Gerbera hybrida “Shenzhen No. 5” seedlings were grown under standard greenhouse conditions at 26/18°C (day/night temperature) and a relative humidity of 65–80%. Ray petals at different developmental stages (Meng and Wang, 2004), and other tissues and organs were sampled for the following experiments.

Arabidopsis thaliana seeds (Col-0) were surface sterilized, plated on Murashige and Skoog medium (MS) (Sigma–Aldrich), and imbibed in darkness at 4°C for 3 days for vernalization. The plates were transferred to a growth room (21–22°C, 60% relative humidity) under long-day conditions (16 h light/8 h dark) for 7 days, then seedlings were transplanted into pots containing peat:vermiculite (3:1) and grown under the same conditions.

The promoter sequence of GEG was obtained by high-efficiency thermal asymmetric interlaced PCR (Hi-TAIL PCR) as previously described (Liu and Chen, 2007). The promoter sequence was analyzed using PLACE¹ and PlantCARE². The primers used for this experiment are listed in Supplementary Table S1.

Protoplasts were isolated from the rosette leaves of 4-week-old Col-0 as previously (Sun et al., 2013). The dual-luciferase assay was performed according to (Hellens et al., 2005). To characterize the core GEG promoter sequences, four fragments of the GEG promoter, named P1365, P880, P580, P260 (with the numbers representing the base pairs to the start of each fragment upstream from the ATG translational start site) were fused into pGREEN0800-LUC to generate reporter vectors. The modified pBluescript vector (pBS) (Paul et al., 2016) was used as an effector. For the interaction study of GhMIF and the GEG promoter, the pGREEN0800-LUC plasmid incorporated with pGEG₃₂₀ (-580∼-261) was used as the reporter, and the pBS-GhMIF was used as an effector. Meanwhile, the pGREEN0800-LUC plasmid incorporated with pGEG₁₇₀ (-430∼-261) was used as the unspecific binding control. The effectors were co-introduced with the reporters into the protoplasts as previously described (Yoo et al., 2007) and the transformed protoplasts were incubated at room temperature for 20–22 h. The dual-luciferase assay was performed as described by the manufacturer (Dual-Luciferase^® Reporter Assay, Promega, United States), and the Firefly and Renilla luciferase activities were detected using an Enspire multi-mode microplate reader (PerkinElmer Inc., United States). Three biological replicates were performed for all experiments.

For the preparation of cDNA library of G. hybrida, total RNA was extracted from ray petals at stage 1 to 6 (S1–S6) (Meng and Wang, 2004) with a RNA extraction kit (Waryong, Beijing) according to the manufacturer’s instructions. A cDNA library of a storage capacity of 1.05 × 10⁷ CFU was established after a series of steps: mRNA purification, cDNA synthesis, 5′ adaptor ligation, homogenization treatment, fusion with AD vector and transformation to Escherichia coli. To obtain the interaction proteins with pGEG₃₂₀, the cDNA library was introduced into bait strains carrying the pGEG₃₂₀-AbAi vectors. Screening and identification of interaction colonies were carried out as described in the Matchmaker Gold Yeast One-Hybrid Library Screening System kit (Clontech, United States). GhMIF, a potential interaction protein, was selected for further analysis.

To confirm the interaction between the pGEG₃₂₀ and GhMIF, the pGEG₃₂₀ region (from -580 to -261 bp upstream of the initial codon ATG) and pGEG₁₇₀ (-430∼-261) were inserted into the pAbAi vector, which was linearized by BstBI digestion and transformed into a Y1HGold strain to generate the Y1H bait strain and the GhMIF ORF (309 bp) was inserted into the pGADT7 vector as the prey plasmid. Y1HGold transformed with pGEG₃₂₀-AbAi was grown in SD-Ura medium to screen for successful transformation, then the pGADT7 prey vector harboring GhMIF was integrated into Y1HGold [pBait-AbAi] yeast strains, and the yeast cells were grown on SD-Leu medium with 200 ng/mL AbA (Aureobasidin A, Clontech, United States) to test the interaction. The plates were cultured at 30°C for 3–5 days. Primers used for the Y1H assay are listed in Supplementary Table S1.

