Flowering time in banana (Musa spp.), a day neutral plant, is controlled by at least three FLOWERING LOCUS T homologues

Banana is an important day neutral food crop with a long flowering/fruiting cycle that is affected by hot summers or cold winters in different places. Manipulating its life cycle requires an understanding of its flowering time machinery to bypass these stresses. Twelve FLOWERING LOCUS T (FT) and two TWIN SISTER OF FT (TSF) members were isolated from banana and their organization and expression pattern studied during development in two varieties that differ in flowering time namely Grand Nain (AAA genotype) and Hill banana (AAB genotype). The expression of at least 3 genes namely MaFT1, MaFT2 and MaFT5 (and to some extent MaFT7) increases just prior to initiation of flowering. These four genes and five others (MaFT3, MaFT4, MaFT8, MaFT12 and MaTSF1 could suppress the delayed flowering defect in the Arabidopsis ft-10 mutant and induce early flowering upon over-expression in the Col-0 ecotype. Most genes are diurnally regulated and differentially expressed during development and in various vegetative and reproductive tissues suggesting roles besides flowering. Subtle amino acid changes in these FT/TSF-like proteins provide interesting insights into the structure/function relationships of banana FTs vis-à-vis Arabidopsis. The studies provide a means for manipulation of flowering in banana for better management of resources and to reduce losses through abiotic stresses.


Results
Banana genome contains 14 FT/TSF-like genes. A total of 14 FT/TSF-like genes were identified from banana (Grand Nain) using a combination of 5′/3′RACE on pooled cDNA of different tissues as well as by genome walking on banana genomic DNA library. The complete genomic sequences of all 14 genes (named MaFT1-MaFT12, MaTSF1 and MaTSF2) with intron and exon boundaries are shown in Fig. 1. Full length cDNAs could be obtained for 13 genes, MaFT10 being the exception.
The coding sequences of all 14 genes were largely conserved, ranging from 516 to 540 bp with the second and third exons being invariant in size. The genomic structure of FT/TSF-like genes, however, varied a lot due to intronic size variations with the largest intron of 3098 bp being present in MaTSF1. Interestingly, while the FT/ TSF-like genes in most plants typically contain a conserved three intron structure, MaFT1, MaFT8, MaFT9 and MaTSF2 contained an additional fourth intron that split the fourth exon into two, thus changing the termination codon from that predicted in the database. To date, the presence of a fourth intron in FT like genes has only been reported in Gypsophila paniculata, a dicot 39 , but not in monocots.
The MaFT/TSF genes are distributed across the different chromosomes. Using the available published banana sequence of DH-Pahang variety 40 , the 14 FT/TSF-like genes were localized on different chromosomes (Fig. 2, Supplementary Table S1). Four of these namely MaFT4, MaFT6, MaFT7 and MaFT10 were localized on chromosome 2 while MaFT2, MaFT8 and MaFT11 were localized on chromosome 10. Of these, MaFT2 and MaFT11 (facing each other) were separated by a distance of only 2515 nt. The two most similar FTs, MaFT3 and MaFT12, were also located close to each other on chromosome 5, separated by 148328 nt. MaFT9 and MaTSF2 were located on chromosome 3 while MaTSF1, MaFT5 and MaFT1 were located on chromosomes 1, 4 and 9 respectively.
In order to understand the relationship between the FT-like genes from banana and other monocots, a phylogenetic tree was constructed (Fig. 3). The various FT-like polypeptides clustered into four subgroups namely IA, IB, IIA and IIB as reported 24,27 . Of these, subgroup IA which had functionally important FT proteins like Hd3aA and RFT1 (rice), HvFT1 and HvFT2 (barley), AcFT1 and AcFT2 (onion) contained MaFT2, MaFT5, MaFT6 and MaFT11. Subgroup IB contained the two TSF-like polypeptides MaTSF1 and MaTSF2 besides MaFT1 and MaFTs7-10 and also included Arabidopsis AtFT and AtTSF (Fig. S1). Interestingly, this sub-group with 87-93% similarity amongst the FTs, included all genes possessing the additional fourth intron. Sub-group IIA with ZCN8    Supplementary Tables S7 and S8. Amongst themselves, the MaFT polypeptides revealed more than 75% similarity (Supplementary Table S2). Of these, MaFT3 and MaFT12 (98.9% identity) differed in only two amino acids at positions 4 and 158 while MaTSF2 showed the lowest similarity (up to 56%) to other FT-like proteins. A comparison with functional FT proteins of other plants revealed MaFT11, MaFT7, MaFT5 and MaFT2 to be most similar to the functional AtFT, AtTSF and OsHD3 while MaFT3 and MaFT12 showed a greater similarity to ZCN8 (Supplementary Table S3).
