Analysis of conifer FLOWERING LOCUS T/TERMINAL FLOWER1-like genes provides evidence for dramatic biochemical evolution in the angiosperm FT lineage


Author for correspondence:

Ove Nilsson

Tel: + 46 90 786 8487



  • In flowering plants, homologs of the Arabidopsis phosphatidylethanolamine-binding protein (PEBP) FLOWERING LOCUS T (FT) are key components in controlling flowering time. We show here that, although FT homologs are found in all angiosperms with completed genome sequences, there is no evidence to date that FT-like genes exist in other groups of plants.
  • Through phylogeny reconstructions and heterologous expression, we examined the biochemical function of the Picea (spruces) and Pinus (pines) PEBP families – two gymnosperm taxa phylogenetically distant from the angiosperms.
  • We have defined a lineage of gymnosperm PEBP genes, termed the FT/TERMINAL FLOWER1 (TFL1)-like genes, that share sequence characteristics with both the angiosperm FT- and TFL1-like clades. When expressed in Arabidopsis, FT/TFL1-like genes repressed flowering, indicating that the proteins are biochemically more similar to the angiosperm TFL1-like proteins than to the FT-like proteins. This suggests that the regulation of the vegetative-to-reproductive switch might differ in gymnosperms compared with angiosperms.
  • Molecular evolution studies suggest that plasticity at exon 4 contributes to the divergence of FT-like function in floral promotion. In addition, the presence of FT-like genes in basal angiosperms indicates that the FT-like function emerged at an early stage during the evolution of flowering plants as a means to regulate flowering time.


Flowering plants, or angiosperms, originated from a massive and unprecedented evolutionary burst that occurred c. 130–90 million years ago – a phenomenon renowned as Darwin's abominable mystery (Crane et al., 1995; Davies et al., 2004). Before the angiosperm radiation, the nonflowering seed plants (gymnosperms), which evolved about 380 million years ago, were the dominant living plants on Earth (Kenrick & Crane, 1997). Instead of flowers, gymnosperms carry reproductive structures that are either male or female. Based on floral-organ-identity gene patterns, it is assumed that reprogramming of an ancestral gymnosperm species could have produced the first flower (Baum & Hileman, 2006; Theissen & Melzer, 2007). Plants of the intermediate stage between gymnosperms and angiosperms, known as ‘proto-angiosperms’, are predicted to have had bisexual flowers with fading borders between the different flower organs (Chanderbali et al., 2009).

In angiosperms, the floral transition is largely dependent on the action of the FLOWERING LOCUS T (FT)-like genes, which play a key role in integrating both exogenous and endogenous signals controlling flowering (Ferrier et al., 2011). FT-like genes encode small globular proteins, almost completely consisting of a phosphatidylethanolamine-binding protein (PEBP) domain (Banfield & Brady, 2000; Ahn et al., 2006). Numerous examples have shown that FT-like genes inherited a conserved and universal floral-promoting function in angiosperms and that altered expression of FT orthologs severely affects the time of flowering (Kardailsky et al., 1999; Kobayashi et al., 1999; Kojima et al., 2002; Bohlenius et al., 2006; Lifschitz et al., 2006; Pin et al., 2010; Laurie et al., 2011; Meng et al., 2011). It has been proposed that flowering is the default developmental program, which is repressed to avoid precocious flowering and allow vegetative growth, and that the FT module simply provides a mechanism for accelerating flowering under favorable conditions (Stebbins, 1974; Hecht et al., 2011).

PEBP genes recently found in Picea abies (Hedman et al., 2009), originally named FT2 and FT4, were initially thought to be the FT counterparts in gymnosperms (Gyllenstrand et al., 2007). However, the cited authors have subsequently renamed these genes FT/TFL1-like1 (FTL1) and FTL2, and changed their original conclusion that they are FT-like genes as they prevented flowering when ectopically expressed in Arabidopsis, in marked contrast to FT (Karlgren et al., 2011).

In the study presented here we confirmed, through combined phylogenetic studies with maximum likelihood (ML), minimum evolution (ME) and neighbor-joining (NJ) analysis, the presence of two eukaryotic PEBP groups in gymnosperms: MOTHER OF FT AND TFL1 (MFT)-like and FT/TERMINAL FLOWER1 (TFL1)-like. As previously suggested (Karlgren et al., 2011), the results show that the FT/TFL1-like genes are as divergent from the FT-like genes as from the TFL1-like genes. Furthermore, heterologous expression in Arabidopsis showed that none of the PEBP genes identified to date in conifers have a biochemically conserved floral-promoting function resembling that of angiosperm FT. FTL1- and FTL2-expressing Arabidopsis plants exhibited a late- and very late-flowering phenotype, respectively, suggesting that the FT/TFL1-like proteins are biochemically more similar to the TFL1-like lineage. Nonsynonymous/synonymous divergence analysis of the coding sites indicates a strong purifying selection at exon 2 for FT/TFL1-, FT- and TFL1-likes, and at segments B-C of exon 4 only for the FT-likes – two regions that have previously been shown to be important for the floral-promoting/repressing functions (Hanzawa et al., 2005; Ahn et al., 2006; Pin et al., 2010). By comparing the degree of divergence between FT/TFL1-likes and FT-likes or TFL1-likes, we mapped onto the entire FT/TFL1-like coding region the conserved FT- and TFL1-like sites. In further experiments, transgenic embryogenic spruce cultures overexpressing either the Arabidopsis FT or spruce FTL1 genes grew normally, but overexpression of FTL2 resulted in growth arrest of the transgenic cultures and death after c. 6 months of tissue culture. These phenotypic results indicate that the spruce FTL2, in contrast to the true FT1/FT2 genes in Populus trees (Bohlenius et al., 2006; Hsu et al., 2006), may be involved in growth suppression. This is consistent with proposals that FTL2 may be involved in the stimulation of short-day-induced growth cessation in spruce (Gyllenstrand et al., 2007). Together, these data suggest that, while the FT/TFL1-like genes in gymnosperms might have a native role in growth regulation, the FT-like genes might have evolved shortly after the gymnosperm–angiosperm split and concurrently acquired a function in the regulation of flowering time.

