Coming into bloom: the specification of floral meristems

Flowering plants, also known as angiosperms (see Glossary, Box 1), were the last of the seed-bearing plant groups to evolve. The most obvious features that distinguish angiosperms from other seed-bearing plants are their reproductive organs, the flowers. In the course of flowering, plants undergo a transition from vegetative to reproductive growth (see Glossary, Box 1), known as the floral transition. Flowers contain reproductive structures, such as stamens and carpels (see Glossary, Box 1;see also Fig. 1), and upon fertilization a subset of carpels develops into fruits. These fruits contain seeds, from which new plants can grow, thus permitting the transfer of genetic information to the next generation.

When plants initiate flowering, the vegetative shoot apical meristem (SAM;see Glossary, Box 1), which gives rise to all the parts of a plant that are above ground, is transformed into an inflorescence meristem (IM; see Glossary, Box 1). The IM, in turn,generates a collection of undifferentiated cells called floral meristems (FMs)that give rise to floral organs. As FMs arise in response to multiple flowering signals and eventually differentiate into various types of floral organ, the regulation of FM development is a crucial and dynamic switch that allows for the successful reproductive development of flowering plants in an unpredictable environment.

Morphological changes of flower development have been monitored in detail in the model plant Arabidopsis thaliana(Smyth et al., 1990), in which flower development is divided into 12 stages according to a series of landmark events. Floral primordia that are present prior to visible floral organogenesis are generally considered to be FMs. Even though floral anlagen are morphologically invisible before stage 1, they have already become distinguishable from other cells in IMs, as is apparent from the expression of certain marker genes. One such marker gene is AINTEGUMENTA(ANT), which encodes a transcription factor of the plant-specific AP2/EREBP family (Fig. 2A). Floral anlagen at this transitional phase are usually referred to as stage 0 FMs (Long and Barton, 2000). Stage 1 FMs emerge as outward bulges on the flank of the IM, with each new FM forming at an angle of ∼130°-150° to previously established ones. From stage 1 to the end of stage 2, FMs enlarge gradually into ball-shaped structures and become separated from the IM(Fig. 2B). The primordia of the first whorl (see Glossary, Box 1) of floral organs, sepals (see Glossary, Box 1), appear at the periphery of the FMs at stage 3 and start to overlie FMs at stage 4, and this is followed by the successive emergence of other floral organs in the internal whorls. The intimate developmental link between FMs and floral organs indicates that the specification of FMs is a key preliminary step for successful flower development.

Recent molecular genetic studies have provided new insights into the specification of FMs in Arabidopsis and other flowering plants. In this review, we focus on the latest progress in our understanding of the regulatory networks that underpin several coordinated programmes of FM development in Arabidopsis and discuss the homologues of key genes that regulate FMs in a variety of flowering plants (angiosperms) to evaluate the conservation of relevant mechanisms across plant species.

FMs are exclusively produced from IMs, the reproductive SAMs into which vegetative SAMs are transformed during the floral transition. By contrast,other organ primordia initiate from plant SAMs during both vegetative and reproductive phases; this suggests a unique role for IMs in specifying FMs. It should, however, be noted that many grasses have evolved more specialized,so-called axillary meristems (see Glossary, Box 1) from IMs that are produced before producing FMs to acquire highly branched inflorescences(Fig. 2C). For instance,indeterminate IMs (see Glossary, Box 1) or their derived branch meristems in maize (Zea mays)give rise to spikelet pair meristems (see Glossary, Box 1), which further differentiate into spikelet meristems and finally into FMs(Barazesh and McSteen, 2008). Regardless of how FMs are ultimately formed, however, the generation of IMs is a prerequisite for FM specification in most flowering plants.

The molecular mechanisms that underlie the transition from vegetative SAMs to IMs have been intensively investigated in Arabidopsis(Fig. 3). This transition is mediated by a complex network of genetic pathways that regulate flowering in response to environmental and developmental signals(Blazquez et al., 2003; Boss et al., 2004; Mouradov et al., 2002; Simpson and Dean, 2002). The autonomous pathway regulates flowering by monitoring endogenous cues from different developmental stages, whereas the gibberellin (GA; see Glossary, Box 1) pathway affects flowering particularly in short-day conditions. The photoperiod and vernalization pathways (see Glossary, Box 1) mediate the responses to environmental signals, such as day length and low temperatures. In addition, some other genetic pathways, such as the ones that respond to changes in light quality and ambient temperature,have been proposed to affect flowering. The flowering signals perceived by these pathways converge on the transcriptional regulation of two major floral pathway integrators, FLOWERING LOCUS T (FT) and SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1 (SOC1). These, in turn, activate FM identity genes, such as LEAFY (LFY) and APETALA1 (AP1), to produce FMs on the flanks of IMs(Blazquez and Weigel, 2000; Kardailsky et al., 1999; Kobayashi et al., 1999; Lee et al., 2000; Lee et al., 2008; Liu et al., 2008; Samach et al., 2000).

