Mutant analyses in different eudicotyledonous flowering plants demonstrated that SEPALLATA-like MADS-box genes are required for the specification of sepals, petals, stamens and carpels, and for floral determinacy, thus defining class E floral organ identity genes. SEP-like genes encode MADS-domain transcription factors and constitute an angiosperm-specific gene clade whose members show remarkably different degrees of redundancy and sub-functionalization within eudicots. To better understand the evolutionary dynamics of SEP-like genes throughout the angiosperms we have knocked down SEP-like genes of rice (Oryza sativa), a distant relative of eudicots within the flowering plants. Plants affected in both OsMADS7 and OsMADS8 show severe phenotypes including late flowering, homeotic changes of lodicules, stamens and carpels into palea/lemma-like organs, and a loss of floral determinacy. Simultaneous knockdown of the four rice SEP-like genes OsMADS1, OsMADS5, OsMADS7 and OsMADS8, leads to homeotic transformation of all floral organs except the lemma into leaf-like organs. This mimics the phenotype observed with the sep1 sep2 sep3 sep4 quadruple mutant of Arabidopsis. Detailed analyses of the spatial and temporal mRNA expression and protein interaction patterns corresponding to the different rice SEP-like genes show strong similarities, but also gene-specific differences. These findings reveal conservation of SEP-like genes in specifying floral determinacy and organ identities since the separation of eudicots and monocots about 150 million years ago. However, they indicate also monocot-specific neo- and sub-functionalization events and hence underscore the evolutionary dynamics of SEP-like genes. Moreover, our findings corroborate the view that the lodicules of grasses are homologous to eudicot petals.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Flowering plants (angiosperms) have evolved a tremendous diversity of floral structures since they originated about 200 million years ago (Wikstrom et al., 2001; Soltis and Soltis, 2004; Endress, 2006). The elucidation of developmental genetic pathways in eudicot model species such as Arabidopsis thaliana, Petunia hybrida and Antirrhinum majus has provided ample evidence that changes in the function of key regulatory genes contributed significantly to the evolution of new morphologies (Cronk, 2001; Frohlich, 2003; Irish, 2003).
In the Arabidopsis sep1/2/3 triple loss-of-function mutant, all petals, stamens and carpels are converted into sepal-like organs, and flower development becomes indeterminate; thus the phenotype of sep1/2/3 triple mutants largely resembles the phenotype of class bc (i.e. ap3 ag or pi ag) double mutants (Pelaz et al., 2000, 2001; Honma and Goto, 2001). In the sep1/2/3/4 quadruple mutant, all flower organs are converted into organs resembling vegetative leaves, are arranged in spiral phyllotaxis and again, appear in indeterminate number, thus resembling class abc triple loss-of-function mutants (Ditta et al., 2004). In the sep mutants, however, the early expression of the class A, B and C floral homeotic genes is not affected. These findings indicate that some SEP-like (AGL2-like) genes are required for class A, B and C gene functions in the specification of flower organ identity and in conferring determinate growth to the flower, but they are not just activators of class ABC genes, but a new class of organ identity genes. Therefore, they were added as ‘class E genes’ to the ABC model (Theissen, 2001).
Phylogenetic analyses revealed that SEP-like genes are monophyletic, and that they can be divided into two subclades, SEP1/2/4-like genes (AGL2/3/4 clade sensu) (Zahn et al., 2005) and SEP3-like genes (AGL9 clade), which were generated by a gene duplication (or whole genome duplication) that probably predated the origin of the most recent common ancestor of extant angiosperms (Theissen et al., 1996; Becker and Theissen, 2003; Zahn et al., 2005). As SEP-like genes are required for specifying the ‘floral state’ by contributing to floral organ and meristem identity (Zahn et al., 2005), they may have played a critical role during the evolutionary origin of the flower and are thus of utmost evolutionary interest.
Despite their evolutionary importance, due to technical limitations the function of few SEP-like genes outside of Arabidopsis have already been studied, including genes from core eudicots, such as tomato (Solanum lycopersicum) (Pnueli et al., 1994; Ampomah-Dwamena et al., 2002), petunia (Petunia hybrida) (Angenent et al., 1994), and gerbera (Gerbera hybrida) (Kotilainen et al., 2000; Uimari et al., 2004). These investigations suggested a general conservation of SEP-like gene function in eudicots. However, two Gerbera SEP-like genes, GERBERA REGULATOR OF CAPITULUM DEVELOPMENT1 (GRCD1) and GRCD2, revealed whorl-specific subfunctionalization, with GRCD2 acting in whorl 4 and GRCD1 acting in whorl 3 (Kotilainen et al., 2000; Uimari et al., 2004). Furthermore, GRCD2 has been found to play a role in regulating inflorescence development in addition to flower development (Uimari et al., 2004).
