Gibberellins regulate the transcription of the continuous flowering regulator, RoKSN, a rose TFL1 homologue.

The role of gibberellins (GAs) during floral induction has been widely studied in the annual plant Arabidopsis thaliana. Less is known about this control in perennials. It is thought that GA is a major regulator of flowering in rose. In spring, low GA content may be necessary for floral initiation. GA inhibited flowering in once-flowering roses, whereas GA did not block blooming in continuous-flowering roses. Recently, RoKSN, a homologue of TFL1, was shown to control continuous flowering. The loss of RoKSN function led to continuous flowering behaviour. The objective of this study was to understand the molecular control of flowering by GA and the involvement of RoKSN in this inhibition. In once-flowering rose, the exogenous application of GA(3) in spring inhibited floral initiation. Application of GA(3) during a short period of 1 month, corresponding to the floral transition, was sufficient to inhibit flowering. At the molecular level, RoKSN transcripts were accumulated after GA(3) treatment. In spring, this accumulation is correlated with floral inhibition. Other floral genes such as RoFT, RoSOC1, and RoAP1 were repressed in a RoKSN-dependent pathway, whereas RoLFY and RoFD repression was RoKSN independent. The RoKSN promoter contained GA-responsive cis-elements, whose deletion suppressed the response to GA in a heterologous system. In summer, once-flowering roses did not flower even after exogenous application of a GA synthesis inhibitor that failed to repress RoKSN. A model is presented for the GA inhibition of flowering in spring mediated by the induction of RoKSN. In summer, factors other than GA may control RoKSN.

Introduction

Gibberellins (GAs), important plant growth regulators, are involved in different aspects of plant development such as growth promotion, seed dormancy, floral initiation, and development (Fleet and Sun, 2005). GAs are cyclic diterpenoid molecules, with GA1, GA3, GA4, and GA7 being the most biologically active in plants. The floral transition from vegetative growth to flowering is an important period in the life of a plant. In long-day (LD) monocarpic plants, it was reported that GA is a floral stimulus and could be substituted for LDs (Lang, 1965) whereas, in most woody plants, GA inhibits flowering (Zeevaart, 1976). For example, GA activates flowering in Arabidopsis thaliana, whereas it represses flowering in woody fruit trees (Wilkie ).

The molecular basis of floral transition and, particularly, the response to GA, was elucidated in the model plant A. thaliana, a facultative LD vernalization-mediated plant. For this transition to be successful, Arabidopsis must integrate different endogenous and environmental signals (reviewed by Srikanth and Smith, 2011). Flowering is controlled by four major genetic pathways: photoperiod, autonomous, GA, and vernalization. These different pathways converge to activate the floral integrators, FLOWERING LOCUS T (FT) and SUPPRESSOR OF CONSTANS1 (SOC1), and to repress FLOWERING LOCUS C (FLC), a floral repressor involved in the vernalization response (Kim ). The floral integrators then activate the meristem identity genes, LEAFY (LFY) and APETALA1 (AP1), leading to activation of the floral organ identity genes known as ABCE model genes.

In Arabidopsis, GA was required for flowering under short days (SDs) and LDs, with a weaker effect under LDs. ga1-3, a GA-deficient mutant due to a mutation in a key GA metabolism gene (Koornneef and van der Veen, 1980), led to a delay in flowering under LDs and was unable to flower under SDs. This phenotype was reversed by exogenous application of GA4 (Wilson ). Under SDs, GA4 was accumulated in the shoot apical meristem (SAM) prior to floral transition. GA4 may be synthesized in the leaves and then move to the apex (Eriksson ).

GAs induce flowering through two different pathways. In leaves, the GA response was FT dependent. Under LDs, GA activated CONSTANS (CO) that in turn induced FT (Hisamatsu and King, 2008). In the apex, GA acted in an FT-independent manner to induce SOC1 and LFY (Hisamatsu and King, 2008). Failure of the ga1-3 mutant to flower under SDs was due to the absence of up-regulation of the floral integrator SOC1 (Moon ) and the meristem identity gene LFY (Blazquez ). In ga1-3 mutants, SOC1 was accumulated after exogenous GA4 application under SDs (Moon ). GA4 regulation of SOC1 expression could be mediated by the down-regulation of SHORT VEGETATIVE PHASE (SVP), a floral repressor involved in the vernalization and ambient temperature pathways (Li ). The GA effect on LFY expression is mediated by a transcription factor, AtMYB33, which is able to link the GAMYB-binding site in the LFY promoter (Gocal ). AtMYB33 expression corresponds to GA4 accumulation in the shoot apex during floral transition (Gocal ; Eriksson ).

