75 Development of floral organ identity: stories from the MADS house Günter Theißen Recent studies on AGAMOUS-LIKE2-, DEFICIENS- and GLOBOSA-like MADS

Transcription

1 75 Development of floral organ identity: stories from the MADS house Günter Theißen Recent studies on AGAMOUS-LIKE2-, DEFICIENS- and GLOBOSA-like MADS-box genes in diverse seed plant species have provided novel insights into the mechanisms by which the identity of the different floral organs is specified during flower development. These advances in understanding may lead to major refinements in the classical ABC model of floral organ identity. Addresses Max-Planck-Institut für Züchtungsforschung, Department Molekulare Pflanzengenetik, Carl-von-Linné-Weg 10, D Köln, Germany Current Opinion in Plant Biology 2001, 4: /01/$ see front matter 2001 Elsevier Science Ltd. All rights reserved. Abbreviations AG AGAMOUS AGL11 AGAMOUS-LIKE11 AP1 APETALA1 CArG-box CC-A rich GG DNA sequence element DEF DEFICIENS FBP7 FLORAL BINDING PROTEIN7 GAGA1 GERBERA AGAMOUS-LIKE1 GLO GLOBOSA GRCD1 GERBERA REGULATOR OF CAPITULUM DEVELOPMENT1 MYA million years ago OSMADS1 Oryza sativa MADS-box1 PI PISTILLATA PLE PLENA SEP1 SEPALLATA1 SI1 SILKY1 SQUA SQUAMOSA ZMM6 Zea mays MADS-box6 Introduction: war of the whorls the ABCs of floral organ identity Flowers owe their unique appearance, their evolutionary success, their beauty and their economic importance to human beings to the distinct identities of the different floral organs that they are composed of. A typical flower of an eudicotyledonous flowering plant consists of four different organ classes arranged in four whorls at the tip of a floral shoot. The first outermost whorl often consists of green, leaf-like sepals. The second whorl is composed of usually coloured and showy petals. The third whorl contains the stamens, that is, the male reproductive organs, which produce pollen. Finally, the fourth, innermost whorl contains the carpels, that is, the female reproductive organs, which are often fused and inside of which the ovules and seeds develop. Although the structures and functions of sepals, petals, stamens and carpels may differ dramatically at maturity, each floral organ starts its development as a little bulge on the floral meristem, a tiny clump of undifferentiated cells. Each cell in the developing floral organ primordium must somehow learn its position within the flower, and differentiate accordingly into an appropriate cell type. About a decade ago, the first working hypotheses were suggested to explain how the specification of organ identity during flower development is controlled by homeotic selector genes [1 3]. These models were based on the study of homeotic mutants in which the identity of floral organs is changed. In the major model plants that were studied, thale cress (Arabidopsis thaliana) and snapdragon (Antirrhinum majus), such mutants come in three classes, A, B and C. Ideal class-a mutants have carpels in the first whorl instead of sepals and stamens in the second whorl instead of petals. Class-B mutants have sepals rather than petals in the second whorl and carpels rather than stamens in the third whorl. Class-C mutants have petals instead of stamens in the third whorl and replacement of the carpels in the fourth whorl by sepals. In addition, these mutants are indeterminate, that is, they have continued production of mutant floral organs inside the fourth whorl. The existence of these mutants suggested that development of the flower is sculpted by homeotic selector genes (i.e. floral organ identity genes) whose expression gives the different floral organs their identity. Such genes can be considered as acting as major developmental switches that activate the entire genetic program for a particular organ. An early ABC model proposed that the state of expression ( on or off ) of three different genes, a, b and c, specifies the identity of the four different floral organs [1]. The proposed states of expression of genes a, b, c in the different floral organs are on, off, off in the sepals, on, on, off in the petals, off, on, on in the stamens and off, off, on in the carpels. If none of the three genes are expressed then leaves develop [1]. Another early model focused on Antirrhinum and considered two developmental pathways, termed A and B, in combination with a floral ground state; pathway A corresponded to gene b, and pathway B to gene c in the early ABC model [2]. In 1991, the classical ABC model, which has been widely accepted until now, was proposed to explain how homeotic genes control floral organ identity (Figure 1a) [3]. Similar to the early ABC model, the classical ABC model maintains that organ identity in each whorl is determined by a unique combination of three organ identity gene activities, called A, B and C. Expression of A alone specifies sepal formation. The combination AB specifies the development of petals, and the combination BC specifies the formation of stamens. Expression of C alone determines the development

2 76 Growth and development Figure 1 (a) (b) Whorl: Petals Stamens Sepals Carpels B A C AP3 PI AP1, AP2 AG or function of these organs. Except for AP2, all ABC genes share a highly conserved, approximately 180 base pair DNA sequence, termed the MADS-box. This sequence encodes the DNA-binding domain of the respective MADS-domain transcription factors (for recent reviews about MADS-box genes in plants, see [5,6 ]). On the basis of studies in petunia (Petunia hybrida), the ABC model was later extended to include a D function, yielding an ABCD model [16,17]. When ectopically expressed, the D-function genes FBP7 (FLORAL BIND- ING PROTEIN7) and FBP11 from petunia induce the formation of ectopic ovules on the perianth organs of transgenic flowers. They have, therefore, been considered as the master control genes of ovules. D-function genes have been defined by mutant phenotypes in only petunia so far, but sequence similarity suggests that the corresponding gene in Arabidopsis is AGL11 (AGAMOUS-LIKE11) [17,18]. FBP7, FBP11 and AGL11 are MADS-box genes, which are comparatively closely related to AG. Current Opinion in Plant Biology The classical ABC model for flower organ identity in