The functions of monocot SEP-like genes are unknown, except for OsMADS1, for which a class E gene function has been described (Agrawal et al., 2005; Prasad et al., 2005). As there is evidence that SEP3 is more important for the class E gene function than any of the other three SEP genes in Arabidopsis (reviewed by Melzer et al., 2009) this is remarkable, given that OsMADS7 and OsMADS8 appear to be more similar to SEP3 than OsMADS1 in terms of phylogenetic relationship and expression patterns (Münster et al., 2002; Becker and Theissen, 2003; Malcomber and Kellogg, 2004, 2005; Nam et al., 2004; Prasad et al., 2005). Thus a priori OsMADS7 and OsMADS8 may have appeared more likely candidates for providing the class E gene function than OsMADS1, and so the question arises as to what functions the other four rice SEP-like genes other than OsMADS1 have. As rice is only quite distantly related to Arabidopsis within the angiosperms, answering that question may tell us a great deal about the evolutionary conservation and dynamics of the SEP-like genes and class E floral homeotic gene function.
Therefore, we explored the functions of the SEP-like genes of rice by a reverse genetics approach employing double-stranded RNA-mediated interference (Chuang and Meyerowitz, 2000; Baulcombe, 2002; Hannon, 2002) by silencing different SEP-like genes individually and in combination, and studying the expression of these genes and the interactions of the proteins encoded by them. Our data demonstrate conservation of SEP-like genes in specifying floral determinacy and organ identities since the separation of eudicots and monocots about 150 million years age, but also reveal monocot-specific neo- and sub-functionalization events, thus indicating an unexpected evolutionary dynamics of SEP-like genes.
Silencing of either OsMADS7 or OsMADS8 produces only mild mutant phenotypes
Sixteen independent transgenic lines transformed with pUJOsM7C, named OsM7-RNAi, and 46 independent transgenic lines with the pUJOsM8 construct, termed OsM8-RNAi, were generated. Real-time PCR revealed that expression of the corresponding genes (OsMADS7 in the OsM7-RNAi lines and OsMADS8 in the OsM8-RNAi lines) was specifically down-regulated, while the expression of other SEP-like genes was surprisingly upregulated compared with the wild type (Fig. 1a). In a relatively high number of flowers of the OsMADS7 knockdown lines a carpel with three stigmas was found (Fig. S1, Table 1). No other obvious alterations were observed in either vegetative or reproductive organs of all these lines. As both the sequences and expression patterns of OsMADS7 and OsMADS8 are quite similar we assumed that functional redundancy between OsMADS7 and OsMADS8 might be one reason why the observed mutant phenotypes deviate only weakly, at best, from the wild type.
Table 1. Characteristics of the florets in transgenic lines
Lodiculesc/No. of lodicules
Stamensd/No. of stamens
Carpelse/No. of styles
anormal; ++, paleas were leaf-like.
bnormal; +, lemmas were longer and narrower than the WT.
cnormal; +, lodicules were slender and papery; ++, lodicules were transformed into palea/lemma-like structures; +++, lodicules were leaf-like.
dnormal; +, stamens were longer and thinner than those of WT; +++, stamens were absent or transformed into leaf-like structures.
enormal; +, carpels were green and palea/lemma-like structures fused partially to form 2–4 abnormal styles; ++, carpels were broader than in the WT and not fused, the stigmas and styles were not visible, some with multi-carpel structures; +++, carpels were transformed into leaf-like structures.
f−/grains, normal and finally seed can be got; +, abnormal organ-like structures inside the carpels with mild phenotypes; ++, stamen- or carpel-like organs developed from the center of the carpels with strong phenotypes; +++, abnormal panicles with an elongated pedicel in the center of the florets.
gThe fertility was affected.
Note: More than 200 spikelets were dissected from each of the transgenic lines.
+ (59%), ++ (41%)
++/2(85%), 3(10%), 4(5%)
+ (55%), ++ (45%)
+/6(50%), 5(22%), 4(28%)
+ (9%), ++ (91%)
+ (24%), ++ (76%)
−/2(60%), 3(20%), 4(20%)
Silencing of both OsMADS7 and OsMADS8 causes severe morphological alterations of floral organs
To test for functional redundancy between OsMADS7 and OsMADS8 we silenced both genes simultaneously, employing pUJOsM7I. This vector contains the region from nucleotides 149 to 454 (corresponding to the translation start codon ATG at 1) of the OsMADS7 coding sequence, which contains 29 nucleotides that are identical between OsMADS7 and OsMADS8 (Table S1). 18 independent pUJOsM7I transgenic lines were generated. The expression levels of all five SEP-like genes in the transgenic lines (named OsM7/8-RNAi) were analyzed by northern blot hybridization or RT-PCR, and real-time PCR. The accumulation of both OsMADS7 and OsMADS8 transcripts were strongly decreased, while the expression of OsMADS1, OsMADS5 and OsMADS34 was upregulated compared with the wild type (Figs 1b and S2). Four of the transgenic lines were analysed further in detail (Table 1).
