• Open Access

A physiological overview of the genetics of flowering time control


  • Georges Bernier,

    Corresponding author
    1. Laboratory of Plant Physiology, Department of Life Sciences, University of Liège, B22 Sart Tilman, B4000 Liège, Belgium
    Search for more papers by this author
  • Claire Périlleux

    1. Laboratory of Plant Physiology, Department of Life Sciences, University of Liège, B22 Sart Tilman, B4000 Liège, Belgium
    Search for more papers by this author

Correspondence (fax 32 4 366 38 31; e-mail gbernier@ulg.ac.be)


Physiological studies on flowering time control have shown that plants integrate several environmental signals. Predictable factors, such as day length and vernalization, are regarded as ‘primary’, but clearly interfere with, or can even be substituted by, less predictable factors. All plant parts participate in the sensing of these interacting factors. In the case of floral induction by photoperiod, long-distance signalling is known to occur between the leaves and the shoot apical meristem (SAM) via the phloem. In the long-day plant, Sinapis alba, this long-distance signalling has also been shown to involve the root system and to include sucrose, nitrate, glutamine and cytokinins, but not gibberellins. In Arabidopsis thaliana, a number of genetic pathways controlling flowering time have been identified. Models now extend beyond ‘primary’ controlling factors and show an ever-increasing number of cross-talks between pathways triggered or influenced by various environmental factors and hormones (mainly gibberellins). Most of the genes involved are preferentially expressed in meristems (the SAM and the root tip), but, surprisingly, only a few are expressed preferentially or exclusively in leaves. However, long-distance signalling from leaves to SAM has been shown to occur in Arabidopsis during the induction of flowering by long days. In this review, we propose a model integrating physiological data and genes activated by the photoperiodic pathway controlling flowering time in early-flowering accessions of Arabidopsis. This model involves metabolites, hormones and gene products interacting as long- or short-distance signalling molecules.


A plant is a sessile organism that has to grow and reproduce in nature at the location at which its seed happens to germinate. The success of reproduction thus requires that plants continuously and accurately monitor their local environment to flower at the correct time. This is all the more critical for monocarpic species which have a single chance to sexually reproduce. The mechanisms elaborated to secure this correct timing were first investigated by physiologists, and have essentially been determined in recent years by molecular geneticists. Whilst physiological work explored diversity (various plant species studied in a broad range of environmental conditions), the genetic approach has focused mostly on the single species Arabidopsis thaliana, usually grown in a more restricted set of environmental conditions. It is not surprising, therefore, that these two approaches have yielded a large body of information, but without achieving comprehensive integration at the whole-plant level (Périlleux and Bernier, 2002). Our attempt in this review is to aid such an integration.

Physiological control of flowering time

Environmental factors

In the natural conditions of temperate areas, many factors in the environment influence flowering time (Lang, 1965; Bernier et al., 1981; Thomas and Vince-Prue, 1997). These factors are either predictable or not predictable, and hence can or cannot be used reliably by plants to time their reproduction. Factors that are highly predictable are considered to be the most specific or ‘primary’ controlling factors; these include the annual change in day length and the period of winter cold (Table 1). Less predictable climatic factors, such as ambient temperature, light integral (day length × irradiance) and water availability, are usually viewed as ‘secondary’ factors that can only modulate the effects of primary factors. Finally, unpredictable or ‘tertiary’ factors are those that the plant has to face locally, such as mineral availability and neighbours (Table 1). The effects of neighbourhood have sometimes been limited to the response to light quality, although they also involve competition for light, water and minerals.

Table 1.  Environmental factors participating in the control of flowering time in temperate areas
FactorMain perceiving organ(s)
  1. SAM, shoot apical meristem.

Highly predictable or primary
 Day length (photoperiodism)Leaves/stem/SAM
 Winter cold (vernalization)SAM
Moderately predictable or secondary
 Ambient temperatureAll organs
 Water availabilityRoots
Unpredictable or tertiary
 Mineral availabilityRoots
 Light qualityLeaves/stem/SAM
 Neighbouring vegetationLeaves/stem/SAM/roots

Although such a classification of controlling factors can account for the predominance of primary factors in many environmental niches, the situation is not so simple: the two primary factors (day length and vernalization) can substitute for each other, and can also be replaced by secondary or tertiary factors (Table 2). The promotion of flowering by a primary factor can also be reduced or even completely suppressed by another factor. Examples include the suppression of flowering, in favourable photoperiodic conditions, by water stress (drought) in the long-day (LD) plant Lolium temulentum and the short-day (SD) plants Xanthium strumarium and Pharbitis nil, or by excess nitrogen input in the LD plant Sinapis alba and the SD plant soybean. In other cases, it has been found that a primary factor might be conditional only in certain circumstances. This is true in Calceolaria, in which low irradiance creates a requirement for vernalization, although vernalization is not required at high irradiance (Bernier, 1988).

