Multiple Pathways in the Decision to Flower: Enabling, Promoting, and Resetting

The Floral Repressor FLC

FLC is a MADS box transcriptional regulator, expressed predominantly in shoot and root apices (Michaels and Amasino, 2000), that acts to quantitatively repress flowering (Michaels and Amasino, 1999a; Sheldon et al., 1999) through repression of the floral pathway integrators FT, SOC1, and LFY (Nilsson et al., 1998; Kardailsky et al., 1999; Kobayashi et al., 1999; Blázquez and Weigel, 2000; Lee et al., 2000; Samach et al., 2000). FLC plays a central role in vernalization requirement and response. Arabidopsis vernalization-responsive mutants and accessions that show a strong vernalization requirement have increased FLC RNA and protein levels, whereas vernalized plants have reduced FLC levels (Sheldon et al., 2000b; Michaels and Amasino, 2001; Rouse et al., 2002). Overexpression of FLC results in very delayed flowering in Landsberg erecta (Ler), which is vernalization insensitive, but in C24 some plants flower late and others flower early (Michaels and Amasino, 1999a; Sheldon et al., 1999).

Analysis of natural variation in different Arabidopsis accessions has shown that allelic variation at FLC contributes to flowering time (Michaels and Amasino, 2000; Schlappi, 2001; Gazzani et al., 2003). In the case of Ler, this appears to be the result of altered regulation of the gene rather than of different FLC protein activities. The FLC allele in Columbia is much stronger than that of Ler, and this is correlated with the presence of a nonautonomous Mutator-like transposable element positioned in the first intron of FLC in Ler, a region known to be required for FLC regulation (Sheldon et al., 2002; Gazzani et al., 2003; Michaels et al., 2003b).

There are five close homologs of FLC in the Arabidopsis genome called MADS AFFECTING FLOWERING1 (MAF1) to MAF5 (Ratcliffe et al., 2001, 2003) (MAF1 is also called FLOWERING LOCUS M [FLM] [Scortecci et al., 2001]). MAF1 to MAF4 appear to act as floral repressors whose expression is affected only slightly, if at all, by vernalization. MAF5 expression is upregulated by vernalization, but the role it plays in determining flowering time is unclear (Ratcliffe et al., 2003).

The Floral Repressors TFL1, SVP, and TOE1/2

TFL1 encodes a protein that is similar to animal Raf kinase inhibitors and is highly homologous with FT (Bradley et al., 1997; Ohshima et al., 1997; Kardailsky et al., 1999; Kobayashi et al., 1999). In contrast to the floral promoting activity of FT, TFL1 extends the vegetative growth phase and maintains the indeterminate nature of the inflorescence (Shannon and Meeks-Wagner, 1991; Alvarez et al., 1992; Bradley et al., 1997). These activities are thought to result from TFL1 both delaying the upregulation of LFY and APETALA1 (AP1)/CAULIFLOWER (CAL) and also preventing a response to these genes in the inflorescence meristem (Bradley et al., 1997; Ratcliffe et al., 1998). Conversely, LFY and AP1/CAL inhibit TFL1 expression in the floral meristems produced on the periphery of the inflorescence meristem (Liljegren et al., 1999; Ratcliffe et al., 1999). The relative activity of TFL1 and LFY/AP1/CAL influences flowering time and the size of the inflorescence (Ratcliffe et al., 1998, 1999). Plants that overexpress TFL1 are late flowering, but flowering can be accelerated in 35S:TFL1 plants by vernalization or overexpression of FCA (O. Ratcliffe, R. Macknight, C. Dean, and D. Bradley, personal communication). This effect is not via FLC because FLC levels are unaltered in 35S:TFL1 plants. Interestingly, crosses between autonomous pathway mutants and tfl1 mutants produce plants that flower later than the autonomous pathway mutants alone (Ruiz-García et al., 1997; Page et al., 1999). This finding suggests some interaction between FLC and TFL1. Perhaps TFL1 affects the spatial expression of FLC in the meristem as it does to LFY and AP1.

SVP also acts as a floral repressor and encodes a MADS domain protein (Hartmann et al., 2000). SVP acts in a dose-dependent manner to delay flowering and does not alter the effects of photoperiod or vernalization on flowering time. svp mutations overcome the late flowering conferred by overexpression of FLM/MAF1, and svp flm double mutants behave like single mutants. Thus, FLM/MAF1 and SVP appear to function in the same floral pathway, which interacts with the photoperiod pathway (Scortecci et al., 2003).