Wild-type plants were flowering at about 73 DAP (days after planting), whereas all the transgenic lines were delayed in heading (Fig. S3) by approximately 2 weeks, with flowering ranging from 84 to 92 DAP. However, the transition stage from shoot meristem to inflorescence meristem was not delayed. The glumes in the spikelets of transgenic plants developed normally, and also palea and lemma largely resembled that of the wild type (Fig. 2a,c). The most significant morphological changes in OsM7/8-RNAi knockdown lines were observed in the organs of the innermost three whorls (Table 1). The lodicules were transformed into lemma/palea-like structures, some of which appeared to be papery and thin (Fig. 2e,f), arrowheads); the number of these organs was sometimes increased and varied from two to four. The anthers and filaments of stamens of the knockdown lines appeared to be longer and thinner (Fig. 2h, arrows) than those of the wild type. The number of stamens in florets varied from three to seven. However, no pollen grains were produced by these stamens (Table 1).
In OsM7/8-RNAi knockdown plants all carpels were aberrant, but showed a range of phenotypes. In mild cases carpels were green and partially fused (Fig. 2f,g). In plants with the strongest phenotypes the carpels were completely unfused (Fig. 2i, black arrows). In addition, the determinacy of the flower meristems in OsM7/8-RNAi knockdown plants was lost, so that inside the carpels additional reproductive organ-like structures were initiated (Fig. 2g,i, white arrows). The alterations of the modified florets were inspected more closely by scanning electron microscopy (SEM). In OsM7/8-RNAi knockdown plants, the abaxial epidermal cells of the lodicules, stamens and carpels were changed in size and shape (Fig. 2l,n,p), compared with those of the wild type (Fig. 2k,m,o). Remarkably, trichome-like structures emerged on the abaxial epidermal surface of the transgenic lodicules (Fig. 2h, arrowhead), stamens (Fig. 2n, arrow) and carpels (Fig. 2p, arrow), a situation similar to that of wild-type palea and lemma (Fig. 2j), suggesting that stamen and carpel identities were transformed homeotically into palea/lemma-like organ identity, at least partially.
Taken together, OsM7/8-RNAi knockdown plants showed both severe meristic and homeotic effects in the inner three floral whorls, while the other floret and spikelet organs appeared much less affected, if modified at all.
Simultaneous silencing of OsMADS1, OsMADS5, OsMADS7 and OsMADS8 transforms all floret organs into leaf-like structures
In order to uncover potential functional redundancy beyond OsMADS7 and OsMADS8 rice lines transgenic for the pUJOsM8M construct were generated in which four SEP-like genes, OsMADS1, OsMADS5, OsMADS7 and OsMADS8 are knocked down. 26 independent transgenic lines showed severe phenotypes. Two independent lines (OsM1/5/7/8-RNAi_6 and OsM1/5/7/8-RNAi_9) in which the expression of OsMADS1, OsMADS5, OsMADS7 and OsMADS8, but not of OsMADS34, were strongly down-regulated (Fig. 1c), were selected for further analysis (Table 1). As revealed by RT-PCR and real-time PCR the expression of other floral homeotic genes or their close relatives (e.g. OsMADS14, OsMADS15, OsMADS4, OsMADS16, OsMADS3, and OsMADS58) in the two lines was not or only weakly affected (Fig. S4a,b).
The OsM1/5/7/8-RNAi knockdown lines showed mutant phenotypes very similar to those of the strong osmads1 mutant (Agrawal et al., 2005) and the dsRNAiOsM1 lines (Prasad et al., 2005), including abnormal panicles (Fig. 3a), under-developed palea and lemma (Fig. 3b,c), and sterility (Table 1). However, there were also interesting differences. The OsM1/5/7/8-RNAi knockdown lines were tremendously delayed by almost 3–4 weeks in their flowering time (Fig. S3). In a few plants with the strongest phenotype the panicles did not appear out of flag-leaves. Moreover, the palea, but not the lemma, were more severely affected in the OsM1/5/7/8-RNAi knockdown transgenic lines (Fig. 3b,c) compared with the osmads1 mutant (Agrawal et al., 2005) and the dsRNAiOsM1 lines (Prasad et al., 2005). All lodicules, stamens and carpels were homeotically transformed into leaf-like structures (Fig. 3c–f). The additional abnormal panicles with an elongated pedicel in the center of the floret (Fig. 3c,d,f) developed from the extra whorl, or simultaneously in the third whorl (Fig. 3d). The different developmental stages of the additional abnormal panicles were inspected by scanning electron microscopy (SEM). Inflorescence-like structures were developed from inside the carpels (Fig. 3g,h) at both early and late developmental stages (Fig. 3i). SEM also showed that hair-like structures are present on the surfaces of the organs that have been homeotically transformed into leaf-like structures in the fourth whorl position (Fig. 3j). These phenotypic features resemble greatly those of the sep1/2/3/4 quadruple mutant of Arabidopsis in which all floral organs are transformed into leaf-like structures and the flower lost determinate growth.
To confirm that the severely abnormal phenotypes of the OsM1/5/7/8-RNAi transgenic plants are the result of RNAi, the presence of OsMADS8-derived small (21–24 nt) RNAs was investigated by Northern blot hybridization. Our data showed that siRNAs were detectable from the transgenic lines with abnormal phenotypes, but not from control plants transformed with the empty vector (Fig. S5b), corroborating the view that the abnormal phenotypes were indeed caused by RNAi.
