MADS reloaded: evolution of the AGAMOUS subfamily genes




AGAMOUS subfamily proteins are encoded by MADS-box family genes. They have been shown to play key roles in the determination of reproductive floral organs such as stamens, carpels and ovules. However, they also play key roles in ensuring a fixed number of floral organs by controlling floral meristem determinacy. Recently, an enormous amount of sequence data for nonmodel species have become available together with functional data on AGAMOUS subfamily members in many species. Here, we give a detailed overview of the most important information about this interesting gene subfamily and provide new insights into its evolution.

I. Introduction

Transcription factors (TFs) containing the MADS domain are present in the majority of eukaryotic organisms. [MADS refers to the four founding members containing the conserved DNA binding domain (Schwarz-Sommer et al., 1990).] Evidence suggests that the MADS domain evolved from a region of topoisomerase IIA subunit A (TOPOIIA-A) in the lineage that led to the most recent common ancestor (MRCA) of extant eukaryotes (Gramzow et al., 2010). A subsequent gene duplication of that ancestral MADS-box gene presumably occurred before the MRCA of eukaryotes and gave rise to the two main clades of SRF (SERUM RESPONSE FACTOR)-like (type I) and MEF2 (MYOCYTE ENHANCER FACTOR 2)-like (type II) MADS-box genes (Gramzow et al., 2010). In Streptophyta (Charophyta algae and land plants), MEF2-like TFs are often termed MIKC-type, as they possess a modular structure where the MADS (M) domain is followed by an intervening (I), a keratin-like (K) and a C-terminal (C) domain (Ma et al., 1991; Theissen et al., 1996; Kaufmann et al., 2005). In land plants, they form the two main groups of MIKC*- and MIKCC-type TFs. The number of MADS-box genes is higher in land plants compared with any other group of eukaryotes, in particular in flowering plants, where diploid species possess c. 100 such genes (Parenicová et al., 2003; Leseberg et al., 2006; Arora et al., 2007). After their appearance, flowering plants diversified enormously during the Cretaceous period and became the largest group of the Earth's flora. Very probably, this extraordinary evolutionary success largely depended on their new reproductive structures and, like gymnosperms, on the use of seeds as a new system of propagation (reviewed by Linkies et al., 2010). In flowering plants, the MIKCC group has significantly expanded (consisting of 39 genes in Arabidopsis thaliana and 38 in Oryza sativa; Parenicová et al., 2003; Arora et al., 2007) and can be divided into 14 different phylogenetic subfamilies, of which 10 are found in all angiosperms and seven also in gymnosperms (Becker & Theissen, 2003; Arora et al., 2007; Heijmans et al., 2012a; Sang et al., 2012). None of these 14 subfamilies are shared with mosses and basal Tracheophyta (Gramzow et al., 2010; Barker & Ashton, 2013). A large body of functional studies strongly suggest that the expansion of the MIKCC group has been critical for the evolution of plant sexual reproductive mechanisms and structures (recently reviewed by Smaczniak et al., 2012a). In other words, the appearance of these genes seems to be closely linked to the successful evolution of flowering plants.

Genetic studies conducted in the early 1990s in the model plants A. thaliana and snapdragon (Antirrhinum majus) resulted in the formulation of the simple genetic ‘ABC model’ that explains the genetic regulation of floral organ identity determination (Coen & Meyerowitz, 1991; for a detailed description of the model, see Supporting Information Notes S1). Strikingly, all the genes involved in the ABC model encode MIKCC-type MADS-domain TFs, with the only exception being the A. thaliana gene APETALA 2 (AP2; Koornneef et al., 1980; Jofuku et al., 1994). A few years later, other MIKCC genes, specifically involved in the regulation of ovule development inside the ovary, have been reported in petunia (Petunia hybrida; Angenent et al., 1995; Colombo et al., 1995) and more recently also in A. thaliana (Favaro et al., 2003; Pinyopich et al., 2003), and based on the studies in petunia a D function for ovule development was added to the model. It was also shown in A. thaliana that all the MIKCC-type proteins involved in floral organ identity determination were dependent for their activity on interactions with ‘E’ function proteins, encoded by the MIKCC SEPALLATA (SEP) subfamily genes (Pelaz et al., 2000; Theissen, 2001; Ditta et al., 2004). Therefore, the original ABC model was updated to the ABC(DE) model of flower development. In the following years, it became clear that, outside A. thaliana and its Brassicaceae family, it is difficult to define a true ‘A function’, whereas the other parts of the model seem to be widely conserved among flowering plants (including monocots; reviewed by Causier et al., 2009; Cui et al., 2010; Dreni et al., 2011; Heijmans et al., 2012a,b; Yun et al., 2013).

This review focuses on the AGAMOUS subfamily of MADS-box genes, which are involved in floral development. This subfamily was named after the AGAMOUS gene of A. thaliana (AG; Yanofsky et al., 1990), which is the only C function gene in this species. In the ABC model, the C function integrates three different roles: stamen identity, carpel identity and floral meristem determinacy (FMD). In other words, the class C TF genes that are involved in the formation of the innermost floral organs have also been recruited to control floral meristem (FM) activity to ensure a fixed number of floral organs as we observe in nature. In A. thaliana, we can also assign a fourth role for the C function, which is the prevention of the misexpression of A function genes in the two whorls of reproductive organs. This is clearly evidenced by the A. thaliana ag mutant, which shows a homeotic conversion of stamens into petals, as a result of the loss of stamen identity and misexpression of the A function. In the fourth innermost whorl, in place of a carpel, a new ag flower develops, which in turn develops in its centre another new ag flower. The phenotype in the inner whorl is a result of the loss of FMD. The new inner flower generally starts with a new whorl of sepals (Bowman et al., 1989, 1991a; Fig. 1c), which can also be interpreted as the homeotic conversion of the carpel.

Figure 1.

Floral and seed phenotypes of AGAMOUS gene subfamily mutants in Arabidopsis thaliana, Antirrhinum majus, Petunia hybrida and Oryza sativa. (a) An A. thaliana wild-type flower. (b) An ag3/+ heterozygous flower showing a supernumerary petal. (c) An ag3 mutant indeterminate flower with ectopic petals replacing stamens in the third whorl, and a new sterile flower in the fourth whorl starting with sepal-like organs with petaloid margins (red arrow). (d) An Antirrhinum far mutant flower where the corolla has been removed to show the reproductive whorls. Although there are no obvious homeotic changes, far mutant flowers show a variable degree of male sterility. (e) A ple mutant flower with second and third whorl organs removed to show the fourth whorl sepaloid/petaloid/carpeloid structure (red arrow) enclosing a new flower. (f) A ple far double mutant flower, with indeterminate petal development in the fourth whorl. (g) A wild-type petunia flower with part of the corolla removed to show the stamens and the pistil; the ovary is marked by the red arrow. (h) An fbp6-1 pMADS3-RNAi flower with complete loss of reproductive organ development. (i) A detail of the new flower developing in the fourth whorl, which starts with sepals and contains in turn a new flower. (j) A wild-type rice flower with the frontal part of the lemma (L) and palea (P) removed to show a second lodicule whorl (white arrow), the six stamens and the pistil (red arrow). (k) Cleared wild-type developing ovules of A. thaliana; the red arrow indicates the funiculus. (l) Arabidopsis thaliana wild-type seed coat at the torpedo stage. Of the five cell layers, the inner one is the endothelium (E). (m) An osmads3 osmads58 double mutant flower 1 month after heading, still producing ectopic lodicules and palea-like organs. This ectopic mass of organs forced the lemma (L) and palea (P) aside and emerged from the flower. (n) An stk mutant ovule at the same stage as in (k), showing a longer funiculus (red arrow). (o) The stk abs double mutant seed coat lacks the endothelium; see for comparison the wild type in (l). Bars represent 1 mm in A. thaliana and rice flowers, 5 mm in Antirrhinum and petunia and 50 μm in A. thaliana ovules and seeds. We thank B. Davies, M. VandenBussche and L. Colombo for the Antirrhinum, petunia and A. thaliana photos, respectively.