A Y2H assay was performed using a Clontech (Matchmaker^® Gold Yeast Two-Hybrid System, Cat. No. 630489) kit with the AH109 yeast strain. The GhMIF ORF was sub-cloned into the EcoRI/BamHI sites of the pGBKT7 vector (Gal4 DNA binding domain, Clontech) as the bait. The bait construct and empty pGADT7 (Gal4 activation domain, Clontech) were co-transformed into AH109 as previously described (Gietz and Schiestl, 2007). Transformed monoclonal yeast cells were identified on SD-Trp/-Leu medium and transferred to SD-Leu/-Trp/-His medium containing 20 mg/ml 5-Bromo-4-chloro-3-indolyl-D-galactopyranoside (X-Gal, Clontech) to test the transcription activation activity and images were acquired after incubation at 30°C for 3 days. Primers used for the Y2H assay are listed in Supplementary Table S1.

Recombinant pET28a-SUMO-GhMIF proteins were expressed in E. coli BL21 cells induced by the addition of 0.5 mM isopropyl-β-D-thiogalactopyranoside (IPTG) overnight at 22°C, and purified on Ni-NTA columns (Qiagen, Germany). The core region of the GEG promoter pGEG₃₂₀ (-580∼-261) was used as a 5′ end biotin labeled probe using T4 polynucleotide kinase and the same fragment, but unlabeled, was used as a competitor. Meanwhile, pGEG₁₅₀ (-580∼-431) and pGEG₁₇₀ (-430∼-261) from the truncation of pGEG₃₂₀ (-580∼-261) were used as 5′ end biotin labeled probes for specific interaction verification. An EMSA analysis was conducted using the LightShift Chemiluminescent EMSA kit (Thermo Scientific, United States). After incubation at room temperature for 20 min, the reaction mixture containing 1 μg of purified fusion protein and 50 nmol/mL biotin-labeled probe for the binding reaction were electrophoresed on a 6% polyacrylamide mini-gel, and then transferred onto a positively charged nylon membrane (GE Healthcare, United States) and the transferred DNA cross-linked to the membrane with an ultraviolet lamp. The primers used for EMSA are listed in Supplementary Table S1.

GhMIF was amplified and cloned into the modified vector pBluescript II SK carrying a yellow fluorescence protein (YFP) to generate the YFP-GhMIF construct. The construct was then transformed into A. thaliana protoplasts (generated as above) and the fluorescence was detected with a laser confocal scanning microscope (LSM710, CarlZeiss, Germany) approximately 12 h after transformation. The empty YFP vector was used as a control. Primers used for the experiment are listed in Supplementary Table S1.

A transient petal transformation assay was performed as previously described (Pei et al., 2013). The GhMIF ORF (309 bp) was used to generate an overexpression vector (pCANG-GhMIF) which is derived by the 35S promoter and a virus-induced gene silencing vector (pTRV2-GhMIF). pTRV2-GhMIF, pTRV1 and pCANG-GhMIF were separately transformed into Agrobacterium tumefaciens strain C58C1. A. tumefaciens was grown in Luria-Bertani broth (LB) containing 75 μg mL^-1 kanamycin and 60 μg mL^-1 rifampicin at 28°C with shaking at 200 rpm overnight. The cultures were then diluted 1:50 (v/v) in fresh LB containing 10 mM of 2-morpholinoethanesulfonic acid (MES), 20 mM acetosyringone, 75 μg mL^-1 kanamycin and 60 μg mL^-1 rifampicin and grown overnight (12–14 h). The cultures were centrifuged at 1,000 g for 10 min and resuspended in an infiltration buffer (10 mM of MES, 10 mM of MgCl₂, 20 mM acetosyringone, pH = 5.6) to an OD₆₀₀ of ∼1.2. A. tumefaciens cultures containing pTRV2-GhMIF and pTRV1 at a ratio of 1:1 (v/v), and a mixture containing pTRV2/pTRV1 as a negative control, were stored at 28°C for 4 h in the darkness prior to infiltration.

Detached ray petals at stage 3 in length of 1.5–2.0 cm (Meng and Wang, 2004) were submerged into infiltration solution with a vacuum of -0.09 MPa for 5 min, then the vacuum was slowly released within 2 min to ensure A. tumefaciens entering ray petals. After infiltration, ray petals were rinsed with sterile distilled water (ddH₂O) to remove the remnants and then placed in the glass dish with filter papers. After incubation at 8°C for 3 days, the transformed petals were transferred to a growth chamber at 23–25°C for 8 days with 50–60% humidity under long-day conditions (16 h light/8 h dark) and sprayed 1 mL ddH₂O once a day. For each treatment, at least 90 petals with three technical replicates were infiltrated and three biological replicates were performed for the experiment. The sequences of the primers used are listed in Supplementary Table S1.

