Seasonal and developmental timing of flowering


For correspondence (fax +608 262 3453; e-mail


The coordination of the timing of flowering with seasonal and development cues is a critical life-history trait that has been shaped by evolution to maximize reproductive success. Decades of studying many plant species have revealed several of the fascinating systems that plants have evolved to control flowering time: such as the perception of day length in leaves, which leads to the production of a mobile signal, florigen, that promotes flowering at the shoot apical meristem; the vernalization process in which exposure to prolonged cold results in meristem competence to flower; and the juvenile to adult phase transition. Arabidopsis research has contributed greatly to understanding these systems at a molecular level.


Over the past decade there has been an explosion of literature on the timing of flowering in Arabidopsis thaliana. The intent of this article is to provide an overview of how Arabidopsis research has contributed to a molecular understanding of some longstanding questions in flowering research: i.e. this is not a comprehensive review, and I apologize in advance to those whose flowering-time papers I have not had space to cite. For more comprehensive coverage of specific aspects of flowering there are many recent reviews (e.g. Turck et al., 2008; Zeevaart, 2008; Adrian et al., 2009; He, 2009; Kim et al., 2009; Michaels, 2009; Yant et al., 2009).

When to initiate flowering is obviously a critical life-history trait that has been shaped, over evolutionary time, to maximize reproductive success in a range of environments. In a broad sense, there are two types of information that plants use to provide input on when to initiate flowering. One is environmental – for example, the changes in temperature and/or day length that consistently reflect seasonal shifts. The other is endogenous. An example of a system involving endogenous cues is the juvenile to adult transition that affects many aspects of plant development, including competence to flower. Specifically, upon germination some plant species enter a juvenile phase in which they are not competent to flower, despite receiving inductive environmental cues. The juvenile phase can be short or long: many annuals become competent to flower after forming just a few leaves, whereas the juvenile phase of certain perennials can be several decades long.

The genetic and biochemical routes by which sensing such cues influences flowering are often referred to as flowering pathways: for example, the photoperiod pathway is the path to flowering caused by exposure to inductive photoperiods. The endogenous changes that permit, or in some cases cause, flowering are sometimes referred to as autonomous pathways. The term ‘autonomous’ is typically used to indicate that the pathway appears to be independent of environmental sensing.

Photoperiod and flowering

Tournois (1914) and Klebs (1918) published the first reports indicating that day length had a role in flowering. The general concept of the photoperiodic induction of flowering (photoperiodism) and the range of response types among plant species was firmly established by Garner and Allard (1920): for example, short-day plants (SDPs) flower when the night length exceeds a critical length, and long-day plants (LDPs) flower as day length increases. There are other response types as well, such as intermediate-day plants and long-short-day plants. A key finding was the demonstration by Knott (1934) that day length is perceived by leaves. Because flowering occurs in the shoot apical meristem (SAM), leaves must transmit a signal to the SAM. This signal is often referred to as florigen (Chailakhyan, 1936). Since the mid-1930s, much effort was devoted to identifying florigen: most investigations involved making extracts of induced leaves and attempting to set up bioassays that would respond to florigen. After several decades without success, the identity of florigen was sometimes referred to as the ‘Holy Grail’ of flowering research (Zeevaart, 2008). As discussed below, the ‘Holy Grail’ was found from the application of genetics and molecular biology to this question, but it is important to emphasize the question was clearly defined by classical physiological studies (Lang, 1965; Zeevaart, 1976). Work in Arabidopsis provided the final piece of this puzzle.

Before discussing the discovery of florigen and progress in understanding photoperiodism, some background information may be useful. Arabidopsis is a quantitative LDP – it flowers more rapidly in LDs, but will eventually flower in SDs (Gregory and Hussey, 1953). Mutants that affect flowering time are readily recognized: for example, mutants in which the LD-promotion of flowering might be compromised will form many more leaves from the primary SAM than their wild-type siblings when grown in LDs, and flower temporally much later (Figure 1a,b). Conversely, mutants in which the photoperiod pathway might be ‘hyperactivated’ will flower more rapidly, and with fewer leaves, than the wild type in non-inductive SD photoperiods. It is important to note that in Arabidopsis research, mutants that just change growth rate but not leaf number at flowering are not considered flowering ‘timing’ mutants; rather, a bona fide flowering-time mutant alters ‘development time’ such that the primary SAM forms more or fewer leaves than the wild type in given conditions (Koornneef et al., 1991). This developmental criterion is not always applied in flowering studies in other species, particularly in crop species in which days to flower is often the only measure of flowering differences among genotypes (e.g. Buckler et al., 2009).

Figure 1.

 Genetic differences in Arabidopsis flowering behaviors.
(a) Wild-type Wassilewskija (Ws) accession in long days (LDs), which are inductive for flowering. The primary shoot apical meristem (SAM) forms approximately eight rosette and cauline leaves before converting to flower formation.
(b) A gigantea mutant in Ws in LDs. The delay in flowering results in the formation of over 25 leaves before flowering. (Leaves formed later in Arabidopsis development are always larger, regardless of whether the delay is genetic or the result of growth in non-inductive photoperiods. Rapid flowering does not provide an opportunity for larger leaves to form.)
(c) Wild-type Columbia accession in LDs. The primary SAM forms approximately 15 rosette and cauline leaves before converting to flower formation.
(d) Columbia plant, into which an active allele of FRIGIDA has been introgressed, grown in LDs without vernalization. The SAM has formed over 60 leaves before flowering.