Interestingly, the D class (ovule identity) genes also belong to the AGAMOUS subfamily. The first genes identified as master regulators of ovule identity were FLORAL BINDING PROTEIN 7 (FBP7) and FBP11 of petunia (Angenent et al., 1995; Colombo et al., 1995). Apart from AG, there are three other AGAMOUS subfamily genes in A. thaliana, namely SEEDSTICK (STK; formerly known as AGAMOUS-LIKE 11, AGL11; Rounsley et al., 1995), which is closely related to FBP7 and FBP11, and SHATTERPROOF1 (SHP1) and SHP2 (formerly AGL1 and AGL5; Ma et al., 1991; Liljegren et al., 2000), which are more closely related to AG. These three genes redundantly control ovule development in A. thaliana (Pinyopich et al., 2003).

Here, we provide an overview of relevant studies that have been performed during the last decade. Several AGAMOUS subfamily genes have been reported and also characterized in other angiosperms, such as the monocots rice (Oryza sativa) and maize (Zea mays). These studies show that these genes retain functional conservation within flowering plants, and that their functions as master regulators of stamen, carpel (precursor of the fruit) and ovule (precursor of the seed) identity, but also FMD, are ancestral. However, the number of AGAMOUS subfamily genes can vary between different species and they typically show various degrees of redundancy and/or subfunctionalization. Furthermore, several studies investigated their upstream and downstream pathways, as well as their genetic and protein–protein interactions. With all this data becoming available, we believe that it is timely to provide a simple but complete overview of these data. This implies also a discussion of the important roles that the AGAMOUS subfamily genes continue to play after fertilization in developing fruit and seed. To date, these roles have not been intensively studied, maybe because they are more difficult to investigate, as AGAMOUS subfamily knock-out and knock-down mutants are often partially or completely sterile.

II. Phylogeny and subfunctionalization within the AGAMOUS subfamily

The phylogenetic relationships among the AGAMOUS subfamily genes have been extensively studied by Kramer et al. (2004) and Zahn et al. (2006). Although members of this family have been isolated in all the extant taxa of nonflowering seed plants (Gingkophyta, Coniferophyta, Gnetophyta and Cycadophyta), they do not seem to form paraphyletic subgroups in these plants (Kramer et al., 2004; Zahn et al., 2006). On the contrary, in flowering plants several paraphyletic lineages arose, which mostly originated from whole-genome duplication events (WGDs) which characterize the evolutionary history of angiosperms. A first WGD, which occurred before the common ancestor of extant angiosperms, probably gave rise to the AG and AGL11 lineages sensu Zahn et al. (2006), previously reported as C and D lineages by Kramer et al. (2004). All the known AGAMOUS subfamily members of flowering plants cluster in one of these two groups. In core eudicots, the AG lineage further divides into the conserved euAG and PLENA (PLE) lineages (Kramer et al., 2004; Figs 2, S1). Members of all these lineages show various degrees of redundancy, subfunctionalization and neofunctionalization (Airoldi & Davies, 2012).

Figure 2.

Phylogenetic tree calculated using AGAMOUS protein sequences from basal Magnoliophyta and eudicots. Of these, 45 proteins were predicted from GenBank expressed sequence tag (EST) sequences of Asteraceae species, and none of these clustered with the PLE lineage. This finding strongly suggests that the PLE lineage is lost in Asteraceae. A subclade of Helianthus sp. AGL11-like proteins having a Q > K amino acid substitution at position 105 is shown. The dichotomy between AGAMOUS and AGL11 lineages is marked with a black star, and that between the euAG and PLE lineages with a black triangle. The analysis was performed using the phylip package version 3.6 (Felsenstein J., distributed by the author. Department of Genome Sciences, University of Washington, Seattle, WA, USA) and the neighbour-joining method (100 replicates). The genomic loci of Amborella trichopoda, peach, grapevine, poplar, tomato, potato and other sequenced species are listed in Supporting Information Table S1.

Our analysis of the currently available large-scale sequencing projects suggests that the euAG lineage is lost in the papaya (Carica papaya) genome, whereas PLE lineage genes are missing in poplar (Populus trichocarpa), Mimulus guttatus (common monkey flower) and probably also Asteraceae species (Fig. 2). The AGL11 lineage is probably lost in the basal eudicot Aquilegia coerulea (Columbine). Analysis of these sequences also showed that the previously identified amino acid position 105 (the number refers to the AGL11 lineage member OsMADS13 of rice; Dreni et al., 2007) is highly conserved in the protein sequences of the AGL11 (a conserved Q residue) and PLE (a conserved R residue) lineages, suggesting that this residue might be fundamental for specific functions. These and other important features of AGAMOUS subfamily genes and proteins are discussed and shown in more detail in Notes S2, Fig. S1 and Table S1.