Amino acid alignment with other FTs revealed a few interesting changes despite overall conservation amongst the different FTs (Fig. 4). The crucial tyrosine and glutamine residues at positions equivalent to Tyr85 and Glu140 in AtFT 41 remained as such in MaFT2, MaFT5, MaFT6, MaFT8-MaFT11, MaTSF1 and MaTSF2. The spacing between these two residues (55 amino acids) was also conserved in all except MaFT6. In MaFT3, MaFT4 and MaFT12 the glutamine (equivalent to Glu140 in AtFT) was replaced by histidine although this appears to be a functionally acceptable replacement as reported in the maize FT homologue ZCN8 42,43 and soybean GmFT5a 44 . Strikingly however, the crucial tyrosine (equivalent to Tyr85 in AtFT) was replaced by histidine in MaFT1 and MaFT7, a change that has been shown to be a characteristic of TFL1-like proteins 41 . The highly conserved B segment, crucial to FT function, differed most in MaFT6 in being 15 residues long (compared to 14 in FTs) and with a few substitutions within this normally invariant region. These included replacement of a conserved G (equivalent to G129 in AtFT) by a P, followed by an additional S that was absent in others. In addition, an A in place of T (T132 in AtFT) and a D in place of a conserved Y/F (Y134 in AtFT) was also noted. All members belonging to group IB, (MaFT1, MaFT7-10, MaTSF1 and MaTSF2) had certain prominent changes that included L to K/E/D at position 82, A to E at position 95, a G to V/I at 129, an E to A at position 149 and S/T to C at 170 (equivalent to positions in AtFT).
Several FT/TSF-like genes are activated prior to flowering initiation in two differently flowering banana genotypes Grand Nain and Hill Banana. In order to get an insight into the flowering related function of the banana FT/TSF-like genes, temporal changes in expression of these genes was studied at different stages of plant development. In Grand Nain, the first visible signs of floral differentiation are seen after day 180 within the pseudostem 45 but not before day 165. The transcript levels of MaFT1, MaFT2 and MaFT5 (as well as MaFT7 and to a lesser extent MaFT12), began to increase just prior to this stage from day 150 onwards with levels going significantly up (~2-6 folds) relative to their expression from 120-150 days (Fig. 5). There was little expression before this period except for MaFT7 and MaFT12 indicating that the rise in transcript level was likely related to flowering initiation. The expression of MaFT4, MaFT10 and to some extent MaTSF1 also began to increase prior to flowering. In addition, a minor peak at 120-135 days was seen for MaFT1, MaFT4, MaFT7 and MaFT12. MaFT11 increased just after flowering initiation from day 180 onwards with peak expression at 240-270 days that coincided with fruit development. Considerable expression during flower/fruit development was also seen for    49 . In view of this, the diurnal expression patterns of the different banana FT/TSF family members were investigated over a 24 hour period. As shown in Fig. 7, all the FT/TSF-like genes except MaFT3 and MaFT11 showed a diurnal or semidiurnal expression pattern. MaFT2, MaFT5, MaFT9, MaFT10, MaFT12, MaTSF1 and MaTSF2 showed a distinct peak around 2 pm. Of these, MaTSF1 expressed to high levels throughout the light period extending up to midnight. MaFT1, MaFT6 and MaFT7 peaked at dusk (around 6 pm) while MaFT4 and MaFT8 peaked in the morning at 10 am. Interestingly 9 of the 14 genes also showed a night time peak. For MaFT3, MaFT4, MaFT8, MaFT11 and MaTSF2 the night time peak was either greater than or as prominent as the day time peak while for MaFT5, MaFT9 and MaFT10 it was less prominent. In all cases except MaTSF2, the night peak was observed at 2 am, four hours before dawn.  The ability of the FT/TSF-like genes to impart early flowering by over-expression in the Col-0 background was also tested. As shown in Fig. 9(C and D) and Supplementary Fig. S3, over-expression of eight of the eleven genes tested under the CaMV35S promoter namely, MaFT1-5, MaFT8, MaFT12 and MaTSF1 led to early flowering MaFT7 expressers flowered earlier than Col-0 but later than the others in 24 days with 8-9 leaves. MaFT6, MaFT9 and MaTSF2 had no effect on flowering while MaFT10 and MaFT11 could not be tested.