Except for Arabidopsis, few data were available about the molecular control of flowering by GAs in other species. In Lolium temulentum, an obligate LD plant, GA activated flowering (Evans ). The leaf GA content increased after LD exposure with transcriptional regulation of GA biosynthesis genes (Gocal ). GA moves from leaves to the shoot apex where the increase in GA content corresponds to floral induction (King ). LD treatment can be mimicked by GA application, especially GA5 and GA6 (Evans ; King ). Lolium perenne, a perennial near-relative of L. temulentum, requires both LDs and vernalization for flowering. LDs activate GA biosynthesis, which then stimulates the flowering of vernalized plants, whereas non-vernalized plants are unable to respond to GA for flowering (MacMillan ). In grapevine, the vvGAI mutant, a gene encoding a DELLA protein, a central repressor of GA signalling, was dwarf, GA-insensitive, and produced inflorescences along the length of the shoot instead of tendrils. GA represses the formation of inflorescences and promotes the initiation of tendrils (Boss and Thomas, 2002).

Rose is an economically important ornamental crop worldwide. Flowers represent the major value of roses, and control of flowering is a major issue. Rose is a perennial woody plant with different modes of flowering. Once-flowering (OF) roses have a single annual flowering period, whereas continuous-flowering (CF) roses have the ability to flower continuously during the season. In rose, the sensitivity of plants to GA was different between OF and CF roses. Exogenous application of GA inhibited flowering in OF roses and had no effect on flowering in CF roses (Roberts ). GA signalling genes such as GID1, a GA receptor, are regulated differently between OF and CF roses (Remay ). Furthermore, GA metabolism was involved in floral initiation. In OF and CF roses, the amount of GA in shoots before floral initiation was low and increased after floral initiation (Roberts ). This GA content was correlated with GA metabolism gene regulation during floral transition. In OF and CF roses, the RoGA20ox gene, encoding an enzyme involved in GA synthesis, was transiently repressed before floral initiation. In contrast, the expression of RoGA2ox, encoding an enzyme involved in GA inactivation, increased before floral initiation and decreased later (Remay ). Recently, the gene encoding the continuous flowering locus was shown to correspond to a floral repressor homologue of TFL1, TERMINAL FLOWER1, and was designated as RoKSN (Iwata ). CF roses had a retrotransposon inserted in the second intron of RoKSN. The presence of a retrotransposon blocked the transcription of RoKSN in CF roses, allowing continuous blooming. In OF roses, RoKSN is weakly accumulated in spring to permit blooming. Later in the season, RoKSN transcripts accumulate and shoots remain vegetative (Iwata ).

The objectives of this study were to understand the molecular control of flowering by GA in rose and to decipher the link between the continuous flowering gene, RoKSN, a TFL1 homologue, and GA. In this study, it was demonstrated that GA inhibited floral initiation in rose by up-regulating the CF gene at the transcriptional level. The promoter of RoKSN was further analysed, and GA-responsive elements were found.

Materials and methods

Plant material

Rosa×wichurana (RW), an OF rose, was obtained from the rose garden, Jardin de Bagatelle (Paris, France). The mutant pair, R. hybrida ‘Félicité et Perpétue’ (FP), an OF rose, and ‘Little White Pet’ (LWP), its CF mutant, were obtained from ‘Loubert Nursery’ (Rosier sur Loire, France). Plants were cultivated as described (Remay ). Nicotiana benthamiana were maintained in a greenhouse (20 °C, 16h light), and 6- to 8-week-old plants were used for agroinfiltration.

GA and paclobutrazol treatments on plants

In December, stems were pruned at six internodes and maintained outside for vernalization (Fig. 1A). At the beginning of the treatment (end of January), plants were placed in a tunnel. Different GA3 concentrations (0, 30, 70, and 140 µM) were sprayed on plants three times a week during different periods and for different durations (Fig. 1B). Paclobutrazol (PCB) treatment (Bonzi®, at 42, 70, or 112 µM) was performed every 2 weeks. After blooming, a new set of plants was used for GA3 and PCB treatments. The inflorescences were removed and growing shoots were pruned at six nodes. Plants, maintained in a greenhouse, were sprayed with GA3 (30 µM) three times a week for 34 d and with PCB (Bonzi®, at 14, 56, or 112 µM) three times during the experiment [0, 15, and 29 days after the beginning of treatment (DAB)].

Fig. 1.

Management of R.×wichurana (RW) after GA3 and PCB treatments. (A) In December, vegetative shoots were pruned at six nodes. Plants remained outside for vernalization. At the end of January, plants were transferred to a tunnel. New shoots developed from axillary buds of the previous years. These shoots were mainly terminated by an inflorescence. After floral development, new shoots grew from the base. Since RW is an OF rose, the shoots remained vegetative. In summer, these shoots were pruned at six nodes and the fate of new emerging shoots from axillary buds was studied. Under controlled conditions, these shoots remained vegetative. RW plants were treated with GA3 (0, 30, 70, or 140 µM) during vegetative growth and floral initiation between January and March (hatched boxes). Another set of RWs was treated with GA3 (30 µM) and PCB (14, 56, and 112 µM) during vegetative growth between August and September. Arrows and asterisks indicate the sampling carried out for RNA extraction and histological analysis, respectively. (B) Experiments with different durations and periods of GA3 (30 µM) treatment. In the ‘Late treatment’, the experiment started at different times and ended at the same time, whereas in the ‘Early treatment’, the experiment started at the same time and ended at different times. Numbers indicate the days after the beginning of treatment (DAB). Broken lines represent the pruned shoots of the previous years, large arrows represent new indeterminate growing shoots, and black circles represent axillary buds. Open circles represent flowers. (This figure is available in colour at JXB online.)