In some of the classic studies of Arabidopsis flowering, two classes of delayed-flowering mutants were identified (Redei, 1962; Koornneef et al., 1991). Mutants delayed only in LDs, but not in SDs, were proposed to affect a day length-sensing pathway, whereas those that were delayed relative to the wild type in both LDs and SDs were classified as affecting a constitutive promotion pathway (Redei, 1962; Koornneef et al., 1991) that has come to be referred to as an autonomous pathway (e.g. Simpson and Dean, 2002). It was also proposed, based on the flowering behavior of the mutants, that three of the genes [CONSTANS (CO), GIGANTEA (GI) and FT1] were involved in the production of a flowering promoter in LDs (Koornneef et al., 1991).

Identification of florigen

FT is now known to be florigen, and CO and GI are key players in the activation of FT expression. FT is a potent inducer of flowering: it is a small protein with some resemblance to RAF kinase inhibitors (Kardailsky et al., 1999; Kobayashi et al., 1999). CO is a zinc-finger protein that is likely to act as a transcription factor (Putterill et al., 1995), and GI is a large plant-specific protein involved in circadian clock function (Fowler et al., 1999; Park et al., 1999). CO expression is affected by the circadian clock (and thus mutants in clock components like GI affect CO expression), and CO is necessary for inducing FT expression in LDs (Suarez-Lopez et al., 2001; Yanovsky and Kay, 2002) (Figure 2). Indeed, the only role of CO may be to activate expression of FT in inductive photoperiods [and to activate an FT relative known as TWIN SISTER OF FT (TSF)] (Wigge et al., 2005; Yamaguchi et al., 2005).

Figure 2.

 Overview of flowering-time regulation in Arabidopsis.
Genes, proteins (represented as ovals), microRNAs and pathways are described in the text. Solid green or red lines with an arrow represent promotion, and those with a perpendicular bar represent repression. Components that overall promote flowering are shown in green, and those that repress flowering are shown in red.

One of the first demonstrations that CO acts in leaves to produce a signal (FT) that moves to the SAM came from the use of various promoters driving CO expression. CO was effective in promoting flowering when expressed in phloem companion cells of leaves, but SAM expression did not cause flowering (An et al., 2004; Ayre and Turgeon, 2004). On the other hand, FT causes flowering when expressed either in leaves or the SAM, and this promotion of flowering does not require CO activity (An et al., 2004). FT is normally expressed in the phloem of leaves (Takada and Goto, 2003); however, a combination of elegant genetic and molecular studies indicated that FT acts in the meristem by partnering with bZIP transcription factor FD (Abe et al., 2005; Wigge et al., 2005) (Figure 2). A direct demonstration that FT does in fact move from young leaves to the SAM was demonstrated with FT:GFP and FT:MYC fusion proteins in Arabidopsis, and in rice using an FT ortholog:GFP fusion (Corbesier et al., 2007; Jaeger and Wigge, 2007; Tamaki et al., 2007). Also, in grafts of Cucurbita moschata, movement of an FT ortholog across a graft junction in the phloem system was correlated with flowering (Lin et al., 2007). These studies provide strong evidence that the long-standing mystery of the identity of florigen is solved.

Studies on the photoperiod pathway in LDP Arabidopsis have been complemented by studies in SDP rice: these studies show how similar components such as CO/HEADING DATE 1 and FT/HEADING DATE 3a can be ‘wired’ to create a photoperiod pathway that responds to SD in rice and to LD in Arabidopsis (for a review see Turck et al., 2008). Moreover, studies in poplar show that a ‘CO/FT module’ is involved in the induction of dormancy in response to perception of changing day length (Bohlenius et al., 2006). Thus, components of the photoperiod pathway first discovered in Arabidopsis are likely to be widely used in higher plants for a range of developmental events that are under day length control. Furthermore, the recent demonstration of a role for CO in circadian outputs in the green algae Chlamydomonas reinhardtii indicates that CO-like genes may be ancient regulators of photoperiod-regulated processes that pre-date the appearance of flowering plants (Serrano et al., 2009).

There has been much progress in understanding the interface between CO transcription, photoperiod and the circadian clock (Imaizumi et al., 2005; Sawa et al., 2007; Fornara and Coupland, 2009; Imaizumi, 2009) (Figure 2). Proteins referred to as CYCLING DOF FACTORS (CDFs) exhibit circadian cycling, and bind to and repress the CO promoter. The abundance of CDFs is controlled by an F-box protein, FLAVIN-BINDING, KELCH REPEAT, F-BOX PROTEIN1 (FKF1), which is a blue-light photoreceptor that appears to be involved in the ubiquitin-mediated degradation of CDFs. The clock protein GI physically interacts with and stabilizes FKF1, promoting CDF degradation and subsequent CO expression at certain times of the circadian cycle in the proper photoperiod. This interface is likely to be conserved throughout the plant kingdom, and its discovery serves as a general paradigm in eukaryotes for how the clock can be linked to outputs. There has been tremendous progress from Arabidopsis research in understanding the molecular workings of the plant circadian clock that is not discussed above, but this progress has been recently reviewed elsewhere (Harmer, 2009; Imaizumi, 2009; Mas and Yanovsky, 2009; McClung, 2009).