An analysis of the expression of the floral identity gene AP1 which is a target of FT and necessary to initiate flowering was carried out in Arabidopsis lines expressing the different banana FT genes in the ft-10 background. A clear up-regulation of AP1 was seen in all lines (Fig. 10) except those expressing MaFT6, MaFT9 and MaTSF2 in spite of the genes being expressed in transgenic plants (Fig. S4). Relatively lower transcript levels were seen in lines expressing MaFT7 that matched the partial restoration phenotype of MaFT7.
Collectively these studies show that at least nine of the 14 FT/TSF like genes namely MaFT1-MaFT5, MaFT7, MaFT8, MaFT12 and MaTSF1 could functionally interact with the flowering components even in Arabidopsis suggesting a structure that is conserved with AtFT and capable of being recognized by the FT/FD complex of Arabidopsis.

Banana genome contains 14 FT/TSF-like genes.
Banana is one of the most important day neutral fruit crops forming the staple food of many African and Asian countries. In the tropics, banana flowers in 7-9 months and completes its first fruit harvest in about 12 months regardless of the season. Although it is grown in most places in Central India under micro-irrigated conditions, its long life cycle makes it susceptible to water stress  under the hot summer conditions or to cold stress in North Indian winter conditions. Hence manipulation of its life cycle for reduction of flowering time is desirable. Yet, in spite of its economic importance, there have been no detailed studies on the PEBP family genes that play a key role in flowering time regulation.
The current studies in the triploid (AAA) variety of dessert banana Grand Nain, aided by the recently completed AA double haploid genome sequence 40 show the presence of 12 FT-like and two TSF-like genes. This is a large number compared to dicots but appears to be a feature of monocots exemplified by the presence 13 in rice 25 , 15 in maize 24 , 13 in sorghum 50 , 9 in barley 25 , at least 6 in onion 27 and 6 each in Brachypodium and wheat 28 . Although varying at the genomic level due to intronic variation, exonic structures are largely conserved with exons 2 (62 nt) and 3 (41 nt) being invariant in size. This feature of the exonic structure has been reported across plants in both monocots such as maize 24 , rice 25,51 , barley 26 , sugarcane 52 and in eudicots, Arabidopsis 5, 6 , apple 23 , tomato 20 , poplar 21 , Medicago 53 etc. Interestingly, MaFT1, MaFT8, MaFT9 and MaTSF2 displayed four introns instead of three that have so far been observed in all plants except the dicot Gypsophila paniculata 39 . The additional intron interrupted the last exon exactly after 217 bp in all the four genes just as observed in G. paniculata 39 which is surprising considering the evolutionary distance between the two. Further studies in other plant FTs may shed light on how widespread the fourth intron is and on the origin of the additional intron/exon in a gene family that has an otherwise highly conserved exonic structure.
