Genetic interaction of OsMADS3, DROOPING LEAF, and OsMADS13 in specifying rice floral organ identities and meristem determinacy

Grass plants develop unique floral patterns that determine grain production. However, the molecular mechanism underlying the specification of floral organ identities and meristem determinacy including the interaction among floral homeotic genes, remains largely unknown in grasses. Here, we report the interactions of rice ( Oryza sativa ) floral homeotic genes, OsMADS3 (a C-class gene), OsMADS13 (a D-class gene) and DROOPING LEAF ( DL ), in specifying floral organ identities and floral meristem determinacy. The interaction among these genes was revealed through the analysis of double mutants. osmads13-3 osmads3-4 displayed a loss of floral meristem determinacy and generated abundant carpelloid structures containing severe defective ovules in the flower center, which were not detectable in the single mutant. In addition, in situ hybridization and yeast two-hybrid analyses revealed that OsMADS13 and OsMADS3 did not regulate each other’s transcription or interact at the protein level. This indicates that OsMADS3 plays a synergistic role with OsMADS13 in both ovule development and floral meristem termination. Strikingly, osmads3-4 dl-sup6 displayed a severe loss of floral meristem determinacy, and produced supernumerary whorls of lodicule-like organs at the forth whorl, suggesting that OsMADS3 and DL synergistically terminate the floral meristem. Furthermore, the defects of osmads13-3 dl-sup6 flowers appeared identical to those of dl-sup6 , and the OsMADS13 expression was undetectable in dl-sup6 flowers. These observations suggest that DL and OsMADS13 may function in the same pathway specifying the identity of carpel/ovule and floral meristem. Collectively, we propose a model to illustrate the role of OsMADS3 , DL and OsMADS13 in the specification of flower organ identity and meristem determinacy in rice.

The proposed genetic ABC model explains how three classes of genes (A, B, and C) work together in specifying floral organ identities (Coen and Meyerowitz, 1991). In Arabidopsis, A (APETALA1, AP1; AP2) alone determines sepals, A and B (AP3; PISTILLATA, PI) together specify petals, B and C (AGAMOUS, AG) define stamens, and C alone defines carpels (Coen and Meyerowitz, 1991). Subsequently, two additional classes of genes (D and E) have been included in the modified ABC model.
As one of the largest families in flowering plants, the grass family (Poaceae) contains many economically important crops such as rice (Oryza sativa), barley (Hordeum vulgare) and maize (Zea mays) (Linder and Rudall, 2005). These crops have unique floral organization and morphology which are distinct from those of eudicots and even other monocots (Grass Phylogeny Working Group, 2001;Rudall et al., 2005;Whipple et al., 2007). Spikelet is the structural unit of grass flowers, and each spikelet consists of a varied number of bract-like organs, glumes, and florets. A rice spikelet consists of two pairs of sterile glumes (i.e. rudimentary glumes and empty glumes) and one floret that contains one lemma, one palea in whorl 1, two lodicules in whorl 2 interior to the lemma, six stamens in whorl 3 and a carpel in whorl 4 (Yuan et al., 2009；Zhang andWilson, 2009).
Although grass flowers are essential for producing grains, the underlying molecular basis that specifies grass floral organs still remains less understood (Clifford, 1987;Whipple et al., 2007). While the ABCDE model is thought to be partially applicable in explaining the grass floral development, grasses have diversified genetic components in specifying the identity of floral organs and meristem (Thompson and Hake 2009). For example, loss-of-function mutants of the orthologs of Arabidopsis AP3, in maize (Silky1) and in rice (SUPERWOMEN1, SPW1 or OsMADS16), display homeotic transformations of stamens to carpels, and lodicules to lemma-or palea-like structures, suggesting the conservation of class B genes between grasses and Arabidopsis (Ambrose et al., 2000;Nagasawa et al., 2003;Whipple et al., 2007).
Grasses have duplicated and subfunctionalized C-class genes (Kramer et al., 2004;Zahn et al., 2006). For example, rice has two AG homologs, OsMADS3 and OsMADS58 (Kramer et al., 2004). OsMADS3 plays key roles in both stamen identity specification and late anther development, while OsMADS58 is crucial for specifying floral meristem determinacy and carpel architecture (Yamaguchi et al., 2006;Hu et al., 2011). Similarly, there is a pair of AG homologs in maize: zag1 (zea agamous1) and zmm2 (Zea mays mads2). The zag1 gene has been shown to determine floral meristem determinacy, while the biological function of zmm2 remains unclear (Mena et al., 1996).
In rice OsMADS13 is a D-class gene which is orthologous to Arabidopsis SEEDSTICK (STK), and FLORAL BINDING PROTEIN 7 (FBP7) and FBP11 in petunia. Co-suppression of FBP7 and FBP11 causes the conversion of ovules into carpelloid organs (Colombo et al., 1995). The osmads13 mutants are associated with homeotic transformation of ovules into carpelloid structures and indeterminate flowers (Dreni et al., 2007;Yamaki et al., 2010). This is in contrast to the mutation of the Arabidopsis STK gene which does not display altered ovule identity (Pinyopich et al., 2003). In Arabidopsis, AG, STK, SHP1 (SHATTERPROOF1) and SHP2 are grouped in the monophyletic AG-like clade and have been shown to be involved in the ovule identity specification. STK is the only D-lineage gene and expressed in the ovule. stk single mutants develop a slightly abnormal ovule with a defect of the funiculus development, while stk shp1 shp2 triple mutant demonstrate the conversion of ovules into leaf-like or carpel-like organs (Favaro et al., 2003;Pinyopich et al., 2003). Furthermore, STK, SHP1, SHP2, and AG were shown to form multimeric complexes in yeast in the presence of SEP MADS-box factors, and the defect of ovule development in sep1/SEP1 sep2 sep3 is similar to that in shp1 shp2 stk triple mutant genes in specifying flower organ identity. More recently AGAMOUS-LIKE6 (AGL6) genes in monocots and dicots have been also shown to play key roles in specifying floral organ and meristem identity (Hsu et al., 2003;Fan et al., 2007;Ohmori et al., 2009;Reinheimer and Kellogg, 2009;Rijpkema et al., 2009;Thompson et al., 2009;Li et al., 2010;Viaene et al., 2010). AGL6-like genes are ancient and widely distributed in gymnosperms and angiosperms and form a sister clade to SEP-like genes (Purugganan et al., 1995;Theissen et al., 2000;Becker and Theissen, 2003;Zahn et al., 2005). Mutations in AGL6 homologous genes in grasses result in defective floral organ identity and meristem determinacy (Ohmori et al., 2009;Thompson et al., 2009;Li et al., 2010).
Although several genes reported to play roles in specifying flower development in rice, their genetic interactions remain largely unknown. In this study, we characterized the genetic interaction of OsMADS3, DL and OsMADS13 in specifying floral organs and floral meristem determinacy and provided new insights into the molecular mechanisms that regulate flower development in rice.