Phenotypic observations

Phenotypic observations were carried out on 10 plants with approximately five pruned stems, 2 months after the end of the treatment. Data were gathered from the three new emerging shoots placed on the distal part of the pruned stem (Fig. 1A). For each new emerging shoot, the percentage of flowering that represents the ratio between the number of flowering shoots and the number of shoots was determined. To calculate the average internode size, three consecutive internodes placed on the apical part of the stem were measured. Statistical analyses were done using the R software, version 2.13.1.

Histological observations

After dissection (removal of leaves), samples were fixed and sections of 2–3 µm were observed under a microscope (Foucher ). Observations were made on four different dates (42, 56, 63, and 70 DAB).

RNA isolation and gene expression analysis

For RNA isolation, the terminal part of the shoot was harvested from five plants per date and per treatment (Fig. 1A). Samples were dissected (removal of expanded leaves and young leaves) and frozen in liquid nitrogen. Total RNA was isolated using the NucleoSpin® RNA plant kit (Macherey-Nagel) according to the manufacturer’s instructions. Elimination of genomic DNA and reverse transcription were performed using 1mg of total RNA; real-time PCRs were performed using 3 µl of reverse transcription product (1/25 dilution), as described by Remay . The amount of plant RNA in each sample was normalized using TCTP and UBC genes (Klie and Debener, 2011), and the relative expression level was calculated according to Pfaffl (2001), from two biological replicates and three technical repetitions per replicate.

DNA constructs

The upstream genomic sequence of RoKSN (promoter region) was obtained using the Genome Walker Universal Kit (Clontech), following the manufacturer’s recommendations from genomic DNA libraries of R. hybrida ‘Knock Out’. Serial nested PCRs were performed using the described reverse primers (Supplementary Table S1 available at JXB online) and the forward primers of the kit. A 1200bp fragment was amplified and sequenced. Then, using Finnzymes’ Phusion® High fidelity DNA Polymerase (Thermo Fisher Scientific), the promoter region was amplified from DNA of Rosa chinensis ‘Old Blush’ and cloned into the pGEM T easy vector (Promega). Next, successive promoter deletions were performed to obtain fragments of different sizes: 200, 400, 800, and 1200bp. PCR products were cloned using the Gateway® system (Invitrogen) with TOPO isomerase mix into the pENTR™/D-TOPO® entry, followed by LR clonase recombination with pKGWFS7 (VIB, Gent, Belgium) (Karimi ). The promoter sequence was fused to a GUS::GFP (β-glucuronidase::green fluorescent protein)-coding gene. Constructions were introduced by electroporation in Agrobacterium tumefaciens EHA105 with the plasmid pbbR.

Transient expression procedure

Agrobacterium tumefaciens C58C1 containing p19 (Voinnet ) and EHA105 containing the different constructions were grown at 28 °C in LB medium supplemented with 50mg l–1 kanamycin, 50mg l–1 rifampicin, and 100mg l–1 gentamicin (only for EHA105) up to the stationary phase. After centrifugation at 5000 g for 10min, bacteria were resuspended in Murashige and Skoog (MS) medium supplemented with 5% (w/v) sucrose, acetosyringone (200 µM), pH 5.6, and with or without GA3 (10 µM) to obtain an optical density (OD600) of 0.4. Two days after agroinfiltration of 6- to 8-week-old N. benthamiana leaves, GFP fluorescence was observed using a Nikon A1 laser scanning confocal microscope.

Sequence analysis

Sequences of promoters of RoKSN (HE863824), MdTFL1, MdTFL1a (AB366639 and AB36640), FvKSN (HQ378595), PpTFL1 (AB636121), and TFL1 (At5g03837) were used to perform the alignment with MVISTA online software (http://genome.lbl.gov/vista/mvista/submit.shtml) (Frazer ). The search for putative binding sites of cis-elements was then carried out using Plantpan online software (http://plantpan.mbc.nctu.edu.tw/seq_analysis.php) (Chang ).

GUS activity

To measure GUS activity quantitatively, MUG assays were conducted on each agroinfiltrated leaf (1–2cm2) as described by Jefferson . The reaction was stopped after 5, 10, 15, 20, and 30min. Fluorescence was measured at these different times on a FLUO-star plate reader (BMG Labtechnologies Inc.) at 460nm when excited at 355nm. Values from the fluorescence assay were standardized by protein concentration, determined as described by Bradford (1976).

Results

Gibberellins block floral initiation in spring in once-flowering rose

To study the effect of GA treatment on plant development, exogenous application of GA3 at different concentrations (30, 70, and 140 µM) was performed in spring on the OF roses, RW and FP, and its CF mutant LWP (Fig. 1).