Recently, it was shown that two AP2-like genes, TOE1 and TOE2, can act as floral repressors (Aukerman and Sakai, 2003). When overexpressed, TOE1 delays flowering, and a toe1 mutant is slightly earlier flowering than is the wild type. The toe2 mutation does not result in an early-flowering phenotype alone, but it does enhance the toe1 early-flowering phenotype. Both the TOE1 and TOE2 mRNA transcripts are targets of EARLY ACTIVATION TAGGED (EAT1), which encodes the precursor of a microRNA of the miR172 family. Overexpression of miR172 causes early flowering (Aukerman and Sakai, 2003; Chen, 2004) and does so by acting as a translational repressor of AP2-like genes (including TOE1 and TOE2). In wild-type plants, miR172 expression increases with developmental age, which makes it tempting to link the TOE1/2–miR172 interaction with a mechanism to control flowering time based on an internal measure of age (Aukerman and Sakai, 2003).

FRIGIDA Upregulates FLC

FRIGIDA (FRI) causes upregulation of FLC expression, which results in Arabidopsis accessions that carry active FRI alleles adopting a winter annual habit—that is, they overwinter vegetatively and, after the vernalization requirement is satisfied, they flower in spring (Napp-Zinn, 1961). Allelic variation at FRI is a major determinant for flowering time variation in Arabidopsis accessions (Johanson et al., 2000). FRI increases FLC levels to such an extent that it overrides the influence of the strong promotive effects of long days. FRI encodes a novel protein containing two coiled-coil domains that suggest interaction with other proteins or nucleic acids (Johanson et al., 2000). FRI levels are very low at all stages of plant development, and expression is not altered by vernalization treatment (C. Lister and C. Dean, unpublished results). Furthermore, extra copies or overexpression of FRI do not greatly increase flowering time; thus, unlike FLC, FRI levels do not have a quantitative effect on flowering time (C. Lister and C. Dean, unpublished results). Conversely, additional copies of FLC in plants containing active FRI significantly delay flowering in the absence of vernalization (Michaels and Amasino, 2000).

Many early-flowering Arabidopsis accessions carry loss-of-function FRI alleles. Two different groups of early-flowering accessions, typified by the well-used laboratory strains Ler and Columbia, carry different FRI deletion events, indicating that the early-flowering habit has arisen independently at least twice in the evolution of Arabidopsis accessions (Johanson et al., 2000). Putative loss-of-function FRI alleles appear to be common (Le Corre et al., 2002; Gazzani et al., 2003), and a high proportion of nonsynonymous changes in exon 1 has been detected. This extensive polymorphism may indicate positive selection for flowering time variation in natural populations (Le Corre et al., 2002).

Positive Regulators of FLC

Other genes that function in the upregulation of FLC have been identified through the analysis of early-flowering mutants that have been genetically linked to FLC or shown to have reduced FLC expression. Some of these mutants emerged from screens for mutants that flowered early in short-day photoperiods: early flowering in short days (efs) (Soppe et al., 1999), early in short days4 (esd4) (Reeves et al., 2002), and photoperiod independent early flowering1 (pie1) (Noh and Amasino, 2003). ESD4 encodes a SUMO-directed protease whose targets are currently unknown (Murtas et al., 2003). pie1 suppresses the late-flowering phenotype of both FRI and autonomous pathway mutants and is similar to ATP-dependent chromatin-remodeling proteins of the ISWI and SWI2/SNF2 family (Noh and Amasino, 2003).

Other FLC positive regulators have been found as suppressors of FRI activity and termed vernalization independence (vip) mutants (Zhang and van Nocker, 2002; Zhang et al., 2003). vip mutants flower earlier than flc null mutants and also show floral abnormalities, suggesting that VIP genes regulate multiple genes that influence flowering time and other developmental processes. VIP4 exhibits sequence homology with yeast Leo1 and proteins from Drosophila and Caenorhabditis elegans. In yeast, Leo1 interacts with the “cap” of the proteasome and is a component of the Paf1 transcriptional complex, which is required for the full expression of a subset of yeast genes (Zhang and van Nocker, 2002). VIP3 encodes a protein consisting almost exclusively of WD repeats and so is likely to function as a scaffold protein in protein complexes (Zhang et al., 2003).

Negative Regulation of FLC: Vernalization

Vernalization, the process that occurs in plants as they overwinter for many weeks in cold temperatures, strongly downregulates FLC levels and so accelerates flowering (Michaels and Amasino, 1999a; Sheldon et al., 1999). Vernalization is quantitative, with longer cold periods accelerating flowering more than shorter exposures, although the flowering time effect saturates after an extensive cold treatment. Vernalization is also a mitotically stable process. The cold signal is perceived many days before the meristem transition, and cuttings regenerated from vernalized plants flower without further vernalization (Metzger, 1988; Burn et al., 1993). This finding suggested that vernalization had an epigenetic basis (Wellensiek, 1962)—that is, a mitotically stable change in gene expression is established by the cold treatment that persists once the plant is returned to warmer temperatures. The vernalized state must be reset subsequently before the end of seed development to ensure that vernalization is again required for flowering in the next generation.

