Plants of eight Rosaceae species were from the experimental plots at Huazhong Agricultural University, Wuhan, P.R. China. Nicotiana tabacum cultivar ‘Xanthi’, Arabidopsis thaliana Col and ft-1 Arabidopsis mutant (Ler ecotype) were used for wild controls.

Genomic DNA from eight Rosaceae species was extracted from young leaves as described previously by Wang Z. et al. (2011). Total RNA was extracted according to a previous protocol (Hu et al., 2002). The initial FT/TFL1 genomic DNA sequences were isolated by homology cloning strategies and genome walking methods (Wang Z. et al., 2011). The degenerated primers were designed according to the FT/TFL1 sequences from other Rosaceae species. For cloning of the FT homologs, the degenerated primers used were: FTF1, 5′-ATGCCTAGGGAHAGGGAYCCYCTTGTT-3′, FTF2, 5′-GCAACAACGGCGGCAAGCTT-3′, and FTR, 5′-CCAGAGCCRCYCTCCCTYTGGCAGTT-3′. For cloning of the TFL1 homologs, the degenerated primers used were: TFL1F, 5′-TTGGNAGAGTGATAGGAGATGTT-3′, TFL1R, 5′-GAGGAAGGTGKGTTGATTGA-3′. Fusion primer and nested integrated PCR (FPNI-PCR) was used to isolate the unknown sequences flanking the core sequences amplified from the degenerated primers. The full-length FT/TFL1 cDNA sequence was isolated by specific primers (Supplementary Tables S1–S3). Amino acid sequences were aligned using CLUSTALW MULTIPLE ALIGNMENT with default parameters. Phylogenetic studies were performed using MEGA5 based on the neighbor-joining method (Tamura et al., 2011). Nodal support was estimated by bootstrap analysis and an interior branch test on the basis of 1000 re-samplings.

Protein structures of FT and TFL1 homologs were obtained using SWISS-MODEL workspace (Arnold et al., 2006¹) and visualized by UCSF Chimera (Pettersen et al., 2004). The three-dimensional structures of 3AXY and 1WKO were used as loading template for FT and TFL1, respectively.

The gene splicing overlap extension PCR method (SOE-PCR) (Ho et al., 1989) was used to get a pre-determined point mutagenic site in FT/TFL1 sequences. We designed a pair of complementary oligo primers in which 1 or 2 base pairs had been altered to introduce a specific mutation into the amplified gene sequence. These mismatch primers mutants (i.e., RoFTmu1F and RoFTmu1R) were paired with unaltered RoFTR and RoFTF primers, respectively, and were used for PCR to generate two DNA fragments with overlapping ends. The two fragments were combined in a subsequent ‘fusion’ reaction PCR using RoFTF and RoFTR primers (Supplementary Table S4). All point mutagenic sequences were introduced into pMD18-T and then pMOG22 vector (Mogen, Leiden, The Netherlands).

The RoFT, RoTFL1, FaTFL1, PhFT, and AtFD genes were amplified by PCR from each RNA with the appropriate specific primers (Supplementary Table S4). The amplified products were cloned into pMD18-T vector (Takara) and sequenced. Then the inserts were subcloned into the modified binary vector pMOG22 containing the cauliflower mosaic virus (CaMV) 35S promoter and the Nos terminator.

For Arabidopsis transformation, the constructs in binary vectors were introduced into Agrobacterium tumefaciens strain GV3101. Transgenic plants were generated by floral dip method and the T1 transformants were selected on hygromycin plates for 1 week in LD (16-h-light/8-h-dark) and then transferred to soil at 20–24°C under long day condition (16-h-light/8-h-dark).

Tobacco was transformed by Agrobacterium tumefaciens strain EHA105 according to previously described method (Ning et al., 2012). All transgenic tobacco plants were kept in the greenhouse under a photoperiod of 12-h-light/ 12-h-dark. Data were collected from at least 20 individuals and evaluated by analysis of variance (one-way ANOVA). Means were compared using Duncan’s multiple range test.

