A petunia transcription factor, PhOBF1, regulates flower senescence by modulating gibberellin biosynthesis

Abstract Flower senescence is commonly enhanced by the endogenous hormone ethylene and suppressed by the gibberellins (GAs) in plants. However, the detailed mechanisms for the antagonism of these hormones during flower senescence remain elusive. In this study, we characterized one up-regulated gene PhOBF1, belonging to the basic leucine zipper transcription factor family, in senescing petals of petunia (Petunia hybrida). Exogenous treatments with ethylene and GA3 provoked a dramatic increase in PhOBF1 transcripts. Compared with wild-type plants, PhOBF1-RNAi transgenic petunia plants exhibited shortened flower longevity, while overexpression of PhOBF1 resulted in delayed flower senescence. Transcript abundances of two senescence-related genes PhSAG12 and PhSAG29 were higher in PhOBF1-silenced plants but lower in PhOBF1-overexpressing plants. Silencing and overexpression of PhOBF1 affected expression levels of a few genes involved in the GA biosynthesis and signaling pathways, as well as accumulation levels of bioactive GAs GA1 and GA3. Application of GA3 restored the accelerated petal senescence to normal levels in PhOBF1-RNAi transgenic petunia lines, and reduced ethylene release and transcription of three ethylene biosynthetic genes PhACO1, PhACS1, and PhACS2. Moreover, PhOBF1 was observed to specifically bind to the PhGA20ox3 promoter containing a G-box motif. Transient silencing of PhGA20ox3 in petunia plants through tobacco rattle virus-based virus-induced gene silencing method led to accelerated corolla senescence. Our results suggest that PhOBF1 functions as a negative regulator of ethylene-mediated flower senescence by modulating the GA production.


Introduction
Senescence is a genetically programmed event that occurs in the terminal phase of individual tissue or organ development in plants. It is typically characterized by wilting, discoloration, and even abscission due to the sequential breakdown of physiological and biochemical activities [1]. Although senescence makes a great contribution to plant survival by allowing nutrient recycling and reallocation [2], it may cause substantially reduced crop yield and biomass production from an agronomic perspective. The retardation of senescence process is essential for many plant species, especially ornamental plants. Numerous studies have been performed to investigate the molecular regulatory mechanisms of senescence in various plant organs, including roots [3,4], leaves [5,6], stems [7], fruits [8,9], and f lowers [10,11]. These studies demonstrate that senescence is a complicated biological process which involves a wide range of degradative, biosynthetic, and regulatory mechanisms coordinated by gene expression.
The initiation of f lower senescence is concomitant with significant alterations in the cellular components. Plant hormones are considered to impose important inf luences on the senescence of f loral organs. It has been revealed that ethylene serves as a major regulator of f lower senescence in ethylene-sensitive plants, whose senescence process shows increased ethylene release in a climacteric-like pattern [12]. Exogenous treatment with ethylene accelerates petal senescence and promotes the transcription of senescence-associated genes [13]. Abscisic acid (ABA) is also known as a promoter of f lower senescence, and its treatment resulted in accelerated senescence of f loral organs and deterioration of postharvest quality in miniature potted rose [14]. Particularly, ABA was thought to be a primary contributor to f lower senescence in ethylene-insensitive gladiolus [15]. Another hormone, jasmonic acid (JA), was shown to have similar function in accelerating senescence progression of Dendrobium orchid flowers [16]. Contrarily to the senescence-promoting effects of the hormones above, cytokinins (CTKs) and gibberellins (GAs) are proposed as important anti-senescence factors in the corollas. It has been indicated that application of GAs extends f lower life of a number of plant species, such as iris [17], rose [18], allamanda [19], tobacco [20], and freesia [21].
It is well recognized that the regulation of f lower senescence depends greatly on a complex hormonal crosstalk rather than individual hormone. For example, ethylene and ABA have been indicated to interact in senescing rose f lowers under water limitations [22]. ABA treatment repressed ethylene production during f lower senescence of hibiscus [23], whereas the treatment of carnation f lowers with an ethylene action inhibitor, silver thiosulfate, restricted ABA-stimulated senescence process [24]. It has been reported that the interplay between JA and ethylene acts in parallel to regulate the timing of f loral organ abscission and senescence [25]. GAs have been shown to be antagonistic to both ethylene and ABA in senescing corollas, and the decline in the accumulation of bioactive GAs promoted ethylene-and ABAmediated rose petal senescence [18]. A recent study suggested that ethylene treatment reduced the expression of GA biosynthetic genes and increased the transcription of GA catabolic genes in rose f lowers [26]. However, how these hormonal interactions are transcriptionally regulated is largely unclear.
The basic leucine zipper (bZIP) transcription factors comprise an extensive and conserved family, which is divided into 10 distinct groups in plants [27]. The group S contains the most bZIP members, which are associated with plant growth and development. Among them, AtbZIP11 from Arabidopsis has been demonstrated to regulate amino acid and sugar metabolism by specifically activating asparagine synthetase 1 and proline dehydrogenase 2 [28]. AtbZIP11 has been identified as a quantitative modulator of auxin-stimulated responses through the regulation of histone acetylation [29]. The functions of some AtbZIP11 homologs from other plant species have also been studied. For instance, an upregulation of tbz17, a tobacco gene that is phylogenetically related to AtbZIP11, was observed during leaf senescence [30], and constitutive expression of tbz17 resulted in increased cell size and sucrose production in the leaves [31]. In kiwifruit, AchnbZIP12 was reported to positively regulate the response to ABA-promoted suberization by affecting the expression of AchnKCS [32]. In maize, a bZIP transcription factor, OBF1, was revealed to form a complex with UNBRANCHED2 and control inf lorescence architecture [33]. To date, the roles of bZIP11 and its homologous proteins in f lower senescence are still not well understood.
Petunia is an important ornamental plant with large showy f lowers, short growth period, and high genetic diversity [34]. In previous studies, petunia has been adopted as a model plant to elucidate the regulatory mechanisms underlying f lower senescence. Some up-regulated transcription factors during the f lowering period of petunia were identified through transcriptional analyses and further functionally characterized [13,[35][36][37]. We have used the tobacco rattle virus (TRV)-based virus-induced gene silencing (VIGS) method with a reporter gene PhCHS to characterize the functions of these genes. It resulted in the revealment of a bZIP transcription factor, annotated as PhOBF1, which plays a critical role in antiviral RNA silencing [38]. In the course of these studies, we have also found that transcription of PhOBF1 was significantly induced by several senescence-related hormones, and PhOBF1-RNAi and -overexpressing transgenic petunia plants showed variable f lower longevity. These findings prompt us to hypothesize that PhOBF1 may function as an important regulator of petal senescence. Here, we elucidate the molecular regulatory mechanism of PhOBF1-mediated f lower senescence. Downregulation of PhOBF1 accelerated corolla senescence, whereas its overexpression delayed the senescence process. Our data demonstrate that PhOBF1 regulates f lower senescence by modulating the GA biosynthesis in petunia.

