMADS21 (Lee et al., 2003). OsMADS3 and OsMADS58 are most similar in sequence to AG, and the expression of these genes is, like that of AG, restricted to stamens and carpels. Mutant analysis showed that the functions of OsMADS3 and OsMADS58 are similar to those of AG. However, there is clear functional diversification between the two rice genes. OsMADS3 is more important for the specification of stamen identity, whereas OsMADS58 contributes more to conferring floral meristem determinacy and to regulate carpel morphogenesis (Yamaguchi et al., 2006). Neither of these two genes seem to determine carpel identity, as single mutants or combined silencing of OsMADS3 and OsMADS58 did not result in the loss of carpel identity. It might be that in grasses the homeotic function for carpel specification is attributed to a YABBY domain gene named DROOPING LEAF (DL) (Nagasawa et al., 2003; Yamaguchi et al., 2004).

Phylogenetic analyses have associated OsMADS13 and OsMADS21 with the STK lineage (Kramer et al., 2004), and the ovule-specific expression profile of OsMADS13 is very similar to the one observed for STK (Lopez-Dee et al., 1999). Besides this similarity in expression profiles and primary amino acid sequence, these proteins also seem to be conserved in their interactions with other MADS-box factors. MADS-box proteins form dimers, and their interactions can be reliably tested in yeast using the two-hybrid system (Davies et al., 1996; de Folter et al., 2005). Yeast interaction assays showed that, similar to the interactions that were observed between STK and SEP proteins, OsMADS13 also interacts with the SEP-like proteins OsMADS24 and OsMADS45. Furthermore, a similar result was obtained with FBP7 and FBP11 from Petunia, which interact with the SEP-like proteins FBP2, FBP5, and FBP9 (Favaro et al., 2002; Immink et al., 2002). The conservation of these interactions became even more apparent from exchange experiments which showed that the SEP proteins of the three species were all interacting strongly with STK, FBP7, and OsMADS13 (Favaro et al., 2002, 2003).

Here we show that, similar to FBP7, FBP11, and STK, OsMADS13 is also a key regulator of ovule identity. The osmads13 knock-out mutant is completely female sterile and ovules are converted into a reiteration of ectopic carpels or into more amorphous structures that have carpel identity. We also functionally analyzed OsMADS21, which is most likely to be a paralog of OsMADS13. Analysis of OsMADS21 expression showed that its mRNA accumulates at very low levels in the inner two whorls of the flower, and that in ovules its expression overlaps with that of OsMADS13. Knock-out analysis showed that the osmads21 mutant has no aberrant phenotype. Analysis of the osmads13 osmads21 double mutant revealed that loss of OsMADS21 function did not enhance the osmads13 phenotype. These data suggest that only OsMADS13 plays a role in ovule identity determination, and that OsMADS21 has probably lost this function during evolution.


Analysis of AG clade proteins

Phylogenetic analysis of MIKC MADS-box genes belonging to the AG clade has shown that this subfamily falls into two groups, which were named C-lineage, with members typically promoting stamen and carpel identity, and ovule-specific D-lineage genes (Kramer et al., 2004; Yamaguchi et al., 2006; Zahn et al., 2006). Both C- and D-lineages contain representatives from throughout the angiosperms, indicating that these two lineages are derived from ancient gene duplication that occurred after the divergence of the angiosperm and gymnosperm lineages (Kramer et al., 2004). Phylogenetic analysis performed by Kramer et al. (2004) and Yamaguchi et al. (2006) showed that OsMADS13 and OsMADS21 group within the D-lineage, whereas OsMADS3 and OsMADS58 form part of the C-lineage. This division was further supported by the absence of intron 8 in the D-lineage genes. In contrast, Zahn et al. (2006) placed OsMADS13, together with its maize orthologues ZMM1 and ZAG2, in the C-lineage and suggested that intron 8 might have been independently lost in their phylogenies.