1. AGAMOUS TFs have conserved but variably partitioned functions during flower development

In A. thaliana, AG (euAG lineage) is the only gene showing a full C function activity. The ag mutant completely loses male and female organ identity and FMD (Bowman et al., 1989), despite the fact that A. thaliana also possesses two PLE lineage genes, SHP1 and SHP2 (Liljegren et al., 2000). Furthermore, the ectopic expression of AG in the perianth whorls, as observed in the ap2 mutants (Drews et al., 1991) or in transgenic plants ectopically expressing AG from the 35S constitutive promoter (Mizukami & Ma, 1995), is enough to homeotically convert sepals into carpels and petals into stamens, thus suppressing the A function. The ectopic expression of other core eudicot AG lineage genes gave very similar results, not only in A. thaliana but also in, for example, transgenic petunia (Tsuchimoto et al., 1993; Kater et al., 1998; Heijmans et al., 2012b), tobacco (Nicotiana tabacum; Mandel et al., 1992; Kempin et al., 1993) and tomato (Solanum lycopersicum; Pnueli et al., 1994; Giménez et al., 2010). In general, these experiments have shown that euAG and PLE lineage proteins show a very similar ability in promoting reproductive organ development and FMD. In A. thaliana, the SHP genes do not seem to regulate these functions under natural conditions; however, SHP and AG ectopic expression causes the same homeotic conversion of perianth organs as described above (Pinyopich et al., 2003). The gene AP2 is required to suppress the expression of AG in the two perianth whorls (Drews et al., 1991; Krogan et al., 2012), and it also negatively regulates SHP1 and SHP2 in the first whorl (Savidge et al., 1995; Flanagan et al., 1996). In ap2 ag double mutants, the homeotic conversion of sepals into carpels occurs despite the absence of AG activity (Alvarez & Smyth, 1999), which is largely a result of ectopic SHP1/SHP2 activity. This was evidenced by the absence of homeotic conversions in the ap2 ag shp1 shp2 quadruple mutant (Pinyopich et al., 2003). A similar phenotype was seen in crc spt ap2 pi ag pentuple mutants (Alvarez & Smyth, 1999), indicating that SPATULA (SPT) and CRABS CLAW (CRC) regulate, in parallel to AG, distinct features of carpel development. Interestingly, it has been suggested that AG, SHP1 and SHP2 might redundantly promote SPT and CRC expression (Pinyopich et al., 2003; Lee et al., 2005). The ectopic expression of SHP2 largely complemented the ag mutant phenotype, as it restored stamen and carpel identity, FMD and also female fertility, indicating that SHP genes potentially still have C class gene activity (Pinyopich et al., 2003). In wild-type plants, the onset of expression of SHP occurs later than that of AG, being restricted to the two early carpel primordia before their fusion, and later mainly in developing ovules together with STK (Rounsley et al., 1995; Pinyopich et al., 2003; Colombo et al., 2010). SHP genes redundantly regulate, in cooperation with CRC and AINTEGUMENTA (ANT), the fusion of the ovary valve margin and the development of the stylar and stigmatic tissues. However, the unfused valves which develop in crc ant shp1 shp2 quadruple mutants still retain carpel identity (Colombo et al., 2010), further indicating that SHP genes are not contributing to the C function in A. thaliana. SHP genes are also expressed, like AG and probably in an AG-independent manner, in nectaries, which are not affected in the ag mutants (Bowman et al., 1991b; Flanagan et al., 1996; Baum et al., 2001). It is therefore possible that the development of nectaries is specified by these three genes. Although ag shp1 shp2 triple mutants have never been described, Lee et al. (2005) showed that nectary development is abolished in the ap2 pi ag mutant background only when SHP genes are also simultaneously removed, and proposed that, like AG, SHP genes can activate CRC, which is necessary for nectary development (Bowman & Smyth, 1999). Furthermore, SHP genes regulate, redundantly with STK, ovule identity and development. In stk shp1 shp2 triple mutants, ovule integuments lose their identity and are homeotically converted into carpeloid tissues (Pinyopich et al., 2003; Brambilla et al., 2007). However, AG also redundantly contributes to the D function, although protein complexes consisting only of AG and SEP are probably more effective in promoting carpel rather than ovule identity, as evidenced by the stk shp1 shp2 triple mutant phenotype (Western & Haughn, 1999; Pinyopich et al., 2003; Brambilla et al., 2007).

In conclusion, all four A. thaliana AGAMOUS subfamily genes act redundantly in regulating ovule identity (D function), although the AGL11-like gene STK seems to have unique functions in funiculus development, as this organ is abnormally increased in length in the stk mutant, indicating that STK controls cell division and expansion (Pinyopich et al., 2003). The only C function factor of A. thaliana is AG, because, although SHP proteins fully maintain a similar potential activity, they are not physically present in the meristem and primordia cells at the appropriate time. Their functions are rather restricted to regulate specific tissue types after carpel identity has been established. This is a clear example of subfunctionalization as a result of a change in expression pattern (Airoldi & Davies, 2012).

It is important to note that, because SHP genes are expressed in early carpel primordia, it remains possible that they redundantly regulate carpel identity with AG. In fact, AG seems to be required for SHP expression, and SHP transcripts were dramatically reduced in ag mutant flowers (Savidge et al., 1995). In contrast, it seems that the AGL11 lineage factor STK has no C function activity. Although in A. thaliana 35S::STK transgenic plants sepals were often converted into carpels, these conversions were associated with the ectopic expression of AG, SHP1 and SHP2, whereas petals were only reduced or completely absent, but never converted into stamens (Favaro et al., 2003). The STK-dependent conversion of sepals into carpeloid organs decreased drastically in an ag mutant background. Furthermore, in these 35S::STK ag plants, the ectopic expression of STK seemed not to be able to complement significantly the ag mutant phenotype. The ability of STK to specify carpel identity seemed to be only indirect, largely as a result of the induction of ectopic expression of AG, SHP1 and SHP2 (Favaro et al., 2003).

In addition to A. thaliana, AGAMOUS subfamily members have been functionally characterized in several other species. For instance, in petunia FBP6 and pMADS3 have been intensively studied. FBP6 belongs to the PLE lineage. The fbp6 class C mutant flowers exhibit partial anther to petal conversions (Heijmans et al., 2012b). In the gynoecium, fbp6 flowers have an incomplete fusion of the style and stigma inner tissues and a partial loss of transmitting tissue, and the stigma is transformed into sepal- or leaf-like structures. These defects reduce the fertility of the gynoecium. The silencing of the other class C (euAG) gene pMADS3 by an RNAi approach produced similar but more severe homeotic conversions of stamens, with a reduction in the amount of pollen, whereas the gynoecium developed normally. Plants homozygous for fbp6 carrying the pMADS3-RNAi construct showed a full C function mutant phenotype, with the nearly complete or complete loss of reproductive organ development and the formation of a new flower in the fourth whorl, starting with sepals, and the reiteration of inner flowers was repeated a few more times (Heijmans et al., 2012b; Fig. 1h,i).

The snapdragon ple-1 class C mutant is not a complete null mutant. Its reproductive organs are replaced by perianth organs and the flower becomes indeterminate, but occasionally still producing small amounts of pollen in the third whorl petaloid organs (Davies et al., 1999). Conversely, mutants in the FARINELLI (FAR, euAG lineage) gene did not show floral homeotic changes (Fig. 1d) and the observed phenotype was only a variable degree of male sterility, attributable to degradation of microspores and tapetal cells. In the ple-1 mutant, the stamens are replaced by unfused petals, still conserving some staminoid appearance, whereas in ple-1 far double mutants these organs are fused and show a more pronounced petaloid morphology. In ple-1, variable sepaloid/petaloid/carpeloid organs were observed in the fourth whorl, inside which a new flower was initiated, composed of mixed sepaloid/petaloid tissue (Fig. 1e). In contrast, the fourth whorl of ple-1 far double mutants consisted of a new tube and corolla of petals followed by further petals in a spiral phyllotaxy (Fig. 1f). These data suggest that, in snapdragon, in contrast to A. thaliana, the C function is mainly regulated by a PLE lineage gene. However, FAR expression decreased consistently in the ple-1 mutant, whereas PLE expression expanded throughout the developing anther and carpel in the far mutants, being more localized in the anther stomium and ovules in the wild type, suggesting that FAR negatively regulates PLE and that in far mutants the expansion of PLE might compensate the loss of FAR activity. Therefore, it is difficult to determine the degree of redundancy and subfunctionalization that exists between the two genes (Davies et al., 1999). Ectopic expression studies in tobacco and snapdragon suggested that FAR has the ability to induce stamen identity, even more so than PLE (Davies et al., 1999; Causier et al., 2005). Furthermore, the ectopic expression of FAR in an A. thaliana ag mutant background restored stamen and carpel identity in the third and fourth whorls, respectively, but not FMD (Airoldi et al., 2010).

