- © 1998 American Society of Plant Physiologists
INTRODUCTION
The general body plan of plants is established during embryogenesis, when the undifferentiated meristematic regions of root and shoot are set aside. However, much of plant development occurs postembryonically, through the reiterative production of organ primordia at the shoot apical meristem (SAM). In most species, the SAM initially gives rise to vegetative organs such as leaves, but at some point the SAM makes the transition to reproductive development and the production of flowers.
This change in the developmental fate of primordia initiated at the SAM is controlled by environmental and endogenous signals (Bernier, 1988; McDaniel et al., 1992). However, unlike many developmental transitions in animals, the SAM of plants is not irreversibly “committed” to reproductive development once flowering commences. In some species and genotypes under certain environmental conditions, leafy shoots are formed after flowers in a phenomenon known as inflorescence reversion (see, e.g., Battey and Lyndon, 1990; Pouteau et al., 1997). This observation implies that the genes and processes involved in the transition to flowering are required to both initiate and maintain reproductive development.
Because many species must reach a certain age or size before they can flower, the vegetative meristem is thought to first pass through a “juvenile” phase in which it is incompetent to respond to internal or external signals that would trigger flowering in an “adult” meristem. The acquisition of reproductive competence is often marked by changes in the morphology or physiology of vegetative structures—leaf shape offers one example—in a process known as vegetative phase change (Poethig, 1990; Lawson and Poethig, 1995). It is likely that some of the genes identified as important in controlling the transition from vegetative to reproductive development are also involved in vegetative phase change.
In some species, the timing of flowering is primarily influenced by environmental factors, which serve to communicate the time of year and/or growth conditions favorable for sexual reproduction and seed maturation. These factors include photoperiod (i.e., day length), light quality (spectral composition), light quantity (photon flux density), vernalization (exposure to a long period of cold), and nutrient and water availability. Other species are less sensitive to environmental variables and appear to flower in response to internal cues such as plant size or number of vegetative nodes. Flowering can also be induced by stresses such as nutrient deficiency, drought, and overcrowding. This response enables the plant to produce seeds, which are much more likely to survive the stress than is the plant itself.
Over the years, physiological studies have led to three models for the control of flowering time (reviewed in Bernier, 1988; Thomas and Vince-Prue, 1997). The florigen concept (reviewed in Lang, 1952; Evans, 1971) was based on the transmissibility of substances or signals across grafts between reproductive “donor” shoots and vegetative “recipients.” It was proposed that florigen, a flower-promoting hormone, was produced in leaves under favorable photoperiods and transported to the shoot apex in the phloem. The identification of a graft-transmissible floral inhibitor also led to the concept of a competing “antiflorigen.” Many research years were consumed hunting for florigen in the phloem sap, but its chemical nature has remained elusive.
The inability to separate the hypothetical flowering hormones from assimilates led to a second model, the nutrient diversion hypothesis. This model proposed that inductive treatments result in an increase in the amount of assimilates moving to the apical meristem, which in turn induces flowering (reviewed in Sachs and Hackett, 1983; Bernier, 1988).
The view that assimilates are the only important component in directing the transition to flowering was superseded by the multifactorial control model, which proposed that a number of promoters and inhibitors, including phytohormones and assimilates, are involved in controlling the developmental transition (Bernier, 1988). According to this model, flowering can only occur when the limiting factors are present at the apex in the appropriate concentrations and at the right times. This model attempted to account for the diversity of flowering responses by proposing that different factors could be limiting for flowering in different genetic backgrounds and/or under particular environmental conditions.
Genetic analysis of flowering time in pea, cereals, and Arabidopsis supports the hypothesis that the transition to flowering is under multifactorial control (reviewed in Snape et al., 1996; Weller et al., 1997; Koornneef et al., 1998b). Indeed, multiple genes that control flowering time have been identified in all three of these species. Moreover, some of these genes act to promote flowering and others to repress it; some interact with environmental variables and others appear to act autonomously.
The most striking recent advances in our understanding of the genetic control of the timing of flowering have come from work on Arabidopsis. This area of research has been extensively reviewed (see Martínez-Zapater et al., 1994; Haughn et al., 1995; Weigel, 1995; Amasino, 1996; Aukerman and Amasino, 1996; Dennis et al., 1996; Hicks et al., 1996; Madueño et al., 1996; Peeters and Koornneef, 1996; Wilson and Dean, 1996; Coupland, 1997; Koornneef et al., 1998b; Levy and Dean, 1998; Piñeiro and Coupland, 1998), and a number of key findings have emerged. Flowering involves the sequential action of two groups of genes: those that switch the fate of the meristem from vegetative to floral (floral meristem identity genes), and those that direct the formation of the various flower parts (organ identity genes). Therefore, genes that control flowering time can be expected to interact with floral meristem identity genes, which in Arabidopsis include LEAFY (LFY), APETALA1 (AP1), CAULIFLOWER (CAL), AP2, and UNUSUAL FLORAL ORGANS (UFO). The floral meristem identity genes are themselves capable of influencing flowering time. For example, overexpression of LFY and AP1 causes early formation of determinate floral meristems (Mandel and Yanofsky, 1995; Weigel and Nilsson, 1995), whereas mutations in TFL1 affect both flowering time and meristem identity (Shannon and Meeks-Wagner, 1991). The regulation of floral meristem identity genes is under intense investigation. However, because of space constraints, this topic is covered here only briefly (for recent reviews, see Ma, 1997; Piñeiro and Coupland, 1998).