The molecular basis of vernalization has been explored through the isolation of mutants showing a defective vernalization response (Chandler et al., 1996; Sung and Amasino, 2004). Characterization of FLC RNA levels in vernalization (vrn) mutants revealed VRN gene function. FLC decreased in the cold, but instead of remaining low during subsequent development in warm temperatures, FLC levels increased in vrn1 and vrn2 mutants (Gendall et al., 2001; Levy et al., 2002). Therefore, the main function of VRN1 and VRN2 is to maintain FLC repression, suggesting that they are part of the cellular machinery that provides a memory of vernalization. By contrast, FLC levels did not decrease in the cold in vin3 mutants, suggesting that VIN3 is involved in the establishment of the repression of FLC (Sung and Amasino, 2004).

VIN3 encodes a protein with a plant homeodomain and a fibronectin type III domain (often involved in protein–protein interactions). VIN3 expression is induced in the later stages of cold exposure and then declines during the subsequent warm growth. Therefore, its function appears to be related to the measurement of the duration of cold exposure and in the establishment of the vernalized state (Sung and Amasino, 2004). VRN2 encodes a protein most similar to Suppressor of zeste 12 [Su(Z)12], a Polycomb group protein required in Drosophila to maintain appropriate repression of certain genes (Birve et al., 2001). Polycomb group proteins usually form part of multisubunit complexes and act to maintain patterns of gene expression set up by other transcriptional regulators. Recently, Su(Z)12 was identified as part of the EXTRA SEX COMBS-E(Z) Polycomb complex, which maintains silent chromatin states through the methylation of specific lysine residues on histone proteins (Cao et al., 2002; Czermin et al., 2002; Müller et al., 2002). Consistent with the idea that VRN2 maintains FLC chromatin in a repressed state, the chromatin structure at FLC intron 1 in vrn2 mutants was shown to be DNaseI hypersensitive, a condition usually associated with transcriptional activity (Gendall et al., 2001). Vernalization has been shown to induce changes in histone modifications at the FLC locus that are characteristic of silent chromatin, specifically an increase in histone H3 Lys-27 and Lys-9 methylation and a decrease in histone H3 acetylation (Bastow et al., 2004; Sung and Amasino, 2004). These changes in FLC chromatin are dependent on mutations in the vernalization pathway (Bastow et al., 2004; Sung and Amasino, 2004). Interestingly, the histone methylation changes occur in discrete regions through the FLC locus, some of which were in similar regions to cis elements required for correct FLC regulation as defined though in vivo expression analysis of constructs carrying FLC deletions (Sheldon et al., 2002). This analysis defined two promoter elements, one positive and one negative, that affected FLC expression in the absence of vernalization. Separate domains also were found for repression and maintenance of repression by vernalization within the large intron 1 (Sheldon et al., 2002). The exact relationship between the histone modifications and the regulatory cis elements has yet to be determined.

Ectopic overexpression revealed a vernalization-independent function for VRN1, predominantly through the regulation of the floral pathway integrator FT (Levy et al., 2002). 35S:VRN1 plants flowered early as a result of increased FT expression, whereas FT expression was reduced in vrn1 mutants. FLC expression was unaffected in these lines until the plants had been vernalized, demonstrating that VRN1 requires vernalization-specific factors to target FLC. Methylation of the Lys-27 residue on histone H3 by the VRN2 complex may be the vernalization-specific process required (Bastow et al., 2004). VRN1 encodes a protein that contains two putative B3 domains (plant-specific DNA binding domains) that bind DNA in vitro in a non-sequence-specific manner (Levy et al., 2002). Whether VRN1 binds specific FLC sequences in vivo is still unknown. To date, little is known about the resetting of FLC expression. The epigenetic marks laid down during vernalization need to be erased in the gametes or developing embryo to allow high FLC levels in the progeny, ensuring a vernalization requirement in every generation.

FLC is not the only flowering time target of vernalization, because an flc null mutant grown in non-floral-promotive conditions (so that flowering time changes can be detected easily) flowers slightly earlier after vernalization (Michaels and Amasino, 2001). A likely additional target is AGL24, a gene with similar properties to SOC1 (Yu et al., 2002; Michaels et al., 2003a). Like SOC1, AGL24 is upregulated by vernalization, but unlike SOC1, this upregulation is independent of FLC activity. agl24 mutants are late flowering, but this is not strongly suppressed by vernalization (Yu et al., 2002; Michaels et al., 2003a). AGL24 appears to act downstream of SOC1 but upstream of LFY, but overexpression of AGL24 increases SOC1 expression, indicating that the relationship between these genes may not be straightforward (Michaels et al., 2003a). The subcellular localization of AGL24 appears to be regulated by phosphorylation, although a role for this process on the effects of AGL24 on flowering time and vernalization has not been established (Fujita et al., 2003).

