The TFL1 homologue KSN is a regulator of continuous flowering in rose and strawberry

SEASONAL FLOWERING (SFL) and RECURRENT BLOOMING (RB) are orthologous loci in strawberry and rose

As rose and strawberry are phylogenetically closely related species (both belonging to the Rosoideae tribe of the Rosaceae family; Potter et al., 2007), and as the genetic control is the same in both (recessive gene in both cases), we attempted to determine by comparison mapping if the two loci controlling continuous flowering in rose and woodland strawberry, i.e. RB and SFL, respectively, are orthologues. RB and SFL were mapped on the rose linkage group RG3 and the strawberry linkage group FG6, respectively (Sargent et al., 2006; Spiller et al., 2011). Eight Rosa markers located on RG3 had one significant match to the Fragaria genome, with a BLASTN score of <E-11. These eight Fragaria matches were located on the Fragaria chromosome FG6 (Figure S1). The colinearity of these loci was conserved, except for two loci that displayed small rearrangements. Therefore, a large segment of rose linkage group RG3 was orthologous to the upper part of the strawberry FG6, which gave rise to the hypothesis of an orthologous relationship between the genes controlling recurrent flowering in Rosa and Fragaria.

RoKSN and FvKSN are TFL1 homologues and co-localise with the RB and SFL loci, respectively

Using degenerate primers in rose and results from the strawberry genome sequence (Shulaev et al., 2011), we isolated four and seven TFL1/FT homologues in rose and strawberry, respectively. We identified two TFL1 homologues in rose and strawberry as good candidates for the continuous flowering gene, termed KSN (Koushin, an old Japanese name for the Chinese CF rose cultivar). RoKSN and FvKSN clearly belonged to the TFL1 clade, along with RoTFL1 and FvATC; whereas RoFT, FvFT, FvFT2 and FvFT3 were included in the FT clade and one (FvMFT) in the MFT clade (Figure 2). We also identified two TFL1/FT homologues (RoTFL1b and FvTFL1b) that are closely related to AtBFT.

Using large populations that segregated for the OF/CF trait, we found that RoKSN and FvKSN co-localized with the RB and SFL locus in rose and woodland strawberry, respectively. No recombinant was found between the RoKSN/FvKSN and RB/SFL loci in 670 individuals from two F₁ progenies obtained via the cross between CF and OF roses, and in 158 S₁ selfed progeny of F. vesca Ilaria, respectively.

Insertion of a retrotransposon in RoKSN is responsible for the recurrent blooming character

By analysing seven CF roses of independent origins, ascertained by pedigree (Cairns, 2000), we detected the presence of a retrotransposon in the second intron of RoKSN (Figure 3). All CF roses were homozygous for the RoKSN allele with the retrotransposon. The 9-kbp long retrotransposon presented characteristics of the copia retrotransposon family (Kumar and Bennetzen, 1999). This large insertion may lead to an inactive RoKSN allele (see results below).

In order to validate the function of RoKSN in continuous flowering control in roses, we studied vegetative mutant pairs showing different flowering behaviours. Frequently, CF roses mutate into climbing roses (Lewis, 1994). These climbing roses have primary shoots with indeterminate vegetative growth, whereas inflorescences are produced by axillary secondary shoots. They bloom in spring, and occasionally they can flower later in autumn (OR; Figure 1c). Only one mutation from OF to CF was described. An OF rose, i.e. ‘Félicité et Perpétue’ (FP), gave a CF dwarf mutant ‘Little White Pet’, LWP (Roberts et al., 1999). Each of the seven mutants studied showed a difference at the RoKSN locus when compared with the wild type. All six climbing mutants had a new allele at the RoKSN locus with a 1-kbp insertion at the same place as the 9-kpb insertion in the second intron (Figure 3). This 1-kbp sequence corresponded to the long terminal repeat (LTR) element of the retrotransposon. Five climbing mutants (‘Gold Bunny Cl.’, ‘Wendy Cusson Cl.’, ‘Pink Chiffon Cl.’, ‘Iceberg Cl.’ and ‘Peace Cl.’) had both the 9-kbp retrotransposon and 1-kbp LTR insertions. One climbing mutant (‘Old Blush Cl.’) only had the 1-kbp insertion (Figure 3). Mutation in climbing roses can be explained by the recombination of the retrotransposon in the climbing mutants. This recombination restores an active RoKSN allele (see discussion below). In the rare OF/CF mutation, FP was found to be heterozygous at the RoKSN locus (allele with and without the retrotransposon), whereas LWP only had the allele with the retrotransposon (Figure 3c). In LWP, by sequencing and Southern blot analyses, only one allele has been found, i.e. that with the retrotransposon (Figures 3c and S2). In this latter genotype, the absence of the active allele (without the retrotransposon) can be explained by the deletion of the allele or a somatic chromatid exchange. Analysis of these independent mutants demonstrated that recombination of the retrotransposon restored an OF phenotype, whereas deletion of the functional allele led to CF roses.

