|Home | About | Journals | Submit | Contact Us | Français|
A huge variety of plant forms can be found in nature. This is particularly noticeable for inflorescences, the region of the plant that contains the flowers. The architecture of the inflorescence depends on its branching pattern and on the relative position where flowers are formed. In model species such as Arabidopsis thaliana or Antirrhinum majus the key genes that regulate the initiation of flowers have been studied in detail and much is known about how they work. Studies being carried out in other species of higher plants indicate that the homologues of these genes are also key regulators of the development of their reproductive structures. Further, changes in these gene expression patterns and/or function play a crucial role in the generation of different plant architectures.
In this review we aim to present a summarized view on what is known about floral initiation genes in different plants, particularly dicotyledonous species, and aim to emphasize their contribution to plant architecture.
A striking feature of plants is the huge variety of forms that can be found in nature. This enormous diversity is due to variation in the shape and size of different plant organs, basically leaves, shoots and flowers (later fruits), and in the proportion of the different kinds of organs and the position where they appear in the plant. The number and arrangement of plant organs are the basis of plant architecture.
Flowers tend to appear clustered in a region of the plant called the inflorescence (Weberling, 1989a). Inflorescence form varies enormously among different species and seems to play a determinant role in reproductive success as it has a strong effect on pollination and fruit set (Wyatt, 1982). Whilst particular forms of inflorescences frequently typify some plant families, the same type of inflorescence architecture can also be found in unrelated families, suggesting that adaptive selection has probably played a role in the evolution of inflorescences (Tucker and Grimes, 1999)
All the aerial organs of the plant derive from the shoot apical meristem (SAM). This meristem generates leaves and shoots during the vegetative phase, and in the reproductive phase – after the floral transition – it becomes an inflorescence meristem and flowers are produced. The architecture of the inflorescence depends on its branching pattern and the position of the flowers: on when and where flowers are formed.
Inflorescence types have been classified following several criteria (Weberling, 1989a). A main parameter for the classification is whether the shoot apices end in terminal flowers or not. When they do not terminate, the inflorescences are classified as indeterminate. A typical example of an indeterminate inflorescence is the raceme, present in species such as Arabidopsis thaliana or Antirrhinum majus. In this type of inflorescence, the apical meristem is able to grow indefinitely, generating a continuous main axis that laterally produces floral meristems (Fig. 1A–C). On the other hand, inflorescences that form terminal flowers are called determinate. A classical type of determinate inflorescence is the cyme. Cymose inflorescences lack a main axis: the main shoot terminates in a flower, while growth continues through lateral axes produced below the terminal flower (Fig. 1D–F). These lateral axes again form terminal flowers and this process is reiterated several times. Data on the developmental control of cymose inflorescences is available for several species such as Silene latifolia or tobacco (Nicotiana tabacum; Fig. 1D, E). A variation of the cymose pattern is found, for example, in tomato (Solanum lycopersicum; Fig. 1F); the inflorescence of this species is also a cyme but, in this case, after the main axis generates the terminal inflorescence, a new axis of growth develops from an axillary meristem that produces a certain number of leaves before again terminating in an inflorescence. This process repeats indefinitely, generating a plant with an apparently continuous growing axis in which the production of leaves and ‘lateral inflorescences’ alternates. This kind of plant architecture is called a sympodium. Finally, as pointed out in an elegant modelling analysis of inflorescence development (Prusinkiewicz et al., 2007), a third main kind of inflorescence architecture, also determinate, is the panicle (Fig. 1G). In contrast to the cyme, in this type of inflorescence a clear main shoot axis exists but this is terminated by a flower, as also occurs in the series of lateral branches produced by the main shoot.
Inflorescences are also classified according to the complexity of their branching. Those inflorescences where flowers are directly formed from the main axis are called simple inflorescences, while compound inflorescences are those where flowers are formed from secondary or higher-order branches. An example of a compound inflorescence is the compound or double raceme present in many Leguminosae species, such as pea (Pisum sativum), Medicago truncatula or Lotus japonicus (Fig. 1C). The inflorescences of arabidopsis and antirrhinum are simple racemes (Fig. 1A, B).
Although the evolution of inflorescences is poorly understood, it is generally accepted that the most primitive inflorescences would have had terminal flowers. This, in part, derives from the idea that the flower is a specialized shoot, and the transition of a vegetative apex to a flower would be direct in a primitive inflorescence. As discussed by Tucker and Grimes (1999), the first authors speculating about inflorescence evolution favoured the idea that a solitary terminal flower would be the ancestral inflorescence form (Parkin, 1914); this supported the idea of woody trees, such as those of Magnoliaceae, being among the most primitive families. However, the primitiveness of the Magnolia type of flower has been challenged by several authors, such as Stebbins (1974), based on questions such as the high complexity of its vasculature, and a more recent view is that the ancestral angiosperms would have had simple cymose inflorescences.
As explained above, the architecture of the inflorescence depends on which meristems give rise to shoots and which to flowers (Coen and Nugent, 1994). The genetic control of the specification of floral meristems has largely been studied in model species, mainly in antirrhinum and arabidopsis, and the main factors have been identified and a lot of information about how they work is available.
In recent years, the homologues of these and other genes with related functions have been identified and studied in many other plant species. These studies suggest that the functioning of the genetic network controlling the initiation of flowers is largely conserved among flowering plants, with key differences often relating to the different inflorescence architecture of each species. In this review we aim to present a summarized view on what is known about floral initiation genes in different species, and we try to emphasize their role in directing plant architecture.
As for many genetic processes in plants, the genetic control of floral initiation is best known in the model plant arabidopsis. However, the aim of this article is not to describe in detail how the specification of floral meristems is controlled in arabidopsis, a question that has been treated in several excellent reviews (Jack, 2004; Vijayraghavan et al., 2005; Blázquez et al., 2006), but to try to describe and compare what is known about the genes controlling this process in other species. Therefore, we will briefly introduce the key elements of the genetic network in arabidopsis as a basis for the comparison.
In arabidopsis, during the vegetative phase the SAM produces on its flanks vegetative primordia that will form leaves with shoot meristems in their axils. Upon transition to the reproductive phase, the SAM becomes an inflorescence meristem (IM) and the new lateral primordia produced after that point develop as floral meristems (FM). Therefore, with the floral transition the fate of these lateral primordia has to be reprogrammed so that they acquire the identity of floral meristems.
