Flowering on time: genes that regulate the floral transition

Introduction

Successful sexual reproduction in plants depends on the recognition of favourable environmental conditions and the integration of that information with endogenous developmental cues. Flowering in higher plants involves the transition of a vegetative meristem, producing leaves and stems, into a floral meristem, producing flowers. Most of our understanding of the regulation of the floral transition has been developed in the past 10 years using molecular genetic approaches in the model plant Arabidopsis thaliana, a rosette plant in which the initiation of flowering is followed by the elongation of the main stem. The two main environmental factors that promote flowering in Arabidopsis are long days (Figure 1) and exposure to low temperatures (vernalization). Genetic and physiological analysis of flowering time in Arabidopsis has led to the identification of a large number of flowering-time genes (>80) that regulate flowering time in response to environmental and endogenous cues (reviewed by Simpson et al., 1999). Regulation occurs through a complex network of genetic pathways, with two main pathways mediating environmental responses (the long-day and vernalization pathways) and two pathways that function independently of environmental cues: the autonomous pathway, which promotes flowering under all conditions, and the gibberellin (GA) pathway, which is needed for flowering under non-inductive short-day conditions. These pathways converge in the induction of floral meristem identity genes and the floral transition (Figure 2). Many flowering-time genes have been cloned recently, and the primary aim of the EMBO workshop in Norwich, and this report, was to highlight the most recent advances in our understanding of the molecular mechanisms regulating the floral transition.

Fig. 1. Flowering in Arabidopsis is promoted by long-day conditions. Arabidopsis plants were grown under short-day conditions (left; 10 h light, 14 h dark) or long-day conditions (right; 16 h light, 8 h dark).

Fig. 2. A simple diagram showing the four major genetic pathways regulating flowering time in Arabidopsis. The two main pathways mediating environmental responses are the long-day and vernalization pathways. The two pathways thought to function independently of environmental cues are the autonomous pathway, which promotes flowering in all conditions, and the GA pathway, which is needed for flowering in non-inductive short-day conditions.

Converging flowering pathways and the FLC floral repressor

Research by the Dennis (Canberra, Australia) and Amasino (Madison, WI) laboratories in the past 2 years has shown that a number of signals controlling flowering converge in the regulation of the FLC gene, which encodes a MADS-box transcription factor and represses flowering (Michaels and Amasino, 1999). Repression of FLC is caused by a set of genes that control the autonomous pathway. FCA is an important component of the autonomous pathway and its expression is characterized by complex regulation that includes alternative splicing. FCA encodes an RNA-binding protein that contains a WW protein-interaction domain. Recent work has shown that one protein that interacts with this WW domain in FCA is the one encoded by FY, another gene placed in the autonomous pathway based on genetic analysis (G. Simpson, Dean laboratory, Norwich, UK). Other genes in this pathway repress FLC independently of FCA/FY and include the recently isolated genes FPA, which encodes RNA recognition motifs (Amasino laboratory), and FVE, which encodes a WD40 repeat protein with similarity to proteins involved in transcription repression complexes (Zapater laboratory, Madrid, Spain).

Cold treatment (vernalization) is another effective repressor of FLC. In late-flowering ecotypes, exposure to cold temperatures at the seedling stage represses FLC and accelerates flowering. This repression is maintained after the inductive treatment ceases and is only reversed in the next generation, suggesting that vernalization is a typical epigenetic event. A very important finding is that FLC expression decreased during vernalization in two vernalization-insensitive (vrn) mutants, but increased again once the plants were returned to normal temperature (T. Gendall, Dean laboratory). Therefore, the VRN genes are needed for the maintenance of the epigenetic state, but not for the establishment of the repression. This is particularly interesting in view of the fact that VRN2 encodes a nuclear-localized zinc-finger protein with similarity to polycomb-group proteins involved in the stable maintenance of the repressed state of chromatin in Drosophila and that VRN1 encodes a b3-domain putative DNA-binding protein.