Full-length RoKSN mRNA is not accumulated in CF roses as a result of retrotransposon insertion

We were unable to detect RoKSN mRNA accumulation in CF roses, whereas the transcript accumulated in OF roses (Figure 4b). mRNA synthesis or accumulation might be blocked by the copia-like retrotransposon insertion. To test this hypothesis, RoKSN transcript accumulation was studied by one-step reverse transcription (RT)-PCR in an OF/climbing mutant pair, i.e. ‘Peace’/’Peace Climbing’. In one-step RT-PCR, cDNA synthesis and PCR were performed at the same time using RoKSN-specific primers. As described before, ‘Peace Climbing’ has an allele with the LTR element at the RoKSN locus (Figure 3c). Using primer combination F9XR11 (surrounding intron 1), one-step RT-PCR products were observed in both ‘Peace’ and ‘Peace Climbing’, suggesting that nascent RNA is transcribed and the first intron is spliced out in both (Figure 4a and c). Using primer combination F9XR13 (surrounding intron 2, where the retrotransposon is inserted), one-step RT-PCR products were observed in ‘Peace Climbing’ and in an OF rose (‘Park Yellow Tea Scent’); no product was detected in the CF rose (Figure 4c). Then, using primer combination F9 and primers in the retrotransposon (InsR3, InsR4, InsR5 or InsR8), we observed one-step RT-PCR products in ‘Peace’ corresponding to PCR products that would be expected if the second intron was not spliced out (Figure 4a and d). We conclude that, as a result of the retrotransposon insertion, the second intron is not spliced out in CF roses and no full-length mRNAs are accumulated. In climbing mutants, the recombination of the retrotransposon restored splicing of the second intron and an active RoKSN allele, leading to indeterminate growth and non-continuous flowering.

FvKSN, the RoKSN orthologue in woodland strawberry controls continuous flowering

Sequencing the entire FvKSN revealed that CF genotypes showed a 2-bp deletion in the first exon that caused a frame shift, which led to the rapid introduction of a stop codon in the putative translated protein and to a non-functional allele named ksn (Figure 5d). In the 158 S₁ seedlings, all homo-zygous individuals with a 2-bp homozygous deletion in the recessive locus (ksn/ksn) had a CF habit, whereas all homozygous individuals without a 2-bp deletion (KSN/KSN) or heterozygous (ksn/KSN) had an OF habit.

In order to confirm the role of the 2-bp deletion of FvKSN in continuous flowering, we conducted a diversity analysis between the 2-bp deletion and flowering phenotypes using 37 accessions of F. vesca, including 13 CF genotypes. The 37 accessions were available from different genetic resources (Table S1). The sequencing of the entire FvKSN gene in all accessions revealed a total of six SNPs located in introns and the 2-bp deletion in the first exon, which led to six haplotypes (Table S1). Within this genetic polymorphism, only the 2-bp deletion was associated with the CF habit (Figure 5b and c). Using 11 neutral microsatellites, we investigated the population structure in these 37 accessions using structure 2.2. The most likely number of potential clusters (K) was estimated as two. All genotypes except one were clearly ranked within one of the two groups. The first group included 17 genotypes from Europe, mainly from the Alps, whereas the second group included 19 genotypes that originated from Europe/North America (Figure 5a). All CF strawberries, with the 2-bp deletion, were included in the former group, suggesting that in F. vesca continuous flowering may have originated from Alpine genotypes.