In arabidopsis, the acquisition of floral meristem identity (FMI) by these primordia is controlled by the interaction of positive and negative regulators. Although several other genes have also been shown to play important roles in the regulation of floral meristem identity in arabidopsis, we will concentrate on LEAFY (LFY), APETALA1 (AP1) and TERMINAL FLOWER1 (TFL1). These genes seem to form the backbone of the network and, consequently, they are the ones whose role in the process has been best analysed in arabidopsis and whose homologues have been studied most in many other species.
The LFY gene is required for the specification of FMI in arabidopsis. This is clearly deduced from the phenotype of lfy mutant plants, where the flowers are replaced by structures with shoot characteristics (Fig. 2A; Schultz and Haughn, 1991; Huala and Sussex, 1992; Weigel et al., 1992). The shoot character of the lfy ‘flowers’ is more marked in the first positions in the inflorescence, while structures formed in more apical positions progressively acquire an increasing degree of floral identity due to independent activation of other floral meristem identity genes such as AP1 (Huala and Sussex, 1992; Bowman et al., 1993). Another aspect of the lfy phenotype is that while in wild-type the flowers are bractless (no subtending leaf; Fig. 2A) many of the lfy transformed ‘flowers’ have bracts, indicating an additional role for LFY in bract suppression during the inflorescence phase (Schultz and Haughn, 1991).
LFY encodes a transcription factor that so far has been only found in the plant kingdom (Maizel et al., 2005). In contrast to most other types of transcription factors, LFY does not belong to a multigene family. Arabidopsis and most angiosperms contain only one LFY gene. Consistent with the phenotype of the mutant, LFY is strongly expressed throughout the young floral meristems from the earliest stages of development (Fig. 2B; Weigel et al., 1992). In fact, upregulation of LFY in these meristems is crucial for them to acquire floral identity, as it activates the expression of AP1 and the floral meristem identity genes (Parcy et al., 1998).
LFY expression is not absolutely confined to floral tissues. Expression can also be detected at low levels in leaf primordia during the vegetative phase, and gradually increases until the floral transition (Blázquez et al., 1997; Hempel et al., 1997). The actual level of LFY expression in the apex is considered to be a critical parameter that determines the time point at which the floral transition takes place (Blázquez et al., 1997). LFY seems to act as an integrator of the pathways controlling flowering time and the initiation of floral meristems (Blázquez and Weigel, 2000; Parcy, 2005). In fact, lfy mutants are slightly delayed in the vegetative-to-inflorescence transition (Blázquez et al., 1997).
In agreement with its proposed roles in floral initiation, constitutive expression of LFY in arabidopsis causes early flowering and the transformation of all shoots into flowers, indicating that LFY is not only necessary, but also sufficient to confer floral identity to emerging shoot meristems (Weigel and Nilsson, 1995).
AP1 is the other main promoter of floral meristem identity. The ap1 mutants show defects in FMI and defects in the identity of the floral organs of whorls 1 and 2. The flowers of ap1 mutants do not have petals and produce bract-like organs instead of sepals. In the axils of those first-whorl organs, new floral meristems are produced that reiterate this pattern, generating ‘branched flowers’ (Fig. 2A; Irish and Sussex, 1990; Bowman et al., 1993).
AP1 also encodes a transcription factor but, in contrast to LFY, it belongs to a large multigene family, the MADS-box gene family (Mandel et al., 1992). Similarly to LFY, it is expressed throughout young floral meristems, shortly after the onset of LFY expression in these meristems (Fig 2; Mandel et al., 1992). In fact, AP1 (as well as CAL; see below) is directly activated by LFY (Wagner et al., 1999). The phenotype of plants constitutively expressing AP1 is also consistent with its role in floral meristem identity: 35S::AP1 plants are early flowering and show shoot-to-flower conversions, a phenotype similar to that of tfl1 mutants and 35S::LFY transgenics (Mandel and Yanofsky, 1995).
CAULIFLOWER (CAL), another MADS-box gene highly related in its sequence to AP1 and with a similar expression pattern, is partially redundant to AP1 in FMI specification in arabidopsis. Single cal mutants show a wild-type phenotype, but simultaneous loss of AP1 and CAL causes a complete transformation of floral meristems into inflorescence-like meristems, which give rise to new inflorescence-like meristems; this pattern reiterates an indefinite number of times to form structures similar to cauliflower heads (Bowman et al., 1993; Mandel and Yanofsky, 1995). As expected, constitutive expression of CAL causes a similar, though weaker, phenotype to that of 35S::AP1 plants (Savidge, 1996; Liljegren et al., 1999). The redundancy of AP1/CAL in specifying floral meristem identity has only been documented in arabidopsis and species from the Brassicacea family. This is consistent with the results of phylogenetic studies showing that AP1 and CAL derive from a recent duplication event, found only within the Brassicaea (Lawton-Rauh et al., 1999; Lowman and Purugganan, 1999).
The role played by TFL1 in floral initiation is opposite to that of LFY and AP1. In tfl1 mutants the shoot meristems are converted into floral meristems: cauline leaves subtend solitary flowers, rather than shoots, and inflorescence shoots are converted into terminal flowers (Fig. 2A; Shannon and Meeks-Wagner, 1991; Alvarez et al., 1992; Schultz and Haughn, 1993). Therefore, while LFY and AP1 specify floral meristem identity, TFL1 would specify shoot identity. Mutations in TFL1 also cause early flowering, indicating that TFL1 also acts as a repressor of flowering (Shannon and Meeks-Wagner, 1991; Schultz and Haughn, 1993).
Constitutive expression of TFL1 driven by the 35S promoter causes a great extension of all developmental phases (Ratcliffe et al., 1998). The 35S::TFL1 plants produce an enlarged vegetative rosette with a high number of leaves and a long inflorescence stem, with many lateral branches, which eventually forms normal flowers. The phenotype of 35S::TFL1 plants led to the proposal that TFL1 acts by retarding the phase transitions at the shoot apex. According to this view, the production of axillary and terminal flowers in tfl1 would be the consequence of the mutant shoot meristems progressing through the phases much faster than the wild type. In this situation, these meristems would make the transition from inflorescence to floral, a phase that would not be reached by the wild-type shoot meristems under normal conditions.