Phylogenetic analysis performed by our laboratory using the amino acid sequences of 53 AG family proteins of different species suggests that OsMADS13 and OsMADS21 both belong to the D-lineage (see Figure S1). To obtain more evidence for our hypothesis we carefully analyzed the alignment of the 53 proteins, and interesting information that clearly distinguished members of the two subfamilies was obtained from the N-terminal part of the K-box in the region of the first α-helix (Figure 1b shows the results for a selected number of proteins, see Figure S2 for the complete alignment). The other regions were not as informative because of the high levels of conservation (MADS-box, I-region) or because they were too diverse (C-region). In the N-terminal part of the K-box there are the following residues that distinguish C-lineage from D-lineage proteins: (i) glutamine 105 (Q105; all amino acid positions indicated here are relative to the OsMADS13 protein sequence) is conserved in all the members of the D-lineage that we analyzed, including ZMM1 and ZAG2, whereas this residue is not present in the C-lineage members; (ii) most of the D-lineage proteins and gymnosperm proteins have a non-polar hydrophobic residue at position 106, whereas the monocot C-lineage proteins have a polar residue at this position; (iii) monocot D-lineage proteins have a specific single amino acid insertion at position 90; and finally, at position 113 there is a histidine residue in all the monocot D-lineage proteins, which is not present in the monocot C-lineage proteins. The observed conservation of certain residues in this region of the K-box supports the positioning of OsMADS13 of rice and ZMM1 and ZAG2 of maize in the D-lineage.

Figure 1.

 Insertions in the OsMADS13 and OsMADS21 loci and comparison of amino acid sequences of the K-box of C- and D-lineage MADS-box genes.
(a) Structure of the OsMADS13 and OsMADS21 loci taken from TIGR. The black and empty triangles represent, respectively, T-DNA and Tos17 insertions in the mutant lines used for this work. The regions used as a probe in Northern blot analysis or for amplification in RT-PCR experiments are indicated by black bars.
(b) Alignment of the N-terminal portion of the K-box of OsMADS13, OsMADS21, OsMADS3, OsMADS58 and another 24 proteins from Petunia hybrida (FBP7), Arabidopsis thaliana (STK, AG, and SHP1), Eschscholzia californica (EScaAGL11 and EScaAG1), Dendrobium thyrsiflorum (DthyrAG1 and DthyrAG2), Lilium longiflorum (LMADS2 and LfMADS1), Asparagus virgatus (AVAG1 and AVAG2), Zea mays (ZMM1, ZMM2, ZMM23, ZMM25, ZAG1, and ZAG2), Elaeis guineensis (EgAG1), Cycas edentata (CyAG), Ginkgo biloba (GBM5), Gnetum gnemon (GGM3), and Antirrhinum majus (FAR and PLE). Amino acid residues discussed in this work are indicated.

Expression patterns of OsMADS13 and OsMADS21

The spatial and temporal expression pattern of OsMADS13 was previously described by our laboratory (Lopez-Dee et al., 1999). OsMADS13 is expressed during all phases of ovule development and remains highly expressed during seed development (Figures 2a and 3a–d). To investigate whether OsMADS21 has a similar expression profile we performed reverse transcription-polymerase chain reaction (RT-PCR) analysis, which revealed that this gene, like OsMADS13, is expressed in inflorescences and developing seeds (Figure 2b), with a higher expression in seeds. To investigate the temporal and spatial expression profile of OsMADS21 during flower development we performed in situ hybridization analysis (Figure 3). Relatively weak hybridization signals were observed in developing anthers, carpels, styles/stigmas, and ovules (Figure 3e–h). This low, diffuse expression throughout the two inner whorls of the flower is different from that of other genes belonging to the D-lineage, such as STK, FBP7, FBP11, and OsMADS13, which are all ovule-specifically expressed (Figure 3a–d). During later stages of flower development, OsMADS21 expression inside the ovary is more evident in the inner cell layers of the ovary wall and in the ovule integuments (Figure 3h), where it overlaps with the expression of OsMADS13. From these expression analyses we can conclude that both OsMADS13 and OsMADS21 have an overlapping expression profile during ovule development, although steady-state mRNA levels of the OsMADS21 transcripts seem to be lower than those of OsMADS13.