On non-treated RW, 80% of the emerging shoots became floral, whereas for the GA3 treatment (30 µM and 70 µM), almost all new emerging shoots remained vegetative (only 5.5% and 1%, respectively, were floral). At higher GA3 concentrations, flowering was completely inhibited (Fig. 2B). GA3 treatment also had an effect on the vegetative development of the stem. The length of internodes increased significantly: without treatment, internodes measured 3cm, at 30 µM GA3 they measured 4.3cm, and they measured 4.5cm for the higher concentrations of GA3 (Fig. 2A). The same inhibitory effect on blooming was shown on FP (Supplementary Table S2 at JXB online). Flowering was not inhibited by GA3 treatment in LWP, whereas internode length was reduced. The effect of the GA synthesis inhibitor PCB was tested in FP and LWP. PCB treatment significantly reduced internode length. Concerning flowering, PCB treatment has no effect on LWP, whereas at high concentrations (70 µM and 110 µM), flowering was significantly reduced in FP. Only 50% of the shoots became floral at 110 µM instead of 79% for untreated plants (Supplementary Table S2).

Fig. 2.

Effect of GA3 treatment on average internode size in centimetres (A) and percentage of shoots that become floral (B) on R.×wichurana (RW). RW plants were treated with GA3 (0, 30, 70, or 140 µM) for 70 d, and with GA3 (30 µM) for different durations and periods. Data were the means of 10 plant observations. There was a statistically significant difference (Waerden’s test, P < 0.05; and the χ2 test, P < 0.0033, for the comparison of percentages) between means with different letters.

To evaluate the effect of GA3 treatment on the fate of the apex, cross-sections of the apex from the vegetative to the floral stage were taken using the OF rose, RW, treated or not with 30 µM GA3 (Fig. 3). In untreated plants after bud outgrowth (42 DAB), the vegetative meristem was narrow with leaf primordia on its flanks (Fig. 3A). The meristem then emerged and enlarged (56 DAB; Fig. 3C). This stage has been defined as the ‘floral initiation’ stage. The meristem then became floral and bracts were visible (63 DAB; Fig. 3E). Later, floral organs (sepals, petals, stamens, and pistils) are visible (70 DAB, Fig. 3G). Under untreated conditions, the floral initiation took place 56 DAB. In GA3-treated plants (Fig. 3B, 3D, 3F, 3H), the meristem was narrow and dome-shaped for all stages. Leaf primordia were visible on the flank of the meristem. This structure was typical of a vegetative meristem. These observations clearly demonstrated that GA3 blocked floral initiation and that the meristem remained vegetative.

Fig. 3.

Longitudinal cross-section through apices treated with GA3 (30 µM; B, D, F, and H) or not (A, C, E, and G). In the untreated plant, the vegetative meristem (A: 42 DAB) progressively changed into a floral meristem (C, E, and G; 56, 63, and 70 DAB, respectively), whereas in the GA3 (30µM)-treated plants, the meristem remained vegetative for the same stage of development (B, D, F, and H, 42, 56, 63, and 70 DAB, respectively). VM, vegetative meristem; LP, leaf primordia; FM, floral meristem; S, sepal; FB, floral bud; P, petal; St, stamen; Pi, pistil. Scale bar=100 µm. (This figure is available in colour at JXB online.)

GA3 is necessary during a short period after bud outgrowth to inhibit flowering.

To characterize the GA response, different durations and periods of GA3 treatment (30 µM) were tested before blooming in spring. To modify the duration and the period of the treatment, GA3 application was started and ended on different dates, the ‘late’ and ‘early’ treatments, respectively (Fig. 1B). The duration varied from 20 d to 70 d, and the period was from late January to the end of March.

The length of internodes significantly increased between untreated plants (3cm) and plants treated with GA3 (from 3.5cm to 4.6cm), except for the GA3 treatment of 20 d (2.3cm) (Fig. 2A). In general, the longer the treatment, the longer the length of the internodes.

Concerning flowering behaviour, the longer the treatment, the higher the GA inhibiting effect. For GA-treated roses, 5.5, 14.8, 25.5, and 40.5% of the shoots became floral for 70, 63, 47, and 33 d, respectively, of treatment begun on different dates (Fig. 2B). For treatment terminated on different dates (‘early’ treatment), the same trend was observed: 80% without treatment; 16.5, 3.6, and 3% for 20, 29, and 48 d, respectively. It is interesting to note that for the same duration of treatment, differences are observed if treatments are not carried out in the same period. For 47/48 d of treatment, the percentage of floral shoots was 3% for ‘early’ treatment, whereas it was 25.5% for ‘late’ treatment. For 33/29 d of treatment, the percentage of floral shoots was 3.6% for ‘early’ treatment, whereas it was 40.6% for ‘late’ treatment (Fig. 2B). ‘Early’ treatments were more effective at inhibiting flowering than ‘late’ treatment. For example, an ‘early’ treatment of 20 d had the same efficiency as a ‘late’ treatment of 47 d (15% blooming, Fig. 2B). In the ‘early’ and ‘late’ treatments, the non-floral shoots remain vegetative and present an indeterminate growth. GA3 blocked the floral initiation and the meristem remained vegetative. In conclusion, a 1 month treatment in February (during floral transition) is sufficient to inhibit floral transition almost completely.