The continuous flowering gene shows seasonal regulation in rose

We monitored RoKSN transcript accumulation during the floral process by quantitative (q)RT-PCR. In OF Rosa wichurana, new shoots arose from axillary buds of shoots from the previous year (Figure 1a). After a short vegetative phase (first three stages, W1, W2 and W3), these new axillary vegetative shoots underwent floral differentiation, with transformation of the apical meristem into a dome-like structure (stages W4 and W5). Then the new shoot was terminated by an inflorescence (stages W6–W9). RoKSN transcripts were barely detectable in axillary vegetative shoots before floral differentiation (stages W1–W3). RoKSN transcripts were transiently accumulated during floral differentiation (stage W4) and later during inflorescence development (stages W7 and W8; Figure 6a). During the floral process, RoFT, a floral activator, and later RoLFY, a homologue of the floral identity gene LEAFY (Remay et al., 2009), were progressively accumulated (Figure 6b and 6c). After the first flowering, new shoots arose from the base of the bush and had indeterminate vegetative growth during the rest of the growing season (Figure 1a). An 80-fold increase in RoKSN transcript accumulation was detected when spring (stage W1) and autumn (stage W10) vegetative shoots were compared (Figure 6a). After flowering, RoFT and RoLFY had the same level compared with spring vegetative shoots (stage W1; Figure 6). Therefore, low RoKSN expression in spring was associated with floral transition and determinate growth, whereas high expression of RoKSN in autumn was associated with vegetative and indeterminate growth. In CF roses, functional RoKSN transcripts were not detected (because of retrotransposon insertion), and all new shoots terminated in an inflorescence.

A floral repressor regulates continuous flowering in rose and woodland strawberry

In rose, we demonstrated that the continuous flowering phenotype was explained by the insertion of a copia-like retrotransposon in the second intron of RoKSN, i.e. a TERMINAL FLOWER 1 homologue. Insertion of the retrotransposon modified RNA maturation. The second intron was not spliced out, and full-length RoKSN RNA was not accumulated (Figure 4). In six independent vegetative mutants, climbing mutant roses differed from CF roses in the size of the insertion, probably as a result of recombination between LTR elements (Figure 3). After retrotransposon recombination, only the solo LTR element remained and the second intron was spliced out, thus restoring a functional RoKSN allele (Figure 4). Such a process has been described in grapevine, where the recombination of a Ty3-gypsy-type retrotransposon restores MYB transcription factors responsible for fruit colour (Kobayashi et al., 2004). In the CF/climbing mutant pair, the restoration of RoKSN in climbing mutants may not be complete as flowering can occur occasionally in the autumn (Figure 1c). The presence of the LTR element in the second intron may modify the level or the spatiotemporal pattern of RoKSN expression. Indeed, LTR elements contain signals for transcription regulation (Kumar and Bennetzen, 1999), which may modify RoKSN expression in climbing mutants. In another vegetative mutant pair (changing from OF to CF behaviour), the functional allele was deleted and only the non-functional allele was present, leading to the continuous flowering phenotype (Figure 3c and S2). Furthermore, the role of KSN in continuous flowering control was strengthened by the absence of recombination between RoKSN and the continuous flowering locus in large F₁ progenies segregating for the CF/OF phenotype (in 670 individuals). These results strongly suggest that in rose the continuous flowering gene is encoded by RoKSN, i.e. a TFL1 homologue. Functional validation still needs to be performed in rose by knocking out the gene in OF roses. Presently, no protocols for the genetic transformation of OF roses are available (Debener and Hibrand-Saint Oyant, 2009). However, the conclusion drawn in rose was reinforced by the results obtained in woodland strawberry.