Materials and Methods

Plant material and growing conditions

Arabidopsis thaliana (L.) Heynh. ecotype Columbia (Col-0), ft-10 mutant (Yoo et al., 2005) and transgenic lines were grown under long-day, 16-h light (150 μmol m−2 s−1), 22°C/8-h dark, 18°C conditions. The tfl1-14 mutant (Schultz & Haughn, 1993) was grown under the same conditions, except before transformation when plants were grown in short days consisting of 9-h light/15-h dark cycles.

Identification of PEBP homologs

PEBP homologs were identified in silico by means of similarity searches using BLAST analysis (Altschul et al., 1990). The protein sequences of the different PEBP family members from Arabidopsis and Populus were used as queries against the nucleotide collection (nr/nt) and nonhuman, nonmouse expressed sequence tag (EST) databases at National Center for Biotechnology Information (NCBI) ( and EST assemblies compiled by the Ancestral Angiosperm Genome Project (; Supporting Information Tables S1 and S2).

Phylogenetic and molecular evolution analyses

Phylogenetic studies were performed using mega5 (Tamura et al., 2011). Multiple protein and codon alignments with pairwise or partial deletion (cut-off 50%), including partial or full coding regions of PEBP genes from various plant species representing all divisions of living plants (Fig. 1 and Table S2), were created using ClustalW (Thompson et al., 1994). A best-fit substitution model was then calculated using ML. Evolutionary reconstruction was inferred using the NJ (Saitou & Nei, 1987), ME (Rzhetsky & Nei, 1992) or ML method based on the best-fit substitution model. Nodal support was estimated by bootstrap analysis and an interior-branch test (Dopazo, 1994) on the basis of 1000 re-samplings. We used Tree Puzzle (Schmidt et al., 2002) to estimate branch lengths and likelihood values for a set of possible tree topologies regarding the relationship between the FT-like, TFL1-like and FT/TFL1-like genes. We also evaluated the different tree topologies using the Shimodaira–Hasegawa test (Shimodaira & Hasegawa, 1999) which is a multi-comparison test that takes into account the fact that any given tree in the test set could be the ML tree.

Figure 1.

Evolutionary history of eukaryote phosphatidylethanolamine-binding protein (PEBP) genes in living plants. A phylogenetic reconstruction was inferred from a codon alignment with partial deletion (cut-off 50%) including exons 2, 3 and 4 of eukaryote MOTHER OF FT AND TFL1 (MFT)-like, FLOWERING LOCUS T/TERMINAL FLOWER1 (FT/TFL1)-like, TERMINAL FLOWER1 (TFL1)-like, BROTHER OF FT AND TFL1 (BFT)-like and FLOWERING LOCUS T (FT)-like genes from nonvascular land plants, nonseed vascular plants, nonflowering seed plants and flowering seed plants using the maximum likelihood (ML) method. An ML tree with 1000 bootstrap re-samplings is shown.

Divergences of and selective pressures acting on the FT/TFL1-, TFL1- and FT-like genes were estimated by comparing the numbers of nonsynonymous substitutions with the numbers of synonymous substitutions. Intraspecific (average pairwise ratio of nonsynonymous substitutions (πa)/average pairwise ratio of synonymous substitutions (πs)) and interspecific (non-synonymous substitutions per non-synonymous site (Ka)/synonymous substitutions per synonymous site (Ks)) diversity analyses were performed using the DnaSP v5.10 program (Librado & Rozas, 2009) with a sliding window mode (window size 12, step 1). The nucleotide sequences used in the divergence analysis are listed in Table S3.

Generation of transgenic Arabidopsis plants

A detailed description of the gene constructs and the generation of transgenic Arabidopsis plants can be found in Methods S1. Full-length coding regions of Pinus taeda (Pit)MFT1, Picea sitchensis (Ps)MFT1, PitMFT2, Picea glauca (Pg)MFT2, PsFTL1, Picea engelmannii × Picea glauca (Peg)FTL2 and PsMFT3 were amplified by PCR from EST clones (Cairney et al., 2006; Ralph et al., 2008). Chimeric PsFTL1-W138, PsFTL1-Bseg, Beta vulgaris (Bv)FT1-Y134, BvFT1-G137, BvFT1-W138, BvFT2-N134, BvFT2-Q137, BvFT2-Q138, BvFT1-Y134-W138 and BvFT2-N134-Q138 genes were synthesized by Millegen (Labège, France). Amino acid positions refer to the Arabidopsis FT positions. The one-letter code in front of the positions represents the amino acid replacing the native amino acid. ‘Bseg’ denotes segment B of the Arabidopsis FT gene.

PsFTL1, PegFTL2, PitMFT1, PsMFT1, PitMFT2, PgMFT2, PsMFT3, PsFTL1-W138, BvFT1-Y134, BvFT1-G137, BvFT1-W138, BvFT2-N134, BvFT2-Q137, BvFT2-Q138, BvFT1-Y134-W138 and BvFT2-N134-Q138 were expressed under the cauliflower mosaic virus 35S promoter (35S) and used to transform Col-0, ft-10 and tfl1-14 according to the floral-dip method (Clough & Bent, 1998). 35S::PsFTL1-Bseg was transformed in Col-0 only. In addition to the nontransgenic control plants, Col-0, tfl1-14 and ft-10 transformed with an empty pB7WG2D vector were used as controls for unspecific effects caused by the transformation and selection procedure.

At least 13 plants from each transformation as well as at least 13 nontransgenic Col-0, tfl1-14 and ft-10 plants were selected and flowering time was monitored by counting rosette and cauline leaves on the main inflorescence. To test the significance of the observed variations in flowering time, two-tailed Student t-tests were used, assuming unequal variance, and we classified P-values < 0.001 as highly significant, values < 0.01, < 0.05 and < 0.1 (which did not meet higher significance thresholds) as highly, moderately and weakly significant, respectively, and values > 0.1 as not significant (ns).