Floral gene expression in response to GA3 treatment

To examine whether the inhibition of flowering observed after GA3 treatment could be associated with the modification of floral gene transcript accumulation, this accumulation in RW shoots that emerged in spring and after blooming using was analysed by qPCR (Figs 1, 4). In spring, in the absence of treatment, the transcripts of the floral genes, RoFT, RoFD, RoLFY, and RoSOC1, were accumulated in shoots, which then become floral. RoFT transcripts were progressively accumulated in two steps as previously described (Remay ). The first accumulation occurred early during the floral transition (15–35 DAB; Fig. 4C), and the second occurred after floral initiation. RoFD was transiently accumulated early before floral initiation (15 DAB; Fig. 4B). The accumulation of RoSOC1 transcripts was progressive during floral transition and occurred before the late accumulation of the RoLFY and RoAP1 transcripts (70 DAB; Fig. 4). The accumulation of RoKSN transcripts was low in spring (Fig. 4A).

Fig. 4.

Transcript accumulation of floral genes in spring and later in summer in GA3-treated (squares), PCB-treated (triangles), or untreated RW (diamonds). Transcript accumulation of floral genes was followed by qPCR in axillary shoots for (A) RoKSN, (B) RoFD, (C) RoFT, (D) RoLFY, (E) RoSOC1, and (F) RoAP1. The x-axis indicates the number of days at which apices were harvested after the beginning of the GA3 or PCB treatment. The transcript accumulation levels are expressed in relation to the first sample, harvested in January, for each gene according to the Pfaffl ratio (Pfaffl, 2001) (base value=1);. The grey box represents the floral initiation determined according to the histological analysis (Fig. 2). Kruskall–Wallis test on two biological replicates (P < 0.05) was realized, and asterisks represent statistical differences between GA3-treated and untreated plants; triangles represent statistical differences between PCB-treated and untreated plants. (G) Transcript accumulation of floral genes in GA3-treated (grey box), and untreated (black box) LWP. The transcript accumulation levels are expressed in relation to the untreated plants. Data with different letters indicate a significant difference (Kruskall–Wallis test P < 0.05). nd, not detected

When GA3 was exogenously applied at 30 µM, the expression patterns of floral genes were modified. GA3 treatment strongly increased the accumulation of RoKSN transcripts. After 70 d of treatment, the RoKSN transcript level increased 40-fold in treated plants (Fig. 4A). RoFT transcript accumulation followed the same expression pattern as in untreated plants, but the accumulation was lower in treated plants than in untreated plants (Fig. 4C). The transient expression of RoFD (15 DAB) was not detected in treated plants. RoSOC1 presented a transient accumulation similar to that of untreated plants, except after 70 DAB when a 2-fold transcript decrease was observed (Fig. 4E). There was no accumulation of RoLFY and RoAP1 transcripts at the end of the experiment.

In FP, RoKSN transcripts were accumulated in response to GA3 after a few hours. RoLFY transcripts were reduced 2-fold and GA3 treatment had no effect on RoFT (Supplementary Table S3 at JXB online). To test whether floral gene transcription was RoKSN dependent, the expression of these floral genes was studied in the CF rose, LWP. After GA3 treatment, the expression of RoFT, RoSOC1, and RoAP1 was not affected, whereas expression of RoLFY and RoFD was reduced 2-fold (Fig. 4G). No RoKSN transcript could be detected.

GA3 and PCB treatment do not modify the fate of indeterminate vegetative shoots after blooming.

In OF roses, new emerging shoots were vegetative after the blooming in spring (Fig. 1A), whereas in CF roses new shoots were always terminated by an inflorescence. Since GA3 inhibits flowering in spring, it was investigated whether a GA synthesis inhibitor, PCB, could allow new blooming later in the season in RW (Fig. 1A). Treatments with PCB were performed at different concentrations (14, 56, and 112 µM), and with GA3 (30 µM) (Table 1). At 30 µM GA3, a significant increase in internode length (2.5cm) was observed compared with untreated plants (2cm). In contrast to the GA3 treatment, PCB treatment significantly reduced the internode length at the three concentrations tested in RW and LWP. In RW, the length of internodes was 2cm for untreated plants and 1.6, 1.5, and 1.4cm for plants treated with PCB at 14, 56, and 112 µM, respectively. The same trend was observed in LWP (Table 1). Neither of the treatments (GA3 or PCB) had an effect on flowering. Under all conditions, in RW, the shoots remained vegetative; no flowers were observed. In LWP, all the shoots were floral (Table 1).

Table 1.