The integration of flowering signals is tightly controlled by a repressor complex that consists of two MADS-box transcription factors, FLOWERING LOCUS C(FLC) and SHORT VEGETATIVE PHASE (SVP)(Hartmann et al., 2000; Li et al., 2008; Michaels and Amasino, 1999; Sheldon et al., 1999). The vernalization and autonomous pathways mainly repress FLC expression through the modulation of its chromatin structure(Michaels, 2009), which promotes flowering by antagonizing the repressive effect of FLC on FTand SOC1 expression (Helliwell et al., 2006; Searle et al.,2006). FLC represses FT expression in leaves; this blocks the translocation of the systemic flowering signals that contain FT protein to the SAMs, an event that is required for activating the expression of SOC1 and AP1 (Abe et al.,2005; Corbesier et al.,2007; Searle et al.,2006; Wigge et al.,2005). FLC also directly represses the expression of SOC1and of the FT cofactor FD in SAMs(Searle et al., 2006), thus further inhibiting the meristem response to flowering signals.

In vegetative seedlings at various ages, the FLC-SVP repressor complex responds mainly to flowering signals that are perceived by the autonomous, the thermosensory and the GA pathways(Hartmann et al., 2000; Lee et al., 2007b; Li et al., 2008). Their mutually dependent function directly regulates SOC1 expression in whole seedlings, as well as FT expression in leaves. Thus, most flowering pathways (with the exception of the photoperiod pathway) appear to promote the expression of FT and SOC1 predominantly through derepression mechanisms (Fig. 3).

Unlike FT, SOC1 is highly expressed in IMs, which makes it a good candidate for contributing to the spatial specificity of FM initiation, during which a small block of cells (normally four cells) on the IM flank acquire progenitor fate for a future FM (Bossinger and Smyth, 1996; Lee et al.,2000; Samach et al.,2000). By contrast, SVP is expressed in both leaves and SAMs during the vegetative phase and is absent from IMs during the reproductive phase (Hartmann et al.,2000). As the repressive effect of SVP on SOC1transcription outweighs the effects of SOC1 activators such as FT and AGAMOUS-LIKE 24 (AGL24) (Li et al.,2008), a decrease in SVP expression is a key event required for the transformation of vegetative SAMs into IMs. The abundance of SVP protein has been found to increase in certain circadian clock mutants under continuous light (Fujiwara et al.,2008), but how SVP expression is gradually downregulated in SAMs during floral transition remains unclear. Overall, the interaction of the above-mentioned flowering regulators in various flowering genetic pathways mediates the transition from vegetative SAMs to IMs, from which FMs are derived.

The regulation of FM initiation not only involves the activation of two well-known FM identity genes, LFY and AP1, but also depends on the control of auxin flux and tissue polarity(Blazquez et al., 2006). Even though the latter two factors have seldom been reviewed in association with FM specification, they are temporally and spatially correlated to the onset of FM development (Blazquez et al.,2006; Heisler et al.,2005). In this section, we discuss how the distribution of auxin,which is affected by its biosynthesis, transport and signalling, influences FM initiation in Arabidopsis and monocotyledons (see Glossary, Box 1; see also Fig. 4). In addition, we also review the regulation of tissue polarity during FM initiation.

In Arabidopsis, the heterogeneous distribution of auxin affects the initiation of all axillary meristems(Benkova et al., 2003),including the initiation of FMs in IMs. Here, auxin accumulates at the positions of floral anlagen, but gradually decreases in concentration with increasing distance from them (Heisler et al., 2005; Oka et al.,1999; Reinhardt et al.,2003). This pattern of auxin distribution is mediated by both auxin biosynthesis and polar auxin transport. At the early stages of reproductive development, FM formation is abolished in Arabidopsisquadruple mutants (yuc1 yuc2 yuc4 yuc6) of the YUCCA(YUC) family of flavin monooxygenases, which are essential for auxin biosynthesis (Cheng et al.,2006). Simultaneous mutations in these four YUC genes result in a naked inflorescence stem. Similar phenotypes are seen in plants with loss-of-function mutations in NAKED PINS IN YUC MUTANTS(NPY) and AGC KINASE genes. Although these genes have been proposed to act in a linear pathway together with YUC genes(Cheng et al., 2008), their exact function in auxin-mediated organogenesis remains to be elucidated further. Loss-of-function mutations in the auxin efflux carrier PIN-FORMED 1 (PIN1), which regulates polar auxin transport, also produce naked inflorescence stems without FMs(Vernoux et al., 2000). The live imaging of Arabidopsis IMs with concurrent monitoring of the expression of PIN1 and of the auxin-responsive reporter DR5 has further revealed that auxin transport is intimately associated with FM initiation(Heisler et al., 2005).