PhOBF1 is phylogenetically related to AtbZIP11 and subcellularly localized in the nucleus
To explore the involvement of bZIP genes in f lower senescence, one up-regulated transcript encoding a putative ocs element binding factor, namely PhOBF1 (SGN accession no. Peaxi162Scf00285g00011), was identified during petunia f lower senescence through transcriptome sequencing analysis [36]. Phylogenetic analysis revealed a close relationship of PhOBF1 with AtbZIP11 in Arabidopsis, NabZIP11 in tobacco, SlbZIP11 in tomato, StOBF1 in potato, and CaOBF1 in pepper. All these proteins belong to the group S of bZIP transcription factors (Fig. 1A). PhOBF1 and its homologous proteins shared conserved basic region and leucine zipper domains. They displayed a remarkably long leucine zipper domain consisting of up to eight heptad repeats (Fig. 1B) [39]. To analyse the subcellular localization of PhOBF1, the fusion protein PhOBF1-GFP was transiently expressed in tobacco leaves with a nucleus-localized marker H2B-mCherry. Green f luorescent signals of PhOBF1-GFP were observed to overlap with red f luorescent signals of H2B-mCherry (Fig. 1C), suggesting that PhOBF1 was located in the nucleus.

PhOBF1 is up-regulated during flower senescence and under different hormone treatments
Expression profile of PhOBF1 was examined during f lower senescence using quantitative real-time PCR. Transcript levels of PhOBF1 continued to increase in the detached f lowers of both petunia cultivars 'Mitchell Diploid' and 'Primetime Blue' from anthesis (D0) to 7 days (D7) after anthesis ( Fig. 2A and B). As plant hormones are involved in f lower senescence, we analysed the transcript profiles of PhOBF1 in petunia detached corollas treated with various growth regulators. PhOBF1 were significantly up-regulated following treatments with ethylene, GA 3 , and ABA but not methyl jasmonate (MeJA). The induced transcription of PhOBF1 by the three hormones was prevented by pre-treatments with their action or biosynthesis inhibitors 1-methylcy-clopropene (1-MCP), paclobutrazol (PAC), and f luridone (FLD), respectively. In addition, the pre-treatment with MeJA biosynthesis inhibitor salicylhydroxamic acid (SHAM) did not affect the transcription of PhOBF1 (Fig. 2C). Next, we examined the effects of exogenous hormone treatments on f lower senescence. Compared to mock control, ethylene and GA 3 applications accelerated and delayed petal senescence, respectively. The treatment with a combination of the two hormones led to a recovery of ethylene-induced f lower senescence (Fig. S1A, see online supplementary material). Specifically, ethylene treatment alone shortened f lower longevity by 2.1 days, whereas GA 3 extended the longevity by 1.9 days. In contrast, the f lowers co-treated with ethylene and GA 3 showed an insignificant variation in f lower longevity compared to mock control (Fig. S1B, see online supplementary material), indicating that GAs play an antagonistic role in ethylene-induced f lower senescence.

Silencing and overexpression of PhOBF1 affect petunia flower senescence
To investigate the role of PhOBF1 in f lower senescence, we generated transgenic petunia lines with PhOBF1 RNAi silencing and overexpression using the cultivar 'Mitchell Diploid'. The f lowering phenotypes among different lines were observed at D0 and D7 after anthesis for attached corollas, or at D0 and D6 after anthesis for detached corollas. Compared to wild-type (WT) lines, PhOBF1-RNAi lines displayed accelerated f lower senescence, while PhOBF1-overexpressing lines showed delayed senescence progress ( Fig. 3A and B). PhOBF1-RNAi and -overexpressing lines displayed a substantial reduction and increase in expression levels of PhOBF1, respectively, compared to WT lines (Fig. 3C). Both attached and detached f lowers showed similar ranges of f lower lifespan variation in transgenic petunia plants. Specifically, silencing of PhOBF1 shortened f lower longevity by 1.0 to 1.2 days. In contrast, overexpression of PhOBF1 extended f lower longevity by 1.4 to 2.2 days (Fig. 3D). Transcript levels of the senescence marker genes PhSAG12 and PhSAG29 increased and decreased substantially in PhOBF1-silenced and -overexpressing lines, respectively, in comparison to WT controls (Fig. 3E).