For real time qRT-PCR analysis, samples were harvested from the shoot apex of 45-day-old seedlings of T1 transgenic tobacco plants or 3-week-old seedlings of transgenic Arabidopsis plants. Three biological replications were performed randomly for each transgenic line. Total RNA was isolated using Trizol reagent (Takara) according to the manufacturer’s instructions. The first strand of cDNA was synthesized using 2 μg of total RNA as a template with the TransScript^TM one-step gDNA Removal and cDNA Synthesis Supermix (Transgen, Beijing, China). The qRT-PCR was performed on 7500 Fast Real-Time PCR System (Applied Biosystems) with SYBR Premix EX Tag^TM (Takara). The tobacco NtEF1α and Arabidopsis AtEF1α transcript were used as an internal standard to calculate the relative expression by the comparative CT (△△CT) method, respectively. The primers for RT-PCR and qRT-PCR are detailed in Supplementary Tables S5, S6.

The coding sequences of AtTFL1, RoFT, RoFTmu1/2/3/4/5, FaTFL1, RoTFL1, and PhFT (all containing the EcoR1 and Sal1 restriction sites at the 5′ and 3′ ends, respectively) were cloned into bait plasmid PGBKT7. Arabidopsis FT (AtFT) was also introduced to the PGBKT7 plasmid, using the Nde1 and Sal1 restriction sites, as a positive control. The full-length Arabidopsis FD coding sequence (AtFD) was cloned into prey plasmid PGADT7 using the Nde1 and BamH1 restriction sites. Yeast cells were transformed using Frozen-EZ Yeast Transformation II^TM kit (ZYMO RESEARCH, USA). Co-transformed yeast cells were selected on SD-Leu/-Trp plates. Interactions were tested on SD-Leu/-Trp/-His/-Ade/X-a-Gal selective media. Three independent clones for each transformation were tested.

Strain of Agrobacterium tumefaciens GV3101 carrying the BiFC constructs were used for the infiltration of 5–6-week-old N. benthamiana leaves, according to the protocol described by Li et al. (2015). Of which, the coding sequences of AtFT, AtTFL1, RoFT, RoFTmu1/2/3/4/5, FaTFL1, RoTFL1, and PhFT were introduced into the vector pFGC-YC155, respectively. The AtFD coding sequence was cloned into the vector pFGC-YN173. All vectors were constructed by Gibson assembly method (Gibson et al., 2009). The primers are detailed in Supplementary Table S7. YFP fluorescence was visualized by confocal laser scanning microscope (LSM510 Meta, Zeiss, Germany).

The coding sequences of AtFT, RoFT, RoFTmu2/3/4/5, PhFT, AtTFL1, RoTFL1, and FaTFL1 were amplified with the primers which were used to construct PGBKT7 vectors before (Supplementary Table S7), and finally cloned into the EcoR1/Sal1 (Sac1/Sal1 for AtFT) sites of PET-32a vector (NOVAGEN) to obtain PET32a-His-FT. The 10 PET32a-His-FT plasmids were transformed into competent Escherichia coli Rosetta (DE3) cells (Transgen, Beijing, China). Fusion protein expression was induced at an OD₆₀₀ of about 0.5 by adding IPTG (isopropyl β-D-1-thiogalactopyranoside) (0.2 mM final concentration), in which the cells were grown overnight and the temperature was shifted from 37 to 16°C. The expressed soluble proteins were purified with Ni-Agarose (CWBIO, Beijing, China) according to the manufacturer’s instructions.

18:1-PC (1, 2-Dioleoyl-sn-Glycero-3-phosphatidylcholine) standards was purchased from Larodan (Sweden). The reaction was performed according to the modified protocol described by Stevenson et al. (1998). Of which, a goat anti-rabbit IgG conjugated to alkaline phosphatase (CWBIO, Beijing, China) against 6X histidine was diluted at a 1:10000 level, and the protein bound to the lipid spot was detected by alkaline phosphatase substrate according to the manufacturer’s instructions (Promega).