Photoperiod and light quality also affect FT expression through stability of the CO protein. CRYPTOCHROME 1 (CRY1) and CRY2, which are blue-light receptors, and PHYTOCHROME A (PHYA), which is a red/far-red receptor, stabilize CO protein at the end of an LD photoperiod, but another red/far-red receptor, PHYTOCHROME B (PHYB), promotes the turnover of CO (Valverde et al., 2004). The involvement in CO turnover of the E3 ubiquitin ligase CONSTITUTIVE PHOTOMORPHOGENESIS 1 (COP1), as well as SUPPRESSOR OF PHYA-105 (SPA) family proteins, which interact with COP1, implicates regulated proteolysis via the proteasome in the control of CO activity (Valverde et al., 2004; Laubinger et al., 2006; Jang et al., 2008; Liu et al., 2008b).

After FT/florigen travels to the SAM, promotion of flowering requires interaction with a meristem-expressed transcription factor of the ‘bZIP’ family, known as FD (Figure 2). The evidence for this critical FT/FD interaction comes from both protein interaction studies as well as genetic studies showing that loss of FD suppresses the effects of overexpressing FT (Abe et al., 2005; Wigge et al., 2005). The FT/FD complex activates downstream floral meristem-identity genes such as APETALA 1 (AP1) (Abe et al., 2005; Wigge et al., 2005), and other floral promoters such as SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1 (SOC1) (Michaels et al., 2005; Yoo et al., 2005): these genes direct certain groups of cells in the flanks of the SAM (primordia) to differentiate into floral meristems. There are other pathways that lead to the increased expression of downstream genes such as AP1: for example, LEAFY (LFY), which itself promotes differentiation of primordia into floral meristems, and can be activated by a gibberellin pathway, also activates AP1 (Blazquez et al., 1998; Liljegren et al., 1999). Consistent with multiple pathways affecting the formation of flowers in the SAM, double mutants of lfy and either ft or fd have a stronger phenotype than single mutants: in the double mutants, primordia produce leaves instead of flowers (Ruiz-Garcia et al., 1997; Abe et al., 2005; Wigge et al., 2005).

Once flowering commences in an Arabidopsis SAM it is irreversible. This irreversibility seems to be to the result of multiple positive feedback loops involving genes expressed in the SAM activating the expression of each other, such as LFY and AP1, or SOC1 and AGAMOUS-LIKE 24, which are activators of LFY (Liljegren et al., 1999; Michaels et al., 2005; Sablowski, 2007; Lee et al., 2008; Liu et al., 2008a) (Figure 2). Not all plant species exhibit this irreversibility of flowering in an individual meristem: for example, in some species reversion to vegetative growth can occur if exposure to inductive photoperiods is not maintained (Zeevaart, 1985; Tooke et al., 2005).

Vernalization: ‘remembering’ winter

Vernalization is the process by which prolonged exposure to cold renders plants competent to flower (Chouard, 1960). As with photoperiodism, aspects of the molecular basis of vernalization were first revealed from studies with Arabidopsis, but there was also a wealth of physiological research that provided a foundation for subsequent genetic and molecular studies (for reviews of physiological studies see Chouard, 1960; Lang, 1965; Bernier et al., 1981).

For most plant species, a requirement for vernalization serves to prevent flowering in the fall season, and then permits flowering the following spring. In temperate climates, for example, many winter annuals and biennials occupy a niche in which they become established during the fall season to enable rapid flowering in spring. Because temperatures often fluctuate widely in the fall, using cold as a reliable environmental cue for flowering requires that plants perceive the duration of cold. For example, it is important that vernalization-requiring plants do not become vernalized by exposure to a short period of cold early in the fall season, because a short cold period followed by warm temperatures later in the fall might result in flowering as winter begins. Thus, exposure to periods of cold that are typical of a winter season in a given region are usually required for effective vernalization of plants well adapted to that region, and, as expected, the duration of cold exposure and the optimal temperature for vernalization vary among species, and within ecotypes of a given species. The photoperiod response often shows a remarkable range of adaptation as well: for example, ecotypes of Xanthium strumarium, an SDP, exhibit a wide range in the length of night required for flowering, and this corresponds to the great latitudinal range of this species (Ray and Alexander, 1966; McMillan, 1974).

Unlike photoperiod, which is perceived in leaves, cells of the SAM directly sense cold and become vernalized. This has been demonstrated in a wide range of vernalization-requiring species by two types of experiments: localized chilling and grafting (Chouard, 1960; Lang, 1965; Bernier et al., 1981). Furthermore, in many species the vernalized state is mitotically stable: i.e. once SAM cells have become vernalized they retain a ‘memory’ of a prior cold exposure and competence to flower. This was first demonstrated by Lang and Melchers working with a biennial variety of henbane (Hyoscyamus niger) that has an obligate requirement for both cold exposure followed by inductive photoperiods. Vernalized henbane plants ‘remember’ a prior cold treatment throughout a long period of growth in non-inductive photoperiods because the vernalized plants flower when shifted to inductive photoperiods (Lang, 1965). Another demonstration that the vernalized state can be mitotically stable is that vernalized plants can be regenerated from vernalized cells. This was first performed in the crucifer Lunaria biennis (Wellensiek, 1962), and is also the case in Arabidopsis (Burn et al., 1993b). However, there are also species, including crucifers, in which the vernalized state is not stable (i.e. the plants do not ‘remember’ winter), and thus a cold-promoted transition to flowering must take place during cold exposure (Bernier et al., 1981; Wang et al., 2009b).