PhOBF1 is involved in the GA biosynthesis and signaling pathways
Genetic transformation of PhOBF1 in petunia led to significant phenotypic differences in plant height and petal diameter, which were lower in the RNAi lines and higher in the overexpressing lines than in the WT lines (Fig. S2, see online supplementary material).
It is commonly known that the plant hormone GAs are responsible for the growth of plant shoots and f lowers [37]. To verify the correlation between the phenotypic differences and the GA pathway, we measured expression levels of GA biosynthetic and signaling genes in WT and PhOBF1 transgenic petunia lines. Three PhOBF1-RNAi lines showed significant reduction in transcript abundances of PhGA20ox2, PhGA20ox3, PhGA3ox1, PhGA3ox2, PhGID1A, and PhGID1B, and their transcript levels increased in PhOBF1-overexpressing lines. By comparison, PhGA2ox2 and PhGAI were up-regulated in the silenced plants and down-regulated in the overexpressing plants (Fig. 4). Silencing and overexpression of PhOBF1 did not change the expression of PhGA20ox1, PhGA20ox4, The treatment with water was used as the control (mock). Transcript levels were standardized to 26S rRNA. Error bars represent standard error of the mean from three biological replicates. Different letters indicate statistical significance as calculated by Duncan's multiple range test at P < 0.05. PhGA2ox1, and PhGA2ox3 (Fig. 4). Consistent with the transcriptional variations, the production of bioactive GAs, GA 1 and GA 3 , was significantly reduced and increased in petunia f lowers with PhOBF1 RNAi silencing and overexpression, respectively (Fig. 5).

Exogenous GA treatment inhibits ethylene-mediated petunia flower senescence
Given that the transcription of GA biosynthetic and signaling genes and GA content were changed in PhOBF1 transgenic petunia plants, we speculated that PhOBF1-mediated f lower senescence may be associated with the GA pathway. We therefore examined the impact of GA 3 treatment on petal senescence in WT and PhOBF1 transgenic lines. At D7 after anthesis, we found that the application of GA 3 alone suppressed the accelerated petal senescence in PhOBF1-silenced line (2), and further promoted the delayed senescence in PhOBF1-overexpressing line (H). In contrast, the pre-treatment with GA biosynthesis inhibitor PAC did not result in significant petal senescence change in transgenic lines, compared to mock control (Fig. 6A). The mock treatment did not affect f lower longevity of WT, PhOBF1-silenced andoverexpressing lines. Under GA 3 treatment, PhOBF1-RNAi line had similar f lower longevity with WT line, and PhOBF1-overexpressing line showed increased f lower life (Fig. 6B). In accordance with f lower longevity, PhSAG12 and PhSAG29 were highly expressed in WT and PhOBF1-RNAi lines, and lowly expressed in transgenic line overexpressing PhOBF1 (Fig. 6C). For the co-treatment with PAC and GA 3 , the overall f lower longevity and transcript abundances of two senescence-associated genes in WT and PhOBF1 transgenic lines were almost identical to those in mock-treated plants ( Fig. 6B and C).
To dissect the antagonistic effect between GAs and ethylene, the ethylene release and expression of its biosynthetic genes were examined. Compared to the mock treatment and cotreatment with PAC and GA 3 , the GA 3 treatment substantially reduced ethylene production in WT and PhOBF1 transgenic plants. Specifically, similar ethylene levels were detected in WT and The f lowers at D6 after anthesis were harvested for quantitative real-time PCR analysis. 26S rRNA was used as an internal control. Error bars represent standard error of the mean from three biological replicates. Statistical significance was determined using Duncan's multiple range test (P < 0.05) and shown as different letters.
PhOBF1-RNAi lines after GA 3 treatment (Fig. 6D), suggesting that GA 3 application reduced corolla senescence in PhOBF1-RNAi line back to the normal levels as in WT line. Correspondingly, the GA 3 treatment significantly decreased the transcription of several ethylene biosynthetic genes, including PhACO1, PhACS1, and PhACS2, in WT and PhOBF1 transgenic lines in comparison to the mock treatment and co-treatment with PAC and GA 3 (Fig. 6E). These findings suggested that PhOBF1 negatively regulates ethylene-mediated corolla senescence by modulating the GA production.

PhOBF1 directly binds to the promoter of GA biosynthetic gene PhGA20ox3
To better elucidate the regulatory function of PhOBF1, we searched the promoter sequences of those GA biosynthetic and signaling genes with altered transcription in PhOBF1 transgenic petunia plants. A G-box motif (CACGTG containing an ACGT core), which is the predicted binding site of PhOBF1's homolog AtbZIP11 [40], was identified in the PhGA20ox3 promoter region with 1.5 kb in length (Fig. S3, see online supplementary material). A short DNA fragment bearing the G-box motif in the PhGA20ox3  promoter served as a probe for electrophoretic mobility shift assay (EMSA). The results revealed a clear binding of PhOBF1 to the biotin-labeled probe, whereas the binding was dramatically suppressed when the unlabeled probe was added (Fig. 7A).
To confirm the transactivation of PhGA20ox3 promoter by PhOBF1, a yeast one-hybrid experiment using the bait and pray constructs was conducted (Fig. 7B). PhOBF1 was found to bind to the PhGA20ox3 promoter and enhance the growth of yeast cells on the SD-Ura-His-Leu medium supplemented with 3-aminotriazole (3-AT) (Fig. 7C). In addition, we also performed a dual luciferase assay to test the interaction between PhOBF1 and PhGA20ox3 promoter. A 14.6-fold increase in firef ly luciferase (LUC) activity was detected after the co-expression of 35S:PhOBF1 and pPhGA20ox3:LUC ( Fig. 7D and E), indicating that PhOBF1 specifically transactivates the PhGA20ox3 promoter.