TFL1 is strongly expressed in the centre of the main and lateral shoot inflorescence meristems, not in the floral meristems. This expression pattern is complementary to that of LFY and AP1, which are present in floral but not in inflorescence meristems (Fig 2B). Action of TFL1 in the inflorescence apex is pivotal to its function, as a main role of TFL1 is to prevent these meristems from assuming the floral identity by inhibiting the expression of FMI genes. Thus, in tfl1 mutants LFY and AP1 expression invades the inflorescence meristems, which are then converted into flowers (Weigel et al., 1992; Bradley et al., 1997).
Conversely, several pieces of evidence suggest that LFY and AP1 prevent TFL1 expression in floral meristems (Liljegren et al., 1999; Ratcliffe et al., 1999; Ferrándiz et al., 2000), although it is not clear whether LFY or AP1 act as direct repressors of TFL1 (Parcy et al., 2002). Correlating with its function in repressing flowering, TFL1 is also expressed, although at a lower level, in the shoot vegetative meristem. Upregulation of TFL1 expression in the shoot apical meristem temporally coincides with commitment to flowering, representing a clear early marker for the floral transition (Bradley et al., 1997).
In contrast to LFY and AP1, TFL1 does not encode a transcription factor. TFL1 is homologous to phosphatidylethanolamine binding proteins (PEBPs; Bradley et al., 1997; Ohshima et al., 1997), a wide group of proteins also found in animals, yeast and bacteria, that play diverse roles related to signalling pathways controlling growth and differentiation (Yeung et al., 1999; Hengst et al., 2001; Chautard et al., 2004). TFL1 belongs to a small gene family (Mimida et al., 2001), one of whose members, FLOWERING LOCUS T (FT), is also a regulator of flowering time. Opposite to TFL1, mutations in FT cause late flowering and 35S::FT plants show a phenotype similar to that of tfl1 mutants (Kardailsky et al., 1999; Kobayashi et al., 1999). The mechanism of action of TFL1 has not been elucidated yet, but recent studies indicate that its homologue FT promotes flowering by acting at the nucleus, as part of a complex with the bZIP transcription factor FD (Abe et al., 2005; Wigge et al., 2005). TFL1 also can bind to bZIP factors. The structure of the TFL1 and FT proteins were recently resolved and are very similar (Ahn et al., 2006). This is highlighted by swapping discrete domains among these proteins, as TFL1 can be converted into FT and vice versa, suggesting that the biochemical function of both proteins is very similar and that differences in their functions could be due to differential binding to interactors (Hanzawa et al., 2005; Ahn et al., 2006).
Although LFY, AP1 and TFL1 are considered to be major regulators of floral initiation, the picture is, of course, not quite so simple and several other genes have also been shown to play important roles in the control of this process in arabidopsis. Among them are, for example, the MADS-box gene FRUITFULL (FUL), highly related in sequence to AP1, which is required for the initiation of the flowers that are eventually formed by the proliferating inflorescence meristems of the double ap1 cal mutant (Gu et al., 1998; Ferrándiz et al., 2000), and AGL24, which has been implicated in the upregulation of LFY expression (Yu et al., 2002). Other examples are the genes APETALA2 (AP2) and UNUSUAL FLORAL ORGANS (UFO) whose mutations enhance the meristem defects of ap1 or lfy mutants, respectively (Ingram et al., 1995; Okamuro et al., 1997).
Comparative studies carried out on FMI genes in other species, however, have been mostly focused on homologues of LFY, AP1 and TFL1. In the following sections we will try to summarize what is known about the homologues of these and other related genes in different species. We will emphasize what changes in function and/or expression have occurred and the possible effects of these changes in the generation of different plant architectures. Although occasionally monocotyledonous species will also be mentioned, for simplicity we will focus on eudicot species. Excellent reviews on the genetics of monocotyledonous inflorescence development have recently been published elsewhere (Bommert et al., 2005; Kellogg, 2007).
LFY and AP1 are the main activators of the cascade of genes initiating floral development. In the last decade, important efforts have been made in order to understand the function and evolution of both factors. LFY is present in all land plants analysed, which have evolved for at least 400 million years. There is no doubting its key role in flower meristem identity acquisition in angiosperms. However, the ancestral function of LFY and its evolution is far from being clear. LFY homologues have been isolated from distant species, such as the moss Physcomitrella patens and different species of ferns and gymnosperms. The LFY proteins have low rates of amino acid substitutions and have been used in the phylogenetic analysis of seed-plant relationships (Frolich and Parker, 2000). In a different study, Maizel et al. (2005) investigated the functionality of different LFY homologues, representing the different taxa from the mosses to angiosperms, by testing their ability to complement the arabidopsis lfy mutant. The degree of complementation of the lfy mutant phenotype correlated with the taxonomic distance from arabidopsis. PpLFY (from Physcomitrella patens) was unable to complement the lfy mutant, while the homologues of ferns and gymnosperms partially complemented the mutation, and the angiosperm homologues fully complemented it. The authors also studied the ability to activate known LFY targets by transcriptional profiling. A major conclusion of these analyses was that the ability of LFY homologues to activate AP1 is restricted to flowering plants.
The results of these experiments agree with a progressive functional divergence of LFY from moss to angiosperms. For example, the moss P. patens contains two LFY homologues, PpLFY1/2, and these are expressed in the main and lateral apices, in the developing archegonium, but not in the antheridium (Tanahashi et al., 2005). The disruption of both PpFLY genes affects the first zygotic division, suggesting an important role of PpLFY in this process, a function that widely diverges from that described in angiosperm species. Gymnosperm species also have two LFY homologues, and both are involved in the development of reproductive tissues. In Pinus radiata, for example, one of the LFY homologues, NEEDLY, is expressed at high levels in female reproductive meristems (Mouradov et al., 1998), while the expression of the second homologue, PRFLL, is detected in buds and male cones (Mellerowicz et al., 1998). The presence of these two paralogous genes, with expression in different reproductive tissues, together with the analysis of LFY homologues from different taxa has led to the proposal of the ‘mostly male’ theory for the origin of the flower (Frolich and Parker, 2000). This theory proposes that a duplication occurred before the separation of flowering plants that gave rise to the LFY and NEEDLY clades. Angiosperm species lost the NEEDLY gene and the theory proposes that LFY would have been recruited to specify female reproductive organs in addition to male reproductive tissues. In this way, the flower would have arisen by the development of ectopic female structures in a LFY-expressing male reproductive shoot.