Arabidopsis. (a) Floral organ identity is specified by homeotic functions A, B and C, which are each active in two adjacent whorls. A alone specifies sepals in whorl 1; the combined activities of A and B specify petals in whorl 2; B and C specify stamens in whorl 3; and C alone specifies carpels in whorl 4. The activities A and C are mutually antagonistic: A prevents the activity of C in whorls 1 and 2, and C prevents the activity of A in whorls 3 and 4. (b) The proteins providing the floral homeotic functions in Arabidopsis. Except AP2, all are MADS-domain proteins, which are expressed in the regions in which they specify organ identity. Antagonistic interactions are indicated by barred lines. A hyphen indicates heterodimer formation, a comma symbolises that the mode of interaction is unknown. of carpels. In contrast to the early ABC model, however, there is not necessarily just one gene behind any of the A, B and C gene activities (i.e. floral homeotic functions). Moreover, in order to explain the three classes of floral homeotic mutants, the classical ABC model proposes that the A- and C-function genes negatively regulate each other, so that the A-function becomes expressed throughout the flower when the C-function is mutated and vice versa (for reviews of the ABC model, see [4,5,6 ]). Arabidopsis genes providing the three homeotic activities A, B and C are known. The A function is contributed by two different genes, APETALA1 (AP1) and AP2, the B function also by two genes, AP3 and PISTILLATA (PI), and the C function by just one gene, AGAMOUS (AG) (Figure 1b). In Antirrhinum, the B function is provided by DEFICIENS (DEF) and GLOBOSA (GLO), and the C function by PLENA (PLE). Cloning of these genes during the 1990s revealed that they all encode putative transcription factors [7 14] (for reviews, see [6,15 ]). Thus, the products of the ABC genes probably all control the transcription of other genes (i.e. target genes) whose products are directly or indirectly involved in the formation This review focuses on the dramatic recent progress in our understanding of the function, conservation and evolutionary origin of floral organ identity genes of the MADS family. The organ identity genes, however, represent just one level in the gene hierarchy that controls inflorescence and flower development [15 ]. I refer to other recent reviews [6,15,19 ] for current information on how the expression of the ABC genes is activated by floral meristem identity genes upstream of the organ identity genes, and what is known about the downstream targets of the floral homeotic genes. Novel floral homeotic functions provided by AGL2-like MADS-box genes Stars of three whorls: SEPALLATA MADS-box genes are required for specifying petal, stamen and carpel identity in Arabidopsis AGL2-like genes have been known as a defined subfamily of the plant MADS-box genes for quite a while [20 23]. The presence of members of this gene subfamily in diverse flowering plants suggests that they have important and conserved functions. However, because no classical mutants have been isolated for any AGL2-like gene so far, the determination of their functions has proved difficult. Transgenic plants in which expression of AGL2-like genes, FBP2 from petunia or TOMATO MADS-box5 (TM5) from tomato (Lycopersicon esculentum), was inhibited by co-suppression or antisense technology developed highly aberrant flowers with modified whorl 2, 3 and 4 organs, including homeotic transformations to sepaloid organs [24,25]. In line with this, wild-type expression of both FBP2 and TM5, as of many AGL2-like genes, was found in petal, stamen and carpel primordia [24 26]. On the basis of these findings and of expression studies in Arabidopsis, it was suggested that the AGL2-like genes act as mediators between the floral meristem and the floral organ identity genes [25,27].

3 Development of floral organ identity: stories from the MADS house Theißen 77 Recently, a thorough investigation of most of the AGL2-like genes of Arabidopsis has corroborated many of the earlier findings, but has also added a fascinating new twist to the tale. Three of the Arabidopsis AGL2-like genes, AGL2, AGL4 and AGL9 [20,28], are expressed in the second, third and fourth whorl organ primordia, although AGL2 and AGL4 are also expressed in the outermost whorl of developing flowers [27 29]. The high sequence similarity and parallel expression patterns among the Arabidopsis AGL2-like genes suggested that they may encode redundant functions and may explain why classical mutations in these genes have not been identified [27]. Indeed, single mutations for each of the three genes generated recently by reverse genetics produced only subtle phenotypes [30 ]. However, in triple mutants in which AGL2, AGL4 and AGL9 had all lost their function, all of the flower organs resembled sepals and the flowers became indeterminate [30 ]. These findings indicate that AGL2, AGL4 and AGL9 have redundant functions that are required for petal, stamen and carpel development and to prevent the indeterminate growth of the flower meristem. The triple mutant phenotype prompted the authors to rename the genes SEPALLATA1 (SEP1), SEP2 and SEP3, respectively, meaning lots of sepals. The sep1 ; sep2 ; sep3 triple mutant phenotype is strikingly similar to that of double mutants that are defective in both a B- and a C-class floral homeotic gene (i.e. ap3;ag or pi;ag), indicating that the floral homeotic B and C functions do not work in the triple mutant. SEP1 3 are, however, still expressed in B and C loss-of-function mutants, indicating that they do not act downstream of the B and C genes [30 ]. Furthermore, the initial expression patterns of B- and C-class genes are not altered in the sep1 ; sep2 ; sep3 triple mutant, suggesting that SEP1 3 are not required for the initial activation of the B-and C-class genes [30 ]. Gerbera s one whorl show of an AGL2-like gene The almost complete functional redundancy of three different genes in the specification of floral organ identity in three different whorls is striking, but may not be a conserved feature of flowering plants. Recent studies in Gerbera hybrida, an ornamental plant species from the Asteraceae