Identification of new alleles of OsMADS13, OsMADS3 and DL
To identify rice mutants with floral defects, we screened a population of rice mutants for defective flowers in the japonica subspecies 9522 background (Oryza sativa L. ssp. Japonica) treated by 60 Co γ-ray (280 Gy) (Chen et al., 2006b). One mutant line displaying complete female sterility was identified. Genetic analysis and map-based cloning indicated that this mutant has one base deletion in the fifth exon in OsMADS13 (Os012g10540) (Figure S1A), causing a frameshift at 132th amino acid and the formation of premature stop-codon. OsMADS13 expression was specifically reduced in pistils of the mutant ( Figure S1B). As the first two mutants of OsMADS13 (osmads13-1 and osmads13-2) have been reported (Dreni et al., 2007;Yamaki et al., 2010), and a genetic analysis indicated that our mutant is allelic to the reported osmads13-1, we named this mutant as osmads13-3. This mutation is not associated with obvious alteration of outer three whorl organs, some osmads13-3 flowers (31%) displayed three or four stigmas (n=121) ( Figures 1A and 1Q), instead of two stigmas in wild-type flowers. Like the osmads13-1 mutant, osmads13-3 showed complete female sterility with aborted ovule development ( Figures 1B and 1Q) and carpelloid structures ( Figures S1F and S1G). In addition, the ectopic expression of DL was observed in the carpelloid structure of osmads13-3 (Figures 2A to 2F), suggesting that these ectopic structures have the carpel identity.
Sequence analysis showed the insertion of one DNA fragment at the second intron of the DL gene ( Figure S2A) which abolished the expression of DL in the mutant ( Figure   S2B). Because of five previously identified strong dl alleles (dl-sup1 to dl-sup5) (Nagasawa et al., 2003;Yamaguchi et al., 2004), we named this mutant dl-sup6. Like the severe dl mutants, dl-sup6 displayed a phenotype of drooping leaves ( Figure S2C To further elucidate the mechanism of OsMADS13 and OsMADS3 in floral development, yeast two hybridization experiment was performed, and we observed no interaction of these two proteins judged by the growth condition in selective culture medium ( Figure S3). RNA in situ hybridization analysis indicated that the OsMADS13 expression pattern was not obviously altered in osmads3-4 at stage Sp8 when the formation of ovule ( Figures 3A-3C), and OsMADS3 mRNA signal was not obviously changed in osmads13-3 ( Figure 3G). Thus OsMADS13 and OsMADS3 do not seem to influence each other at transcriptional level.  Figure 4E). This phenotype implies a severe loss of floral meristem determinacy which was further confirmed by the in situ hybridization of OSH1 mRNA ( Figure 4H). SEM observation showed that at early stage of Sp8, the osmads3-4 dl-sup6 flower violated from the normal development process and formed indeterminate floral meristem in the flower center ( Figure 4G).Transverse section analysis indicated that these underdeveloped tissues were morphologically close to those of lodicules with the characteristic pattern of vascular bundles (Figures 4I to 4K). Also this indication was confirmed by the SEM observation that the morphology of epidermal cells of these underdeveloped tissues appeared similar to those of lodicules ( Figures 4L and 4M). Meanwhile, the mRNA of rice B-class gene SPW1 (OsMADS16), which accumulates in wild-type lodicules and stamens ( Figure 4N; Nagasawa et al., 2003), was detectable in these undifferentiated organs ( Figure 4O).
This was combined with the presence of transcripts of the putative class A gene