Banana has more than one FT/TSF-like flowering inducer genes. The 14 FT/TSF-like proteins show greater than 75% similarity in their amino acid sequence amongst themselves and more than 60% with AtFT, AtTSF1 and OsHd3a suggesting high conservation. Indeed, functionality studies carried out in Arabidopsis show that eight of the twelve genes studied, namely MaFT1-MaFT5, MaFT8, MaFT12, and MaTSF1 could efficiently suppress the delayed flowering phenotype of the ft-10 mutant while MaFT7 weakly suppressed the defect. These genes also imparted early flowering in the Col-0 ecotype upon over-expression indicating the ability of the encoded proteins to recognize and interact with components of the flowering machinery even in a distantly related plant like Arabidopsis. The results thus highlight the remarkable conservation in FT structure/function in plants as diverse as banana and Arabidopsis. Interestingly, MaFT6, MaFT9 and MaTSF2 failed to suppress the ft-10 phenotype in spite of high similarity to other FTs suggesting that certain amino acid differences within these proteins might prevent their recognition by the flowering machinery components in plants. An exhaustive analysis of amino acids and determinants necessary for FT function has already been performed through several studies 22,41,54,55 . The differences due to genetic diversity in banana FTs provides additional interesting study material to reveal further intricacies in the FT structure and function. As an example, MaFT9 and MaTSF2 in spite of being very similar to MaFT8 and MaTSF1 respectively (Figs 3 and 4) are unable to suppress the flowering defect in ft-10 or induce early flowering in Col-0 upon over-expression. This implies that certain residues that differ between the pairs MaFT8/MaFT9 and MaTSF1/MaTSF2 may be important in deciding FT functionality at least in the context of the Arabidopsis flowering machinery. An important finding of our studies is that the Y to H substitution in both MaFT1 and MaFT7 at the position equivalent to amino acid 85 in AtFT, is not deleterious to their function as FT unlike that reported by Hanzawa et al. 41 . This is evident from the ability of both these proteins to suppress the flowering defect in ft-10 (albeit weakly in case of MaFT7) as well as induce early flowering through over-expression in Col-0 despite this change. This suggests that both FTs can tolerate this change so long as there are other simultaneous changes (as yet unidentified) within the proteins that suppress the effect of the Y to H change and allow them to continue to function as FTs. At least the crucial amino acids that supposedly interact with Y through hydrogen bonding 55 , do not appear to differ between MaFT1/MaFT7 and other FTs indicating other subtle changes within these two FTs. So far, MaFT1 and MaFT7 represent the only two proteins in literature with the Y to H substitution at position 85 (equivalent to Arabidopsis AtFT) where early flowering has been reported. The differences in the abilities of these two proteins to suppress the ft-10 defect might be due to subtle differences within their sequences. The ability of MaFT5 to induce flowering earlier than any other banana gene in both the Col-0 background and the mutant ft-10 background also highlights the fact that differences in regions other than the B segment may influence FT function. Simultaneously, the importance of the conserved B segment, which is invariant with 14 residues in FTs 54 , is also apparent from the inability of MaFT6 to promote flowering. MaFT6 contains a 15 amino acid B segment with crucial changes in the conserved residues. B segments longer than 14 residues and varied in amino acid composition are features of plant TFL1-like genes. However, unlike TFL1s, MaFT6 did not delay flowering over that of control upon over-expression in Arabidopsis. Nevertheless, one cannot rule out MaFT6 as a flowering gene since minor differences in the FT proteins (that prevent interactions in Arabidopsis as seen for MaFT6, MaFT9 and MaTSF2) may be associated with complementing differences in the interacting partners such as FD thus still allowing an efficient complex to be formed in banana.
Although a large number of the FT/TSF-like genes in banana can function like AtFT, only a few can be actually correlated with the transition from vegetative to reproductive growth that occurs between 170-190 days in Grand Nain and a little later in Hill banana. As expected from such a function, transcript levels of MaFT1, MaFT2 and MaFT5 begin increasing between 150-180 days, going up by 2-6 fold compared to their levels at 120-150 days not only in Grand Nain but also in the delayed flowering Hill banana. Five others namely MaFT4, MaFT7, MaFT10, MaFT12 and MaTSF1 also showed an increase in transcript accumulation between 150/165-180 (MaFT7, MaFT10 and MaFT12) and 120-150 (MaFT4 and MaFT7) days indicating a possible flowering related function. The expression of most of these was largely conserved even in Hill banana, and delayed slightly in tune with the delay in appearance of the inflorescence. Since the Hill banana is taller than Grand Nain at least some delay in the visible inflorescence may be due to the greater distance that has to be covered by the growing inflorescence in Hill banana rather than entirely due to a delay in development of the inflorescence meristem. This might explain the lesser delay in expression pattern of the FT genes in Hill banana compared to the actual delay observed between Grand Nain and Hill banana. The coordinated increase in transcript levels of seven of the 14 FT/TSF genes prior to flowering may provide a robust means for initiation of flowering. However, identifying the relative contribution of individual genes to flowering will require further studies through genetic manipulation which is somewhat difficult in banana. Whether the diurnal cycling of the transcripts with day and night time peaks might also be linked to flowering is difficult to say as of now.