Recently, the PLE and euAG-like genes NbSHP and NbAG of Nicotiana benthamiana have also been functionally characterized by virus-induced gene silencing (VIGS; Fourquin & Ferrándiz, 2012). The development of efficient VIGS protocols allowed the characterization of AG lineage genes even in plant taxa not readily transformable, such as the basal eudicots opium poppy (Papaver somniferum; Hands et al., 2011) and California poppy (Eschscholzia californica; Yellina et al., 2010), both belonging to the Papaveraceae family. These experiments led again to typical C mutant phenotypes. Interestingly, a single AG lineage gene has been identified in opium poppy, but encoding two alternative transcripts, PapsAG-1 and PapsAG-2, producing distinct proteins with different lengths of the C-terminal domain downstream of the typical AG motif II. These two proteins retain both functional redundancy and unique functions (Hands et al., 2011).

AGAMOUS subfamily members have also been analysed in monocot species. Functional analysis using loss-of-function mutants have only been reported in the domesticated grasses rice and maize. In rice, OsMADS3 and OsMADS58 belong to the AG lineage. OsMADS3 appears to be more important for stamen identity and stamen development (Yamaguchi et al., 2006). In osmads3 mutants, carpel identity and FMD are just weakly impaired, and only when the osmads3 mutant was combined with the osmads58 mutant was a full C function mutant phenotype observed (Dreni et al., 2011). The osmads3 osmads58 double mutant flowers completely lose FMD and reproductive organ identity, with ectopic lodicules replacing stamens and one or more small palea-like organs developing in place of the pistil (Fig. 1m). Thus, this double mutant largely mimics the A. thaliana ag mutant. In maize, ZMM2 (ZEA MAYS MADS 2; Theissen et al., 1995) and ZAG1 (ZEA AGAMOUS 1; Schmidt et al., 1993) are orthologous to OsMADS3 and OsMADS58, respectively, and they appear to be much more subfunctionalized in regulating stamen development in the tassel and carpel development in the ear, respectively. This is supported by their expression profiles and by the phenotype of the zag1 mutants (Mena et al., 1996; Ambrose et al., 2000). Unfortunately, zmm2 mutant alleles have not yet been reported in the literature. Furthermore, a second OsMADS3 orthologue exists in maize, named ZMM23 (Münster et al., 2002); however, this gene has not yet been functionally characterized.

In conclusion, all these functional analyses across flowering plants strongly support the idea that duplicated AG lineage genes have different degrees of subfunctionalization of the C function (see also Notes S3).

There is also quite extensive information available about the AGAMOUS subfamily members controlling ovule identity. As mentioned in the Introduction, in A. thaliana the four AGAMOUS subfamily genes all redundantly regulate ovule identity. In petunia, based on co-suppression studies, the AGL11-like genes FBP7 and FBP11 were initially reported to be essential in controlling ovule identity (Angenent et al., 1995; Colombo et al., 1995). However, recently it was shown that significant homeotic conversions of ovules into carpeloid structures were only obtained when the fbp7 fbp11 double mutant was combined with the fbp6 class C gene mutant, or with a pMADS3-RNAi silencing construct (Heijmans et al., 2012b). This suggests that a high redundancy between AG and AGL11 lineage genes in regulating ovule identity seems to be maintained within core eudicots. Recently, we reported the functional characterization of the rice AGL11 lineage genes OsMADS13 and OsMADS21 (Dreni et al., 2007, 2011). Despite the fact that this monocot species has two genes in this clade, only OsMADS13 seems to be an ovule identity gene. Nevertheless, preliminary experiments indicated that OsMADS21 is partially able to complement the osmads13 mutant phenotype, when expressed under the control of the OsMADS13 putative promoter. Furthermore, even though the two AGAMOUS homologues OsMADS3 and OsMADS58 are also expressed in developing ovules, the osmads13 single mutant shows strong homeotic ovule to carpeloid organ conversions. This phenotype is similar to those of the A. thaliana stk shp1 shp2 triple mutant and the petunia fbp6 fbp7 fbp11 or pMADS3-RNAi fbp7 fbp11 lines, indicating a different scenario from that in core eudicots. OsMADS13 also regulates, redundantly with OsMADS3 and OsMADS58, the identity of the ovary wall adaxial epidermis, thus retaining some aspects of the C function (Dreni et al., 2011).

Despite the regulation of FMD being originally assigned to AG homologues, our experiments (Dreni et al., 2011) and those of Li et al. (2011a) showed that in rice OsMADS13 strongly contributes to FMD. A redundancy between AG and AGL11 lineage genes in this function was also recently proposed for petunia (Ferrario et al., 2006; Heijmans et al., 2012b), thus suggesting that it might be a common feature of flowers with a central placentation type, while in plants like A. thaliana, with a parietal placentation, the FM terminates in the carpel primordium, and thus those AGAMOUS subfamily genes with ovule-specific expression do not participate in the process of FMD (Colombo et al., 2008).

2. Expression domain of the AGAMOUS subfamily genes

The spatial and temporal expression profile of AGAMOUS subfamily genes is closely consistent with their conserved function and subfunctionalization. The expression of AG lineage genes in the floral meristem after the emergence of the perianth organ primordia, and in both the third and fourth floral whorls, is probably the ancestral expression profile of these genes (Kramer et al., 2004; Zahn et al., 2006), as it is the most frequently observed pattern in basal eudicots (Di Stilio et al., 2005; Zahn et al., 2006; C; Yellina et al., 2010; Hands et al., 2011; Hu et al., 2012), core eudicots and monocots (Notes S4). However, there are clear exceptions to this conserved pattern, as observed in poplar and apple tree (Malus domestica), where a weak expression of AG lineage genes was observed in vegetative tissues (Brunner et al., 2000; van der Linden et al., 2002) Furthermore, in apple the euAG gene MdMADS15 is also clearly expressed in sepals and in the receptacle (van der Linden et al., 2002).

Within the AGL11 lineage, gene expression is usually specific for the ovary placenta, ovule primordium and all the stages of ovule development. The maize gene ZAG2 is also expressed in the central domain of the developing silk (Schmidt et al., 1993). This expression pattern of AGL11-lineage genes is consistent with a role restricted to ovule development. In mature ovules, the expression is often predominant in the integuments. However, there are also exceptions; for instance, the cucumber (Cucumis sativus) gene CUM10 (also named Cucumis AGAMOUS 1, CAG1) is strongly expressed in the stamens, carpels and nectaries of both male and female flowers (Kater et al., 1998; Perl-Treves et al., 1998). Although the cotton (Gossypium barbadense) gene GbAGL1 is mainly expressed in developing ovules, weak expression was also shown in the early petal and stamen primordia (Liu et al., 2010a). In rice, OsMADS21 is expressed in reproductive organs at early stages of their development, later becoming predominantly expressed in the ovule integuments, but its expression level is significantly lower than those of OsMADS3, OsMADS13 and OsMADS58 (Arora et al., 2007; Dreni et al., 2007, 2011). Kramer et al. (2004) hypothesized that the ovule-specific expression of many STK lineage genes has somehow evolved from an ancestral situation in which these genes were broadly expressed in the male and female reproductive organs, which is further supported by the observation that in gymnosperms the AGAMOUS genes are expressed in microsporophylls, megasporophylls, and ovules (reviewed by Kramer et al., 2004; Englund et al., 2011).