The vernalization requirement in Arabidopsis

The physiological studies described above raised the question of how cold exposure renders the SAM competent to undergo the floral transition. Part of the answer to this question in Arabidopsis came from studies of the genetic basis of natural variation in flowering. Typical ‘lab strains’ of Arabidopsis flower rapidly without a vernalizing cold treatment – in fact, such types were chosen as lab models because the lack of a vernalization requirement results in a more rapid life cycle (Figure 1a,c). However, there are many Arabidopsis accessions that require a vernalizing cold treatment of several weeks to flower rapidly (Figure 1d). Analyses of segregating populations derived from crosses of vernalization-requiring and rapid-flowering types provides a way to assess the number of genes involved in the natural variation for vernalization requirement. This approach was first taken by Napp-Zinn, who found that the vernalization requirement is often caused by a dominant gene that he named FRIGIDA (FRI) (Napp-Zinn, 1987). Many years after Napp-Zinn’s natural variation studies, several groups found dominant alleles of FRI associated with a vernalization requirement in a broad range of accessions (Burn et al., 1993a; Lee et al., 1993; Clarke and Dean, 1994; Koornneef et al., 1994).

The molecular cloning of FRI demonstrated that dominant alleles of FRI encoded a full-length protein, whereas recessive alleles contained mutations that would render the protein non-functional (Johanson et al., 2000). Thus, the ancestral state of A. thaliana was vernalization-requiring, and rapid-flowering types arose independently several times from loss-of-function mutations (Johanson et al., 2000). There have been a number of studies examining the possible adaptive value of the presence or absence of an active FRI allele in different environments (e.g. McKay et al., 2003; Brock et al., 2009; Scarcelli and Kover, 2009; Wilczek et al., 2009).

Natural variation studies also revealed a second gene, FLOWERING LOCUS C (FLC), that, with FRI, confers a vernalization requirement (Koornneef et al., 1994; Lee et al., 1994). The molecular cloning of FLC, which encodes a MADS-box protein, provided insight into the molecular basis of vernalization-mediated competence of the Arabidopsis SAM (Michaels and Amasino, 1999; Sheldon et al., 1999). Overexpression of FLC, even in the absence of a functional FRI allele, prevents flowering, demonstrating that FLC is a potent suppressor of flowering. The block to flowering in the fall season in vernalization-requiring strains of Arabidopsis is largely a result of the FRI-mediated elevation of FLC expression to a level that suppresses flowering. Vernalization turns off FLC expression, and thereby creates an SAM that is competent to undergo the floral transition. Vernalization has no effect on FRI expression (Michaels et al., 2004); rather, as discussed below, vernalization appears to render FLC insensitive to the activating effects of FRI.

FLC suppresses flowering, at least in part, by repressing the expression of the floral activators SOC1, FT and FD (Figure 2): FLC has been shown to bind directly to these activators (Hepworth et al., 2002; Helliwell et al., 2006; Searle et al., 2006). There is much redundancy in the genome/proteome of most plant species, and in Arabidopsis there is a clade of FLC relatives in which most members have been shown to be flowering repressors (Scortecci et al., 2001; Ratcliffe et al., 2003). As noted above, there is an FT relative, TSF, that appears to be redundant with FT. In most situations, FLC seems to be the major repressor of flowering, and FT seems to be the major contributor to florigen, but it is likely that other FLC relatives contribute to the repression of SOC1 and FT, and that FLC and its relatives also directly repress the FT relative TSF (Michaels et al., 2005; Yamaguchi et al., 2005).

MADS-domain proteins often work in multimeric complexes with other MADS-domain proteins. FLC (and perhaps other members of the FLC clade) appears to interact with the MADS-domain protein SHORT VEGETATIVE PHASE (SVP). SVP binds to FT and SOC1– the same targets as FLC (Lee et al., 2007; Li et al., 2008). The SVP–FLC interaction is important for flowering, as loss of SVP partially suppresses the ability of FLC to inhibit flowering (Fujiwara et al., 2008; Li et al., 2008). That loss of SVP causes only partial suppression of the effect of FLC indicates that there may be other redundant FLC partners. The level of FLC expression, but not that of SVP, plays a key role in flowering-time control. FLC mRNA levels change in response to environmental cues like vernalizing cold, whereas SVP levels do not, and increasing FLC alone is sufficient to cause extremely late flowering (Michaels and Amasino, 2001), indicating that the levels of potential FLC binding partners are not limiting.

As noted above, allelic variation at FRI is a major determinant of natural variation for the vernalization requirement. FRI is a member of a small family of proteins found only in the plant kingdom, and is located in the nucleus (Johanson et al., 2000). In Arabidopsis, two other members of the FRI family [FRI-LIKE 1 (FRL1) and FRL2] are required for FLC to be expressed to levels that create a vernalization requirement (Michaels et al., 2004; Schlappi, 2006). That the FRI family is found in all sequenced higher plant genomes, regardless of whether the species has a vernalization response, indicates that members of the family are likely to play roles in aspects of plant growth and development other than flowering. However, FRI itself appears to specifically affect FLC expression: in microarray studies the major gene that appears to be upregulated by the presence of FRI is FLC (see