Comparison of expression patterns of OsMADS7 and OsMADS8
Previously, northern blot analysis showed that the expression of OsMADS7 and OsMADS8 is restricted to reproductive organs such as inflorescences, and developing kernels (Pelucchi et al., 2002). In situ hybridization showed that during flower development both genes are expressed in the part of the floral meristem where the lodicule and stamen primordia will originate. Later on expression is found in lodicules, developing stamens and pistils until flowers mature (Pelucchi et al., 2002). To better understand the mutant phenotypes described above with respect to gene expression the temporal and spatial expression patterns of the paralogous genes were analyzed by in situ hybridization in more detail involving a series of panicle and spikelet developmental stages. To facilitate the description of the observed expression patterns, the designation of developmental stages proposed previously was used (Itoh et al., 2005).
At very early stages of inflorescence development no OsMADS7 signal was visible (Fig. 4a,b); meanwhile the OsMADS8 transcripts were first detected in the primary branch meristems and at the tip of the bracts (bract 1 and 2) (Fig. 4c,d). Subsequently, an OsMADS7 signal became visible but was very restricted to the spikelet meristems (Fig. 4e,f), while OsMADS8 was expressed strongly and broadly both in the branch shoots and spikelet meristems (Fig. 4g). The expression domains of OsMADS8 and OsMADS7 are overlapping during spikelet development (Fig. 4h–m), but appear to be slightly different spatially during the early development of floral meristems (Fig. 4h,i). Subsequently, the transcripts of OsMADS8 and OsMADS7 were localized in the developing lodicules, stamens and pistils (Fig. 4j,k). In the mature florets, the transcripts of both OsMADS8 and OsMADS7 were confined to the reproductive organs (i.e. the stamens and ovary) (Fig. 4l,m).
Interaction patterns among rice SEP-like proteins
Previous studies showed that the OsMADS7 and OsMADS8 proteins are able to interact with candidate class A, class B and class C floral organ identity proteins, a situation similar to that of the SEP proteins in Arabidopsis, whereas OsMADS1 only interacts with putative class A (AP1-like) proteins (Moon et al., 1999; Lim et al., 2000). We investigated the interaction between all the rice SEP proteins under study in this manuscript in vitro and in vivo. Yeast two-hybrid assays showed that OsMADS7, OsMADS8 and OsMADS1 share similar interaction patterns at different strengths. As shown in Figure 5, both OsMADS7 and OsMADS8 are able to form homodimers (Fig. 5k,p), while OsMADS1 can homodimerize only weakly (Fig. 5a). These proteins also interact with each other (Fig. 5c,d,l,n,q,s). However, OsMADS5 can neither homodimerize, nor heterodimerize with the other SEP proteins (Fig. 5b,f,g,h,I,r). In addition, we analysed the interaction of rice class E proteins in Arabidopsis mesophyll protoplasts. Coimmunoprecipitation assays showed that OsMADS7 and OsMADS7, OsMADS7 and OsMADS1, OsMADS7 and OsMADS8, OsMADS8 and OsMADS8, OsMADS8 and OsMADS1, and OsMADS1 and OsMADS1 undergo protein–protein interactions in Arabidopsis mesophyll protoplasts, while OsMADS5 does not interact with OsMADS1, OsMADS7 or OsMADS8 (Fig. 6).
OsMADS7 and OsMADS8 have a class E floral homeotic gene function in rice
In Arabidopsis, SEP1, 2 and 4 are expressed throughout the floral meristems already at developmental stage 2, while SEP3 expression starts a bit later in the region in which floral organ primordia will initiate soon; subsequently, expression of SEP1 and 2 is restricted to floral organs of all four whorls and SEP3 expression to the inner three whorls, whereas SEP4 expression is localized in the central dome and only weakly in sepals (Flanagan and Ma, 1994; Mandel and Yanofsky, 1998). This finding, together with their considerable sequence similarity, suggests that the SEP genes encode proteins with redundant functions. This situation is indeed the case, as single mutants for each of the SEP genes generated by reverse genetics display only very weak mutant phenotypes, if any, while triple and quadruple mutants show homeotic transformations of floral organs and a dramatic loss of floral determinacy (Pelaz et al., 2000, 2001; Ditta et al., 2004).
In rice, the expression patterns of SEP–like genes are quite heterogeneous. OsMADS1 is first expressed in the spikelet meristem before the glume primordia emerge, and then restricted to lemma and palea, with weak expression in the carpel (Chung et al., 1994; Prasad et al., 2001). Expression of OsMADS5 was detected in primordia of stamens and (weakly) carpels, but not in those of lemma and palea (Kang and An, 1997). OsMADS34 is expressed throughout the plant, but expression is not detectable in the organs of the second and fourth whorls at later stages of flower development (Pelucchi et al., 2002). The expression patterns of these SEP–like genes in rice make them no obvious candidates for being the only class E genes in a strict sense (Becker and Theissen, 2003), but do not exclude subfunctionalization of class E genes. Expression of OsMADS8 and OsMADS7 was first detected in spikelet meristems, and then not detected in lemma and palea primordia, but transcripts of both genes were found to accumulate in developing lodicules, stamens and carpels during spikelet development (Fig. 4) (Greco et al., 1997). The expression patterns of these two genes are thus quite similar to those of SEP1, SEP2, and SEP3 in Arabidopsis, suggesting that OsMADS8 and OsMADS7 play a corresponding role in rice.