Analysis of the interaction between OsMADS13 and DL
To determine the relationship between OsMADS13 and DL, we constructed the Therefore we proposed that OsMADS13 and DL may function in the same pathway in specifying carpel/ovule identity and floral determinacy, and DL may act upstream of OsMADS13, positively regulating OsMADS13 expression, while OsMADS13 may repress the ectopic expression of DL in the ovule.

Rice has conserved and diversified mechanism controlling the ovule identity
The ovule development is of importance in plant life cycle. Ovule is the source of the megagametophyte and the precursors of seeds, consisting of the nucleus, integument(s) and funiculus (Reiser and Fischer, 1993;Colombo et al., 2008). Previous studies in Petunia, Arabidopsis and rice revealed that the MADS-box genes belonging to the AG clade are necessary for specifying ovule identity.
In rice, the AG clade contains four MADS-box members: two C-lineage genes OsMADS3 and OsMADS58, two D-lineage genes OsMADS13 and OsMADS21 (Kramer et al., 2004;Zahn et al., 2006). The expression of OsMADS13 is restricted in the ovule, which is very similar to that of STK, FBP7 and FBP11. In contrast, OsMADS21 is mainly expressed in developing seeds (Lee et al., 2003;Dreni et al., 2007), and was thought to play a minor role in controlling ovule development (Dreni et al., 2007). Grass species including maize, wheat (Triticum aestivum), barley and rice have duplicated C class genes (Mena et al., 1996;Kramer et al., 2004;Yamaguchi et al., 2006;Zahn et al., 2006). To date, there is no evidence indicating that class C genes are required for carpel identity in grasses ( Thompson and Hake, 2009). In rice, analyses of mutations of OsMADS3 and knockdown of OsMADS58 OsMADS3 was also shown to directly regulate the expression of MT-1-4b, which encodes a type 1 small cysteine-rich and metal-binding protein with superoxide anion and hydroxyl radical scavenging activity, suggesting that OsMADS3 is a key Similarly, two duplicated AG homologs (zag1 and zmm2) are present in the maize genome, and mutations in zag1 cause loss of floral meristem determinacy in the ear, without obvious alteration of floral organ identity (Mena et al., 1996). Currently no mutants of zmm2 have been identified, but the expression pattern of zmm2 is in agreement with that of class C function (Mena et al., 1996). Here, our genetic analysis of double mutant osmads13-3 osmads3-4 indicated that OsMADS3 plays a critical role in ovule formation and floral meristem determinacy redundantly with OsMADS13 ( Figure 6). These data also support that the C-class and D-class genes probably retain their function even though they underwent multiple subfunctionlization events and several neofunctionalization after duplication within AG clade (Rijpkema et al., 2010).
In rice a YABBY domain gene DL was shown to be crucial for carpel specification (Nagasawa et al., 2003;Yamaguchi et al., 2004), that is different from the well-known ABC genes. In addition, the role of DL is distinct from the closely related YABBY gene CRC of Arabidopsis, which plays a mild role in carpel development (Yamaguchi et al., 2004;Alvarez and Smyth, 1999;Bowman and Smyth, 1999). Analysis of Furthermore, in this study, our finding suggests that OsMADS13 and DL specify carpel/ovule and floral meristem identity in the same pathway. Beside the observation that osmads13-3 dl-sup6 displayed flower defects similar to that of dl-sup6, no obvious OsMADS13 expression was detectable in dl-sup6 flowers, and DL transcripts were ectopically detected in osmads13-3 flowers, suggesting that DL may directly or indirectly regulate OsMADS13 expression. In other words, loss of OsMADS13 expression in dl-sup6 may be resulted from the altered carpel/ovule identity in dl-sup6, or DL regulates carpel/ovule and meristem identity by controlling the OsMADS13 expression. Furthermore, the ectopic expression of DL in osmads13-3 is likely caused by the altered identities of ovule and meristem, and OsMADS13 may indirectly restrict the expression of DL in the ovule (Figure 6).