LFY could have evolved from a different, broader, function in more distant species before recruitment in flowering plants for the acquisition of the floral fate. Expression of LFY in tissues other than reproductive meristems such as leaves, tendrils or vegetative meristems (as will be discussed below) could be a remnant of this broader function that has been retained in certain cases.
In contrast with the presence of LFY orthologues in all land plants, AP1 orthologues have only been found in angiosperm species. Arabidopsis has two additional genes, FUL and CAL, which have high sequence homology with AP1 and share functions in floral meristem identity specification. AP1 and FUL belong to different gene clades, which were generated as the result of a duplication event at the base of the core eudicots (Litt and Irish, 2003). The fact that both genes belong to the MADS-box family of transcription factors and that they present high sequence homology frequently makes it difficult to clearly ascribe the homologues isolated from other species to one of the two clades. The study of phylogenetic relationships between AP1 and FUL homologues from a variety of angiosperm species has led to the identification of specific C-terminal motifs characteristic of each clade. It has been suggested that this duplication together with the appearance of a new C-terminal motif in the AP1 clade contributed to fix the floral structure observed in core eudicots (Litt and Irish, 2003).
LFY and AP1 are key regulators of flower and inflorescence development. For that reason, many groups have become interested in the comparative study of their function in different species. Such studies are helping us to understand how diversity in plant architecture has been generated. Below, we discuss significant examples of homologues that have been studied in some detail in different dicot species. Relevant data from these species, and from others that we have been not able to describe due to space limitations, are summarized in Table 1.
Antirrhinum majus is a euasterid from the order Lamiales (Fig. 3) and, with arabidopsis, was a key model species for the initial studies on the genetic control of inflorescence and flower development. In fact, the first member of the LFY gene family to be isolated and characterised was FLORICAULA (FLO) from antirrhinum (Coen et al., 1990) and the isolation of AP1 and its antirrrhinum homologue SQUAMOSA (SQUA) was reported almost simultaneously (Huijser et al., 1992; Mandel et al., 1992). The architecture of both species is very similar, both having inflorescences that are simple racemes. However, in antirrhinum all the stem internodes elongate during the vegetative phase, whilst in arabidopsis these internodes remain compressed, forming a rosette. In addition, in antirrhinum all the flowers are subtended by a bract, while the arabidopsis flowers are bractless (Fig. 1).
Correlating with similar inflorescence architectures, FLO and SQUA seem to work in a very similar way in antirrhinum as do their arabidopsis counterparts. FLO and SQUA essentially exhibit the same expression patterns as LFY and AP1, respectively. In addition, the phenotype of the flo and squa mutants also indicates that the functions of the anthirrhinum genes are similar to their arabidopsis homologues (Coen et al., 1990; Huijser et al., 1992). As in lfy mutants, flo mutants exhibit conversion of flowers into inflorescences, confirming the role of FLO in floral meristem identity. One notable difference is that LFY is also involved in bract suppression while FLO does not have this function.
Mutations in SQUA also cause conversion of flowers into shoots. However, while in the ap1 mutant these shoots consist of branched flowers, lacking petals but bearing normal stamens and carpels, the FMs of the squa mutant are replaced by vegetative shoots that only rarely produce flowers (Huijser et al., 1992). The weaker inflorescence phenotype of the ap1 mutant in comparison to squa can probably be explained by the redundant activity of the AP1 paralogue CAL, a gene possibly only present in Brassicaceae (Lawton-Rauh et al., 1999; Lowman and Purugganan, 1999).
Many species from the large Fabaceae family, also known as Leguminosae (Fig. 3), have compound double racemes (Weberling, 1989a, b). After floral transition, the SAM of these legume species becomes a primary inflorescence meristem (I1) that rather than producing flowers, generates second-order inflorescence meristems (I2) that produce the flowers. These I2 usually produce a certain number of flowers, depending on the species, before they are consumed in forming a rudimentary stub (Fig. 1C; Singer et al., 1999). This generates a compound raceme architecture where the main axis, rather than subtending individual flowers, subtends small racemes. The lateral secondary inflorescences of these legume species share morphological features with the simple racemes of arabidopsis or antirrhinum, in the sense that both consist of a main axis that laterally produces flowers (from 1–2 flowers in the case of most pea cultivars, to many more as, for example, in some Trifolium species) and do not differentiate into a terminal flower. As we will see, the analysis of the legume LFY and AP1 homologues confirms that they work as functionally equivalent structures.
Homologues of LFY and AP1 have been isolated and characterized from several model legume species (Hecht et al., 2005; Domoney et al., 2006) and a general conclusion is that, in spite of the differences between the inflorescence of legumes and arabidopsis, these genes play similar functions in the legume lateral secondary inflorescences as do LFY and AP1 in the arabidopsis inflorescence. The pea LFY homologue, UNIFOLIATA (UNI), although with a wider expression pattern than its arabidopsis homologue, is expressed in floral meristems and its mutations cause flower-to-inflorescence conversions (Fig. 4; Hofer et al., 1997). A similar expression pattern and mutant phenotype have also been described for the Lotus japonicus LFY homologue, LjLFY (Dong et al., 2005). On the other hand, the pea AP1 homologue, PROLIFERATING INFLORESCENCE MERISTEM (PIM, also known as PEAM4; Berbel et al, 2001; Taylor et al., 2002), is expressed in floral meristems with a pattern essentially identical to that of AP1 and SQUA. Mutations in PIM also cause flower-to-shoot conversions and its functional homology with AP1 is also supported by the phenotypes of PIM over-expression in arabidopsis. As in the case of the antirrhinum squa mutant, the mutations in PIM, as well as in MtPIM, the M. truncatula homologue, cause a phenotype more severe than that of the arabidopsis ap1 mutant. SEM analysis has shown true conversion of the mtpim floral meristems into proliferating secondary inflorescence meristems, which generate structures that resemble the proliferating inflorescence meristems produced by the double ap1 cal mutants (Benlloch et al., 2006).