Repressing flowering and altering chromatin structure

In contrast to many of the late-flowering mutants, which represent genes with relatively specific effects on flowering, early-flowering mutants show a large diversity of pleiotropic effects. This suggests that these genes may encode more general repressors with a variety of functions and that these are recruited to control flowering time. The cloning of several of these genes is in agreement with this view. ESD4 is homologous to a SUMO-specific protease (P. Reeves, Coupland laboratory, Norwich, UK). The genes corresponding to three other early-flowering mutants encode polycomb-group proteins (EMF2, R. Sung, Berkeley, CA; TFL2, K. Goto, Okayama, Japan) and a chromatin-remodelling protein (EBS, M. Piñeiro, Madrid, Spain). When this list is added to the previously described CLF (encoding a polycomb-group-like protein) and the VRN2 genes, it appears that silencing mechanisms that involve chromatin remodelling are frequently used in the control of flowering.

Among the genes that are repressed in the vegetative phase are several MADS-box genes that encode floral organ identity proteins. These genes are ectopically expressed in early-flowering mutants such as emf and clf. Furthermore, overexpression of such genes often results in early flowering, despite the fact that the floral organ identity mutants do not have altered flowering time. A teleological explanation for the control of flowering time that is consistent with this observation and that takes into account the fact that MADS proteins generally act as multimers could be that there is an advantage in regulating different stages of flowering through transcription-factor complexes involving MADS proteins, whose composition varies with developmental stage (P. Huijser, Cologne, Germany).

Responding to day length: the plant circadian clock and flowering

Plants use their circadian clock (an endogenous oscillator that regulates rhythms with a period of ∼24 h) in combination with signals from photoreceptors to sense day length and to regulate photoperiodic flowering. In the last 5 years, genetic components of the day-length sensing and response mechanism have been identified in Arabidopsis, through mutations that affect both circadian-clock activity and flowering time. These genes include CCA1 and LHY, which both encode transcription factors with single myb repeats. Their overexpression leads to arrhythmia, a hallmark of genes involved in the circadian clocks of other organisms, and to late day-length-insensitive flowering. The LHY and CCA1 proteins can physically interact by forming homo- or hetero-dimers, and phosphorylation of CCA1 is needed for its DNA-binding activity. A protein kinase, CK2, is proposed to phosphorylate CCA1. Consistent with this idea, the loss of CK2 activity resulted in a longer circadian period and late flowering, while overexpression of one of the kinase regulatory subunits causes the opposite phenotype (E. Tobin, Los Angeles, CA).

Another gene, TOC1, was also proposed as a key component of the plant clock, because in toc1 mutants all circadian rhythms tested have a shorter period, regardless of light conditions. The toc1-1 mutant allele also causes plants to flower early under short-day conditions in some ecotypes, and this has been shown to be due to their circadian defect. However, while the molecular details of the autoregulatory feedback loops that comprise the oscillators of other organisms, such as Neurospora, Drosophila, Synechococcus and the mouse have been known for some time, this has not been the case so far in plants.

At this meeting, the first molecular models for a plant-clock loop mechanism were presented (M. Yanovsky, Kay laboratory, San Diego, CA, and G. Coupland, Norwich, UK). The loop is proposed to begin in the morning with expression of the LHY and CCA1 genes, which are light- and circadian-regulated and peak at dawn. The LHY and CCA1 proteins then repress the expression of evening genes during the day. TOC1 is proposed as a key target for this repression. The repression of TOC1 by CCA1 and LHY may be direct, as these proteins bind the upstream ‘evening element’ (present in at least 31 evening-expressed genes) in the TOC1 promoter and are essential for circadian regulation of the TOC1 gene (Yanovsky). Once LHY and CCA1 levels drop, repression of TOC1 is relieved and it is expressed in the evening. Finally, the loop is completed by TOC1 activating, directly or indirectly, the expression of LHY and CCA1 (Yanovsky) (Albadini et al., 2001). Genetic support for this model comes from the observation that the cca1 lhy double mutant flowers early, has very short period rhythms that dampen rapidly in constant light conditions, and expresses many evening genes in the morning (Coupland). This model raises the possibility of the existence of multiple parallel feedback loops, with LHY and CCA1 repressing many evening-expressed genes during the day, and those, when finally expressed in the evening, positively regulating LHY and CCA1 expression.