Figure 2.

 Northern blot and RT-PCR expression analysis of OsMADS13 and OsMADS21.
(a) Northern blot analysis of OsMADS13 expression in roots (R), leaves (L), inflorescences (I), and 5–10 days after pollination (DAP) kernels (K).
(b) RT-PCR analysis (25 cycles) of OsMADS21 expression in roots (R), leaves (L), inflorescences (I), 5–10 DAP kernels (K), inflorescences of the osmads13 mutant (I13), and lane (G) is the control using genomic DNA.
(c) Northern blot analysis of OsMADS13 expression in wild-type (WT) and osmads13 mutant (M) panicles.
(d) RT-PCR analysis (30 cycles) of OsMADS21 expression in wild-type (WT) and osmads21 mutant (M) kernels (5–10 DAP).

Figure 3.

In situanalysis of OsMADS13 and OsMADS21 expression in developing rice flowers.
(a–d) OsMADS13 and (e–h) OsMADS21 expression by in situ hybridization at different stages of rice flower development: a, anther; c, carpel wall; o, ovule; s, style. Scale bars represent 100 μm for all panels.

OsMADS13 controls ovule identity in rice

To analyze the function of OsMADS13 we screened by PCR the rice mutant population at the Pohang University of Science and Technology, Korea, for T-DNA and Tos17 insertions in OsMADS13 (Hirochika, 2001; Jeong et al., 2002). We obtained one line for osmads13 that has a Tos17 insertion in the leader intron 821-bp upstream of ATG (Figure 1a). RNA blot analysis showed that no OsMADS13 transcript could be detected in the homozygous osmads13 mutant, confirming that the Tos17 insertion causes silencing of this gene (Figure 2c).

Osmads13 mutant plants were completely female sterile. Phenotypic analysis showed that approximately 30% of the mutant flowers had a variable number of ectopic styles and stigmas, and that the ovaries were clearly swollen (Figure 4f). To investigate the observed morphologic changes in more detail we performed a detailed histologic analysis, which revealed that the abnormal phenotypes were caused by changes in ovule development. In wild-type plants, the ovule primordium is directly derived from the floral apex and develops into a bitegmic hemianatropous ovule (Figure 4c–e; Lopez-Dee et al., 1999). In contrast with wild type, ovule primordia in osmads13 plants developed into carpelloid structures that later grew out of the carpel, giving rise to the ectopic styles and stigmas (Figure 4h,i). In 70% of the flowers the ovule primordia developed into more amorphous cell structures, of which the identity was difficult to establish by histologic analyses (Figure 4j). These data show that osmads13 plants are sterile as a result of a complete disruption of normal ovule development.

Figure 4.

 Analysis of the osmads13 mutant phenotype.
(a) Wild-type carpel at about 1 day before flowering.
(b) Wild-type kernel at 6 days after pollination.
(c–e) Histologic analysis of wild-type developing carpels.
(f) osmads13 carpel with extra stigmas protruding from the carpel.
(g) osmads13 carpel at later stages (after flowering) in which the indeterminate formation of carpels can be observed (indicated with an arrowhead).
(h, i) Histologic analysis of osmads13 pistils in which the ovule is homeotically converted into a carpel (indicated with arrowheads). The ectopic style–stigma structure that grows out of the pistil is indicated with a red arrow.
(j) Histologic analysis of osmads13 pistils in which the ovule is converted into more amorphous structures (indicated with an arrowhead).
(k–r) In situ hybridization analysis to investigate DL expression in wild-type and osmads13 mutant flowers at different stages of development: (k–n) wild-type flowers, (o–r) mutant flowers.
(p–r) Ectopic DL expression is visible (indicated with arrowheads), showing that this tissue has carpel identity.
(p, q) Carpel structure develops instead of an ovule.
(r) More amorphous structures develop instead of an ovule. a, anther; c, carpel wall; o, ovule; ov, ovary; s, style; st, stigmas. White and black scale bars represent 1 mm and 100 μm, respectively.