Negative Regulation of FLC: The Autonomous Pathway

fca, fld, fpa, fve, fy, ld, and fld mutants are classified as functioning in an autonomous pathway because they flower late in all photoperiods (Koornneef et al., 1991; Lee et al., 1994; Chou and Yang, 1998). Genes of the autonomous pathway function to reduce FLC mRNA accumulation (Michaels and Amasino, 1999a, 2001; Sheldon et al., 1999, 2000b). The late-flowering phenotype of the mutants can be overcome by either vernalization or growth in far-red-enriched light, so these environmental cues are considered to function in parallel to the autonomous pathway (Martínez-Zapater, 1990; Koornneef et al., 1991). However, unlike vernalization, far-red light does not repress FLC RNA levels in these mutants, indicating that vernalization and far-red floral promotion have different downstream targets (Sheldon et al., 2000a; A. Gendall and C. Dean, unpublished results).

The genes of the autonomous pathway do not function in a linear hierarchy but as a series of subgroups sharing a common target, FLC (Figure 2). FCA and FY form one epistatic group, with fca and fy mutants having additive effects on flowering time when combined with fve and fpa mutants (Koornneef et al., 1998) (Figure 2). FCA encodes a protein with two RNA binding domains and a WW protein interaction motif at the C terminus (Macknight et al., 1997). The FCA transcript is alternatively processed, with four detectable transcripts, only one of which produces a protein active in flowering time control (Macknight et al., 1997). FCA negatively regulates its own expression by promoting cleavage and polyadenylation within intron 3 (Quesada et al., 2003). This causes the production of one of the inactive forms of FCA transcript at the expense of the full-length FCA mRNA, thus limiting the expression of active FCA. The negative autoregulation is under developmental control, requires the FCA WW interaction domain, and is dependent on FY, which interacts with FCA through the WW domain (Macknight et al., 2002; Quesada et al., 2003; Simpson et al., 2003). FY encodes a protein with WD repeats and a C-terminal extension carrying two PPLP motifs and belongs to a highly conserved group of eukaryotic proteins. The yeast homolog Pfs2p is an essential component of protein complexes involved in RNA 3′ end processing (Ohnacker et al., 2000). The current model is that FCA binds target RNA through its two N-terminal RNA binding domains and tethers the 3′ end–processing machinery to this RNA via interaction between the FCA WW domain and the PPLP domain of FY. One target for this interaction is FCA pre-mRNA, but whether FLC is another pre-mRNA targeted directly by FCA/FY is not yet known.

The epistasis analysis also grouped the activities of FPA and FVE in the repression of FLC (Koornneef et al., 1998). FPA also encodes an RNA binding protein (Schomburg et al., 2001), whereas FVE encodes a WD-repeat protein (Blázquez et al., 2001), but how they interact functionally has not been established. A fy fpa double mutant has not been recovered and so may be lethal (Koornneef et al., 1998), which would suggest that FPA and FY are redundant for an essential role in plant development. This is not the case for the fve fca double mutant, which can be recovered (Koornneef et al., 1998).

Two other genes, LUMINIDEPENDENS (LD) and FLOWERING LOCUS D (FLD), are negative regulators of FLC that show genotype-dependent effects on flowering time (Lee et al., 1994; Sanda and Amasino, 1996). LD encodes a homeodomain protein that is targeted to the nucleus (Lee et al., 1994). Some homeodomain proteins function in RNA processing (Dubnau and Struhl, 1996); therefore, LD also may interact with RNA as predicted for FCA and FPA (Simpson et al., 1999). FLD encodes a protein that is homologous with a member of a human histone deacetylase complex, which functions to remodel chromatin into a silent form via the removal of acetyl groups from histone tails (Hakimi et al., 2003; He et al., 2003). In a similar manner, FLD acts to deacetylate FLC chromatin, preventing the transcription of FLC and promoting flowering (He et al., 2003). Thus, epigenetic regulation via chromatin modification is emerging as a major mechanism to modulate FLC levels.

Integrating the Antagonistic Activities That Regulate FLC

A key question when antagonistic pathways control the expression of a common target is how the predominance of those activities is exerted. With respect to FLC, FRI activates the expression and overcomes the repression caused by genes of the autonomous pathway. This predominance may be attributable to differing temporal expression of the genes. Expression of FCA is maintained at low levels during early seedling development as a result of FCA negative autoregulation promoting the use of an internal polyadenylation site within FCA intron 3 (Quesada et al., 2003). However, as judged by levels of an FCA:β-glucuronidase (GUS) translational fusion, which specifically monitors when an active transcript is generated, FCA levels increase in the shoot and root meristematic regions by ∼4 to 5 days after germination (Macknight et al., 2002). This may reflect a specific developmental mechanism to increase FCA levels and thus decrease FLC levels in the meristem at later stages of development. Removal of introns from FCA bypasses the autoregulation and results in increased levels of FCA much earlier in development. This overcomes the repression of flowering conferred by FRI (Quesada et al., 2003). Thus, altering the timing or levels of FCA and FRI accumulation can reverse the outcome of their antagonistic activities. Therefore, the negative autoregulation of FCA may have evolved to limit FCA activity early in development and hence avoid precocious flowering too soon after germination.