The regulation of PIN1 activity also affects FM initiation. Intercellular auxin fluxes are controlled by the phosphorylation status of PINs, which is mediated through the antagonistic regulation of an AGC kinase, PINOID (PID),and PROTEIN PHOSPHATASE 2A (PP2A)(Michniewicz et al., 2007). As pid mutants fail to produce FMs(Cheng et al., 2008), the modulation of the PIN1 phosphorylation status appears to play a role in FM initiation. Interestingly, additional factors that regulate PIN1 function have been identified recently. For example, P-glycoprotein (PGP) transport proteins, which have been suggested to form another group of auxin efflux carriers, genetically interact with PINs in a concerted fashion during organogenesis (Mravec et al.,2008), whereas AUXIN RESISTANT 1 (AUX1), an auxin influx carrier,and its paralogues LIKE AUX 1, 2 and 3 (LAX1, LAX2 and LAX3) are required for mediating coordinated PIN1 polarization(Bainbridge et al., 2008). Whether these factors are also involved in PIN1-mediated FM initiation,however, remains to be elucidated.

Consistent with the roles of auxin biosynthesis and transport in FM initiation discussed above, auxin signalling also has a crucial function in this process. Auxin response factors (ARFs) are considered to be key components of the auxin signalling pathway, and loss-of-function mutations in the ARF gene MONOPTEROS (also known as ARF5) abolish FM initiation (Przemeck et al.,1996). The phenotype seen in these mutants is similar to that observed in yuc, pin1 and pid mutant plants. These results clearly show that auxin plays an indispensable role in FM initiation.

Several lines of evidence have provided a molecular link between auxin and FM specification. First, LFY expression is reduced and changed into a ring-like pattern that encircles the IM of pin1 mutants, and the expression of LFY downstream targets, such as AP1 and AP3, also decreases in pin1(Vernoux et al., 2000). Second, the dynamic expression of PIN1 protein corresponds to LFYexpression at the sites of FM initiation(Heisler et al., 2005). Third,an auxin response element has been identified in the LFY promoter that might be recognized by an ARF (Bai and DeMason, 2008). Taken together, these observations indicate that the initiation of FMs, which is regulated by auxin, might be integrated with the specification of FM identity by LFY.

Recent progress suggests that the regulatory mechanisms of FM initiation through auxin biosynthesis and transport might be partially conserved from Arabidopsis to monocotyledons (see Glossary, Box 1). The maize gene sparse inflorescence 1 (spi1) encodes a YUC-like flavin monooxygenase that is involved in local auxin biosynthesis and in the regulation of axillary meristems, including the initiation of spikelet meristems and FMs (Gallavotti et al.,2008a). In addition, PIN1-like genes have been identified in maize and rice (Carraro et al.,2006; Paponov et al.,2005). Zea mays PIN1a (ZmPIN1a), a PIN1 homologue in maize, is localized in the L1 layer of axillary meristems and IMs(Gallavotti et al., 2008b),which is comparable to the localization of PIN1 in Arabidopsis. Moreover, ZmPIN1a activity rescues Arabidopsis pin1-3, resulting in the re-establishment of auxin maxima and the re-formation of FMs, which indicates that the auxin transport mechanism during FM initiation might be conserved between Arabidopsis and grasses(Gallavotti et al., 2008b). Interestingly, the phosphorylation and localization of ZmPIN1a is also regulated by a homologue of PID, BARREN INFLORESCENCE2 (BIF2)(McSteen et al., 2007; Skirpan et al., 2009), which suggests similarities in the regulation of auxin transporter trafficking between maize and Arabidopsis. In rice (Oryza sativa),OsPID, another orthologue of PID, has been suggested to function in polar auxin transport (Morita and Kyozuka,2007), but its role in FM initiation is so far unknown.

FM initiation inherently involves the establishment of tissue polarity, as illustrated by the fact that several polarity genes were found to mark the abaxial and adaxial sides (see Glossary, Box 1) of FMs. The Arabidopsis FILAMENTOUS FLOWER (FIL) gene, which encodes a member of the YABBY family of transcription factors, is specifically expressed on the abaxial side of emerging FMs (Sawa et al., 1999b; Siegfried et al., 1999). In fil mutants, FMs differentiate into various structures, including flowerless pedicels (see Glossary, Box 1) and curled sepals(Chen et al., 1999). Moreover,crossing fil mutants with ap1 or lfy mutants results in plants with enhanced defects in FM formation(Sawa et al., 1999a). These observations suggest that properly established tissue polarity regulated by FIL is required, together with AP1 and LFY, for FM specification.

It has been shown that the initial asymmetric development of leaf primordia is controlled by a mutual antagonism between the PHABULOSA(PHB)-like genes, which promote adaxial cell fate, and the abaxial-fate-promoting KANADI (KAN) genes. This antagonism,in turn, affects polar YABBY expression, which promotes abaxial cell fate (Eshed et al., 2001; Eshed et al., 2004). Whether this mechanism also regulates the function of FIL in FMs remains unknown. PHB, PHAVOLUTA (PHV) and REVOLUTA(REV) are a group of class III homeodomain/leucine zipper (HD-ZIP)genes that regulate adaxial cell fate in lateral organs(Emery et al., 2003; McConnell et al., 2001). Among these genes, REV has been demonstrated to play an important role in FM formation (Otsuga et al.,2001). In rev mutants, some FMs develop with reduced size. Notably, fil rev double mutants show greatly enhanced floral defects, with FMs transforming completely into pedicels(Chen et al., 1999). Thus, the interaction between adaxial-promoting genes, such as REV, and abaxial-promoting genes, like FIL, might determine tissue polarity in a way that is important for the proper initiation of FMs.

