Identification of cotton MOTHER OF FT AND TFL1 homologs, GhMFT1 and GhMFT2, involved in seed germination

Plant phosphatidylethanolamine-binding protein (PEBP) is comprised of three clades: FLOWERING LOCUS T (FT), TERMINAL FLOWER1 (TFL1) and MOTHER OF FT AND TFL1 (MFT). FT/TFL1-like clades regulate identities of the determinate and indeterminate meristems, and ultimately affect flowering time and plant architecture. MFT is generally considered to be the ancestor of FT/TFL1, but its function is not well understood. Here, two MFT homoeologous gene pairs in Gossypium hirsutum, GhMFT1-A/D and GhMFT2-A/D, were identified by genome-wide identification of MFT-like genes. Detailed expression analysis revealed that GhMFT1 and GhMFT2 homoeologous genes were predominately expressed in ovules, and their expression increased remarkably during ovule development but decreased quickly during seed germination. Expressions of GhMFT1 and GhMFT2 homoeologous genes in germinating seeds were upregulated in response to abscisic acid (ABA), and their expressions also responded to gibberellin (GA). In addition, ectopic overexpression of GhMFT1 and GhMFT2 in Arabidopsis inhibited seed germination at the early stage. Gene transcription analysis showed that ABA metabolism genes ABA-INSENSITIVE3 (ABI3) and ABI5, GA signal transduction pathway genes REPRESSOR OF ga1-3 (RGA) and RGA-LIKE2 (RGL2) were all upregulated in the 35S:GhMFT1 and 35S:GhMFT2 transgenic Arabidopsis seeds. GhMFT1 and GhMFT2 localize in the cytoplasm and nucleus, and both interact with a cotton bZIP transcription factor GhFD, suggesting that both of GhMFT1, 2 have similar intracellular regulation mechanisms. Taken together, the results suggest that GhMFT1 and GhMFT2 may act redundantly and differentially in the regulation of seed germination.


