Manipulation of flowering time and branching by overexpression of the tomato transcription factor SlZFP2

Summary Flowering of higher plants is orchestrated by complex regulatory networks through integration of various environmental signals such as photoperiod, temperature, light quality and developmental cues. In Arabidopsis, transcription of the flowering integrator gene FLOWERING LOCUS T (FT) that several flowering pathways converge to is directly regulated by more than ten transcription factors. However, very little is known about the transcriptional regulation of the FT homolog SINGLE FLOWER TRUESS (SFT) in the day‐neutral plant tomato (Solanum lycopersicum). Previously, we showed that the zinc finger transcription factor SlZFP2 plays important roles in regulation of seed germination and fruit ripening in tomato and also found that overexpression of SlZFP2 impacted flowering and branching. Here, we characterized in detail the early flowering and high branching phenotypes by overexpression of this transcription factor. Our data showed that overexpression of SlZFP2 accelerated flowering in an SFT‐dependent manner as demonstrated by elevated SFT expression in the leaves and the transcription factor's binding ability to SFT promoter in vitro and in vivo. Furthermore, overexpression of the SlZFP2 gene in the sft plants failed to rescue the mutant's late flowering. Through analysis of grafting phenotype, growth response of branches to auxin application and transcriptome profiling by RNA sequencing, we also showed that overexpression of SlZFP2 affected shoot apical dominance through multiple regulatory pathways. Our results suggest that the transcription factor SlZFP2 has potential applications in genetic modification of plant architecture and flowering time for tomato production and other crops as well.