family, suggest that there are also single AGL2-like genes with whorl-specific phenotypes [31 ]. Downregulation of the expression of the AGL2-like gene GRCD1 (GERBERA REGULATOR OF CAPITULUM DEVELOPMENT1) in transgenic Gerbera resulted in homeotic transformations in only one whorl: sterile staminodes, which normally develop in whorl 3 of marginal female florets, were transformed into petals, suggesting that GRCD1 is involved in specifying stamen identity [31 ]. The whorl 3 of GRCD1 loss-offunction mutants is strikingly similar to that of plants in which one of the Gerbera floral homeotic C-function genes, GAGA1 (GERBERA AGAMOUS-LIKE1) or GAGA2, is downregulated [32]. Hence, Kotilainen et al. [31 ] proposed that GRCD1 is required to provide the floral homeotic C function in whorl 3 of Gerbera [31 ]. In analogy to results obtained with the sep1 ; sep2 ; sep3 mutant, Kotilainen et al. [31 ] also observed that the expression of the C-function genes is not reduced in the GRCD1 mutant. Likewise, GRCD1 is not downregulated in Gerbera C-function mutants. Thus, GRCD1 does not act downstream of the C genes and is not required for the initial activation of the C genes. Like the SEP genes, GRCD1 is expressed in all floral whorls [31 ]. The difference between the phenotype of the grcd1 single mutant and that of both sep single and sep triple mutants could have several explanations [31 ]. For example, GRCD1 may show complete functional redundancy with other AGL2-like genes from Gerbera, except in whorl 3. It may also be, however, that Gerbera has AGL2- like genes that are functionally whorl-specific. If so, there should be also AGL2-like genes specialised in petal or carpel development. Finally, GRCD1 may be functionally important in all four floral whorls, but the investigated transgenic plants may not have a complete loss-of-function for GRCD1 and the stamens may be more susceptible than organs in other whorls to a reduction in the amount of GRCD1 protein. AGL2-like genes in grasses Further insights into the functional importance of AGL2- like MADS-box genes have been obtained from studies on monocotyledonous plants. Although the oldest known fossils of monocots were deposited just 90 million years ago (MYA) [33], molecular estimates indicate that the monocot lineage separated from the other angiosperms about MYA [34,35], that is, not long after establishment of the clade of extant angiosperms. Studies on certain kinds of MADS-box genes in monocots can thus tell us a great deal about the conservation of floral homeotic genes and their functions within the flowering plants. An especially important group of monocots are the grasses (Poaceae), which include the important cereals and the model systems maize (Zea mays ssp. mays) and rice (Oryza sativa). The tiny, wind-pollinated flowers of the grasses are quite distinct from the flowers of other taxa. They have carpels and stamens like their eudicot relatives, but they lack petals and sepals. Instead, lodicules, palea and lemma surround the reproductive organs, thus constituting structures called florets. Lodicules are small glandular-like organs that swell at anthesis to spread the lemma and palea apart so that the wind can disperse the pollen. Although grass florets are simplified and small, they are assembled into complex higher-order structures. One or more florets are surrounded by glumes, thus constituting spikelets. Spikelets may be formed as single units (as in rice) or as pairs (as in maize), and are assembled into often large inflorescences (such as the maize tassel and ear). The AGL2-like genes in grasses show relatively heterogeneous expression patterns, strongly suggesting that they are not a functionally homogeneous class of genes. It seems that this subfamily of genes underwent a relatively

4 78 Growth and development rapid functional diversification within the grasses shortly after they separated from the rest of the monocots, which possibly reflects the evolution of complex inflorescence structures [22,36]. In maize, there are at least eight different AGL2-like genes with distinguishable expression patterns [6 ]. For example, expression of the gene ZMM6 (Zea mays MADS-box6) is initially restricted to just one primordium in each pair of developing spikelet primordia, suggesting that this gene is involved in determining the identity of the pedicellate spikelet primordium (in contrast to the sessile spikelet primordium) [37,38]. The expression of two other AGL2-like genes, ZMM8 and ZMM14, is detectable only in the upper floret of each developing spikelet, suggesting that these genes determine upper floret identity (in contrast to lower floret identity). These genes may also be involved in conferring determinacy to the spikelet or upper floret meristem [36], a function that has recently been confirmed by transgenic loss-of-function mutants of ZMM8. More than the normal two florets often develop in the spikelets of these mutants (W Deleu, G Theißen, unpublished data). Similar observations have recently been reported for rice plants in which the ortholog of ZMM8, OSMADS1 (Oryza sativa MADS-box1), is mutated [39,40 ]. Wild-type rice spikelets develop only one floret, but in strong osmads1 mutants, new florets are generated within the spikelets. This observation suggests that OSMADS1 plays an important role in floral meristem determination during the early development of rice florets [40 ]. Other phenotypic characteristics of plants with missense mutations in OSMADS1 are elongated papery leafy paleas and lemmas, palea- and lemma-like lodicules, reduced stamen number, and an increase in carpel number. The changes in lodicules and reproductive organs resemble those of B-function mutants in grasses (see below), suggesting that OSMADS1 is required for B function. The deep: on the molecular basis of floral homeotic gene interactions According to the results outlined above, AGL2-like genes may be required for the floral homeotic B function (OSMADS1 in rice), for the C function (GRCD1 in Gerbera) or for both B and C function (SEP in Arabidopsis), and for regulating floral or spikelet determinacy (in Arabidopsis and rice) [30,31,40 ]. These findings raise, once more, the