To complement earlier reviews, we describe here the current view of the control of flowering time and discuss the classic physiological studies in the context of recent molecular genetic advances. We begin by introducing the genes and mutations identified in Arabidopsis that are known to influence the timing of flowering. On the bases of their phenotypes under different growth conditions and genetic epistasis experiments, these mutants and genes are grouped into separate pathways that either promote or repress flowering. The role of DNA methylation in flowering is covered in two places to discuss separately its possible role in repression of flowering and its hypothesized role in vernalization.
In the second section, we examine the role of substances such as phytohormones that classically have been implicated in the control of flowering time and attempt to place these substances in the promotive and repressive genetic pathways. In the final section, we discuss recent data on genetic interactions that control the floral transition, and we present an updated model that attempts to summarize some of the known interactions.
GENETIC CONTROL OF FLOWERING
Arabidopsis is a facultative long-day plant; thus, long-day photoperiods are inductive, and short-day photoperiods are noninductive. The majority of Arabidopsis ecotypes are winter annuals, that is, they flower late unless they have experienced a vernalization period. This feature allows them to overwinter vegetatively and to delay flowering until favorable conditions arrive in the spring. Genes that affect flowering time in Arabidopsis have been identified through analyses of natural variation in different ecotypes and through characterization of induced mutations. The currently identified genes that are considered to play a role in flowering-time control are summarized in Figure 1 and Table 1.
Most of the genes identified by mutagenesis are derived from three rapid-cycling progenitor ecotypes: Landsberg erecta (Ler), Wassilewskija (WS), and Columbia (Col). The analysis of flowering-time variation in the naturally late-flowering ecotypes therefore complements the mutagenic approach, particularly regarding repressors of the floral transition. A number of genes—FRI, FLC, FKR, JUV, and KRY—and quantitative trait loci (QTLs) that are not represented in the mutant collections have been identified by this approach (Figure 1 and Table 1; reviewed in Koornneef et al., 1998b). Taken together, there are currently ~80 loci in Arabidopsis that are known to affect flowering time.
The response of flowering-time mutants to environmental treatments, such as vernalization and photoperiod (Table 1), combined with genetic analyses of epistasis, have established the existence of at least four pathways that control flowering time in Arabidopsis (Figure 2). Two of these pathways appear to monitor the endogenous developmental state of the plant. The floral repression pathway(s) may be a built-in mechanism that prevents flowering until the plant has reached a certain age or size, whereas the autonomous promotion pathway is believed to increasingly antagonize this repression as the plant develops. The other two pathways mediate signals from the environment: the photoperiodic promotion pathway is responsible for floral induction in response to inductive photoperiods, and the vernalization promotion pathway allows flowering to occur after experiencing an extended period of cold temperature (Figure 2).
Floral Repression Pathways
The identification of loss-of-function mutations that accelerate flowering in rapid-cycling ecotypes such as Ler reveals that even in early-flowering ecotypes, some genes act to repress flowering. Most early-flowering mutants have been categorized by their response to photoperiod (Table 1); some (e.g., clf, elf1, elf2, elg, esd4, pef1, pef2, pef3, phyB, speedy, tfl1, tfl2, and wlc) retain a response to photoperiod, whereas others (elf3, emf1, emf2, and pif) do not. Because this division is not absolute, the early-flowering mutants are considered here collectively, and the products of the corresponding wild-type genes are thought to act in repression of flowering.
Genetic Map Showing the Approximate Positions of the Genes and Quantitative Trait Loci That Affect Flowering Time in Arabidopsis.
This map, which has been updated from that shown in Koornneef et al. (1998b), shows the five chromosomes as vertical bars, with the centromeres indicated by gray ellipses. Mutant loci are given in lowercase, whereas loci identified in natural populations are given in uppercase. The QTLs were initially described in the following publications: QLN1-12, Jansen et al. (1995); QFT1-5 and QTL1-7, Koornneef et al. (1998b); FDR1-2, Mitchell-Olds (1996); RLN1-5, Clarke et al. (1995); and DFF1-2, Kowalski et al. (1994).
The EMF genes have been considered to play a major role in repression of flowering because emf1 and emf2 mutants flower with essentially no preceding vegetative phase (Sung et al., 1992; Yang et al., 1995). The EMF genes may mediate the repression of flowering via their interactions with certain floral meristem identity genes (Figure 2). For example, AP1 and AG are expressed very early in germinating emf seedlings, and constitutive expression of LFY enhances the phenotype of weak emf1 alleles, These observations suggest that the EMF genes and AP1 and AG1 reciprocally regulate each other in a negative fashion (Figure 2; Chen et al., 1997).
Some gene products that promote flowering may act, in part, by directly or indirectly repressing EMF function. For example, emf1 and emf2 are, respectively, epistatic to gi and co (two late-flowering mutants in the photoperiodic promotion pathway; Figure 2) (Yang et al., 1995). However, when the emf mutations are combined with fca and other mutations that result in late flowering, the double-mutant plants flower after they have produced an intermediate number of leaves (Haung and Yang, 1998), which suggests that the corresponding wild-type products of these genes do not act by repressing EMF function.
TFL1, another floral repressor (Table 2), was cloned recently on the basis of its similarity to its Antirrhinum ortholog CENTRORADIALIS (CEN) (Bradley et al., 1997) and by T-DNA tagging (Ohshima et al., 1997). The tfl1 mutant flowers early, and the normally indeterminate shoot apex terminates with a flower. Ordinarily, therefore, TFL1 must function to suppress flower formation at the apex and to delay the transition from vegetative to reproductive development. Consistent with this role, overexpression of TFL1 greatly extends the vegetative and inflorescence growth phases (Ratcliffe et al., 1998). It is likely that TFL1 exerts this delay in flowering by repressing the function of genes such as FCA, FVE, and FPA, which operate in the autonomous promotion pathway (Figure 2). This is because the late-flowering phenotype conferred by mutations in these genes is epistatic to tfl1 (Ruiz-García et al., 1997; T. Page and C. Dean, unpublished results).