Studies of the molecular components required, along with FRI and FRI family members, for active FLC expression in the fall have identified a range of genes, most of which, perhaps not surprisingly, encode components of chromatin-modifying complexes involved in covalent modifying histone residues to promote active chromatin. That many of these genes are involved in chromatin modification is inferred from prior studies of homologous genes in other species, and from studies of the changes in the spectrum of modifications at FLC chromatin when such genes are mutated. In brief, these studies have revealed the requirement for methylation of histone 3 at lysines 4 and 36 by the Arabidopsis equivalent of the RNA Polymerase-II Associated Factor-1 (PAF1) complex, mono-ubiquitination of histone 2B at lysine 123 by a RAD6-BRE1-type complex, and H2A.Z deposition by a SWR1/SRCAP-type complex. Several recent reviews discuss the components of these complexes (He, 2009; Kim et al., 2009). This research direction has provided much insight into how chromatin modifications are used in developmental contexts in plants. Not surprisingly, FRI affects the state of FLC chromatin (Doyle and Amasino, 2009; Jiang et al., 2009), but whether or not FRI is involved directly in interactions with chromatin-modifying complexes remains to be determined.

In addition to the FRI family and chromatin-modifying complexes, other components required for active FLC expression have been identified, such as the plant-specific zinc-finger proteins FRIGIDA-ESSENTIAL 1 (FES1) (Schmitz et al., 2005) and SERRATE (SE), and the mRNA cap-binding complex (Bezerra et al., 2004). FES1 appears to be specific to flowering (Schmitz et al., 2005), whereas lesions in SE or genes encoding mRNA cap-binding complex subunits have pleiotropic phenotypes (e.g. Prigge and Wagner, 2001; Bezerra et al., 2004), as expected for proteins involved in mRNA processing as well as the production of microRNAs (Gregory et al., 2008; Laubinger et al., 2008). That loss of SE or the mRNA cap-binding complex results in lower levels of FLC mRNA could be because FLC RNA processing is quite sensitive to a lack of these factors (Bezerra et al., 2004). A recent report indicates that FRI interacts with the mRNA cap-binding complex, and may elevate FLC expression through a co-transcriptional mechanism linking FLC transcription with RNA processing (Geraldo et al., 2009).

A cold-mediated epigenetic switch

As discussed above, the vernalized state is mitotically stable in many plant species, including Arabidopsis. Although the definition of an epigenetic switch is not universally agreed upon, one could consider the vernalization-mediated repression of FLC epigenetic in the following sense: FLC is repressed in response to a transient signal (cold), and the ‘off switch’ is maintained through mitotic cell divisions in the absence of the cold signal (e.g. Amasino, 2004).

The mechanism by which vernalization results in the mitotically stable repression of FLC has been extensively reviewed (Adrian et al., 2009; Greenup et al., 2009; He, 2009; Kim et al., 2009; Michaels, 2009), and thus only an overview is provided below. First, as with other aspects of flowering, much of the current understanding has genetics at its foundation: in this case screens for mutants in which vernalization is impaired. The first ‘vernalization gene’ reported was VERNALIZATION 2 (VRN2), which encodes an Arabidopsis homolog of the Drosophila protein SUPPRESSOR OF ZESTE 12 (Gendall et al., 2001). SUPPRESSOR OF ZESTE 12 is part of a Polycomb repression complex that modifies chromatin (Muller et al., 2002). In vrn2 mutants, FLC expression decreases during long cold exposure, but the repression of FLC is not maintained. Although chromatin modification was a candidate for the epigenetic nature of FLC repression, the identification of VRN2 was the first indication that this was likely to be the case. Subsequently, using chromatin immunoprecipitation (ChIP), it was indeed shown that during and after vernalization the levels of two repressive histone modifications, histone H3 Lys 9 (H3K9) and H3K27 methylation, increase at FLC chromatin (Bastow et al., 2004; Sung and Amasino, 2004).

A mutant screen also revealed a gene, VERNALIZATION-INSENSITIVE 3 (VIN3), that is only expressed during the prolonged cold required for vernalization: i.e. VIN3 has a ‘vernalization-specific’ expression pattern (Sung and Amasino, 2004). In vin3 mutants, vernalization-mediated changes in FLC chromatin, such as H3 deacetylation, and H3K9 and H3K27 methylation, do not occur (Sung and Amasino, 2004). Although the cold induction of VIN3 is necessary for the vernalization process, overexpression of VIN3 is not sufficient to cause vernalization (Sung and Amasino, 2004). VIN3 contains a plant homeodomain (PHD) domain: PHD domains can bind to specific histone modifications and are often found in proteins involved in chromatin modifications in a range of eukaryotes (Mellor, 2006). There is a family of VIN3-like proteins in Arabidopsis, and certain family members can dimerize via interactions of their carboxy-terminal domains (Sung et al., 2006; Greb et al., 2007). Another member of the family that can dimerize with VIN3, VERNALIZATION 5 (VRN5)/VIN3-LIKE 1 (VIL1), is also required for vernalization, and for chromatin changes at FLC (Sung et al., 2006; Greb et al., 2007).

VRN2, VIN3 and VRN5/VIL1 appear to be part of a repressive complex that is similar to the Polycomb repression complex 2 (PRC2), which has been described in a range of eukaryotes (Wood et al., 2006; De Lucia et al., 2008). PRC2 adds methyl groups to H3 at K27 residues (H3K27me3) via the enzyme activity of the ENHANCER of ZESTE [E(Z)] subunit (Muller et al., 2002). The complex with VIN3 and VIL1/VRN5 has been shown to contain at least two Arabidopsis E(Z) homologs: CURLY LEAF (CLF) and SWINGER (SWN) (Wood et al., 2006; De Lucia et al., 2008).