Ray petal and cell lengths were measured as previously described (Li et al., 2015). After transient transformation, petals from different treatments (mock, 35S::GhMIF, and pTRV2-GhMIF/pTRV1) were cultivated in normal growth conditions for 8 days. And photographs were taken each day with a digital camera (Nikon, D7200, Japan). A total of 90 petals for each treatment with three technical replicates were analyzed using ImageJ software³ (NIH, Bethesda, MD, United States) and data were used to evaluate the elongation rate of the basal and whole petals. The elongation rate was calculated according to the following equation: Elongation rate = (Lt–Li)/Li × 100%, where Lt is the petal length at the 8th day after normal cultivation while Li is the initial petal length at the 1st day after transferred to the normal growth conditions. For measurements of epidermal cell length and number, the basal regions of 10 petals from each treatment were stained with propidium iodide (0.1 mg mL^-1) for 5 min at room temperature, then immediately washed with deionized water. Images of the epidermal cells were taken using a laser confocal scanning microscope (LSM710, CarlZeiss, Germany). To determine the average cell length, 20 epidermal cells from each photograph were measured using ImageJ. To determine the mean cell number, all epidermal cells in the visual field were counted. The average cell number was used to represent the total number of epidermal cells in each petal.

The tissue samples of G. hybrida were lyophilized and stored at -80°C until use. Total RNA was extracted with a RNA extraction kit (Waryong, Beijing) according to the manufacturer’s instructions. Extracted RNA was treated with DNase to remove contaminating DNA and the gene specific primers were designed by the Primer Premier 5. The first-strand cDNA was synthesized from the 0.5 μg of total RNA using the 5 × RT Master Mix (Toyobo, Japan). The synthesized cDNA (10 μl) was diluted into 50 μl with ddH₂O and then used as template for qRT-PCR with the 2 × Realstar Green Fast Mixture (GeneStar, Beijing). Each PCR reaction (20 μl) contained 0.4 μl cDNA template, 10 μl 2 × RealStar Green Mixture, 0.4 μl (0.2 μM) gene-specific primers and 8.8 μl ddH₂O. qRT-PCR was performed on a Bio-Rad CFX96^TM qRT-PCR system using the following procedure: 95°C for 60 s, followed by 40 cycles of 95°C for 15 s, 60°C for 15 s and 72°C for 45 s. Melt curve analysis was performed on the end of PCR using the following procedure: 65°C for 5 s, then to 95°C with the increment of 0.5°C/s, to determine the specificity of reactions. The PCR amplification efficiency of ACTIN, GhMIF and GEG is 93.5, 94.8, and 98.2%, respectively. The relative expression levels were calculated using the 2^-ΔΔC_T method (Livak and Schmittgen, 2001). All reactions were performed with three technical and three biological replicates. The data were normalized to the ACTIN (AJ763915) gene as previously described (Kuang et al., 2013). The primers used are shown in Supplementary Table S1.

Ray petals detached from inflorescences at stage 5 were placed on two layers of Whatman filter papers that were immersed in GA₃ (10 μM), ABA (50 μM), GA biosynthesis inhibitor paclobutrazol (PAC) (10 μM) and ABA biosynthesis inhibitor fluridone (FLU) (0.1 μM), respectively, for 0, 4, 8, 12, 16, and 24 h at 23°C with 50–60% humidity under the long-day condition (16 h light/8 h dark). Petals treated with 0.1% ethanol in deionized water were used as a control. Approximately 0.1 g petal tissue from the different treatments for each time point were sampled in parallel then frozen in liquid nitrogen prior to processing for qRT-PCR analysis. Hormones and inhibitors were acquired from the Sigma–Aldrich Chemical Co. (Shanghai, China). The experiment was performed with three biological replicates.

One-Way ANOVA was used to analyze the data from the three biological replicates with SPSS 13.0 software (SPSS Inc., Chicago, IL, United States). Tukey’s honestly significant difference (HSD) test was applied to evaluate the statistical significance (^∗p < 0.05, ^∗∗p < 0.01).

GEG (AJ005206) from G. hybrida has been identified as a suppressor of ray petal growth in the later developmental stages (Kotilainen et al., 1999). However, the underlying molecular regulatory mechanism is not clear. Here, we first cloned 1,365 bp of the GEG promoter region by Hi-TAIL PCR (Supplementary Figure S1A), and showed by histochemical staining of P1365::GUS transgenic Arabidopsis seedlings that GEG expression was induced by GA (Supplementary Figure S1B) as previously reported (Kotilainen et al., 1999).