CLF and WLC (Table 2) act to delay flowering by repressing certain floral meristem identity genes. The clf mutant expresses AG ectopically in leaves, inflorescence stems, and flowers (Goodrich et al., 1997), and wlc expresses AG and AP3 ectopically in leaves. Thus, the wild-type function of CLF and WLC is to prevent the expression of the floral meristem identity genes in vegetative tissue. The CLF gene shares sequence homology with the Drosophila polycomb group of genes, which are involved in maintaining the repression of homeotic genes (Goodrich et al., 1997). The wlc mutant displays hypomethylation of repetitive sequences associated with the centromeres (C. Hutchison and C. Dean, unpublished results); thus, reduced methylation may directly alleviate the repression of AG and AP3 expression in leaves. Similarly, induced hypomethylation resulting from constitutive expression of an antisense methyltransferase gene resulted in ectopic expression of AG and AP3 and early flowering (Finnegan, 1996). Thus, methylation may play an important role in the repression of the floral transition.
Genes and Mutations That Affect Flowering Time in Arabidopsisa
Genetic Pathways That Control Flowering Time in Arabidopsis and Proposed Interactions among Some of the Genes Involved.
The horizontal line symbolizes the vegetative (V) to floral (F) transition, with the promotive and repressive pathways exerting their influence on this switch. Four pathways are shown: repression (green), autonomous promotion (red), photoperiodic promotion under long days (LDs; dark blue) and short days (SDs; light blue), and vernalization promotion (pink). Genes that influence both floral meristem identity and flowering time are shown in black. Promotive (arrows) and repressive (T-bars) interactions are based on genetic epistasis experiments and analysis of gene expression in mutant and overexpressing lines. Not all interactions have been tested directly, and little is known about how the floral repressors interact with the various promotive pathways; thus, most of the repressors have simply been represented below the horizontal line. Therefore, this model, which is an updated combination of those published by Koornneef et al. (1998b) and Nilsson et al. (1998), does not fully represent the complexity of the interactions between genes and pathways that control flowering time in Arabidopsis.
Methylation appears to play a role in the regulation of flowering time by the FWA gene. Working with the ddm1 mutant, which has decreased DNA methylation but unaltered methyltransferase activity (Richards, 1997), Kakutani et al. (1996) noted late flowering as a frequently appearing phenotype in repeatedly self-pollinated ddm1 lines. FTS, the dominant locus conferring this late-flowering phenotype, was mapped genetically (Kakutani, 1997) and localized close to FWA, which was previously characterized by Koornneef et al. (1991) as a dominant mutation conferring late flowering. Subsequent analysis of the methylation status of the genomic region surrounding the FWA locus in ddm1 and in EMS-induced fwa alleles showed the region to be hypomethylated (Koornneef et al., 1998b). Therefore, the wild-type product of the FWA gene may encode a repressor of flowering that normally is downregulated by methylation. However, because there is precedence for local hypermethylated sites within a hypomethylated region of a gene (see, e.g., Jacobsen and Meyerowitz, 1997), it is difficult to predict whether or not FWA expression will be up- or downregulated in the fwa mutant. Ronemus et al. (1996) speculated that a general and gradual increase in methylation during development could serve to change meristem competency and determinacy as a plant ages. It will interesting to test whether such a gradient of methylation exists in Arabidopsis and whether alleviation of the autonomous repression of flowering depends, at least in part, on changes in methylation at specific loci such as FWA.
Analysis of the natural variation in flowering time has revealed that the early-flowering ecotypes such as Ler and Col can themselves be considered as mutants in genes conferring strong repression of the floral transition. Crosses between a number of winter and spring Arabidopsis ecotypes revealed that late flowering and a requirement for vernalization segregated as a dominant monogenic trait (Sanda et al., 1997) that mapped to the FRI locus (J.E. Burn et al., 1993; Lee et al., 1993; Clarke and Dean, 1994). The recent map-based cloning of FRI has revealed that Ler and Col are likely to carry loss-of-function FRI alleles (U. Johanson and C. Dean, unpublished data).
Dominant alleles at a second locus, FLC, are required for the full repression of flowering by FRI (Lee et al., 1994b; Aukerman and Amasino, 1996). Most ecotypes carry dominant alleles at FLC, but Ler and the C24 ecotype carry recessive alleles (Michaels and Amasino, 1995). Map-based cloning of FLC is nearing completion (S.D. Michaels and R.M. Amasino, personal communication), and therefore, the basis of this variation can soon be analyzed at the molecular level. Future studies will also be able to address how the vernalization promotion pathway (see below) is able to bypass the repression of flowering mediated by FRI and FLC (Figure 2).
Cloned Arabidopsis Genes That Affect Flowering Timea
Given that so many genes are involved in the regulation of flowering time in Arabidopsis, it is interesting that a major determinant of both the natural variation in flowering time and the requirement for vernalization is allelic variation at FRI. FRI maps close to one of the two major QTLs that confer a vernalization requirement in Brassica spp (Osborn et al., 1997). Thus, an important question to address in the future is whether FRI orthologs correspond to flowering-time loci in a number of plant species.