Table 2.  Examples of factors capable of substituting for a primary factor in the control of flowering time
Primary factorAlternative factor acting in conditions in which the primary factor is unfavourableExamples of plant species
Long daysVernalizationSilene armeria, Blitum sp., Anagallis arvensis, Sinapis alba
High ambient temperatureRudbeckia bicolor, Samolus parviflorus, Silene armeria
High irradianceSinapis alba, Brassica campestris
Low nitrogen availabilitySinapis alba, late-flowering pea lines
Short daysVernalizationPerilla sp., Pharbitis nil, Maryland Mammoth tobacco
High ambient temperatureChenopodium polyspermum
High irradiancePharbitis nil
Low irradiancePerilla sp., Salvia occidentalis
Low nitrogen availabilityPerilla sp., Pharbitis nil
VernalizationLong daysLate-flowering pea lines, Geum urbanum
Short daysSome winter cereals and perennial grasses, kohlrabi
High ambient temperatureScrofularia sp., Festuca arundinacea
High ambient temperature + high irradianceShuokan chrysanthemum

It is important to emphasize here that these interactions have been found in experiments in which only two factors were changed, and each was given at optimal levels. Such situations are probably infrequent in natural environments in which several factors change simultaneously and mutually influence the optimal values of each other. For example, it is known that, in many photoperiodic plants, the critical day length for flowering is affected by the ambient temperature. It is therefore no surprise that, when variations in more than two factors were tested on time to flowering, very complex situations were found. Thus plants, as sessile opportunists, can make use of alternative factors to control their flowering time in natural environments.

Organs involved in environmental perception

Environmental factors participating in the control of flowering time are not all perceived by the same organ(s) (Lang, 1965; Bernier et al., 1981; Bernier, 1988) (Table 1). Vernalization is generally perceived by the shoot apex [shoot apical meristem (SAM) plus leaf primordia], as shown by the fact that cold treatment applied to excised apices induces their floral transition, provided that they are supplied with nutrients. In pea and a few other species, it has been reported that vernalization is also perceived by leaves.

Day length and light quality are usually believed to be essentially perceived by expanded leaves but, in the absence of leaves, they can also be perceived by the stem. However, excised shoot apices of the SD plants Perilla frutescens, Xanthium and Pharbitis and the LD plant Anagallis arvensis respond to day length in the same way as intact plants, suggesting that this plant part is also involved in the response to photoperiod, provided that apices are supplied with sucrose and minerals in the culture medium. Experiments with optical fibres have further demonstrated that the apex of Pharbitis seedlings is capable of perceiving light quality treatments controlling flowering time, namely a red night-break or a far-red end-of-day extension.

Thus, all aerial organs participate in the perception of day length and light quality. This is also the case for ambient temperature, which is, of course, sensed by all plant parts, including the roots. For example, high temperature, promoting flowering in Silene armeria and inhibiting it in Brassica pekinensis, is perceived by the roots. Irradiance is essentially perceived by photosynthetic organs, expanded leaves and stem, whereas water and mineral availability are perceived by the roots. Roots have been found in some studies to promote or inhibit flowering depending on the species and environmental conditions. However, the role of the ‘hidden half of the plant’ has often been disregarded in physiological studies and genetic work on flowering.

Endogenous cues

In addition to environmental signals, plants are also known to monitor endogenous cues related to their flowering time, such as size, node number or age. Size rather than age was demonstrated to be particularly important in biennials, as well as in long-lived monocarps, in polycarps with long-lived monocarpic ramets and other polycarps (Lacey, 1986). Biennials, for example, generally flower during their second year of growth when they are cultivated in resource-rich conditions, such as experimental growth areas, gardens or agricultural fields. In natural environments, however, they often flower only during their third or fourth year, or even later (Lacey, 1986; Klinkhamer et al., 1987). They should then be called ‘delayed’ biennials or, more appropriately, ‘monocarpic perennials’. As concluded in many field studies, the best predictor of flowering onset in these plants is the reaching of a threshold size, although this threshold may vary greatly amongst species and ecotypes (Lacey, 1986; Wesselingh et al., 1993). This conclusion is in line with physiological observations showing that partial or complete removal of foliage, i.e. plant trimming, may decrease or even abolish the response of many plants to vernalization or favourable day lengths (Bernier et al., 1981). Size is directly related to the amount of resources accumulated, and thus depends on the ambient temperature, irradiance, water/mineral availability and presence/absence of neighbours. In other words, in natural environments in which many factors are far from optimal, secondary and tertiary factors (Table 1) are often predominant over the primary factors for the control of flowering time. By contrast, flowering in natural populations of annuals is often principally controlled by a primary environmental cue, such as day length, and occurs independently of size or age (Lacey, 1986).

Long-distance signals

Although the fact that most plant parts participate in the sensing of the environmental factors that control flowering time clearly indicates that inter-organ, long-distance signalling must be involved in the triggering of flowering of the SAM, most of the physiological work to date has favoured the study of the unidirectional signalling event linking, in photoperiodic plants, the leaves to the SAM. The leaf-to-SAM signal is called ‘florigen’ when the leaves are exposed to day lengths favourable to flowering and ‘antiflorigen’ when the leaves are exposed to day lengths unfavourable to flowering. Numerous grafting experiments have shown the movement of such signals in several plant species (Thomas and Vince-Prue, 1997), but progress in identifying them has been extremely slow. A first success has been recorded recently in the rosette LD plant Lolium temulentum (King and Evans, 2003). The exposure of plants to LDs causes an increase in the level of several gibberellins, especially GA5, in the leaves, and their further transport to the SAM. When applied to leaves at early stages of the floral transition, GA5 markedly stimulates flowering, and hence may be the ‘Lolium florigen’. At later stages of the transition, other compounds come into play, such as GA1/GA4 (King and Evans, 2003) and sucrose (Périlleux and Bernier, 1997). GA5 is unlikely, however, to be a universal ‘florigen’, as it is not detected in extracts of florally induced Arabidopsis and Sinapis plants (Xu et al., 1997; Corbesier et al., 2004).