Previously, data on gene function of grass SEP–like genes based on mutant phenotypes were only available for OsMADS1 and OsMADS5. Some missense mutations in OsMADS1 result in the lhs1 mutant phenotype, suggesting that OsMADS1 plays a role in specifying meristem, palea and lemma identities (Jeon et al., 2000). More recently it was shown that knockdown of OsMADS1 affects the differentiation of specific cell types in lemma and palea, thus generating glume-like features, with severe derangements in the lemma (Prasad et al., 2005). In contrast, ectopic expression of OsMADS1 causes conversion of the glumes into lemma-/palea-like organs, but does not result in morphological alteration of floral organs (Prasad et al., 2001, 2005). The organs of the inner whorls in many OsMADS1 knockdown florets are converted into glume-like organs (Prasad et al., 2005; Chen et al., 2006).
Severe loss-of-function mutants affected in OsMADS1 display complete homeotic transformations of the organs of the three inner whorls (lodicules, stamens and carpels) into lemma- and palea-like structures, and a loss of determinacy of the flower meristem (Agrawal et al., 2005). If one equates palea/lemma with sepals, this phenotype strikingly resembles that of the sep1/2/3 triple mutant of Arabidopsis (Pelaz et al., 2000), which led to the definition of the class E floral homeotic function (Theissen, 2001). Such a function may appear difficult to reconcile with the fact that OsMADS1 is not expressed in lodicules and stamens, which might be explained by very early effects of OsMADS1, when the gene is expressed in the spikelet primordium before organ primordia develop. A function of OsMADS1 at this stage may be required for the specification of floral organs at later stages of ontogeny (Prasad et al., 2005).
In case of the loss-of-function of OsMADS5, the closest relative of OsMADS1 in the rice genome, the only deviation from wild-type phenotype found in the whole plant was that lodicules were attached to the lemma and palea, suggesting that the gene is not required for class E function (Agrawal et al., 2005), even though it may contribute to it in a redundant way.
Like in case of osmads1 mutants, the morphogenetic alterations in our studies, caused by knockdown of both OsMADS7 and OsMADS8, were also restricted to the organs of the innermost three whorls. Specifically, the lodicules, stamens and carpels were transformed into palea/lemma-like structures, whereas glumes, lemma and palea were not affected. Flower apical meristems (AMs) in the center of the transgenic carpels were reverted into reproductive meristems, which developed into stamen-like structures or carpel-like structures. As either of the single gene knockdowns does not cause the morphological change, our results strongly suggest that OsMADS7 and OsMADS8 are functionally redundant, but that normal function of at least one of these genes is required for proper development of organ identities in the inner three floral whorls, and for maintenance of flower meristem determinacy. Furthermore, the knockdown OsMADS7/8 transgenic plants were delayed in flowering time by approximately 2 weeks, but not in the transition from shoot meristem to inflorescence meristem, thus resembling sep1/2/3 loss-of-function mutants in Arabidopsis (Jack, 2001; Theissen and Saedler, 2001). Considering that OsMADS7/8 is expressed at every developmental stage from branch meristems to floral organs, we hypothesize that late heading caused by loss-of-function of OsMADS7/8 may be due to the affected stem elongation and bolting.
The lack of redundancy to other SEP-like genes may explain why at least two ‘natural’ mutants, leafy hull sterile1 (lhs1) and naked seed rice (nsr), have been found for OsMADS1. This is in contrast to the redundantly acting OsMADS7 and OsMADS8 and the SEPALLATA genes of Arabidopsis, which required the generation of double or even triple mutants obtained by reverse genetics to display a mutant phenotype. The mutant phenotypes of osmads1 and osmads7/8 are strikingly similar, indicating that these genes do not act redundantly such as SEP1/2/3, but in an interdependent way.
Evolutionary implications of class E genes in rice
Phylogenetic analyses revealed that SEP-like genes can be further subdivided into several well supported clades (subfamilies), and that many duplications of SEP-like genes occurred during angiosperm evolution (Malcomber and Kellogg, 2005; Zahn et al., 2005). For each of the five rice SEP-like genes, there are putative orthologs identified from other grass species, such as maize (Münster et al., 2002) and ryegrass (Petersen et al., 2004), indicating that the most recent common ancestor of maize and rice, which existed quite close to the base of extant grasses about 50–70 million years ago, had already at least five different SEP-like genes (Münster et al., 2002; Becker and Theissen, 2003). A recent phylogenetic analysis provided evidence for an ancient duplication of SEP-like genes before the origin of the most recent common ancestor of extant angiosperms that generated two subclades, SEP1/2/4-like genes (or AGL2/3/4 clade) and SEP3-like genes (AGL9 clade), with OsMADS8 and OsMADS7 being SEP3-like genes, and OsMADS1, OsMADS5 and OsMADS34 being SEP1/2/4-like genes (Zahn et al., 2005). The expression patterns of SEP1/2/4-like genes vary considerably among cereals, suggesting that these genes evolved diverse functions which may have facilitated the origin of diverse inflorescence structures in grasses (Malcomber and Kellogg, 2004; Shitsukawa et al., 2007).