Autonomous Promotion Pathway
The identification of loss-of-function mutations that delay flowering of rapid-cycling ecotypes reveals genes that act to promote flowering. Many of these late-flowering mutants have been categorized by their response to vernalization and photoperiod and in epistasis experiments (Table 1; Koornneef et al., 1991, 1998a). One group of mutants (co, fd, fe, fha, ft, fwa, and gi) show little response to photoperiod or vernalization, and the corresponding genes are thought to act in the photoperiodic promotion pathway (Figure 2). A second group of mutants (fca, fpa, ld, fve, and fy) respond strongly to vernalization but flower even later under noninductive photoperiods. Because the products of the corresponding wild-type genes appear to promote flowering independently of photoperiod, these genes are considered to act in the autonomous promotion pathway (Figure 2). Moreover, the fact that these mutants respond to vernalization suggests that the vernalization promotive pathway acts redundantly with the autonomous promotion pathway in these early-flowering ecotypes.
Two genes of the autonomous promotion pathway encode proteins whose function may be to regulate the expression of other genes (Table 2). LD encodes a putative homeodomain protein, and although the LD transcript is expressed throughout the plant, it is most abundant in the shoot and root apices (Lee et al., 1994a; Aukerman and Amasino, 1996). FCA encodes a protein with RNA binding and protein–protein interaction domains (Macknight et al., 1997). The RNA binding domains of FCA are similar to those of the Drosophila proteins SX-1 and ELAV, which regulate alternative splicing of pre-mRNA transcripts important for sex determination and neuronal differentiation (Macknight et al., 1997). The FCA transcript is itself alternatively spliced, and increasing the levels of specific FCA transcripts results in earlier flowering (R. Macknight and C. Dean, unpublished results).
Analysis of the interaction of FCA with meristem identity genes indicates that FCA function is required for both activation and competence to respond to LFY and AP1 (T. Page and C. Dean, unpublished results). FCA, or downstream gene products, appear to act in a cell non-autonomous manner, because even in plants in which a large proportion of the two inner layers of the SAM (i.e., L2 and L3) are genotypically fca, bolting and flowering are normal (Furner et al., 1996).
Transmissible signals that promote flowering are also the focus of recent work by Colasanti et al. (1998). The maize id1 mutation confers late flowering and altered floral development. ID1 encodes a protein with zinc finger motifs, suggesting that it acts as a transcriptional regulator. Several observations led Colasanti et al. (1998) to propose that ID1 may be involved in the production or transport of a transmissible signal. For example, id1 plants do not flower under field conditions, and plants containing an increasing proportion of transposon-induced wild-type ID1 sectors in a mutant id1 background flower progressively earlier (Colasanti et al., 1998). Taken together, these experiments suggest that ID1 is required to produce and/or modulate the activity of a signal that originates in immature leaves and influences reproductive development in the SAM.
That leaves are required to determine the developmental potential of the apex has also been established using cultured maize apices. Excised apices revert to producing a full set of leaves before they produce flowers, irrespective of how many leaves had been produced before they were placed in culture (Irish and Jegla, 1997). However, leaving the four to six youngest leaf primordia on the excised apices prevents the resetting of the developmental program, indicating that some signal from the leaves influences development of the apex.
Photoperiodic Promotion Pathway
Plants detect light in at least five regions of the visible spectrum by using at least three classes of photoreceptors. Blue light and ultraviolet-A are detected by the cryptochromes, red (R) and far-red (FR) light are detected by the phytochromes, and ultraviolet-B is detected by an as-yet-unidentified photoreceptor (Thomas and Vince-Prue, 1997). In Arabidopsis, there are at least five phytochromes (PHYA to PHYE) and two cryptochromes (CRY1 and CRY2) (Thomas and Vince-Prue, 1997). These photoreceptors typically have been characterized by the effect they have on seedling morphogenesis under different light conditions. Several Arabidopsis mutants that were originally isolated based on abnormal seedling photomorphogenesis are also affected in flowering time. These include cop1, det1, det2, hy1, hy2, hy4, phyA, phyB, pef1, pef2, and pef3 (Table 1). Conversely, several mutants isolated based on their flowering-time phenotypes were subsequently found to exhibit abnormal seedling photomorphogenesis. These include elf3, elg, fha, and lhy (Table 1).
The role of photoperiod in flowering was conclusively demonstrated by Garner and Allard in the 1920s in their classic experiments with the Maryland Mammoth mutant of tobacco and the Biloxi variety of soybean (reviewed in Thomas and Vince-Prue, 1997). Recent genetic studies have begun to identify molecular components of the photoperiodic promotion pathway (Figure 2), and an overall picture of how Arabidopsis perceives and responds to inductive photoperiods is beginning to emerge.
The pathway begins with photoreceptors (such as PHYA and CRY2), which initiate signals that interact with a circadian clock and entrain the circadian rhythm. Somehow, day length is measured, and when the length of the dark period decreases below a critical length, genes that promote flowering (such as CO) are activated. This activation leads, in turn, to the upregulation of floral meristem identity genes and, thereafter, flowering.
In Arabidopsis, light quality affects flowering time, with R light inhibiting and FR light promoting flowering (Martínez-Zapater et al., 1994). The phenotype of phyB mutants (Table 1) suggests that PHYB normally plays a role in inhibiting flowering under high R to FR conditions but is not involved in day-length perception (Koornneef and Peeters, 1997). Physiological studies on multiple mutant combinations suggest that in addition to PHYB, other light-stable phytochromes also regulate flowering in response to light quality (Koornneef and Peeters, 1997). In contrast, mutations in PHYA, which encodes a light-labile photoreceptor, prevent perception of low-fluence-rate, FR-enriched day-length extensions that promote flowering. These observations suggest that PHYA is involved in both day-length perception and promotion of flowering by inductive photoperiods (Figure 2; Koornneef and Peeters, 1997).