Photoperiod Pathway

To perceive and respond to changes in photoperiod, plants must be able to detect light duration (period) and couple this to an internal timer or oscillator (circadian clock) (Thomas and Vince-Prue, 1997). In Arabidopsis, the red/far-red light–absorbing phytochromes (Quail, 2002) and the UV/blue light–absorbing cryptochromes (Lin, 2000) are the main photoreceptors involved in daylength sensing. Light perceived by these photoreceptors acts to reset the oscillator so that it remains in synchrony (entrained) with its environment. Entrainment is achieved via a number of mechanisms, one of which involves PHYTOCHROME-INTERACTING FACTOR 3 (PIF3), a basic helix-loop-helix transcription factor. PHTYOCHROME (PHY) B is activated by light and then binds to PIF3, resulting in the upregulation of both CIRCADIAN CLOCK ASSOCIATED1 (CCA1) and LATE ELONGATED HYPOCOTYL (LHY), two proteins postulated to be constituents of the central oscillator (Martínez-García et al., 2000). Other genes that influence entrainment and flowering time include the ZTL/FKF/LKP family (Nelson et al., 2000; Somers et al., 2000; Schultz et al., 2001). zeitlupe (ztl) mutants are late flowering in long days and disrupt circadian rhythms, causing an increase in period length. This phenotype is strongly dependent on light intensity, suggesting that ZTL may act as a photoreceptor (Somers et al., 2000). Overexpression of LOV KELCH PROTEIN2 (LKP2) induces late flowering in long days and abolishes a number of circadian rhythms (Schultz et al., 2001). The third member, FLAVIN BINDING KELCH REPEAT F-BOX1 (FKF1), causes late flowering when mutated but has less influence on the circadian system (Nelson et al., 2000). These three genes encode proteins that contain a PAS domain, an F-box (involved in protein degradation), six repeated kelch motifs (a domain involved in protein–protein interactions), and a LOV domain (a light-sensing domain of the blue light receptor phototrophin). It has been suggested that the presence of these motifs, along with protein interactions in vitro between ZTL, PHYB, and CRYPTOCHROME1 (CRY1), indicates that the ZTL/FKF/LKP2 family functions to recruit and degrade clock components in a light-dependent manner (Jarillo et al., 2001; Kim et al., 2003). However, recent studies of the LOV domain of FKF1 have shown that ZTL, FKF1, and LKP2 form a new family of blue light photoreceptors, suggesting that these proteins function in a unique light-signaling pathway (Imaizumi et al., 2003).

EARLY FLOWERING3 (ELF3) also is a component of the signal transduction pathway between the photoreceptors and the oscillator. It was identified originally as an early-flowering mutant that disrupted the circadian system and caused arrhythmia, but only under constant light conditions (Hicks et al., 1996). The reason for this light-conditional arrhythmia has been explained by a number of recent studies. ELF3 has been found to encode a nuclear protein whose expression is regulated in a circadian manner with a peak at the beginning of the night (Covington et al., 2001; Liu et al., 2001). The presence of ELF3 during the night period has been shown to antagonize light input to the oscillator and the circadian-regulated CAB gene. Therefore, it has been proposed that ELF3 sustains rhythmicity in constant light by repressing phototransduction at subjective dusk (McWatters et al., 2000). In the elf3 mutant, the absence of ELF3 under constant light exposes the oscillator to light signaling during the subjective night phase, causing the oscillator to stop. Therefore, ELF3's effects on the photoperiodic regulation of flowering time result from its role in the circadian system. Exactly how ELF3 functions in repressing light signaling is unknown, but it does exhibit a number of phyB-like phenotypes and has been shown to interact physically with PHYB (Liu et al., 2001). Aberrant PHYB-mediated responses also are observed in the sensitivity to red light reduced1 (srr1) mutant. Like elf3, srr1 mutants are pale with elongated petioles and hypocotyls and are early flowering (Staiger et al., 2003). The late-flowering mutant gigantea (gi) is also implicated in the PHYB pathway, and an allele of GI was isolated from a screen for inhibitors of hypocotyl elongation in red light, a PHYB-dependent phenotype (Huq et al., 2000). In addition to these phyB effects, gi and srr1 mutants disrupt the circadian system, shortening the period and reducing the amplitude of a number of circadian outputs (Huq et al., 2000; Staiger et al., 2003). Together, these data indicate that ELF3, SRR1, and GI function downstream of PHYB and act to maintain the viability of the circadian system so that it generates the correct period and amplitude. Therefore, when mutated, these genes disrupt the circadian system and consequently cause aberrant regulation of the photoperiodic timing of the floral transition.