A particular feature of the uni mutant is that it is also affected in its leaf morphology (Hofer et al, 1997). While the pea wild-type leaves are compound odd-pinnate, with a rachis supporting several pairs of leaflets, the leaves of the uni mutant are much simpler, having a shorter petiole bearing only one-to-three leaflets. Accordingly, UNI is also expressed in developing leaves. It has been suggested that UNI could have a function controlling the indeterminacy during leaf or flower development, reminiscent of an ancestral broader function of LFY genes in meristem control. The expression of UNI during pea leaf development would temporarily inhibit leaf determination, allowing the development of a complex leaf (Hofer and Ellis, 2002). The role of LFY homologues in leaf complexity also extends to other legume species, as mutations in LjLFY transform the Lotus trifoliate leaf into unifoliate (Dong et al., 2005).
Most species from this family of rosids (Fig. 3) have a simple raceme type of inflorescence, similar to arabidopsis. However, a few species, such as Ionopsidium acaule (violet cress), Idahoa scapigera and Laevenworthia crassa, exhibit ‘rosette flowering’. In such plants the main stem does not elongate and flowers are produced on long pedicels that emerge from the axils of rosette leaves, positions that in arabidopsis would produce coflorescences. Evidence suggests that changes in the LFY homologues of these plants might have played an important role in the evolution of this different inflorescence architecture (Yoon and Baum, 2004).
These LFY homologues show expression patterns that differ to that of arabidopsis LFY. The LFY homologue of violet cress is strongly expressed in its SAM (Shu et al., 2000) and the promoter of the L. crassa LFY directs expression to the SAM in transgenic arabidopsis plants (Yoon and Baum, 2004). Moreover, lfy mutant plants transformed with a genomic construct of LcrLFY exhibit morphological features that are reminiscent of rosette flowering. Finally, expression induced by AtLFY and LcrLFY has been compared by microarray analysis (Sliwinski et al., 2006). Analysis of genes up- or down-regulated, showed that the TFL1 gene was over-expressed in plants containing an LcrLFY transgene in comparison with those carrying an arabidopsis LFY transgene. Therefore, the nature of the interaction between LFY and TFL1 could have changed between these species, generating differences in the architecture of the inflorescence. This study suggests that changes in cis-regulatory elements, leading to ectopic expression in axillary meristems and also in the protein coding region of LcrLFY, could be at the origin of rosette flowering (Sliwinski et al., 2006). The idea that morphological evolution involves changes in the regulation of developmental genes has already been suggested (Doebley and Lukens, 1998). This example and others discussed in this review place LFY, AP1 and TFL1 as master developmental genes whose changes in expression pattern could play a key role in determining plant and inflorescence architecture.
Another dicot family where LFY homologues have been studied in detail is the Solanaceae (Fig. 3). Many species from this family have determinate cymose inflorescences, including tobacco (Fig. 1E). Tomato and Petunia hybrida plants, in addition to cymose inflorescences, also exhibit a sympodial growth habit (Fig. 1F). Again, the behaviour of the LFY homologues in these plants show particular features that relate to the architecture of their inflorescences.
The function of the LFY homologues in tomato and petunia in FMI seems to be similar to that in arabidopsis. Mutations in FALSIFLORA (FA) in tomato, or in ABERRANT LEAF AND FLOWER (ALF) in petunia, cause the conversion of the floral meristem into an inflorescence meristem similar to lfy (Souer et al., 1998; Molinero-Rosales et al., 1999). However, the expression of FA and ALF significantly differ from that of LFY in arabidopsis. In addition to being expressed in floral meristems, in petunia ALF is expressed in the inflorescence meristem and in tomato FA is also expressed in the sympodial meristem. The expression of LFY homologues in the shoot meristems of these plants is very likely related to the formation of terminal flowers by their inflorescences. In both species LFY transcripts can be also found, as in pea, in leaf primordia. In agreement with this, tomato fa mutants show certain reductions in leaf complexity; however, no leaf phenotype can be observed in petunia alf mutants.
Tobacco contains two LFY homologues, NFL1 and 2, possibly due to the allotetraploid origin of this species (Kelly et al., 1995). Expression of NFL genes is similar to that of their petunia and tomato homologues, with their transcripts also being found in vegetative and axillary meristems. However, the function of the tobacco homologues could be different. NFL1 over-expression in tobacco promotes terminal flower formation and inhibits inflorescence branching but does not cause early flowering, such as 35S::LFY causes in tobacco. On the other hand, the co-suppression of NFL genes produces unregulated initiation of lateral meristems. All these data suggest that NFL1 has an additional role compared to LFY in the allocation and placement of meristematic cells (Ahearn et al., 2001).
This species (from the order Caryophyllales) also has a determinate inflorescence, of a type named a dichasium, where the apical meristem forms a terminal flower and two inflorescences meristems are formed on its flanks (Fig. 1D). In S. latifolia, two AP1 homologues, SLM4 and SLM5, have been characterized (Hardenack et al., 1994). Only SLM4 is a likely AP1 orthologue, while SLM5 is closer to FUL (Litt and Irish, 2003). SLM4 is expressed in flowers with a similar pattern to that of AP1 or SQUA but, as with the LFY homologues in the Solanaceae, the S. latifolia AP1 homologue is also expressed in the inflorescence meristem. Again, this is another example where the generation of a determinate inflorescence seems to be related to expression of a FMI gene in the shoot apical meristem.
In this species, indeterminate and determinate inflorescence varieties are found. For determinate varieties, the first nodes produce leaves subtending axillary vegetative shoots, the next nodes produce axillary inflorescences also subtended by leaves, and the last nodes before the terminal flower produce flowers with subtending bracts. Indeterminate varieties continue producing axillary inflorescences indefinitely (Pouteau et al., 1998; Ordidge et al., 2005). Flowering in I. balsamina is dependent on short-day (SD) conditions and the plants remain vegetative under long days (LD). A specific feature of I. balsamina is that the terminal flower of determinate varieties reverts to vegetative growth after transfer to LD non-inductive conditions, a phenomenon that is termed floral reversion.