Blue light alone promotes flowering in Arabidopsis, and the product of the FHA gene has recently been shown to encode CRY2, one of the two cryptochromes thus far identified in Arabidopsis (Guo et al., 1998). Transgenic plants overexpressing CRY2 flowered earlier than did the wild type and had increased levels of CO mRNA (Guo et al., 1998), suggesting that blue light promotes flowering via CRY2 and CO (see below). Furthermore, the level of CO mRNA was found to be reduced in cry2 mutants grown under long days but not under short days (Guo et al., 1998), thereby providing a possible explanation for the basis of the original fha late-flowering phenotype. Because the levels of both PHYA and CRY2 proteins drop rapidly and dramatically in the light (Thomas and Vince-Prue, 1997), they could fulfill the role of providing information about light/dark transitions to the circadian clock.
CRY1, the other cryptochrome in Arabidopsis, was originally identified as the hy4 mutant, which has a long hypocotyl under blue light (Table 1). hy4 is sensitive to photoperiod and is not delayed in flowering in a Ler background under white light and inductive photoperiods. However, in the presence of non-Ler alleles of FLC and in blue-enriched light, hy4 is late flowering and exhibits photoperiodic sensitivity (Bagnall et al., 1996; Koornneef and Peeters, 1997). Therefore, CRY1 is involved in the promotion of flowering, but its interaction with floral promotion pathways is unclear.
Several genes that affect photoperiodic sensitivity and that may encode components of the circadian clock itself have been identified. CCA1 and LHY RNA levels oscillate in a rhythmic fashion, and overexpression of either gene results in long hypocotyls and late flowering (Schaffer et al., 1998; Wang and Tobin, 1998). Constitutive expression of either CCA1 or LHY also abolishes or alters the circadian expression of their own transcripts as well as several other genes, which suggests that CCA1 and LHY negatively regulate their own expression (Wang and Tobin, 1998).
Another likely component of the circadian clock is TOC1, which was identified as a semi-dominant mutation that shortened the period length of the circadian clock by 2 to 3 hr (Somers et al., 1998). The toc1 mutation reduces the sensitivity of plants to photoperiod and causes early flowering under short days, indicating that quantitative changes in the pace of the circadian clock, not rhythmicity/arhythmicity alone, can alter flowering time.
ELF3 may mediate the interaction of light signals generated by the photoreceptors with the circadian clock (Figure 2). The phenotype of the elf3 mutant (Table 1) suggests that the wild-type product of this gene is involved in repressing flowering under noninductive photoperiods. However, the conditional arhythmicity of the elf3 mutant suggests that ELF3, which has recently been cloned (Table 2), does not function in the circadian clock itself (Hicks et al., 1996; Koornneef and Peeters, 1997).
The circadian clock is believed to affect the expression of downstream genes that operate in the photoperiodic promotion pathway, including CO (Table 2) (Putterill et al., 1995). CO mRNA is expressed throughout the plant and is more abundant in plants grown under long days compared with short days (Piñeiro and Coupland, 1998). GI, which has recently been cloned (Table 2), probably acts upstream of CO (Figure 2), because the phenotype of plants that overexpress CO is epistatic to the gi mutation (Piñeiro and Coupland, 1998).
Several lines of evidence suggest that the level of CO activity in Ler plants is directly correlated with flowering time (reviewed in Piñeiro and Coupland, 1998). Using a glucocorticoid-inducible system, Simon et al. (1996) demonstrated that induction of CO activity is sufficient to rapidly cause flowering under short days and to initiate transcription of LFY and TFL1 as rapidly as when these genes are induced by transfer to inductive photoperiods. However, levels of AP1 mRNA increase more slowly after CO activation than they do in response to inductive photoperiods (Simon et al., 1996). These data suggest that CO acts in a pathway that is sufficient to activate LFY and TFL1 transcription but that rapid activation of AP1 requires an additional pathway (Figure 2). Interestingly, genetic analyses by Ruiz-García et al. (1997) have placed CO and TFL1 in different genetic pathways, so the rapid activation of TFL1 transcription remains to be explained.
Vernalization Promotion Pathway
Another seasonal cue in temperate zones is a winter period, and many species require exposure of imbibed seeds or vegetative plants to a period of cold temperature (typically 2 to 8 weeks at ~4°C) in order to flower. This process, known as vernalization, is slow and quantitative but requires active metabolism (reviewed in Chouard, 1960; Vince-Prue, 1975). The site of perception of vernalization is the shoot apex (e.g., Curtis and Chang, 1930; Metzger, 1988), but all actively dividing cells, not only those at the shoot apex, may be capable of responding to vernalization (Wellensiek, 1964). Unlike photoperiodic induction, vernalization prepares the plant to flower but does not itself evoke flowering. That is, there is a clear temporal separation between cold treatment and flowering, which commonly occurs after a period of growth at warmer temperatures. Vernalization is required in each generation for winter annuals and biennials and each growth year for perennials, which suggests that meiosis or some other aspect of reproductive growth resets the requirement for vernalization.