Regulation of rice floral meristem termination
Floral organs are formed by a floral meristem, a pool of pluripotent and dividing cells (Prunet et al., 2009). The regulation of floral meristem seems to be widely conserved among angiosperms (Ferrario et al., 2004;Prunet et al., 2009). In Arabidopsis, AG is a master regulator terminating floral meristem by turning WUSCHEL (WUS) off (Sieburth et al., 1998;Sun et al., 2009). In addition to homeotic transformations of stamens into petals, strong ag alleles (ag-1 to -3) showed a -complete loss of floral meristem determinacy, and the carpel is replaced by a new flower (Bowman et al., 1989;Yanofsky et al., 1990;Bowman et al., 1991). The genomes of both eudicot and monocot species including Antirrhinum, rice, maize and barley contain duplicated and subfunctionalized AG homologs (Zahn et al., 2006). Recent analysis of osmads3-4 osmads58 double mutant suggests that two rice C class genes OsMADS3 and

Plants materials
The mutants osmads13-3 and dl-sup6 were identified from M2 population of 9522 Prior to the analysis, osmads13-3, osmads3-4 and dl-sup6 were all crossed with wild-type 9522 three times respectively. Double mutant plants were isolated by phenotype observation and verified by genotyping with primers3TPF/3TPR and 13TPF/13TPR for osmads3-4 and osmads13-3, respectively (Supplemental Table 1).
Mutant and wild-type rice plants were planted in paddy fields under normal condition in shanghai or greenhouse in Shanghai Jiao Tong University, China.

Histological analysis and microscopy observation
Materials were fixed and dehydrated as described by Li. et al., (2006). For histological analysis, tissues were substituted by xylene and embedded in paraplast plus. Then, materials were sectioned to 8 μm thick and stained with toluidine blue and photographed using a Nikon E600 microscope (Nicon Corporation) and a Nikon DXM1200 digital camera (Nicon Corporation). Scanning electron microscopy (SEM) observation was performed with JSM-6360LV (JEOL) as described previously (Li et al., 2006). The dividing of the ovule stage refers to previous report (Lopez-Dee et al., 1999).

Yeast two hybridization
The MATCHMAKER GAL4 Two-Hybrid System (CLONTECH, Japan) was used to detect the interaction between OsMADS3 and OsMADS13. cDNA fragments encoding IKC domain of OsMADS3 and OsMADS13 were amplified by RT-PCR with primers 3YF/3YR, 13YF/13YR respectively (Supplemental Table 1), and cDNA fragment encoding IKC14 domain of OsMADS6 was amplified by RT-PCR with primers 6YF/6YR (Supplemental Table 1). Then, these cDNA fragments were cloned into pGBKT7 and pGADT7 to fuse with the BD (bait domain) and AD (activation domain) of GAL4 respectively. Recombinant vectors were named AD-13, BD-13, AD-3, BD-3, AD-6, BD-6 respectively. Self activation was assayed on SD plates (-Leu/-His/+3-AT or -Trp/-His/+3-AT). Then, combination of AD-3, BD-13 and BD-3, AD-13 was transformed into yeast strain AH109 simultaneously according to the protocol. The transformants co-transformed with plasmids encoding OsMADS6 and OsMADS13 were used as a positive control (Favaro et al., 2002), and the transformants containing plasmids pGADT7 and pGBKT7 were used as a negative control. The interaction was judged by the growth condition on selective mediums (-Trp/-Leu/-His/+3-AT) according to the protocol from the company.     Q and R, Floral diagrams of dl-sup6 (Q) and osmads3-4 dl-sup6 (R).
Supplemental Figure 3 Supplemental Figure 3. OsMADS3 does not interact with OsMADS13 in yeast cells.
The transformants co-transformed with plasmids encoding OsMADS6 and OsMADS13 as the positive control, could grow on the selective medium plate. While the transformants containing both plasmids pGADT7 and pGBKT7 as the negative control, could not grow on the selective medium plate in the same condition. Like the negative control, transformants harboring plasmids encoding OsMADS3 and OsMADS13 could not grow on the selective medium plate.
SD/2-represents the SD medium containing no Trp and Leu, SD/3-/3-AT+ represents the SD medium without His, Trp and Leu, but containing 3-AT with optimized concentration.