Generation of transgenic Norway spruce embryogenic cultures

The full coding sequences of the Arabidopsis FT and spruce FTL1 and FTL2 genes were assembled behind a constitutive ubiquitin promoter from Zea mays (Christensen et al., 1992) into the pBGWSXA binary vector. The pBGWSXA vector was constructed by turning the pAHC25 vector (Christensen & Quail, 1996) into a Gateway Destination Vector by exchanging the GUS gene for an RfA Cassette using the Gateway Vector Conversion System (Life Technologies, Grand Island, NY, USA). An embryogenic cell suspension culture of Norway spruce (Picea abies) line 61:21 was spread onto 47-mm-diameter discs of sterile paper. The discs were soaked for 2 h in modified leaf protoplast (LP) proliferation medium (Clapham et al., 1995), containing 0.25 M myoinositol, and bombarded with vector DNA-coated gold particles as previously described (Clapham et al., 2000). Directly after bombardment, the filter papers were transferred to semi-solid proliferation medium containing 0.25 M myoinositol and 0.18% phytagel (Sigma) for 8 d. The cells were then transferred to a second proliferation medium containing 3 mg l−1 5-azacytidine and 1 mg l−1 DL-phosphinotricin (Basta® active ingredients (Hoechst Schering AgrEvo GmbH, Berlin, Germany)) and myoinositol for 7 d, and then moved to a third proliferation medium with 3 mg l−1 DL-phosphinotricin and myoinositol, which was renewed every month. Basta®-resistant embryogenic tissue arose after 3–5 months. When the resistant embryogenic tissues were c. 2–4 mm in diameter, they were subcultured onto a new proliferation medium without DL-phosphinotricin and extra myoinositol, which gave rise to sublines within 1–2 months. In order to confirm that the Basta®-resistant lines were transgenic, genomic DNA was prepared from lines that survived the initial selection and was genotyped by PCR using vector-specific primers.


Identification and evolutionary history of the PEBP genes in conifers

By means of BLAST homology searches against nucleotide sequence collection and EST databases at GenBank, several eukaryote-type PEBP genes from conifers were identified (Table S1). All genes contained full-length coding regions.

To assess the relationships among the PEBP genes from conifers and other living plants, a phylogenetic reconstruction was performed using PEBP coding sequences representing most of the plant divisions. We included in the analysis sequences from green algae (Chlorophyta, represented by Coccomyxa), nonvascular land plants (Bryophyta/mosses, represented by Physcomitrella, and Marchantiophyta/liverworts, represented by Marchantia), nonseed vascular plants (Lycopodiphyta/clubmosses, represented by Selaginella), nonflowering seed plants (Pinophyta/conifers, represented by Picea and Pinus, Cycadophyta/cycads, represented by Zamia, and Ginkgophyta/ginkgo, represented by Ginkgo) and flowering seed plants (ancestral angiosperms including Amborella, Nymphaeales, represented by water-lily (Nuphar advena), and Magnoliales, represented by tuliptree (Liriodendron tulipifera) and avocado (Persea americana); monocots, represented by rice (Oryza sativa); and eudicots/true dicotyledons, represented by various asterids, caryophyllids and rosids; Fig. 1 and Table S2). The tree topology showed different evolutionary groups with, in addition to the FT-like, TFL1-like, BROTHER OF FT AND TFL1 (BFT)-like and MFT-like clades, a fifth group consisting of PEBP genes found only in gymnosperms (cycad, ginkgo, pine and spruce; Figs 1 and S1). This last group was previously described and named FT/TFL1-like (Karlgren et al., 2011). Multiple phylogenetic methods were used to analyze the nucleotide and amino acid data sets and these analyses all suggest that the relationship between angiosperm FT- and TFL1-like and the gymnosperm FT/TFL1-like genes cannot be clearly resolved (Fig. 2). NJ (Fig. 2a) and ME (Figs 2c and S1) analyses placed the FT/TFL1-like group as a sister clade of the TFL1- and BFT-likes, whereas in ML trees (Figs 1 and 2e) FT/TFL1-like genes are sisters of the FT-like genes, and lastly NJ and ME analyses on nucleotide alignment (Fig. S2) placed the FT/TFL1-likes as an ancestral clade to the TFL1- and FT-likes.

Figure 2.

Phylogeny tests of the multiple full protein sequence alignment used to reconstruct the evolutionary history shown in Supporting Information Fig. S1, using neighbor-joining (NJ), minimum evolution (ME) and maximum likelihood (ML) methods. (a, b) NJ trees with 1000 bootstrap and 1000 interior-branch re-samplings, respectively. (c, d) ME trees with 1000 bootstrap and 1000 interior-branch re-samplings, respectively. (e) ML tree with 1000 bootstrap pseudoreplicates. (f) Condensed tree. FLOWERING LOCUS T (FT), TERMINAL FLOWER1 (TFL1), BROTHER OF FT AND TFL1 (BFT), MOTHER OF FT AND TFL1 (MFT).

The inferred phylogenetic trees indicate that the gymnosperms contain MFT-like and FT/TFL1-like genes, but may lack homologs of FT-, TFL1- and BFT-like genes. If so, the FT-, TFL1- and BFT-like genes would be angiosperm-specific, and, based on recently completed EST sequencing, represented already in the very early angiosperm lineages, such as Amborella, and Nuphar (Fig. 1). FT/TFL1-like orthologs were identified in pine (Pinus), cycad (Zamia) and ginkgo (Ginkgo) while a pair of FT/TFL1-like paralogs, named FTL1 and FTL2 (Karlgren et al., 2011), were found in several species of spruce (Picea) (Fig. 1 and Table S4). Both genera of conifers present three copies of MFT-likes: a pair of paralogs, named MFT1 and MFT2, which apparently derived from a duplication predating the spruce/pine bifurcation, and a third and more divergent gene, MFT3 (Fig. S1). The ancient lineages (i.e. chlorophytes, bryophytes and pteridophytes), which are exclusively represented, in contrast to the gymnosperms and the angiosperms, in the MFT-like group, also present a divergent form of MFT-like gene that has most overall evolutionary similarity to the conifer MFT3 gene (Fig. S1). Although the node is poorly supported, these ancestral copies of PEBP genes seem to form a separate subclade within the MFT-like group that we named ‘ancestral MFT-like’ (Fig. S1). This observation is in agreement with findings of a previous study (Hedman et al., 2009) and suggests that MFT represents the ancestral form of the eukaryote PEBP genes in plants.