Introduction
In the life cycle of higher plants, the transition from vegetative to reproductive growth (floral transition) is controlled by endogenous cues and external cues [1][2][3]. Previous studies on the PLOS  molecular mechanisms of flowering initiation in Arabidopsis thaliana have uncovered six major pathways to orchestrate flowering time, such as photoperiod, vernalization, ambient temperature, gibberellin (GA), autonomous and age pathways [4]. Another transition in a flowering plant is from embryonic growth to germination growth [5]. Seed dormancy can form a soil seed bank and open a protective mechanism for the seed in an unfavorable environment [6]. Seed germination plays an essential part during the growth of the descendants and it is a prerequisite for crop yield [7]. Breaking dormancy is affected by internal cues and external cues [8], which are primarily controlled by two major antagonistic phytohormones, abscisic acid (ABA) and GA [9]. It is well known that ABA functions in developmental processes of many plants like seed germination and dormancy. ABA inhibits seed germination by hampering water absorption as well as the rupture of seed coat and endosperm [10,11]. The extensive molecular genetics analyses reveal that ABA-INSENSITIVE3 (ABI3) and ABI5 are the essential regulators in regulating ABA to boost seed dormancy and suppress seed germination [12][13][14][15]. GA can counteract the effects of ABA and promote seed germination, plant growth and early flowering. During seed germination, GA accumulation is accompanied by a decrease of ABA content, indicating that GA and ABA play opposite roles in seed germination [16,17]. GA3-oxidase (GA3OX) is a pivotal enzyme controlling GA biosynthesis, which is encoded by four genes GA3OX1, GA3OX2, GA3OX3 and GA3OX4. The ga3ox1 ga3ox2 double mutant exhibits serious defects in seed germination, indicating that GA3OX1 and GA3OX2 are crucial genes regulating seed germination [18,19]. DELLA proteins, consisting of REPRESSOR OF ga1-3 (RGA), GA-INSENSITIVE (GAI), and RGA-LIKE1-3 (RGL1-3) [20-24], act as repressors in regulating GA signaling, among which RGA and RGL2 play roles in inhibiting seed germination by mediating the interaction between GA and ABA [14,22,25]. MOTHER OF FT AND TFL1 (MFT), which encodes a phosphatidylethanolamine-binding protein (PEBP), has an evolutionary and functionally conserved role in most plants [26][27][28][29]. In A. thaliana, the PEBP family genes are mainly composed of three clades: FT-like, TFL1-like and MFT-like [30]. It is generally considered that MFT-like branch is the evolutionary ancestor of the other two branches [26,29]. FT-like and TFL1-like are two important regulators with opposite functions in the control of flowering time and plant architecture [31][32][33][34][35][36]. Compared with FT and TFL1, the exact biological function of MFT is not well understood. There is only one MFT gene in Arabidopsis genome [37]. Overexpression of MFT leads to slightly early flowering compared with that of wild-type plants, and loss-of-function mutation in MFT does not exhibit observable phenotypes, suggesting that MFT acts redundantly in regulating flowering time in Arabidopsis [37]. MFT homologs have been characterized from several plant species, and recent studies have shown that they have different roles in controlling flowering time in respective species. For example, Hou et al. [38] reported that Adiantum capillus-veneris MFT (AcMFT) accelerated the fioral transition and partially rescued the late flowering phenotype of Arabidopsis ft-1 mutant. However, two groups reported that both Dendrobium nobile MFT (DnMFT) [ [43], Glycine max [44], Actinidia chinensis [45] and Citrus latifolia [46]. In addition to regulating flowering, MFT homologs have also been found to be involved in seed dormancy and germination [5,40,44,[47][48][49][50][51][52][53]. In Arabidopsis, MFT functions as a negative regulator of germination under far-red light conditions by modulating ABA and GA signaling [50]. Expression of MFT is directly regulated by two key transcription factors, ABI3 and ABI5, in response to ABA [5]. Triticum aestivum MFT (TaMFT) also inhibits seed germination and functions as a positive regulator of dormancy [47].
The allotetraploid cotton (Gossypium hirsutum and Gossypium barbadense) with a complex genome is the world's leading sources of natural fiber crops and crude oil [54]. In recent years, the whole genome has been sequenced successfully in cotton (Gossypium spp.), which provides a resource for characterization of gene family [55][56][57][58][59][60][61][62]. Genome-wide analysis reveals that there are at least eight PEBP homoeologous gene pairs in tetraploid G. hirsutum [63][64][65]. Our group has demonstrated that the FT homolog GhFT1 controls flowering time and TFL1 homolog SELFING-PRUNING (GhSP) regulates fruiting branches architecture in cotton [66][67][68], and similar findings have also been reported in other groups [69][70][71]. However, the exact biological functions of GhMFT homologs in Upland cotton remain unclear. We characterized two MFT homoeologous genes, GhMFT1 and GhMFT2, from G. hirsutum in this study. We found that their expressions were upregulated significantly in the developing ovule and their expressions were obviously declined and responded to ABA and GA during seed germination. Furthermore, ectopic overexpression of GhMFT1 and GhMFT2 in Arabidopsis inhibited seed germination at the early stage. These preliminary results suggest that GhMFT homologs may be involved in ovule development and serve as the potential negative regulators in seed germination.

Plant materials and growth conditions
The seeds of G. hirsutum cv. Xinluzao 33 were planted in the experimental fields of Shihezi University (Xinjiang, China). A. thaliana mft-2 (in the wild type (Columbia, Col-0) background) was ordered from the Arabidopsis Biology Resources Center (ABRC, Columbus, OH, USA). Seeds of wild type and mft-2 were surface sterilized and planted as described previously [66]. After 10 days, the seedlings were transplanted to soil in a growth chamber under longday (LD) conditions (16-h-light/8-h-dark, ambient temperature of 22˚C, light intensity of 200 μmol photons m -2 s -1 ). For tissue expression analysis, roots, stems, shoot apical meristems (SAM) and leaves were collected at the third true leaf expanding stage (20 days after planting). A whole flower was collected at the flowering stage. Cotton bolls were harvested at the following time-points during development: -3 and 0 d of anthesis (DOA) ovules, 3 and 8 d of postanthesis (DPA) ovules which contain initiating fiber cell, and 12-30 of DPA ovules. All samples collected were immediately frozen in liquid nitrogen and stored at -80˚C.

Sequence alignment and phylogenetic analysis
Cotton MFT homologous genes were obtained through tBLSATn searches using AtMFT as query against the Upland cotton (AD) 1 tetraploid genome of G. hirsutum [58,60], the sea land cotton (AD) 2 tetraploid genome of G. barbadense [59], the A 2 genome of diploid G. arboreum [57] and the D 5 genome of diploid G. raimondii [56] at COTTONGEN (http://www. cottongen.org; S1 Table). The amino acid sequences of MFT proteins from other plant species used in this study were downloaded from NCBI non-redundant database (S2 Table). Multiple sequence alignment was performed with Clustal W using the default parameters. A phylogenetic tree was constructed using Molecular Evolutionary Genetics Analysis software 6.0 [72] (neighbor-joining, Poisson correction distance model). The nodal reliability in the tree was evaluated by bootstrap analysis with 1000 replicates. Putative cis-acting regulatory elements of MFT promoters were analyzed based on the description of Xi et al. [5].