Introduction
The timing of flowering is crucial for higher plants to complete their life cycles in response to environmental conditions. Manipulation of flowering time has also great potential applications in plant breeding and is being under extensive investigation. In many plant species, flowering time is orchestrated by complex regulatory networks through integration of various environmental signals such as photoperiod, temperature, light quality and developmental cues. Several flowering pathways perceiving different environmental and developmental stimuli converge to few flowering integrators, for example, FLOWERING LOCUS T (FT), SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1 (SOC1) and AGAMOUS-LIKE 24 (AGL24) in Arabidopsis (Michaels, 2009;Navarro et al., 2011;Pin et al., 2010;Tsuji et al., 2010). FT, mainly expressed in leaf, is transported to shoot meristems (SAM) where it interacts with FLOWERING LOCUS D (FD) to induce the transition of SAM to floral meristems (FM) (Wigge et al., 2005). At transcription level, FT is directly regulated by a number of transcription factors in response to different stimuli. For example, FT transcription is directly activated by CONSTANS (CO), CRYPTOCHROME-INTERACTING BASIC-HELIX-LOOP-HELIX1 (CIB1), WRKY71, PHYTOCHROME-INTERACTING FACTOR 4 (PIF4) and Morf-related Gene 2 (MRG2) (Kumar et al., 2012;Liu et al., 2008Liu et al., , 2013Tiwari et al., 2010;Xu et al., 2014;Yu et al., 2016). In addition, FT transcription is also directly repressed by TEMPRANILLO (TEM) 1 and 2, TARGET OF EAT (TOE) 1 and 2, SHORT VEGETATIVE PHASE (SVP), CYCLING DOF FACTOR1 (CDF1), EARLY-FLOWERING MYB PROTEIN (EFM), SCHLAFMUTZE (SMZ) and SCHNARCHZAPFEN (SNZ) (Marin-Gonzalez et al., 2015;Mathieu et al., 2009;Song et al., 2012;Yan et al., 2014;Zhang et al., 2015). Thus, transcriptional regulation of FT expression plays a crucial role in flowering time control in Arabidopsis.
Flowering is also regulated by phytohormones either through FT or other flowering regulators. For example, gibberellins (GAs) promote flowering through increasing FT expression in the vascular tissue under inductive long-day condition (Porri et al., 2012). Abscisic acid (ABA), acting antagonistically with GA during seed germination, is also implied to play a role in regulation of flowering time (Wilmowicz et al., 2008). ABA-deficient and insensitive mutants of Arabidopsis flower earlier under shortday conditions (Martinez-Zapater et al., 1994), indicating ABA has a repressive role in flowering time control. Some evidences show that flowering inhibition by ABA is likely through transcriptional regulation of the flowering repressor FLOWERING LOCUS C (FLC) because two ABA signalling components ABA INSENSITIVE 4 and 5 (ABI4 and ABI5) directly activate FLC transcription (Shu et al., 2016;Wang et al., 2013;). However, under stress conditions, ABA may delay Arabidopsis flowering through GA pathway because in the quadruple-della mutant, the transcription of LEAFY (LFY) was elevated (Achard et al., 2006), likely due to diminished DELLA repression on LFY and SOC1 (Achard et al., 2007). On the other hand, FT seems to regulate stomatal opening and seed dormancy in Arabidopsis through ABA signalling pathway (Chen et al., 2014;Kinoshita et al., 2011). Nevertheless, there is an evident interconnection between ABA signalling and flowering pathways.
SINGLE FLOWER TRUESS (SFT), the homolog of FT, is a major player in flowering time control of the day-neutral plant tomato (Solanum lycopersicum) (Lifschitz and Eshed, 2006;Lifschitz et al., , 2014Molinero-Rosales et al., 2004). Functional conservation has also been revealed for the FD homolog SUPPRESSOR OF SP/SP-Interacting G-BOX (SSP/SPGB) and the flowering repressor TERMINAL FLOWER1 (TFL1) homolog SELF-PRUNING (SP) (Park et al., 2014;Pnueli et al., 1998). SFT and SP have additional functions in regulation of shoot architecture because mutations in the two genes cause either altered shoot growth pattern or reversion of inflorescence into leaves (Lifschitz et al., 2014). When sp mutation is present, sft heterozygosity exerts yield heterosis in an SFT dosage-dependent manner (Jiang et al., 2013;Krieger et al., 2010). This indicates that transcription of SFT is tightly regulated. Despite its crucial roles in regulation of flowering and shoot architecture, how SFT is transcriptionally regulated in tomato is unclear. In addition to the abovementioned flowering genes, mutations in UNIFLORA (UF), Blind (Bl), JOINTLESS (J) and COMPOUND INFLORESCENCE (S) also affect flowering in tomato (Dielen et al., 2001;Quinet et al., 2006aQuinet et al., ,b, 2010. But the molecular mechanisms whereby these genes regulate flowering are unknown. Plant architecture is largely determined by shoot branching. The formation of lateral shoots in tomato requires Lateral suppressor (Ls), Bl and its homologs Bli1 and Bli3 (Busch et al., 2011;Schmitz et al., 2002;Schumacher et al., 1999). After the formation of axillary meristems for developing lateral shoots, the outgrowth of shoot branches is regulated by auxin, cytokinins and strigolactones (Rameau et al., 2015). Auxin synthesized in shoot tips is associated with apical dominance; the apex-derived auxin inhibits the growth of axillary buds below and its depletion from stem after decapitation releases bud dormancy. Because the movement of auxin in the main stem is directional-only moving downwards but not upwards into buds, the action of auxin on shoot branching is thought to be indirect (Muller and Leyser, 2011). Nevertheless, apical dominance is reduced in the auxin perception mutants of Arabidopsis tir1 and cul1 (Moon et al., 2007;Ruegger et al., 1998). As mentioned above, Bl is involved in flowering time control in tomato, its pepper homolog CaBLIND also regulates flowering in addition to shoot branching (Jeifetz et al., 2011). Moreover, genetic analysis of multi-parent recombinant inbred lines (AMPRILs) in Arabidopsis has revealed that flowering time genes FLC, FRIGIDA (FRI) and FT have pleiotropic effects on shoot branching (Huang et al., 2013). Thus, these findings suggest that there are possible interconnections among the genetic networks regulating flowering and shoot branching.
Previously, we showed that overexpression of the zinc finger transcription factor SlZFP2 affected multiple traits including flowering, branching, seed germination and fruit ripening (Weng et al., 2015). We demonstrated that SlZFP2 regulates seed germination through direct transcription repression on ABA biosynthetic genes and it controls ripening by preventing Colorless non-ripening (CNR) expression before the onset of ripening process, but how overexpression of SlZFP2 affected flowering and branching was not addressed. In this study, by phenotypic, gene expression and biochemical analysis, we showed that the earlyflowering phenotype by overexpression of SlZFP2 was resulted from elevated SFT transcription in the leaves, and the increased branching was due to weakened apical dominance. Moreover, gene expression analysis demonstrated that SlZFP2 is required for SFT expression during fruit development. Because SlZFP2 directly binds to SFT promoter in vivo and in vitro, our study provides a promising strategy to manipulate flowering time for improvement of tomato production.