Because of the unclear position of the FT/TFL1 evolutionary group with respect to the FT- and TFL1-likes, and the fact that both ancient and modern angiosperms seem to lack FT/TFL1-likes, it is tempting to speculate that FT/TFL1 may be the ancestral form of the FT-like and TFL1-like genes, as recently proposed (Karlgren et al., 2011). Nevertheless, the current phylogenetic data (Fig. 2) also suggest another possible evolutionary model where FT- and TFL1-likes would have emerged from a gene duplication predating the last common ancestor of the seed plants (i.e. gymnosperms and angiosperms). In this second scenario, one copy resulting from the gene duplication (i.e. FT-like or TFL1-like copy) would have been lost in the gymnosperm lineage, while the biochemical differentiation between FT-likes and TFL1-likes would have occurred in the angiosperm lineage after the divergence of the gymnosperms. We used Tree Puzzle (Schmidt et al., 2002) to evaluate different tree topologies involving the FT/TFL1-like, FT-like and TFL1-like genes. The DNA sequence alignment support a tree grouping the FT/TFL1-like genes with FT-like genes (log L = −14670.35) but the alternative topologies, either grouping FT/FTL1-like with TFL1-like or grouping FT-like with TFL1-like, have similar likelihoods (log L = −14672.93 and log L = −14672.36, respectively). Neither of these topologies can be rejected using a Shimodaira–Hasegawa test (Shimodaira & Hasegawa, 1999) (pSH = 0.386 and pSH = 0.313, respectively). Similar results were obtained using the protein sequence alignment, where grouping FT/TFL1-like genes with FT-like genes has a slightly, but not significantly, greater likelihood (log L = −5927.21) than grouping FT/TFL1-like genes with TFL1-like genes (log L = −5931.85, pSH = 0.112) or when grouping FT-like genes with TFL1-like genes (log L = −5931.90, pSH = 0.109).

Overall, the phylogenetic reconstructions of the PEBP genes suggest that, while gymnosperms do have simple orthologs of the MFT clade, we are not able to resolve the relationship between the gymnosperm FT/TFL1-like genes and the distinct FT-like and TFL1-like clades found in angiosperms (Figs 2 and S2). This is probably a result of the fact that the gymnosperm sequences contain a mix of FT-like, TFL1-like and novel amino acid characteristics (Table S3). However, regardless of whether the FT vs TFL1 duplication event predated or postdated the last common ancestor of extant seed plants, it is very clear that the angiosperm genes underwent considerable sequence divergence following the separation of the flowering plant lineage. Given that these angiosperm-specific amino acids are known to be critical to the distinct biochemical functions of FT and TFL1-like proteins, this raises questions as to what the biochemical nature of the gymnosperm FT/TFL1-like genes might be.

Biochemical characterization of the conifer PEBPs by heterologous expression in Arabidopsis

Numerous studies have found that heterologous expression of angiosperm FT-like genes can act to promote flowering in a wide array of taxa (Kojima et al., 2002; Endo et al., 2005; Bohlenius et al., 2006; Hsu et al., 2006; Sreekantan & Thomas, 2006; Carmona et al., 2007; Hayama et al., 2007; Lin et al., 2007; Hou & Yang, 2009; Blackman et al., 2010; Kotoda et al., 2010; Pin et al., 2010; Traenkner et al., 2010; Danilevskaya et al., 2011; Fukuda et al., 2011; Imamura et al., 2011; Laurie et al., 2011; Lazakis et al., 2011), a function that appears tied to a specific set of highly conserved amino acid residues (Hanzawa et al., 2005; Ahn et al., 2006; Pin et al., 2010). We used heterologous expression of MFT and FT/TFL1-like genes from spruce and pine to assess their biochemical similarity to either FT- or TFL1-like genes in angiosperms. In contrast to Karlgren et al. (2011), who characterized the effects of expressing PEBPs from Norway spruce in a few fixed transgenic T2 generation lines, we screened at least 25 independent transgenic T1 generation plants, allowing a broader evaluation of the genes' effects, especially in terms of strong repression of flowering which might lead to a lack of fertility. We also included genes from several conifer species to further extend the analysis. Transgenic plants expressing MFT3 from Sitka spruce (Picea sitchensis) did not shown any phenotypic deviations from wild-type counterparts in flowering (Figs S3 and S4). The heterologous expression of MFT1 and MFT2 from white and Sitka spruce (P. glauca and P. sitchensis, respectively) and loblolly pine (Pinus taeda) led to a weak but significant delay in flowering time (Figs S3 and S4). These results are consistent with previous findings that expression of MFT1 has little or no effect on flowering in Arabidopsis (Yoo et al., 2004; Pin et al., 2010). By contrast, ectopic expression of FTL1 or FTL2 from two different spruces (P. sitchensis and P. engelmannii × glauca) severely delayed flowering onset (Figs S3 and S4) and perturbed floral formation (data not shown). These effects have previously been described in 35S::TFL1 Arabidopsis plants (Ratcliffe et al., 1998).

To test for the ability to complement the early-flowering phenotype observed in the tfl1 mutant, both FTL1 and FTL2 genes were expressed in the tfl1-14 genetic background. FTL1- and FTL2-expressing tfl1-14 plants flowered very late (Figs 3–5a,h,i), suggesting that the FT/TFL1-like genes from conifers can functionally replace TFL1 in repressing floral initiation when expressed from the 35S promoter in transgenic Arabidopsis. As observed in the wild-type genetic background, transgenic FTL1- and FTL2-expressing tfl1-14 plants showed, in addition to the late-flowering phenotype (Fig. 5a,h,i), several morphological aberrations, such as the formation of floral meristems that reverted to vegetative meristems (Fig. 5b), the formation of inflorescences that developed into vegetative shoots (Fig. 5c), inflorescences that developed intermediate vegetative/reproductive structures (Fig. 5d,e), a lack of petals (Fig. 5f), the development of late inflorescences that were not subtended by cauline leaves, and the formation of clusters of flowers at the end of that phase, as previously shown in transgenic 35S::TFL1 Arabidopsis plants (Ratcliffe et al., 1998). In conclusion, the phenotypes observed in Arabidopsis plants expressing either FTL1 or FTL2 very closely resemble those observed in 35S::TFL1 Arabidopsis (Ratcliffe et al., 1998) with shoots progressively becoming more flower-like with age and towards the apex. MFT1, MFT2 and MFT3 genes were also expressed in the tfl1-14 mutant but did not complement the tfl1 mutant phenotype. MFT1- and MFT2-expressing tfl1-14 plants showed a small, but again significant, delay in flowering time in comparison with the nontransgenic control plants (Fig. 3).