In Sinapis, analyses of changes in the contents of phloem and xylem saps during the floral transition have disclosed a complex shoot-to-root-to-shoot signalling loop involving both nutrients and hormones (Havelange et al., 2000; Bernier et al., 2002). As shown in Figure 1, exposure of leaves to an inductive LD results in the rapid export of extra-sucrose (Suc) and extra-cytokinin (CK) of the isopentenyladenine (iP) type in the phloem. Suc was found to move both upwards and downwards. When reaching the roots, Suc causes an increased and early upward export of CKs [mainly zeatin riboside (ZR)] and nitrate in the xylem. Recent experimental evidence has indicated that the major function of root CKs is to bring information to the shoot (essentially the leaves) on nitrogen availability (Schmülling, 2002). Apparently, root CKs are then rapidly metabolized in the leaves (Letham, 1994; Faiss et al., 1997). Extra-nitrate is, on the other hand, presumably converted into glutamine (Gln) and putrescine (Put), as these compounds are later exported in greater amounts in the phloem by induced leaves. GAs do not apparently participate in signal movements in Sinapis as their complements and levels remain unchanged in both leaves and shoot tip following floral induction (Corbesier et al., 2004).

Figure 1.

Diagram showing long-distance movements of floral signals in plants of Sinapis alba induced to flower by one long day (LD). Early signals, moving very soon within the photoextension period of the LD, are sucrose and isopentenyladenine (iP) in the phloem and zeatin riboside (ZR) and nitrate in the xylem. Other signals, moving several hours later in the phloem, are glutamine and putrescine.

All the leaf-exported compounds eventually enter the SAM, and we have shown that Suc and CK cause events there that are specifically related to the induction of flowering. For example, Suc increases invertase activity and, later, energy metabolism; both CK and Suc stimulate cell division (Bernier et al., 1993, 2002). In fact, Suc unloaded from the phloem into the SAM is known to be hydrolysed into hexoses by invertases, which in turn have been found to be activated by sugars and/or CKs (Roitsch and Ehneß, 2000; Koch, 2004). Hence, the co-arrival of extra-amounts of Suc and iP at the Sinapis SAM at floral induction provides a mechanism for the amplification of hexose production. Further, the increased input of hexoses may not only stimulate energy-consuming processes in the SAM, but also trigger, together with the CKs, cell divisions via their action on D-type cyclins (Potuschak and Doerner, 2001). An interesting result, in this context, is that over-expression of the CYCLIN D2 gene of Arabidopsis in tobacco plants causes an increase in cell division in the SAM and early flowering (Cockcroft et al., 2000).

The flowering response of SAM to the inductive LD is abolished when the downward movement of Suc is interrupted by girdling, or when the upward movement of ZR and nitrate is prevented by plant exposure to saturating humidity (Havelange et al., 2000), indicating that this long-distance signalling loop is essential for flowering in Sinapis. Interestingly, in plants in which flowering is inhibited by the interruption of long-distance signalling, an application of a CK or Gln directly to the apex substantially restores the flowering response (Havelange et al., 2000; A. Havelange and G. Bernier, unpublished data 2002).

As the levels of Suc and CKs in the sap exported by induced leaves change in the SD plant Xanthium, as in Sinapis (Bernier et al., 1998), it is quite possible that the co-ordinated movement of some nutrients and hormones is part of the controlling system of flowering in this species also.

Long-distance signalling also occurs in so-called ‘self-inductive’ plants which flower irrespective of vernalization and day length conditions. These plants, when grown continuously in defined environmental conditions, produce a constant number of nodes before starting to initiate flowers. Such is the case in day-neutral tobaccos which, depending on the genotype, may produce from 14 to 60 nodes before flowering. Grafting studies have shown that the specific node numbers in these genotypes are in fact controlled by a combination of two different long-distance signals, one of root and the other of leaf origin, as well as by SAM competence to respond to these signals (McDaniel, 1996). Unfortunately, the chemical nature of the signals and the mechanisms of SAM sensitivity were not determined in these studies.

Genetic control of flowering time

Work in this area has been conducted mainly in Arabidopsis, a rosette plant whose flowering is accelerated by LDs, vernalization, a rise in ambient temperature (from about 15 to about 25 °C), a low red/far-red ratio in incoming light and a low mineral availability. The flowering response to these environmental factors involves several signalling pathways that converge towards the regulation of floral meristem identity genes (Mouradov et al., 2002). Two of these genes [LEAFY (LFY) and APETALA 1 (AP1)] were identified first because their mutation clearly perturbs the fate of the SAM productions. Other genes, called ‘integrators’, act upstream of LFY/AP1 and their mutation severely delays flowering in different growing conditions. These genes include FLOWERING LOCUS T (FT) and SUPPRESSOR OF OVEREXPRESSION OF CO1 (SOC1) (see Figure 2).

Figure 2.

Interactions between genes promoting or repressing flowering in Arabidopsis thaliana. Environmental factors which set the different pathways in motion are shown at the top of the figure. The full names of the genes are listed in Table 3. GA, gibberellin.