long-standing question of how the different floral organ identity genes interact at the molecular level. For example, if the Arabidopsis floral organ identity genes AP3, PI, AG and SEP must interact to specify stamens, as the phenotypes of their respective mutants indicate, at what molecular level do the interactions among these genes become operative, and how do the AGL2-like genes fit in with respect to the classical B- and C-function genes? Obviously, the known ABC models do not answer these questions. All of the classical MADS-domain floral homeotic proteins from Arabidopsis (i.e. AP1, AP3, PI, AG) can interact in vitro [41]. Both electrophoretic mobility shift assays and the yeast two-hybrid system have, however, revealed that the formation of DNA-binding dimers shows a high degree of partner-specificity in case of proteins from both Antirrhinum and Arabidopsis [41 43]. For example, B-function genes, such as DEF and GLO in Antirrhinum or AP3 and PI in Arabidopsis, and C-function genes, such as PLE in Antirrhinum or AG in Arabidopsis, interact genetically to specify stamen identity. The B-function proteins do not, however, form DNA-binding dimers with the C-function proteins. In fact, the different B-function proteins form obligatory DNA-binding dimers with each other (DEF GLO or AP3 PI), but not with MADS-domain proteins from any other class [41,42]. It was observed relatively early, however, that C-function proteins form DNA-binding dimers with AGL2-like proteins [42,43]. An early hypothesis was therefore that the C-function in Arabidopsis may require interaction between AG and the AGL2-like proteins, a suggestion that has now been fully corroborated [30 ]. Similarly, Kotilainen et al. [31 ] observed that the AGL2-like protein GRCD1 interacts with the C-class proteins GAGA1 or GAGA2, suggesting, together with other data (see above), that a GRCD1 GAGA1/2 protein heterodimer is required to provide the floral homeotic C-function in whorl 3 of Gerbera. The capacity for specific heterodimerization between AG-like and AGL2-like proteins has probably been conserved for at least 300 million years: the GGM3 (Gnetum gnemon MADS-box gene3) protein, which is encoded by an AG ortholog from the gymnosperm Gnetum gnemon [44], forms specific DNA-binding heterodimers with all of the AGL2-like proteins from Arabidopsis (C Weiser, C Kirchner, G Theißen, unpublished data). This finding suggests a conserved and hence important function for this interaction, but its exact nature was not immediately clear. Early studies of floral homeotic protein interactions did not lend support to the idea that the combinatorial mode of action of ABC gene activities is achieved through direct interactions between the corresponding proteins. This mode of action seemed unlikely because of the limited partner-specificity that these proteins exhibit in the formation of dimers [41]. Multimeric protein complexes were not, however, considered in these early studies. Recently, evidence for the formation of larger complexes of MADS-domain proteins has been found that may be directly relevant to the problem of floral homeotic protein interactions [45 ]. Egea-Cortines et al. [45 ] report genetic interaction in Antirrhinum between the floral meristem identity gene SQUAMOSA (SQUA) (an ortholog of AP1) and the B-function genes DEF and GLO. This interaction is required for the establishment of the whorled phyllotaxis of floral organs [45 ]. Remarkably, in electrophoretic mobility shift assays and the yeast two-hybrid system, this interaction is also seen at the protein level, where DEF,

5 Development of floral organ identity: stories from the MADS house Theißen 79 GLO and SQUA form a multimeric complex. The DNA-binding affinity of this complex differs from that of the individual dimers. Egea-Cortines et al. suggest a model according to which the protein complex is actually a protein tetramer, composed of a DEF GLO heterodimer and a SQUA SQUA homodimer. Within this tetramer, the DEF GLO and SQUA SQUA dimers recognise different DNA sequence elements termed CArG-boxes (CC-A rich GG DNA sequence elements) [45 ]. If the formation of such complexes were conserved in different flowering plant species, it could easily explain why some Arabidopsis proteins (such as AP1 and AP3) interact in vitro but do not form DNA-binding dimers [41]: they have protein interaction domains employed in multimeric complexes but of the six possible dimers generated by up to four different proteins in a tetramer, only two (e.g. AP3 PI but not AP1 AP3) form a DNA-binding surface. The characterisation of multimeric complexes of MADSdomain proteins may represent a breakthrough in understanding the interaction among floral homeotic genes. Interestingly, there is evidence, cited by Pelaz et al. [30 ], that the MADS-domain proteins of Arabidopsis form multimeric complexes that include SEP proteins. Specifically, it seems that AP3 PI interacts with SEP3 and AP1, as well as with SEP3 and AG, in larger complexes. The war of the words: are floral homeotic functions still a useful concept? The exciting findings by Pelaz et al. [30 ] demonstrate that SEP1 3 represent a class of floral organ-identity genes that is required for development of petals, stamens and carpels in Arabidopsis. Likewise, GRCD1 is a novel floral-organ identity gene required for stamen development in Gerbera [31 ], and OSMADS1 appears to be required for the correct development of lodicules and stamens in rice [40 ]. How do the functions provided by these genes fit into the ABC model? To answer that question one should note that in the classical ABC model, the terms A, B and C are used in different ways. A, B, and C denote: first, three regions of the floral meristem (i.e. the expression domains of floral homeotic genes), with A representing whorls 1 and 2, B whorls 2 and 3, and C whorls 3 and 4; second, three classes of floral homeotic mutants, with class A affected in whorls 1 and 2, class B in whorls 2 and 3, and class C in whorls 3 and 4; third, three classes of regulatory (i.e. homeotic) functions, with A specifying sepals, A+B petals, B+C stamens, and C carpels; and fourth, three classes of floral homeotic genes, with class A genes