CCA1, LHY, and TIMING OF CAB EXPRESSION1 (TOC1) have been identified as possible candidates for oscillator components in Arabidopsis (Millar et al., 1995; Schaffer et al., 1998; Wang and Tobin, 1998) (Figure 3). When mutated, TOC1, CCA1, and LHY all confer early flowering. CCA1 and LHY both encode very similar proteins with a single MYB repeat. The expression of CCA1 and LHY RNA and protein is regulated by the circadian clock, with a peak in the early morning. Mutations in either gene shorten the circadian period, whereas overexpression of either one results in downregulation of the other and general arrhythmia (Schaffer et al., 1998; Wang and Tobin, 1998; Green and Tobin, 1999; Mizoguchi et al., 2002). TOC1 encodes a pseudoresponse regulator whose expression also is regulated by the circadian clock, with a peak of expression in the evening. Mutations in TOC1 shorten the period of circadian outputs, and overexpression causes arrhythmia (Somers et al., 1998; Strayer et al., 2000).

From the expression profiles of LHY, CCA1, and TOC1, a model has been proposed whereby these three proteins function to produce an autoregulatory transcriptional and translational negative feedback loop (Alabadí et al., 2001). During the late evening, TOC1 activates CCA1 and LHY, resulting in a peak of expression at the start of the day. Subsequent increases in CCA1 and LHY during the day act to repress TOC1 expression; consequently, CCA1 and LHY mRNA levels decay by the end of the day as a result of the removal of their activator, TOC1. Reduction in CCA1 and LHY relieves the repression of TOC1, and the cycle begins again, forming an oscillatory loop. However, it must be stated that many aspects of this model have yet to be proven, which has led to the development of an alternative model in which the members of the TOC1 family form the circadian oscillator (Makino et al., 2002; Matsushika et al., 2002a, 2002b). This model involves the sequential expression of a quintet of TOC1-like genes in circadian waves, independent of the entrained photoperiod conditions (Matsushika et al., 2000). The expression profile of these genes changes constantly throughout the 24-h cycle, allowing some indication of the progression of time. Recent data have demonstrated that ELF4 also may be involved in the oscillator. elf4 mutants cause arrhythmia and, like toc1-2, cause the downregulation of CCA1 mRNA. ELF4 expression peaks during the night, suggesting that it may function along with TOC1 to activate CCA1 and LHY expression (Doyle et al., 2002).

The CONSTANS (CO) gene plays a key role in integrating light and temporal information essential for determining how daylength sensing is achieved. CO encodes a transcription factor with two B-box zinc fingers and directly activates FT. co mutants flower late under long days, but flowering time is unaffected in short days. Overexpression of CO causes early flowering under either photoperiod (Putterill et al., 1995; Suárez-López et al., 2001). Altered levels of CO expression are observed in a number of circadian mutants. In late-flowering, gain-of-function lhy and cca1 mutants, CO mRNA levels are low, whereas in the early-flowering elf3 mutant, CO levels are high. The effects of these mutations on the circadian clock appear to be translated into a flowering effect via CO, an idea that is consistent with the circadian regulation of CO expression. CO mRNA also is modulated by photoperiod, and under short days, CO expression peaks in the night. Under long days, there is a peak in expression at the end of the day, an expression profile that is mirrored by FT. Therefore, a model has been proposed whereby CO expression provides a light-sensitive rhythm for the perception of daylength. Only under long days would high CO mRNA levels coincide with light, leading to an increase in FT expression (Suárez-López et al., 2001). This description of CO function provides molecular support for the theoretical photoperiodic model of external coincidence, first proposed by Bünning (1936). This model predicts that the circadian clock generates a photoperiodic response rhythm that is sensitive to light at certain phases of the cycle. When light coincides with these sensitive phases, a photoperiodic response results. In Arabidopsis, the photoperiodic response rhythm is provided for by the rhythm of CO expression, and light input to FT is perceived by PHYA and CRY2 (Yanovsky and Kay, 2002). In long-day conditions, CO expression coincides with the light input, resulting in the activation of FT expression and the promotion of flowering. Despite these advances in understanding the photoperiod regulation of flowering, the mechanism that generates the daylength-dependent expression profiles of CO remains unclear. However, recent studies suggest that FKF1 may be involved (Imaizumi et al., 2003). It is interesting that PHYA and CRY2 are light-labile proteins that exhibit a diurnal pattern in protein abundance under short days but not long days. This may act as part of the daylength-sensing mechanism (Mockler et al., 2003). The importance of protein stability in flowering time has been shown through quantitative trait loci analysis. In the Arabidopsis accession Cvi, a single amino acid substitution causes an increased stability of CRY2-Cvi, and this allele confers an early-flowering phenotype in short days (El-Assal et al., 2001).