As expected, the pathways controlling the flowering time in Arabidopsis that were first identified concerned ‘primary’ promoting factors (LDs and vernalization), as the criteria used to screen and characterize flowering time mutants included these environmental controls (Martínez-Zapater et al., 1994). Late-flowering mutants that were delayed in LDs led to the identification of genes of the ‘LD pathway’, while mutants which were still responsive to the photoperiod but impaired in the vernalization response were included in a ‘vernalization pathway’. Mutants that remained sensitive to both environmental factors were then classified as affected in ‘autonomous’ flowering. In addition to this trio, a ‘GA pathway’ was proposed based on the extremely late-flowering phenotype of GA-deficient mutants in SDs.

On the other hand, genes involved in the repression of flowering were identified from early-flowering mutants (Pouteau et al., 2004). These mutants are often very pleiotropic, indicating that the corresponding genes affect many processes, and hence it is difficult to integrate them into specific pathways. Moreover, their sensitivity to environmental factors has not been so well characterized; this is why the corresponding genes are regarded as ‘modulating’ the primary promoting pathways, which still form the framework of Figure 2, rather than identifying specific repression pathways.

Thus, until recently, most models proposed four promoting pathways (Mouradov et al., 2002). In the last couple of years, models have become more and more complex because of an increased interest in ‘secondary’ environmental factors, such as ambient temperature (Blázquez et al., 2003). Some of the most recent reviews no longer show the original quartet of pathways, but separate schemes (Boss et al., 2004). Clearly, because the environmental factors controlling flowering time influence each other, as discussed above, cross-talks are continually being discovered between intervening genes, reflecting plant plasticity. For clarity, genes are confined to their main, or best supported, function (Figure 2) and, to limit quotations, we refer to recent reviews (Mouradov et al., 2002; Périlleux and Bernier, 2002; Boss et al., 2004) unless another reference is given. The full names of the genes are listed in Table 3.

Table 3.  Full names of genes

Primary promoting pathways

The first signalling pathway promotes flowering in response to LDs. It includes genes encoding the photoreceptors phyA and cry2 and the clock components needed for correct circadian time measurement, several clock-associated genes, such as GI, and the downstream gene CO. The abundance of CO is photoperiod dependent and only in LDs does a large amount of protein coincide with the presence of light, sensed by phyA and cry2 (Valverde et al., 2004). This ‘external’ coincidence allows CO to activate its target FT. Other targets of CO are SOC1 and TFL1.

A second or ‘vernalization’ pathway, activated by a period of cold of sufficient duration, acts through repression of the FLC gene, which is itself a strong repressor of flowering. After optimal vernalization, FLC expression is abolished and this repression is stable. The vernalization pathway involves the functions of the VIN3 and MAF2 genes, which ensure that cold periods of insufficient duration will not cause flowering, and of the VRN genes, which are necessary for the maintenance of FLC repression after return to warm temperatures (Ratcliffe et al., 2003; Sung and Amasino, 2004a).

A third pathway was initially found to promote flowering independently of environmental factors and hence was called ‘autonomous’ (Boss et al., 2004). This pathway includes several subsets of independent genes, such as FCA/FY, FVE/FPA, LD and FLD. Their promoting effect on flowering is mainly exerted through their common repressive function on FLC. Thus, the autonomous and vernalization pathways co-operate to down-regulate FLC expression. In contrast, FLC is positively regulated by FRI: in so-called ‘winter’ (late-flowering) accessions having an active FRI gene, repression of FLC by the autonomous pathway is entirely overcome (thus vernalization is absolutely required), whereas ‘summer’ (early-flowering) accessions with a loss-of-function FRI gene are not dependent on vernalization for flowering. FLC-dependent pathways also interact with the photoperiod pathway, as FLC has been shown to regulate negatively CRY2 expression (El Assal et al., 2003). In addition, the VRN1 gene apparently acts in an FLC-independent pathway to positively regulate the downstream gene FT (Boss et al., 2004).

A fourth or ‘GA promotion’ pathway includes genes, such as GA1, GAI, RGA, FPF1 and AtMYB33, with roles in GA biosynthesis or signalling. Applications of physiologically active GAs are known to promote flowering in Arabidopsis, especially in SDs, and their amounts increase in the shoot of plants after transfer from SDs to LDs (Bagnall, 1992; Chandler and Dean, 1994; Xu et al., 1997). Curiously, however, mutations impairing the GA pathway are most inhibitory to flowering in plants grown under SDs, indicating that the importance for flowering of these promotive hormones is greater under unfavourable than under favourable day length conditions. However, recent evidence links the GA and LD pathways. Mutations at the EBS locus cause early flowering in SDs; this phenotype requires GA biosynthesis (Gómez-Mena et al., 2001) and is due to the fact that EBS normally represses the expression of FT (Piñeiro et al., 2003). Another example is given by the SPY gene, which was first demonstrated to be a negative regulator of GA signalling, and which has now been found to interact with the LD pathway upstream of CO (Tseng et al., 2004).

Secondary pathways

Superimposed on these four ‘primary’ pathways, secondary factors have been investigated recently and have been found to involve the same actors (Figure 2).

Ambient temperature has recently been suggested to act through the autonomous pathway, as the fca and fve mutants were found to flower as late at 23 °C as at 16 °C (Blázquez et al., 2003). In the promotive effect of higher ambient temperature, night temperature may be more important than day temperature (Thingnaes et al., 2003).