providing the floral homeotic A function, class B the B function, and class C the C function. In the classical ABC model the different meanings of ABC almost perfectly coincide. A typical class-c gene, for example, provides the floral homeotic C function and is expressed in region C. Mutation in a class-c gene leads to a class-c mutant, which has homeotic transformations of the organs of region C (i.e stamens and carpels) into perianth organs (i.e. petals and sepals, respectively). Pelaz et al. [30 ] suggest that it might be useful to think of SEP (meaning here SEP1 and/or SEP2 and/or SEP3) as a combined B C-class of genes in Arabidopsis. Similarly, Kotilainen et al. [31 ] argue that GRCD1 participates in the floral homeotic C function in whorl 3 in Gerbera. If these arguments are accepted, however, the coincidence among the different meanings of A, B and C is abandoned, so that the ABC system no longer appears to be satisfactory. Why, for example, are the activities provided by AP3/PI and AG considered as distinct floral homeotic functions B and C, respectively, whereas the SEP genes are interpreted to merely contribute to these functions? If, for example, petal identity requires genes AP1+AP3+PI+SEP and stamen identity AP3+PI+AG+SEP, why should we unite the different genes AP3 and PI as contributors to a single B function, define AG as providing a C function, but consider the SEP genes as mere contributors to a combined B C-function? If a characteristic expression domain or mutant phenotype is the criterion for a distinct function, then the SEP genes clearly deserve their own function because their expression domain and mutant phenotype differ from those of both the class-b and the class-c genes. One may argue, however, that a sep1 ; sep2 ; sep3 mutant is almost indistinguishable from a class-b and -C double mutant. Although one could say the same about b ; c double mutants with respect to sep mutants, it is not possible to specify four different types of floral organs with SEP and the A-function genes alone. Nevertheless, the criteria that are used to categorise the functions provided by the floral homeotic genes (including SEP1 3) into floral homeotic functions A, B and C appear somewhat arbitrary. As an alternative, Pelaz et al. [30 ] suggest that we consider SEP as a novel D class of genes. Unfortunately, this suggestion may lead to confusion with the D function (specifying ovule identity) of the ABCD model sensu Colombo, Angenent and colleagues [16,17], which was published much earlier. One could, however, consider the SEP genes as providing yet another floral homeotic function, for which I suggest the term E function here, thus yielding an ABCDE model, or A E model (Figure 2). Unfortunately, after having already lost its beautiful symmetry with the introduction of the D function, the involvement of a fifth floral homeotic function would ensure that the ABC model has finally lost its simplicity, which has certainly been one reason for its wide acceptance. The situation gets even worse if we assume that the Arabidopsis E function is divided into several other, whorlspecific functions (E1, E2, and so on) in Gerbera (and possibly other flowering plants). I wonder, therefore, whether the floral homeotic functions are still a useful concept given that our understanding of the relationships among these functions, the genes which provide them, and the different organ identities has become

6 80 Growth and development Figure 2 (a) Whorl: a Petals Stamens Sepals Carpels 4b Ovules The quartet model of flower organ identity The quartet model tries to explain how different combinations of floral homeotic genes (or their gene products) specify the identity of the different floral organs, without referring to floral homeotic functions. The model suggests that four different combinations of four different floral homeotic proteins determine the identity of the four different floral organs. The recent information about the interactions of floral organ identity proteins (as outlined above) is directly reflected by the model. (b) A AP1, AP2 B AP3 PI The A E model, or ABCDE model, of flower organ identity in Arabidopsis. (a) The functions A, B and C are the same as in the classical ABC model (Figure 1). In addition, there is a D function specifying ovules, and an E function, which is required for the specification of petal, stamen and carpel identity, and possibly also for ovule identity. The involvement of the C function in ovule formation is still unclear and is therefore indicated by a dashed line. (b) The proteins providing the floral homeotic functions in Arabidopsis. The functions A, B and C are provided by the same proteins as in the classical ABC model. The role of AGL11 in providing the D function is hypothetical and thus highlighted by a question mark; this hypothesis is based on the close phylogenetic relationship between AGL11 and the D-function genes from petunia, and on the similar expression patterns of these genes. The E function is provided by SEP1, and/or SEP2, and/or SEP3. Because the accumulation of SEP1 transcript is high in developing ovules (and, after fertilisation also in developing embryos and seed coats) [29], an involvement of the SEP genes (i.e. an E function) in ovule formation is assumed. Antagonistic interactions are indicated by barred lines. A hyphen indicates heterodimer formation, a comma symbolises that the mode of interaction is unknown. increasingly complex as our knowledge of the involvement of AGL2-like genes in specifying floral organ identity has expanded. In my view, it would suffice to refer to the different states of floral organ identity on the one hand and to the combinations of floral homeotic genes that specify these organ identities on the other hand. There may be no need for a definition of floral homeotic functions. These considerations lead us now to a novel model of floral organ identity specification. C E AG SEP1/SEP2/SEP3 D AGL11? Current Opinion in Plant Biology For simplicity, let us ignore ovules as extra organs. Thus, we have to explain how the identities of four different floral organs, the sepals, petals, stamens, and carpels, are specified. For that, we need four