Results
Overexpression of SlZFP2 promotes flowering in an SFTdependent manner We previously showed that the zinc finger protein SlZFP2 negatively regulates ABA biosynthesis during fruit development and ripening (Weng et al., 2015). In addition to its role in regulation of seed germination and ripening, we also found that the transgenic plants overexpressing SlZFP2 fused with HA tag (HA-SlZFP2) or its coding sequence alone displayed (SlZFP2) earlyflowering phenotype (Figure 1a). To investigate the early-flowering phenotype by overexpression of SlZFP2 in more detail, we further quantified the flowering time of these overexpression lines in both genetic backgrounds of Solanum pimpinellifolium LA1589 and the cultivated tomato M82, respectively. The leaf number formed before the first inflorescence is predictable and consistent in a given growth condition, making it a good indicator for flowering time. In our growth conditions, the wild-type (nontransgenic) plants of LA1589 and M82 form 11-12 and 7-8 leaves before the first inflorescence, respectively (Figure 1b-d). In contrast, three to four transgenic lines from LA1589 and M82 overexpressing either HA-SlZFP2 or SlZFP2 had significantly fewer leaves formed before the first inflorescence; several transgenic lines from LA1589 and M82 produced only six or seven leaves before their first inflorescences formed ( Figure 1). This indicates that overexpression of either HA-SlZFP2 or SlZFP2 can effectively shorten flowering time in both of LA1589 and M82.
Flowering time in tomato is mainly governed by SFT Shalit et al., 2009). It is plausible that overexpression of SlZFP2 promotes flowering through SFT pathway. To test the possibility, we generated sft plants overexpressing HA-SlZFP2 by crossing between the HA-SlZFP2 overexpression lines and the sft mutant. Overexpression of HA-SlZFP2 in the sft mutant failed to rescue the mutant's late-flowering phenotype to wild type, although it flowered earlier than the mutant did, but the effect was weak; the sft plants overexpressing HA-SlZFP2 formed 14.6 leaves on average before the first inflorescence, compared to 15.4 leaves of the sft mutant ( Figure 2a). This suggests that promotion of flowering by overexpression of SlZFP2 requires a functional SFT pathway. We then performed a quantitative reverse-transcription PCR (qRT-PCR) analysis of SFT expression in the mature leaves of four HA-SlZFP2 overexpression lines from LA1589 and the wild type. All the four overexpression lines had significantly higher SFT expression by more than twofolds ( Figure 2b). Because SFT expression increases with leaf maturation , we found overexpression of HA-SlZFP2 did not alter the temporal expression pattern of the florigen gene SFT; instead, it elevated its expression at all stages ( Figure 2c). The results further support that overexpression of SlZFP2 accelerates flowering through SFT pathway by activating its expression in the leaves.
Previously, we have shown that the transcription factor SlZFP2 recognizes cis-elements containing (A/T)(G/C)TT motifs (Weng et al., 2015). SFT also contains multiple (A/T)(G/C)TT elements within its 1.5-kb promoter region. Using HA antibody, we  confirmed by EMSA using Escherichia coli expressed GST-SlZFP2 on the same regions (Figure 3a,b). We further conducted a transient gene expression analysis to test whether high SlZFP2 expression can enhance SFT transcription in tobacco leaves. Indeed, the expression of YFP-SFT under the control of the 2.0 kb native SFT promoter was activated in Nicotiana. benthamiana leaves transiently overexpressing SlZFP2 under the control of the 35S promoter ( Figure 3c). Moreover, GUS expression driven by the 1.8 kb SFT promoter was increased dramatically in Arabidopsis protoplasts by overexpression of SlZFP2 ( Figure 3d). Thus, SlZFP2 has transcriptional activation activity to enhance SFT expression in vitro and in vivo and it can directly bind to the (A/T) (G/C)TT elements of the SFT promoter.