Figure 3.

Flowering time of transgenic Arabidopsis plants (tfl1-14 genetic background) ectopically expressing conifer phosphatidylethanolamine-binding protein (PEBP) genes. Rosette leaves were continuously scored until bolting time and cauline leaves on the main inflorescence were scored until the apical meristem turned into a flower or flowerlike structure. Bracts (‘cauline leaves’ on single flowers that never turned into proper side shoots) were not counted. The total numbers of leaves produced by the transgenic (gray) and nontransformed tfl1-14 plants (black) were compared. The last graph shows the flowering time of nontransgenic and transgenic plants harboring an empty binary vector that does not contain any gene of interest but carries the Basta® resistance gene. Graph headings show labeling, number of plants (n), mean flowering time and significance level of differences in the means between lines calculated by two-tailed Student t-tests assuming unequal variance, where ****, < 0.001; ***, < 0.01; **, < 0.05; *, < 0.10 and ns (not significant); > 0.10. Picea sitchensis FLOWERING LOCUS T/TERMINAL FLOWER1-LIKE1 (PsFTL1), Picea engelmannii x Picea glauca (Peg) MOTHER OF FT AND TFL1 (MFT), Pinus taeda (Pit), Picea glauca (Pg).

Figure 4.

Effect of ectopic expression of the conifer phosphatidylethanolamine-binding protein (PEBP) genes in the Arabidopsis tfl1 mutant. Three independent experiments (a–c) were carried out under long-day conditions. The photos show representative phenotypes of 50-d-old tfl1-14 mutants, transgenic tfl1-14 mutants expressing PEBP genes from spruce and pine, and transgenic controls obtained from empty vector transformation. Picea sitchensis FLOWERING LOCUS T/TERMINAL FLOWER1-LIKE1 (PsFTL1), Picea engelmannii x Picea glauca (Peg) MOTHER OF FT AND TFL1 (MFT), Picea glauca (Pg), Pinus taeda (Pit).

Figure 5.

Growth and flower phenotypes of FT/TFL1-like1 (FTL1)- and FTL2-expressing Arabidopsis plants. (a) Seven-week-old tfl1-14 mutant (left) and 35S::Picea engelmannii x Picea glauca (Peg) FTL2 tfl1-14 (right) plants. (b) Reversion of floral to vegetative meristems in FTL-expressing Arabidopsis plants, which results in aerial rosette formation (white arrow). (c–g) Floral phenotypes observed in 35S::PsFTL1 and 35S::PegFTL2 plants. Severely phenotypically perturbed inflorescence showing continuous leaf development (c), moderately (d) and weakly (e) phenotypically perturbed inflorescences developing intermediate vegetative/reproductive shoots, inflorescence showing floral buds lacking petal formation (f), and normal floral development (g). (h, i) Continuous vegetative growth observed in FLOWERING LOCUS T/TERMINAL FLOWER1 (FT/TFL1)-like-expressing Arabidopsis plants. (h) Severe phenotype, observed more frequently in 35S::PegFTL2 transgenics, developing small round/dark green leaves and impaired shoots, as illustrated in (c) and (d). (i) A less severe phenotype, observed more frequently in 35S::PsFTL1 transgenics, which developed shoots as illustrated in (e) but may, at very late stages of growth, develop normal flowers (g).

Collectively, the transgenic data showed that none of the spruce and pine PEBP genes had the ability to induce flowering when expressed in Arabidopsis, suggesting that the conifer PEBP genes lack the FT biochemical function.

Molecular evolution of the FT and TFL1 functions

Although sharing an average nucleotide and protein identity of 57.5% and 60%, respectively (Table S3), the FT-likes and the TFL1-likes have opposite functions in flowering control (Kardailsky et al., 1999; Kobayashi et al., 1999). Certain motifs of the conserved PEBP domain have been shown to be determinants of this functional divergence, such as amino acid position 85 in exon 2 (Hanzawa et al., 2005) and segment B at exon 4, which encodes an external P-loop (Ahn et al., 2006).

As FT/TFL1-like genes repress flowering when expressed in Arabidopsis (Karlgren et al., 2011; Figs 3–5, S3 and S4), we expected them to have signatures that are more similar to TFL1-like than FT-like signatures. Surprisingly all FT/TFL1-like proteins carry a tyrosine at position 85, which is specific to the FT-like proteins (Fig. 6; Hanzawa et al., 2005). Nevertheless, it has previously been shown that genes with Y85 could still have TFL1 function, for instance in beets, where flowering time is regulated by two FT-like genes that have evolved opposite transcriptional regulation patterns and remarkably antagonistic functions (Pin et al., 2010). BvFT2 promotes flowering, whereas its close paralog, BvFT1, suppresses it. Despite its inhibiting function, BvFT1 has a tyrosine at position 85 (Fig. 6). In addition, by expressing chimeric BvFT1-BvFT2 genes in Arabidopsis, the opposite functions of BvFT1 and BvFT2 were mapped to the P-loop, where BvFT1 differs from the Arabidopsis FT, BvFT2 and other flowering-promoting FT orthologs by three amino acid substitutions (Pin et al., 2010; Fig. 6). Although analysis of the P-loop domain suggested that FT/TFL1-likes share more identity with FT than with TFL1, several amino acid changes were detected between the FT/TFL1 and FT-like proteins; spruce FTL1 and FTL2 proteins carry three and six amino acid changes within their P-loop, respectively (Fig. 6). FTL2 shares with the flowering repressor BvFT1 a glutamine at position 137 (Q137) – one of the three amino acid positions, together with positions 134 and 138, responsible for the inhibiting function of BvFT1 (Pin et al., 2010).

Figure 6.