Two major physiological models have been proposed to explain how day length regulates photoperiodic flowering. The ‘external coincidence’ model proposes that the coincident timing of light exposure (the external signal) with an internal circadian-regulated phase promotes flowering in long-day plants (such as Arabidopsis) and delays flowering in short-day plants (such as rice). The ‘internal coincidence’ model, on the other hand, proposes that two circadian rhythms coincide to promote flowering under floral-inductive conditions.

Several Arabidopsis genes that promote flowering are cyclically expressed and circadian-regulated and may correspond to the internal circadian-regulated factors predicted by the models. These include CO, whose phasing to the light period at the end of long days may be needed to activate a key target gene, FT, and hence to promote flowering (Suarez-Lopez et al., 2001). Consistent with this, a CO transcript was rephased into the light in short days in the early-flowering toc1-1 mutant (Yanovsky). Surprisingly, while CO appears to be activated by light, the CO protein is more stable in the dark than in white light, although it does accumulate to high levels in plants exposed to blue light (Coupland). Differential stability of a second floral protein, the blue light receptor CRY2, also appears to regulate flowering time (M. Koornneef, Wageningen, The Netherlands, and Chentao Lin, Los Angeles, CA). The presence of CRY2 protein under long-day conditions correlates with rapid flowering, but the protein is unstable on non-inductive short days. The exception to this is the EDI allele of CRY2, which causes early flowering on short days and increases protein stability of the photoreceptor.

Response to day length in Arabidopsis and rice

How well does the Arabidopsis model hold up in rice? Molecular characterization so far indicates that the genes controlling photoperiodic flowering in Arabidopsis also control it in rice. Great progress has recently been made in the isolation of rice genes that correspond to quantitative trait loci (QTLs) for flowering time (called ‘heading date’ in rice). Three of these QTLs, Hd1, Hd3a and Hd6 have recently been cloned and found to encode similar proteins to the Arabidopsis genes encoding CO, FT and the CK2 alpha subunit, respectively (M. Yano, Fukuoka, Japan). A rice gene similar to the Arabidopsis flowering-time gene GI has also been characterized and found to be circadian-regulated and to promote CO expression as it does in Arabidopsis (Ryousuke Hayama, Shimamoto laboratory, Ikoma, Japan). However, there must be differences in regulation of some of these flowering-time genes, since rice and Arabidopsis respond quite differently to day length. The explanation may lie in CO regulation of FT. Under long-day conditions, CO appears to repress flowering in rice via repression of FT. The opposite is true in Arabidopsis, where CO upregulates FT under long-day conditions and promotes flowering.

Hormonal regulation of flowering

Alteration in hormonal and nutrient balance has long been known to affect flowering time. However, proof that some of these compounds are indeed involved in the determination of flowering time has been obtained only in recent years.

In terms of the role of hormones, the GAs, which regulate multiple aspects of plant vegetative growth, have been shown to be essential for flowering under non-inductive short-day conditions (Wilson et al., 1992), and the mechanisms by which they influence flowering are beginning to be unravelled. Under short-day conditions, GAs activate downstream molecular targets such as the floral meristem identity gene LFY through a signalling pathway that is independent of the long-day pathway (Blazquez and Weigel, 2000). Newly identified components of the pathway include FPF1 (S. Melzer, Zurich, Switzerland) and FOG1 (M. Blázquez, Valencia, Spain), whose characterization has unveiled interactions between these long-day- and GA-dependent flowering pathways. A dominant mutation at the FOG1 locus accelerates flowering under short days, even in the absence of GAs, and increases GA sensitivity during germination and hypocotyl elongation, solidifying the role of this protein in the GA-dependent pathway. On the other hand, expression of the long-day pathway gene FT and the FLC floral repressor is altered in the fog1 mutant. This indicates that the FOG1 pathway feeds into the long-day pathway, and that the FT and FLC genes may serve as integration points for many different floral-inductive signals.