Sequence data from this article can be found in NCBI under the following accession numbers: Arabidopsis AtFT (AF152096); Beta BvFT1 (HM448910); Beta BvFT2 (HM448912); Citrus CiFT (AB027456); Fragaria FaFT (CBY25183); Malus MdFT1 (BAD08340); Malus MdFT2 (ADP69290); Nicotiana NtFT1 (JX679067); Nicotiana NtFT2 (JX679068); Nicotiana NtFT3 (JX679069); Nicotiana NtFT4 (JX679070); Oncidium OnFT (ACC59806); Oryza Hd3a (AB052944); Petunia PhFT (ADF42571); Photinia PsFT (AEO72028); Platanus PaFT (ACX34055); Populus PnFT1 (AB106111); Populus PnFT2 (AB109804); Populus PnFT3 (AB110612); Prunus mume PmFT (CBY25181); Prunus persica PpFT (AEO72030); Pyracantha PfFT (AEO72029); Pyrus pyrifolia PpFT (KF240775); Rosa RoFT (CBY25182); Spiraea ScFT (AEO72031); Vitis VvFT (ABF56526); Zea ZmFT (ABW96237); Arabidopsis TFL1 (U77674); Antirrhinum CEN (CAC21564); Citrus CiTFL1 (AY344245); Fragaria FaTFL1 (AEO72027); Malus MdTFL1-1 (AB162040); Malus MdTFL1-2 (AB366643); Oryza FDR1 (AF159883); Oryza FDR2 (AF159882); Photinia PsTFL1 (AEO72024); Populus PnTFL1 (AB181183); Prunus mume PmTFL1 (AEO72021); Prunus persica PpTFL1 (ADL62867); Prunus yedoensis PyTFL1 (AEO72023); Pyracantha PfTFL1 (AEO72026); Pyrus pyrifolia PpTFL1-1 (BAD10962); Pyrus pyrifolia PpTFL1-2 (BAK74839); Rosa RoTFL1 (AEO72022); Spiraea ScTFL1 (AEO72025); Vitis VvTFL1 (AF378127); Zea ZmTFL1 (ABI98712).

FT/TFL1 orthologs of eight Rosaceae species, namely, Spiraea cantoniensis, Pyracantha fortuneana, Photinia serrulata, Fragaria ananassa, Rosa hybrida, Prunus mume, Prunus persica (only for FT) and Prunus yedoensis (only for TFL1), were isolated. Two TFL1 copies were isolated from Fragaria ananassa genomic DNA, but only one gene copy was isolated from all other genotypes. Each of the isolated FT/TFL1 sequences contained four exons and three introns. In all isolated genes, the sizes of the second and third exons were the same, i.e., 62 and 41 bp, respectively (Figures 1A,B). The seven FT/TFL1 sequences share 92.09 and 90.59% identity, respectively (Supplementary Figure S1). All FT/TFL1 homologs from the eight Rosaceae species were found to contain the (putative) crucial amino acid residues of Tyr-85 (for FT) and His-88 (for TFL1). Based on the construction of the phylogenetic tree, it was deduced that all seven FT orthologs were clustered within the FT-like group and all seven TFL1 orthologs were clustered within the TFL1-like group (Figure 1C).

For functional study of FT/TFL1 homologs from eight Rosaceae species, we constructed over-expression vectors harboring FT and TFL1 homologs (cDNA) of Prunus mume, Rosa hybrida, and Fragaria ananassa. The three species represent different vegetative growth and flowering habit. Two TFL1 copies were isolated from Fragaria ananassa genomic DNA, namely, FaTFL1-1 and FaTFL1-2. There are three single-base differences between the two predicted CDS regions. But only one copy was amplified from the cDNA which shared the same sequence with the predicted CDS region of FaTFL1-1 gDNA sequence.

