How Plants Decide When to Flower: The Strange Photoperiodism of Florigen
A plant in autumn must somehow know it is autumn. The signal it uses is daylength: most temperate plants are short-day species that flower as days shorten, or long-day species that flower as days lengthen. The mechanism for this calculation was hypothesized by Mikhail Chailakhyan in 1937, who proposed that leaves measure daylength, produce a signal he called florigen, and transmit it through the plant to the shoot apex where flowering would begin. The hypothesis was strong: leaves of a single plant could be exposed to short days while the rest of the plant remained on long days, and flowering would still occur, demonstrating that the signal was mobile. Various plant species could even be grafted together, and a leaf on one plant could trigger flowering on another via the graft junction. The signal was clearly a chemical messenger, transmissible through the vascular system.
The molecular identity of florigen, however, resisted identification for nearly 70 years. Decades of work by hundreds of researchers narrowed it down to a small mobile molecule, but extraction attempts failed repeatedly. The chemistry was wrong, the candidate molecules failed to recapitulate the effect, and by the 1990s the field had partially given up on the chemical-signal hypothesis and started suspecting that florigen might be a more complex signal, perhaps involving sugars or hormones acting in concert. The actual answer, identified by 2005 work from George Coupland's lab and confirmed by 2007 work demonstrating mobility, was that florigen is a protein, specifically the product of the FT (FLOWERING LOCUS T) gene, and that 70 years of failed chemical extraction reflected the difficulty of detecting a protein produced in vanishingly small quantities in specific cells of specific leaves at specific times of day.
The architecture of the calculation
The full mechanism, as understood by 2026, is among the more elegant pieces of plant molecular biology. The cells of the leaf contain a circadian clock that produces rhythmic expression of the CO (CONSTANS) gene with a peak in the late afternoon. The CO protein is rapidly degraded in darkness, so on short days the protein never accumulates. On long days, the late-afternoon peak of CO expression coincides with daylight, the protein accumulates rather than being degraded, and accumulated CO activates expression of the FT gene. FT protein is then loaded into the phloem and transported to the shoot apex, where it binds to a transcription factor called FD and triggers the cascade of gene expression that converts vegetative meristem to flowering meristem.
The calculation has the structure of a daylength threshold detector with built-in adjustability. The threshold is set by the timing of the CO expression peak relative to dusk, which is itself controlled by the circadian clock and can be tuned by selection. Different plant varieties have evolved slightly different clock parameters and slightly different CO-promoter sequences, which produce slightly different daylength thresholds. The threshold can be sharp enough that a 30-minute change in daylength is the difference between flowering and not flowering, which is what allows plants at different latitudes to flower at roughly the same calendar date despite experiencing different daylengths.
The short-day plants do something more subtle. The same machinery is present, but the regulatory connections are inverted: CO accumulation is suppressed by light rather than enabled by it, so FT expression occurs at night and only when night is long enough. The genetic architecture is conserved across plant species; what varies is the wiring of CO to FT and the threshold parameters of the circadian clock. This is why a single set of molecular components can produce the full range of photoperiodic responses observed across plant species, from strict short-day plants like rice and soybean to strict long-day plants like Arabidopsis and wheat to day-neutral plants like tomato that have largely decoupled flowering from photoperiod.
The graft experiments that established mobility
The strongest evidence that florigen is FT protein came from a series of grafting experiments in the mid-2000s. The principle was simple: if FT protein is the mobile signal, then a plant unable to produce FT should still flower if grafted to a plant that can produce FT, and the FT protein produced in the donor should be detectable in the recipient. The experiments were technically demanding because FT is produced in such small quantities and because plants of different species do not all graft easily, but by 2007 multiple labs had independently demonstrated that GFP-tagged FT protein could be detected in the shoot apex of recipient plants after grafting to FT-expressing donors. The signal moved through the phloem in measurable quantities, and the recipient plants flowered on a timeline matching when the donor plant's photoperiodic conditions would have triggered flowering.
The follow-up experiments showed that FT protein from one species can trigger flowering in another species at the graft junction, which is the molecular explanation for the classical observation that cross-species grafts can induce flowering. The mechanism is conserved well enough that Arabidopsis FT can trigger flowering in tobacco, and tobacco FT can trigger flowering in Arabidopsis. The receptors at the shoot apex are similarly conserved, and the downstream transcription factor cascades have enough cross-species compatibility to make the system functionally interchangeable across most flowering plants.