Interestingly, ETTIN, which is also known as AUXIN RESPONSE TRANSCRIPTION FACTOR 3 (ARF3), regulates organ asymmetry through the modulation of KAN activity(Pekker et al., 2005). This links auxin signalling with the regulation of tissue polarity and indicates that tissue polarity is fine-tuned through certain ARFs that are stimulated by auxin gradients. Furthermore, during FM initiation, PIN1 expression marks a domain between abaxial and adaxial cell identities, as marked by FIL and REV expression, respectively. This lends further support to the notion that auxin transport patterns influence organ polarity in FMs (Heisler et al., 2005). It will be instructive to investigate how auxin is involved in FM initiation. One possibility is that it affects FM identity through LFY and mediates FM polarity by regulating the expression of abaxial and adaxial genes(Fig. 4).

The emerging FMs are specified by the so-called FM identity genes,including LFY and AP1. The characterization of FM identity genes in Arabidopsis and the isolation of their homologues in different plant species suggest that some conserved mechanisms underlie FM specification, even though the homologues of FM identity genes might have evolved various functions in different taxonomic groups.

LFY and AP1 are two major FM regulators that specify FM identity on the flanks of IMs in Arabidopsis(Bowman et al., 1993; Mandel and Yanofsky, 1995; Weigel et al., 1992). When the activity of either gene is lost, FMs that would normally develop into flowers are partly converted into IMs. It has long been known that the shoot identity gene TERMINAL FLOWER 1 (TFL1) antagonizes LFY and AP1 and thus counteracts the establishment of FM identity (Liljegren et al.,1999; Ratcliffe et al.,1999). However, this antagonistic interaction does not explain the puzzle of how LFY and AP1 are regulated in response to upstream flowering signals to specify FMs in IMs, as the mechanism by which TFL1 is integrated into the flowering regulatory networks remains unclear. Recent studies on the integration of flowering signals have, however,shed some light on the regulation of LFY and AP1(Fig. 3).

LFY plays a dual role in regulating FM identity and floral organ patterning (Parcy et al.,1998), and its expression is affected by several flowering pathways (Blazquez and Weigel,2000). Among all the known flowering-time factors, SOC1 is currently the only transcription factor known to bind to the LFYpromoter in vivo, and this binding process is partly mediated through the interaction of SOC1 with AGL24 (Lee et al., 2008; Liu et al.,2008). SOC1 expression gradually increases in SAMs during the floral transition in response to multiple flowering signals(Lee et al., 2000; Samach et al., 2000). This increase could provide temporal and spatial cues for promoting LFYexpression in the incipient floral primordia to the threshold levels that are required for FM specification.

Three closely related MADS-box genes, AP1, CAULIFLOWER(CAL) and FRUITFULL (FUL), also appear to be potential activators of LFY during the floral transition(Ferrandiz et al., 2000). A combination of mutations in these three genes produces leafy shoots in place of flowers (Ferrandiz et al.,2000). The abolishment of LFY upregulation is partially responsible for this phenotype, which indicates that FUL, AP1 and CAL act redundantly upstream of LFY in determining FM identity. The functional redundancy between FUL and SOC1also masks their roles in FM formation(Melzer et al., 2008). These two genes share a similar expression pattern in both IMs and FMs. soc1 ful double mutant plants show strongly delayed flowering when grown under long day conditions when compared with the single mutants. Interestingly, the apical IMs of soc1 ful revert into vegetative SAMs after the plants enter the reproductive phase (Melzer et al., 2008). This pattern is recurrent, which is reminiscent of the lifestyle of perennial plants. These observations demonstrate that SOC1 and FUL not only control flowering time, but also play an important role in meristem determinacy, which might be partly attributed to their function in modulating LFY expression. Another key floral pathway integrator, FT, and its cofactor, FD, activate SOC1expression in IMs (Abe et al.,2005; Corbesier et al.,2007; Wigge et al.,2005) and promote FUL expression in leaves as well as,potentially, in IMs (Teper-Bamnolker and Samach, 2005). Therefore, FT could control LFY expression through both SOC1 and FUL during the floral transition.