Integrating the Antagonistic Activities of the Photoperiod Pathway and FLC on the Expression of Floral Pathway Integrators

The photoperiod and FLC-regulated pathways ultimately converge on a few targets that include FT and SOC1. The molecular basis of the antagonistic action between the two pathways has been investigated by studying the levels of these integrator genes in a number of mutant combinations. For example, the activation of SOC1 by overexpression of CO was blocked completely by overexpression of FLC (Hepworth et al., 2002), whereas the photoperiod deficiency in co and gi mutants was partially restored when combined with increased levels of FLC generated by fca (Koornneef et al., 1998). In a similar manner, the daylength-insensitive phenotype of CRY-Cvi can be overcome by the presence of high levels of FLC, which decrease the levels of CRY2 and restore photoperiod response (El-Assal et al., 2003). To determine how the interactions between these pathways are achieved, the cis elements involved in SOC1 repression by FLC and activation by CO were examined in a deletion series of SOC1 promoter:GUS fusions in transgenic plants. A 351-bp promoter sequence was defined that mediates both activation by CO and repression by FLC. A MADS box binding element (CArG) is present within the 351-bp region, and this was shown to bind FLC in vitro (Hepworth et al., 2002). Thus, when both CO and FLC proteins are present, FLC binds to the 351-bp region in vivo, possibly at the CArG element, impairing the functioning of CO in the activation of SOC1 transcription (Hepworth et al., 2002). In addition, high levels of FLC cause the downregulation of CRY2, which prevents the CRY2 promotion of CO (El-Assal et al., 2003). Thus, FLC acts at several levels to downregulate the effects of the photoperiod promotive pathway.

Light Quality and Ambient Temperature

Light also affects flowering time independently of photoperiod via a light quality pathway (Simpson and Dean, 2002). Light quality is affected by shading, which results in a reduction in the ratio of red to far-red light. PHYB represses flowering via the downregulation of FT (Halliday et al., 2003), and this is mediated by PHYTOCHROME AND FLOWERING TIME1, a gene that encodes a nuclear protein that activates FT expression (Cerdan and Chory, 2003). PHYD and PHYE also repress flowering, but their effects are observed only when the predominant effect of PHYB is absent (Halliday et al., 1994; Aukerman et al., 1997; Devlin et al., 1998, 1999). In contrast to red light, far-red and blue light promote flowering. phyA is late flowering in conditions in which the light given at the end of the light period is far-red enriched or when the night period is interrupted by a short light period (Reed et al., 1994), and cry2 is late flowering under long days (Lin, 2000). It has been proposed that CRY2 and PHYA promote flowering via two independent pathways, one directly on floral pathway integrators and one by repression of PHYB (Mockler et al., 2003). Therefore, the CRY2 and PHYA pathways act in an antagonistic manner to the PHYB pathway, forming a network that functions to regulate flowering under varying light qualities, providing a readout to the plant of how crowded the local environment is with other plants.

Ambient temperature significantly influences the phenotypic effects caused by the loss of different photoreceptors. phyB flowers earlier than the wild type at 22°C, whereas at 16°C, the mutation has no effect (Halliday et al., 2003). At 16°C, PHYE appears to act as the predominant phytochrome regulating flowering: phyA phyB phyD phyE quadruple mutants flower earlier than phyA phyB phyD triple mutants, an effect that is mediated by FT levels (Halliday et al., 2003). The effect of the cry2 mutation on flowering also is temperature sensitive. At 23°C cry2 mutations delay flowering, but this effect is enhanced greatly at 16°C. The increased delay in flowering appears to be caused by the loss of PHYA activity at 16°C. At 23°C, PHYA functions redundantly with CRY2 to promote flowering, so cry2 mutants are only slightly delayed. However, at 16°C, cry2 is as late flowering as the phyA cry2 double mutant at 23°C, suggesting a loss of PHYA activity at the lower temperature (Blázquez et al., 2003).

The role of ambient temperature on flowering time control also has been analyzed in the wild type and flowering time mutants (Blázquez et al., 2003). The flowering of wild-type plants is delayed at 16°C, although fca and fve mutants show little difference in flowering time at 16 or 23°C. FCA and FVE are thought to mediate a temperature-dependent repression of FT expression that does not involve FLC, the usual target of these genes within the pathways that enable the floral transition (Blázquez et al., 2003). This has been interpreted as showing that FCA and FVE function in a pathway that mediates ambient temperature effects, in addition to their role in the autonomous pathway. However, it is possible that this is just a side effect of those mutants containing high levels of FLC, which will effectively suppress FT expression, ameliorating any regulation by the light quality pathway. The effects of ambient temperature on flowering may all be through the light quality pathway activation of FT and may not directly involve FCA or FVE function.