Gene expression analysis
Total RNA for each sample was isolated using the RNAprep Pure Plant Kit (Polysaccharides & Polyphenolics-rich) (TIANGEN, Beijing, China) according to the manufacturer's protocol. Total RNA was reversed to cDNA using an M-MLV Reverse Transcriptase Kit (Bioteke Corporation, Beijing, China). Quantitative Real-time PCR (qRT-PCR) was carried out on an Applied Biosystems 7500 Fast Real-Time PCR System (Life Technologies, Carlsbad, CA, USA) in a 20 μL volume containing 100 ng of cDNA, 4 pM of each primer, and 10 μL SYBR Green PCR Master Mix system (TaKaRa). The PCR conditions and the calculation method of gene expression were the same as what had been described previously [68]. Information on the qRT-PCR primers for gene expression analysis and gene cloning used in this study was listed in S3 Table. The nucleotide sequences of GhMFT homoeologous genes marked with primer location for qRT-PCR were shown in S1 Fig. A cotton Ubiquitin7 (GhUBQ7, GenBank accession no. DQ116441) gene and an Arabidopsis Actin2 (AT3G18780) gene were used as internal controls, respectively. Three replicate assays were conducted with separately isolated RNA, and three technical triplicates were performed for each PCR reaction.

Plasmids construction and Arabidopsis transformation
Complete open reading frame (ORF) cDNAs of GhMFT1 and GhMFT2 were amplified by RT-PCR using gene specific primers (S3 Table). The ORF cDNAs were separately cloned into pMD TM 19-T Vector Cloning Kit (TaKaRa). Two constructs confirmed by sequences analysis were then introduced into pCAMBIA2300-35S-OCS [66] binary vectors containing downstream of the cauliflower virus 35S promoter to generate 35S:GhMFT1 and 35S:GhMFT2. Agrobacterium tumefacien strains GV3101, harboring 35S:GhMFT1 and 35S:GhMFT2 constructs, were used to be transformed into wild type and mft-2 mutant plants using the floral dip method [73], respectively. Homozygous transgenic plants were screened and identified as described by Guo et al. [66]. Flowering time was monitored as the number of rosette leaves per plant and days to flowers for the first flower bloomed after transplanting the seedlings into the soil [68].

Seed germination assays
100 mature cotton seeds with uniform size were sterilized with 0.1% (w/v) mercuric chloride for 10 min, and then rinsed several times in sterile water. After removing the seed coats, the sterilized seeds were sown on the Petri dishes with half-strength Murashige and Skoog (MS) salt (Murashige & Skoog, Duchefa, pH 5.7) mixture, 1% (w/v) sucrose and 0.8% (m/v) agar, and the number of germinating seeds was counted. For the abiotic treatments, the sterilized seeds were planted on half-strength MS medium supplemented with 50 μM GA 3 (Sigma-Aldrich) and 100 μM ABA (Sigma-Aldrich) according to the previous studies [74,75], respectively. Petri dishes were placed in a phytotron at 28˚C under dark conditions. When the primary root length reaches 1 mm, the seed is considered to be germinated [76]. Seeds were collected at 12 h and 24 h of imbibition for RNA isolation.
Seeds of wild type, mft-2 and each transgenic homozygote were sterilized as described previously [66]. Sterilized seeds were then plated on the Petri dishes with half-strength MS medium. Petri dishes were placed in a phytotron at 22˚C under LD conditions (16-h-light/ 8-h-dark). The method of abiotic stress treatments was as described above. The concentrations of GA 3 and ABA are 5 μM and 10 μM, respectively. For the germination assay, at least 100 seeds for each genotype were observed, and the germination percentage was calculated according to three independent experiments.

Subcellular localization analysis
To analyze the subcellular localization of the GhMFT1 and GhMFT2 proteins, we utilized the 35S:GFP vector constructed by Guo et al. [66]. The coding regions of GhMFT1 and GhMFT2 without stop codon were separately amplified by PCR and inserted into the Kpn I and BamH I sites of the 35S:GFP vector to generate 35S:GhMFT1-GFP and 35S:GhMFT2-GFP in-frame fusions which were transformed into A. tumefaciens strains GV3101, respectively. The abaxial surface of leaf blade of Nicotiana benthamiana was then infiltrated with A. tumefaciens strains according to the description of Si et al. [68]. The infiltrated leaves were used to detect GFP fluorescence using a confocal laser scanning microscopy (CLSM510; Zeiss, Jena, Germany).

