Activation Tagging of the Floral Inducer FT

The transition from the vegetative to the flowering phase ofArabidopsis is controlled by several genetic pathways that monitor the developmental state of the plant as well as environmental conditions (1). Despite the cloning of severalArabidopsis genes participating in these pathways, substantial gaps remain in our knowledge of how the signals controlling flowering are transduced and integrated. To complement other approaches to the study of floral induction, we applied activation tagging to whole plants. Activation tagging is a random overexpression screen that was developed several years ago for isolated plant cells. In this scheme, transcriptional enhancers from the viral 535S promoter are randomly inserted in the genome with transferred DNA (T-DNA) ofAgrobacterium (2). Using activation tagging, we identified a mutant, 1733, that flowered early, independently of day length (3). In addition, it had terminal flowers. Adjacent to the 1733 T-DNA insertion was an overexpressed gene that, when linked to either the original 35S enhancers (Fig. 1A) or the full 35S promoter and reintroduced into wild-type plants (3), recapitulated the 1733 phenotype (Table 1and Fig. 2). Because the 1733 insertion mapped close to the flowering-time gene FT, we sequenced genomic DNA corresponding to the tagged gene from three ethylmethane sulfonate (EMS)–induced ft alleles (4). All three contained mutations in the open reading frame of the tagged gene (Fig. 1A) (3), indicating that the 1733 mutant carried a dominant, early flowering allele ofFT, whose recessive alleles cause late flowering (4).

• Download high-res image
• Open in new tab
• Download Powerpoint

• Download high-res image
• Open in new tab
• Download Powerpoint

Late-flowering mutants have been functionally grouped by their environmental responses and their genetic interactions (4, 5). One class, which flowers much later than wild type in long days, includes the recessive co andft mutants and the dominant fwa mutants. In contrast to co mutants, ft and fwa not only flower late in long days but are also moderately delayed in short days. The two groups, FT/FWA and CO, also interact differently with the meristem-identity gene LFYbecause only co mutations affect transcriptional induction of LFY (6, 7). In addition,ft lfy and fwa lfy but not co lfydouble mutants have a phenotype that is associated with loss of expression of the meristem-identity gene APETALA1(AP1), indicating that FT and FWAact redundantly with LFY to regulate AP1(8, 9). AP1 was expressed precociously in 35S::FT plants, but in contrast to 35S::LFY, AP1 expression was confined to floral primordia (Fig. 3, A and D), suggesting that FT regulates AP1expression less directly than the LFY transcription factor (10).

• Download high-res image
• Open in new tab
• Download Powerpoint

Because changes in FT levels affected flowering, as deduced both from the 35S::FT phenotype and the semidominant nature of ft mutants (Table 1), we determined whether CO or FWA regulates FT mRNA accumulation. In both long and short days, FT levels in wild type increased from young seedlings to older plants, with higher overall levels in long days (Fig. 1C) (3). WhereasFT expression profiles were similar in wild-type andfwa-2 plants grown in long days, FT expression was reduced in co-2 seedlings, rising to wild-type levels in older plants (Fig. 1, D and E). These data suggest that COfunctions partially upstream of FT and that FWAacts downstream of or in parallel with FT (3).

We complemented the expression studies by testing how constitutive FT expression affected the co-2 orfwa-2 mutant phenotypes. Even though co mutants have a more severe phenotype in long days than ft mutants,35S::FT could completely suppress theco-2 phenotype. 35S::FT also masked the effects of CO overexpression (6), confirming that changing CO activity had no effect in a35S::FT background (Table 1 and Fig. 2B). Although these interactions would normally suggest that FT is the only downstream effector of CO, FT andCO interact differently with LFY(6–9), which argues against a simple linear hierarchy from CO through FT to flowering. A possible explanation is that increased activity of the FT-dependent pathway can compensate for reduced activity of a parallel, normallyFT-independent pathway in co-2 mutants. Consistent with such a scenario, 35S::FT caused precocious induction of LFY mRNA (Fig. 3, B and E), even though FT is not normally required for LFYexpression (7, 8). Unlike35S::FT co-2 plants, 35S::FT fwa-2 plants flowered much later than wild type (Table 1 and Fig. 2B), consistent with FWA affecting events downstream ofFT (3).