Most of the carpels developing in osmads13 flowers aborted rapidly after anthesis; however, in some cases they stayed green and more ectopic carpels emerged from the carpel as a reiterated set of carpels (Figure 4g). This phenotype shows that OsMADS13 is not only important for the determination of ovule identity, but also that loss of OsMADS13 activity leads to indeterminacy of the innermost part of the floral meristem.

In the osmads13 mutant ovules are homeotically transformed into carpelloid structures

In 30% of the osmads13 mutant flowers ovules were clearly transformed into carpelloid structures. However, in most flowers more amorphous structures developed instead of ovules, and the identity of these structures was indeterminable. To understand whether also these structures have carpel identity we performed in situ hybridizations using DL as a probe. In wild-type plants DL is specifically expressed in the leaf midrib and in carpels (Yamaguchi et al., 2004). DL expression is visible from the first appearance of the carpel primordia, whereas no expression was observed in ovules (Figure 4k–n). Our in situ hybridization experiments showed that during early stages of flower development, DL expression in wild-type flowers is identical to the expression profile observed in the osmads13 mutant (Figure 4k,o). Expression was detected in carpel primordia and during carpel development. As expected, in wild-type flowers no expression in the ovule was observed. In contrast DL expression was clearly visible in the osmads13 mutant in both the carpelloid and the more amorphous structures that developed instead of the ovule (Figure 4p–r). This analysis suggests that in osmads13 mutant flowers the ovule is homeotically transformed into tissues that have carpel identity, although they do not always have clear carpel morphology.

Assessing redundancy between OsMADS13 and OsMADS21 during ovule development

To functionally analyze OsMADS21 we obtained from Pohang University a rice line that has a Tos17 insertion in the OsMADS21 leader intron, 491-bp upstream of ATG. RNA blot analysis revealed that this insertion does not silence the OsMADS21 gene (not shown). We obtained another line from the OryGenesDB T-DNA collection of Gènoplante Valor (France) that had a more upstream T-DNA insertion, positioned 13 bp from the 5′-end of the predicted transcription start site (Figure 1a). RT-PCR analysis revealed that no OsMADS21 mRNAs could be detected, indicating that this T-DNA insertion inhibits the expression of this gene (Figure 2d). Morphologic analysis revealed that osmads21 mutants are indistinguishable from wild-type plants. Furthermore, osmads21 mutant plants were fully fertile.

OsMADS13 and OsMADS21 are similar in sequence, and both genes have an overlapping expression profile in ovules, although OsMADS21 seems to be lower expressed than OsMADS13. This suggests that these putative paralogous genes might have a redundant function during ovule development. To assess redundancy we combined the osmads13 with the osmads21 mutant and analyzed a segregating population in the F2. In the osmads13 single mutant there were two distinct phenotypes: (i) flowers in which ovules were homeotically converted into carpels with styles and stigmas, and (ii) flowers with ovules converted into carpel tissue with a more amorphous structure. Analyzing flowers of four osmads13 osmads21 double mutant plants showed that the loss of OsMADS21 function did not have any additive effect on the osmads13 floral phenotype. The two phenotypes were observed in the same frequencies as in the osmads13 single mutant, and no other phenotypic effects were observed in these flowers. This result indicates that OsMADS21 does not control ovule identity in rice flowers.

To exclude the possibility that the observed lack of redundancy between these two genes results from the fact that in the osmads13 single mutant OsMADS21 is also silenced, which means that the expression of OsMADS21 is regulated by OsMADS13, an RT-PCR experiment was performed. This analysis showed that OsMADS21 is expressed in inflorescences of the osmads13 mutant (Figure 2b). However, the expression was lower than in wild type, which is probably because of the absence of ovules in osmads13 flowers, which is the tissue where OsMADS21 transcripts are most abundant.