AP1 itself is another major FM identity gene that is specifically expressed in emerging FMs (Mandel et al.,1992). During the floral transition, AP1 expression is directly activated by LFY and by a complex consisting of FT and FD(Abe et al., 2005; Wagner et al., 1999; Wigge et al., 2005). AP1 function overlaps with that of CAL genes, as ap1 cal1 mutants show a complete transformation of FMs into IMs(Bowman et al., 1993). LFY determines FM identity by directly controlling the expression of at least three transcription factors, namely AP1, CAL, and LATE MERISTEM IDENTITY 1 (LMI1), which encodes a class I HD-ZIP transcription factor (Saddic et al., 2006; William et al., 2004). Together with LFY, LMI1 controls CAL expression directly. This interaction is suggested to form a coherent feed-forward loop that fine-tunes the FM identity switch in response to environmental stimuli(Saddic et al., 2006). These data suggest that the network that converges on the regulation of LFYand AP1 by SOC1 and FT might be an essential molecular link that translates the multiple flowering signals integrated by FT and SOC1 into the actual specification of FMs by LFY and AP1 (Fig. 3).

Since the isolation of the LFY homologue FLORICAULA(FLO) and of the AP1 homologue SQUAMOSA(SQUA) in snapdragon (Antirrhinum majus)(Coen et al., 1990; Huijser et al., 1992),additional homologues of LFY and AP1 have been identified in many other plant species. LFY homologues are present in all the land plants that have been analyzed to date, including moss(Chujo et al., 2003). A LFY homologue appears to have been recruited to flower development in the ancestor of all angiosperms, as it is involved in this process in all angiosperm species tested so far (Benlloch et al., 2007; Blazquez et al.,2006). The extent of phenotypical complementation of Arabidopsis lfy mutants by different LFY homologues seems to be related to the taxonomic distance from Arabidopsis, ranging from no complementation by moss homologues to full complementation by angiosperm homologues (Maizel et al.,2005). In some instances, however, LFY homologues seem to have been recruited to play additional roles along with their conserved function in FM specification. For example, some LFY homologues, such as UNIFOLIATA in pea and FALSIFLORA in tomato, regulate leaf development (Hofer et al.,1997; Molinero-Rosales et al.,1999), whereas studies on the function of maize and rice FLO/LFY genes have revealed a role for monocotyledonous LFY homologs in inflorescence branching(Bomblies et al., 2003; Kyozuka et al., 1998).

Phylogenetic analyses of AP1/FUL-like MADS-box genes reveal the presence of two gene clades within the core eudicotyledons (see Glossary, Box 1),euAP1 (e.g. AP1) and euFUL (e.g. FUL)(Litt and Irish, 2003). The homologues of the euAP1 gene clade are found only in core eudicotyledons, which includes the majority of extant angiosperm species(Litt and Irish, 2003). This suggests that euAP1 function might be specific to flower formation in core eudicotyledons. Similar to LFY homologues, however, some homologues of FUL-like genes, which are not restricted to eudicotyledons, show novel functions in certain plant species in addition to their role in specifying FM identity(Benlloch et al., 2007). In grasses, for example, FUL1 and FUL2 have evolved additional functions in regulating the floral transition(Preston and Kellogg,2007).

Apart from the homologues of Arabidopsis FM identity genes, other meristem identity genes that are unique to grass species have been isolated. FM initiation in maize is controlled by an APETALA2(AP2)-like gene, indeterminate spikelet1 (ids1),and by its related gene sister of indeterminate spikelet1(sid1). Loss-of-function mutations in either of these genes abolish FM initiation, which indicates that in grasses, the AP2 genes might replace LFY to function in FM identity(Chuck et al., 2008). Therefore, the mechanisms that underlie the specification of FMs in grasses are partly similar to those in Arabidopsis; however, grasses, which frequently have complex floral and inflorescence structures, might also have evolved some unique genetic and molecular programmes of FM specification.

In the course of flowering, the emerging FMs can potentially take a developmental step backwards to turn into inflorescence shoots, a phenomenon called floral reversion, or precociously differentiate to produce abnormal floral organs. Therefore, simply establishing FM identity is not sufficient for securing normal flower development. Additional mechanisms that are responsible for the active maintenance of floral identity in FMs appear to be required until normal floral patterning occurs at a later stage. In this section, we discuss the evidence in favour of the existence of such mechanisms.

In Arabidopsis, floral reversion often occurs in FM identity mutants, such as lfy and ap1, which indicates that the mutated genes play key roles in maintaining FM identity by repressing floral reversion.

ap1 mutants are characterized by the generation of secondary flowers or inflorescences in individual FMs, which signifies a partial reversion from FMs to IMs (Bowman et al.,1993). These phenotypes appear to be partially attributable to the activity of three flowering-time genes, AGL24, SVP and SOC1(Liu et al., 2007; Yu et al., 2004), because loss-of-function mutations in these three genes, either individually or combined, alleviate the FM defects seen in ap1 by lowering the frequency of secondary structure production. Indeed, the expression of these genes is upregulated in ap1 FMs. Consistently, the transgenic expression of AGL24 under the control of the constitutive 35S promoter (35S:AGL24) promotes the transformation of FMs into IMs, a phenotype that is enhanced by 35S:SOC1, whereas the transgenic expression of 35S:SVP promotes the transformation of FMs into vegetative shoots (Liu et al.,2007; Masiero et al.,2004; Yu et al.,2004). It has been shown that induced AP1 activity represses the expression of AGL24, SVP and SOC1(Liu et al., 2007; Wellmer et al., 2006; Yu et al., 2004), and that AP1 binds directly to the promoters of these three genes(Gregis et al., 2008; Liu et al., 2007). These results suggest that the suppression of these flowering-time genes by AP1 is one of the processes involved in maintaining FM identity(Fig. 5).