Overexpression of SlZFP2 increased branching
In addition to early flowering, we also observed high branching or fast bud outgrowth phenotype in these transgenic lines overexpressing HA-SlZFP2 or SlZFP2 during vegetative development (Figures 1a and 4a,b;Weng et al., 2015). We quantified the numbers of branches formed on one-and-half-month-old plants overexpressing HA-SlZFP2 or SlZFP2 in LA1589 and M82 backgrounds. Except for the line L102 from LA1589, branch numbers were significantly increased in the other three HA-SlZFP2 overexpression lines (Figure 4c,d). In addition, three of the four SlZFP2 overexpression lines also had more branches (Figure 4e). Furthermore, the transgenic plants overexpressing HA-SlZFP2 produced side-shoots much earlier than the wild type did on both of the main shoots and branches ( Figure 4a). The results suggest that overexpression of HA-SlZFP2 or SlZFP2 impacts bud outgrowth in both of LA1589 and M82.
The outgrowth of shoot branches is regulated by plant hormones including auxin, cytokinin and strigolactones (Rameau et al., 2015). Strigolactones are mainly synthesized in roots and transported to axillary buds to inhibit their outgrowth. To determine whether the increased branching phenotype observed on these transgenic plants overexpressing either HA-SlZFP2 or SlZFP2 was caused by impaired strigolactone biosynthesis in the roots, we performed reciprocal grafting between the seedlings of the HA-SlZFP2 overexpression line L103 and the wild type. The wild-type rootstocks did not suppress the branching of the scions overexpressing HA-SlZFP2; they still produced more branches than the wild-type scions grafted on the rootstocks of either L103 or the wild type ( Figure 5). Self-grafted shoots of the HA-SlZFP2 overexpression line recovered slowly, but the plant stature looked very similar with its nongrafted plants. The slow recovery of selfgrafted plants overexpressing HA-SlZFP2 was likely due to its less developed root system as shown previously (Weng et al., 2015). Nevertheless, the grafting results suggest that the high branching phenotype by overexpression of HA-SlZFP2 is likely independent of strigolactone biosynthetic pathway.
In addition to root-derived strigolactones, the growing shoot apex may inhibit the activation and outgrowth of axillary buds formed below, and the apical dominance is classically linked to shoot apex-derived auxin (Leyser, 2005;Muller and Leyser, 2011;Teichmann and Muhr, 2015). We then tested the possibility that overexpression of SlZFP2 weakens apical dominance by monitoring the growth response of the first and second branches below the decapitation site to NAA application. We first measured the branch growth response using 15-day-old plants of the HA-SlZFP2 overexpression line L103 and the wild type. As the first branches were completely suppressed, only the length of second branches was measured. Application of 0.5% NAA to the decapitated stumps almost completely suppressed the outgrowth of the second branches of the wild-type plants. In contrast, the branch outgrowth of the HA-SlZFP2 overexpression line was not inhibited by 0.5% NAA (Figure 6a). Next, we tested the auxin sensitivity of branch outgrowth using relatively older plants of 30 days old. Without NAA applied on the decapitated stumps, the first and second branches of the HA-SlZFP2 overexpression line L103 from 30-day-old plants grew much slower than the wild-type branches did (Figure 6b,c). When 0.5% NAA was applied, the growth of the first branch was similarly effectively inhibited in both L103 and the wild type ( Figure 6b). However, NAA application had weaker inhibitory effect on the outgrowth of the second branch of the HA-SlZFP2 overexpression line (Figure 6c). These results imply that the branch outgrowth of the HA-SlZFP2 overexpression lines is less sensitive to auxin. Therefore, overexpression of HA-SlZFP2 likely weakens apical dominance.

Transcriptional regulation by SlZFP2
To further understand how overexpression of SlZFP2 increases branching and accelerates flowering at the transcription level, we conducted an RNA-seq analysis on the shoot apices of two HA-SlZFP2 overexpression lines (L103 and L104) and the wild type (nontransgenic siblings from the two respective overexpression lines). RNA-seq libraries were made of total RNA isolated from the shoot apices with small visible leaves from 45day-old plants. After sequencing, the reads were mapped to the tomato reference genome (version SL2.4) by Tophat (Trapnell et al., 2009). After the uniquely mapped reads were assembled by Cufflinks (Trapnell et al., 2010), differentially expression genes (DEs) between the two HA-SlZFP2 overexpression lines and the wild type were first selected using a cut-off of adjusted P-value at 0.05 and fold change of 1.5. In addition, genes with expression changes larger than twofolds in the two lines and FPKM ≥0.1 were also considered to be differentially expressed. In total, we identified 707 and 569 genes, respectively, up-regulated and down-regulated by overexpression of HA-SlZFP2 (Table S1).
Consistent to the role of SlZFP2 in regulation of ABA pathway (Weng et al., 2015), the GO ontology analysis of DEs revealed that genes involved in stress responses were significantly  (Figure 7a). In addition, function category of floral development was also over-represented. Particularly, SFT transcripts were only detected in the two overexpression lines, further confirmed that overexpression of HA-SlZFP2 induces SFT expression. SlFPF1, the putative homolog of Arabidopsis flowering promoter FPF1, was also only detected in the two overexpression lines, whereas the flowering repressors SP and SELF-PRUNING 2G (SP2G), the putative ortholog of Arabidopsis MOTHER OF FT (MFT), were down-regulated (Figure 7b). In addition, transcription of self-pruning interacting protein 1 (SIP1) was drastically elevated. Likely due to the elevated SFT and SlFPF1 expression, genes involved in floral meristem formation were activated in the two overexpression lines, which then led to transcriptional activation of genes involved in floral organ formation (Figure 7c).
Auxin is mainly synthesized in shoot apex and its downward transport is believed to repress the outgrowth of axillary buds. The hormone is perceived by the SCF TIR1 -SKP1-CUL protein complex (Salehin et al., 2015). Mutation in the auxin receptor gene TIR1 of Arabidopsis leads to weakened apical dominance, and the cul1 mutant is also short and bushy (Moon et al., 2007;Ruegger et al., 1998). Overexpression of HA-SlZFP2 downregulated Solyc01 g067200 and Solyc05 g009260, encoding proteins showing high similarity with CUL1 and TIR1, respectively (Table S1). This suggests that overexpression of HA-SlZFP2 likely attenuated auxin signalling in the shoot apex. Furthermore, Solyc08 g016060, the SPIKE1 (SPK1) homolog, which its mutation has been shown to induce PIN2 internalization in Arabidopsis (Lin et al., 2012), was also repressed in the two overexpression lines. However, the weakened apical dominance of the HA-SlZFP2 overexpression lines may only be explained partially by the downregulated expression of these auxin-related genes because another half of 16 DE genes involved in auxin pathway were up-regulated by overexpression of HA-SlZFP2. The up-regulated genes by overexpression of HA-SlZFP2 include auxin biosynthetic gene Solyc09 g091090 (YUC3 homolog), Solyc01 g111640 (SKP1 homolog) and six auxin response genes. In addition, the accelerated branch outgrowth in these HA-SlZFP2 overexpression lines may also be resulted from down-regulated expression of the Blind like 3 (Bli3) gene, encoding a MYB transcription factor that has been shown to regulate shoot branching (Busch et al., 2011).