OsMADS13 and OsMADS21 are paralogous MADS-box genes

Phylogenetic analysis of MADS-box transcription factors belonging to the AG clade shows that gene duplications were critical for shaping this subfamily. An ancient duplication after the divergence of the angiosperm and gymnosperm lineage resulted in two groups that were named the C- and the D-lineages (Kramer et al., 2004; Yamaguchi et al., 2006), referring to the C-function genes involved in stamen and carpel development (Coen and Meyerowitz, 1991) and to the D-function genes involved in ovule identity determination (Colombo et al., 1995). As this classification is used in many phylogenetic studies we decided to use the same nomenclature, although, as suggested correctly by Zahn et al. (2006), because orthology does not always coincide with functional equivalence, this type of classification might be a bit misleading.

The latest most extensive phylogenetic analyses of the AG clade MADS-box genes, presented by Kramer et al. (2004) and Zahn et al. (2006), were contradictory as to whether OsMADS13 belongs to the C- or D-lineage, as both research groups obtained a position for the locus but these were in disagreement. However, both studies positioned OsMADS21 in the D-lineage. To understand whether these two genes were more ‘recent’ paralogs, or were derived from the ancient duplication after the divergence of the angiosperm and gymnosperm lineages, we performed a phylogenetic analysis that produced a tree in which OsMADS13 grouped together with OsMADS21. To obtain more evidence to confirm that the position of these two genes is correct we analyzed the sequence of C- and D-lineage proteins. Amino acid sequence features observed in the K-box sustain our phylogenetic analysis placing both OsMADS13 and OsMADS21 in the D-lineage, and indicate that they are most likely to be the result of a more recent gene duplication event that occurred within the D-lineage.

Conservation of class-D MADS-box gene function in monocot and dicot plants

Based on studies in Petunia, Colombo et al. (1995) proposed to extend the ABC model of flower development by a new class of genes that they named ‘D’ class. Two Petunia MADS-box genes, FBP7 and FBP11, were shown to control ovule identity because they were both necessary and sufficient to control ovule development in Petunia flowers. Overexpression of these genes resulted in ectopic ovule formation on sepals and petals, whereas co-suppression of both genes caused the homeotic transformation of ovules into carpelloid structures (Angenent and Colombo, 1996). In Arabidopsis three closely related MADS-box genes STK, SHP1, and SHP2 redundantly control ovule identity (Pinyopich et al., 2003). In the stk shp1 shp2 triple mutant ovules are converted into leaf-like or carpel-like structures, whereas in single or double mutant combinations of these genes ovule identity was not affected (Pinyopich et al., 2003). Furthermore, ectopic expression of these MADS-box genes also resulted in ovule formation on sepals in Arabidopsis (Battaglia et al., 2006; Favaro et al., 2003; Pinyopich et al., 2003). These data show that in Arabidopsis both C-lineage genes, SHP1 and SHP2, and the D-lineage gene STK are controlling ovule identity.

Phylogenetic analysis showed that OsMADS13 belongs to the D-lineage. Furthermore, OsMADS13 is, like the other ovule identity D-lineage genes (i.e. FBP7, FBP11, STK), specifically expressed in ovules, which suggests that there might be functional conservation between these genes. This hypothesis was confirmed by analyzing the osmads13 mutant. In this mutant ovules are converted into carpelloid structures, although differences in phenotypes were observed between flowers on the same plant. In some flowers there was a complete conversion of ovules into carpels, whereas in other flowers the ovules were converted into structures that did not clearly resemble a carpel organ, although in situ hybridizations using DL confirmed that these structures have carpel identity. The variation in phenotype between osmads13 flowers might result from a partial redundancy with other genes. The most likely candidate for such redundant function is OsMADS21. However, analysis of the osmads21 single and the osmads13 osmads21 double mutant clearly showed that OsMADS21 has no function in ovule identity determination, and is therefore not redundant with OsMADS13. In Arabidopsis also, C-lineage genes AG, SHP1, and SHP2 are involved in ovule identity determination. Therefore, it might well be that in rice the C-lineage genes OsMADS3 and/or OsMADS58 also contribute to the establishment of ovule identity. Combining the osmads3 and osmads58 mutants with the osmads13 mutant might reveal redundancy between these genes.