In contrast to AP1, LFY might not directly repress AGL24,SVP or SOC1. The repression of AGL24 by induced LFY activity could be mediated through certain unknown mediator(s)(Yu et al., 2004). Moreover, SVP and SOC1 are not upregulated in lfy FMs(Gregis et al., 2008; Liu et al., 2007). As LFY directly upregulates AP1 in FMs, it is possible that LFYspecifies FMs partly through AP1(William et al., 2004).

Similar to ap1 mutants, secondary flowers have also been observed in Arabidopsis plants that carry mutations in three of the SEPALLATA (SEP) floral identity genes, SEP1, SEP2and SEP3 (Fig. 5),and, at a higher frequency, in sep1 sep2 sep3 sep4 quadruple mutants. This indicates that, in addition to their role in specifying floral identity, SEP1, SEP2, SEP3 and SEP4 are also involved in FM specification (Ditta et al.,2004). Both AGL24 and SVP are expressed in the ectopic FMs of sep1 sep2 sep3 mutants. Chromatin immunoprecipitation(ChIP) results have further demonstrated the direct binding of SEP3 to AGL24 and SVP promoters, which indicates that SEP3 is involved in directly repressing AGL24 and SVP in FMs(Gregis et al., 2008). SEP1, SEP2 and SEP4 are expressed throughout stage 2 FMs(Ditta et al., 2004; Flanagan and Ma, 1994; Savidge et al., 1995); this expression pattern overlaps with AP1 expression. Although SEP3 transcripts start to accumulate in the upper portion of late stage 2 FMs (Mandel and Yanofsky,1998), protein localization analysis has recently shown the presence of SEP3 protein in FMs from stage 1 onwards(Urbanus et al., 2009). Furthermore, AP1 interacts with SEP proteins (except SEP2) in yeast(de Folter et al., 2005). These results indicate that AP1 and SEPs might form protein complexes to maintain FM identity by directly suppressing the expression of AGL24and SVP. In addition, ectopic AGL24 and SVPexpression is also detectable in mutants with loss-of-function mutations in the floral identity gene AGAMOUS (AG), which display defects in FM termination and in the growth of reproductive organs(Gregis et al., 2008; Lenhard et al., 2001; Mizukami and Ma, 1997). Therefore, the precise control of AGL24 and SVP expression seems to be a consistent mechanism that is required for FM specification and flower development (Fig. 5).

However, the role played by AGL24 and SVP in FMs is still unclear. Based on the alleviation of ap1 floral phenotypes by agl24 or svp, it has been proposed that, in the absence of AP1, AGL24 or SVP might recapture its function during the floral transition to promote either inflorescence or vegetative shoot identity in FMs, respectively (Liu et al.,2007; Yu et al.,2004). The observation that FMs are transformed into IMs or shoot meristems upon overexpression of AGL24 or SVP supports this possibility. By contrast, based on the FM-to-IM transition that is observed in ap1agl24 svp triple mutants(Gregis et al., 2008), AGL24 and SVP have been suggested to promote FM fate. Although the function of AGL24 and SVP in FM specification needs to be elucidated further, the current consensus appears to be that,overall, these two factors promote FM indeterminacy(Fig. 5).

AGL24 and SVP are members of the StMADS11 clade of MADS-box genes (Becker and Theissen,2003). Recent studies have identified members of this clade in a wide range of plant species. Notably, the overexpression of many StMADS11-like genes from dicotyledons and monocotyledons in Arabidopsis results in floral phenotypes similar to those produced by the transgenic expression of 35S:SVP or 35:AGL24(Table 1), which indicates that these genes could share common functional properties in FM development. In Antirrhinum, the StMADS11 member INCOMPOSITA(INCO) acts with FLO and SQUA to specify FM identity (Masiero et al.,2004). Interestingly, INCO has been found to either inhibit or promote FM identity in different Antirrhinum mutant backgrounds (Masiero et al.,2004). As squa inco double mutants produce more flowers than do squa, INCO appears to prevent the development of reproductive axillary meristems into flowers in this context. Furthermore, the overexpression of INCO in Arabidopsis produces flowers with vegetative characters that are similar to the flowers of 35S:SVPtransgenic plants. These results suggest that INCO represses FM identity. By contrast, in Antirrhinum flo-662 inco double mutants, inco enhances the FM defect shown in the weak flo-662mutant, with axillary inflorescences being generated instead of flowers, which suggests that INCO and FLO act together to promote FM identity. These contradictory functions of INCO in the regulation of FM identity could be due to the interaction of INCO with additional protein partners. It has been suggested that, in the presence of SQUA, the INCO-SQUA heterodimer might act together with FLO to specify FM identity whereas, in the absence of SQUA, the INCO homodimer might inhibit FM identity(Masiero et al., 2004). It is noteworthy that protein-protein interactions between AP1 homologues (e.g. SQUA) and StMADS11-clade regulators (e.g. INCO) have been detected in several plant species, such as Arabidopsis(de Folter et al., 2005), Pharbitis nil (Kikuchi et al.,2008) and wheat (Triticum aestivum L.)(Kane et al., 2005). Thus, it will be important to investigate whether the function of StMADS11-like genes in regulating FM identity is modulated through protein interactions with additional FM identity genes, particularly with members of the AP1/SQUA gene clade.