Light quality is obviously perceived by the photoreceptors involved in the LD pathway (phyA and cry2), which mediate the positive effects of far-red and blue light, respectively, on flowering. Interestingly, the mutant phenotype of cry2 is strongly accentuated at 16 °C (Blázquez et al., 2003), and hence CRY2 function is influenced by ambient temperature. On the other hand, PHYB acts as a repressor of flowering, as the phyB mutant exhibits a strong early-flowering phenotype. The divergent effects of these different photoreceptors find some explanation in the fact that they exert mutual control on each other (Mockler et al., 2003). PhyB is the major contributor to the physiological responses of green seedlings to red light and is involved in shade avoidance (Smith and Whitelam, 1997). Interestingly, the early-flowering phenotype of phyB is also temperature sensitive: at 16 °C, the phyB phenotype is completely lost; hence, phyB is completely inactive and replaced by phyE (Halliday et al., 2003). PHYB is thus believed to act in ‘light quality’ and ‘ambient temperature’ pathways, and it has recently been suggested that PHYB does so by regulating FT activity via an intermediate gene PFT1 (Cerdán and Chory, 2003). However, PHYB is also believed to interact with the autonomous pathway, as the early-flowering phenotype of phyB requires FCA function (Koornneef et al., 1998).

Photoreceptors may also affect flowering time indirectly through their effects on the biological clock. Clock entrainment is known to involve different photoreceptors in different light conditions (quality and irradiance), and novel blue light photoreceptors of the ZTL/FKF1/LKP2 family have recently been shown to regulate clock components and downstream genes acting in the photoperiod pathway (Boss et al., 2004).

Light irradiance may also influence flowering through photosynthesis. The phyA and fca mutants respond much more to increased irradiance than do wild-type plants (Bagnall, 1992; Bagnall and King, 2001), indicating that the promotive effect of the corresponding wild-type genes may be, at least partly, mediated through photosynthesis. Unfortunately, we still do not know how other late-flowering mutants would react to irradiance.

Integrator and floral meristem identity genes

As shown in Figure 2, the four ‘primary’ promotion pathways regulate the activity of ‘integrator’ genes, which are also under the influence of repressors, ensuring fine tuning of the process.

FLC is a repressor whose activity is balanced between the genetic reinforcing effect of FRI and the weakening effects of the autonomous and vernalization pathways.

FT is a promoter whose activity is mainly up-regulated by LDs through CO, but may also be up-regulated by the interconnected autonomous, light quality and ambient temperature pathways. Genes involved in the vernalization pathway also act on FT, which is down-regulated by FLC, but up-regulated by VRN1 independently of FLC. TFL2 is another repressor of FT activity (Kotake et al., 2003).

SOC1 is up-regulated by CO and the GA pathway and down-regulated by FLC. Recently, SOC1 has also been found to cross-talk with AGL24, a promoter of inflorescence fate, whose activity is induced by vernalization independently of FLC (Michaels et al., 2003). Interestingly, AGL24 is close in sequence to SVP, which exerts an antagonizing repression effect on flowering. SVP and another gene, MAF1 (or FLM), are thought to be part of an autonomous pathway exerting its function in repressing the floral transition independently of FLC and interacting with the photoperiod pathway (Hartmann et al., 2000; Scortecci et al., 2003).

Further downstream, FT and SOC1, together with the GA pathway, co-operate in the up-regulation of the floral meristem identity genes LFY/AP1. These genes up-regulate each other and are essentially expressed in the lateral productions of the SAM where individual flowers form. In the centre of the SAM, LFY/AP1 activity is antagonized by TFL1, whose expression maintains the indeterminacy of the SAM (Ratcliffe et al., 1999). LFY/AP1 also repress AGL24, which promotes inflorescence fate rather than flower formation (Yu et al., 2004). Thus, spatial patterning is finely regulated in the SAM on floral transition to organize the inflorescence architecture.

Localization of pathway activities within the wild-type plant

Most genes controlling flowering time are expressed across a wide range of organs and tissues, but a survey of available data on their spatial expression patterns reveals that many genes show preferential expression in more limited areas. Table 4 shows the result of a careful re-examination of published pictures (including references); some uncertainty remains, however, as localization studies were occasionally inconsistent, possibly because of the different techniques used.

Table 4.  Localization of expression of genes controlling flowering time in wild-type plants
 Gene expression
High in SAM and/or surrounding tissuesHigh in RAMIn vascular tissues
Long day pathwayPHYA1, CRY21, CO2,3PHYA1, CRY21CO3,4
Autonomous/vernalization pathwayFCA5, FY6, LD7, FLD8, VIN39, VRN26, FLC9, FRI6, PIE110, AGL2411FCA5, FY6, LD7, FLD8VIN39, VRN26, FLC9,FRI6, AGL2411FCA5, AGL2411
Gibberellin pathwayGA112, FPF113, AtMYB3315GA112, SPY14, AtMYB3315GA112, AtMYB3315
Integrator and floral meristem identity genesSOC1 (AGL20)16,17, LFY18,19, AP118,20 SOC117, FT4, LFY18,19
RepressorsPHYB1, TFL24,21, MAF1 (FLM)22, SVP23, TFL12,20PHYB1, TFL24,21, MAF1 (FLM)22, SVP23TFL24,21

Genes of the photoperiod promotion pathway could have been expected to show preferential expression in expanded leaves as day length is classically believed to be principally perceived by these organs (see above). The situation is in fact different: the genes encoding PHYA and CRY2 are most expressed in the SAM and root apical meristem (RAM) of the vegetative seedling and less expressed in the hypocotyl and cotyledons. The downstream gene CO is expressed relatively weakly in SAM and leaf primordia and also, quite strikingly, strongly in the vascular tissues (mainly phloem) of the hypocotyl, expanded leaves and roots.