The expression of LFY and AP1 homologues, IbLFY and IMP-SQUA, have been analysed in the determinate variety (Pouteau et al., 1997). As occurs in other species such as arabidopsis, IMP-SQUA expression is only expressed after floral induction. However, the situation with IbLFY is somewhat unusual, as it is expressed in the vegetative and in the flowering terminal meristem of I. balsamina. This is in contrast to arabidopsis or antirrhinum but similar to what occurs in tobacco, and could be related to the determination of the shoot meristem into a terminal flower. However, expression of IbLFY is also detected in the shoot meristem after floral reversion and in the meristem of non-induced LD-growing plants, suggesting that IbLFY expression in the meristem is not sufficient to specify floral identity. Nevertheless, the phenotype of IbLFY over-expression in arabidopsis indicates functional homology between IbLFY and LFY (Ordidge et al., 2005). Therefore, a possibility is that the interaction of LFY with other regulators of FMI could be different in I. balsamina (see IbTFL1, below).
Woody species have two important developmental characteristics that make them different from annual herbaceous plants. First, they have long juvenile phases (from several years to decades) during which they produce only vegetative organs. Second, the flowering process often extends to two consecutive seasons – during the first season buds are formed that during the second season will develop and produce flowers or inflorescences. Genetic analysis in perennials is a complicated task and, consequently, our understanding of the function of the LFY and AP1 homologues from this type of plants is not as precise as in herbaceous species. Nevertheless, homologues have been analysed in several perennial species and the available data indicate that these genes affect both these characteristics of woody plant development.
As expected, expression of LFY and AP1 homologues in perennials is also associated with floral and inflorescence buds (see Table 1). Expression of these genes appears to follow a bimodal pattern related to the two seasons that are needed to flower. This has been studied in detail in the case of grapevine (Vitis vinifera). During the first season, the SAM produces lateral meristems that will generate inflorescence meristems. These inflorescence meristems form inflorescence branch meristems before the buds enter dormancy. In the second growing season, these buds form additional inflorescence branch meristems before dividing into 3–4 floral meristems. VFL, the LFY homologue, is expressed in lateral meristems independent of their fate, although VFL expression increases in young floral meristems. The expression level of VFL reaches two peaks, one at the time of flowering induction during the first growing season, and a second peak at the time of bud reactivation and flower initiation during the second growing season (Carmona et al., 2002). The grapevine AP1 homologue, VAP1, is expressed in early stages of inflorescence development during the first season, and later on in inflorescence branch meristems. During the second season, VAP1 expression is detected in floral meristems and it is maintained during flower development. Expression patterns of the grapevine LFY/AP1 homologues suggest that both genes are also involved in other processes in addition to flower development. Thus, VFL is also expressed in leaf primordia and in the growing margins of developing leaves, where it has been suggested that it maintains the cell proliferation needed for the typical palmate morphology of the grapevine leaves (Carmona et al., 2002). On the other hand, VAP1 seems to be involved in tendril development as it is expressed during the development of these organs, independent of the flowering process, even in very young plants that have not undergone the floral transition (Calonje et al., 2004).
Another example of expression of FMI genes associated with the two growing seasons in perennials is that of BpMADS3, the likely AP1 orthologue of birch (Betula pendula; Elo et al., 2001). BpMADS3 also exhibits a bimodal expression pattern during inflorescence development. Birch has separate male and female inflorescences and BpMADS3 shows different expression patterns in each of them, according to their different timing of development. A peculiarity of the B. pendula AP1 homologue is that expression also continues at a high level during late flower development and even during seed development.
TFL1 has an opposite function to LFY and AP1 and belongs to the group of PEBP proteins. PEBP genes have been found in many angiosperm species, dicots and monocots, and constitute gene families whose number varies in different species – from six members in arabidopsis or tomato to 19 in rice. Plant PEBP proteins can be grouped into three main clades: the MFT-, FT- and TFL1-like subfamilies (Mimida et al., 2001; Carmel-Goren et al., 2003; Chardon and Damerval, 2005). Those TFL1-like genes for which a function has been found have roles in the control of plant development, usually in flowering. As we will see in the examples that follow, many TFL1-like genes are key controllers of flowering time and inflorescence architecture.
As was the case with LFY and FLO, CENTRORADIALIS (CEN) from Antirrhinum majus was the first member of the plant PEBP gene family that was characterized (Bradley et al., 1996); thereafter TFL1 was isolated as an arabidopsis CEN homologue. In agreement with the similarities between arabidopsis and antirrhinum inflorescences, mutations in the antirrhinum homologue also cause the conversion of the SAM into a terminal flower, changing the inflorescence from indeterminate to determinate. As with TFL1 in arabidopsis, CEN is expressed in the subapical region of the shoot meristem, somehow inhibiting the expression of the LFY homologue in this meristem. However, while TFL1 is expressed both in vegetative and inflorescence shoot meristems of arabidopsis, CEN is only expressed in the inflorescence meristem. The absence of CEN expression in the apex before floral transition has been used to explain the fact that, in contrast to tfl1, cen mutations do not affect flowering time (Bradley et al., 1996; Cremer et al., 2001). Nevertheless, although the expression patterns could explain the different mutant phenotypes, differences in the function of the two proteins can not be discarded. In fact, while 35S::CEN causes an extreme delay of flowering in tobacco, 35S::TFL1 did not show any effect in this species (Amaya et al., 1999).
Interestingly, the most likely arabidopsis orthologue of CEN is not TFL1 but the ATC gene (Arabidopsis thaliana CEN homologue; Mimida et al., 2001), another member of the TFL1 clade. However, while ATC can functionally substitute for TFL1, as suggested by its ability to complement the tfl1 mutant phenotype when constitutively expressed, ATC loss-of-function mutations do not cause any obvious phenotype, indicating that ATC could be involved in a function different to IM identity. ATC expression is very low and restricted to the tissues around the vasculature of the hypocotyl. The striking differences between the expression patterns of ATC and TFL1 have been considered to be the basis of their different functions (Mimida et al., 2001).