Discussion
Shoot branching involving initiation of axillary meristems, bud development and outgrowth is regulated by complex interaction of plant hormones and transcription factors. In tomato, the MYB transcription factor Bl and the GRAS family member Ls regulate axillary meristem formation (Schmitz et al., 2002;Schumacher et al., 1999). Overexpression of HA-SlZFP2 did not affect Bl and Ls expression, but down-regulated the expression level of the Bl homolog Bli3. The down-regulated Bli3 may attribute, at least in part, to the branching phenotype in these SlZFP2 overexpression lines because increasing branching phenotype was observed in its RNAi lines (Busch et al., 2011). However, genes involved in auxin, BR and cytokinin pathways were also impacted in the shoot apices of the HA-SlZFP2 overexpression lines, suggesting that the increased branching phenotype caused by overexpression of SlZFP2 may also be related to impaired hormone signalling and/or crosstalks among these hormones. As the branch outgrowth of the HA-SlZFP2 overexpression lines was less sensitive to NAA application to the decapitated stumps, overexpression of HA-SlZFP2 likely weakened apical dominance. In agreement with the notion, the expression of the two genes encoding homologs of Arabidopsis TIR1 (Solyc05 g009260) and CUL1 (Soly-c01 g067200) were down-regulated in the shoot apices of the HA-SlZFP2 overexpression lines. It has been shown that loss-offunction mutations in the two genes caused weak apical dominance in Arabidopsis. Therefore, it is plausible that the weakened apical dominance was resulted from transcriptional repression of auxin signalling involved in regulation of bud outgrowth. Nevertheless, overexpression of SlZFP2 increases branching likely independent of the strigolactone biosynthetic pathway as demonstrated by reciprocal grafting between the HA-SlZFP2 overexpression lines and the wild type.
SFT and SP are two key flowering regulators in tomato (Lifschitz et al., 2014). Our results demonstrated that overexpression of SlZFP2 accelerates flowering mainly through activation of SFT expression, not by transcriptional repression of SP. Although SP expression was significantly down-regulated in the shoot apices of the two HA-SlZFP2 overexpression lines, there was no obvious defect in sympodial shoot formation observed on any of these SlZFP2 overexpression lines. Instead, the flowering phenotype of the plants overexpressing HA-SlZFP2 or SlZFP2 resembles the SFT overexpression lines; the extreme earlyflowering lines overexpressing HA-SlZFP2 produced similar leaf number before the first inflorescence as the SFT overexpression lines did. EMSA and ChIP assay as well as transient gene expression analysis further confirmed that SlZFP2 activated SFT expression through direct binding to its promoter region containing (A/T)(G/C)TT element in vivo and in vitro. Thus, SFT is likely a direct target of SlZFP2. However, SlZFP2 unlikely plays a major role in flowering time control mediated by SFT because suppressing this transcription factor by RNAi did not delay flowering (Weng et al., 2015). The notion is further supported by the different expression patterns between the two genes; SlZFP2 is expressed in young leaves and shoot apex during vegetative development, whereas SFT is mainly expressed in mature leaves. Although SFT is unlikely a direct target of SlZFP2 in the leaves, this transcription factor regulates SFT expression in the fruit. By qRT-PCR analysis, we found that in the wild-type SFT was not only expressed in leaves but also during fruit development and its expression reached its maximal level at breaker stage when the fruit started to ripen and seed maturation was almost completed ( Figure S1a). Overexpression of HA-SlZFP2, like in leaves, activated SFT expression in mature green fruits, whereas downregulated SlZFP2 expression by RNAi dramatically decreased SFT expression ( Figure S1b,c). Although the role of SFT in fruit development remains to be unravelled, we speculate that during fruit development, the activation of the florigen gene by SlZFP2 might be involved in fruit and/or seed development because we previously demonstrated that SlZFP2 is required for fruit ripening and seed development (Weng et al., 2015). In Arabidopsis, it has been shown that FT can store the temperature memories in the fruit to control progeny's seed dormancy (Chen et al., 2014). It is plausible that SFT has similar function in tomato seed development because high expression of SFT was detected at breaker stage during fruit development. Furthermore, there are increasing evidences indicating that FT or its homologs in other species regulates diverse developmental processes other than flowering including lateral shoot outgrowth and stomatal opening in Arabidopsis as well as the formation of potato tuber and onion bulb (Lee et al., 2013;Navarro et al., 2011).
The transcription factor SlZFP2 is a negative regulator of ABA biosynthesis; overexpression of SlZFP2 decreases ABA production (Weng et al., 2015). When compared with the sft mutant, the sft plants overexpressing HA-SlZFP2 flowered slightly but substantially earlier, indicating that high HA-SlZFP2 expression likely affected other flowering pathway independent of SFT. One explanation for this observation is that the suppressed ABA biosynthesis by overexpression of HA-SlZFP2 promoted early flowering because the tomato ABA-deficient mutant sitiens (sit) and flacca (flc) also flowered slightly earlier; the two mutants formed fewer leaves before the first inflorescence occurred ( Figure S2a). The result is consistent with previous observations that Arabidopsis ABA-deficient and insensitive mutants flower earlier (Martinez-Zapater et al., 1994). Recently, it has been