For most genes of the autonomous and vernalization pathways, including FLC, the highest expression was observed in both the SAM and RAM, i.e. in areas in which cell division is mostly active in vegetative plants. These observations are consistent with the fact that vernalization is known to be perceived by the SAM. At the molecular level, events are also well confined, as many genes of these pathways are involved in the stable modification of FLC chromatin structure, leading to its mitotically stable repression (Boss et al., 2004; Sung and Amasino, 2004a). We have also seen above that genes of the autonomous pathway seem to be involved in sensing ambient temperature, and it is known that temperature influences the cell division process, and hence meristematic zones.

Work on the expression pattern of genes involved in the GA pathway has revealed that the GA biosynthesis gene GA1 is essentially expressed in vegetative plants just below the SAM, and also in the veins of expanding and mature leaves and in the RAM. During the transition to flowering, GA1 is strongly activated in the inflorescence SAM. Expression of the GA sensitivity FPF1 and GAMYB-like gene, AtMYB33, also strongly increases in the SAM during this transition. These findings suggest that the amount of GAs and the activity of the GA pathway are tightly restricted in the SAM at the vegetative state, but become unrestricted at the floral transition.

Of the integrator genes, FT is active exclusively in the phloem tissues of the leaves and inflorescence stem. SOC1 and LFY are weakly active in the apical tissues of vegetative plants, but their expression is strongly up-regulated during floral transition and occurs in the SAM and surrounding leaf primordia. Later, LFY and AP1, but not SOC1, are expressed in incipient floral meristems; SOC1 and LFY are also expressed in the vasculature of the incipient inflorescence stem.

Some floral repression genes are ubiquitously expressed, whereas others are mostly expressed in the SAM, RAM and leaf primordia (Table 4), as is the case for PHYB, MAF1, SVP and TFL2. Contrary to TFL2, which is similarly expressed at all developmental stages, MAF1 and SVP are down-regulated and TFL1 is up-regulated in response to floral induction. Up-regulation of TFL1 is early as it precedes LFY and AP1 activation and, remarkably, is restricted to a small area located just below the SAM.

In conclusion, it appears that: (i) of the genes acting in the promotion or repression of flowering, none (except FT) is exclusively expressed in expanding and/or expanded leaves before and during the floral transition; (ii) most genes (except CO and FT) are preferentially expressed inside and/or around the SAM, which thus appears to be fairly autonomous for flowering time control.

Given the large number of genes expressed in the RAM (Table 4), the root system is presumably capable of reacting to some critical environmental changes and, as a result, influences to some extent the flowering process. Alterations in root size, morphology or function are indeed observed in the fca, gai, spy and tfl2 mutants, but not in co (Macknight et al., 2002; Swain et al., 2002; Kotake et al., 2003). Strikingly, the root alterations in fca are suppressed by vernalization, just as the delay in flowering time, suggesting that both phenotypes are intimately related, possibly through FLC regulation. Expression of FLC in the RAM was unexpected, as there is no indication that cold treatment of roots alone is capable of promoting flowering in intact plants. Recently, PIE1 (a positive regulator of FLC) was found to be expressed in the SAM only, shedding some light on a possible discrimination in FLC activity in SAM and RAM (Noh and Amasino, 2003). However, more work on the root functions in the flowering process in Arabidopsis is warranted.

Of particular interest are the genes active in the vascular tissues (Table 4), especially the phloem, as it is known that florigenic and antiflorigenic signals are transported in these tissues (Lang, 1965; Bernier et al., 1981). This raises the question of whether the products of these genes might be the signals themselves, or precursors. Such an hypothesis makes sense only for early expressed genes, such as CO, FT, and GA1. For others, such as AGL24, SOC1 and LFY, which are clearly expressed in the vasculature after completion of the floral transition, the start of this expression pattern has not been determined precisely.

Long-distance signals

At this point, physiological and genetic work have arrived at different conclusions. On the one hand, the physiological studies summarized above have demonstrated the existence of long-distance signals moving up and down the plant in the phloem and xylem saps and participating in the flowering process at the SAM. In Arabidopsis, the movement of one or several floral signals from the leaves to SAM was inferred from sequential defoliation experiments (Corbesier et al., 1996). On the other hand, the Arabidopsis SAM seems to be fairly autonomous as most genes controlling flowering time are active there. The autonomy of the SAM is further supported by the fact that excised shoot apices of various plants are capable of perceiving several environmental factors controlling flowering, but only if provided with sugar and minerals in the culture medium (see above). Thus, a simplistic model of flowering control would be that the plant only provides nutrients to the SAM. The predominant role attributed classically to expanded leaves in day length perception could then be explained on the basis that they are the major providers of nutrients to the SAM (Périlleux and Bernier, 2002).

As far as minerals are concerned, it has been known for some time that a reduction in their supply promotes flowering in Arabidopsis (Lang, 1965). This has been confirmed more recently by the observation that increasing mineral supply to the roots delays flowering in several mutants of the photoperiod and autonomous pathways, as well as in wild-type plants (van Tienderen et al., 1996). An important part of this inhibition is presumably due to nitrogen (Bernier et al., 1981). An explanation for this puzzling effect might come from our own result in Sinapis, where the increased export of Suc towards the SAM at the floral transition (see Figure 1) is greatly reduced when plants are grown on high nitrate-supplemented medium (Corbesier et al., manuscript in preparation).