According to the results from 20 independent transgenic tobacco lines, the majority of over-expressing RoFT and PmFT tobacco lines (Figures 2A–C), exhibited strongly advanced flowering traits, this was consistent with an earlier preliminary analysis (Ning et al., 2012). At time of flowering, the wild-type had generated 28.6 ± 1.1 leaves, compared with 6.8 ± 1.0 and 5.9 ± 1.1 leaves in the 35S::RoFT lines R0-4 and R0-15, respectively (Table 1). In contrast to the strongly advanced flowering of RoFT and PmFT lines, the over-expression of FaFT in line F0-1 produced a moderately late flowering time (almost 30 days later relative to wild-type). The number of leaves and height remained comparable to the wild-type (Table 1). One of the transgenic line F0-9’s flowering time was approximately 50-days later than the wild-type. Thus, there was clearly some functional divergence with respect to the control of flowering between the FT orthologs from the different plant species.

The majority of 35S::PmTFL1, 35S::RoTFL1, and 35S::FaTFL1 transformants flowered much later than wild-type plants. Most transformants did not flower in less than 7 months after sowing, as compared to approximately 5.5 months seen in wild-type plants. In some extreme cases, flowering in transformed plants was delayed to over 12 months after sowing (Figures 2G,H). As shown in Table 1, the two selected lines transformed with 35S::FaTFL1 had produced as many as over twice leaves on the main stem to wild-type plants by the time of flower initiation. Transformants expressing 35S::PmTFL1 and 35S::RoTFL1 showed very similar results to those shown for 35S::FaTFL1 transgenic lines. Therefore, tobacco plants overexpressing the three TFL1 orthologs from Prunus mume, Rosa hybrida, and Fragaria ananassa had an extended vegetative phase and a strongly delayed transition to the reproductive phase.

A similar phenotype to this late flowering imposed by Rosaceae TFL1 homologes also resulted from the over-expression of a FT homolog which was isolated from Petunia hybrida (Figures 2D–F). The PhFT gene contained the Tyr-85 residue and LYN/IYN triplet motif as typical FT sequences, but a Lys-139 residue replaced the normal amino acid in FT (i.e., Gln-140); the corresponding residue in TFL1 was Asp-144 (Supplementary Figure S2). The resulting 35S::PhFT transgenic tobacco reached over 2 m in height because of extremely late flowering. Thus, it demonstrated a new role of TFL1 although it was identified as an FT homolog in our phylogenetic analysis.

Since Rosa FT (RoFT) and Fragaria FT (FaFT) exhibited quite different effects on flowering time in transgenic tobacco, we compared their sequences in more detail. The two proteins share approximately 88% identity with 13 non-conserved substitutions amongst 20 different amino acids (Supplementary Figures S1, S2), to be key in their flowering time function. We focused on five amino acids, which corresponding to residues 7, 65, 116, 153, and 154 in RoFT. The amino acids at positions 7, 65, 116, and 153 in RoFT were changed individually to correlate with the corresponding amino acid residues encoded by FaFT (Figures 3A,B). In addition, we mutated the amino acid N-154 which is identical between RoFT and FaFT within the IYN triplet motif of segment C in exon 4. The five resulting mutants were respectively named RoFTmu1-5 and each was over-expressed under the control of the constitutive CaMV 35S promoter (Figure 3C).