3. AGAMOUS subfamily protein–protein interactions

Genetic and molecular studies in A. thaliana revealed that floral organ identity MADS-domain factors are dependent for their function on the interaction with SEP (E function) MADS-domain proteins. SEP proteins are thought to act as a ‘bridge’ allowing the formation of higher order complexes and to add transcriptional activator activity to these complexes (Honma & Goto, 2001; Pelaz et al., 2001). Based on this knowledge, the ABCDE model was translated into the ‘quartet model’ of MADS-domain proteins (Theissen, 2001; Theissen & Saedler, 2001). The model predicts that a dimer composed of two MADS-domain proteins binds a specific target sequence, named the CArG box (CCA/T6GG; reviewed by Kaufmann et al., 2005), in a promoter region and that another MADS-dimer binds another CArG box. Subsequently, these two dimers interact to form a tetramer (‘quartet’) and loop the DNA that lies between the two CArG boxes (Melzer et al., 2009; Melzer & Theissen, 2009; Smaczniak et al., 2012b; Notes S5).

AGAMOUS subfamily proteins efficiently dimerize with SEP proteins. According to the quartet model, in the third floral whorl of A. thaliana, the B function heterodimer APETALA3-PISTILLATA (AP3-PI) interacts with the AG-SEP heterodimer to form the functional stamen identity tetrameric complex (Fig. 3). In the fourth innermost whorl, the interaction of two AG-SEP heterodimers forms the FMD and carpel identity complex. Ovule identity complexes are established by complexes of SEP, STK and/or AG and SHP proteins (Favaro et al., 2003; Fig. 3). Very recently, Mendes et al. (2013) provided in vitro and in vivo evidence for the existence of SEP3-STK quartets and for the importance of these tetramers in the regulation of the expression of target genes during ovule and seed development.

Figure 3.

Conserved MADS-box tetrameric complexes regulate stamen (yellow), carpel (green) and ovule (purple) identity in Arabidopsis thaliana (left) and rice (right). SEPALLATA (SEP) proteins are shown in black, AP3/PI in blue and AGAMOUS in red. In rice, the names of MADS-domain factors are abbreviated and only the number is shown. In rice, the AGL6-like factor OsMADS6 is not present in anthers but seems redundant with SEP proteins in the fourth whorl. However, as the single osmads6 mutant already shows obvious phenotypes, it could also form alternative complexes and have specific functions. The floral meristem determinacy (FMD) is regulated by the carpel identity complexes and in rice probably also by the ovule identity complex.

Based on the high functional conservation of the AGAMOUS subfamily members in many species and the fact that they show similar interaction patterns (Davies et al., 1996; Ferrario et al., 2003; Vandenbussche et al., 2003), analogous tetrameric complexes are predicted to be formed, at least within core eudicots. The complexes, however, do not seem to contain only MADS-domain proteins. For instance, in A. thaliana, molecular and genetic evidence suggests that AG-SEP complexes interact with the homeodomain TF BELL1 (BEL1) to regulate ovule development, in particular to repress WUSCHEL in the chalaza to control outer integument development (Brambilla et al., 2007).

In contrast to the AGAMOUS subfamily, the SEP subfamily is only found in angiosperms, and is probably closely related to the AGL6 subfamily, which in contrast is found in all seed plants (Zahn et al., 2005). The involvement of SEP in the floral meristem and floral organ identity complexes suggests that the rise of this subfamily represented a crucial step in the evolution of the flower (Zahn et al., 2005). The four SEP factors in A. thaliana are SEP1, SEP2 and SEP4 (AGL2/3/4 or LOFSEP group) and SEP3 (AGL9 or SEP3 group; Malcomber & Kellogg, 2005; Zahn et al., 2005). Despite the fact that these four factors are highly redundant in specifying the floral state, it is evident that SEP3 is the most important and that it forms the most effective B, C and D function complexes (Honma & Goto, 2001; Favaro et al., 2003; Ditta et al., 2004). In rice, OsMADS3 and OsMADS13 seem to interact mainly with the two SEP3 orthologues OsMADS24 and OsMADS45 (Favaro et al., 2002; Cooper et al., 2003; they are allelic to OsMADS8 and OsMADS7, respectively; Kang et al., 1997), rather than with LOFSEP orthologues (reviewed by Dreni et al., 2013). This suggests that SEP3-like factors might be the most important direct partners of AGAMOUS factors in all flowering plants. Besides OsMADS24 and OsMADS45, a third strong interactor with OsMADS13 is OsMADS6 (Favaro et al., 2002), which belongs to the AGL6 subfamily; for more information about this, see Notes S6.

In 1995, Sieburth et al. showed that in A. thaliana the functions of AG in the specification of stamen identity, carpel identity and FMD are genetically separable, as they depend on specific parts of the K-box or specific amino acid residues within it. The most likely explanation for this finding is that different parts of the K-box are required for specific protein–protein interactions. Thus, particular mutations could specifically affect the formation or the functionality of just one complex.

Interestingly, the ectopic expression of FAR (35S::FAR) in both snapdragon and A. thaliana mostly results in the homeotic conversion of petals into stamens, but sepals do not turn into carpeloid organs like they do when other AG lineage genes are used (Causier et al., 2005). Airoldi et al. (2010) were able to elucidate the molecular mechanism behind this peculiar behaviour of FAR through the heterologous expression of artificial variants in A. thaliana. These experiments revealed that the inability of FAR to induce carpel identity in the first whorl is caused by the presence of an additional glutamine in the K3 helix of the protein. This additional amino acid limits FAR to interact in A. thaliana only with SEP3, which is not expressed in the first whorl at early developmental stages. The result of this is that the ectopic expression of FAR cannot lead to the formation of carpel identity complexes in the first whorl of A. thaliana, as a consequence of the absence of the only SEP partner of FAR. However, the ectopic expression of FAR in an ag mutant background was able to restore both stamen identity in the third whorl and carpel identity in the fourth whorl (but not FMD), because of the concomitant presence of SEP3.

Conversely, a variation in the Medicago truncatula SHP homologue protein, MtruSHP, strengthened its interaction with SEP3 in yeast and was associated with the coiled fruit formation observed in this and related species. Other Medicago species having the threonine residue develop uncoiled pods (Fourquin et al., 2013). This threonine-to-alanine variation is located in the C terminus immediately upstream of the conserved AG motif I that was identified by Kramer et al. (2004). Therefore, it is possible that this conserved motif has an important function in mediating protein–protein interactions.

The data as described in this paragraph suggest that MIKCC MADS-box factors establish interactions to form higher order complexes, which determine the function of the individual proteins. These data therefore indicate that proteins forming evolutionarily conserved complexes must have constraints in the modification of their sequence as they need to ‘coevolve’, because a mutation in one of the proteins that impairs their interaction, or the activity of the complex, will be negatively selected, when it is not masked by redundancy or compensated by some other change in the partner protein. These observations also provide a reason why duplicated copies of MADS-box genes are frequently observed in plant genomes. The presence of redundant homologous gene copies removes their functional constraint, leaving one or both of them free to sub- or neofunctionalize (Airoldi & Davies, 2012). Furthermore, the ‘balance hypothesis’ suggests that duplicated copies of transcription factor-encoding genes, obtained by WGDs as often occurred in plants, are maintained to prevent imbalance in the concentration of the subcomponents of these protein complexes (Maere et al., 2005). This is probably also the case for the AGAMOUS subfamily members, where concentrations of the AGAMOUS subfamily proteins and their SEP partners are probably important (Brambilla et al., 2007).