In monocotyledons, StMADS11-like genes also affect FM identity. The overexpression of barley MADS1 (BM1) and BM10inhibits floral development and results in floral reversion in both barley and Arabidopsis (Trevaskis et al.,2007). In addition, the overexpression of Oryza sativa MADS22 (OsMADS22) or OsMADS47 in Arabidopsiscauses floral reversion and floral defects that are similar to the phenotypes observed when overexpressing SVP or AGL24(Fornara et al., 2008). These results indicate that StMADS11-clade genes might play conserved roles in regulating FM identity. Although the mechanisms of action of these genes still need to be investigated further, the appropriate control of their expression in FMs seems to be crucial for the maintenance of FMs, which lays the foundation for further normal floral patterning.

Another key aspect in the maintenance of FM identity is the prevention of precocious differentiation triggered by the onset of expression of floral homeotic genes that specify floral organ identity. In Arabidopsis,each whorl of floral organs is determined by the combinatorial action of the class A, class B and class C floral homeotic genes(Box 2)(Bowman et al., 1991; Coen and Meyerowitz, 1991). Three out of four Arabidopsis floral organ identities, namely petals(see Glossary, Box 1), stamens and carpels, are controlled by class B and/or class C homeotic genes, which are only activated in early stage 3 FMs. In many mutants in which class B or class C genes are precociously activated in a deregulated pattern, FMs prior to stage 3 that contain insufficient numbers of meristem cells are compelled to enter the floral organogenesis program, which results in a reduced number of floral organs and in the deregulation of floral organ identities. It appears that the repression of floral homeotic genes is the `default'programme in emerging FMs to ensure that floral anlagen fully develop into stage 3 FMs that contain sufficient cells for the proper patterning of whorled organs by floral homeotic genes. During the past few years, efforts have been made to understand how floral homeotic genes are appropriately repressed in emerging FMs (Table 2).

SEUSS (SEU) and LEUNIG (LUG) are transcriptional co-regulators that negatively regulate the expression of the class C floral homeotic gene AG(Franks et al., 2002; Liu and Meyerowitz, 1995). Flowers of Arabidopsis seu lug double mutants exhibit severe floral homeotic transformation, with ectopic AG expression throughout the FMs. SEU interacts with LUG to form a protein complex(Sridhar et al., 2004), and ChIP assays have shown that this SEU-LUG complex directly associates with the AG promoter (Sridhar et al.,2006). Because neither SEU nor LUG contains a DNA-binding domain,an interesting question is how they are directed to the promoters of their target genes. AP1 has been identified as an interacting partner of SEU(Sridhar et al., 2006). Comprehensive yeast two-hybrid assays among Arabidopsis MADS-box proteins show that AP1 can also interact with AGL24 or SVP(de Folter et al., 2005). Moreover, the LUG-SEU co-repressor complex interacts with AP1-AGL24 and AP1-SVP dimers (Gregis et al.,2006), which indicates that these proteins might form a higher-order protein complex to control target gene expression. The observations that agl24 svp ap1 triple mutants show lug-like floral defects and that AG is ectopically expressed in FMs of agl24 svp double mutants support a common role for AP1,AGL24 and SVP in preventing the ectopic expression of AG in FMs (Gregis et al.,2006).

Recently, it has been shown that the expression of class B and class C homeotic genes in FMs before stage 3 is redundantly repressed by AGL24,SVP and SOC1 through the direct repression of SEP3(Liu et al., 2009). In soc1 agl24 svp triple mutants, strong ectopic SEP3 activity interacts with LFY activity to synergistically activate class B and class C floral homeotic genes in floral anlagen and emerging FMs, resulting in striking floral defects, such as the loss of most floral organs and the generation of chimeric floral structures (Liu et al.,2009). Thus, AGL24, SVP and SOC1 suppress SEP3 to regulate the timing of floral organ patterning by inhibiting the ectopic expression of floral homeotic genes in young FMs. These results suggest that the regulation of AGL24, SVP and SOC1 at an appropriate level is crucial for the maintenance of FMs because their elevated expression causes FM indeterminacy, whereas lower expression causes the precocious differentiation of FMs (Fig. 5).