GEG promoter was analyzed using PLACE⁴ and PlantCARE⁵ and the result showed that there distribute hormones (GA, ABA, and Auxin) responsive elements, binding elements for transcription factors (MYB, MADS, WRKY71, C₂H₂ ZFP, and TCP-domain protein) and cell division related elements in the 1,365 bp promoter (Figure 1A). To identify the core region in the 1,365 bp promoter that influence the expression level of GEG, four 5′ truncated fragments (P1365, P880, P580, and P260) of the GEG promoter (Figure 1A) were used to generate reporter constructs for a dual-luciferase assay in A. thaliana Col-0 protoplasts. The result showed that P1365, P880, and P580 all exhibited similar LUC/REN ratios (2.14- to 2.19-fold), which were significantly different from the control. However, P260 showed much lower reporter activity than P580 although it was higher than the control (1.73-fold) (Figure 1B). These data indicated that the region from -580 to -261 bp (pGEG₃₂₀) in the GEG promoter is of importance for GEG expression, which was also supported by the results from the transient expression assay in rice callus (Supplementary Materials and Methods and Figure S1C).

To identify TFs that interact with the GEG promoter, a Y1H screen was performed (Supplementary Figures S2A,B). Using the core GEG promoter region pGEG₃₂₀ (-580 to -261 bp) as a bait, an interacting mini zinc-finger protein (MF370885, GhMIF) was identified, which shares 69% sequence identity with the AtMIF2 protein from A. thaliana (Supplementary Table S2). To confirm the interaction between GhMIF and the bait, a Y1H assay was performed, which further indicated that GhMIF could induce the expression of the reporter gene driven by the GEG promoter (Figure 2A), and suggested that GhMIF specifically binds to the GEG promoter.

We then performed an EMSA analysis to validate the physical interaction between GhMIF and the GEG promoter in vitro. The core GEG promoter region pGEG₃₂₀ (-580 to -261 bp) was used as a biotin-labeled probe, and the same oligonucleotide, but unlabeled, was used as a competitor. The competition assay was carried out by adding excess amounts of the unlabeled probe. As shown in Figure 2B, the GhMIF protein bound to the DNA probes. Furthermore, increasing the concentration of unlabeled probe in the binding reactions resulted in weaker bands, suggesting that the GhMIF protein directly bound to the core region of the GEG promoter in vitro. The truncation assay further found that GhMIF protein binds to the pGEG₁₅₀ region that contains the C₂H₂ ZFP binding motif (TACAAT), not to the pGEG₁₇₀ (Supplementary Figure S3), which suggested that GhMIF has a specific binding to GEG promoter.

We also performed a dual-luciferase assay with A. thaliana Col-0 leaf protoplasts. The recombinant pGREEN0800-GEG₃₂₀ and pGREEN0800-GEG₁₇₀ as reporters, respectively, and the fused protein with GhMIF as an effector, were co-introduced into the protoplasts. Co-transformation of pGEG₃₂₀ and the GhMIF protein resulted in a higher LUC/REN ratio compared with the control, but the LUC/REN ratio for pGEG₁₇₀ and GhMIF protein was similar to the control (Figure 2C). The result indicated that GhMIF specifically binds to the GEG promoter in vivo and can activate the expression of the reporter gene. Meanwhile, we performed a subcellular localization analysis of GhMIF and a transactivation assay using a Y2H assay. The results showed that the YFP-GhMIF fusion protein accumulated in both the nucleus and cytoplasm (Figure 2D) and GhMIF acts as a transcriptional activator by the detection of β-glucosidase activity (Figure 2E). Taken together, these data indicated that GhMIF acts as a transcriptional activator of GEG.

To characterize the function of GhMIF in petal growth, transient transformation assays of ray petals were performed. After transformation and incubation at 8°C for 3 days, ray petals were transferred to a growth chamber at 23–25°C for 8 days. From days 1 to 8, the elongation rate in petals over-expressing GhMIF (35S::GhMIF) was lower than in control petals, but was higher in petals in which GhMIF expression was suppressed using virus induced gene silencing (VIGS) (pTRV2-GhMIF/pTRV1). For the whole petals (including top, middle, and basal regions), the elongation rates of ray petals were 0.23 ± 0.06 in 35S::GhMIF and 0.55 ± 0.05 in pTRV2-GhMIF/pTRV1, which corresponds to a decrease of 47% and an increase of 27%, respectively, compared with the elongation rate of 0.43 ± 0.03 in the mock control (Figures 3A,B). Because the basal region of ray petals was the main zone of elongation during treatment with GA3 (Li et al., 2015), we then focused on this area. As shown in Figure 3B, the elongation rates of basal petals were 0.22 ± 0.02 in 35S::GhMIF and 0.54 ± 0.04 in pTRV2-GhMIF/pTRV1, with a decrease of 50% and an increase of 24%, respectively, compared with that of the mock control.