III. Genetic interactions between B and C function genes

A typical C function knock-out phenotype is described as the conversion of stamens to petals and the formation of a new sterile flower in place of the fourth whorl, which starts with a whorl of sepals. However, this is observed only in A. thaliana ag and petunia fbp6 pMADS3-RNAi mutants, whereas in snapdragon ple-1 far double mutants there are not true inner flowers, but only new whorls of petals (Davies et al., 1999). Thus, it has been proposed that in snapdragon the AG lineage genes are required to repress the B function genes in the fourth whorl, a function that might be common in basal and core eudicots, as similar phenotypes have also been reported in California poppy, opium poppy and N. benthamiana (Yellina et al., 2010; Hands et al., 2011; Fourquin & Ferrándiz, 2012). In A. thaliana, the repression of B genes in the fourth whorl may be exclusively performed by the C2H2-type zinc finger protein SUPERMAN/FLO10 (SUP; Schultz et al., 1991; Bowman et al., 1992; Sakai et al., 1995). Interestingly, AG, AP3 and PI are all positive regulators of SUPERMAN, which once activated acts in defining the boundaries between the third and fourth whorls (Sakai et al., 2000). However, it is important to note that, in the A. thaliana ag mutants, the repetition of concentric flowers is not really regular but quite variable in the number of ectopic petals that develop between two adjacent whorls of sepals, and the number of ectopic sepals in each whorl is often reduced to three or in most cases two (perhaps reflecting the number of carpel primordia in the wild-type flower). Moreover, the ectopic sepals are mostly chimeric organs composed of petal-like white tissue on the margins and green sepal-like tissue in the middle, which is sometimes greatly reduced (Bowman et al., 1989, 1991a; L. Dreni et al., pers. obs.; Fig. 1c). Therefore, even in A. thaliana AG could play a role in the repression of the B function in the fourth whorl. The hypothesis that SHP genes are redundant with AG in this function is difficult to test, because they are not thought to be expressed in the fourth whorl of ag mutants (Savidge et al., 1995; Flanagan et al., 1996).

In the rice osmads3 osmads58 double mutant, the carpel is replaced by an ectopic palea-like organ, which never develops lodicule-like tissues. This suggests that the rice AG lineage genes are probably not required to exclude the expression of B function genes in the fourth whorl (Dreni et al., 2011). In rice, the YABBY gene DROOPING LEAF seems to be a determinant for the repression of B genes in the fourth whorl (Nagasawa et al., 2003).

In A. thaliana, AP3 and PI repress genes involved in carpel development (Wuest et al., 2012). The snapdragon B function mutants deficiens and globosa (def and glo) lack a fourth whorl, indicating that B genes are required to prevent the premature FM termination orchestrated by the AG genes (Tröbner et al., 1992). Such a function is not observed in A. thaliana, and conversely in maize and rice the B genes could even have an opposite role, as indeterminacy increased significantly in bc double mutants (Ambrose et al., 2000; Yun et al., 2013). In grasses, other mechanisms might act against the premature FM suppression, such as the fine-tuning of AGAMOUS gene expression in the FM, which seems to differ between rice and A. thaliana. In A. thaliana, AG expression is first detected in the whole FM between late stage 2 and early stage 3 (Drews et al., 1991), well before stamen primordia arise at stage 5. In rice, OsMADS3 activation in the FM occurs significantly later than AG in A. thaliana and is initially observed only in the lateral cells which will give rise to the stamen primordia (Yamaguchi et al., 2006; Dreni et al., 2011). At the same stages, OsMADS58 expression was weakly observed throughout the FM, but our experiments clearly showed that this gene alone, without the support of OsMADS3 and OsMADS13 (which is activated later), is not able to lead to FMD. In fact, we showed that in rice, like in A. thaliana, FMD is most sensitive to the levels of AGAMOUS gene activities (Mizukami & Ma, 1995; Dreni et al., 2011; Fig. 1b).

The bc double mutants of A. thaliana have indeterminate flowers composed of only sepals (Bowman et al., 1991a; Coen & Meyerowitz, 1991; Meyerowitz et al., 1991). This phenotype is very similar to that of the sep1 sep2 sep3 triple mutant (Pelaz et al., 2000) and in agreement with the ABCDE model. Partial bc double mutants of snapdragon, maize and rice show comparable phenotypes (Davies et al., 1999; Ambrose et al., 2000; Yun et al., 2013).

In conclusion, several modifications of the genetic interactions occurring between MIKCC and other genes underlie the generally highly conserved B and C functions. The monocot Lacandonia schismatica (Triuridaceae) is, among flowering plants, the only bisexual species bearing carpels in the third whorl surrounding central stamens, thus representing a homeotic phenotype of the general floral model. This surprising floral organization is correlated with the inversion of the expression domain of the AP3 orthologous gene LsAP3 (ABC to ACB model; Álvarez-Buylla et al., 2010; Garay-Arroyo et al., 2012). It will therefore be really interesting to functionally characterize in this plant species B and C function genes, their genetic interactions occurring within the reproductive whorls and the way in which they regulate FMD.

IV. Regulatory pathways during flower development

The similar expression domains and functions of the AGAMOUS subfamily genes in different species suggest that both their upstream regulators and target pathways are conserved. Within the highly similar gene structure of AGAMOUS genes, it has been shown that the first and second introns are essential for their correct expression pattern. The A. thaliana AG second intron fused to a 35S minimal promoter was sufficient to confer the AG expression pattern to a GUS reporter gene (Busch et al., 1999; Hong et al., 2003). The activation of AG in the centre of the FM is orchestrated by a number of chromatin remodelling factors, transcription factors and miRNAs (reviewed by Causier et al., 2009; Airoldi et al., 2010), acting as enhancers or repressors. Among the enhancers, the transcription factors LFY and WUS are essential for AG activation and bind to cis-elements located in the second intron (Lenhard et al., 2001; Lohmann et al., 2001; Ikeda et al., 2009). Causier et al. (2009) found that previously identified cis-elements located in the AG second intron, a 70-bp region characterized by repeats of CCAATCA, an aAGAAT box and a LFY-binding site (Davies et al., 1999; Hong et al., 2003), are mostly conserved in the second intron of AG lineage genes of core eudicot species and rice. Recent evidence suggests that the tritorax factors ULTRAPETALA 1 (ULT1) and ULT2 are direct or indirect activators of AG (Monfared et al., 2013). In 1991, Drews et al. reported that AG is ectopically expressed in the two outer whorls of ap2 mutants. However, only recently it has been demonstrated that AP2 is a direct repressor of AG that binds to elements in the second intron (Yant et al., 2010; Dinh et al., 2012).

The MIKCC factors AP1, SVP and AGL24 are also direct repressors of AG, AP3 and PI during stages 1 and 2 of flower development, and the AP1-SVP and AP1-AGL24 dimers act by recruiting the LEUNIG-SEUSS (LUG-SEU) co-repressor complex (Sridhar et al., 2004, 2006; Gregis et al., 2006, 2009). In the AGL11 lineage gene STK, the first (leader) intron is essential for ovule-specific expression (Kooiker et al., 2005). We recently reported that the BASIC PENTACYSTEINE (BPC) factors bind target GAGA boxes across the STK upstream region and the leader intron to restrict the STK expression domain (Kooiker et al., 2005; Simonini et al., 2012). Interestingly, BPCs were shown to be essential for recruiting the SVP-AP1-LUG-SEU repressor complex to the STK promoter. This mechanism may be conserved, as we found that GAGA boxes are also abundant in the corresponding region of OsMADS13, which is also important to confer ovule-specific expression (Dreni et al., 2011).