The activation of the class C gene AG is required for the termination of FMs through the repression of a meristem gene, WUSCHEL(WUS), which is necessary for maintaining FMs in a proliferative and indeterminate state (Lenhard et al.,2001; Lohmann et al.,2001); this repression was recently found to be mediated by the C2H2-type zinc finger protein KNUCKLES(Sun et al., 2009). LFY and WUS act together to induce AG expression in FMs, and the participation of SEP3 in AG induction implies that the direct regulators of SEP3, namely AGL24, SVP and SOC1, play a role in preventing FM termination through mediation of the timing of AGexpression (Fig. 5).

Another important group of regulators involved in repressing floral homeotic gene expression are chromatin regulators, many of which are Polycomb Group (PcG) proteins that affect chromatin states to inhibit the transcription of floral homeotic genes (Table 2). The PcG proteins EMBRYONIC FLOWER 2 (EMF2), CURLY LEAF (CLF),SWINGER (SWN), FERTILIZATION-INDEPENDENT ENDOSPERM 1 (FIE1) and MULTICOPY SUPRESSOR OF IRA1 (MSI1) form a putative Polycomb Repressive Complex 2 (PRC2)that catalyzes the tri-methylation of lysine 27 of histone H3 (H3K27me3) of target genes, leading to their transcriptional silencing(Chanvivattana et al., 2004; Farrona et al., 2008; Goodrich et al., 1997; Hennig et al., 2003; Katz et al., 2004). Mutations in clf, emf2, fie or msi1 cause the ectopic expression of floral homeotic genes, even in embryos or vegetative seedlings, indicating that these PRC2 components are required for the repression of floral homeotic gene expression during plant development. EMBRYONIC FLOWER 1 (EMF1), a potential PRC1-like factor that maintains the transcriptional repression of targets by recognizing H3K27me3, acts together with the EMF2 complex to repress AG expression during vegetative development(Calonje et al., 2008).

TERMINAL FLOWER 2 (TFL2; also known as LIKE HETEROCHROMATIN PROTEIN 1,LHP1) is probably also an Arabidopsis PRC1-like factor and is homologous to Heterochromatin Protein 1 (HP1) in metazoans and yeast, a protein that plays important roles in chromatin packaging and gene silencing(Gaudin et al., 2001; Kotake et al., 2003). TFL2 is expressed in proliferating cells, including those of FMs, and the encoded TFL2 protein preferentially binds to chromatin marked with H3K27me3 in vivo (Zhang et al.,2007). TFL2 is directly associated with the regulatory sequences of a group of floral homeotic genes, such as AP3, PISTILLATA(PI), AG and SEP3, and suppresses their expression during vegetative growth (Kotake et al.,2003; Turck et al.,2007; Zhang et al.,2007). The protein interaction between TFL2 and INCURVATA2, a DNA polymerase subunit probably involved in DNA replication(Barrero et al., 2007),indicates a role for the replication machinery in the maintenance of gene silencing.

It is noteworthy that almost all of the above-mentioned chromatin regulators are ubiquitously expressed in Arabidopsis. Thus, an important question is how they are specifically regulated to permit the onset of floral homeotic gene expression in FMs at stage 3. A recent study has revealed that the orchestrated repression of SEP3 by SVP, AGL24 and SOC1 is mediated by recruiting two interacting chromatin regulators, TFL2 and SAP18, a member of the SIN3 histone deacetylase complex(Liu et al., 2009). The downregulation of AGL24, SVP and SOC1 in FMs disrupts the histone-modification function of TFL2 and SAP18 at the SEP3 locus,thus derepressing SEP3, which in turn contributes to the activation of other floral homeotic genes such as AP3, PI and AG. This finding suggests that the developmental specificity of chromatin regulators could be achieved by regulating the levels of their interacting transcription factors.

Over the past few years, an ever-expanding list of regulators has emerged that form regulatory hierarchies to govern the successive developmental programmes involved in FM specification. The translation of flowering signals into actual flower formation requires the coordinated control of these regulatory hierarchies by certain common factors. In Arabidopsis, the FM identity genes AP1 and LFY, and the flowering-time genes AGL24, SVP and SOC1, participate in this control by linking the molecular events that regulate their activity with the regulation of their downstream targets. Investigating the homologues of these Arabidopsisgenes in other plant species has unravelled the functional conservation and divergence of their counterparts in governing the specification of FMs. These advances in understanding FM specification have, however, raised some additional questions to which answers are still outstanding. For example,although we know that auxin contributes to FM initiation, it is unclear how flowering-time genes affect the auxin pathway to trigger FM formation, and how the auxin pathway interacts with known FM regulators, such as LFY and AP1, to specify FM identity. In addition, the biological significance of FM polarity regulation remains to be elucidated. Addressing these questions by comprehensive molecular, genetic and biochemical approaches will greatly contribute to our understanding of the combinatorial control of FM specification.

Two recent studies (Wang et al.,2009; Yamaguchi et al.,2009) provide evidence for the involvement of miRNA-regulated SQUAMOSA PROMOTER BINDING PROTEIN-LIKE transcription factors in the regulation of flowering.