Our results provide strong evidence that an ancestral function of class E floral homeotic genes in being involved in specifying the identity of all kinds of floral organs is conserved through the angiosperms, although the functional partitions among the paralogs vary in different species. While the situation in Arabidopsis is largely dominated by functional redundancy among the SEP genes (Pelaz et al., 2000; Ditta et al., 2004), a whorl-specific subfunctionalization was observed in another eudicot, Gerbera (Uimari et al., 2004). In the monocot rice, however, the functional interdependence of two types of genes, OsMADS1 and OsMADS7/8, evolved, a situation that can somewhat be considered being the opposite of functional redundancy. The fact that osmads1 and osmads7/8 mutants have similar, but not identical functions, such as in the control of flowering time, could indicate that individual proteins have additional functions beyond their activities in complexes involving both types of proteins.
The obligate functional interdependence of floral homeotic genes, as observed here for OsMADS1 and OsMADS7/8, is not unprecedented, but well-known from the class B floral homeotic genes of eudicots. In these cases the formation of obligate heterodimers between DEF- and GLO-like proteins upregulating the expression of the DEF-like and GLO-like genes in an autoregulatory loop explains why both genes are functionally interdependent, so that mutants in the DEF-like gene are almost identical to mutants in the GLO-like gene (Schwarz-Sommer et al., 1992; Lenser et al., 2009). This system possibly evolved to increase the robustness of important decisions during the development of flower organ identity (Lenser et al., 2009). A similar explanation may apply to the SEP-like genes of rice. Here, the class E gene function specifying the identity of lodicules, stamens and carpels may require the direct interaction of OsMADS1 and OsMADS7/8 during spikelet meristem development, when the expression patterns of these genes overlap. For example, protein complexes involving OsMADS1–OsMADS7 dimers may act redundantly with those involving OsMADS1–OsMADS8 dimers in specifying lodicules, stamen and carpel identity, with OsMADS1 being required in both complexes. In contrast, complexes involving OsMADS7 or OsMADS8 but not OsMADS1 may have a specific role in determining heading date or palea development.
However, there is a clear difference between the functional interdependence of DEF-like and GLO-like class B floral organ identity genes and the case discussed here. The interdependent positive autoregulatory loops in which DEF-like and GLO-like genes are involved imply that expression of the one gene is almost abolished when the function of the other gene is compromised, and vice versa (Schwarz-Sommer et al., 1992; Lenser et al., 2009). However, knockdown of OsMADS7/8 expression does not reduce but upregulates OsMADS1 expression. Thus the remaining activity of OsMADS7/8 or OsMADS1, respectively, either suffices to sustain the autoregulatory loop, or the respective genes do not exhibit a regulatory relationship in such a way at all. The upregulation seen may even suggest that the genes are connected by negative rather than positive feed-back loops, either directly or indirectly. It is becoming more and more clear that many MADS-box genes are subject to positive or negative cross- and autoregulatory control (see, e.g., Gómez-Mena et al., 2005; Lenser et al., 2009; Liu et al., 2009; Ohmori et al., 2009), so that such a scenario may not appear too far fetched.
Alternatively, the similarity of OsMADS1 and OsMADS7/8 loss-of-function phenotypes may reflect a dosage effect during spikelet meristem development. Assuming that the genes are functionally almost equivalent, but dosage dependent (incomplete) knock-down of either OsMADS7 or OsMADS8 may still leave sufficient protein, while knockdown of both genes, or of OsMADS1, may bring the OsMADS1/7/8 amount in the spikelet meristem cells below a critical threshold for proper SEP-like protein function. Both osmads1 and osmads7/8 mutants reveal some aspects of the severe and complete sep phenotype when the four SEP-like genes (OsMADS1/5/7/8) are down-regulated, suggesting that these class E paralogs have undergone subfunctionalization resulting in partial overlapping functions.
It may be worthwhile to note that OsMADS34, also a SEP-like gene in rice that is expressed in all plant tissues, was not negatively affected in its expression in all the transgenic rice lines studied by us, as revealed by real-time PCR and RT-PCR. Remarkably, its expression level in the transgenic lines was even higher than that in wild-type plants, suggesting that the transcription of OsMADS34 may be usually repressed by the other SEP-like genes, either directly or indirectly. However, our data show that the upregulation of OsMADS34 is insufficient to compensate for the decreased expression of the other SEP-like genes.