Tobacco plants over-expressing RoFTmu1 (R7Q), RoFTmu2 (T65I), and RoFTmu3 (A116S) displayed an early-flowering phenotype, comparable to the native RoFT in transgenic tobacco. In contrast, 35S::RoFTmu4 (Y153C) and 35S::RoFTmu5 (N154D) transgenic plants showed a strong late flowering phenotype (Figures 4A,B). As shown in Table 2, 35S::RoFTmu1, 35S::RoFTmu2, and 35S::RoFTmu3 tobacco plants flowered after producing approximately 8–10 leaves over 2 months of growth. By contrast, the majority of the 35S::RoFTmu4 and 35S::RoFTmu5 transformants had a much delayed flowering time, requiring 210 ± 27.1 and 248.1 ± 32.7 days of growth, respectively. We also ectopically expressed 35S::Roftmu3 (A116S) and 35S::RoFTmu4 (Y153C) in Arabidopsis Col. 35S::RoFTmu3 (A116S) plants showed a marked early flowering phenotype, with approximately 50% the number of leaves as found in the wild-type Col at floral initiation (Figures 5A–C). Transgenic 35S::Roftmu4 (Y153C) Arabidopsis flowered slightly later than the corresponding wild-type Col (Figures 5A–C). In addition, overexpressing RoFTmu1, RoFTmu2, and RoFTmu3 within ft-1 mutant (Ler ecotype) resulted in significant early flowering compared to ft-1 plants (Figures 5E,F). As shown in Figure 5G, ft-1 mutant harboring 35S::RoFTmu3 possessed 9.1 ± 0.9 rosette leaves at the time of bolting, which is almost consistent to that resulted from 35S::RoFT (8.9 ± 0.7), while, ft-1 mutant had produced as many as >3-fold leaves (30.2 ± 2.5) until flowering. Meanwhile, the flowering time was much earlier than those ft-1 plants (Figure 5H).

It has been reported that the opposite roles of FT and TFL1 are related to the conserved amino acids His-88 and Asp-144 in TFL1 (Hanzawa et al., 2005; Ahn et al., 2006). To examine whether these amino acids is also conserved in other plant species, we constructed mutants RoTFL1mu1 (H82Y), RoTFL1mu2 (D137Q), FaTFL1mu1 (H84Y), and PhFTmu1 (K139Q) (Figure 3B), and transferred them into tobacco plants. As shown in Figures 4C–F, no early flowering phenotype was observed in any of these transformants, as compared to wild-type tobacco. In fact, some of these transgenic plants remained in the vegetative growth phase for over 11 months (Table 3).

According to previous studies (Abe et al., 2005; Wigge et al., 2005; Searle et al., 2006), the FT protein activates the floral meristem identity genes APETALA1 (AP1), SOC1, and LFY. These have been identified as important floral pathway integrators in Arabidopsis. The expression of the LFY, AP1, and SOC1 orthologs, NtNFL, NtAP1, and NtSOC1 of tobacco was evaluated by real-time RT-PCR in the shoot apex of 45-day-old seedlings of T1 transgenic lines and wild-type (Smykal et al., 2007; Zhang et al., 2014). NtNFL (Figure 4G), NtAP1 (Figure 4H) and NtSOC1 (Figure 4I) were highly up-regulated in 35S::RoFT, 35S::RoFTmu1, 35S::RoFTmu2, and 35S::RoFTmu3 transgenic tobacco plants, which all showed an early-flowering phenotype. There was no obvious change in transcript levels of these endogenous genes in the 35S::RoFTmu4 and 35S::RoFTmu5 transgenic plants, which showed a late-flowering phenotype. Similarly, the expression of AtAP1, one of a downstream gene of FT, was up-regulated in 35S::RoFTmu3 transgenic Arabidopsis plant (Figure 5D).

According to the literature, both FT and TFL1 can interact with the bZIP transcription factor FD, which regulates the expression of several flower meristem (FM) identity genes (Abe et al., 2005; Benlloch et al., 2011). In order to examine whether Rosaceae FT/TFL1 homologs are able to interact with FD, and whether single amino acid substitutions in RoFT can affect the interaction, we performed yeast two-hybrid assays. Arabidopsis FD (AtFD) was used as a prey, and various FT/TFL1 homologs were cloned as the bait. Transformed yeast cells growing on SD/-Leu-Trp selection medium were shown in Supplementary Figure S3. The results indicated that in yeast, AtFD was able to interact with AtFT, RoFT and five RoFTmu1-5 point-mutated forms. However, no interaction was observed of AtFD with AtTFL1, FaTFL1, RoTFL1, or PhFT (Figure 6A). To further verify the interaction of FT/TFL1 homologs and AtFD, the N-terminal half of YFP fused to AtFD (AtFD-YFP^N) and the C-terminal half of YFP fused to FT (FT-YFP^C) were employed to perform BiFC test. YFP fluorescence was obviously observed in the nucleus (Figure 6B). The two results indicated that, except FaTFL1 and RoTFL1, the other FT/TFL1 homologs were able to interact with AtFD in the nucleus.