S l A P 1 S l S E P 2 S l S E P 3 S l S E P 4 S l A P 3 S l P I S l P I -B T A G 1 S l S E P 4 L 1 S l S E P 4 L 2
Floral organ formation Fold change (log2)

Express ion level (FPKM)
Percentage of genes shown that ABA delays flowering in Arabidopsis through transcriptional regulation of the FLC gene mediated by ABI4 and ABI5 (Shu et al., 2016;Wang et al., 2013). The repressed FLC expression releases its inhibition on FT transcription to induce flowering. It is unclear that tomato has a functional FLC pathway although there are members of MADS genes sharing high similarity with FLC. But, the action of ABA on tomato flowering time control is complex because no early-flowering phenotype was observed on the ABA-deficient mutant notabilis (not) and the flowering time of these mutants was not well associated with SFT expression in the leaves ( Figure S2b). Developing new elite crop varieties with optimal flowering time and plant architectures is a prerequisite to meet the increasing demand for food, feed and biofuel production. For many crops, the transition of vegetative to reproductive phase is governed by the conserved FT pathway, and natural allelic variations at FT orthologs have been widely used in traditional breeding programmes (Blumel et al., 2015). Thus, manipulating the transcription of these FT orthologs has great potential applications in breeding programmes of crops whose flowering time is mainly governed by this conserved pathway. The transcription factor SlZFP2 has the activity to activate the expression of the tomato FT homologous gene SFT by direct binding to the core (A/T)(G/C)TT elements in the latter's promoter region, making it an excellent candidate to manipulate SFT expression for flowering time control in tomato. In addition, it has also been shown that SFT not only controls flowering time but also regulates heterosis depending on the presence of a mutation in SP (Krieger et al., 2010). The yield heterosis observed in heterozygous sft hybrids is likely due to dosage-dependent SFT action on shoot and inflorescence development resulting from the modification of flowering time because reducing SFT expression by artificial microRNA recapitulated the shoot architecture phenotypes of sft/+ heterozygotes (Jiang et al., 2013). It has been proposed that meristem-specific SFT/SP ratio mediates plant growth balance, which high ratio causes growth arrest and termination in shoot meristems (Lifschitz et al., 2014). In the genetic background without the sp mutation such as LA1589, overexpression of either HA-SlZFP2 or SlZFP2 has no obvious negative impact on fruit weight (Weng et al., 2015). Thus, manipulation of SlZFP2 expression might also be used for yield improvement through optimizing the balance between growth and flowering in tomato and other crops.

Plant materials and growth conditions
The wild relative of tomato Solanum pimpinellifolium LA1589, the mutants flc, sit, not, sft and the cultivars Ailsa Craig, LA0534 and LA0535 used in this study were obtained from Tomato Genetics Resource Center at University of California, USA. The tomato cultivar M82 was provided by Dr. Daniel Zamir at the Hebrew University of Jerusalem, Israel. Generation of the transgenic lines overexpressing HA-SlZFP2 or SlZFP2 in LA1589 and M82 as well as the SlZFP2 RNAi lines from LA1589 has been described in our previous study (Weng et al., 2015). Because most of the M82 lines overexpressing HA-SlZFP2 had no or very few seeds, we conducted most of the experiments described in this study on the HA-SlZFP2 overexpression lines from LA1589. The transgenic lines together with the wild-type plants were grown in phytotron under 70-80% relative humidity at 20-25°C and illuminated for 16 h daily by light intensity of 150 mE/m 2 /s from metal halide lamps and high pressure sodium lamps. To maintain optimal growth, the three ABA-deficient mutants (not, sit and flc) were sprayed 50 lM ABA at 10-day intervals. Plants were fertilized weekly with all-purpose fertilizer and watered as needed.