The features of vernalization suggest that an epigenetic mechanism may be responsible for the establishment, persistence, and resetting of whatever self-perpetuating changes occur during or subsequent to exposure to cold. The observations that the flowering of late-flowering, vernalization-sensitive Arabidopsis mutants is accelerated by azacytidine treatment (J.B. Burn et al., 1993) and that cold treatment leads to specific changes in gibberellin (GA) metabolism (Hazebroek and Metzger, 1990; Hazebroek et al., 1993) led J.B. Burn et al. (1993) to propose that vernalization causes a specific reduction in cytosine methylation. This reduction, J.B. Burn et al. (1993) hypothesized, results in the activation of the gene encoding kaurenoic acid hydroxylase, an enzyme that catalyzes an early step in GA biosynthesis. Indeed, when general levels of methylation were reduced in wild-type plants by introducing a transgene expressing an antisense version of a methyltransferase gene (antisense-MET1), developmental abnormalities and early flowering were observed (Finnegan, 1996; Finnegan et al., 1998). However, the role of methylation in vernalization is still unclear, because substantial demethylation did not prevent vernalization from fully accelerating flowering in these lines, nor did it prevent resetting of the vernalization requirement in the progeny of antisense-MET1 plants (Finnegan et al., 1998).
One approach to understanding the molecular basis of vernalization has been to isolate mutants of Arabidopsis that are specifically impaired in their response to cold treatment (Chandler et al., 1996). The starting point for this genetic screen was fca, a late-flowering mutant whose phenotype can be completely corrected by a period of vernalization. fca plants were mutagenized, and a population of progeny plants were vernalized and screened for individual plants that flowered late, that is, which no longer exhibited a strong response to vernalization. Of these candidate double mutants, those that flowered no later than fca itself without cold treatment were selected for further characterization (Chandler et al., 1996). Such vrn mutants may be defective either in the perception of cold temperature or in the transduction of the cold signal by the vernalization promotion pathway (Figure 2). An initial screen identified five independent recessive vrn mutations in at least three complementation groups (Chandler et al., 1996), and a second screen identified five additional mutants, which have not yet been assigned to complementation groups (Y.Y. Levy and C. Dean, unpublished results). Two mutants, vrn1 and vrn2 (Table 1), have been characterized in some detail and are being cloned by chromosome walking. Both vrn1 and vrn2 have a normal acclimation response, indicating either that they are downstream of a cold-perception pathway common to acclimation and vernalization or that cold perception occurs via independent pathways in these two responses (Chandler et al., 1996). Analysis of the VRN genes should reveal some of the molecular components involved in promotion of flowering by vernalization.
INTEGRATING PHYSIOLOGY AND GENETICS: FLORAL SIGNALS AND GENETIC PATHWAYS
Considerable physiological analysis has led to certain compounds and processes being implicated in controlling the floral transition. These include the role of sugars, cytokinins, and GAs. In this section, we discuss the role of these substances in flowering and try to place them within the promotive and repressive pathways.
The Role of Carbohydrates in Flowering
Compelling evidence that sucrose may function in long-distance signaling during floral induction comes from studies of Sinapis alba, a long-day plant in the mustard family. After induction of flowering in S. alba by either a single long day or a displaced short day, the concentration of sucrose in the phloem reaching the apex increases rapidly and transiently (Bernier et al., 1993). Furthermore, this pulse of sucrose precedes the increase in cell division that is normally observed in the SAM upon floral induction. The sucrose reaching the apex appears to be derived from the mobilization of stored carbohydrates, most likely starch in the leaves and stems, because plants induced by a displaced short day receive the same photosynthetic input as plants maintained under noninductive photoperiods (Bernier et al., 1993).
In Arabidopsis, Ler plants grown in darkness with their apices in contact with sucrose-containing medium flower with the same number of leaves as do plants grown under long days (Roldán et al., 1997). In contrast, sucrose has a significant effect on the flowering of vernalization-requiring ecotypes Leiden and Stockholm, which flower early when grown under these conditions and with approximately the same number of leaves as Ler (Roldán et al., 1997). Furthermore, sucrose alone, whether supplied in the dark or in the light, is responsible for most of this acceleration. Therefore, supplying sucrose to these late-flowering ecotypes bypasses the inhibition of flowering normally conferred by the existence of dominant alleles at FRI and FLC (Table 1). Sucrose also accelerates the flowering of fve, fpa, fca, co, and gi but not of ft and fwa (Roldán et al., 1997). This result implies that FVE, FPA, FCA, CO, and GI function in processes that are either upstream of or separate from control of sucrose availability to the vegetative apex, whereas FT and FWA function in processes downstream of this control point.
Further genetic evidence connecting carbohydrate metabolism with control of flowering is available, but the nature of this connection is unclear. For example, there are at least five Arabidopsis mutants, adg1, cam1, gi, pgm, and sex1, which are altered in starch synthesis, accumulation, or mobilization and which flower late under some conditions (Table 1). The flowering time of cam1 and gi is not influenced by photoperiod, and therefore, both are likely to act in the photoperiodic promotion pathway (Eimert et al., 1995). pgm and sex1 mutants flower later in short days than they do in long days and so fall into the autonomous promotion pathway. Flowering of these mutants is accelerated by cold treatment, suggesting that vernalization does not depend on normal starch metabolism (Bernier et al., 1993).
Phytohormones
The role of GAs in the transition to flowering has been difficult to establish. On the one hand, there are many examples in which the abundance or composition of endogenous GAs changes under conditions that induce flowering (Pharis and King, 1985). Furthermore, because applying certain GAs can induce flowering in some species, there has been an emphasis on the study of GAs in floral initiation and in the search for florigen (reviewed in Chouard, 1960; Evans, 1971; Zeevaart, 1983; Thomas and Vince-Prue, 1997). On the other hand, applied GAs are rarely effective at inducing flowering in short-day plants. Moreover, they generally inhibit flowering of woody angiosperms, although they do promote flowering of conifers (Pharis and King, 1985). Even within long-day plants, the same GA can have a different effect in different species. For example, 2,2-dimethyl GA4 has potent florigenic activity when applied to Lolium temulentum but has no effect on flowering in S. alba (Bernier et al., 1993).