unique combinations of floral homeotic genes or proteins. In Arabidopsis, these combinations may be based on the formation of four different protein complexes. First, sepals: AP1 AP1??. Second, petals: AP1 AP3 PI SEP. Third, stamens: AP3 PI AG SEP. Fourth, carpels: AG AG SEP SEP. Unknown components are denoted by? (Figure 3). These protein complexes, assumed to represent transcription factors, may exert their function by specific binding to the promoters of target genes, which they activate or repress as appropriate for the development of the different floral organ identities. Specificity of target gene selection may be achieved by the four complexes having different binding affinities for pairs of diverse DNA-sequence elements (probably CArG-boxes) and by the characteristic distribution of these pairs of sequence elements in the promoter regions of the different target genes. The antagonism between the A and C function in the classical ABC model could be explained by the fact that protein complexes that contain AP1 repress the expression of the AG gene, and complexes that contain AG repress the AP1 gene. These interactions could, however, be more indirect. Assuming that the formation of multimeric complexes of MADS-domain proteins is a conserved feature of flowering plants, more general quartet models could be obtained by substituting the names of Arabidopsis proteins by the names of the protein clades that they are members of. Sepals would be identified by SQUA SQUA??, petals by SQUA DEF GLO SEP, stamens by DEF GLO AG SEP, and carpels by AG AG SEP SEP. To achieve an unambiguous system for nomenclature, these clades of orthologs are named after the first clade members to have been molecularly characterised [22]. To B or not to B: on the origin and evolution of DEF- and GLO-like genes As floral organ identity strictly depends on the function of floral homeotic genes, the phylogeny of the floral organ identity genes must have played an important role in the evolution of floral organs [6,21,46]. More insights into the phylogeny of the floral homeotic genes may, thus, help us to better understand the evolution of flowers. Recently, tremendous progress has been made in understanding the

7 Development of floral organ identity: stories from the MADS house Theißen 81 Figure 3 Whorl: Sepals Petals Stamens Carpels?? AP3 PI AP3 PI SEP AG AP1 AP1 + AP1 SEP + AG SEP + AG SEP + AG AG AP1 AP1 Current Opinion in Plant Biology Target genes Target genes Target genes Target genes The quartet model of flower organ identity in Arabidopsis. The model suggests that four different combinations of four different floral homeotic proteins determine the identity of the four different floral organs. These protein complexes, which are assumed to represent transcription factors, may exert their function by specifically binding to the promoters of target genes, which they either activate (+) or repress ( ) as appropriate for the development of the identities of the different floral organs. Binding probably occurs to pairs of diverse DNAsequence elements termed CArG-boxes (not highlighted here), which are brought into close vicinity by DNA bending. The antagonism between the A and C function in the classical ABC model could be caused by the repression of AG gene expression by protein complexes that contain AP1. Similarly, complexes that contain AG repress the AP1 gene. These interactions could, however, also be more indirect. Note that the exact structures of the multimeric complexes of MADSdomain proteins controlling the identity of flower organs is still highly hypothetical. The model presented here is mainly intended as a working hypothesis for future research. phylogeny of the class-b genes (or better, the DEF- and GLO-like genes). For previous insights into the evolution of the class-a and -C genes, see [6,15 ]. The grace of the grasses: what the DEF- and GLO-like genes of rice and maize tell us about organ homology and the conservation of the floral homeotic B-function The structure of grass flowers deviates so strongly from those of eudicot flowers that homology is no longer obvious for some organs. The identity of the eudicot organs to which the palea, lemma and lodicules are homologous has been widely debated for a long time. Recently, studies of B-function genes in grasses might have provided the answer to this question. Loss-of-function of the GLO-like gene OSMADS4 in transgenic rice plants results in homeotic transformations of lodicules into palea-/lemma-like organs and of stamens into carpelloid organs [47]. Loss of OSMADS4 function thus leads to a phenotype that is similar to that of the B-function mutants of eudicots, in which second whorl petals are transformed into sepals and third whorl stamens into carpels. This evidence suggests that the classical ABC model is applicable to grasses as it is to eudicots, except that palea and lemma (rather than sepals) are specified by the A function and lodicules (rather than petals) by a combination of A and B function (provided that an A function is operative in grasses). The cloning of the SILKY1 (SI1) gene has stimulated considerable progress in understanding the B function in maize [48 ]. The si1 mutant has a phenotype that is strikingly similar to the OSMADS4 loss-of-function phenotype described above. In line with this, SI1 is expressed in organ primordia that give rise to lodicules and stamens, and according to phylogeny reconstructions, SI1 is an ortholog of the B-class gene DEF [48 ]. Taken together, the data from maize and rice [47,48 ] suggest that, as in higher eudicots, both a DEF- and a GLO-like gene are required to provide the floral homeotic B function (sensu the classical ABC model ) in monocots. Characterisation of the DEF-like gene from rice [49] and the GLO-like genes from maize are needed, however, to test that hypothesis. Evidence showing that the DEF- and GLO-like proteins from maize and rice form heterodimeric complexes, as their orthologs from eudicots do, is already available [48,49]. After all that has been said above, one should not be too surprised to find these proteins exerting their function as part of multimeric complexes that also include AGL2-like proteins.