Other genes like MaFT7, MaFT9, MaFT3 and MaTSF2 that show a peak in transcript levels soon after transplanting may have additional functions. This period coincides with the formation of ratoons. In plants like onion and potato, where bulbing and tuberization is observed, the expression of FT genes like AcFT1 and AcFT4 in onion and StSP3A in potato has been correlated with bulb 27 and tuber formation 37 respectively. An interesting possibility could be a similar function for these genes in formation of suckers or ratoons. Many genes are also transcribed in tissues like fruit pulp, peel, flowers and bracts at levels higher than those in mature leaves and persist long after flowering initiation. The expression in these tissues is similar to reports in apple 23 , maize 24 and Medicago 53 and clearly suggests a role that must be different from flowering and possibly more diverse in monocots given the large expansion of the family. For instance, in maize only 8 of the 15 FT like genes were transcribed in leaves and only ZCN8 has so far been associated with flowering suggesting other roles for the other members. In contrast to maize, where tissue/stage specific splicing differences were observed, most of the banana FT/TSF like members showed only the completely spliced form in all tissues studied (data not shown) suggesting that regulation by alternative splicing is unlikely in banana. Knocking out these genes, although time consuming and difficult due to the triploid nature of dessert banana and its inability to set seeds, may provide a clue to their function in these tissues and is currently in progress. Nevertheless, the structural similarity amongst the different proteins (that act in flowering and non-flowering functions) suggests that they might have similar interacting surfaces for other partners (possibly other members of the bZIP and 14-3-3 family) even when they are active in non-flowering related functions.
Taken Genomic DNA isolation and Genome walking library preparation. Genomic DNA was isolated from leaves by the modified plant DNA mini-preparation method 57 . The genome walking library was prepared with isolated genomic DNA using the Genome Walker Universal kit (Clontech Laboratories, Inc. USA) as per manufacturer's instructions.
RNA extraction, cDNA and SMART RACE library preparation. RNA was isolated from mature leaves of tissue cultured plants at the hardening stage, six and nine month old field growing plants, developing apical inflorescence (inside the pseudostem of seven month old plants) and various flower parts as described 58 with minor modifications and purified using RNeasy Plant Mini Kit (Qiagen, Germany). A mixed cDNA library was prepared from this RNA.
For tissue specific studies, total RNA was extracted from juvenile leaves of hardening stage plants, mature leaves, bracts, apical inflorescence, mature flowers (from 8 month old flower bearing plants), mature fruit skin and pulp (from 9-10 month old fruit bearing plants) as described by Chaurasia et al. 45 .
For developmental expression studies, leaf samples were collected every 15 days (Grand Nain) and 45 days (Hill Banana) after planting for a period up to 255 days. For each sample, the fully mature, expanded, newly emerged leaf was chosen. Samples were collected from ten independent field plants between 1 to 2 PM and two pools of five independent leaves were prepared.
For diurnal analysis of gene expression, leaf samples (fully mature, expanded, newly emerged leaves) were collected every four hours from 150 day old field growing plants as described above in triplicates.
First strand cDNA was synthesized using the iScript cDNA synthesis kit (Bio-Rad) along with the 3′AP (5′ AAGCAGTGGTATCAACGCAGAGTACTTTTTTTTTT TTTTTTTT3′; Invitrogen) as per manufacturer's instructions. Separate SMART cDNA libraries (5′ and 3′ SMART) were prepared using SMART RACE cDNA Amplification Kit (Clontech Laboratories, Inc. USA). A pooled mix of total RNA from various tissues was used for library preparation. Sequences of all the primers used are provided in Supplementary Table S6 Table S6) based on these partial sequences were used on SMART cDNA libraries and Genome Walker libraries (Clontech Laboratories, Inc. USA) to obtain the flanking sequences for cDNA (inclusive of 5′ and 3′ UTRs) as well as genomic DNA. Once the composite sequence was available, primers were designed