Silencing of PhGA20ox3 by VIGS accelerates petunia flower senescence
To further verify the involvement of PhGA20ox3 in f lower senescence, a TRV-based VIGS method was employed to knockdown PhGA20ox3 in purple-f lowered petunia cultivar 'Primetime Blue'.  revealed the accumulation of TRV RNA1 and RNA2 in the f lowers systemically infected with different TRV constructs (Fig. S5, see online supplementary material). At D6 after anthesis, the plants infected with TRV-PhCHS/GA20ox3 displayed accelerated petal senescence compared to the mock-, empty vector-, and TRV-PhCHS-infected plants (Fig. 8A). Silencing of PhGA20ox3 decreased f lower longevity by 1.6 days in comparison to the controls (Fig. 8B). The suppression of PhCHS and PhGA20ox3 transcription in TRV-PhCHS/GA20ox3-infected f lowers was ver-ified by quantitative real-time PCR (Fig. 8C). PhGA20ox3 silencing resulted in a remarkably reduced production of bioactive GAs, GA 1 and GA 3 , in TRV-PhCHS/GA20ox3-infected f lowers (Fig. 8D). Two GA-inducible genes expansin 1 (PhEXP1) and cysteine proteinase 1 (PhCP1) were down-regulated in the f lowers infected with TRV-PhCHS/GA20ox3 (Fig. 8E). Expression levels of PhSAG12 and PhSAG29 in TRV-PhCHS/GA20ox3-infected corollas were detected to be significantly higher than those in the corollas infected with the mock, empty vector, and TRV-PhCHS (Fig. 8F). These The activation is expressed as a LUC/REN ratio. Error bars represent the standard error of the means from three biological replicates. Statistical significance was determined using Student's t test ( * P < 0.05, * * P < 0.01) and denoted by asterisks.
observations revealed the critical role of PhGA20ox3 in PhOBF1regulated f lower senescence.

Discussion
Flower senescence is closely associated with the economic significance of ornamental plants [41]. The delay of senescence progress is essential for improving the f lower quality. It has been indicated that endogenous plant hormones participate in the modulation of f lower senescence [42]. In this study, we revealed that a bZIP transcription factor gene, designated PhOBF1, plays a crucial role in regulating the senescence process of petunia corollas. PhOBF1 was found to specifically affect the expression of GA biosynthetic gene PhGA20ox3, which was functionally characterized as an inhibitor of f lower senescence (Figs 7 and 8). Our data suggest a PhOBF1-PhGA20ox3 regulatory module that interferes with the ethylene-mediated senescence process in petunia corollas (Fig. S6, see online supplementary material).