To test whether RoFT, point mutated RoFT and PhFT have the lipid-binding property, we performed a Fat-Western blotting using membrane-lipid overlay assays. All of the AtFT, RoFT, RoFTmu2/3/4/5 PhFT, AtTFL1, RoTFL1 and FaTFL1, with a C-terminal histidine tag, were expressed and purified (Figure 7A). The fusion proteins were hybridized with PC-spotted nitrocellulose membrane and detected using anti-His antibodies respectively. A clear binding of His-FT/TFL1 to PC was detected (Figure 7B) though these FT/TFL1 proteins have, not have or even in verse roles in flowering modulating.

The plant PEBP family can be divided into three major clades, i.e., the FT-like, MFT-like, and TFL1-like clades. The first two act as promoters of flowering, whereas TFL1-like clade acts as strong repressors of the response. Within the eight Rosaceae species, the FT/TFL1 homologs show high sequence identity (Supplementary Figure S1). Ectopic expression of PmTFL1, RoTFL1, and FaTFL1 in tobacco extended the vegetative phase and resulted in a significant delay in flowering. It is indicated that TFL1 homologs play a conservative role in controlling flowering time as previously reported for AtTFL1. However, most tobacco overexpressing PmFT and RoFT, displayed extremely advanced flowering. Contrarily, overexpression of FaFT did not promote flowering but, instead, caused slightly delayed by 1–2 months than the wild-type (Figures 2A,B and Table 1). The results demonstrated a divergence role of FT homologs between different species.

It has been reported that AtFT and AtTFL1 may demonstrate interchangeable roles by replacing a single amino acid (Hanzawa et al., 2005; Hou and Yang, 2009) or a larger protein segment (Ahn et al., 2006; Pin et al., 2010). Tyr-85 in AtFT and His-88 in AtTFL1 have been identified as two key residues that determine the respective FT and TFL1 functions (Hanzawa et al., 2005). It is interesting that Tyr-85 and His-88 are conserved in all FT and TFL1 proteins from the eight Rosaceae species, respectively (Supplementary Figure S1). Sequence comparison analyses showed that there are only 13 non-conserved substitutions between Rosa FT (RoFT) and Fragaria FT (FaFT), but nevertheless the two genes demonstrated opposite functions in controlling flowering time in transgenic plants (Figure 2A). In Arabidopsis, protein segment B, in conjunction with the adjacent segment C, has been implicated as essential for FT-like activity (Ahn et al., 2006). However, within this segment B we found only one residue is different between RoFT and FaFT, i.e., Glu-139 in RoFT compared to Gly-139 in FaFT and other FT homologs (Supplementary Figure S2). Thus, we suggest that protein segment B is not critical to the activity of FaFT as a flowering repressor. Previous study showed that FT protein is transported from the leaves, where it is synthesized, to the shoot apex where it then interacts with FD, and so leads to the activation of floral meristem identity genes AP1, LFY, and SOC1 (Abe et al., 2005; Wigge et al., 2005; Searle et al., 2006). The expression of the endogenous genes NtNFL, NtAP1, and NtSOC1 were highly up-regulated (49-, 127-, and 22-fold, respectively) in 35S::RoFT transgenic tobacco line #1 (Figures 4G–I). The three site-directed mutants RoFTmu1-3 acted as promoters of flowering in transgenic tobacco lines and ft-1 plants (Figures 4A, 5E), and resulted in the elevated expression of the endogenous genes, the same as seen in response to RoFT. By contrast, RoFTmu4 and RoFTmu5 demonstrated TFL1-like function in the flowering time, and the expression of NtNFL, NtAP1 and NtSOC1 in tobacco transformed with these constructs was about twofold higher than that of the control (Figures 4G–I). While we cannot rule out complexities that might arise from co-suppression in specific constructs, considering the consistent phenotypes between different ectopic transformants, it suggests that the phenotypes were due to the over-expression of different site-mutated RoFT.