OsMADS13 has a function in meristem determinacy

In Arabidopsis loss-of-function mutations in the class-C gene AG result in the homeotic conversion of stamens into petals, and in the center of the flower the carpel is replaced by a new indeterminate ag flower (Yanofsky et al., 1990). This phenotype shows that AG has two functions: one is the determination of stamen and carpel identity, and the other is to establish meristem determinacy in the center of the floral meristem. AG seems to establish determinacy of the floral meristem by repressing the meristem maintenance factor WUSCHEL (WUS; Laux et al., 1996). It has recently been proposed that the repression of WUS is established via a negative feedback loop in which WUS and the floral meristem identity gene LEAFY (LFY; Weigel et al., 1992) activate AG, which in turn represses WUS (Lenhard et al., 2001; Lohmann et al., 2001). However, the mechanism by which AG represses WUS is not yet clear.

In Arabidopsis after establishment of carpel identity the ovules arise from the placenta, which develops from the septum that separates the two fused carpels. Therefore, as the ovules are not directly formed from the floral meristem, at this phase of flower development, a meristem determinacy function is not required for genes that determine ovule identity. Rice flower anatomy and its developmental program clearly differs from that of Arabidopsis, and therefore the situation seems to be different. As in Arabidopsis the two C-lineage genes OsMADS3 and OsMADS58 both have a floral meristem determinacy function, although the latter seems to be the main player in controlling meristem activity (Yamaguchi et al., 2006). However, when the carpel develops, the floral meristem is not yet completely consumed in the center. From this innermost part of the rice flower the ovule primordium develops directly from the floral meristem. The analysis of the osmads13 mutant showed clearly that OsMADS13 has a function in the determinacy of this part of the meristem, as the loss of its activity gave rise to indeterminate carpel formation in the center of the flower. This indeterminacy differs from the type that is observed in ag mutant flowers, where loss of determinacy results in a complete reiteration of the floral meristem. The same holds for the rice osmads58 knock-down mutant, although in this mutant there is not a complete reiteration of the floral meristem but repeated sets of lodicules, stamens/ectopic lodicules, and carpel-like organs are formed (Yamaguchi et al., 2006). The carpel in a carpel phenotype that we observe in the osmads13 mutant might be explained by the fact that the ovule is transformed in a carpel, which itself tries to make an ovule, but this is transformed into a carpel, and so on, which means that the indeterminacy phenotype is a consequence of the homeotic phenotype, although this type of indeterminacy was not observed in similar mutants in Arabidopsis and Petunia. Furthermore, the inactivation of class-D gene expression probably does not allow the expression of genes controlling perianth organ identity, including class-B genes that control stamen identity in the central part of the flower, which could explain why the indeterminacy leads to a repetition of only carpels.

Recently, studies in transgenic Petunia plants by Ferrario et al. (2006) showed that ectopic co-expression of the class-D and -E genes, FBP11 and FBP2, respectively, caused early arrest in development at the cotyledon stage. Molecular analysis of these transgenic plants revealed a possible combined action of FBP2 and FBP11 in repressing the Petunia WUS homolog, TERMINATOR. This also suggests that in Petunia, where multiple ovules develop from the placenta that is directly formed from the floral meristem (Angenent et al., 1995), the situation might be similar to the one that we observed in rice. It will be interesting to test whether OsMADS13 also controls this function, together with the SEP-like proteins, by repressing WUS homologous genes in rice. Furthermore, it would, from an ‘evo-devo’ point of view, be interesting to investigate if this meristem determinacy function of class-D genes is conserved in all species where ovules more or less directly develop from the floral meristem.