Yeast two-hybrid assays
The coding sequences of GhMFT1 and GhMFT2 were amplified and cloned into pGBKT7 (Clontech) to produce BD-GhMFT1 and BD-GhMFT2, respectively. AD-GhFD was constructed by Si et al. [68]. Yeast two-hybrid assays were carried out according to the method of Si et al. [68].

Bimolecular fluorescence complementation (BiFC) assays
The coding regions of GhMFT1 and GhMFT2 were separately amplified and cloned into the pDONRZeo vector (Invitrogen) for fusion with the N-terminus of PVYNE [77] by LR reaction. GhFD coding region has been amplified and fused into the C-terminus of PSCYCE vectors [68]. The BiFC assays were performed as described by Si et al. [68].

Identification and phylogenetic analysis of GhMFT homologs from G. hirsutum
Genome-wide identification revealed that there were two MFT loci in each A 2 and D 5 diploid cotton genome and four in each tetraploid genome with two homoeologous genes in each A and D subgenome of the (AD) 1 and (AD) 2 tetraploid (S1 Table). Our findings are in good agreement with those of published papers [64,65,71]. To explore their functions in Upland cotton, their ORF sequences were successfully cloned from G. hirsutum using gene specific primers (S3 Table).  Table) indicated that GoMFT proteins contain the conserved D-P-D-x-P and G-x-H-R motifs (Fig 1A), which are also present in both FT-like and TFL1-like proteins. The conserved key amino acid residues Tyr85 (Y)/His88 (H) in FT/ TFL1-like protein are replaced by Trp83/85 (W) in GoMFT1 and GoMFT2, respectively, suggesting that GoMFT homologs do not play a central role in flowering time control. There is a conserved proline (P) at the end of the carboxyl group, which is only found in MFT-like, but no similar amino acid has been found in FT/TFL1-like [26]. Phylogenetic analysis revealed that GhMFT1 showed a closer genetic relationship to Glycine max MFT (GmMFT), Citrus unshiu MFT (CuMFT) and Jatropha curcas MFT1 (JcMFT1); whereas GhMFT2 showed a closer genetic relationship to Citrus limon MFT (CiMFT), JcMFT2 and HbMFT1 ( Fig 1B).

Expression of GhMFT homoeologous genes in G. hirsutum
To further understand the potential functions of GhMFT homoeologous genes, expression patterns of two Upland cotton homoeologous MFT genes were analyzed in different tissues including roots, stems, leaves, the SAM, fiowers, and at different developmental stages of ovules using qRT-PCR. The transcripts for two homoeologous MFT gene pairs were detected in all the investigated samples with a certain difference in their expression levels (Fig 2). The expression level of GhMFT1-A in roots, leaves and the SAM was higher than that of GhMFT1-D, whereas GhMFT1-D was highly expressed in ovules (Fig 2A). During ovule development, GhMFT1 homoeologous genes had similar expression patterns. Their expression levels were significantly increased at the ovule of 16 DPA, and were continuously upregulated and peaked at ovule of 30 DPA. The expression of GhMFT1-D was higher than that of GhMFT1-A during ovule development. GhMFT2 homoeologous genes were also expressed in roots, stems, leaves, the SAM and flower, with the highest expression level in flower ( Fig 2B). Furthermore, the expression level of GhMFT2-D was significantly higher than that of GhMFT2-A in each tissue except in SAM. During ovule development, the expression level of GhMFT2-D was significantly higher than that of GhMFT2-A. GhMFT2-D was upregulated during ovule development from -3 DOA to 12 DPA and peaked at 12 DPA, and then downregulated. However, the expression level of GhMFT2-D in the ovule of 25 DPA reached the highest point during the of ovule development. Our results revealed that the expression patterns of GhMFT1 and GhMFT2 homoeologous genes were slightly different, suggesting that they may play differential roles in regulating ovule development.
To explore whether GhMFT homologous genes played roles in cotton seed germination, we tested their expression levels during seed imbibition. The results revealed that expressions of both GhMFT homologous genes decreased during cotton seed imbibition. The mRNA levels of GhMFT1 homoeologous genes gradually decreased during 0-12 h of imbibition, after that they maintained at relatively stable levels ( Fig 2C). However, expressions of GhMFT2 homoeologous genes decreased immediately in the early stages of seed germination (0-3 h), and downregulated gradually during 3-9 h, and kept at a relatively stable level after 9 h ( Fig 2D). Furthermore, our results showed that the expression levels of the A subgenome were higher than that of D subgenome, implying that GhMFT1-A and GhMFT2-A may play major roles during seed germination.
Six putative ABA-responsive elements (ABREs) were found to be located~1850 bp upstream of the start codon of AtMFT (S3 Fig). These ABREs are necessary for ABA to regulate AtMFT expression, and AtMFT acts as a negative regulator in response of ABA [5]. Several putative ABREs were also identified in the approximate 1.8 kb upstream of the initiation codon of two GhMFT homoeologous genes through promoter analysis. There is a single ABRE located~153 bp upstream of the translational start site of GhMFT1 homoeologous genes, whereas a separate cluster of three ABREs located~142 bp upstream of the start codon of GhMFT2 homoeologous genes (S3 Fig). The presence of these ABREs in the promoters of GhMFT homoeologous genes hints that ABA might regulate their expressions. We next investigated whether the expression levels of GhMFT homoeologous genes were influenced by GA or ABA. To answer this question, we detected changes in their transcription levels in response to exogenous GA and ABA treatments at 12 and 24 h of imbibition by qRT-PCR. As is shown in Fig 3B, the control seeds and seeds treated by GA started to germinate after 9 h of imbibition, and GA treatment slightly promoted seed germination, whereas this process was obviously inhibited by ABA. These results indicated that GA and ABA treatments have certain effects on cotton seed germination. In this process, the expression levels of GhMFT homoeologous genes were significantly increased in response to ABA treatments (Fig 3C-3F). However, their expressions showed different expression patterns in response to GA treatment. The expression of GhMFT1-A was notably decreased in response to GA treatment (Fig 3C).
Amino acids shown in red and green indicate three conserved amino acid His (H)/Tyr (Y)/Trp (W) residues for FT/TFL1/MFT-clade and Pro (P) residue for MFT-clade, respectively. (B) Phylogenetic analysis of plant PEBP family based on amino acid sequences. The bootstrap consensus tree was inferred from 1000 replicates using MEGA 6.0 [72].
However, there were no significant changes in the expression levels of GhMFT1-D and GhMFT2 homoeologous genes under GA treatment (Fig 3D-3F). In summary, these results indicated that GhMFT homoeologous genes may be involved in the control of seed germination of cotton in response to ABA.

Overexpression of GhMFT1 and GhMFT2 do not affect the flowering time in transgenic Arabidopsis
To investigate whether GhMFT homologous genes were involved in the control of flowering time and plant architecture, we generated homologous transgenic Arabidopsis plants in wild type or mft-2 background. Eight independent 35S:GhMFT1 homozygotes and 10 independent 35S:GhMFT2 homozygotes in wild type background were generated; and eight independent

Overexpression of GhMFT1 and GhMFT2 inhibits seed germination in transgenic Arabidopsis
Considering the expression patterns of GhMFT1 and GhMFT2 during seed germination, we guessed that they may play important roles in the regulation of seed germination. To confirm this hypothesis, the seed germination rates of different 35S:GhMFT1 and 35S:GhMFT2 transgenic lines were compared with those of wild type and mft-2, respectively. The result revealed that the germination rate of 35S:GhMFT1 transgenic seeds in wild type background was much lower than that of wild type at the early stage of seed germination (1 d after imbibition) ( Fig  4A), and the germination rate of 35S:GhMFT2 transgenic seeds remained much lower than that of wild type on the second day (Fig 4B). Similarly, the 35S:GhMFT1 and 35S:GhMFT2 transgenic seeds in mft-2 background also had lower germination rates than those of mft-2 within 3 d of imbibition (Panels A and B in S7 Fig). In addition, we noticed a correlation between the expression levels of GhMFT1 and GhMFT2 and the germination rates of transgenic lines not only in wild type but also in mft-2 background (Fig 4A and 4B To investigate whether the seed germination of 35S:GhMFT1 and 35S:GhMFT2 was influenced by GA and ABA treatments, we next analyzed the germination rates of all transgenic plants and controls under two treatments, respectively. The results revealed that GA treatment boosted germination of all plants, but the germination rate of wild-type seeds was higher than those of 35S:GhMFT1 and 35S:GhMFT2 transgenic plants in wild type or mft-2 background (Fig 4C and 4D; Panels C and D in S7 Fig). The germination rates of the 35S:GhMFT1/2 mft-2 transgenic lines were increased when exogenous GA was applied, but their germination rates could not reach the same level as that of mft-2 in normal conditions (Panels C and D in S7  Fig). However, the seed germination rate of mft-2 showed little difference in response to exogenous GA compared with that of wild type. Under ABA treatment, 35S:GhMFT1 and 35S: GhMFT2 transgenic plants in the wild type background began to germinate from the third day, and the germination rates were significantly lower than that of wild type (Fig 4E and 4F), whereas the transgenic plants seeds in the mft-2 background began to germinate from the fifth day, and the germination rates were significantly lower than that of mft-2 (Panels E and F in S7  Fig). Moreover, the germination rates of wild type and mft-2 were lower under ABA treatment than that in normal conditions. These results revealed that ectopic expression of GhMFT1 and GhMFT2 repressed seed germination of Arabidopsis at the early stage.
Germination percentages of cotton seeds treated with 50 μM GA and 100 μM ABA. Expression of GhMFT1-A (C) and GhMFT1-D (D) in cotton seeds at 12 h and 24 h of treatment with 50 μM GA and 100 μM ABA. Expression of GhMFT2-A (E) and GhMFT2-D (F) in cotton seeds at 12 h and 24 h of treatment with 50 μM GA and 100 μM ABA. A cotton Ubiquitin7 gene (GhUBQ7, GenBank accession no. DQ116441) was used as an internal control. Data represent mean ± SEM of three independent assays. Different lowercase letters represent statistically significant differences as determined by one-way ANOVA (P < 0.05, Duncan's multiple range tests). Scale bar, 3 mm. https://doi.org/10.1371/journal.pone.0215771.g003 Identification of cotton MOTHER OF FT AND TFL1 homologs, GhMFT1 and GhMFT2

GhMFT1 and GhMFT2 influences the expression of ABA and GA signaling pathway genes in transgenic Arabidopsis
To explore the possible mechanisms of GhMFT1 and GhMFT2 in repressing seed germination, we further analyzed the expression levels of germination-related genes, including ABA and GA signaling pathway genes among wild type, 35S:GhMFT1 and 35S:GhMFT2 transgenic plants. AtABI3 and AtABI5 showed higher expression levels in all the 35S:GhMFT1 and 35S: GhMFT2 transgenic plants than those of wild-type seeds (Fig 5), suggesting that GhMFT1 and GhMFT2 may be involved in the ABA accumulation in transgenic Arabidopsis seeds and control seed germination by upregulating AtABI3 and AtABI5 expression.
GA3OX1 and GA3OX2 are the key rate-limiting enzymes in GA synthesis pathway in Arabidopsis, whereas RGA and RGL2 belong to DELLA family members involving in repressing seed germination [14,22,25]. qRT-PCR results showed there were slightly differences in the expression of AtGA3OX1 and AtGA3OX2 among all detected plants (Fig 5), implying that GhMFT1 and GhMFT2 may affect GA synthesis. However, the expression levels of AtRGA and AtRGL2 were significantly upregulated in all transgenic plants, suggesting that more DELLA proteins may be accumulated in all overexpressing GhMFT1 and GhMFT2 lines, resulting in inhibiting seed germination.

Both GhMFT1 and GhMFT2 interact with GhFD protein
In the SAM of Arabidopsis and rice, FT interacts with florigen receptor 14-3-3 proteins and a bZIP transcription factor FD to induce transcription of floral meristem identity genes [3,78,79]. Recent studies reveal that cotton FT homolog GhFT and TFL1/SP homolog GhSP interact with a cotton bZIP transcription factor GhFD [64,68,69,71]. To analyze the subcellular mechanisms of cotton MFT homologs and their interaction with FD, we generated constructs containing GhMFT1 and GhMFT2 C-terminal fusions with GFP under control of the CaMV 35S promoter, which were transiently expressed in leaf epidermal cells of N. benthamiana, subsequently, the fluorescence was observed by CLSM510. As is shown in S8 Fig, green fiuorescence of the GhMFT1-GFP and GhMFT2-GFP fusion proteins was found in the peripheral cytoplasm (surrounding the vacuole) as well as in the nucleus, which was similar to that observed in cells expressing GFP alone, showing that the subcellular localization of GhMFT1 and GhMFT2 appears to be similar to that of GhFT1 [66,67] and GhSP [68]. Yeast two-hybrid analysis confirmed that both GhMFT1 and GhMFT2 interacted with GhFD ( Fig 6A). Furthermore, we observed strong fluorescence in the nucleus of Arabidopsis protoplasts that co-expressed GhFD with GhMFT1 or GhMFT2 by BiFC (Fig 6B), which further confirmed these results of proteins interaction. Sequencing of cotton genome has expanded our exploration of the PEBP family. Two MFT homoeologous gene pairs, GhMFT1 and GhMFT2, were identified according to cotton genome sequences dataset in the study [58,60]. In Arabidopsis, loss-of-function mutant in MFT does not show defects in flowering time, and its overexpression slightly accelerates flowering, suggesting that it functions as a weak inducer [37]. Similarly, in the present study, we found that  [43], Glycine max [44], Actinidia chinensis [45], and Citrus latifolia [46]. A comparison of protein sequences has revealed that the key amino acid residue W in cotton MFT homologs replaces Y in FT or H in TFL1 (Fig 1A), respectively, wherein this residue lies at the entrance to a ligand-pocket and plays a crucial role in determining protein specificity as flowering inducer or repressor [32,33], suggesting that two cotton MFT homologs may not be involved in regulating floral transition. Furthermore, GhMFT1 and GhMFT2 have a conserved P residue near the C terminus (Fig 1A), which is absent in FT-like or TFL1-like proteins [26]. Phylogenetic analysis revealed that GhMFT1 and GhMFT2 belong to the MFT-like subfamily (Fig 1B). These results suggest that MFT homologs

Two GhMFT homoeologous genes are differentially expressed during cotton ovule development and seed germination
Like AtMFT expression, most of the identified MFT-like genes in various plant species display a high expression level in seeds, suggesting a highly conserved function for MFT-like genes is involved in seed development, seed dormancy and seed germination across the plant kingdom, such as TaMFT [47], JcMFT1 and JcMFT2 [80,81], and GmMFT [44]. In this study, we discovered that the expression levels of GhMFT1 homoeologous genes in ovules were higher than those in other tissues, and increased significantly at the later stage of ovule development (Fig  2A). Compared with GhMFT2-A, GhMFT2-D is highly expressed during ovule development (Fig 2B), suggesting that GhMFT2-D played a major role in ovule development. However, the expression levels of GhMFT1 homoeologs were higher than those of GhMFT2 homoeologous genes during ovule development (Fig 2) and their expression patterns were very similar to that of MFT in Arabidopsis and soybean [5,44]. These results imply that two cotton MFT homologs may be involved in the regulation of ovule development redundantly or differently.
During seed germination, the expression levels of GhMFT1 and GhMFT2 decreased significantly, which were similar to that of GmMFT [44]. The expression levels of GhMFT1 homoeologous genes were gradually decreased at the early stage of seed germination (Fig 2C), whereas GhMFT2 homoeologous genes were sharply reduced (Fig 2D). We found that GhMFT1-A expression was higher than GhMFT1-D during seed germination and GhMFT2-A was also higher than GhMFT2-D at the early stage of seed germination, suggesting that A subgenome of GhMFT plays more important roles in the regulation of seed germination. To sum up, the increase of expression during ovule development and the decrease of expression during seed germination suggest that GhMFT homologs may be involved in the regulation of ovule development and seed germination. The differential expression patterns of GhMFT homologous genes support this view that the homoeologous gene pairs from A subgenome and D subgenome complementarily contribute to Upland cotton agronomic traits [60].

Expression of cotton MFT homologs is mediated by ABA and GA
Seed germination is regulated by two major antagonistic plant hormones, ABA and GA [9]. Expression of AtMFT is upregulated throughout seed germination when seeds are treated with exogenous ABA [5]. There were several ABREs in upstream of the start codon of AtMFT, which regulated expression of AtMFT (S3 Fig). We found that there were also several ABREs in upstream of the promoters of GhMFT homoeologous genes, imply that ABA might regulate their expressions, just like AtMFT in Arabidopsis. In this study, the time-point expression patterns of GhMFT1 and GhMFT2 homoeologous genes during seed imbibition demonstrated that they were changed in response to ABA and GA (Fig 3). We found that the expressions of MFT homoeologous gene pairs were remarkably enhanced in response to ABA, which were similar to those of GmMFT and Dimocarpus longan MFT (DlMFT) treated by 10 μM ABA [44,53]. GhMFT1-A expression was markedly downregulated with exogenously applied GA, which is similar to AtMFT treated by 10 μM GA [5], implying that cotton MFT homologs play roles in seed germination through mediating the interaction between ABA and GA signals.

GhMFT homologs redundantly and differentially regulate seed germination
In Arabidopsis, mft-2 mutant does not exhibit observable defects compared with wild type under normal conditions [37]. Seeds of mft-2 are hypersensitive to exogenous ABA and exhibit lower germination rate than that of wild type [5]. The germination rate of mft-2 is not significantly different from that of wild type in normal conditions, but its germination rate is much lower than that of wild type in the presence of exogenous ABA indicating that seeds of mft-2 are hypersensitive to exogenous ABA and exhibit lower germination rate than that of wild type [5]. However, seeds of mft-2 are more insensitive to GA than that of wild type and the seed germination rate shows little difference in response to exogenous GA compared with that of wild type due to the regulation of intrinsic ABI3, ABI5 and DELLA proteins in seed germination [5]. Ectopic overexpression of soybean GmMFT in Arabidopsis does not affect flowering time, but inhibits the seed germination at the early stage [44]. Furthermore, TaMFT repressed seed germination [47]. Under far-red light, AtMFT represses seed germination by modulating ABA and GA responses in A. thaliana [50]. In this study, we found that ectopic overexpression of GhMFT homologs in A. thaliana also significantly inhibited seed germination by generating 35S:GhMFT1 and 35S:GhMFT2 transgenic lines, respectively (Fig 4 and S7 Fig). The germination rates of the 35S:GhMFT1/2 mft-2 transgenic lines could not reach the same level as that of mft-2 in normal conditions when exogenous GA was applied (Panels C and D in S7 Fig), indicating that the phenotypes of mft-2 on seed germination are not fully complemented by GhMFT1/2 overexpression. Moreover, overexpression of cotton MFT homologs could not rescue the ABA-sensitive phenotype of mft-2, and the germination rates of the transgenic lines in mft-2 background were much lower than mft-2 with ABA (Panels E and F in S7 Fig), implying that GhMFT homologs aggravated the sensitivity of mft-2 to ABA. To sum up, we guess that GhMFT1 and GhMFT2 may act as negative regulators of seed germination as AtMFT in farred light [50], GmMFT [44] and TaMFT [47] do.
In Arabidopsis germinating seeds, AtABI5 upregulates MFT expression but AtABI3 suppresses MFT expression, whereas DELLA proteins RGA and RGL2 also upregulate MFT expression [5]. Under far-red light conditions, the accumulation of the transcription factor PHYOCHROME-INTERACTING-FACTOR1 (PIF1) promoted the expression of genes encoding ABI5 and DELLA growth-repressor proteins. MFT gene expression was promoted by farred light through the PIF1/SOMNUS (SOM)/ABI5/DELLA pathway, and then MFT repressed seed germination [50]. In this study, we found that the expression levels of AtABI3 and AtABI5 were notably elevated in GhMFT1 and GhMFT2 transgenic lines (Fig 5). Moreover, the expression levels of AtRGA and AtRGL2 were also found to be remarkably elevated in all transgenic plants. Based on these results, we speculate that GhMFT1 and GhMFT2 may inhibit seed germination by increasing the concentration of ABA in germinating seeds. However, the GA synthetic pathway genes, AtGA3OX1 and AtGA3OX2 showed differential expression patterns in different transgenic lines of 35S:GhMFT1 and 35S:GhMFT2, suggesting that GhMFT1 and GhMFT2 regulate seed germination differentially in GA pathway. This hypothesis needs further confirmation.
Recent studies showed that cotton FT and TFL1/SP-like proteins compete for interaction with GhFD, involving in the regulation of plant architecture and flower morphology [64,68,69,71]. As Prewitt et al. [71] reported that cotton GhMFT1 can interact with G. raimondii FD (GrFD) in yeast cells, we also found that both GhMFT1 and GhMFT2 can interact with GhFD not only in yeast cells but also in Arabidopsis protoplast cells (Fig 6). Since GhMFT1 and GhMFT2 are located in the cytoplasm and nucleus (S8 Fig), we hypothesize that two MFT-FD proteins may act as transcription co-regulators that regulate the expression of genes related ABA pathway, but the roles for these complexes in ovule development and seed germination of cotton needs further study.

Conclusions
Genome-wide analysis identified two MFT homoeologous gene pairs in G. hirsutum, namely, GhMFT1-A/D and GhMFT2-A/D. Expression of GhMFT1 and GhMFT2 homoeologs remarkably increased during ovule development but quickly decreased during seed germination. Their expression levels were significantly affected by ABA. Ectopic overexpression of GhMFT1 and GhMFT2 in Arabidopsis repressed seed germination at the early stage. Moreover, the expression levels of AtABI3, AtABI5, AtRGA and AtRGL2 were obviously upregulated in transgenic Arabidopsis seeds. Both GhMFT1 and GhMFT2 interact with a bZIP transcription factor GhFD. Taken together, we speculate that GhMFT1 and GhMFT2 may act redundantly in the regulation of seed germination.