PhOBF1 functions as a regulator of ethylene-mediated flower senescence in petunia
bZIP family is known to have regulatory functions in many plantspecific processes, such as f lower development [43]. In Arabidopsis, the bZIP genes have been reported to control f loral inductive signals during the f lowering period [44]. AtbZIP14 also known as FD can interact with FLOWERING LOCUS T (FT) to cooperatively regulate f lowering time via a direct activation of APETALA1 (AP1) [45]. The PERIANTHIA gene encoding a bZIP transcription factor has been demonstrated to be involved in the evolution of f lower pattern in the mustard family [46]. A recent report revealed that two bZIP genes, PaFDL1 and PaFDL2, from Platanus acerifolia functioned as important regulators of f lower initiation and morphology [47]. However, little is known about the roles of bZIP genes in controlling petal senescence. In our transcriptome data, a member of the bZIP family, PhOBF1, was significantly upregulated in senescing f lowers, suggesting its potential role in modulating f lower senescence. The present results validated the crucial involvement of PhOBF1 in the regulation of f lower senescence. Our findings uncovered a novel role of bZIP transcription factor in regulating petunia petal senescence. Aside from PhOBF1, transcriptome data showed that the transcription of several other members of the bZIP family was changed during f lower senescence of petunia [36]. Further studies should be performed to clarify whether other bZIP members also play regulatory roles in corolla senescence.
Many pieces of evidence have suggested that ethylene is the key hormone controlling the senescence process of ethylenesensitive f lowers [48]. A sharp rise in ethylene production commonly results in the onset of f lower senescence [49]. The prominent role of ethylene in controlling petunia corolla senescence has been verified in previous reports [50,51]. Based on the data presented here, we conclude that PhOBF1 is implicated in the ethylene-mediated f lower senescence in petunia. The increased transcription of PhOBF1 in petunia petals under natural senescence and ethylene treatment supported this assumption (Fig. 2). We noted that GA 3 treatment also induced the transcription of PhOBF1 apart from ethylene. This observation seems to contradict the notion that ethylene and GAs promote and delay f lower senescence, respectively. One explanation is that PhOBF1 may be implicated in ethylene-and GA-mediated stress responses. It is well known that ethylene and GAs are not only senescencerelated hormones but also stress-associated ones [52]. Much evidence has suggested that these two hormones are involved in defense responses to various biotic and abiotic stresses, including virus [53,54], fungus [55,56], low temperature [57,58], and dehydration [59,60]. We have previously revealed PhOBF1 to be a positive regulator of antiviral RNA silencing. Inoculation with TRV and treatments with cold and drought significantly increased the expression of PhOBF1 [38]. The potential role of PhOBF1 in defense responses needs to be further examined in the subsequent work. In addition, ethylene and GAs serve as direct hormonal signals for affecting corolla senescence. As a transcription factor, PhOBF1 was presumed to have regulatory function upstream of ethylene and GAs here. However, the impacts of ethylene and GA 3 treatments on PhOBF1 expression seem to be a more complex regulatory mechanism, possibly involving hormonal interplay and feedback regulation. Thus, although ethylene and GAs promote and delay f lower senescence, respectively, it does not mean that they will have opposite impacts on PhOBF1 expression. Our results are consistent with a previous report, showing that both treatments with ethylene and CTK, an inhibitor of petal senescence, promoted the transcription of RhPR10.1 [11].
We also examined the PhOBF1 expression following the treatment with a combination of ethylene and GA 3 . A slight 1.8fold increase in transcript abundance of PhOBF1 was detected at 24 h after the co-treatment compared to mock control (Fig. S7, see online supplementary material). It suggests that the simultaneous application of ethylene and GA 3 may interfere with their respective inducing effect on PhOBF1 expression. This result is in accordance with the fact that there is an antagonistic effect between ethylene and GAs [18,61]. It is highly likely that the antagonism of these two hormones interrupted the functions of ethylene and GAs in the f lowers after the combined treatment, thus resulting in an impaired induction of PhOBF1 expression.
Furthermore, PhOBF1 was identified to be an important regulator of petunia f lower senescence through stable genetic transformation (Fig. 3). In most cases, the transcripts with increased expression during f lower senescence play positive roles in regulating petal senescence [35]. For example, silencing of PhFBH4, one up-regulated gene at different stages of f lower senescence, was reported to delay petunia petal senescence, while its overexpression had the opposite effect [37]. However, the contradictory results were shown in this study with reduced and increased expression of PhOBF1 promoting and suppressing the senescence of petunia petals, respectively (Fig. 3). This finding is consistent with the accelerated senescence in RhHB6-and RhPR10.1-silenced f loral organs in rose [11], and the delayed senescence in AtERF019overexpressing Arabidopsis f lowers [62]. In a recent study, a Bbox gene RhBBX28 with increased transcription during rose f lower senescence has also been shown to negatively modulate petal senescence [63]. These studies may reveal important mechanisms for inhibiting the senescence process in the f lowers.

PhOBF1 regulates ethylene-induced flower senescence by modulating GA content
The changed accumulation of bioactive GAs (GA 1 and GA 3 ) and transcription of GA biosynthetic and signaling genes were observed in PhOBF1 transgenic petunia plants (Figs 4 and 5), implying that PhOBF1-mediated f lower senescence may be dependent on the GA pathway. A number of reports have suggested the roles of bZIP transcription factors in regulating the responses to GA signals. For instance, the tobacco RSG protein with a bZIP domain has been demonstrated to modulate the GA homeostasis via the feedback regulation of NtGA20ox1, a gene responsible for synthesizing bioactive GAs [64]. The latest finding showed that the translocation of SWIZ, a bZIP protein from Brachypodium distachyon, into the nucleus was negatively affected by GA biosynthesis, and its overexpression has been found to increase transcript abundance of GA20ox4 [65]. In the present study, we revealed that PhOBF1 regulated the GA production by specifically activating the downstream gene PhGA20ox3, whose silencing resulted in shortened f lower longevity (Figs 7 and 8). Our data supported the hypothesis that PhOBF1 participates in ethylene-induced f loral senescence through the modulation of GA pathway. Application of GA 3 was observed to block the ethylene function in petal senescence by reducing the transcription of a few ethylene biosynthetic genes: PhACO1, PhACS1, and PhACS2 ( Fig. 6D and E). It is in accordance with the earlier research that the role of GAs in corolla senescence was antagonistic to that of ethylene [18,66]. Besides, we found that the combined treatment with ethylene and GA 3 resulted in an insignificant change in f lower longevity compared to mock control (Fig. S1, see online supplementary material). It demonstrated that GA 3 application inhibited ethylene-induced petal senescence, further validating the antagonism between ethylene and GAs. However, GAs and ethylene have been indicated to co-regulate the epidermal cell death of rice [67] and bud viability of larch [68] in a synergistic manner, suggesting that the interaction between these two hormones may vary among different plant organs or tissues.
It has been reported that Arabidopsis plants constitutively expressing AtbZIP11, a homolog of PhOBF1, showed inhibited plant growth compared to WT plants [28,40]. However, we found that PhOBF1 overexpression resulted in increased plant height and petal size (Fig. S2, see online supplementary material). We re-checked the sequence alignment of PhOBF1 and AtbZIP11 (Fig. 1B) and observed that the similarity between them is only 47.5%. In comparison with AtbZIP11, PhOBF1 displayed a large number of amino acid variations even in its conserved domain, suggesting a high sequence polymorphism between the two proteins. Many pieces of evidence have revealed that amino acid variations commonly result in functional divergence among different proteins in plants [69,70]. This may explain the phenotypic differences between PhOBF1-and AtbZIP11overexpressing transgenic plants. In addition, AtbZIP11 or its homolog has been shown to function as an important regulator of sugar and proline metabolism in plants [28,31,40]. It has been uncovered that sugar contributes to the prolongation of f lower life due to the suppression of ethylene biosynthesis or sensitivity to ethylene [71]. Proline may play a crucial role in the regulation of petal senescence by affecting energy depletion and reactive oxygen species accumulation [72]. Thus, whether PhOBF1 is involved in the regulation of f lower senescence by modulating sugar and proline metabolism requires further exploration.
The bZIP transcription factors are known to activate or inhibit the expression of genes whose promoter sequences contain a consensus ACGT cis-acting element [73]. Consistent with this notion, PhOBF1 was found to directly bind to the G-box motif (CACGTG) containing a core ACGT sequence (Fig. 7A). In Arabidopsis, it has been shown that some bZIP proteins, such as AtABI5 [74], AtbZIP11 [40], AtbZIP53 [75], AtbZIP67 [76], AtHY5 [77], and GBFs [78] also target the G-box motif. Other members of bZIP family exhibit variable DNA binding sites with different nucleotides f lanking the core ACGT element. For example, the ABRE-binding factors belonging to one group of bZIP transcription factors can specifically bind to the PyACGTGG/TC motif [79]. AtbZIP10 was reported to shuttle between the nucleus and the cytoplasm for binding to G-and C-box (GACGTC) motifs, respectively [80]. ThbZIP1 from Tamarix hispida was also found to have multiple binding properties to the C-, G-, and A-box (TACGTA) sequences [81]. Indeed, we identified a number of core ACGT elements in the PhGA20ox3 promoter region besides the G-box motif. The possibility of binding to those additional ACGT elements by PhOBF1 requires further examination in the future.