There is plenty of evidence that Suc supply to the SAM is essential for flowering in Arabidopsis. First, the Suc level in the phloem sap exported by leaves increases early and markedly during floral induction and, when this increase is prevented, as in the starch-deficient pgm mutant, flowering is inhibited (Corbesier et al., 1998). Second, Suc application to wild-type plants grown in suboptimal conditions for photosynthesis, as well as to the late-flowering phyA mutant, promotes flowering (King and Bagnall, 1996; Bagnall and King, 2001). Third, Suc supply to the aerial part, presumably the SAM, of plantlets grown in vitro almost completely suppresses the late-flowering phenotype of mutants, such as gi, co and fca, but is unable to rescue ft (Roldán et al., 1999). This suggests that Suc is involved somewhere in between the LD and autonomous signalling pathways, but upstream of FT. Fourth, Suc is required for the up-regulation of LFY by exogenous GA (Blázquez et al., 1998).

However, Suc may not be the only missing signal needed by the SAM to achieve flower initiation, and the fact that the CO and FT genes are active in the phloem of mature leaves has motivated careful experiments to see whether their products (transcripts or proteins) could be translocated in the plant.

With regard to CO, different promoters were fused to the coding region to target CO activity in complementation experiments. Remarkably, although CO is expressed in the SAM of wild-type plants, misexpression of CO in the SAM does not rescue the late-flowering phenotype of the co mutant, but its expression in the companion cells of the phloem does (An et al., 2004). The effect of CO misexpression in the phloem has been found to involve cell autonomous activation of FT. On the other hand, FT misexpression, either in the phloem or in the SAM, can correct the late-flowering phenotype of co plants, indicating that CO promotes flowering by up-regulation of FT in the phloem, but that the effectiveness of FT in promoting flowering is not restricted to these cells (An et al., 2004). The FT protein is only 23 kDa, which is below the size exclusion limit of plasmodesmata (Imlau et al., 1999); hence, it might move freely between cells from the ends of provascular strands towards and inside the SAM (An et al., 2004). FT then interacts in the SAM with the product of the FD gene (Daimon et al., 2004), and this interaction eventually up-regulates AP1. Plants misexpressing CO in phloem cells, in a background in which FT is inactive, are still capable of flowering, confirming that CO can promote flowering through FT-independent processes, e.g. through SOC1 and LFY (Figure 2; An et al., 2004).

Although these results suggest that FT may be a mobile signal in the LD signalling pathway, micrografting experiments came to a different conclusion. Indeed, it was observed that, although the late-flowering phenotype of a gi or co receptor was partially rescued by grafting with a wild-type donor shoot, the response of an ft receptor was much less convincing (An et al., 2004; Turnbull and Justin, 2004). Together with the previous observation that Suc corrects the flowering time of gi or co in vitro, but fails to do so for ft, these data support the idea that the floral-promoting material provided by the wild-type donor is Suc and not FT. This discrepancy emphasizes that results of misexpression experiments must be interpreted cautiously, especially when misexpression is higher than native expression in the target tissue (Turnbull and Justin, 2004).

The participation of other compounds, particularly GAs, CKs and abscisic acid (ABA), in long-distance signalling in Arabidopsis should also be considered. The importance of GAs in the control of flowering is well established (Figure 2), but whether these hormones originate from distant plant parts and/or are synthesized in the SAM is not clear. Indeed, it was found that the GA biosynthetic GA1 gene is expressed in the veins of expanding and expanded leaves, in the RAM and in tissues below the SAM before the transition to flowering. Thereafter, expression extends to the SAM itself (Silverstone et al., 1997).

CKs are also promoters of flowering in Arabidopsis, as shown by the fact that transgenics that are deficient in CKs are late flowering (Werner et al., 2003), whereas plants that are enriched in CKs are early flowering (Chaudhury et al., 1993). Because CKs affect the rate of leaf initiation more than the flowering time, these results have sometimes been overlooked when the flowering time is expressed in terms of the leaf number, but this difficulty must not mask the real promotive effect of these hormones on flowering, as discussed elsewhere (Bernier, 2003). Plants deficient in CKs may even stay vegetative until death, as found with some transgenics over-expressing a CK oxidase/dehydrogenase gene (Werner et al., 2003), or with some triple mutants lacking three histidine kinase CK receptors, whilst single or double mutants have no flowering phenotype (Nishimura et al., 2004). In wild-type plants, CK application accelerates flowering, but only when irradiance is low, indicating that this promotive effect is dependent on sugars (Dennis et al., 1996). The promotive effect of CKs is also more important in LDs, as shown by multiple approaches: (i) the CK-enriched amp1 mutant rescues the late-flowering phenotype of the gi mutant, but not of fca, suggesting that these hormones act essentially in relation with the LD pathway downstream of GI (Dennis et al., 1996); (ii) the amounts of iP-type CKs increase in LD-induced Arabidopsis plants in the leaves, in the phloem sap and in the SAM (Corbesier et al., 2003), where they may activate cell proliferation, a very precocious event of the SAM transition to flowering (Jacqmard et al., 2003). Contrary to the situation presented above for GAs, CKs detected in the SAM are probably transported from other plant parts, as none of the CK biosynthetic genes encoding ATP/ADP isopentenyltransferases (IPT) are expressed in the SAM itself (Miyawaki et al., 2004). The main source of CKs arriving at the SAM is probably the expanded leaves as the IPT3 gene is specifically expressed in the phloem cells of these organs.