Phenotypic analysis of SlZFP2 overexpression lines
Flowering time and branch number was recorded mainly at 45 days postgermination on 3-4 independent transgenic lines and their corresponding siblings segregated from heterozygous transgenic lines (using as the wild-type controls). Flowering time was recorded as leaf number before first inflorescence. For branching measurements, only branches of 0.5 cm or longer were counted. Flowering time and branching phenotype were recorded for three to four independent transgenic lines overexpressing HA-SlZFP2 in LA1589 and M82 as well as four lines overexpressing SlZFP2 in LA1589.
For apical dominance analysis, 15-and 30-day-old plants of the representative HA-SlZFP2 overexpression line L103 and the wild type were decapitated, and 0.5% (w/w) NAA mixed with lanolin was applied to the decapitated stumps. Same amount of lanolin without NAA was used as mock control. The lengths of the first and second branches below the decapitated sites were measured daily after treatment.
To test whether the high branching phenotype observed on these SlZFP2 overexpression lines was caused by impacted strigolactone biosynthesis, reciprocal grafting was made between 15-day-old seedlings of the representative HA-SlZFP2 overexpression line L103 and the wild type and the number of branch (≥0.5 cm) was counted at 50 days after grafting.

Profiling of global gene expression by RNA sequencing
Two homozygous transgenic lines L103 and L104 showing very similar phenotypes in flowering and branching were chosen for profiling gene expression regulated by SlZFP2 via RNA sequencing (RNA-Seq). Their respective nontransgenic siblings (L103N and L104N) were served as controls. Total RNA was extracted by Trizol reagent (Invitrogen, Carlsbad, California, USA) from shoot tips including one small visible leaf was collected from one-and-halfmonth-old plants as described previously (Xiao et al., 2009). Paired-end sequencing libraries were created and sequenced by Shanghai Hanyu Bio on Illumina's Genome Analyzer IIx system using 100-bp reads. Reads were mapped to tomato genome using Tophat program v.1.3.2 (Trapnell et al., 2009), and the transcripts were assembled using cufflinks (Trapnell et al., 2010) based on transcripts predicted in SL2.40 assembly from SGN (version ITAG2.3). In total, the numbers of fragments (one or both of the paired-end reads) mapped to SL2.40 assembly were 13 560 517 (90.4%) for L103, 13 778 522 (91.9%) for L104, 13 592 147 (90.6%) for L103N and 13 834 382 (92.2%) for L104N.
The expression values of the transcripts were calculated in fragments per kilobase of transcript per million mapped reads (FPKM) (Mortazavi et al., 2008). Differentially expressed genes were identified using t statistics with P values adjusted for false discovery rate using Benjamini and Hochberg method (Benjamini and Hochberg, 1995). Adjusted P value of 0.05 or less was considered as statistically significant. Total of 1205 differentially expressed genes were identified. We verified transcript levels of 33 genes with different FPKM values by real-time quantitative RT-PCR (qRT-PCR) on the same batch of RNA samples used for RNA-seq and found the two methods are comparable (Table S2) was barely detected (CT >35). Therefore, the differentially expressed gene list was further filtered with arbitrary FPKM cutoff at 0.1 and fold change at 1.5. Genes with twofold or higher change but not identified by t statistics were also included. In total, 1276 genes were considered differentially expressed between the HA-SlZFP2 overexpression lines and nontransgenic controls.
The raw reads and gene expression data have been deposited in the National Center for Biotechnology Information under accession number GSE45243.

Real-time quantitative RT-PCR
Total RNA was extracted from various tomato tissues as described above. Residual genomic DNA in RNA samples was removed by RNase-free DNase (New England Biolabs (Beijing), Beijing, China) at 37°C for 10 min, and 1 lg of DNase-treated total RNA was converted to first-strand cDNA using First Strand cDNA Synthesis Kit (Fermentas, Vilnius, Lithuania, EU). qRT-PCR was performed in three biological replicates using SYBRâ Premix ExTaqTM (Takara DaLian, China) on an ABI Applied Biosystems StepOnePlus machine, except the data shown in Table S2 were from three technical replicates. Transcript level was expressed as relative expression normalized with SleIF4a6 signals.