The H3K27 methylation activity of PRC2 complexes in animals is enhanced by an association with PHD-domain proteins (Cao et al., 2008; Sarma et al., 2008), and recent work indicates that during vernalization VIN3 and VRN5/VIL1 are likely to serve in that enhancing role in Arabidopsis, with respect to activity towards FLC (De Lucia et al., 2008). Thus, the molecular events that initiate FLC repression during vernalization include the induction of VIN3 by cold, and probably the enhancement of the repressive activity of a PRC2-like complex at FLC by a dimer of VIN3 with VRN5/VIL1 (De Lucia et al., 2008). After cold exposure, VIN3 expression ceases (Sung and Amasino, 2004), but the levels of H3K27me3 at FLC continue to increase (Finnegan and Dennis, 2007; De Lucia et al., 2008), and the association of VRN5/VIL1 with FLC chromatin becomes widespread (De Lucia et al., 2008).

Mitotically stable maintenance of repression requires a mechanism to maintain the repressed state through rounds of DNA replication. A general maintenance mechanism is the establishment of feedback loops in which chromatin-modifying complexes recognize certain modifications, and maintain those modifications through cycles of DNA replication. PHD domains can bind to specific chromatin modifications, and thus PHD-domain proteins could ‘read’ the histone code (Mellor, 2006). It is tempting to speculate that the PHD domain proteins involved in vernalization, VIN3 and VRN5/VIL1, might enhance Polycomb-mediated repression through interactions with modified histones at the FLC locus (De Lucia et al., 2008; Kim et al., 2009). Because VIN3 is transiently expressed only during cold and is required for all vernalization-mediated chromatin modifications, it may be part of the system that, along with VRN5/VIL1 (Sung et al., 2006; Greb et al., 2007), initiates the switch of the FLC expression state. VRN5/VIL1, which is constitutively expressed, remains associated with FLC chromatin after cold exposure, and thus is likely to contribute to maintenance (De Lucia et al., 2008). Of course the chromatin changes at FLC are likely to be a progression with no clear demarcation between the initiation and progression of gene repression (e.g. De Lucia et al., 2008; Kim et al., 2009).

Certain chromatin-modifying complexes and chromatin modifications do, however, appear to be primarily involved in maintenance. As discussed above, H3K9 methylation is an additional repressive chromatin mark that accumulates at FLC as a result of vernalization-mediated repression, and appears to be involved in maintenance (Bastow et al., 2004; Sung and Amasino, 2004; Sung et al., 2006; Greb et al., 2007). For example, in vrn1 mutants, FLC is repressed during cold and H3K27 methylation increases during cold. However, H3K9 methylation does not increase in vrn1 mutants, and FLC silencing cannot be maintained after plants return to warm conditions (Levy et al., 2002; Bastow et al., 2004; Sung and Amasino, 2004). VRN1 is a plant-specific protein that contains DNA-binding domains (Levy et al., 2002), and it is likely to be part of a feedback loop that maintains H3K9 methylation. Another likely player in an H3K9 methylation feedback loop is LIKE HETERCHROMATIN PROTEIN 1 (LHP1)/TERMINAL FLOWER 2 (TFL2). A related protein from animals, HP1, binds to chromatin by interacting with methylated H3K9, and is involved in gene silencing (e.g. Maison and Almouzni, 2004). In lhp1 mutants, FLC repression also cannot be maintained after plants resume growth in warm conditions (Mylne et al., 2006; Sung et al., 2006), and H3K9 trimethylation only transiently increases during cold (Sung et al., 2006).

In animals, maintenance of repression that initiates with PRC2 typically requires another Polycomb complexL Polycomb repression complex 1 (PRC1). PRC1 binds to H3K27me3 and helps to maintain that histone modification (e.g. Muller et al., 2002). However, the Arabidopsis genome does not contain genes encoding PRC1 components (e.g. Schubert et al., 2005), and, as discussed above, LHP1 and VRN1 may serve in the analogous role for ‘locking in’FLC repression. It is intriguing that the creation of a stable silenced state at FLC appears to require PRC2-mediated H3K27 methylation, as well as H3K9 methylation. LHP1 has been shown to bind H3K27me3 and H3K9me3 in vitro, and, in broad genome studies, to be associated with regions of the Arabidopsis genome enriched in H3K27me3, but not with regions enriched in H3K9me3 (Turck et al., 2007). However, at certain regulated loci such as FLC, ‘crosstalk’ between H3K9 and H3K27 methylation that involves LHP1 and VRN1 may be necessary for mitotically stable repression. Although LHP1 and VRN1 are required to maintain H3K9 methylation, these proteins are not methyltransferases, and the actual enzymes that carry out this modification at FLC remain to be determined.

A major question is how silencing is targeted to FLC. There is no evidence that VIN3 has the ability to recognize a specific DNA sequence, and, as noted above, VIN3 expression alone is not sufficient to initiate FLC silencing. Recently, it was found that the level of FLC antisense transcripts increases substantially and rapidly during cold exposure (Swiezewski et al., 2009). Perhaps the increase in FLC antisense transcripts is a key initiating event in FLC silencing: for example, the increase in antisense transcripts might be necessary for the increased activity of a VIN3-containing Polycomb complex at FLC. However, there is not yet direct evidence for a role for the antisense transcripts in vernalization, such as a demonstration that eliminating the production of antisense transcripts abrogates FLC silencing.