The rice F-box protein ABERRANT PANICLE ORGANIZATION 1 (APO1) is very probably an indirect positive regulator of the rice C function genes OsMADS3 and OsMADS58 (Ikeda et al., 2005, 2007), as apo1 flowers show in the two innermost whorls phenotypes similar to those of mild C function mutants (Dreni et al., 2011). Its closest A. thaliana homologue is UNUSUAL FLORAL ORGANS (UFO; Samach et al., 1999), which conversely is a positive regulator of the B class genes AP3 and PI and is probably directly targeted by AG (O'Maoiléidigh et al., 2013).

The onset of A. thaliana B, C and E gene expression does not seem to be interdependent (Pelaz et al., 2000); however, feedback regulatory loops are subsequently established. The third whorl complex AP3-PI-SEP3-AG has been proposed to be involved in a feedback auto-activation ensuring the proper expression level of its subunit-encoding genes. This is supported by the observation that AG binds in vivo to CArG boxes in its own regulatory region but also to AP3, PI, SEP3 and CRC (Gómez-Mena et al., 2005; O'Maoiléidigh et al., 2013). A similar model has also been proposed for the SEP3-AG-SEP3-AG complex in the fourth whorl. Sridhar et al. (2006), hypothesized that the binding of AG to SEP3 might also be responsible for sequestering SEP3 from the SEP3-LEU-SEU repressor complex, to switch off the repression of AG in the two innermost whorls. In rice, the AGL6-like factor OsMADS6, which has SEP-like functions, is reported to genetically interact with B, C and D class genes, and to bind in vivo a region of the OsMADS58 second intron (Li et al., 2011b), which indicates regulatory interactions also in this species.

Changes in the concentrations of AGAMOUS factors often cause variations in the expression level and/or pattern of other members of the subfamily. For example, the ectopic expression of STK in A. thaliana was able to induce AG, SHP1 and SHP2 in the first floral whorl (Favaro et al., 2003). In snapdragon, FAR expression is strongly reduced in the ple-1 mutant, whereas PLE expands its expression domain in far mutants (Davies et al., 1999). In contrast, the expression of the PLE orthologues SHP1 and SHP2 in A. thaliana is largely AG-dependent (Savidge et al., 1995; Flanagan et al., 1996), and the direct binding of AG to a CArG box located in the SHP2 promoter strongly suggests that this regulation is direct (Savidge et al., 1995). Several AG downstream genes have already been identified. AG terminates FM activity by repressing its activator WUS. This repression is either direct, through the recruitment of the chromatin remodelling Polycomb complex (Liu et al., 2011), or indirect, through activation of other WUS repressors such as KNUCKLES (Sun et al., 2009) which encodes a C2H2 zinc-finger protein. The molecular mechanism by which AG suppresses WUS needs further study, and in this regard it is interesting to note that, despite AG starting to be expressed in the FM between late stage 2 and early stage 3 of flower development, the transcription of WUS is suppressed only after stage 6 (Drews et al., 1991; Mayer et al., 1998; Lenhard et al., 2001). Despite the function of AGAMOUS factors in promoting FMD being universal in flowering plants, it is not clear if it is produced via similar mechanisms across different species. In particular, it is not clear to what extent the well-characterized WUS/CLV (CLAVATA) pathway is conserved in monocots, as discussed in our recent review (Dreni et al., 2013).

After establishing stamen identity, A. thaliana AG is also involved in late stamen development and microsporogenesis, and directly promotes the expression of the genes encoding DEFECTIVE IN ANTHER DEHISCENCE1 (DAD1), a catalytic enzyme of jasmonic acid (Ito et al., 2007), and NOZZLE/SPOROCYTELESS (Ito et al., 2004). Although in rice OsMADS3 seems to play a similar role in late stamen development, this might be achieved by regulating different downstream pathways, as it is involved in ROS (reactive oxygen species) homeostasis (Hu et al., 2011).

To date, several other direct targets of AG have been identified (Notes S7) and, during the preparation of this manuscript, a genome-wide analysis of AG direct targets was published which showed that AG regulates a plethora of different developmental pathways (O'Maoiléidigh et al., 2013). Interestingly, the authors showed that AG is important for the repression of the leaf developmental programme in floral primordia by directly controlling key regulatory genes.

Chromatin immunoprecipitation-DNA sequencing (ChIP-Seq) experiments indicated that SEP3 targets in vivo nearly 3400 genes (Kaufmann et al., 2009). A subset of these genes are thought to be direct targets of the AP3-PI-SEP3-AG stamen identity and SEP3-AG-SEP3-AG carpel identity complexes in the third and fourth whorls, respectively. Kaufmann et al. found that the identified SEP3 targets in the wild-type and ag mutant backgrounds overlap by nearly 70% (FDR < 0.001). However, this overlap increased to 90% when considering only highly enriched genes, thus suggesting that only a small fraction of the strongly enriched SEP3 targets are specific for the stamen and carpel identity complexes containing AG.

To date, the only direct target gene identified for STK is the REM (REPRODUCTIVE MERISTEM) transcription factor VERDANDI (VDD; Matias-Hernandez et al., 2010). VDD is required for embryo sac development and synergid cell identity, showing that STK also controls more downstream processes such as female gametophytic development.

V. Fruit and seed development

The function of AGAMOUS subfamily TFs is not limited to flower development, as many of them are still active in the ovaries after fertilization, where they play important functions in seed and fruit development. Colombo et al. (1997) provided the first evidence of such functions. They showed that transgenic petunia plants in which FBP7 and FBP11 were down-regulated by cosuppression could develop some normal ovules which, however, generated abnormal seeds. FBP7 and FBP11 are both expressed in the ovule integument and, after fertilization, in the derived seed coat. In contrast to the round wild-type seeds, transgenic seeds were shrunken as a result of endosperm degeneration. Furthermore, embryo development was delayed or, in some of these seeds, arrested. These defects have a maternal (sporophytic) origin and are related to the degeneration of the endothelium layer. Recently, the analysis of stable fbp7 fbp11 double knock-out mutants gave similar phenotypes to the cosuppression lines, indicating that the petunia AGL11-like genes have unique functions in seed development.

In A. thaliana, stk mutant seeds are also abnormal, as they are rounder and smaller (Pinyopich et al., 2003). Early during ovule development, an enlargement of the funiculus is visible but no other defects are evident (I. Roig-Villanova, C. Mizzotti & L. Colombo, pers. comm.; Fig. 1n), suggesting that the seed defects occur after fertilization. In the stk mutant, the mature seeds also fail to detach from the funiculus, as the abscission zone does not differentiate properly during seed development (Pinyopich et al., 2003). Recently, it has been shown that STK and ARABIDOPSIS BSISTER (ABS; a MIKCC gene belonging to the Bsister subfamily; Nesi et al., 2002) redundantly regulate endothelium development (Mizzotti et al., 2012). In the stk abs double mutants, mature ovules and seeds lack the endothelium (Fig. 1o), and a massive starch accumulation is observed in the embryo sac. This abnormal amount of starch probably explains, at least in part, the strong reduction in embryo sac fertilization that has been observed in stk abs double mutants. Furthermore, seed abortion often occurred.