Chromatin immunoprecipitation assay
Chromatin immunoprecipitation assay was performed on the four HA-SlZFP2 overexpression lines as described previously (Weng et al., 2015). Essentially, young leaves with shoot apices of 45-day-old plants were used for chromatin isolation. Chromatin was immunoprecipitated with HA monoclonal antibody (Sigma St. Louis, Missouri, USA) coupled on Dynabeads Protein G (Invitrogen, Carlsbad, California, USA). Eluted DNA samples were further purified by phenol/chloroform extraction after Protease K treatment. Then, 2 lL ChIPed DNA per sample dissolved in 100 lL TE was used for qPCR analysis.

Electrophoretic mobility shift assay
Electrophoretic mobility shift assay (EMSA) was performed on E. coli expressed GST-SlZFP2 fusion protein and DNA fragments of SFT promoter containing putative binding sites for SlZFP2 as described previously (Weng et al., 2015). The same primers used for ChIP-qPCR analysis were used for PCR amplification of probe template from LA1589 genomic DNA. Purified PCR fragments from SFT genomic DNA were labelled by DNA 3 0 End Biotinylation Kit (Pierce, Waltham, Massachusetts, USA). The binding reactions were conducted at room temperature in binding buffer [10 mM Tris (pH7.5), 50 mM KCl, 1 mM DTT, 2.5% glycerol, 0.05% NP-40, 5 mM MgCl2, 0.5 mM EDTA, 50 ng/mL poly (dIÁdC)] containing 1.5 lg purified GST-SlZFP2 fusion protein and 50 fmol probes. Protein-DNA complex was separated on 6% native polyacrylamide gel in 0.5xTBE and then was transferred onto Hybond-N+ nylon membrane (GE Healthcare Life Sciences, Chicago, USA). Protein-DNA interaction was detected using Light Shift Chemiluminescent EMSA Kit (Thermo Fisher Scientific, Waltham, Massachusetts, USA).

Transient gene expression in N. benthamiana leaves and Arabidopsis protoplasts
For transient gene expression in N. benthamiana leaves, the reporter construct for expression of the YFP-SFT fusion protein driven by the native SFT promoter was made by placing the 2.0kb SFT promoter amplified using primer XP2845 and XP1038 to the upstream of the YFP-SFT coding sequence, which was prepared by cloning the full-length cDNA of SFT amplified using primers XP1830 and XP2849 to the 3 0 end of the YFP coding sequence (information of all primers used in this study can be found in Table S3). The expression cassette was then cloned into the pZH001 vector derived from pBI121. The two constructs for expression of SlZFP2 (effector) and YFP (control) under the control of the 35S promoter have been described previously (Weng et al., 2015). After verification by sequencing, the plasmids were introduced into the Agrobacterium tumefaciens strain GV3101, respectively. A. tumefaciens GV3101 strain containing reporter or control plasmid was co-infiltrated with the agrobacteria containing the effector plasmid into N. benthamiana leaves. After three days postinfiltration, transient expression of YFP-SFT fusion protein was monitored using an Olympus FV1000 confocal microscope. Transient gene expression in Arabidopsis protoplast was conducted as described previously with a plasmid containing GUS expression cassette driven by 1.8kb SFT promoter amplified by primers XP1773 and XP2070 (Weng et al., 2015). expression profiles, and the effect of SUN on fruit shape. BMC Plant Biol. 9, 49. Xu, Y., Gan, E.S., Zhou, J., Wee, W.Y., Zhang, X. and Ito, T. (2014) Arabidopsis MRG domain proteins bridge two histone modifications to elevate expression of flowering genes. Nucleic Acids Res. 42, 10960-10974. Yan, Y., Shen, L., Chen, Y., Bao, S., Thong, Z. and Yu, H. (2014) A MYB-domain protein EFM mediates flowering responses to environmental cues in Arabidopsis. Dev. Cell, 30, 437-448. Yu, Y., Liu, Z., Wang, L., Kim, S.G., Seo, P.J., Qiao, M., Wang, N. et al. (2016) WRKY71 accelerates flowering via the direct activation of FLOWERING LOCUS T and LEAFY in Arabidopsis thaliana. Plant J. 85, 96-106. Zhang, B., Wang, L., Zeng, L., Zhang, C. and Ma, H. (2015) Arabidopsis TOE proteins convey a photoperiodic signal to antagonize CONSTANS and regulate flowering time. Genes Dev. 29, 975-987.

Supporting information
Additional Supporting Information may be found online in the supporting information tab for this article: Figure S1 Positive regulation of SFT expression by SlZFP2 in mature green fruits. Figure S2 Flowering time and SFT expression in the leaves of ABA-deficient mutants and their wild types. Table S1 Differentially expressed genes in p35S:HA-SlZFP2 lines and nontransgenic controls.