Among legume species, TFL1 homologues have been described for pea and Lotus japonicus. Pea contains at least three TFL1/CEN homologues. No function has been assigned to TFL1b, a likely orthologue of CEN. TFL1a and TFLc are most closely related to TFL1 and play TFL1-related functions. Mutations in PsTFL1a, also known as DETERMINATE (DET), cause the determination of the main apex without affecting flowering time, in a similar manner to what occurs in cen mutants of antirrhinum. On the other hand, mutations in PsTFL1c, also known as LATE FLOWERING (LF), cause early flowering without affecting determination (Foucher et al., 2003). Therefore, it seems that in pea the ‘two functions’ of the arabidopsis TFL1 gene, flowering time and apex determinacy, are controlled by two different genes. Whether the different functions of DET and LF are due to differences in their encoded proteins or to differential expression patterns is unclear. Expression patterns might at least partly explain the differences; thus, similar to CEN, DET is expressed in the shoot apex only after the floral transition, while LF expression is observed also in the vegetative apex. Expression of pea TFL1 homologues have only been analysed by RT-PCR, but in situ hybridization analysis has indicated that the L. japonicus DET homologue, LjCEN1, is expressed in the inflorescence meristems (Guo et al., 2006).
It is interesting to point out that the phenotype of the pea det mutant underlines the differences between the arabidopsis simple raceme and the pea compound raceme. The shoot apex of det mutants does not form a terminal flower, but instead it is transformed into an I2 and produces 1–2 flowers before terminating in a stub (Fig. 4); thus, in det mutants the I1 is transformed into an I2 (Singer et al., 1990). Conversely, pea plants mutated in either the VEGETATIVE1 (VEG1) or VEGETATIVE2 (VEG2) genes show very similar phenotypes, where I2s are replaced by I1s and thus are opposite to det (Reid and Murfet, 1984; Murfet, 1992; Singer et al., 1994; Reid et al., 1996). During the vegetative phase, veg1 mutant plants look identical to wild-type plants, and both produce several vegetative nodes, each with a leaf bearing an axillary vegetative meristem that stays dormant. In the wild-type plant, the meristems in the nodes produced after the floral transition develop as secondary inflorescences. In contrast, in the veg1 and veg2 mutants, the meristems of nodes in equivalent apical positions also start growing but, rather than producing secondary inflorescences, they generate indeterminate vegetative shoots identical to I1s. This generates plants with a unique non-flowering phenotype, never observed in any other species as the result of a single recessive mutation. We can explain the phenotypes of det and veg mutants by considering that VEG1/2 and DET genes control the identity of the primary and secondary inflorescence meristems. DET would specify I1 identity, whilst I2 identity would be specified by VEG1 and VEG2. Such phenotypes do not occur in arabidopsis as compound racemes have additional levels of regulation of meristem identity.
In summary, we can easily establish a parallelism between the repression network of TFL1-LFY/AP1, which maintains the indeterminacy of the shoot meristem, and the DET-VEG1/2 network, which prevents the conversion of the shoot apical meristem into an I2. The molecular characterization of VEG1 and VEG2 will represent a major advance in understanding the molecular mechanism underlying the development of compound inflorescences.
TFL1/CEN homologues have been studied from two Solanaceae, tobacco and tomato. From tobacco, seven CEN-like genes (CET genes) have been isolated (Amaya et al., 1999). Sequence analysis indicates that CET2/4, CET5/6 and CET1/7 are pairs of genes probably representing single copy genes in the diploid progenitors of this allotetraploid species. CET2/4 are the most closely related to CEN and, like TFL1, are expressed in axillary shoot meristems. However, expression of CET2/4 was not detected in the apex of the main shoot, which expresses NFL (the LFY orthologue) and forms a terminal flower (Ahearn et al., 2001). In fact, CET2/4 expression is restricted to vegetative axillary meristems and is not detected in those axillary meristems just below the terminal flower that will give rise to terminal cymes, and which express NFL. As the vegetative axillary meristems develop into flowering shoots, expression of CET2/4 decreases and NFL is upregulated, suggesting that the CET genes act to maintain the vegetative character of these meristems, delaying their transition to an inflorescence phase. These observations also suggest that the antagonism between TFL1 and FMI genes observed in arabidopsis also occurs in tobacco and again supports the view that in some species with determinate inflorescences the formation of terminal flowers depends on the balance between TFL1 and FMI genes in the shoot meristems.
Interestingly, the CET2/4 genes from tobacco are most closely related to CEN and ATC. Actually, the expression of CET1 or CET6 (the most similar to TFL1) was not detectable by in situ hybridization. This might indicate that in tobacco, as in antirrhinum (also from the asterids clade), the CEN-related genes have a more prevalent function than the TFL1-related ones, as also suggested by the different effects of CEN and TFL1 over-expression in tobacco (Amaya et al., 1999).
Tomato has a sympodial growth habit and the tomato TFL1-like gene SELF PRUNING (SP) acts specifically on the sympodial meristem (Pnueli et al., 1998). Mutations in SP do not affects the floral transition of the vegetative shoot meristem, which produces the same number of nodes as the wild type before forming the primary inflorescence. However, in sp mutants, the number of leaves per sympodial unit progressively decreases with age until the last unit generates only an inflorescence, so that the shoot is terminated by two consecutive inflorescences. The SP expression pattern is somewhat unusual. While expression of the arabidopsis, antirrhinum and tobacco homologues (TFL1, CEN and CET2/4) is localized to specific shoot meristems, SP expression occurs in all meristems (vegetative, inflorescence floral and axillary), and in leaf and floral organ primordia (Pnueli et al., 1998). It is intriguing how SP only affects vegetative-to-reproductive transitions of sympodial meristems.
As in tobacco, SP is more closely related to CEN than to TFL1, and 35S:CEN (as 35S:SP) rescues the indeterminate phenotype in sp mutants (Pnueli et al., 1998). In tomato again, therefore, a CEN homologue would be playing a ‘TFL1-related’ function, although affecting only the lateral sympodial meristem. SP9D, the tomato homologue most closely related to TFL1, is expressed in roots and in the shoot apex (Carmel-Goren et al., 2003). It would be interesting to see whether mutations in this gene cause a phenotype similar to tfl1, and accelerated flowering.