Even more exciting than the idea of a connection between the long-day- and GA-dependent flowering pathways is the hypothesis, based on physiological observations from the Bernier and other laboratories, that GAs and nutrients act through a common signalling pathway to promote flowering. Careful analysis of the concentrations of sucrose and GA₄ (the most active endogenous GA that promotes flowering in Arabidopsis) in plants growing under short-day conditions reveals that both compounds start accumulating at the apex of the plant, shortly before flowering occurs (O. Nilsson, Umeå, Sweden). Interestingly, GAs and sucrose synergistically activate the LFY promoter.

Certainly, GAs are not the only hormones involved in the control of flowering time. As reported by S. Davis (Millar laboratory, Warwick, UK), brassinosteroids (BRs) become limiting for flowering in certain genetic backgrounds. Davis reported that mutations that prevent BR synthesis or perception severely enhanced the late-flowering phenotype caused by loss of function in an autonomous pathway gene called LD, raising the possibility that BRs and GAs share signalling elements in the control of flowering.

The final players: meristem identity genes

The decision to flower involves a dramatic change in plant architecture, which is governed by the interplay of genes that promote shoot identity, such as TFL1 and TFL2, and those that promote floral identity, such as LFY and AP1. Mutations in these genes affect both processes. For instance, tfl mutations confer early flowering and premature consumption of the shoot meristem into a flower, and a reduction in LFY activity causes delayed flowering and inability to construct normal flowers. The similar phenotypes of tfl1 and tfl2 can now be easily explained, as both mutants show increased expression of FT. Similarly, AP1, but not LFY, is expressed precociously in tfl2 mutants (K. Goto, Okayama, Japan), an effect that is in agreement with the observation that AP1 is genetically downstream of FT.

Understanding how floral meristem identity genes activate flower development programmes is a key issue that will require the identification of elements that link floral identity specification with floral patterning. For instance, it will be important to determine how LFY activates floral homeotic genes in particular regions of the floral meristem if LFY itself is expressed throughout the primordium. The hypothesis that floral-patterning programmes ‘recruit’ shoot-patterning elements would be one explanation, and this idea has gained strong experimental support recently. WUS, a gene required for the organization of the central zone of the shoot meristem, acts as a coactivator of LFY in the transcriptional regulation of AG in the centre of the flower meristem (D. Weigel, San Diego, CA).

Perspectives

The molecular data reported at Norwich have strengthened the existing complex genetic model that has been developed and continuously refined in Arabidopsis. The involvement of chromatin-remodelling factors and epigenetic changes in the control of flowering time is an important emerging theme. However, many of the details of how environmental signals are integrated with endogenous regulatory systems such as the circadian clock, photoreceptors and plant hormones remain to be elucidated, as do the biochemical functions of some of the cloned genes. The rapid progress made in the monocot model rice, especially by Japanese research groups, was impressive and in fact showed that many components of the floral transition are similar to those used for the same purpose in Arabidopsis. The next important question is to what extent the complex molecular mechanisms that govern flowering in Arabidopsis can be generalized to other plant species with different reproductive habits in order to allow manipulation of reproductive traits.

The workshop was held at the John Innes Centre, Norwich, July 11–14, 2001. http://www.jic.bbsrc.ac.uk/events/embo/. Workshop Administrator: Dee Rawsthorne, John Innes Centre, Norwich Research Park, Colney, Norwich NR4 7UH, Norfolk, UK (Tel: +44 1603 450527; Fax: +44 1603 450025; E-mail: dee.rawsthorne@bbsrc.ac.uk).

The authors* and organizers of the workshop (from left to right): Jo Putterill*, Caroline Dean, Gordon Simpson, Tony Gendall, Maarten Koornneef* and Miguel Blázquez*