Alignment of amino acid position 85 at exon 2, and segment B and partial segment C at exon 4 of TERMINAL FLOWER1 (TFL1)-like, FLOWERING LOCUS T (FT)-like and FT/TFL1-like proteins.

Characterization of amino acid substitutions

Because of the dramatic effects caused by small variations within the P-loop of PEBPs on flowering, we decided to investigate the possible role of the amino acid changes observed in the FT/TFL1-like proteins. As the spruce FTL2 proteins have a Q137, we decided first to acquire a complete picture of the amino acids in the beet BvFT1 gene responsible for its repressing function. Six different chimeric BvFT1 and BvFT2 genes in which a single amino acid was swapped between the two proteins were expressed in Arabidopsis. None of the three BvFT1 chimeras promoted flowering; instead, they maintained their flowering repression function, although the expression of the chimeric version with W138 resulted in a less pronounced delay in flowering time (Fig. S5). These data suggest that more than one amino acid change in BvFT1 compromises the FT function. Likewise, none of the chimeric BvFT2 genes with a single amino acid swapped showed a repressing function but, in contrast to the chimeric BvFT2 carrying Q137 (which promotes flowering), chimeric BvFT2-N134 and BvFT2-Q138 did not promote flowering, or promoted it very weakly (Fig. S5). These results indicate that at least two positions are essential for FT function and that substitution of the glycine at position 137 in the FT protein by a glutamine is not the cause of the repressor function of BvFT1. In that respect, it is also unlikely that the observed flowering repression function of the spruce FTL2 is attributable to Q137. The single amino acid swapping experiment also strongly suggests that the Y134N and W138Q substitutions are responsible for the shift in function between BvFT2 and BvFT1. In order to verify our hypothesis, we generated and expressed two additional chimeric genes, BvFT1-Y134-W138 and BvFT2-N134-Q138, in Arabidopsis. Strikingly, these amino acid swaps resulted in conversion of BvFT1 from a repressor to an activator and conversion of BvFT2 from a promoter to a repressor of flowering (Fig. S5), corroborating the hypothesis that Y134 and W138, in addition to Y85, are essential positions for the activating function of FT-like genes (Fig. 6).

In light of these new data, we excluded the possibility of Q137 being responsible for the floral repressing FTL2 function in transgenic Arabidopsis. Regarding the other important amino acid positions in FTL1 and FTL2, both positions 85 and 134 have a tyrosine, whereas position 138 shows a serine; interestingly, this is shared with nearly all TFL1-likes (Fig. 6). Of the three amino acids (Y85/Y134/W138) that the beet observations indicate are important for the FT function, FTL1 and FTL2 only lack the tryptophan at position 138. Although a small number of characterized FT-likes encode other amino acids at 138, for example, corn (Zea mays) Zea mays CENTRORADIALIS8 (ZCN8) and barley (Hordeum vulgare) HvFT3, W138 is highly conserved in FT-likes (Fig. 6). To assess whether FT/TFL1-like proteins can be converted to flowering promoters in angiosperms by substituting S138 with W138, we generated and expressed chimeric FTL1-W138 in Arabidopsis. The resulting transgenic plants did not have an early-flowering phenotype, but the time of flowering was accelerated in comparison with the FTL1-expressing plants in the different genetic backgrounds (i.e. wild-type, tfl1-14 and ft-10; Fig. S6). These results indicate that S138, as well as other sites outside the P-loop domain, is involved in the floral-repressing activity of FTL1. Swapping the entire segment B, to that of Arabidopsis FT (PsFTL1-Bseg), did not accelerate the time of flowering further (data not shown). Besides Y85 and the B segment, the C segment of exon 4 was also found to have a role in the floral-promoting activity. Swapping both segments B and C of TFL1 for segments B and C of FT turns the protein into a floral activator, whereas swapping of either segment B or C of FT for segment B or C of TFL1 annuls the FT function (Ahn et al., 2006). A feature of the C segment is a highly conserved L/I150-Y151-N152 triad in the FT-likes (Fig. 6). Although positions 150 and 152 can vary in TFL1-likes, N151 is found in almost all TFL1-likes. Intriguingly, all FT/TFL1-likes carry an N151 residue, which is the only site distinguishing the FT- from the FT/TFL1- and TFL1-likes at the C segment (Fig. 6).

Divergence analysis of the coding sequences, based on synonymous/nonsynonymous substitution ratios, showed that exon 2 of all three gene classes is under strict purifying selection (Fig. 7a). By contrast, segment B of exon 4 is subject to purifying selection only for the FT-like genes (Fig. 7a). This selection pressure pattern is consistent with the finding that some sites at exons 2 and 4 of the FT-likes are essential to maintain a floral-promoting activity. Similar analysis of the divergence of FT/TFL1- from TFL1- and FT-likes allowed us to reconstruct the original FT/TFL1-like coding region by mapping the FT/TFL1-like-, TFL1-like- and FT-like-specific and conserved segments (Fig. 7b,c). While nearly half of the whole coding sequence has been conserved during the FT- and TFL1-like radiations, amino acid changes observed in the other half are evenly distributed between FT/TFL1-like-, TFL1-like- and FT-like-specific segments (Fig. 7c,d), supporting the hypothesis that FT/TFL1-likes contain a mix of TFL1- and FT-like character states. The apparently weaker purifying selection observed for TFL1-likes (Fig. 7a) indicates that the exact sequence of TFL1 is less important for its repressing function than the exact sequence is for the activating function of FT. Overall, these molecular evolutionary analyses suggest that TFL1- and FT-likes are closely related to the FT/TFL1-likes and that the observed floral-repressing activity of FT/TFL1-likes, in an angiosperm system, results from TFL1-like-specific segments that mapped to the important B and C segments of exon 4 (Fig. 7c).

Figure 7.