To determine whether the change in petal length was caused by cell elongation or cell division, we measured the number and size of the basal cells in ray petals after cultivation for 8 days. As shown in Figures 3C,D, the average length of the epidermal cells was 32 μm in 35S::GhMIF petals compared with 60 μm in the mock controls. In contrast, the average length of the epidermal cells from pTRV2-GhMIF/pTRV1 lines was approximately 89 μm. No significant differences in cell numbers were observed in the transiently transformed petals (Figure 3E). The data suggested that GhMIF inhibits petal elongation primarily by affecting cell elongation, rather than cell division.

Quantitative RT-PCR assays were also performed to detect the expression level of GhMIF and GEG in the transient transformation lines (Supplementary Figure S4). The result showed that the expression level of GhMIF was significantly upregulated (∼2.40-fold) in the overexpression lines and downregulated (∼0.40-fold) in the silenced lines compared with the mock control (Figure 3F), which indicated that the GhMIF expression level was inversely proportional to the petal elongation rate and cell length (Figures 3B–D). Besides, the GEG expression was significantly upregulated (∼1.95-fold) in petals of the GhMIF overexpression lines, but decreased (∼0.43-fold) in silenced lines, compared with the mock control (Figure 3F). Taken together, these results showed that GhMIF inhibits petal growth by activating the expression of GEG.

Previously GA was reported to activate the expression of GEG in ray petals and regulate petal growth antagonistically with ABA in G. hybrida (Kotilainen et al., 1999; Li et al., 2015). We evaluated the expression levels of GhMIF and GEG following treatments with GA, ABA and their respective biosynthetic inhibitors, PAC and FLU, using qRT-PCR. The expression levels of GEG and GhMIF increased after GA₃ treatment and decreased in response to PAC treatment over the course of times. While ABA treatment attenuated and FLU treatment enhanced the expression of GEG and GhMIF (Figures 4A,B).

As shown in Figure 4A, the expression of GEG was upregulated significantly after 8 h and 4 h treatment with 10 μM GA₃ and 0.1 μM FLU, respectively, compared with the mock control. Although the expression trend of GhMIF in response to GA₃ and FLU was similar to that of GEG, GhMIF increased to a lesser extend (Figure 4B). In contrast, the expression of GEG was slightly suppressed from 0 to 16 h treated with 10 μM PAC and 50 μM ABA, which was in accordance with that of GhMIF. These results suggested that GA, ABA and their biosynthetic inhibitors affected the expression of GEG and GhMIF in a similar manner, but to a different extent.

Meanwhile, the expression level of GEG and GhMIF was investigated in the petal developmental stages (S1–S6) and different floral organs and tissues. The spatial–temporal expression assay showed that the expression level of GEG increased in parallel with the expression of GhMIF from S1–S6 (Figure 4C), and these two genes were expressed at higher levels in floral organs than in other tissues (Figure 4D). Taken together, these data are consistent with the expression of GEG and GhMIF being coordinated during petal development, which is also supported by the fact that the expression level of GEG was up-regulated and suppressed significantly in GhMIF over-expression and silenced lines, respectively (Figure 3F).

Petal growth is a complex physiological process that is regulated by complicated GRNs. Previous studies have suggested that zinc finger proteins are important components of GRNs. For example, in A. thaliana, a C₂H₂-zinc finger transcription factor, RABBIT EARS (RBE), is required for petal development of Arabidopsis, and its action results in the transition from cell division to post-mitotic cell expansion through repression of the CIN-TCP genes (TCP5, TCP13, TCP17) during early petal development (Huang and Irish, 2015). Meanwhile, in the Asteraceae plants, the CYC genes encoding TCP transcription factors have a conserved function in control of the identity of flower type, probably through cell division, cell elongation or organ fused events (Broholm et al., 2008; Kim M. et al., 2008). In addition, in Chrysanthemum morifolium ‘jinba,’ transcriptome and hormone analyses on petals revealed that some zinc finger proteins exhibit obvious up-regulation; a result that is congruent with them influencing petal growth (Wang et al., 2017). Mini zinc-finger proteins (MIF), a subgroup of the ZFP family, have been identified from a few species, namely Arabidopsis and tomato so far, and some of them were found to be involved in the development of floral organs and leaves, consistent with the functions of ZFPs family (Cheuk and Houde, 2016). In respect to the function of MIFs in petal growth, only AtMIF1 has been reported to inhibit the development of floral organs. However, the target genes of MIF or other MIF proteins still need to be explored (Hu and Ma, 2006).