ABA is regarded as a general inhibitor of flowering (Bernier, 1988), and this is confirmed in Arabidopsis where mutants deficient in or insensitive to ABA are early flowering in SDs (Martínez-Zapater et al., 1994). Interestingly, two ABA-deficient mutants, aba2 and aba3, and an ABA-insensitive mutant, abi4, are allelic to sugar-insensitive mutants, indicating that signalling pathways mediated by ABA and sugars interact to regulate plant development (Gibson, 2004).

Other substances might be involved in signalling during the transition to flowering. Mutants deficient in salicyclic acid or insensitive to ethylene are indeed late flowering (Boss et al., 2004). Brassinosteroids have recently been suggested to be involved in the autonomous pathway, but also to co-operate with GAs in controlling flowering time (Domagalska et al., 2004).

An integrated model of flowering time control

If we summarize the achievements identifying the elusive ‘florigen’ involved in the floral transition of wild-type Arabidopsis in LDs, it appears that it could be formed of both long-distance and short-distance signalling molecules: potential long-distance signalling molecules (Suc and CK) have been identified by the analysis of phloem sap exported by leaves in response to floral induction and by genetic approaches (FT may move in the phloem from leaf veins to SAM); short-distance signalling within the SAM is suggested by the fact that most components of the genetic machinery controlling flowering (except CO and FT) are highly active in or nearby the SAM (Table 4). As discussed earlier, the question of whether GAs act as a long-distance signal of leaf origin or as a short-distance signal produced locally, or both, is unresolved.

At this stage, we propose a model, shown in Figure 3, based on the data presented above and integrating events occurring in wild-type Arabidopsis plants in LDs; this model holds for plants with non-inhibitory FLC expression levels (summer early-flowering accessions or winter vernalized plants).

Figure 3.

A model of control of the shoot apical meristem (SAM) floral transition in Arabidopsis thaliana plants exposed to an inductive long-day (LD) treatment. This transition is supposed to occur in response to arrival at the SAM of long-distance signals of leaf origin and their interactions with short-distance signals produced in the SAM itself. Long-distance signals include sucrose (Suc), the cytokinin isopentenyladenine (iP) and the FT protein produced following activation of the CO gene. In the case of gibberellins (GAs), it is unclear whether they are imported from leaves or produced locally.

In our model, Suc has a dual role. In addition to the direct role of Suc in the floral transition, we hypothesize that Suc plays a crucial, albeit indirect, role in flowering, as it is the major component of phloem sap in most plants and so controls the mass flow of solutes operating between source leaves and sinks. We believe that Suc loading in leaf phloem and unloading at the SAM are important checkpoints in the control of SAM flowering, and it has previously been reported that the increased export of Suc by Arabidopsis leaves in response to LD induction might be due, at least in part, to increased efficiency of Suc loading (Corbesier et al., 1998).

After unloading in the SAM, the long-distance signals entrained in the Suc streaming stimulate a number of cellular and molecular events (Bernier, 1988). Suc is first hydrolysed by local invertases. Although the general belief is that vacuolar invertase plays a prominent role in meristematic areas (Koch, 2004), participation of cell wall invertase activity may have been overlooked. Indeed, it has been observed recently that the flowering of Arabidopis plants in LDs can be accelerated by the over-expression of cell wall invertase in the SAM, while flowering in SDs is not modified (Heyer et al., 2004). On the contrary, transgenics having an increased cytosolic (vacuolar) invertase activity are delayed in flowering, in both LDs and SDs.

CKs activate invertase activity and, together with the products of Suc hydrolysis, increase the rate of cell division (see above). Hexoses also participate with GAs in the up-regulation of LFY expression, whilst the other floral meristem identity gene, AP1, is activated by FT, which is itself positively regulated by CO. In Figure 3, FT is speculated to move from leaf to SAM in the phloem, whilst CO is unable to move out of the phloem (An et al., 2004). A difficulty remains, however, in the fact that flowering can be promoted by misexpression of CO in the phloem, independently of FT. This suggests that CO may activate other targets in the phloem. SOC1 is a candidate but, although expression in the wild-type plant was reported in the vasculature of the nascent stem at the floral transition (Borner et al., 2000), SOC1 activity is much higher in the SAM. Activation of SOC1 in the SAM might thus be due to other signals, possibly a CK or a GA (Bonhomme et al., 2000; Borner et al., 2000).

The model of Figure 3 is obviously not complete. More long-distance signals presumably operate during LD induction – for example Gln (Corbesier et al., 2001), whose function is unknown – as well as more short-distance signals. As several genes expressed in the SAM were found to control its spatial patterning, the short-distance signalling presumably involves some of their products (RNA and protein), as suggested recently (Wu et al., 2003). However, it is clear that we are getting closer and closer to the identification of the elusive multifactorial ‘florigen’, at least in Arabidopsis.


Most of the work on long-distance signalling in Sinapis alba and Arabidopsis thaliana performed in our laboratory was supported by Interuniversity Poles of Attraction Programmes (Belgian State, Prime Minister's Office – Federal Office for Scientific, Technical and Cultural Affairs; P4/15 and P5/13), the European Community Contract BIO4 CT 97-2231 and the University of Liège (Fonds Spéciaux).