Effect of GA3 and the gibberellin inhibitor (PCB) on floral development and vegetative growth in late summer of OF rose (R.×wichurana, RW) and CF rose (‘Little White Pet’, LWP).

Genotype		Treatment		Second treatment (August–September)
				Untreated		GA3 30 µM		PCB 14 µM		PCB 56 µM		PCB 112 µM
RW		Average internode size (cm)		2±0.5 b		2.5±0.7 a		1.6±0.5 c		1.5±0.8c		1.4±0.4c
		Percentage of shoots that become floral		0		0		0		0		0
LWP		Average internode size (cm)		2.3±0.7 b		2.6±0.5 a		2.1±0.5 b		1.7±0.5c		1.5±0.5d
		Percentage of shoots that become floral		99.2 a		99.1 a		99.1 a		100 a		100 a

Data were the means of five plant observations. There was a statistically significant difference (Waerden’s test; P < 0.05) between means with different letters. The experimental design is described in Fig. 1.

In the absence of treatment, transcript accumulation of RoFT, RoSOC1, RoFD, RoAP1, and RoLFY remained low, and RoKSN transcripts were accumulated (Fig. 4). After exogenous application of GA3, RoKSN expression increased 3-fold. GA3 treatment had no effect on RoFD and RoFT (Fig 4B). RoLFY transcript decreased 3-fold after 34 d of GA3 treatment (Fig. 4D), while RoAP1 was accumulated later (34 DAB). No significant changes were detected after PCB (at 56 µM) treatment, except for RoFD and RoFT, which were transiently induced at 14 DAB (Fig. 4B).

RoKSN promoter contains GA-responsive cis-elements

The induction of RoKSN by GA was further investigated by studying the upstream genomic region of RoKSN. A 1200bp sequence of RoKSN (upstream of the ATG codon) was isolated. By comparing promoters of TFL1 of other Rosaceae (Malus domestica, Prunus persica, and Fragaria vesca) and Arabidopsis thaliana, two conserved regions were detected (Fig. 5A). Block A (–350bp to –281bp) was specific to the Rosoideae subfamily, whereas block B (–177bp to –61bp) was common to the Rosaceae family. One GAMYB cis-element, known to be involved in the GA response (Gocal ), was identified in block A. No GA response elements were found in block B. The non-conserved regions of the promoter contained three PIF3, three GAMYB, 10 WRKY71, one CARE, and one GARE element (Supplementary Fig. S1 at JXB online)

Fig. 5.

Analysis of GA-responsive elements in the RoKSN promoter. (A) Alignment of the four TFL1 homologue promoters using MVISTA. Conserved regions (>70% similarity on a 100bp window) are in red and surrounded by dotted lines (blocks A and B). Alignment was performed with TFL1 (Arabidopis), FvKSN (F. vesca), MdTFL1 (M. domestica), MdTFL1a (M. domestica), and PpTFL1 (P. persica). (B) Successive promoter deletions fused with a GFP::GUS gene. (C) GFP fluorescence and GUS activity. GFP fluorescence observed under a confocal microscope for the four different promoter deletions. Untreated and treated plants are in the left and right panel, respectively. GUS activity for the different deletions in the presence (grey) or absence (black) of GA3 (10 µM) by transient expression in tobacco leaves after Agrobacterium infiltration. Data are expressed relative to the minimal untreated promoter (p200). Data with different letters indicate a significant difference (Kruskall–Wallis test P < 0.05). (This figure is available in colour at JXB online.)

To determine the functional region for GA response expression, successive promoter deletions were performed: p200, p400, p800, and p1200, which corresponded to the –200, –400, –800, and –1200bp of the upstream sequence of RoKSN, respectively. No fluorescence or GUS activity was detected with the p200 construct that contains only a TATA box element (Fig. 5C). Weak expression was detected with the p400 promoter in the absence of GA. In the presence of GA, the promoter activity was significantly increased by 14-fold (Fig. 5C). With the p800 and p1200 constructs, the promoter activity was not significantly different between GA-treated and non-treated plants. In both conditions, GUS activity was high (Fig. 5C). The 800bp fragment may contain a positive cis-element, which provokes a constitutive expression of the promoter under the conditions used here (heterologous system). Due to this constitutive expression, GA induction may no longer be detectable.

Discussion

Physiological studies have demonstrated that GAs are floral repressors in woody perennials such as fruit trees (Wilkie ). However, how GA mediates its inhibitory effect at the molecular level remains largely unknown. In this study, the role of GAs in flowering on OF roses was investigated at the physiological and molecular level. It was demonstrated that the inhibitory effect of GA in rose is principally mediated by a floral repressor homologue to TFL1, RoKSN.