Experimental procedures

Phylogenetic analysis

M-, I-, K-, and C- regions of MADS-box proteins were aligned using the multiple alignment mode of clustalx (Thompson et al., 1997), and were revised manually with genedoc (Nicholas et al., 1997). Phylogenetic tree construction was performed as described previously (Pelucchi et al., 2002).

Plant materials

Oryza sativa L. ssp. japonica of cultivars Nipponbare and Dongjin, and mutant lines 2D-003-37 (Tos17 insertion in OsMADS13, cv. Dongjin), 2D-406-05 (Tos17 insertion in OsMADS21, cv. Dongjin), and AEHH03 (T-DNA insertion in OsMADS21, cv. Nipponbare) were grown in a greenhouse under cycles of 12-h light at 28°C and 12-h dark at 22°C.

Genotyping of mutant plants

For genotyping the knock-out line 2D-003-37 we used the following primers: Osp66 (forward) with Osp67 (reverse) to detect the OsMADS13 wild-type allele, and Osp66 with Osp14 (Tos17R) for the osmads13 mutant allele. For the mutant line AEHH03 we used the primers Osp65 (forward) with Osp41 (reverse) to detect the OsMADS21 wild-type allele, and Osp58 (T-DNA LB) with Osp41 for the osmads21 mutant allele detection.

Northern blot analysis

Total RNA was isolated as described by Verwoerd et al. (1989). Total RNA (15 μg) was loaded on a 0.03% formaldehyde, 1.5% agarose gel into 3-(N-morpholino)propanesulphonic acid (MOPS) electrophoresis buffer, blotted onto a Hybond N+ membrane (Amersham, and hybridized with a cDNA probe labeled with [ α-32P]dATP. Probe primers: Osp18 (forward) and OL160 (reverse) for OsMADS13, and Osp20 (forward) and Osp44 (reverse) for OsMADS21.


cDNA was synthesized from a maximum of 1 μg of total purified RNA according to the instructions provided with i-Script kit (Bio-Rad,; the primers used for the OsMADS21 RT-PCR were Osp20 (forward) and Osp44 (reverse).

In situ hybridization

Digoxygenin-labeled antisense RNA probes were generated by in vitro transcription according to the instructions provided with the DIG RNA Labeling Kit (SP6/T7; Roche, cDNA used for probe transcription was synthetized with the following primers: Osp18 (forward) and Osp19 (reverse with T7 promoter) for OsMADS13; Osp20 (forward) and Osp21 (reverse with T7 promoter) for OsMADS21; Osp70 (forward) and Osp71 (reverse with T7 promoter) for DL. Developing inflorescences of wild-type rice plants were fixed and embedded in Paraplast Plus embedding medium; a 6-μm-thick section was hybridized as described by Cañas et al. (1994). Stringent formamide washings were performed after hybridization, as described by Kouchi and Hata (1993). OsMADS13 and DL-hybridized sections were mounted with a coverslip and subsequently observed using a Zeiss Axiophot D1 microscope ( equipped with differential interface contrast (DIC) optics. Images were captured on an Axiocam MRc5 camera (Zeiss) using the axiovision program (version 4.4). OsMADS21-hybridized sections were mounted with a coverslip, observed with a Leica DM LB microscope and images were captured using a Leica DFC280 camera (

Light microscopy

For histologic analysis, materials were vacuum infiltrated and fixed in 3.7% formaldehyde, 50% ethanol, and 5% acetic acid overnight, washed twice for 5 min in 70% ethanol, and dehydrated in ethanol series. Infiltration and polymerization in Technovit resin ( was performed according to the manufacturer’s instructions; 6-μm-thick sections were stained in 0.1% toluidine blue, mounted with a coverslip and observed with a Leica DM LB microscope. Images were captured using a Leica DFC280 camera.

Primers used



We thank Silvia Pozzoli, Dong-Hoon Jeong and Simona Masiero for their help in some of the experiments. LD was supported by the University of Milan through a fellowship from the PhD school in Genetics and Biomolecular Sciences.