PhOBF1 may be involved in the complex hormonal crosstalk
Apart from ethylene and GA 3 , exogenous treatment with ABA increased expression levels of PhOBF1 in petunia petals (Fig. 2C). ABA is generally considered as a promoter of senescence in many ethylene-sensitive f lowers such as rose [14] and petunia [13]. The interplay between ABA and ethylene has been suggested in some previous reports. ABA treatment inhibited ethylene production in hibiscus f lowers [23], while the ABA-induced f lower senescence in carnation was attributed to the activation of ethylene synthesis [82]. However, a contrasting report showed that the ABA treatment increased f lower sensitivity to ethylene and transcription of ethylene pathway-related genes [83]. In petunia, ABA seems to be a downstream product of the ethylene pathway during f lower senescence, because the interruption of ethylene signaling was shown to block ABA accumulation in senescing petals [84]. In addition, an antagonistic effect of GAs on ABA existed in rose f lowers, with the reduced GA accumulation accelerating ABApromoted petal senescence [18]. Based on the induced expression of PhOBF1 by ABA, we suppose that ABA may play a role in PhOBF1-regulated petunia corolla senescence. It remains to be investigated whether PhOBF1 participates in the ABA-stimulated senescence process, and the crosstalk among ABA, ethylene, and GAs during f lower senescence of petunia is still unknown.
It is worth mentioning that endogenous hormones have proven to be critical factors in regulating plant responses to biotic and abiotic stresses. Given the impacts of several stress-associated hormones including ethylene, GA 3 , and ABA on PhOBF1 expression, we hypothesized that PhOBF1 may be essential for the defense against various environmental stimuli. In support of this hypothesis, we have previously reported an essential role of PhOBF1 in RNA silencing-mediated antiviral defense possibly by activating the salicylic acid biosynthesis pathway [38]. We also found that silencing and overexpression of PhOBF1 reduced and increased, respectively, VIGS silencing efficiency and resistance to TRV and tobacco mosaic virus infections. In particular, the down-regulation of PhOBF1 by TRV-based VIGS resulted in a substantially impaired development of PhCHS-silenced white-petal phenotype [38]. That is the reason why we did not use the TRV-PhCHS system to characterize the function of PhOBF1 in petunia corolla senescence. Moreover, the treatments with cold and dehydration dramatically increased the transcription of PhOBF1 [38]. Interestingly, AtbZIP11 from Arabidopsis has been revealed to negatively modulate the resistance to Pseudomonas syringae infection [85]. It appears likely that the involvement of PhOBF1 in plant defense responses may be a complicated process.
In conclusion, our findings demonstrate that PhOBF1 plays a crucial role in f lower senescence by modulating the GA biosynthesis. In vitro and in vivo promoter-binding tests confirmed the specific regulation of PhGA20ox3 by PhOBF1. The identification of PhOBF1, a differentially-expressed gene from transcriptome data, provides a valuable genetic solution to extend f lower longevity of petunia through genetic engineering. An integrative method of high-throughput transcriptome sequencing and VIGS system is needed to find more promising targets for controlling the senescence of petunia f lowers. To further dissect the regulatory network of PhOBF1 in f lower senescence, an extensive multiomics data analysis using PhOBF1 transgenic petunia materials should be performed in future work.

Plant materials and growth conditions
Two petunia (Petunia × hybrida) cultivars, 'Mitchell Diploid' and 'Primetime Blue' were used as experimental materials. Their seeds were obtained from Goldsmith Seeds Inc. (Santa Clara, CA, USA). Of them, 'Mitchell Diploid' was used for stable genetic transformation assay, while 'Primetime Blue' was used for TRVbased VIGS assay. Petunia seeds were sown in a tray containing the soil mix with a ratio of peatmoss to perlite being 2:1. After germination, they were placed into an artificial climate chamber under conditions of 16 h light/8 h dark and 25 • C day/20 • C night. Four-leaf-stage plantlets were transferred to small plastic pots for a continuous cultivation. Upper leaves from petunia plants were used as the explants for Agrobacterium-mediated stable genetic transformation. The plantlets at the six-leaf stage were used for VIGS inoculation with mock control and various TRV constructs. Flower longevity of each plant was determined as the duration from full opening of the petal to its complete wilting [13]. The f lowers at D0 were used for various hormone treatments. The corollas at various stages were sampled to determine the levels of gene expression and endogenous hormones.