In cotton (Gossypium barbadense), the orthologous gene GbAGL1 is strongly expressed in the outer integument and in the commercially important fibres that develop from the ovule epidermal cells (Liu et al., 2010a). In grapevine (Vitis vinifera), the AGL11-like gene VvMADS5 is specifically expressed in the carpels of female flowers (Boss et al., 2002), and is probably allelic to VvAGL11, which has been reported to be the major candidate gene for seedlessness (Mejía et al., 2011). Thus, AGL11-like genes appear to have unique functions in core eudicots during seed development. However, many euAG- and PLE-like genes such as AG, SHP1 and SHP2 in A. thaliana, STAG1 (strawberry AGAMOUS 1) in Fragaria × ananassa and PsAG in Prunus serotina are also preferentially expressed in the endothelium of mature ovules and developing seeds, although their functions are still mostly unknown (Bowman et al., 1991b; Savidge et al., 1995; Flanagan et al., 1996; Rosin et al., 2003; Liu et al., 2010b).

In A. thaliana fruits, SHP1 and SHP2 promote the lignification of valve margin cells adjacent to the dehiscence zone. In the shp1 shp2 double mutant plants, siliques are characterized by a reduction in valve margin cell lignification, which results in the absence of dehiscence zones, and thus fail to disperse seeds (Liljegren et al., 2000). A similar function for PLE-like genes has been reported for N. benthamiana, where the VIGS-induced silencing of NbSHP led to an absence of lignification along the fused carpel margins, which resulted in indehiscent capsules (Fourquin & Ferrándiz, 2012). The PLE-like gene of peach (Prunus persica), PpPLE, has also been proposed as a possible regulator of fleshy fruit formation (Tadiello et al., 2009) and lignification (Tani et al., 2007). Interestingly, an amino acid change in a PLE-like gene seems to be responsible for the strong lignification in the valve margins and the coiled pod formation typical of several Medicago species (Fourquin et al., 2013). The tomato counterpart, TAGL1 (TOMATO AGAMOUS-LIKE 1), is necessary for fruit development and ripening, and its ectopic expression is also sufficient to induce ripening in sepals (Itkin et al., 2009; Vrebalov et al., 2009; Giménez et al., 2010; Pan et al., 2010). These reports suggest that PLE-like genes were recruited early in the evolution of core eudicots in regulating essential aspects of fruit development and dehiscence. Furthermore, it seems that lignification and ripening are controlled by a similar regulatory pathway downstream of PLE lineage genes, suggesting a common evolutionary origin, and indeed a switch between dry dehiscent and fleshy fruit occurred frequently during plant evolution (Pabón-Mora & Litt, 2011; Fourquin & Ferrándiz, 2012) even within the same families, for example in Nicotiana and Solanum.

The apple euAG gene MdMADS15 is also expressed in sepals and the receptacle, which is unusual for an AGAMOUS homologue. In apple, the accessory fleshy fruit tissues partially originate from that organ and therefore it has been proposed that MdMADS15 might be essential for this process (van der Linden et al., 2002).

Molecular studies in the monocot banana (Musa acuminata) revealed that one of its AGL11 lineage genes, MuMADS1/MaMADS5, is probably involved in processes of fruit ripening (Liu et al., 2009; Elitzur et al., 2010; Roy Choudhury et al., 2012; Liu et al., 2013), whereas a related gene of oil palm (Elaeis guineensis) regulates the development of the lignified shell surrounding the kernel, with important implications for oil yield (Singh et al., 2013). In rice, all four AGAMOUS subfamily members are highly expressed in early stages of kernel development (Arora et al., 2007), suggesting that also in rice they might play a role in seed development.

It is very interesting that some gymnosperms independently evolved fruit-like structures, such as the aril in Taxus baccata and the ovule outer integument of Ginkgo biloba that grows and becomes fleshy. Apparently, AGAMOUS subfamily genes and molecular pathways leading to fruit ripening and maturation might also have been recruited for the development of these fruit-like structures (Lovisetto et al., 2012).

VI. Conclusions

The AGAMOUS gene subfamily probably arose in the common ancestors of extant seed plants, where it was recruited to regulate the identity and development of sexual reproductive organs. The ancestral function in regulating male and female organ development and FMD is highly conserved in angiosperms, but in these plants, with the evolution of novel structures and mechanisms of seed dispersion, these genes also acquired new functions, for example the control of fruit dehiscence and ripening and of seed detachment. Genomes of nonmodel systems are being sequenced, and this information will reveal further insights into the evolution of this important gene family. Future functional analysis in model and nonmodel plants and the analysis of the pathways that are controlled by these key regulators will provide further insights into the regulatory pathways they control and will probably bring us closer to understanding how the AGAMOUS subfamily members contributed to the spectacular evolutionary success of flowering plants.

It is not easy to explain why multiple AGAMOUS lineages have been maintained in flowering plants, as large redundancy is often observed. However, it seems that different MADS-domain protein complexes are recruited to specify carpel and ovule identity, which are probably composed of the same SEP factors but contain different AGAMOUS subfamily proteins. AGL11 lineage proteins are preferentially recruited for ovule development and, at least in core eudicots, for seed coat development. AGL11 lineage genes have so far not been identified in Ranunculaceae, and the only whole-genome sequence available from this family, from A. coerulea, apparently lacks AGL11 lineage members. Ovule development in several genera of Ranunculaceae has been studied in detail by Wang & Ren (2008). It will be interesting to compare ovule development between basal eudicots that have and do not have AGL11 lineage genes, as well as to express AGL11 lineage genes in A. coerulea. In this species, AG lineage members probably compensate for the lack of AGL11 relatives, and interestingly AG lineage orthologous genes showing ovule-specific expression have been reported in the genera Thalictrum and Aquilegia (Di Stilio et al., 2005).

In core eudicots, the AG lineage is further divided into two branches, the euAG and PLE lineages, which show in different species different levels of redundancy and subfunctionalization in regulating the C function. However, PLE-like genes regulate specific functions in fruit ripening and/or lignification and dehiscence. The lack of PLE lineage representatives in the Asteraceae expressed sequence tag (EST) and cDNA databases suggests that this lineage has been lost in this large eudicot family, although this needs further confirmation. The absence of PLE lineage genes might be explained by the observation that daisy family species develop a dry indehiscent achene-like fruit, where the unique functions of PLE-like genes are probably not needed. However, we also observed a loss of the PLE lineage in distant species such as poplar and M. guttatus, both producing a dry dehiscent capsule.

Despite the fact that AGAMOUS genes from many plant species have been studied extensively, it still remains an open field with many questions, especially regarding their functions after fertilization and how they regulate the various processes via their target genes. This information is of interest because the functions of AGAMOUS subfamily members in flower and fruit development seem to be widely conserved across flowering plants. This conservation can probably even be extended to some gymnosperm species where they have also been recruited for the formation of fruit-like structures.

Surprisingly, recently it was shown that in A. thaliana STK, SHP1 and SHP2 are also involved in the formation of lateral roots (Moreno-Risueno et al., 2010), which is a nice reminder for those scientists working with the aerial parts of the plant to look sometimes also into the soil and of course vice versa!


We thank Brendan Davies and Michiel Vandenbussche for the pictures of wild-type and mutant snapdragon and petunia flowers, respectively. We are also grateful to Chiara Mizzotti, Irma Roig Villanova and Eva Zanchetti for the pictures of A. thaliana wild-type and stk mutant ovule primordia and seeds. Many of the new sequence data presented in this review were produced by the US Department of Energy Joint Genome Institute (see Table S1). We apologize to the authors of the many valuable works that we could not discuss in this review because of the vastness of the subject and insufficient space.