8 82 Growth and development During evolution, orthologous genes (DEF- and GLO-like genes) may have been recruited independently for the specification of organs that are not homologous. There is, however, a more parsimonious explanation for the fact that DEF- and GLO-like genes specify lodicule identity in grasses, and for the homeotic transformations seen in osmads4 and si1 mutant plants: it just may be that lodicules are homologous to eudicot petals and that palea and lemma are homologous to eudicot sepals (and are not a prophyll or bract, respectively, as other interpretations have it). Thus, it seems likely that the developmental program that specifies the identity of petaloid organs in the second floral whorl is a synapomorphy of all extant monocots and eudicots (or even all extant angiosperms). Monocots separated from the lineage that led to eudicots earlier than the speciation events that gave rise to lower and higher eudicots. The loss of the requirement for DEF- and GLO-like genes for the maintenance of petal identity during late developmental stages in some species of lower eudicots [46] may, therefore, represent a derived state that was established within some lineages that led to extant lower eudicots [6,48,50], rather than a basal condition that reflects several independent evolutionary derivation events of petals from stamens [46]. The battle of the sexes: on the origin of floral homeotic B-function genes The sudden appearance and strong diversification of the flowering plants within the fossil record of the Early Cretaceous, about MYA, was an abominable mystery to Charles Darwin more than a century ago and still awaits an explanation. One important reason for our difficulties in understanding how flowers originated is the problem of assigning homology between the reproductive organs of flowering plants and those of their putative ancestors. It is not clear from which organs of angiosperm relatives, and by what series of morphological transformations, the typical floral organs sepal, petal, stamen and carpel originated. Nevertheless, because the identity of these organs is specified by conserved organ identity genes, clarifying the origin and evolution of these genes may give us valuable clues about the origin of floral organs and flowers as a whole [6,15,21]. The available evidence suggests that the last common ancestor of extant ferns and seed plants, which existed about 400 MYA, did not yet have orthologs of floral homeotic genes [6 ]. The floral homeotic gene lineages must, therefore, have been established in the lineage that led to the angiosperms after the separation from the fern lineage, but probably before the radiation of the flowering plants. The critical taxon for understanding the origin of floral homeotic genes are, thus, the extant gymnosperms, comprising conifers, gnetophytes, cycads and Ginkgo [6 ]. The extant gymnosperms are probably a monophyletic group that separated from the lineage that led to angiosperms about 300 MYA [51,52]. Phylogeny reconstructions of the MADS-box gene family have previously suggested that orthologs of floral homeotic B-function genes (i.e. DEF/GLO-like genes) exist in gymnosperms [53,54]. The isolation of such genes from three different gymnosperm species, that is, the gnetophyte Gnetum gnemon, and the conifers Norway spruce (Picea abies) and Monterey pine (Pinus radiata) [44,55,56 ], was achieved only recently. This raises the question of what might be the function of orthologs of floral homeotic B-function genes in taxa that form neither stamens nor petals. It turned out that the orthologs of DEF-, GLO-, or GGM2-like genes (i.e. B genes) from Gnetum and conifers are exclusively expressed in male reproductive units [44,55,56 ], suggesting that these organs are homologous to stamens. If, in addition, one takes into account that gymnosperms do not have petals, one realises that the B genes from gymnosperms and angiosperms have similar expression patterns focused on the male reproductive organs. This work led to the hypothesis that a major function of the expression of B genes in seed plants may be to distinguish between male reproductive organs (in which B gene expression is on ) and female reproductive organs (in which B gene expression is off ) [6 ]. Differential expression of B genes may, thus, represent the primary sex-determination mechanism of all seed plants [44]. Moreover, as functionally conserved orthologs of C- and D-function genes are also present in diverse gymnosperms [44,53,57], it seems that the system for specification of reproductive organ identity in angiosperms was recruited from a similar system that was already present in the last common ancestor of all extant seed plants (about 300 MYA) [6 ]. At the molecular level, therefore, flower origin seems less abominable than the morphological difference between flowers and gymnosperm cones or strobili may suggest. Recently, some MADS-box genes were detected in both gymnosperms and angiosperms that are more closely related to B genes than to any other known MADS-box gene subfamily, although they clearly constitute a separate clade from that of the B genes [58 ]. In contrast to the B genes, which are mainly expressed in male reproductive structures (and angiosperm petals), these Bsister genes are predominantly expressed in female reproductive structures ([58 ]; A Becker, G Theißen, unpublished data). There is hope, therefore, that the B and Bsister genes will provide molecular genetic clues concerning the evolution of distinct male and female sporophylls in the lineage that led to extant seed plants. Conclusions To explain how the different floral organs adopt their unique identities during flower development, combinatorial interactions among three classes of floral homeotic genes, termed A, B, and C, were proposed about a decade ago, with A specifying sepals, A and B petals, B and C stamens and C carpels. Later, D-function genes, specifying ovules, were added to the classical ABC model. Recent studies on MADS-box genes revealed an additional class of floral homeotic genes, termed E-function genes here,