Hormonal Inputs

Genetic studies have confirmed the physiological findings that gibberellins (GAs) accelerate Arabidopsis flowering (Langridge, 1957). Plants overexpressing GA-20 oxidase, a gene late in the GA biosynthesis pathway, are early flowering in both long days and short days (Huang et al., 1998; Coles et al., 1999). The spindly mutation is considered to exhibit constitutively active GA signaling and is early flowering (Jacobsen and Olszewski, 1993), whereas plants overexpressing FLOWERING PROMOTIVE FACTOR1, a gene involved in GA signal transduction or responsiveness to GAs, also flower early (Kania et al., 1997). Conversely, a decrease in GA levels or insensitivity to GA signaling delays flowering, although this effect is significant only in short days (Wilson et al., 1992). For example, a mutation in GA1, the first committed step in GA biosynthesis, makes Arabidopsis an obligate long-day plant because it no longer flowers under short days. gai mutants are insensitive to GAs and also are late flowering in short days (Wilson et al., 1992). Double mutant analyses have established that the GA pathway is genetically distinct from the photoperiod promoting pathway and the enabling pathway (Figure 3) and have confirmed that the GA pathway has less influence on flowering time in long days than in short days (Putterill et al., 1995; Chandler et al., 2000; Reeves and Coupland, 2001). However, in the absence of the long-day promotion pathway, the GA pathway is an important promoter of flowering (Reeves and Coupland, 2001). Some physiological experiments have suggested that vernalization works via the GA pathway, and in some plant species, GA application can substitute for a vernalization treatment (Zeevaart, 1983; Sheldon et al., 1999). However, in Arabidopsis, vernalization and GA action are clearly independent, because the fca-1 ga1-3 and FRI FLC ga1-3 genotypes, in which GA biosynthesis is blocked, still respond to vernalization (Michaels and Amasino, 1999b; Chandler et al., 2000).

How changes in GA biosynthesis or signal transduction result in altered flowering time is an area of active research. One target of the GA signal is LFY, because LFY promoter activity is reduced in a ga1-3 mutant and increased by exogenous GA application in both the wild type and ga1-3 (Blázquez et al., 1998). The promoter region of LFY, which is responsible for GA-induced expression, was found to include an 8-bp motif that is a consensus binding site sequence for MYB transcription factors (Blázquez and Weigel, 2000). A MYB-like transcription factor, AtMYB33, has an overlapping expression pattern with LFY in the floral meristem and can bind the 8-bp LFY promoter motif in vitro (Gocal et al., 2001). AtMYB33 expression is induced upon GA application, suggesting that it may play a role in the GA responsiveness of the LFY promoter, so GAs may alter flowering time by increasing LFY expression. The phenotype of the atmyb33 mutant has not been reported. Constitutive expression of LFY accelerates flowering but cannot fully complement the flowering time effect of ga1-3 (Blázquez et al., 1998). This may be attributable to another floral integrator, SOC1, also being regulated by GAs (Moon et al., 2003a). The study of a mutation in EARLY BOLTING IN SHORT DAYS (EBS) suggests that GAs also may influence FT expression (Gómez-Mena et al., 2001; Piñeiro et al., 2003). EBS encodes a protein with a bromoadjacent homology domain and a plant homeodomain zinc finger and so is assumed to be part of a chromatin-remodeling complex (Piñeiro et al., 2003). ebs mutants are significantly early flowering in short days and slightly early flowering in long days, but the early-flowering phenotype is lost in ebs ga1-3 and ebs ft double mutants (Gómez-Mena et al., 2001; Piñeiro et al., 2003). This finding suggests that GAs can alter FT expression patterns but that this effect is repressed by EBS via chromatin remodeling. Therefore, the inability of LFY overexpression to fully complement the late-flowering phenotype of ga1-3 may be attributable to the altered expression of SOC1 and/or FT in ga1-3.

There is considerable research currently analyzing the cis elements in FT, SOC1, and LFY that mediate the convergence of the GA, photoperiod, and light quality activation pathways. The cis elements through which GAs activate the LFY promoter are distinct from those required for photoperiodic activation, demonstrating that at least some of the convergence of these different promotion pathways occurs at the level of the LFY promoter (Blázquez and Weigel, 2000).

Other hormones are implicated in the control of flowering time in Arabidopsis. Mutants that reduce abscisic acid biosynthesis are earlier flowering under noninductive conditions, suggesting that abscisic acid inhibits flowering (Martínez-Zapater et al., 1994). In support of this, abi1 and abi2 (abscisic acid signaling mutants) have been shown to reduce the flowering time of fca-1 mutants (Chandler et al., 2000). Despite the large amount of physiological evidence implicating cytokinins in flowering time control (Bernier et al., 1993), genetic evidence for a role for cytokinins in the promotion of Arabidopsis flowering is lacking. Ethylene signaling mutants (Guzmán and Ecker, 1990; Ogawara et al., 2003), the brassinosteroid biosynthesis mutant det2 (Chory et al., 1991), and plants altered in salicylic acid biosynthesis (Martínez et al., 2004) are late flowering, implicating these plant growth regulators in floral promotion pathways.