GA treatment modified the phenology of flowering in once-flowering rose

Under the experimental conditions without GA treatment, the OF roses (RW and FP) initiated their blooming in mid-March. The histological analysis showed that the vegetative meristem was progressively converted into a floral meristem (Fig. 3). As previously described by Roberts , GA3 treatment inhibited flowering in OF roses. High levels of GA3 (>30 µM) completely blocked flowering (Fig. 2). The histological analysis showed that GA3 treatment inhibited floral initiation. After GA3 treatment, the apex remained vegetative and none of the morphological modifications observed during floral initiation (such as enlargement and doming of the meristem) was observed (Fig. 3). To specify the effect of GA3, different times and durations of GA3 application were used. A 1 month treatment before floral initiation (in February) was sufficient to obtain a strong inhibition of flowering (Fig. 2A). This period was referred to as floral transition, the phase when the meristem is still vegetative but receptive to signals that stimulate or inhibit flowering. During this phase, GA3 acts as a floral repressor. This is consistent with the role of GA, which is known to promote vegetative growth instead of reproductive development in woody plants. The exogenous application of GA in fruit trees blocked the floral process and partially reduced the intensity of flowering, for example in apple (Tromp, 1982; Bertelsen ) or in avocado (Salazar-Garcia and Lovatt, 2000). In peach, this inhibitory effect has been linked to the date of treatment (Painter and Stembridge, 1972). In strawberry, GA treatment induced the initiation of runners and repressed the formation of crown branches (Hytönen ). None of these studies explained how GA mediates its inhibitory effect on flowering.

A floral repressor, RoKSN, is transcriptionally regulated by GAs

To look for the molecular targets of GA during the rose floral process, molecular analyses were carried out in spring and after blooming in GA3-treated and untreated OF rose. During floral transition in untreated plants, floral activators were progressively accumulated (RoFT and RoSOC1). Later, RoLFY and RoAP1 were accumulated at the end of floral initiation when the meristem started to become floral (Figs 3, 4). RoKSN transcripts remained low during floral transition and initiation. Similar patterns of transcript accumulation have already been observed in FP, an OF rose (Remay ; Iwata ), and in other Rosaceae. In apple, where floral initiation takes place in June after blooming, expression patterns of floral activators (homologues of SOC1, FT, LFY, and AP1) and repressors (homologues of TFL1) are similar (Hättasch ).

When GA3 was exogenously applied, the strongest effect was the large accumulation of RoKSN transcripts and the absence of the accumulation of RoLFY and RoAP1 transcripts (Fig. 4). In RW, RoKSN was accumulated 20 times more in GA3-treated plants than in untreated plants in spring (Fig. 4). The same pattern was observed in summer. Furthermore, RoKSN was induced by GA3 in two independent OF roses, RW and FP, suggesting a conservation of this regulation in rose (Supplementary Table S3 at JXB online).

To analyse the induction of RoKSN by GA3 further, promoter analysis was performed. A 200bp fragment (located between –200bp and –400bp) is sufficient to induce expression of RoKSN in response to GA3. Indeed, the 200bp promoter is not responsive to GA3, whereas the 400bp promoter revealed a stronger activity in the presence of GA3 (Fig. 5C). This part of the promoter contained GAMYB and PIF3 cis-elements (Supplementary Fig. S1 at JXB online). It is hypothesized that the transcription factor PIF3 (PHYTOCHROME INTERACTING FACTOR 3) integrates different endogenous and environmental factors such as light and GA to induce stem elongation (Feng ). In Arabidopsis, AtMYB33, a GAMYB transcription factor, is able to link to the GAMYB-binding site in the LFY promoter (Gocal ). This 200bp fragment corresponds to block A, conserved between rose and strawberry (Fig. 5A), suggesting that GA regulation might be conserved in strawberry where GA is also known to inhibit flowering (Thompson and Guttridge, 1959).

By studying transcript accumulation and promoter activity, the induction by GA3 of RoKSN, a gene of the TFL1 family, was clearly demonstrated. Further investigations are necessary to understand the signalling between GAs and RoKSN. This study represents the first demonstration of the induction of a floral repressor by GA. In Arabidopsis, TFL1 belongs to a small family (Kobayashi ). It was recently demonstrated that other members of the TFL1 family were regulated by hormones. The paralogues of FT, TWIN SISTER OF FT, were activated by cytokinin (BAP) to induce flowering under SDs (D’Aloia ). In A. thaliana, MOTHER OF FT and TFL1 regulated seed germination via abscisic acid and the GA pathway (Xi ). During floral initiation, it was hypothesized that FT was induced by GA via a positive regulation of CO (Hisamatsu and King, 2008; Porri ).

RoKSN mediates floral inhibition by GA

GA blocked the floral transition in OF roses (Fig. 2), and RoKSN, a floral repressor, is induced in response to GA3 (Fig. 4). It was uncertain whether this induction by GA might be the principal component in GA floral inhibition. During the floral process, the GA response between CF and OF roses was different: after GA3 treatment, OF roses did not flower, while CF roses flowered (Roberts ; this study). The CF roses had no functional RoKSN allele, due to insertion of a retrotransposon (Iwata ). As a consequence, CF roses did not accumulate RoKSN and were no longer able to respond to GA. No floral repression occurred. In OF roses, during the floral transition, RoKSN is weakly accumulated and floral initiation can happen (Fig. 4A). The low accumulation may be explained by the low GA content (Roberts et al., 1999) and the regulation of GA metabolism genes: transient accumulation of RoGA2OX, a gene encoding an enzyme involved in GA inactivation, and low accumulation of RoGA20OX, a gene encoding an enzyme involved in the last steps of active GA synthesis (Remay ).