9 Development of floral organ identity: stories from the MADS house Theißen 83 that is necessary for the specification of the organ identity of petals, stamens and carpels. The molecular mode of the interaction of the floral homeotic genes remained a mystery for quite a while, but now it seems that the capacity of MADS-domain proteins to form multimeric complexes provides a molecular basis for the combinatorial interactions of the A, B, C and E genes. On the basis of these findings, alternatives to the classical ABC model, such as the A E model and the quartet model presented here, have been developed. The quartet model directly links floral organ identity to the action of four different tetrameric transcription-factor complexes composed of MADS-domain proteins. Major goals of future research will be to define the exact structures of the transcription factor complexes that work as switches during the development of floral organ identity. Which target genes do they control, and which upstream regulators control their formation? How conserved are these complexes in evolution, and to what extent did gene duplications within the MADS-box gene family contribute to the origin of these complexes? What are the implications of complex formation for the co-evolution of MADS-box genes? By cloning and characterisation of B genes from several gymnosperms and monocots, significant progress was recently made in understanding the origin and evolution of DEF- and GLO-like proteins, which are important components of some multimeric complexes and providers of the floral homeotic B function sensu the classical ABC model. The study of the B genes, and of their recently detected sister genes in both gymnosperms and angiosperms, promises to provide novel insights into the evolution of distinct male and female sporophylls in the lineage that led to extant seed plants. Acknowledgements Many thanks to Mika Kotilainen and Teemu Teeri for sending me an in press manuscript, and to the whole Teeri laboratory for interesting discussions. Many thanks also to Annette Becker, Wim Deleu, Charlotte Kirchner, Christof Weiser and Kai-Uwe Winter from our laboratory for sharing unpublished data. I would also like to thank Thomas Münster and Wim Deleu for helpful comments on the manuscript, and Heinz Saedler for stimulating discussions and continuous support. References and recommended reading Papers of particular interest, published within the annual period of review, have been highlighted as: of special interest of outstanding interest 1. Haughn GW, Somerville CR: Genetic control of morphogenesis in Arabidopsis. Dev Genet 1988, 9: Schwarz-Sommer Z, Huijser P, Nacken W, Saedler H, Sommer H: Genetic control of flower development by homeotic genes in Antirrhinum majus. Science 1990, 250: Coen ES, Meyerowitz EM: The war of the whorls: genetic interactions controlling flower development. Nature 1991, 353: Weigel D, Meyerowitz EM: The ABCs of floral homeotic genes. Cell 1994, 78: Riechmann JL, Meyerowitz EM: MADS domain proteins in plant development. Biol Chem 1997, 378: Theissen G, Becker A, Di Rosa A, Kanno A, Kim JT, Münster T, Winter K-U, Saedler H: A short history of MADS-box genes in plants. Plant Mol Biol 2000, 42: By reviewing current knowledge of MADS-box genes in ferns, gymnosperms and different types of flowering plants (i.e. basal angiosperms, monocots and eudicots), the authors demonstrate that the phylogeny of MADS-box genes is strongly correlated with the origin and evolution of plant reproductive structures such as ovules and flowers. It seems likely, therefore, that changes in MADS-box genes have been a major source of innovation in reproductive development during the evolution of land plants. Accordingly, reasonable models of the origin and evolution of the ABC system of flower organ identity specification can be suggested. In more general terms, MADS-box genes are advertised as attractive models for the study of evolutionary developmental genetics (i.e. evodevotics), a novel scientific endeavour which assumes that changes in developmental control genes are a major aspect of evolutionary changes in morphology. 7. Sommer H, Beltrán J-P, Huijser P, Pape H, Lönnig W-E, Saedler H, Schwarz-Sommer Z: Deficiens, a homeotic gene involved in the control of flower morphogenesis in Antirrhinum majus: the protein shows homology to transcription factors. EMBO J 1990, 9: Yanofsky MF, Ma H, Bowman JL, Drews GN, Feldmann KA, Meyerowitz EM: The protein encoded by the Arabidopsis homeotic gene agamous resembles transcription factors. Nature 1990, 346: Tröbner W, Ramirez L, Motte P, Hue I, Huijser P, Lönnig W-E, Saedler H, Sommer H, Schwarz-Sommer Z: GLOBOSA: a homeotic gene which interacts with DEFICIENS in the control of Antirrhinum floral organogenesis. EMBO J 1992, 11: Mandel MA, Gustafson-Brown C, Savidge B, Yanofsky MF: Molecular characterization of the Arabidopsis floral homeotic gene APETALA1. Nature 1992, 360: Jack T, Brockman LL, Meyerowitz EM: The homeotic gene APETALA3 of Arabidopsis thaliana encodes a MADS box and is expressed in petals and stamens. Cell 1992, 68: Bradley D, Carpenter R, Sommer H, Hartley N, Coen E: Complementary floral homeotic phenotypes result from opposite orientations of a transposon at the plena locus of Antirrhinum. Cell 1993, 72: Goto K, Meyerowitz EM: Function and regulation of the Arabidopsis floral homeotic gene PISTILLATA. Genes Dev 1994, 8: Jofuku KD, den Boer BGW, Van Montagu M, Okamuro JK: Control of Arabidopsis flower and seed development by the homeotic gene APETALA2. Plant Cell 1994, 6: Theißen G, Saedler H: The golden decade of molecular floral development ( ): a cheerful obituary. Dev Genet 1999, 25: This review describes the major breakthroughs in understanding the molecular genetics of floral development during the 1990s, especially the cloning of the floral meristem and organ identity genes in model systems such as Arabidopsis and Antirrhinum. 16. Colombo L, Franken J, Koetje E, van Went J, Dons HJM, Angenent GC, van Tunen AJ: The petunia MADS box gene FBP11 determines ovule identity. Plant Cell 1995, 7: Angenent GC, Colombo L: Molecular control of ovule development. Trends Plant Sci 1996, 1: Rounsley SD, Ditta GS, Yanofsky MF: Diverse roles for MADS box genes in Arabidopsis development. Plant Cell 1995, 7: Ng M, Yanofsky MF: Three ways to learn the ABCs. Curr Opin Plant Biol 2000, 3: This article reviews recent evidence showing that the floral meristem identity gene LEAFY is an immediate upstream regulator of at least some of the ABC genes in Arabidopsis. 20. Ma H, Yanofsky MF, Meyerowitz EM: AGL1 AGL6, an Arabidopsis gene family with similarity to floral homeotic and transcription factor genes. Genes Dev 1991, 5: Theißen G, Saedler H: MADS-box genes in plant ontogeny and phylogeny: Haeckel s biogenetic law revisited. Curr Opin Genet Dev 1995, 5: Theißen G, Kim JT, Saedler H: Classification and phylogeny of the MADS-box multigene family suggest defined roles of MADS-box gene subfamilies in the morphological evolution of eukaryotes. J Mol Evol 1996, 43: