All plants were grown in the greenhouse of the Institute of Tropical and Subtropical Cash Crops, Yunnan Academy of Agriculture Sciences. We used 2-year-old lemon trees (Femminello) propagated by bud grafting to trifoliate orange rootstocks, 80 and 15 healthy trees were selected for water deficit-treated and control, respectively. The trees ranged in height from 1 to 1.5 m growing in 52-cm plastic pots containing potting mix of commercial medium and perlite (3:1). Trees in the greenhouse were exposed to natural variations in photoperiod throughout the experiment during Summer (from June to July) 2015. Before the water deficit treatment, trees were kept in a greenhouse and watered every 2 days to saturation. Minimum and maximum temperatures were 23 and 31°C, respectively. At the beginning of the water deficit, trees were fully watered and soil moisture was measured 4 h later by Theta probe type ML2x (Delta-T Devices Ltd., Cambridge, England), and the soil water naturally evaporated until reaching the set level (around 15%). Then the soil moisture was maintained by accurately watering according to the daily loss of water (pots were weighed and watered thrice per day). Leaf water potential was measured with a pressure chamber (PMS Instrument Co., Corvallis, OR, United States), thrice per day (10:00 am, 2:00 pm, and 6:00 pm). After 2 weeks, trees were fully watered. In this study, the soil moisture was measured three times per pot and leaf water potential was detected once per plant with the values indicated as the mean ± SE of three plant leaves, all treated plants were investigated. When the plants were fully watered, the soil moisture was about 30%, and the corresponding leaf water potential was near –1 MPa. Under water deficit treatment, soil moisture levels were maintained at around 15% corresponding to approximately 50% of the control, and the corresponding water potentials were around -1.5 MPa. In this study, terminal bud and the following five buds from flushes were collected at three stages (stage 1: 1 week before water deficit; stage 2: 1 week after the beginning of water deficit; and stage 3: 1 week after releasing from water deficit). It is worth noting that flower buds were visible to the naked eye at stage 3. To analyze the expression of CiFT, leaf sample was also collected. Only mature healthy fully expanded leaves were randomly collected from the three most apical nodes of shoots. Considering that CiFT might be affected by light, all leaves were sampled at 10:00 am. In this study, bud and leaf samples were collected from three groups of trees used as three biological repeats, each group containing three trees. All plant tissues were sampled, immediately frozen in liquid nitrogen, and stored at –80°C until used.

For morphological observation, 15 water deficit-treated and 15 untreated trees were selected and tagged, respectively. For flowering evaluation, each shoot as a statistical unit (Figure 1A). All shoots of each tree were counted, no flowering shoots as vegetative shoots. The data were processed using one-way analysis of variance (ANOVA), and significant differences were compared based on Student’s t-test. P < 0.01 was considered significant. Paraffin section analysis was performed as described by Yao et al. (2007), approximately 500 buds were selected and tagged under a similar growing condition. Ten shoots from lemon derived from these buds were collected every 4 days before the water deficit treatment and every 2 days beginning of water deficit.

Trees were randomly divided into five groups with 15 plants per group and subjected to treatments in a growth chamber with adjustable temperature (Minimum and maximum temperatures were 23 and 31°C, respectively). These trees were treated by spraying 40 mg L^-1 of GA₃, 100 mg L^-1 of ABA, 40 mg L^-1 of NAA, and 800 mg L^-1 of paclobutrazol (PBZ, a GA biosynthesis inhibitor) onto entire trees with a hand-gun sprayer, using approximately 0.5 L per tree and wetting the tree to the point of run-off. Meanwhile, a non-ionic wetting agent (Tween 20, 20% w/v) was added at a rate of 0.05% v/v in all treatments. Water-treated trees served as control. Samples were collected at 15-days intervals until flower bud formation of PBZ-treated trees (visible to the naked eye). The CiFT expression was measured in leaves, whereas other genes were measured in buds.

Total RNA was extracted as previously described (Zhang et al., 2009b). The libraries were produced and sequenced using Illumina’s Genome Analyzer (Solexa; Illumina, San Diego, CA, United States). The sequencing and data analysis were carried out essentially as described in previous studies (Mortazavi et al., 2008; Conesa et al., 2016). In this study, three biological replicates were carried out for the RNA Seq analysis. Raw sequence reads were filtered for quality using the FASTX-Toolkit (Blankenberg et al., 2010). All clean reads were mapped to the citrus genome¹ (Xu et al., 2013) using Burrows–Wheeler Aligner (BWA) software (Li and Durbin, 2010). In this study, the NOIseq was used to identify differentially expressed genes, probability ≥ 0.8 and the absolute value of log₂^Ratio ≥ 1 were used as the threshold (Tarazona et al., 2011). Gene annotation was conducted using the Blast2GO program (Conesa et al., 2005). The alternative splicing (AS) events from different stages were identified using TopHat (Trapnell et al., 2009). All the junction sites of the same gene were used to distinguish the type of AS events, including alternative 5′-splice sites (A5SS), alternative 3′-splice sites (A3SS), skipped exons (ES), and retained introns (RI). During the analysis of the AS CiFT, we found that two PCR products were generated by a pair of CiFT primers, implying the existence of CiFT AS in lemon. By comparing the two CiFT cDNA sequences, we discovered that one splicing product contained 88 amino acids of open reading frame (ORF) because of intron 2 and 3 retention compared with a typical CiFT gene. Thus, we designated this form as CiFT2β and the full-size transcript as CiFT2α.

Real-time PCR was performed with SYBR Green I chemistry (Qiagen, Hilden, Germany) as described previously (Zhang et al., 2009b, 2014). All primers are listed in Supplementary Table S1. Data were evaluated by using the LightCycler 480 software version 1.5 (Roche Applied Science, Mannheim, Germany) and normalized to expression of β-actin (Zhang et al., 2014). Three independent biological replications of each sample and four technical replications of each biological replication were used for real-time PCR.

The samples for ABA, IAA, and GA₃ quantification were prepared according to a previously reported protocol (Pan et al., 2010; Wu et al., 2014). The Icon Isotopes of internal standard (d₆-ABA, d₅-IAA, and d₂-GA₃) were used for ABA, IAA, and GA₃, respectively. An Agilent 1100 HPLC (Agilent Technologies, Santa Clara, CA, United States), C18 column (150 mm × 2.1 mm, 5 μm; Waters, Milford, MA, United States), and API3000 MS-MRM (Applied Biosystems, Foster City, CA, United States) were used for the analysis (Pan et al., 2010; Wu et al., 2014). Three biological replications were assayed for each sample in this study.

To generate the over-expression vectors of CiFTα, CiFTβ, CiBFT, CiTFL1, and CiMFT, each full-length cDNA sequence was cloned into the binary vector pBI121 with the ClonExpress One Step Cloning Kit (Vazyme Biotech Co., Ltd., Nanjing, China) according to the manufacturer’s instructions, respectively. Meanwhile, the promoter of CiFT, CiBFT, CiTFL1, and CiMFT was also cloned into the DX218 vector, respectively. The floral dipping transformation method was used in this experiment (Clough and Bent, 1998). The surviving plants were transplanted into soil and grown under long-day conditions (16 h light/8 h dark) at 22 ± 1°C. Morphological analyses were investigated in the transgenic plants T₃. The number of rosette leaves and days to flowering were counted when transgenic plants bore a 1-cm-long inflorescence (Li et al., 2010; Sun et al., 2014).

Seeds from transgenic plants T₃ were vernalized in the dark at 4°C for 1 week before sowing. Approximately 200 vernalized seeds were sown on a plastic plate (9-cm), then the plates with seeds were transferred into a Percival growth chamber (Percival Scientific, Percival AR41L2), which was set at standard conditions (22°C, 60% humidity, 16 h light/8 h dark, ∼300 μmol m^-2 s^-1 photon flux) until two-leaf stage. When the seedlings reach two-leaf stage, extra seedlings were removed, leaving approximately 100 plants per plate under similar growth status. For the drought treatments, the relative air humidity in the chamber was adjusted to 30% to accelerate the water evaporation at the beginning of the water deficit. Meanwhile, treated plants stop watering and soil moisture was measured by Theta probe type ML2x. Soil moisture levels were maintained at around 15% corresponding to approximately 50% of the control. After 3 days, whole plant material was collected. GUS staining was performed as previously described (Zhang et al., 2014). In this study, three independent transgenic lines for each promoter were analyzed.

Lemon trees produced more flowers than control trees under water deficit condition (Figure 1A). Under control condition, very few flowering shoots (10% of total shoots) were produced in irrigated trees, whereas 72% of shoots flowered when trees were exposed to water deficit (Figure 1B). Most of the flowers formed in trees were of the axillary flower under water deficit condition (flower from axillary buds, Figures 1C,D). These results indicated that the water deficit treatment was effective method at inducing lemon flowering.

To examine cytological changes of buds during floral inductive water deficits, buds were collected, fixed, and stained with hematoxylin for microscopic examination (Figures 1E–L). The paraffin sections analysis of buds showed that lemon do not flower bud differentiation, begin to produce vegetative buds before water deficit (Figure 1E). Under water deficit conditions, the floral buds rapidly initiated differentiation (about 3–4 days after the beginning of water deficit, Figures 1F–H). Floral development hastened differentiation and produced the primordia of the floral organ including sepal, petal, stamen, and pistil (Figures 1H–L). The whole integrated flower bud was formed in 2 or 3 weeks, and then part of the flower bud began to flowering.

To analyze dynamic changes in the lemon bud transcriptome during floral inductive water deficits, RNA sequencing was performed on lemon buds at three stages (stage 1: 1 week before water deficit; stage 2: 1 week after the beginning of water deficit; and stage 3: 1 week after releasing from water deficit). After removing low-quality reads, approximately 28 million clean reads were obtained for each biological replicate. From stage 1 to stage 2, of the 21,947 read-mapped genes detected, more genes were up-regulated (11,490) than down-regulated (10,457). From stage 2 to stage 3, of the 21,763 read-mapped genes detected, more genes were down-regulated (11,956) than up-regulated (9807). Similarly, of the 21,947 read-mapped genes detected from stage 1 to stage 3, more genes were down-regulated (11,055) than up-regulated (10,892). A total of 22,354 non-redundant genes were expressed during floral inductive water deficits when the data from the three stages were combined. Of these, 406 were not observed from stage 1 to stage 2, 326 were not found from stage 2 to stage 3, and 591 were not found from stage 1 to stage 3.

.

To explore potential AS events during floral inductive water deficits, gene structure analyses of AS genes were performed. When combining cDNA of the sweet orange genome (Xu et al., 2013) and RNA sequencing data, more than 27% (12,041) of the multi-exon genes contained at least one AS event. In this study, we detected 45,752 splice junctions in 10,689 genes at stage 1, 43,365 splice junctions in 10,193 genes at stage 2, and 42,734 splice junctions in 10,125 genes at stage 3 (Figure 2A). We also identified the following four common types of AS: A3SS, A5SS, ES, and RI. At stage 1, classification of the 45,752 splice junctions showed that 807 (1.8%) splice junctions corresponded to ES, 2488 (5.4%) corresponded to A5S, 4563 (10.0%) corresponded to A3S, and 38,894 (85.0%) corresponded to IR. At stage 2, of the 43,365 splice junctions 610 (1.4%) splice junctions corresponded to ES, 2213 (5.1%) corresponded to A5S, 4043 (9.3%) corresponded to A3S, and 36,503 (84.2%) corresponded to IR. At stage 3, of the 42,734 splice junctions 685 (1.6%) splice junctions corresponded to ES, 2038 (4.8%) corresponded to A5S, 3743 (8.6%) corresponded to A3S, and 36,268 (84.9%) corresponded to IR (Figure 2B). Intron retention was the most prevalent mechanism. There might be two possible explanations for the high percentage of IR transcripts: at first, some novel transcriptional events may have occurred during under water deficit condition. Secondly, the ORF from reference transcripts may be inaccurate because their ORFs were predicted based on triplet code (Xu et al., 2013).

We noted many AS events of flowering-related genes, such as RELATIVE OF EARLY FLOWERING 6 (REF6), FLC, FCA, EMBRYONIC FLOWER 1 (EMF1), FT, and FY. Almost all reported AS events of citrus flowering time genes were found, including FLC, FCA, and FY (Zhang et al., 2009a; Ai et al., 2012, 2016). Here, we report the splicing of lemon FT (CiFT) as an example. CiFT comprises four exons and three introns according to DNA analysis of lemon; two AS forms were detected based on RNA sequencing. Further expression analysis of CiFT was conducted by reverse transcriptase PCR (Figure 2C). Two AS transcripts of CiFT were isolated from lemon. One of the AS transcripts showed high identity with trifoliate orange and Citrus unshiu FT and was named CiFTα, and only some SNPs (single nucleotide polymorphism) were found compared with the protein-coding sequences of the previously published trifoliate orange and sweet orange FT (Kobayashi et al., 1999; Endo et al., 2005; Zhang et al., 2009b). The other AS transcript contains four exons and two introns (because of intron 2 and 3 retention) and was named CiFTβ. The isolated cDNA of CiFTβ is 2429 bp long, and 267 nucleotides of an ORF contained 88 amino acids based on bioinformatics prediction (Figure 2D) because the second intron has a termination codon (Figure 2C).

To identify flowering-related genes, the number of normalized gene reads of different stages was calculated using FPKM. Genes were considered to be differentially expressed based on probability ≥0.8 and an absolute value of log₂^Ratio≥1 as a threshold (Tarazona et al., 2011). According to these criteria, 944 genes were DEGs from stage 1 to stage 2, and 455 were up-regulated and 489 were down-regulated. From stage 2 to stage 3, 922 genes were DEGs, and 596 were up-regulated and 326 were down-regulated. From stage 1 to stage 3, 686 DEGs were identified, with 395 up-regulated and 291 down-regulated (Figure 3A). By combining results from the three stages, 1638 DEGs were identified as candidates that may represent the common flowering-related genes, and 59 DEGs were shared in all three stages that may represent the common flowering-responsive genes (Figure 3B). And then, a homology search was conducted using the NCBI database to investigate the biological processes possibly regulated by the 1638 DEGs (Supplementary Table S2). We detected 1519 DEGs as having homology with known proteins and the remaining 19 did not in the NCBI database. Based on Gene Ontology annotation of these genes performed by Blast2GO analysis, 1436 DEGs were divided into the three principal GO organization categories (Supplementary Figure S1): molecular function, biological process, and cellular components.

Based on the similarity of the expression patterns of the DEGs, 1638 DEGs were classified into four clusters. The gene from cluster 1 (including 172 genes) was induced at stage 1 and most maintained high expression levels at stage 3 compared with stage 1 and stage 2 (Figure 3C). This cluster featured genes encoding transcription factors (TFs), biotic/abiotic response proteins, and heat shock proteins. These genes were significantly induced at the beginning of water deficit, indicating that the gene cluster might play a key role in the necessary development of lemon. The gene from cluster 2 (including 181 genes) was suppressed at stage 1 and maintained low expression levels at stage 3 compared with stage 1 and stage 2 (Figure 3D). These genes were involved in transcriptional regulation, protein metabolism, and ABA signaling, according to BLAST analysis. These genes may involved in meristem gene regulation and development of vegetative buds. Meanwhile, hormone-related genes and stress proteins were featured in this cluster, indicating these genes might be related to the water deficit response. The gene from cluster 3 (including 776 genes) was transiently suppressed at stage 2 and was then induced at stage 3 (Figure 3E). This cluster also featured genes encoding hormone (GAs, ABA, auxin, and ethylene) signaling/biogenesis and flowering control proteins. These genes shows up-regulated at stage 3, indicating these genes involved in flowering and recovery of vegetative growth. The gene from cluster 4 (including 509 genes) that was transiently induced at stage 2 and was then suppressed at stage 3 (Figure 3F). The suppression of the gene cluster may imply possible involvement in drought stress response, floral induction, and flower bud differentiation of lemon.

Many previously reported flowering-related genes were found among DEGs such as GI, FLC EARLY FLOWERING 3 (ELF3), and EARLY FLOWERING 4 (ELF4) (Supplementary Table S2). The expression pattern of these genes was closely correlated with flowering during floral inductive water deficits. In Arabidopsis, drought and ABA promote transcriptional up-regulation of PEBP family members such as FT and TSF leading to flowering under long days (Riboni et al., 2013). In this study, the two genes showed high identity with MOTHER OF FT (MFT) and BFT from the PEBP gene family. Moreover, some additional related flowering genes that have not been placed in any specific flowering pathway were also identified in this study, such as the SQUAMOSA PROMOTER BINDING PROTEIN family genes (SPL6/8/9/13) and MADS TF family genes. Many genes involved in different hormone synthesis and signaling pathways showed significant expression changes during floral inductive water deficits (Supplementary Table S2). For example, four ABA-related genes (two ABA 8-hydroxylase genes and two ABA stress-related proteins) showed significant up-regulation under water deficit condition, which encode key enzymes in ABA biosynthesis and metabolism. Eleven auxin-related genes were significant altered included four auxin-induced proteins, three auxin response factors, three auxin transporters, and one auxin canalization protein. The auxin response factors were down-regulated and auxin-induced proteins were up-regulated at the beginning of water deficit. After the water deficit treatment, cluster 3 genes for ethylene biosynthesis and perception were up-regulated, including 15 ethylene-responsive TFs and 1 ethylene response protein. In addition, six genes involved in the response to GA stimulus pathways were up-regulated at the beginning of water deficit. Furthermore, the biological interpretation of the DEGs was further investigated by KEGG pathway analysis. The most frequently represented pathways are plant hormone signal transduction and metabolic pathways at different stages (Supplementary Figure S2). These results suggest that plant hormone are involved in the regulation of flowering under water deficit conditions.

To validate the expression profiles obtained by RNA sequencing, a total of 26 genes (The expression pattern of 14 genes belong to cluster 3 and 12 genes belong to cluster 4) were chosen to real-time PCR analysis; these genes included 24 DEGs, four genes of no differential expression. The results from real-time PCR were compared with RNA sequencing data. For 22 of the 26 genes, the same expression patterns was shown between the RNA sequencing data and real-time PCR analysis (Figures 4A–C), the remaining four genes showed different expression patterns according to RNA sequencing and real-time PCR.

In this study, the expression of CiTFL1 was not detected during floral inductive water deficits. One possible explanation is that its mRNA levels were too low to be detected by RNA sequencing. These results consistent with our previous reports on TFL1 in citrus (Zhang J.Z. et al., 2011). Therefore, the expression of lemon CiTFL1 was investigated in lemon bud by real-time PCR. The expression of CiTFL1 was transiently suppressed under water deficit treatment and was then induced at the resumption of irrigation. On the other hand, citrus FT was mainly expressed in leaves, with little or no expression in citrus buds based on our previous study (Zhang et al., 2009b). Therefore, the expression of lemon FT (CiFTα; CiFTβ; tCiFT: Total CiFT including CiFTα and CiFTβ) was investigated in lemon leaves, and the results showed that the accumulation of tCiFT significantly increased under water deficit treatment (Figure 4D). However, when water deficit was interrupted, the expression levels returned to the initial levels. The expression patterns of CiFTα and CiFTβ were similar to that of tCiFT. These results indicated that the expression of CiFT was induced by water deficit.

The RNA sequencing data revealed that hormone-related genes changed greatly during the water deficit process, especially genes for ABA biosynthesis and metabolism, followed by the genes for GA and IAA biosynthesis and signal transduction. Therefore, the ABA, IAA, and GA₃ contents of buds were measured in water deficit-treated and control trees. IAA increased immediately at the beginning of the water deficit treatment and maintained high expression levels until the resumption of irrigation (Figure 5A). ABA content was transiently induced during floral inductive water deficits and was then suppressed at resumption of irrigation (Figure 5B). Unfortunately, we were unable to detect GA₃ successfully. Therefore, trees were sprayed with PBZ, a GA biosynthesis inhibitor, and untreated trees served as the control. The number of flowers in the PBZ-treated trees was increased compared to the control (Supplementary Figure S3). PBZ yielded a similar trend with water deficit treatment. These results indicate that the GAs content may be suppressed during water deficit.

To investigate the effect of exogenous hormones on citrus flowering and the expression of flowering genes, GA₃, ABA, and NAA were also sprayed onto entire trees. Some new vegetative shoots were formed by GA₃ and NAA treatment compared with the control. However, no significant differences were observed by ABA treatment. Samples were collected at 15-days intervals until flower buds of PBZ-treated trees were visible to the naked eye. In this study, plant materials from PBZ-treated, GA₃-treated, ABA-treated, NAA-treated, and water-treated trees were collected at 4 stages (stages 1, stage 2, stage 3, and stage 4). It appears that stage 1 was the bud induction period and stage 2 and 3 were flower bud differentiation period because part of the flower bud from PBZ-treated trees were visible to the naked eye at stage 4 compared with the control trees. The expression of related flowering DEGs (tCiFT, CiBFT, CiTFL1, CiELF, and CiSPLs) was investigated at different treated stages. Compared with the control trees, the expression of tCiFT in leaves was significantly increased by ABA and PBZ treatment (Figure 5C). The expression level of tCiFT in GA₃- and NAA-treated trees was significantly suppressed from stage 3 to stage 4. Under PBZ treatment condition, the expression patterns of CiFTα and CiFTβ were similar to that of tCiFT, indicating that they may perform a similar function role during the treatment process. For different hormone treatments, the relative expression of CiGI, CiTFL1, and CiBFT showed significantly higher values throughout the entire treated period (Figure 5D). They increased at the beginning of the treatment, then maintaining a high level of expression during treatment, and tended to decrease at the end of the treatment. The levels of CiBFT1 and CiTFL1 accumulation in buds were induced during the entire treatment period by ABA. A high transcript level of CiMFT was seen as the beginning of treatment, and it was then maintained at a low level except when treated with NAA. No differences in the expression of CiELF3 and CiELF4 were detected between the control and the treated trees; they showed high expression levels throughout the entire period studied. The relative expression levels of CiSPL6, CiSPL8, and CiSPL9 were not significantly altered. They were detected and present at low levels during the treatment process, and their expression levels were independent of treatments throughout the study period. Relative expression of CiSPL13 was different from other CiSPL genes at stage 4; the accumulation of CiSPL13 was induced at stage 4 by NAA and GA treatments (Figure 5D).

To assess the functional characteristics of CiFTα, CiFTβ, CiBFT, CiTFL1, and CiMFT, these genes were over-expressed in Arabidopsis. Sixteen, 20, 18, 14, and 26 independent T₁ transgenic lines were generated, respectively. Interestingly, CiFTβ transcribed several new transcripts (Figure 6A). Comparison of the various CiFTβ cDNA revealed that four AS transcripts of CiFTβ were identified; Among these AS transcripts, one of the transcripts was the same as CiFTα, other transcripts could not encode complete protein compared with CiFTα because of intron retention or A3SS (Supplementary Figure S4).

To further analyze the function of these genes, three transgenic lines were randomly selected for each gene. We selected 15 T₃ plants for each transgenic line. Compared with control plants, the 35S::CiFTα and 35S::CiFTβ transgenic lines flowered significantly earlier than control plants in terms of both number of leaves and days to flowering (Figure 6B). In 35S::CiFTα and 35S::CiFTβ, the average time to flowering ranged from 23.7 to 26.1 days in six transgenic lines, whereas that of the control plants was 33.5 days. The average number of leaves at flowering ranged from 7.1 to 9.1 and was 12.1 control plants (Figure 6I). The 35S::CiFTβ flowered earlier than 35S::CiFTα. Transgenic plants from 35S::CiTFL1 and 35S::CiBFT showed late flowering (Figures 6C,D). The average time to flowering of the transgenic plants ranged from 41.1 to 52.1 days, while that of the control plants was 34.1 days. The average number of leaves at flowering ranged from 15.8 to 20.1 in the transgenic plants and was 12.1 in the control plants (Figure 6J). Two transgenic lines from 35S::CiTFL1 plants flowered significantly later than the control plants (80- to 120-day delay in flowering, Figure 6E). It is worth noting that some lines showed flowering defects including alterations in floral organ number, pale green sepals, vestigial petals, and persistent sepals around the fruit (Figures 6F–H). However, no difference in the appearance of flowering time, flower and inflorescences was observed between 35S::CiBFT and control plants. Twenty-six plants were obtained in the T₁ generation. However, CiMFT did not affect flowering and flower inflorescences in transgenic Arabidopsis (data not shown).

To further study the expression of CiFT, CiTFL1, and CBFT, the promoters (about 1.5–2.0 kb) of the three genes were amplified from lemon by using the genome walking method. We confirmed that these sequences were the promoters of the three genes by comparing them with the sweet orange genome (Xu et al., 2013). The 5′ upstream region of the above three genes were analyzed by using PLACE software. The results showed that the common elements were found in these promoters such as the putative transcriptional start site, TATA box, CAAT box, and different binding motifs (hormone response elements, light regulation, and drought response elements) (Supplementary Table S3). To examine the spatial patterns of CiFT, CiTFL1, and CBFT, we generated transgenic Arabidopsis with the GUS reporter gene driven by these putative promoters. Consequently, we obtained >30 independent transgenic lines for each promoter.

Histochemical staining indicated that the expression of CiFT promoter was seen first in the vascular tissues of cotyledons. With the development of plants, GUS staining was present throughout the whole plant except roots. In inflorescences, CiFT promoter was also detected in the vascular tissues of pedicels and floral organs but not in the inflorescence meristem (Figures 7A–D). The GUS signal from CiTFL1 promoter can be observed at different stages. In juvenile transgenic plants, GUS staining was present throughout the whole plant. Further analysis of GUS activity in various organs of transgenic plants revealed that GUS staining was observed in rosette leaf, cauline leaf, flowers, silique pods and roots (Figures 7E–H). GUS expression gradually decreased at the adult stage. The CiBFT promoters were found only in the cell elongation zone and around the vascular bundles of the roots except root tip in the entire life cycle of transgenic plants. The same pattern is observed in lateral roots (Figures 7I–L). Closer analysis revealed low GUS activity in the leaves. In addition, the expression of CiFT, CiTFL1, and CiBFT promoter response to water deficit treatments was also investigated; the expression of CiBFT and CiTFL1 promoters was decreased and that of CiFT promoter was increased under water deficit conditions (Figures 7M–O).