In Arabidopsis, signaling mediated by GAs appears to play a promotive role in flowering, particularly under noninductive photoperiods (Figure 2). Application of GAs accelerates flowering of wild-type plants under short days (Langridge, 1957) and of the late-flowering mutants fb, fca, fd, fe, co, fpa, ft, fve, and fwa (Table 1) under long days (Chandler and Dean, 1994). Under noninductive photoperiods, the ga1 mutant (Table 1) does not flower unless provided with GAs (Wilson et al., 1992), and the gai mutant (Table 1) flowers very late. Furthermore, spy (Table 1), a mutant considered to exhibit constitutive GA-mediated signal transduction, flowers early (Jacobsen and Olszewski, 1993), as do plants constitutively expressing FPF1, a gene that appears to be involved in GA-mediated signal transduction or responsiveness to GAs (Table 1; Kania et al., 1997).
The role of GAs in activation of the LFY promoter has recently been analyzed (Blázquez et al., 1998). The basal level of LFY promoter activity is lower in ga1 mutants, and the upregulation by long days is delayed. In contrast, LFY activity is slightly higher in a spy mutant grown in short days, correlating with an acceleration of flowering. A cauliflower mosaic virus 35S–LFY transgene was also found to rescue flowering in ga1 mutant plants in short days. Thus, GAs promote flowering in Arabidopsis at least in part by activating LFY expression. Blázquez et al. (1998) also analyzed the direct effect of GA3 with and without sucrose on LFY promoter activity. GA3 alone had no effect, sucrose produced a small increase, and both together had a synergistic effect. This requirement for two activation signals for maximal effect may account for observations with excised Lolium apices (McDaniel and Hartnett, 1996). In this study, photoperiodic induction was found to result from two signals acting at the apex. One of these signals has not been identified (but from this analysis, it is possibly sucrose), and the other is GA (McDaniel and Hartnett, 1996).
The role of GAs in vernalization has received particular attention because in some species, application of GAs to vegetatively growing plants can substitute for cold treatment (see Chouard, 1960; Lang, 1965; Evans, 1971; Zeevaart, 1983; Martínez-Zapater et al., 1994). However, in the majority of species examined, including most cereals and nonrosette plants, application of GAs is not sufficient to overcome a requirement for vernalization (Chouard, 1960; Lang, 1965; Evans, 1971; Zeevaart, 1983). Because GAs are involved in flowering processes such as floral evocation (McDaniel and Hartnett, 1996) and bolting (Metzger, 1990), which occur well after the cold treatment, it is possible that application of GAs can simply bypass vernalization completely. Consistent with this possibility is the notion that vernalization may increase the sensitivity of plants to GAs but that GAs have no direct role in the process of vernalization itself (Chouard, 1960).
Further indication that GAs may not play a role in vernalization in Arabidopsis comes from experiments with ga1-3 (Table 1), a mutant severely impaired in GA biosynthesis (Sun and Kamiya, 1994). When combined with fca, which responds strongly to vernalization, the ga1-3 fca double mutants still exhibit a robust vernalization response (J. Chandler and C. Dean, unpublished data). However, because ga1-3 plants still contain residual GAs (T.-p. Sun, personal communication; Zeevaart and Talón, 1992), this result must be interpreted with caution. In summary, the precise role of GAs in the transition to flowering is unclear. Potential tissue-specific changes in GA biosynthesis and sensitivity need to be addressed, as does the potential existence of as-yet-undiscovered florigenic GAs (for a discussion of this possibility, see Evans, 1971; Zeevaart, 1983).
GAs are not the only class of phytohormones that has been implicated in affecting the floral transition. For example, there is evidence from studies on S. alba that long-distance signaling by cytokinins might play a role in the transition to flowering in response to inductive photoperiods (reviewed in Bernier et al., 1993). As discussed above, inductive photoperiods cause the rapid and transient export of sucrose from the leaves to both the shoot and root meristems. In the root, this sucrose leads to the export of cytokinin, primarily zeatin riboside, to the shoot and leaves, presumably via the xylem. Subsequently, another cytokinin, isopentenyladenine riboside, moves out of the leaves, and some makes its way to the shoot apex, where its levels increase within 16 hr of induction (Bernier et al., 1993).
The relative importance of the cytokinin and sucrose fluxes to the floral transition in Arabidopsis remains to be established. Application of cytokinins provokes a phenotype similar to that of deetiolated 1 mutants—early flowering and severe pleiotropic effects on growth (Chory et al., 1994). emf2 has been shown to be allelic (Z.R. Sung, personal communication) to the cytokinin resistance mutant cyr1 (Deikman and Ulrich, 1995), but the apparent lack of mutations that implicate cytokinins in flowering may be due to a high degree of redundancy in the genes involved. Alternatively, the mutant phenotypes may be so pleiotropic that such mutants have not been classified as cytokinin mutants.
In addition to GAs and cytokinins, other phytohormones, such as abscisic acid (ABA), ethylene, and polyamines, may be involved in flowering under certain circumstances and in some species (Martínez-Zapater et al., 1994). The ethylene-insensitive mutant ein2 is slightly delayed in flowering, and ABA-deficient mutants flower somewhat early under noninductive photoperiods (Martínez-Zapater et al., 1994), suggesting a role for ethylene and ABA in floral promotion and repression, respectively.