It is also not known how FLC is reset to the ‘on state’ as it is passed to the next generation. Because germ cells are not set aside in plant development as they are in animals, meristem cells in which FLC has been repressed by vernalization produce the next generation in which FLC becomes active. Although there have been reports investigating when in development the resetting of FLC occurs (Sheldon et al., 2008; Choi et al., 2009), the mechanism by which this occurs has not yet been identified.

A final area of vernalization that is ripe for exploration is the mechanism of long-term cold sensing. At present, nothing is known about the mechanism of cold sensing during vernalization in Arabidopsis, or in any other plant species. In Arabidopsis, the induction of VIN3 and increased FLC antisense transcription are formally outputs of a cold-sensing system, but there have not been any reports of genetic variation in these outputs that might provide an entrée to a better understanding of cold sensing.

Vernalization in Arabidopsis: a paradigm?

As discussed above, because of the ancient nature of the photoperiod pathway, progress in understanding the molecular basis of how photoperiod affects flowering in Arabidopsis has provided a molecular foundation for photoperiod sensing in all flowering plants. This is not likely to be the case for vernalization. Flowering plants diversified a few hundred million years ago (Soltis et al., 2008) during an era, and in regions, in which it is likely that temperate-type winters were not encountered (i.e. when the climate overall was much warmer and the continents had not drifted to their present locations). Subsequently, continental drift, climate change and radiation resulted in new environments in which a vernalization response would have had adaptive value.

That vernalization pathways arose by convergent evolution is supported by studies in cereals (for reviews see Distelfeld et al., 2009; Greenup et al., 2009; Kim et al., 2009). For example, although the vernalization requirement in both Arabidopsis and cereals is to the result of a repressor of flowering that is turned off during cold exposure, the repressors are unrelated proteins. As discussed above, the Arabidopsis repressor FLC is a MADS-domain protein, whereas in cereals the repressor (VRN2) is a zinc-finger protein that has no ortholog in the Arabidopsis genome (Yan et al., 2004). There are other differences in the ‘circuitry’ of vernalization between Arabidopsis and cereals as well (e.g. Distelfeld et al., 2009; Greenup et al., 2009; Kim et al., 2009). As expected, however, in both Arabidopsis and cereals the vernalization pathway appears to be superimposed onto a more ancient and conserved photoperiod pathway. Thus, Arabidopsis has served as a paradigm for how a vernalization pathway can be constructed, and how such a pathway can interface with other flowering pathways (Figure 2). Other long-term cold responses, such as the breaking of bud dormancy by cold exposure, are also likely to have arisen independently in different groups of plants, and it will be interesting to explore how vernalization and bud dormancy systems evolved in a range of plant groups.

Autonomous flowering

As noted in the Introduction, the term ‘autonomous’ is typically used in flowering for pathways that are independent of environmental cues. In many plant species, for example, flowering cannot occur until a transition is made from the juvenile to adult phase (e.g. Poethig, 2003): this transition is an example of an autonomous pathway that results in competence to flower.

In the Arabidopsis flowering literature, a specific autonomous pathway has been operationally defined in genes that when mutated cause delayed flowering in both LDs and SDs (as opposed to the photoperiod pathway, which was defined by mutations that affect flowering only in LDs) (e.g. Simpson and Dean, 2002). Such autonomous-pathway mutants have arisen in screens for delayed flowering in ‘lab strains’ of Arabidopsis, which, as discussed above, do not have an active allele of FRI, and flower rapidly without a vernalizing cold treatment. However, mutations in autonomous-pathway genes in such lab strains create a robust vernalization response similar to that of FRI-containing lines (Martinez-Zapater and Somerville, 1990; Koornneef et al., 1991; Simpson and Dean, 2002). Furthermore, as described earlier for FRI-containing lines, the delayed flowering of these autonomous-pathway mutants is caused by elevated levels of FLC expression, and the robust vernalization response is caused by the repression of FLC (Michaels and Amasino, 2001).

Several autonomous-pathway genes that fit the above definition, namely FCA, FLOWERING LOCUS K HOMOLOGY DOMAIN (FLK), FPA and FY, have been shown to, or are predicted to, encode proteins involved in RNA metabolism (Macknight et al., 1997; Schomburg et al., 2001; Simpson et al., 2003; Lim et al., 2004; Manzano et al., 2009), and it is possible that the autonomous-pathway gene LUMINIDEPENDENS (LD), which encodes a homeodomain-containing protein, is also involved in RNA metabolism, as homeodomain-containing proteins have been shown to bind RNA (Chan and Struhl, 1997). Other autonomous-pathway genes encode components of chromatin-remodeling complexes (for reviews see He, 2009; Kim et al., 2009; Michaels, 2009). RELATIVE OF EARLY FLOWERING 6 (REF6) encodes a jumonji protein belonging to a certain class of histone demethylases (Noh et al., 2004; Agger et al., 2008). FLOWERING LOCUS D (FLD) encodes a protein belonging to a different class of histone demethylases that were first found in histone deacetylase (HDAC) complexes in other systems (He et al., 2003; Shi et al., 2004). FVE encodes a member of an MSI1-like protein family (Ausin et al., 2004; Kim et al., 2004); MSI1s are conserved WD-repeat proteins found in several chromatin-modifying complexes in eukaryotes, including HDAC complexes (Hennig et al., 2005). Histone acetylation patterns at FLC are perturbed in fve and fld mutants (He et al., 2003; Ausin et al., 2004; Kim et al., 2004).