As shown above, OsM1/5/7/8-RNAi knockdown lines resemble greatly those of the sep1/2/3/4 quadruple mutant of Arabidopsis. This finding suggests that genes OsMADS1/5/7/8 cover the full class E floral homeotic function, even though we cannot completely exclude the possibility that a quintuple mutant also comprising OsMADS34 would not reveal an even more severe class E loss-of-function phenotype.
As summarized in the ‘floral quartet model,’ the SEP proteins form higher order complexes together with class A, B or C proteins to control various transcriptional programs required and sufficient for specifying floral organ identity during development in eudicots (Theissen, 2001; Theissen and Saedler, 2001). Our data show that some phenotypes caused by silencing of the SEP orthologs (OsMADS7 and OsMADS8) resemble those of the class B (OsMADS16) or class C (OsMADS3 and OsMADS58) gene mutants (Nagasawa et al., 2003). Taking together, these data suggest that ‘floral quartet’ complexes, similar to those in eudicot species (Honma and Goto, 2001; Favaro et al., 2003), may also be formed to control flower development in rice. More data on protein–protein and protein–DNA interactions, as provided, e.g. by electrophoretic mobility shift assays and yeast two-hybrid assays (lmmink et al., 2009; Melzer and Theissen, 2009, Melzer et al., 2009; this study) are required to clarify the issue.
Previously, comparison of loss-of-function mutants of class B genes in maize (Ambrose et al., 2000; Whipple et al., 2004) and rice (Nagasawa et al., 2003) with those in eudicots have shown that the lodicules of grasses are homologous to eudicot petals. A recent study provided evidence that class B genes have a conserved function involved in specifying second whorl organ identity across the angiosperms (Whipple et al., 2007). Our findings that lodicules are transformed into palea/lemma-like organs after knockdown of OsMADS7/8 in transgenic plants, and that lodicules are transformed into leaf-like structures after silencing of OsMADS1/5/7/8 in transgenic plants provide further evidence that there is a common conserved mechanism for specification of second whorl organ identity throughout flowering plants.
Plants of O. sativa ssp. Japonica variety ‘Zhonghua 11’ were grown in local paddy-fields and the greenhouse of the Institute of Botany, Chinese Academy of Sciences, Beijing (latitude of 39˚48′ N and a longitude of 116˚28′ E).
Generation of knockdown vectors for rice SEP-like genes
Specific regions of either OsMADS7 or OsMADS8 used for the RNAi constructs (Fig. S5a) were defined by the alignment of the rice candidate floral homeotic genes (Table S1). Cloning of the OsM7C region in sense and antisense orientation yielded vector pUJOsM7C, predicted to specifically knockdown OsMADS7; likewise, pUJOsM8C containing the OsM8C region was expected to knockdown OsMADS8, pUJOsM7I to knockdown both OsMADS7 and OsMADS8, and pUJOsM8M to simultaneously knockdown OsMADS1, OsMADS5, OsMADS7 and OsMADS8 by RNAi in transgenic rice plants. The intron and nos terminator cassette of pJawohl3–RNAi (GenBank Accession no. AF404854) were transferred with BamHI/NotI sites to the pBluescript SK (Stratagene, La Jolla, CA, USA), termed pBJWl3 as an intermediate vector. Then the desired coding regions (OsM7C, OsM8C, OsM7I, OsM8M in Fig. S5a) were amplified with specific primers (Table S2) and cloned into two sides of the intron region of the pBJWl3 vector in the sense and antisense orientation, respectively. Finally, the dsRNAi cassette containing the intron and two oppositely orientated coding sequences were mobilized with restriction enzymes BamHI and SacI and introduced into vector pCAMBIA1301–Ubi, in which the maize (Zea mays) Ubi (ubiquitin) promoter (Cornejo et al., 1993) was inserted into the HindIII and BamHI sites, thus resulting in final plasmids pUJOsM7C, pUJOsM8C, pUJOsM7I and pUJOsM8M. All these constructs were verified by restriction mapping.
Rice calli, which had been induced from mature embryos of variety Zhonghua11 (O. sativa ssp. Japonica), were transformed by the Agrobacterium strain EHA105 harboring one of the RNAi constructs or the control vector pCAMBIA1301, as described previously (Hiei et al., 1994; Huang et al., 2000). Transgenic calli were selected on Murashige and Skoog (MS) medium containing 50 mg L−1 hygromycin B (Roche, cat.No.10843555001). Hygromycin-resistant plants regenerated from calli were transplanted into soil and grown at a greenhouse or local paddy-fields. For measurement of flowering time, the transgenic and wild-type plants were transplanted under natural short day conditions (in paddy-fields in Beijing).
Scanning electron microscopy
Inflorescences from transgenic or wild-type rice were fixed in fresh FAA solution (50% ethanol, 5% acetic acid, and 3.7% formaldehyde), dried, coated as described (Shan et al., 2006), and photographed with a Hitachi S-800 scanning electron microscope (SEM).