In this study, for the first time we present the experimental evidences that GhMIF, a member of mini zinc-finger protein family, inhibit petal elongation by activating GEG in G. hybrida. This is supported by the following results: (1) For the whole petal, the elongation rate was decreased by 47% in GhMIF overexpression lines and increased by 27% in GhMIF silenced lines, compared with the mock control (Figure 3B), indicating that GhMIF is a negative regulator of petal elongation. Furthermore, our work also revealed that the basal region is the main elongation part of ray petals (Figure 3B), in agreement with previous studies (Li et al., 2015; Wang et al., 2017). (2) It was verified that GhMIF bound directly to the core region of the GEG promoter (Figures 2A,B). Moreover, the expression of GEG was significantly upregulated in petals of the GhMIF overexpression lines and decreased in those of GhMIF silenced lines, compared with the mock control (Figure 3F), which suggested that GhMIF could activate the expression of GEG consistent with the result in Figure 2C.

It’s also well known that petal growth is a complex process that integrates the cell division and cell expansion. And previous studies have revealed that petal size mainly depend upon cell expansion at the later stage of flower opening in G. hybrida (Meng and Wang, 2004; Zhang et al., 2012). In our work, GhMIF was confirmed to limit the cell expansion thereby affecting the petal size (Figure 3C), in which only the cell length is significantly changed, but the cell number remains unchanged compared with the mock control (Figures 3D,E) as previously described (Laitinen et al., 2007). Taken together, GhMIF plays important roles in the growth of ray petals by affecting cell elongation. Although some transcription factors with putative specificity to individual flower types have been identified by a 9k gerbera cDNA microarray, it is found that in disk flowers there only existed several basic TFs related to chromatin assembly, regulation of transcription initiation and mRNA processing (Laitinen et al., 2006). So, whether GhMIF is involved in the growth of disk florets remains to be investigated through various experimental methods, such as the RNA-blot and in situ hybridizations.

Gibberellin and ABA are known to antagonistically regulate various developmental processes throughout the plant life cycle, including stress responses (Yuan et al., 2017), seed germination and seedling growth (Lin et al., 2015; Zhong et al., 2015; Liu et al., 2016), and the formation of arbuscular mycorrhizal symbiosis (Martín-Rodríguez et al., 2016). For example, in rose (R. hybrida), the activity of Rh-PIP2;1, which is involved in ethylene-mediated rose petal expansion, was reported to be enhanced by GA₃ at the later stages but suppressed by ABA during the early stages (Ma et al., 2008; Li et al., 2009; Pei et al., 2013). Our previous studies also demonstrated GA and ABA have antagonistic effects on the petal growth in G. hybrida (Li et al., 2015), but the nature of the cross-talk between hormones and elongation promoting genes in this system is still not well defined. Findings from the current study suggest that GA and ABA have antagonistic effects on the expression of GhMIF which acts as a transcription factor to inhibit petal elongation. GA treatment promotes and ABA suppresses the expression of GhMIF with a similar degree. When treated by the biosynthesis inhibitor of endogenous GA and ABA (PAC and FLU), the expression level of GhMIF are reversed (Figure 4B). These results suggest that the disturbance of endogenous GA and ABA biosynthesis contributes, at least partially, to the antagonistic effect of these two hormones on the expression of GhMIF in G. hybrida.

However, the response of GEG to GA and ABA were not congruent with those of GhMIF. GEG was activated dramatically by GA application but was suppressed weakly by ABA treatment, which is similar to the expression trend of GhMIF with a different extent. One explanation for these observations is that there might be other unknown components that could regulate the expression of GEG during GA and ABA signaling pathways. Actually, we also found three other proteins that putatively interact with the GEG promoter in the Y1H screen: GhBZR1 (MF370884), GhEIL1 (MF370883), and GhMBF1 (MF370886) (Supplementary Table S2). Based on their sequence homology and annotations in A. thaliana, these three proteins may be the components of phytohormone signaling pathways, but whether these proteins influence the expression of GEG need to be verified by further researches.