After exogenous application of GA3, floral genes were down-regulated (such as RoSOC1, RoFD, RoLFY, and RoAP1) (Fig. 4). To determine whether these regulations were RoKSN dependent, the expression of these genes in response to GA3 was studied in a CF rose, LWP (Fig. 4G). In the absence of RoKSN, transcript accumulation of RoFT, RoSOC1, and RoAP1 was not modified after GA3 treatment. In contrast, RoLFY and RoFD transcripts were reduced 2-fold after GA3 treatment. These results suggest that RoFT, RoSOC1, and RoAP1 regulation by GA is RoKSN dependent, whereas the GA effect on RoFD and RoLFY is, at least partially, independent of RoKSN.

The following model is proposed to explain the floral inhibition by GA in OF roses. During the floral transition, exogenous application of GA3 leads to a strong accumulation of RoKSN that blocks induction of floral genes such as RoAP1, RoFT, and RoSOC1 (Fig. 4). As proposed in Arabidopsis, it can be hypothesized that RoKSN is interacting with RoFD to block induction of downstream floral genes such as RoSOC1 and RoAP1 (Hanano and Goto, 2011). The RoFT repression could be due to a positive feedback loop that is blocked in the presence of RoKSN. Such a positive feedback loop has been shown for TSF, the paralogue of FT in Arabidopsis (Yamaguchi ). RoKSN is not the only target of GA. RoFD and RoLFY are down-regulated in the absence of RoKSN, suggesting an independent pathway for these genes. However this RoKSN-independent pathway seems to be minor, as in roksn mutants (such as as LWP, a CF rose), no GA floral inhibition happens even at high GA3 concentrations (Supplementary Table S2 at JXB online). The transient expression of RoFD (15 DAB, Fig. 4B) is early and may suggest a role for RoFD in processes other than floral initiation. FD is interacting with TFL1 and FT (Wigge ; Hanano and Goto, 2011). However, a role for FD in other developmental process cannot be excluded as FD homologues can interact with proteins controlling processes such as growth (Pnueli ; Mimida ).

This is the first model that focuses on the inhibition of flowering by GA. Previous studies concerning the role of GA in flowering mainly targeted plants where GA had a stimulating effect. In these plants, for example Arabidopsis and L. perenne, GA acts via activation of floral activators (Blazquez ; Gocal ; Moon ; MacMillan ).

GA might not be the only factor that regulates RoKSN. After flowering in spring, the RoKSN level increases and blooming is blocked (Iwata ; this study). Treatment with PCB, a GA synthesis inhibitor, fails to induce new flowering (Table 1). Reduction of GA content is not sufficient to induce flowering. The pattern of floral gene accumulation is not modified after PCB treatment (Fig 4). It cannot be excluded that PCB treatment did not reduce GA content enough to induce blooming (as no GA dosage study has been performed). However, PCB treatment is sufficient to reduce internode length (Table 1; Supplementary Table S2 at JXB online). It is proposed that after the blooming in spring, RoKSN could be regulated by GA but also by other pathways. Vernalization is a potentially interesting pathway as OF roses maintained in a greenhouse without a cold period do not flower the next spring. Control of TFL1 homologues by another environmental pathway has already been proposed. In diploid strawberry, the expression of FvKSN, the orthologue of RoKSN, is regulated by the photoperiodic pathway. Under SDs, expression of FvKSN is repressed to permit floral initiation, whereas under LDs, expression of FvKSN increases to inhibit flowering (Koskela ). Furthermore, in Arabis alpina, AaTFL1 is involved in an ‘age-dependent response to vernalization’ (Wang ). In young shoots, AaTFL1 acts as a floral repressor after vernalization and blocks LFY expression. In old shoots, AaTFL1 increases the duration of vernalization required for flowering (Wang ).

In conclusion, a model has been proposed that reveals how GA can inhibit flowering in rose by positively regulating a floral repressor, RoKSN. Pathways other than GA such as vernalization may control blooming in rose, and further experiments are required to explore the molecular interactions between these different pathways.

Supplementary data

Supplementary data are available at JXB online.

Figure S1. Location of GA response elements in the promoter of RoKSN.

Table S1. List of the names and sequences of primers used in the study.

Table S2. Effect of GA3 and a gibberellin inhibitor (PCB) on OF rose (Félicité et Perpétue; FP) and CF rose (Little White Pet; LWP) on floral development and vegetative growth in spring during floral transition

Table S3. Induction of RoKSN, RoLFY, and RoFT by gibberellins (GA3) in R. hybrida ‘Félécité et Perpétue’.