These and other findings indicate that FT andLFY have parallel functions downstream of the long-day–dependent and –independent pathways of floral induction (1, 7, 8, 11). Indeed, in contrast to plants that overexpressed only FT orLFY (12), the vegetative phase was bypassed in35S::FT 35S::LFY plants, which produced a terminal flower immediately upon germination. The only leaves produced by these plants were the first two leaves, which are already initiated in the embryo (Fig. 2, C, E, and I, and Table 1). A less marked effect was seen in 35S::FT 35S::AP1 plants (Fig. 2C and Table 1), even though35S::AP1 plants on their own flowered considerably earlier than did 35S::LFY plants (7,12–14) (Table 1). The 35S::FT 35S::LFY phenotype was also more severe than that of 35S::AP1 35S::LFY plants (14), indicating that FT does not only induce AP1, which is confirmed by a failure of anap1 mutation to suppress early flowering of35S::FT plants (Table 1).

The deduced FT protein belongs to a small family ofArabidopsis proteins, which includes the TFL1 protein, whose amino acid sequence is more than 50% identical to that of FT (3, 15). FT and TFL1 have opposite effects on flowering. Loss of FT function causes late flowering (4), whereas loss of TFL1causes early flowering along with the formation of terminal flowers (16). However, FT and TFL1effects are not entirely mirror images of each other, because35S::FT plants flower much earlier thantfl1 loss-of-function mutants, particularly under short days, and 35S::TFL1 plants not only flower late, as do ft loss-of-function mutants, but they also show transformation of individual flowers into shootlike structures (17).

To clarify the relation between FT and TFL1, we tested whether FT promotes flowering by eliminatingTFL1 activity. 35S::FT tfl1-1 plants flowered even earlier than 35S::FT plants and often formed only a single, terminal flower on the main shoot, indicating that TFL1 is still active in35S::FT (Table 1 and Fig. 2, A and H). Consistent with this finding, TFL1 was expressed in35S::FT plants (Fig. 3, C and F). Independent action of FT and TFL1 was likewise evident from the fact that 35S::TFL1 attenuated the early flowering of 35S::FT, even though the attenuation was modest (Table 1 and Fig. 2A). Together, these observations suggest that FT and TFL1 act at least partially in parallel.

TFL1 mRNA is highly expressed in a small group of shoot meristem cells (15) (Fig. 3F). Using reverse transcriptase polymerase chain reaction (RT-PCR), we detected FT mRNA throughout the aerial part of the plant (Fig. 1B). In situ hybridization revealed no specific concentration of FTtranscripts at the shoot apex, suggesting that FT andTFL1 do not have to be expressed in the same pattern to antagonize each other's effects.

FT and TFL1 are related to a membrane-associated mammalian protein that can bind hydrophobic ligands (18). This protein also gives rise to hippocampal cholinergic neurostimulating peptide (HNCP), which is generated from its precursor by cleavage after amino acid 12 (19). Comparison of the FT and TFL1 sequences (3) with the crystal structure (20,21) of HCNP precursor, also called phosphatidylethanolamine binding protein (PEBP), revealed several interesting features. The conserved residue Arg¹¹⁹has been proposed to activate the bond between Leu¹² and Ser¹³ for cleavage of HCNP (20). Arg¹¹⁹ is important for FT function as well, because this residue was changed to histidine in the strong ft-3 allele (Fig. 1A). It has also been proposed that access to the PEBP ligand-binding site is regulated by a COOH-terminal α helix (20, 21). A missense mutation in ft-1close to the COOH-terminus indicates that this region is critical for FT as well (Fig. 1A).

In summary, FT and TFL1 encode related proteins with opposite effects on flowering. Similarly to FT, its antagonist TFL1 is positively regulated by CO(6), suggesting that the balance betweenFT and TFL1 activity serves to fine tune the response to floral inductive signals (3). It remains to be determined how far the sequence similarity between FT, TFL1, and mammalian PEBP reflects similar biochemical modes of action.

Note added in proof: Human PEBP was recently shown to be identical to RKIP (Raf kinase inhibitor protein), which regulates the activity of the RAF/MEK/ERK signal transduction pathway (22).