Results of analysis of molecular divergence between the FLOWERING LOCUS T/TERMINAL FLOWER1 (FT/TFL1)-, TFL1- and FT-likes. (a) Intra-evolutionary group divergences (πas) of the coding region. (b) Inter-evolutionary group divergence (Ka/Ks) between FT/TFL1- and TFL1- or FT-likes across the whole coding region. (c) Reconstruction model of the FT/TFL1-like according to the inter-evolutionary group divergence analysis between FT/TFL1- and TFL1- or FT-likes. FT-like, TFL1-like, FT/TFL1-like and conserved segments were assigned based on a 0.10 Ka/Ks cut-off as illustrated by the threshold in panel (b). (d) Percentages of FT-like, TFL1-like, FT/TFL1-like and conserved segments across the FT/TFL1-like coding region with different Ka/Ks cut-offs. (e, f) Model of evolution of important sites potentially involved in FT- and TFL1-like functions. Amino acids are represented by a one-letter code and numbers represent amino acid positions according to the Arabidopsis FT protein. Blue, red and black colors represent FT-like-, TFL1-like- and FT/TFL1-like-specific sites, respectively, ‘X’ represents unknown or varying amino acids. (e) The ancestral FT/TFL1-like protein is a repressor of flowering but is predisposed to evolve into a variant with FT-like function. Gene duplication resulted in two copies of FT/TFL1-like genes that diverged into FT-like and TFL1-like variants. (f) The ancestral FT/TFL1-like protein duplication predated the last common ancestor of seed plants. One of these copies gave rise to the angiosperm FT-like genes and the other to the TFL1-like genes. One of the ancestral copies is then lost in the gymnosperms (in gray). In this figure we have arbitrarily depicted the copy giving rise to the FT-like genes to be lost in the gymnosperms.

Insights regarding FTL2 function in spruce

Although both FTL1 and FTL2 genes act as floral repressors when expressed in Arabidopsis, the endogenous function of the FT/TFL1-like genes in conifers is unclear. To acquire the first insights into the functions of FTL1 and FTL2, we overexpressed the spruce FTL1 and FTL2 genes, and the Arabidopsis FT gene, in transgenic spruce. In addition to the control plants, which were transformed solely with the selective gene, both transgenic AtFT- and FTL1-expressing spruce embryogenic cultures developed normally (Fig. 8a,b). By contrast, all embryogenic spruce cultures expressing FTL2 stopped growing and died after a few months of tissue culture (Fig. 8c). Although preliminary, these first transgenic data suggest that FTL2 may function as a growth repressor, in accordance with a previous proposal that FTL2 is involved in short-day-induced growth cessation in spruce (Gyllenstrand et al., 2007).

Figure 8.

Transgenic spruce embryogenic culture overexpressing FLOWERING LOCUS T/TERMINAL FLOWER1 (FT/TFL1)-like and Arabidopsis FT genes. (a, b) Close-ups of Arabidopsis FT- (a) and FTL1- (b) expressing embryogenic cultures. (c) Overview of numbers of obtained transgenic spruce embryogenic cultures after 5 months on selective media, and long-term survival rates after transfer to nonselective media. Picea sitchensis (Ps), Picea engelmannii x Picea glauca (Peg).


Among the PEBP gene family members, TFL1 and FT have been by far the most intensively studied genes in Arabidopsis. Both of these genes are represented in a wide range of flowering plants and during evolution they have been the subject of several duplication events (as illustrated in Fig. 1), potentially resulting in functional retention, loss of function, subfunctionalization or neofunctionalization. An example of the plasticity and gain of function of FT genes is found in Populus trees, where the FT1/FT2 paralogous pair mediates growth control. When simultaneously down-regulated by RNA interference (RNAi), double PtFT1-PtFT2 RNAi Populus trees show growth arrest in long days (Bohlenius et al., 2006), while overexpression of either FT1 or FT2 leads to precocious flowering and abolishes the short-day-induced growth cessation (Bohlenius et al., 2006; Hsu et al., 2006). Although ectopic expression of both genes results in the same growth/flowering phenotypes, native gene expression analyses suggest that FT2 is the most important gene controlling vegetative growth vs growth cessation, while the FT1 gene might be more closely related to the induction of flowering (Hsu et al., 2011). In Norway spruce, growth is also photoperiodically controlled, and short days promote growth cessation and bud set (Heide, 1974).

In contrast to Populus, the currently available conifer EST sequence data suggest that spruce and pine trees do not have FT-like genes (Fig. 1; Karlgren et al. (2011)), raising the possibility that the reproductive transition is controlled by other genes in conifers. Instead, the gymnosperms carry one or two copies of FT/TFL1-like genes, which represent a clade that is either sister or ancestral to the FT and TFL1 genes found in angiosperms, according to our detailed molecular evolution analyses, combining: phylogenetic reconstruction; domain swapping; and nonsynonymous/synonymous substitution-based divergence analyses, as well as previous results (Karlgren et al., 2011).

We found that about 50% of the coding region has been evolving under strong purifying selection during the divergence of the FT/TFL1- and both the FT- and TFL1-like copies. In contrast, the other half displays an even distribution of segments that are conserved between FT/TFL1-like and either TFL1-like or FT-like genes, respectively (Fig. 7c,d); thus, the FT/TFL1-like genes are highly distinct in sequence from either the angiosperm FT- or TFL1-like clades, a situation that makes resolving their phylogenetic relationships very difficult. The gymnosperm genes possess a mix of FT- and TFL1-like amino acid character states, which may be more suggestive of the first model presented (Fig. 7e), where the gymnosperm sequences are pre-duplication ancestors.

Based on previous mapping of important domains for the FT-like function – Y85 at exon 2 and segments B and C at exon 4 (Hanzawa et al., 2005; Ahn et al., 2006; Pin et al., 2010) – we concentrated on differences in the amino acid composition in these regions between the FT/TFL1- and TFL1- or FT-like proteins (Fig. 6). FT/TFL1-likes carry both Y85 and Y134, which are required for the FT-like function (Hanzawa et al., 2005; Fig. S5). However, they lack W138, which, according to our swapping data from the beet FT genes (Fig. S5), is also important for this function. In addition, several positions within the C segment (150–152 triad, 164) differ quite dramatically among the FT/TFL1-, TFL1- and FT-likes. In particular, N151Y and A164C are in strong linkage with FT/TFL1- and TFL1- vs FT-likes (Fig. 6). It is likely that more than one position in segment C contributes to its floral-repressing activity observed in our heterologous expression analyses using an angiosperm system. In contrast to all other TFL1-likes analyzed in our study which carry N151, Arabidopsis TFL1 carries Y151. A previous report showed that switching only the C segment of Arabidopsis FT for Arabidopsis TFL1 leads to a protein with a repressing function (Ahn et al., 2006), suggesting that Y151 cannot account solely for the promoting function at segment C, and that other sites, possibly 150, 152 and 164, are also involved. Further swapping experiments would help in pointing out the additional amino acids involved in the floral repressive function (in an angiosperm system) of the gymnosperm FT/TFL1-like genes.