PhFT from Petunia hybrida shares 71.0 and 54.4% identity with AtFT and AtTFL1, respectively, and it encodes a typical FT residue Tyr-85 and an important IYN triplet motif located in segment C. However, Lys-139 of PhFT differs from both counterparts from Arabidopsis FT (Gln-140) and TFL1 (Asp-144). Phylogenetic analysis placed PhFT in a cluster with FT-like genes (Figure 1C), suggesting a putative FT-like function. Over-expression of PhFT in tobacco did not promote early flowering (Figures 2D,E) instead, strongly suppressed flowering of the transgenic tobacco. With a mutant PhFTmu1 (K139Q), ectopic expression of PhFTmu1 in tobacco was found with late-flowering (Figure 4C). These results of transgenic analysis were highly reminiscent of the FT-like repressor activity of BvFT1 in sugar beet (Beta vulgaris subsp. vulgaris), which exists alongside its antagonistic paralog BvFT2. Although both of these Beta vulgaris genes encode Tyr-85/Gln-140 residues and the IYN triplet, they demonstrate a naturally evolved antagonistic function (Pin et al., 2010). Similar findings have also been found in the FT gene family of tobacco and Dimocarpus longan (Harig et al., 2012; Heller et al., 2014). Thus, the presence of Tyr-85, Gln-140 and triplet IYN residues is not sufficient to indicate whether the FT-like proteins undertake the role of flowering promoter or not. It has been reported that the three differing amino acids in segment B, forming an external loop, are the major cause of the BvFT1 and BvFT2 antagonistic function (Pin et al., 2010). However, analysis of the 14-amino-acid segment B of PhFT by crystal structure analysis indicated a close resemblance to the tertiary structure of Arabidopsis FT. Thus, further investigations are needed to elucidate the real reason why both the PhFT and PhFTmu1 proteins did not function to promote flowering in tobacco plants, as predicted according to their key sequence traits.

Previous reports described transgenic plants expressing the site-directed mutant TFL1 genes 35S::AtTFL1-H88Y (Hanzawa et al., 2005) and 35S::OnTFL1-H85Y (Hou and Yang, 2009) to show an early flowering phenotype, similar to that of Arabidopsis plants overexpressing native FT. Here, we have described transgenic tobacco plants over-expressing Rosa TFL1 (RoTFL1) and Fragaria TFL1 (FaTFL1) to show a late-flowering phenotype (Figure 2H). Specific mutations were introduced into these Rosaceae genes, corresponding to the putative key functional His-88 and Asp-144 residues of AtTFL1. However, these mutated genes did not result in early-flowering phenotypes in the transgenic plants (Table 3), which is thereby inconsistent with previous report. Based on our study in transgenic tobacco, key amino substitution is not sufficient to promote flowering via RoTFL1 and FaTFL1 (Figure 4D).

According to a previous report (Ahn et al., 2006), exon 4 of Arabidopsis FT plays a critical role in determining FT/TFL1 function. The exon 4 sequence contains four segments, A–D, and segments B and C are necessary for FT-like activity. These segments are also found in the TFL1 protein but, whereas the B and C sequences are highly conserved in many FT orthologs, they appear to have diverged in proteins with TFL1-like activity (Ahn et al., 2006). In the segment B encoded by RoFT, a single residue (Glu-139) is different from other FT homologs (Supplementary Figure S2). Thus, considering that the consensus sequence of FT orthologs contains a Gly residue at this corresponding site in the B segment and, despite this, RoFT still functions as a flowering promoter, we suggest that the contrary action of the FaFT gene-product as a floral repressor does not hinge on the sequence of segment B in exon 4. Among our five RoFT mutants, three mutants outside of IYN triplet led to an early flowering phenotype, similar to that mediated by over-expression of the unaltered RoFT gene. By contrast, two mutants within the IYN triplet motif of segment C, were not effective in the promotion of flowering and even to some extent, appeared to act similarly to a TFL1-like floral repressor (Figures 4A, 5A,E).