Identification of PhOBF1
A cDNA fragment of PhOBF1 containing the complete coding region was identified among up-regulated genes from transcriptome analysis of petunia f lower senescence [36]. Its nucleotides were converted into amino acids using the translation tool on the ExPASy web server (http://web.expasy.org/translate/). Amino acids of PhOBF1 were used for searching against Sol Genomics Network (https://solgenomics.net/organism/Petunia_axillaris/ genome) and NCBI GenBank (http://www.ncbi.nlm.nih.gov/ structure/cdd/wrpsb.cgi) databases to identify its homologous proteins from petunia and other plant species, respectively. The conserved bZIP domains within the amino acids of PhOBF1 and its homologs were identified through the Conserved Domain Search Service on the NCBI web server. Comparison of multiple protein sequences was conducted using the DNAMAN alignment tool (version 8.5). Phylogenetic relationship was analysed through the MEGA tool (version 6.0.6).

Subcellular localization
The full-length coding region of PhOBF1, whose termination codon was removed, was amplified from WT petunia cDNA. It was then introduced into the binary pCAMBIA1301-GFP vector between BamHI and SalI restriction sites. The 35S:PhOBF1-GFP fusion construct under the control of CaMV 35S promoter was generated. The pCAMBIA2300-35S-H2B-mCherry vector with the red f luorescent marker mCherry fused to histone 2B was used as a reference for nuclear localization. 35S:PhOBF1-GFP/35S:H2B-mCherry or 35S:GFP/35S:H2B-mCherry were co-transformed into tobacco leaves using a biolistic PDS-1000He delivery system (Bio-Rad, Hercules, CA, USA). The plants were incubated at 25 • C for 16 h in the dark. Subsequently, the transformed cells were visualized for f luorescent signals under a TCS SP8 confocal microscope (Leica, Solms, Germany).

Treatments with various hormones
The whole f lowers with short pedicels were cut from the petunia plants at D0. They were placed immediately into small vials with distilled water or hormone solutions. For ethylene treatment, 10 μL·L −1 gaseous ethylene was insuff lated into an airtight transparent box containing the f lowers. For 1-MCP and ethylene treatments, the corollas were initially exposed to 50 nL·L −1 1-MCP for 3 h prior to the application of ethylene. For other non-gaseous hormones, the f lowers were inserted into the vials containing 50 μM GA 3 , 50 μM ABA, and 100 μM MeJA. For the combined treatments with their respective inhibitors, 20 μM PAC, 20 μM FLD, and 40 μM SHAM were firstly applied to the f lowers for 3 h, respectively. The corollas were harvested at intervals (0, 3, 6, 12, and 24 h) post treatment. For the combined treatment with ethylene and GA 3 , the f lowers inserted into the vials containing 50 μM GA 3 were simultaneously exposed to 10 μL·L −1 gaseous ethylene. To examine the effects of hormone treatments on petal senescence, the f lowers were treated with ethylene, GA 3 , and a combination of the two hormones as mentioned above, and photographed at 0 and 6 days after treatments. To check the impact of GA treatment on PhOBF1-mediated f lower senescence, the corollas from WT and PhOBF1 transgenic lines at D0 were sprayed with 50 μM GA 3 or 20 μM PAC before GA 3 application. The f lowers were harvested at D6 after anthesis. The treatment with deionized water was used as mock control.

Generation of transgenic petunia plants
The complete open reading frame (ORF) region of PhOBF1 was amplified and ligated into the pGSA1403 vector to generate the overexpression construct, while a 329-bp fragment in both forward and reverse orientations was introduced into the pGSA1285 vector to form the RNAi construct as previously described [38]. Electroporation was used to transform the recombinant plasmids into Agrobacterium tumefaciens strain LBA4404 using a Gene Pulser apparatus (Bio-Rad, Richmond, CA, USA) at 2.5 kV and 400 Ω. Agrobacterium-mediated leaf-disk method was used for stable genetic transformation according to a previously described protocol [86]. The positive transgenic plants with kanamycin resistance were screened on the MS plates. PCR amplification was performed to confirm the integration of PhOBF1 cDNA fragment into the petunia genome. Expression levels of PhOBF1 in the f lowers of WT and transgenic petunia plants at D0 were examined by quantitative real-time PCR. Three representative lines based on transcript levels of PhOBF1 in the RNAi or overexpression assay were selected for further experiments.

Semi-quantitative RT-PCR and quantitative real-time PCR
Total RNA isolation was conducted on petunia f lowers through a TRIzol-based method. For RNA purification, RNase-free DNase I (Promega, Madison, WI, USA) was added to eliminate contaminating DNA at 37 • C for 30 min. About 2-5 μg of total RNA were used for reverse transcription to synthesize cDNA samples using a PrimeScript RT reagent kit (Takara, Otsu, Shiga, Japan). Semi-quantitative RT-PCR was performed with Taq DNA polymerase (TaKaRa, Otsu, Shiga, Japan) in 25 μL volumes following the manufacturer's instructions. The resulting products were analysed by 1.5% agarose gel electrophoresis using GelRed as a nucleic acid staining dye. The gene bands were visualized and photographed under ultraviolet light with a Gel Doc XR+ imaging system (Bio-Rad, Hercules, CA, USA). Abundance of mRNA was detected by quantitative real-time PCR. The reactions were run in a LightCycler instrument (Roche Diagnostic, Basel, Switzerland) using the SYBR Green Master Mix (Applied Biosystems, Foster City, CA, USA). Transcript levels of genes were standardized to the constitutively expressed 26S rRNA. Analysis of relative expression data was performed based on the comparison of threshold cycles in PCR reactions [87]. Three independent RNA samples were used for expression analysis.