The downstream cascade
What happens at the shoot apex after FT protein arrives is the next chapter of the same story, and it is almost as elegant. FT binds to FD, a transcription factor expressed specifically in the shoot apical meristem. The FT-FD complex activates expression of several flowering-control genes, including SOC1 (SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1) and the floral meristem identity genes LFY (LEAFY) and AP1 (APETALA1). The combined action of these transcription factors switches the shoot apical meristem from vegetative growth, where each new cell becomes part of a leaf or stem, to floral growth, where each new cell becomes part of a flower.
The transition is essentially irreversible. Once the meristem has committed to flowering, it cannot return to vegetative growth in the same place. A plant can produce multiple inflorescences from different meristems, and most plants do, but each meristem makes the transition once and stays committed. The reason is that the floral meristem identity genes activate each other in positive feedback loops, and once the loops are engaged the cells lock into the new identity. Disrupting any one of the genes does not break the loop because the others maintain it.
The same machinery also drives the production of the floral organs themselves: the four-whorl arrangement of sepals, petals, stamens, and carpels that is the canonical structure of a flower. The ABC model of flower development, established in the 1990s by work in Arabidopsis and snapdragon, identifies three classes of transcription factors whose combinatorial expression determines which organ forms in which whorl. A whorl with only A-class expression becomes sepals; A plus B becomes petals; B plus C becomes stamens; C alone becomes carpels. Mutants that lose one class of expression produce flowers with predictable substitutions, and the pattern is conserved across angiosperms broadly enough that the same genes can be identified in essentially every flowering plant.
The applied surface
The agricultural implications of understanding the FT mechanism have been substantial. Crop varieties bred for different latitudes are mostly varieties with different FT alleles or different CO promoter sequences. Tropical varieties of rice, soybean, and other major crops are typically short-day plants that fail to flower at high latitudes; temperate varieties have been bred to flower at longer daylengths. Modern breeding programs use molecular markers in or near FT to select for desired daylength responses, and several major commercial varieties have been engineered to express FT under controllable promoters that can be triggered by an applied chemical to synchronize flowering for hybrid seed production.
The same understanding has been used in greenhouse horticulture to manipulate flowering for ornamental and commercial purposes. Chrysanthemums, poinsettias, and many other ornamentals are short-day plants whose flowering is triggered by interrupting the long days of summer with dark cycles, or whose flowering is suppressed under long days by extending photoperiod with artificial light. The molecular understanding of why this works has not changed the practice, which predates the molecular work by decades, but it has refined the timing.
The wheat case is particularly interesting because wheat flowering is influenced both by photoperiod (via the FT homolog VRN3) and by vernalization, the cold-requirement system that prevents wheat from flowering until after it has experienced winter. The vernalization pathway involves a different gene (VRN1), but VRN1 and VRN3 interact with each other and with the photoperiod sensing system to produce the integrated response that lets wheat flower in spring after winter cold exposure regardless of latitude. The molecular details of this integration were largely worked out between 2003 and 2015, and the resulting understanding has fed back into breeding programs that produce varieties optimized for specific climates.
What this is and is not
The honest summary: a plant in autumn knows it is autumn because the cells of its leaves contain a circadian clock that produces rhythmic protein expression, the timing of the rhythm relative to dawn and dusk encodes daylength, the photoperiod-sensitive protein activates production of a mobile signal protein, and the signal protein travels through the phloem to the shoot apex where it triggers a transcription factor cascade that commits the meristem to flowering. The molecular components have been identified, the mechanism has been validated by grafting and genetic experiments, and the system is conserved enough across plant species that the same set of genes explains the photoperiodic responses of essentially all flowering plants.
The mechanism is elegant in a way that 70 years of botanists could not have predicted from the phenomenological description, which is itself a useful case study in how molecular biology can transform a field. The pre-molecular literature on photoperiodism is enormous and largely correct in its phenomenological observations but mostly wrong about the chemistry. The molecular work since 2005 has both confirmed the older observations and revealed that the actual implementation is a particular kind of digital signal processing on top of an analog daylength input, which is not the kind of architecture the field was looking for. The lesson is one that recurs across biology: the systems are more sophisticated than the phenomenology suggests, and the only way to know what they are doing is to look at the molecules.