Using yeast two-hybrid assays, Jang et al. (2009) reported that Arabidopsis FT, but not TFL1, interacted with FD. However, Hanano and Goto (2011) used the BiFC technique to demonstrate that both TFL1 and FT can interact with FD within the plant cell nucleus (Hanano and Goto, 2011). In our yeast two-hybrid assays, FaTFL1 was found not to interact with FD, consistent with the findings of Jang et al. (2009) but different with Abe et al. (2005). However, we also found that PhFT, in spite of having high sequence similarity to FT, showed the same interaction pattern as FaTFL1. Our system was able to verify that native Arabidopsis FT interacted with FD. RoFT and the five corresponding point mutated protein forms were all shown to interact with AtFD in a similar way to the native Arabidopsis FT, which is also strongly supported by our BiFC system (Figure 6). In addition, ectopic overexpression of AtFD led to 2–3-months early-flowering in tobacco (Supplementary Figure S4 and Table S8), which showed that the AtFD is functionally active in tobacco as is the case of 35S::AtFD in Arabidopsis (Abe et al., 2005; Wigge et al., 2005). Since over-expression of the RoFTmu4 did not promote flowering in tobacco or Arabidopsis, we conclude that the physical interaction of FT homologs with the FD protein is not sufficient to bring about the promotion of flowering. These results also indicate that the substitution of a single amino acid residue of RoFT does not necessarily have a major impact on its interaction with FD but may, nevertheless, change its role in the control of flowering. Other interaction partners specific to FT or TFL1 are likely to exist, and this is supported by other studies (Jang et al., 2009; Taoka et al., 2011; Ho and Weigel, 2014). On the other hand, the diversity of interaction with AtFD in TFL1 homologs, verified by yeast two-hybrid and BiFC system, also show no correlation to their roles in flowering delaying. Though FT/TFL1 share a similar 3D structure with animal PEBP with an anion binding pocket, neither FT nor TFL1 were shown to bind any phospholipids in vivo. In another study, point mutation of the Arabidopsis FT at Asp71 located in the deep pocket did not affect FT activity (Ho and Weigel, 2014). So the significance of the pocket is unclear.

It has been reported that FT binds the phospholipid phosphatidylcholine (PC), a component of cellular membranes whose higher level accelerates flowering. Two models have been proposed to explain the effect of PC on flowering control (Nakamura et al., 2014). As a component of the nuclear membrane, PC may attract free FT from the cytosol into nucleus to promote flowering. Alternatively, PC-containing vesicles could help trafficking of FT to FD. Our FT-lipid assay result shows that whether they promote flowering or not, all FT/TFL1 homologs have the lipid-binding properties (Figure 7B). Thus, it is also deduced that lipid-binding and flowering promotion were two independent events. Considering TFL1 homologs have opposite function in controlling flowering, the PC-binging ability may imply other functions such as in mobile signaling. The TFL1 gene is transcribed in the central region of the SAM, and the protein spreads throughout the IM (dose not reach FM). By contrast, FT is produced in leaves and then is moved into SAMs (Bernier and Périlleux, 2005; Conti and Bradley, 2007; Wickland and Hanzawa, 2015). TFL1 was reported to play a role in endomembrane trafficking to protein storage vacuoles (PSVs) (Sohn et al., 2007). In addition to the fact that TFL1 protein is located in both the nucleus and cytoplasm, thus, TFL1 maybe shuttle FD from nuclei to PSVs, in nuclei where FT recruits FD, to block FD-dependent transcription occurs (Hanano and Goto, 2011). It also implies the TFL1 functions obviously in protein trafficking to PSVs from that the PC binding of His-TFL1 looks stronger than His-FT.

Collectively, beside description of the functional divergences in many FT/TFL1 homologs, our data have also shown that many novel amino acids change can switch FT-like activity to TFL1-like activity. On the other hand, it is also verified that the divergence of flowering time modulating by FT/TFL1 homologs is independent to its interaction and binding activities.