GENETIC INTERACTIONS THAT CONTROL THE FLORAL TRANSITION
The genetic interactions that control the floral transition in Arabidopsis have been described in a model that is constantly updated and revised as new data become available (Figure 2; see, e.g., Schultz and Haughn, 1993; Martínez-Zapater et al., 1994; Coupland, 1995; Yang et al., 1995; Koornneef et al., 1998a). This model fits well with the multifactorial control model, which was developed on the basis of physiological analyses of flowering time (Bernier, 1988). Its essential feature is that the time at which flowering occurs is determined by antagonism between the promotive action of parallel pathways that monitor developmental age and environment and the repressive action of floral inhibitors. The promotive pathways are functionally redundant, explaining why no single mutation that prevents flowering has yet been found.
How the long-day, autonomous promotion, and GA pathways integrate to activate the meristem identity genes is one of the most active areas of research in this field. Quantitative increases in LFY expression are clearly required, with flowering occurring only after a threshold concentration of LFY has been reached (Blázquez et al., 1998). Expression of AP1 is more qualitatively linked to floral determination (Hempel et al., 1997). Unlike LFY and AGL-8, expression of AP1 is upregulated after the point of floral determination. The connection between the flowering-time genes and LFY has been directly addressed (Blázquez et al., 1998; Nilsson et al., 1998). Indeed, CO, GI, FCA, FVE, GA1, and GAI all play a role in activation of LFY (Figure 2) and are required to some extent for full expression of LFY function. In contrast, FWA, FE, and FT appear to be necessary for plants to respond to LFY expression (Nilsson et al., 1998). FT has recently been cloned independently by T-DNA tagging (Araki et al., 1998) and activation tagging (D. Weigel, personal communication); it encodes a protein with pronounced similarity to another meristem identity gene, TFL1 (Bradley et al., 1997). Despite their similarity, TFL1 and FT have opposing functions, with one repressing and the other promoting flowering.
Genetic analyses by Ruiz-García et al. (1997) have distinguished FWA and FT from the other flowering-time genes, and it has been proposed that these two genes function to activate AP1 in a pathway that runs parallel to the pathway leading to LFY activation (Figure 2). This separation of FT and FWA was also observed by Roldán et al. (1997) in their study of the sucrose-dependent acceleration of flowering in Arabidopsis flowering-time mutants (see The Role of Carbohydrates in Flowering, above). Thus, FWA and FT appear to act as intermediaries between some of the other floral promoters and floral meristem gene activation (Figure 2). How the many known floral meristem genes fit into this picture remains to be seen, but it is clear that different promotive pathways converge to redundantly activate a large set of floral meristem identity genes, which are themselves at least partially redundant in function. As stated previously, this area has been extensively reviewed recently and so is not covered in great detail here (see Figure 2; Koornneef et al., 1998b; Piñeiro and Coupland, 1998).
PERSPECTIVES
In summary, very rapid progress is being made in elucidating the molecular control of the floral transition. The next phase of the work will require the use of genetic screens designed, for example, to identify suppressors and enhancers of existing mutations. Creative genetic strategies that take advantage of the ability to constitutively express individual flowering-time genes or that use specific mutant backgrounds will help to identify both genes that operate downstream in the same pathway and genes with redundant functions. As more flowering-time genes are cloned, biochemical and cellular characterization of their products will become increasingly important. Several flowering-time genes that have already been cloned appear to encode regulators of gene expression (Table 2); identification of the upstream and downstream targets of these gene products will help to establish their regulatory role and, perhaps, to confirm genetically defined steps in the various signaling pathways.
As the genes controlling flowering time in Arabidopsis become better defined, an important question will be to address how they correspond to genes that regulate flowering time in other species. A focused effort on comparative mapping will be required to establish the potential correspondence of different genes in different species. With this goal in mind, we have assembled a list of possible orthologs from Arabidopsis, pea, sugar beet, barley, and wheat (all vernalization-responsive, quantitative, long-day plants), based on the physiological characteristics of the mutants or allelic variants and genetic dominance for late- or early-flowering phenotypes (Table 3).
Establishing correspondence among these different genes would clearly accelerate their cloning, and it would also provide useful information on gene function in Arabidopsis. The ability to combine grafting with genetic analysis in peas has provided important information on the role of the flowering-time genes. For example, the Gigas gene product is involved in the production of a graft-transmissible floral promoter, whereas the products of Late flowering and Vegetative 2 are not graft transmissible and are thought instead to alter the threshold sensitivity of the meristem to the transmissible signals. Determining whether Gigas, Late flowering, and/or Vegetative 2 correspond to FCA and/or FRI would significantly add to our understanding of the function of these Arabidopsis genes. Although gene function may have diverged during evolution, the identification of orthologs in different species would inform a working model, which could then be tested.
Possible Orthologs of Arabidopsis Flowering-Time Genesa
ACKNOWLEDGMENTS
We thank Tony Gendall and Gordon Simpson, whose comments greatly improved the manuscript. We thank our colleagues who allowed us to include information prior to publication, and we apologize to those authors whose work could not be included due to space limitations. Research in the Dean laboratory on control of flowering time in Arabidopsis is funded by grants from the Biotechnology and Biological Sciences Research Council (Grant No. PG208/0606), the European Commission (Grant No. BIO4-CT97-2340), and the Human Frontier Science Program (Grant No. RG0303/1997-M).
- Received July 2, 1998.
- Accepted October 13, 1998.
- Published December 1, 1998.