Collectively, our data lead to a model of molecular evolution where at least three sites played an essential role in the TFL1- and FT-like fates – sites 85, 134 and 138 – as well as one or several positions in segment C. S138W, and possibly X150L/I-N151Y-X152N triad +/− A164C, have led to the FT-like function, whereas Y85H and Y134X have further stabilized the TFL1-like function, making it veritably impossible to go in the opposite direction and gain FT-like function. Although our study provides insight into the molecular evolution leading to the current FT-like and TFL1-like forms, the data do not preclude one of the two possible evolutionary scenarios where: the gymnosperm FT/TFL1-like genes are sister to a clade that contains both the FT- and TFL1-like genes, which were produced by an angiosperm-specific duplication event (Fig. 7e); or the gymnosperm FT/TFL1-like genes are in fact sister to either FT- or TFL1-like, which would suggest a pre-seed plant duplication followed by loss of one of the lineages in gymnosperms (Fig. 7f). In the second evolutionary scenario, the two gene copies present in the angiosperms evolved into two biochemically distinct forms, that is, FT-like and TFL1-like, while in gymnosperms they maintained a TFL1-like function (Fig. 7f). The difference in gene copy numbers between spruce and pine, ginko and zamia carrying, respectively, two and one FT/TFL1-like gene copies, suggested that either a recent gene duplication event occurred in spruce, or that the pine, ginko and zamia lineages have lost one of the gene copies and that spruce is the only gymnosperm that maintained the two original copies (Figs 1 and 7f). When specifically testing these alternative relationships among FT-like, TFL1-like and FT/TFL1-like genes (using TREE PUZZLE (Schmidt et al., 2002)) we find that we cannot distinguish statistically between alternate topologies, probably because of low statistical power, as the genes contain relatively few phylogenetically informative sites.

Regardless of when the FT vs TFL1 duplication occurred, it appears that one paralogous lineage evolved into the FT-like sequence through specific amino acid substitutions (S138W, 150–152 triad +/− A164C), while another copy reinforced its ancestral biochemical TFL1-like function through the gain of additional changes (Y85H, Y134X; Fig. 7e,f). Both TFL1- and FT-like copies are important, as their activities balance vegetative and reproductive growth in angiosperms. Some of the important sites involved in flowering fate (e.g. segment B) are under strong purifying selection in the FT-likes, suggesting that nearly all alternative FT-like alleles have deleterious effects. By contrast, at the same positions, TFL1-likes are under closer to neutral selection, indicating that the exact nature of the amino acid composition is not as crucial to maintain TFL1-like function.

Our data, together with earlier findings (Karlgren et al., 2011), provide insights into the biochemical differentiation with regard to flowering in an angiosperm system of the TFL1- and FT-like genes. However, the native role of FT/TFL1-likes in a gymnosperm system remains poorly understood. Although the heterologous expression studies show that there is biochemical similarity between conifer FT/TFL1-like genes and the angiosperm TFL1-like genes, they do not provide any information regarding the actual endogenous functions in gymnosperms. A certain protein function is evolving together with the function of other proteins in a regulatory context, and we cannot therefore completely exclude the possibility of the presence of a more FT-like function in the conifers; this will have to await a more complete functional characterization in transgenic conifers. However, several lines of evidence indicate that the biochemical conservation shown here might be related also to a functional conversation in the conifers. Gene expression studies in Norway spruce have shown that PaFTL1 transcription increases after the winter and during the spring in the microsporophyll of male cones (Karlgren et al., 2011). Other studies have shown that transcription of PaFTL2 in Norway spruce is induced in needles and shoots in response to short days, and thus correlates with bud set and growth cessation (Gyllenstrand et al., 2007; Asante et al., 2011; Karlgren et al., 2011). In addition, PaFTL2 expression was found to be down-regulated before bud burst. Together, these data suggest that FTL1 and FTL2 have opposite transcriptional regulation patterns and are likely to have sub- or neo-functions. Our transgenic experiments, where the FTL1 and FTL2 genes were ectopically expressed in Norway spruce, indicate that FTL2 is a repressor of growth (Fig. 8c). It is therefore conceivable that FTL2 participates in growth-cycle control in spruce by promoting bud set and growth cessation, essentially in an opposite way to PtFT2 in Populus trees (Hsu et al., 2011), while FTL1 may act in a similar way to TFL1-likes in flowering plants, preventing a precocious vegetative-to-reproductive development switch. Additional experiments, in particular functional validation using knock-down transgenic approaches and inducible promoters, are required to obtain further insights regarding the function of the FT/TFL1-like genes in gymnosperms. However, a plausible scenario is emerging where, in gymnosperms, the FT/TFL1-likes are involved in growth control, and possibly in the repression of reproduction, but not in the reproductive stage promotion; a feature that could have evolved only with the FT-likes. Our evolutionary data also show that FT-likes are exclusively found in angiosperms, including very basal and modern lineages, corroborating the hypothesis that the split between nonflowering seed and flowering seed plants coincided with the evolution of an FT/TFL1-like gene into one with an FT-like-flowering promoting function. The FT function might possibly have contributed to the rapid radiation of the flowering plants on Earth, and thus be part of the answer to Darwin's abominable mystery.


We thank Dr Joerg Bohlmann for providing the spruce ESTs and The Institute for Genomic Research, a division of the J. Craig Venter Institute, USA for providing the Pine ESTs used in this study. We also thank Flanders Interuniversity Institute for Biotechnology (VIB) for providing the pB7WG2D vector and Sofie Johansson for performing the spruce transformations. This work was supported by grants from the Swedish research Council (VR), the Knowledge Foundation (KK-stiftelsen) and the Swedish Governmental Agency for Innovation Systems (VINNOVA).