In I. balsamina, the expression of IbTFL1, a functional homologue of TFL1, has been analysed in both determinate and indeterminate varieties (Ordidge et al., 2005). Expression of IbTFL1 in determinate plants resembles that of the tobacco TFL1-homologues CET2/4, being found in axillary meristems that produce inflorescences, but not in the shoot apical meristem nor in axillary meristems of the upper nodes of the inflorescence, which develop as flowers. This has been used as a basis to suggest that IbTFL1 may be involved in maintaining the inflorescence state of axillary shoot meristems and axillary inflorescences (Ordidge et al., 2005). However, the regulation of the terminal inflorescence seems to be different. While, in principle, the absence of expression in the apex of determinate plants would appear to be linked to terminal flower formation, IbTFL1 is not expressed either in the apex of indeterminate plant varieties or in the apex of plants of the determinate line grown under non-inductive conditions. The meristems of these indeterminate apices express IbLFY, which is therefore insufficient to specify them as floral, even in the absence of IbTFL1 (Pouteau et al., 1997). Consequently, it has been proposed that in I. balsamina the fate of the terminal inflorescence is controlled by an integration system not depending on LFY and TFL1. The AP1 pathway has been suggested as a more likely candidate for this control (Ordidge et al., 2005).
TFL1 homologues have been studied in some detail in a few perennial dicots; species such as orange tree (Citrus sinensis; Pillitteri et al., 2004), apple (Malus domestica; Kotoda and Wada, 2005), Metrosideros excelsa (Sreekantan et al., 2004) and grapevine (Carmona et al., 2007) are good examples (Table 1). As for LFY homologues, no mutants have been described for any of the TFL1 perennial homologues and conclusions on their functions are based on expression studies and analysis of transgenic plants.
In general, expression of the TFL1 homologues in these woody perennials is detected in vegetative tissues, such as apical buds and stems. In situ hybridization has also shown that MeTFL1 is expressed in subapical regions of the inflorescence meristems of M. excelsa, in a pattern very similar to that of TFL1 in arabidopsis. There seems to be a relationship between the two seasons required for flowering in woody plants and the expression of their TFL1 homologues. Thus, the TFL1-like genes from M. excelsa and grapevine have been shown to follow a bimodal expression pattern associated with reproductive development; expression is high in latent buds of the first season, then disappears during the dormancy period, and is then observed again in the second season when bud development is resumed (Sreekantan et al., 2004; Carmona et al., 2007). In general, expression of the TFL1 homologues in perennials does not coincide in space and/or time with that of the LFY or AP1 homologues; instead they are largely complementary. Thus, in grapevine and apple expression of the TFL1 homologues in developing buds is high during the initial stages of inflorescence development, but is absent later during flower development, when expression of the LFY and AP1 homologues becomes high.
For the TFL1-like genes of apple and citrus, constitutive expression in arabidopsis has been shown to cause a late-flowering phenotype, similar to that of plants over-expressing the arabidopsis TFL1 gene. This, and their expression patterns, suggests a role for the TFL1-like genes of these perennials in maintaining indeterminacy of the shoot meristems within the developing bud.
Regarding the juvenile phase, the TFL1 homologues from citrus and apple have been suggested also to be involved in the regulation of its duration. For citrus, this is based on the high levels of CsTFL transcripts found in juvenile tissues (Pillitteri et al., 2004), and in apple on the strong reduction of the juvenile phase in antisense MdTFL1 apple trees (Kotoda et al., 2003). Whether this is also a likely function for the other perennial TFL1 homologues remains to be investigated.
A major conclusion of this review is that LFY, AP1 and TFL1 genes are major factors controlling the behaviour of reproductive meristems, not only in arabidopsis but in most plant species. The emerging picture from the comparative studies described here is that differences in their expression patterns or in the activity of their proteins could explain a great part of the variation in the basic inflorescence architecture found among species.
Something that seems generally conserved is the antagonism between the function of LFY and TFL1 homologues, clearly shown by their generally mutual complementary expression patterns. Although they are often considered regulators of floral identity specification, both genes seem to play a relatively general role in the control of meristem fate. TFL1 homologues apparently act by ‘repressing’ transitions between developmental phases: all phase transitions at the arabiodpsis SAM, floral transition of the sympodial meristem of tomato or, in legumes, I1 to I2 transition. In addition to floral meristem identity, LFY genes also appear to be involved in phase transitions, though with a promoting role. For example, in arabidopsis LFY integrates genetic pathways inducing the floral transition.
LFY genes encode a type of transcription factor, unique to the plant kingdom, which is found from mosses to eudicots. Remarkably, the LFY genes have dupicated very rarely, usually being represented by only one or two homologues in the different plant species. Nevertheles, during its long evolutionary history, the LFY genes appear to have been recruited for other functions not related to flowering, such as control of leaf development in legumes and tomato.
AP1 is a MADS-box type transcriptional regulator. The MADS-box gene family has greatly diversified in the plant kingdom, leading to the appearance of a high number of genes with very diverse roles in the regulation of plant development. The AP1 genes are only found in the core eudicot clade, probably linked to the origin of the eudicot flower. Accordingly, the expression and function of this group of genes are generally related to floral meristems.
Finally, knowledge on the evolution of TFL1 homologues is much more limited than that of LFY- or AP1-like genes. In the different eudicot species analysed, the TFL1 gene family has no more than six members and the available data suggest that its function as a phase-change repressor could have been adopted by two different members of the TFL1 clade, TFL1-homologues in rosids and CEN-homologues in asterids, possibly as consequence of evolution of their expression patterns.
Our knowledge of the genetic control of inflorescence development in different plants is still limited. It seems clear that more studies, systematically comparing the function of the relevant genes in species representing the major clades of the plant phylogenetic tree, need to be done in order to have a more comprehensive view. Nevertheless, it is remarkable that, despite the available information still being fragmentary, the analysis in different species of a few genetic functions initially identified in model plants is showing that such a very simple regulatory network could be the basis of the generation of a huge variety of inflorescence architectures. This is a clear example of how comparative studies of genetic regulatory functions are helping us to understand the evolution of plant development.
We thank Cristina Ferrándiz and Pedro Fernández-Nohales for critical reading of the manuscript. We also thank two anonymous referees and to the editor for their comments and suggestions to improve the manuscript. The work of our lab is funded by grants from the Secretaría General del Plan Nacional de Investigación Científica y Desarrollo Tecnológico (Spain) (grant Nº BIO2006–10994, the Conselleria d'Empresa Universitat i Ciencia from the Generalitat Valenciana, and by the European Union Grain Legumes Integrated Project (grant no. FP6-2002–FOOD–1–506223).