In situ hybridization
Freshly collected young panicles were fixed immediately in FAA solution. The gene-specific C-terminal regions of OsMADS8 (nucleotides 616–960 counted from the start codon ATG) and OsMADS7 (548–911, nucleotide positions from ATG) were used as templates for synthesizing sense and antisense digoxigenin-labeled RNA probes using the DIG Northern Starter Kit (Roche Diagnostics, Mannheim, Germany). The plant material was dehydrated, embedded, sliced (8 μm), pretreated, hybridized, washed and detected as previously described (Kouchi and Hata, 1993).
RNA preparation and gel blot analysis
Total RNA was extracted from rice inflorescences ranging from 3 to 6 cm by using TRIzol reagent (Gibco-BRL, Gaithersburg, MD) according to the manufacturer’s instructions. The gene-specific probes for OsMADS7 and OsMADS8 (Table S2 were labeled with the Prime-a-Gene® Labeling System (Promega Madison, WI, USA). Procedures of gel blot analysis were performed as previously described (Lu et al., 2007).
Small RNA gel blot analysis was performed as described (Liu et al., 2005). A probe corresponding to endogenous MADS-box siRNAs, which contained the OsMADS8 cDNA coding sequence (nucleotides 48–316 counted from the start codon ATG), was amplified with primers of OsM8M1 and OsM8M2 (Table S2) and cloned into the pGEM-T easy vector (Promega Madison, WI, USA), then transcribed with T7 polymerase in vitro, and labeled with [α-32P]ATP.
RT-PCR and real-time PCR expression analysis
Total RNA was isolated as mentioned above. Reverse transcription was performed by using Superscript™ III Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA) with an oligo(dT)18 primer. The diluted cDNA samples were used as templates for RT-PCR and real-time PCR.
For RT-PCR analysis, the PCR conditions were 94°C for 2 min followed by 22–25 cycles of 94°C for 30 sec, 56°C for 30 sec, and 72°C for 1 min; then 10 min of extension at 72°C. For real-time PCR analysis, triplicate quantitative assays were performed using the QuantiTect® SYBR® Green PCR kit with a Rotor-Gene 3000 (Corbett Research, QIAGEN, Hilden, Germany) detection system and software according to the manufacturer’s instructions. The amplifying program with the gene-specific primers (Table S2) was as follows: 95°C for 15 min, followed by 40 cycles of 94°C for 30 sec, 56°C for 30 sec, and 72°C for 30 sec. The relative expression levels of genes were analysed using the two Standard Curves Relative Quantification method. The amplification of ACTIN was used as an endogenous control to normalize all data. All results represent the means of more than three independent experiments.
Yeast two-hybrid (Y2H) analyses
The full-length cDNAs of OsMADS1, OsMADS5, OsMADS7, OsMADS8 and OsMADS13 were amplified with gene-specific primers (Table S2). Then the PCR sequences were fused into the activation-domain (AD) vector pGADT7 and the DNA-binding-domain (BD) vector pGBKT7 at the EcoRI and XhoI/SalI sites, respectively. All constructs were verified by restriction enzyme analyses and sequencing. These constructs were transformed into Saccharomyces cerevisiae strain AH109 (BD Biosciences, Palo Alto, CA, USA) according to the manufacturer’s protocol. No detectable self-activation for each of the single construct was observed on SD selective medium (SD-His-Leu+5 mm 3-AT or SD-His-Trp+5 mm 3-AT). The transformants co-transformed with plasmids encoding OsMADS7 (or OsMADS8) 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. Interaction analyses were performed as previously described (Shan et al., 2006).
Coimmunoprecipitation (Co-IP) analyses of the rice SEP-like proteins in Arabidopsis mesophyll protoplasts
Eight vectors, 2×35SP::6×myc-OsMADS1/5/7/8 and 35SP::3×flag-OsMADS1/5/7/8 were constructed to transiently express OsMADS1/5/7/8 in Arabidopsis protoplasts for coimmunoprecipitation. Full-length cDNAs of OsMADS1/5/7/8 were cloned as mentioned above (Y2H), and the fragments were cloned into the vector pRT107–6×myc under the control of double Cauliflower Mosaic Virus (CaMV) 35S promoter (Töpfer et al., 1993) and also into the plasmid pRT105–3×flag driven by a 35S promoter (Zhao et al., 2007), respectively. All final constructs were verified by sequencing.
Protoplasts of A. thaliana ecotype Columbia (Col-0) were isolated and transformed as described (Yoo et al., 2007). Coimmunoprecipitation was done using methods described elsewhere (Zhao et al., 2007).
We thank Dr Bin Liu and Xiaofeng Cao (Institute of Genetics and Developmental Biology, Chinese Academy of Sciences) for help with the small RNA hybridizations, and Dr Yan Guo (National Institute of Biological Sciences, Beijing) for his kindly gifts of plasmids and excellent technical assistance on Co-IP analysis. Many thank also to two anonymous reviewers whose comments helped considerably to improve our manuscript. This work was supported by the Ministry of Science and Technology of China (Grants 2006CB100202; 2006AA10Z190) and the Chinese Academy of Sciences (Grant KSCX2-YW-R-135).