In a recent series of papers, certain histone-modifying protein arginine methyltransferases (PRMT4a, PRMT4b, AtPRMT5 and AtPRMT10) have also been shown to be autonomous-pathway genes involved in FLC repression (Wang et al., 2007; Niu et al., 2008; Schmitz et al., 2008). In prmt mutants, FLC expression is elevated, and the level of histone arginine methylation at FLC chromatin is reduced. Interestingly, PRMT5 also appears to be required for FLC silencing during vernalization in an FRI backgound, but is not required for vernalization-mediated silencing in the absence of FRI (Schmitz et al., 2008).

It is important to note that the autonomous-pathway genes discussed above are involved in more than FLC regulation and flowering-time control. One of the first indications of this was the demonstration that a double mutant between a weak fy allele and fpa was lethal (Koornneef et al., 1998), and the later demonstration that a null allele of fy is lethal (Henderson et al., 2005). More recently it has been shown that double mutant combinations of many autonomous-pathway mutants have severe pleiotropic phenotypes (Veley and Michaels, 2008). Furthermore, many autonomous-pathway genes have recently been shown to play a broad role in RNA-mediated gene silencing, and in the expression of certain transposons and their DNA methylation patterns (Baurle et al., 2007; Baurle and Dean, 2008; Veley and Michaels, 2008). Another connection between autonomous-pathway effects on FLC expression and RNA-mediated gene silencing is the observation that double mutants of DICER-LIKE 1 and DICER-LIKE 3 (dcl1 dcl3; dicers are involved in small RNA formation) exhibit FLC-dependent delayed flowering that is vernalization responsive: i.e. the dcl1 dcl3 double mutant has a phenotype similar to typical autonomous-pathway mutants (Schmitz et al., 2007). Thus, small RNAs are likely to be involved in FLC repression, perhaps by guiding chromatin-modifying complexes to FLC chromatin (e.g. Baurle and Dean, 2008).

In summary, the ‘classical’ Arabidopsis autonomous pathway does not appear to be a linear pathway, but rather represents a collection of genes that are: (i) broadly involved in repression of gene expression, and (ii) participate in setting the ‘basal’ levels of FLC expression.

Recently, an additional flowering pathway has been described that is ‘autonomous’ in the sense that it represents a developmental pathway, parts of which are independent of environmental variables (for reviews see Fornara and Coupland, 2009; Poethig, 2009; Yant et al., 2009). Studies of the juvenile to adult phase transition revealed that the role of a particular microRNA, miR156, is to promote the juvenile phase in Arabidopsis and maize, and one consequence of this promotion is to prevent precocious flowering (Wu and Poethig, 2006; Chuck et al., 2007). Conversely, expression of another microRNA, miR172, promotes flowering, at least in part, by repressing APETALA 2-like repressors of FT (Figure 2) (Aukerman and Sakai, 2003; Jung et al., 2007; Mathieu et al., 2009). The levels of miR156 and miR172 exhibit contrasting age/development-specific expression patterns: miR156 levels decline during development whereas miR172 levels correspondingly increase (e.g. Chuck et al., 2007; Wu et al., 2009). In Arabidopsis, the age/development-specific expression patterns of these microRNAs are independent of day length: i.e. they are under autonomous control (Wang et al., 2009a). Recently, it has been shown that there is a regulatory circuit consisting of these microRNAs and transcription factors of the SQUAMOSA PROMOTER BINDING-LIKE (SPL) class: part of this circuit consists of miR156 repressing expression of certain SPL genes that are positive regulators of miR172 expression (Wu et al., 2009). Some miR156-repressed SPL genes are also positive regulators of other flowering promoters (Figure 2) (Cardon et al., 1999; Wang et al., 2009a; Yamaguchi et al., 2009). Thus, as development proceeds, the decline of miR156 levels, and the increase in levels of miR172 and certain SPLs, leads to the activation of FT in leaves, and the increase in SPLs in the meristem leads to the activation of several genes that promote the floral transition, including SOC1, a SOC1 homolog AGAMOUS-LIKE 42 and FUL (Wang et al., 2009a), as well as the meristem-identity genes LFY and AP1 (Yamaguchi et al., 2009). In contrast to the particular autonomous pathway that sets basal levels of FLC expression and appears to be specific to a subset of flowering plants, the mir156/miR172/SPL system is likely to represent an autonomous flowering pathway that is conserved in flowering plants, as it is present in maize, rice and Arabidopsis (e.g. Poethig, 2009).


  • 1

    Several Arabidopsis ‘flowering-time genes’ do not have ‘full’ names, only two or three letter abbreviations (see Koornneef et al., 1991). Over time, FT has come to be called FLOWERING LOCUS T, but others such as FCA, FD, FPA, FVE and FY lack full names.


I thank Thomas Ream, Amanda Wollenberg and Sibum Sung for helpful comments, and I thank Amanda Wollenberg for assistance in producing Figure 2. I also thank the College of Agricultural and Life Sciences and the Graduate School of the University of Wisconsin for their support of our research infrastructure. Our work on flowering has been supported by the National Institutes of Health, the National Science Foundation, the Korean Global Research Laboratory Program and the US Department of Agriculture National Research Initiative Competitive Grants Program.