Measurement of gibberellins and ethylene
GA levels were measured according to a previously described protocol [88]. In brief, the f loral samples were frozen and powdered using a milling tool in the presence of liquid nitrogen. Extraction was conducted using 80% methanol with internal standards GA 1 and GA 3 at 4 • C for 30 min. The aqueous phase was collected and further extracted with the solution containing EtOAc (pH = 3.0) and K-Pi buffer (pH = 8.5). After purification by chromatography on Sep-Pak C18 column, the eluant was evaporated to dryness with nitrogen at 40 • C and re-dissolved in 80% methanol. After purification through a 0.45 μm filter, the solution was detected by HPLC electrospray ionization tandem mass spectrometry (HPLC-ESI-MS/MS). The production of gaseous ethylene was measured as previously described [13]. The f lowers with the pedicels were harvested and inserted into the vials with sterile water. They were placed into a 250-mL container and maintained at 25 • C for 3 h. A 2-mL hypodermic syringe was used for injecting the samples into a gas chromatographic apparatus (Agilent, Palo Alto, CA, USA) for quantification of ethylene release. Three biological replicates were used with at least five f lowers in each replicate.

Electrophoretic mobility shift assay
EMSA experiment was conducted using a previously reported method [89]. The ORF sequence of PhOBF1 was introduced into the pET28a vector to form the pET28a-PhOBF1 plasmid, which was then transformed into Escherichia coli Rosetta (DE3) cells. Next, 0.1 mM isopropylthio-β-galactoside was added to induce the expression of His-tagged PhOBF1 protein. An ultrasonic device was applied to disrupt the bacterial cells for protein release. The protein samples were extracted and purified by passing through the HisTrap HP column. A 35-bp biotin-labeled DNA fragment in the promoter of PhGA20ox3 was synthesized and annealed, which was used as WT or mutant probe. The WT probe with no labeling was referred to as the binding competitor (Table S1, see online supplementary material). A LightShift EMSA Optimization and Control Kit (Pierce, Thermo Fisher Scientific, MA, USA) was used for the interaction of protein and DNA. The protein-DNA complex was separated by 6% non-denaturing polyacrylamide gel electrophoresis, and then transferred to a 0.45 μm nylon membrane using an electrophoretic transfer apparatus (Bio-Rad, Hercules, CA, USA). After conjugation and purification, the binding signals on the membrane were visualized and photographed in the imaging system as mentioned above.

Yeast one-hybrid assay
A 493-bp DNA fragment harboring PhOBF1-bound cis-acting element in the PhGA20ox3 promoter was amplified using a primer pair (Table S1, see online supplementary material). The product was cloned into the pHIS2 vector for the bait construction. The coding sequence of PhOBF1 with full length was introduced into the pGADT7-Rec vector harboring the GAL4 activation domain, which was regarded as the pray construct. The recombinant plasmids were co-transformed into Saccharomyces cerevisiae strain Y187. The positive colonies were selected and identified for liquid culture, which was diluted with LB media to 10 and 100 times. For each dilution, a small amount of cells were spotted on the SD-Ura-His-Leu plates by addition of 100 mM 3-AT. The activation of PhGA20ox3 promoter by PhOBF1 was verified based on the proliferation rates of yeast cells.

Dual luciferase assay
The assay was carried out according to a previously described method [90]. The complete coding sequence of PhOBF1 and 1500-bp sequence of PhGA20ox3 promoter were jointed with the pGreenII62-SK and pGreenII0800-LUC vectors, respectively. For the effector, the CaMV 35S promoter was used to drive the expression of PhOBF1. For the reporter, PhGA20ox3 promoter and 35S promoter were employed to activate the transcription of LUC and Renilla luciferase (REN), respectively. The specific primers used are listed in Table S1 (see online supplementary material). The cells of A. tumefaciens strain GV3101 containing the constructed plasmids were used to co-inoculate young petunia leaves. The enzymatic activities of both LUC and REN were examined through a luminometer (Männedorf, Switzerland), and expressed as the LUC/REN ratio.

Virus-induced gene silencing
The TRV-PhCHS plasmid containing a 194-bp fragment of PhCHS was constructed as previously described [91]. To obtain the TRV-PhCHS/GA20ox3 construct, a PhGA20ox3 fragment with 244 bp in length was inserted into the SacI-XhoI sites of TRV-PhCHS vector. The plasmids were electro-transformed into A. tumefaciens strain GV3101. Agrobacterium cells with kanamycin resistance were cultured in 15-mL LB media by addition of 40 mg·L −1 kanamycin and other reagents at 28 • C for 48 h. The cultures were centrifuged at 4000 rpm for 15 min to harvest the cells. The reagents containing 200 μM acetosyringone, 10 mM MES, and 10 mM MgCl 2 were used to resuspend the harvested cells, whose concentration was then adjusted to an OD 600 of 4.0. After 3 h of gentle shaking at room temperature, the mixture of TRV1-and TRV2-transformed bacteria was used to inoculate the leaves of petunia plantlets as previously described [38]. Three biological replicates were used with five plantlets in each replicate.