In F. vesca, which also showed continuous flowering in certain genotypes, we identified FvKSN, a gene showing high similarity with RoKSN (Figure 2). Using comparative mapping, we showed that the two genes were located in a syntenic and co-linear region, suggesting that they are orthologues. Furthermore, all CF individuals from an S₁ segregating population had a 2-bp deletion in the coding sequence of the gene, whereas the deletion was absent in all OF homozygous genotypes. This deletion induced a frame shift and a stop codon leading to a truncated protein (Figure 5d). We propose that this 2-bp deletion in the coding sequence is responsible for continuous flowering in F. vesca. As no mutant was available in strawberry, we performed a diversity analysis. Among the 37 accessions studied, the 2-bp deletion and the continuous flowering phenotype were always associated (Figure 5). We concluded that FvKSN was the orthologue of RoKSN, and that both genes control the continuous flowering trait.

KSN and the seasonality of flowering

The main consequence of the continuous flowering mutation is the modification of flowering seasonality. Perennial plants repeatedly cycle between vegetative and floral development. Most perennial plants have a single reproductive phase that is synchronized to the changing seasons. OF plants bloom in spring or early summer, and further developments are vegetative. This seasonality might be explained by seasonal regulation of KSN. We showed a seasonal regulation of RoKSN transcript accumulation in OF roses: before the first bloom, RoKSN was barely expressed; in parallel, RoFT and RoLFY, homologues of the floral integrator FT and the floral identity gene LFY (Remay et al., 2009), were progressively accumulated (Figure 6). The low RoKSN level in spring may increase the flowering competence of the apical meristem. Opposite results have been found in other Rosaceae (Japanese pear and quince) where TFL1 homologues are expressed before floral differentiation and are repressed during floral differentiation (Esumi et al., 2007). In all cases, flowering is associated with LFY induction (Figure 6c; Esumi et al., 2007). TFL1 might be differently regulated between rose and pear. The RoKSN expression during floral differentiation (stages W4, W7 and W8; Figure 6a) suggests a role of RoKSN in inflorescence development. Genetic analysis reveals that a quantitative trait locus (QTL) involved in the inflorescence architecture is centred on the RB locus in rose (Kawamura et al., 2011).

After the first blooming, newly arising shoots remained vegetative and RoKSN transcripts accumulated. No RoFT and RoLFY transcripts accumulated. In continuous flowering roses, this regulation was disrupted by the absence of RoKSN transcript accumulation. The floral repressor was absent and flowering occurred during all seasons. This report and a previous study in Arabis alpina (Wang et al., 2009) highlighted the importance of floral repressors in the control of flowering seasonality. In A. alpina, PERPETUAL FLOWERING 1 (PEP1), a homologue of the floral repressor FLOWERING LOCUS C (FLC), limited the duration of flowering and facilitated the return to vegetative development. PEP1 is transiently repressed by low temperature, thereby allowing seasonal flowering of A. alpina (Wang et al., 2009).

KSN and other aspects of plant development

Origin of continuous flowering in horticulture

Our results highlighted the origin of CF in two important horticultural plants. In woodland strawberry, CF was well described from the 18th century, and was proposed to have originated in the Alps (Duchesne, 1766). The hypothesis of only one mutation is strengthened by the presence of only one haplotype in CF genotypes with the 2-bp deletion in the FvKSN coding sequence, even though the OF displayed five haplotypes (Figure 5; Table S1). Only one mutation might be responsible for the CF habit in the wood strawberry genotypes studied, despite the fact it has been found growing wild at various sites in the European Alps (De Vilmorin, 1898). This hypothesis has to be validated on a larger number of genotypes. In rose, CF flowering was thought to arise from a mutation in wild Chinese rose and then introduced in old Chinese cultivated roses (Hurst, 1941; Ogisu, 1996). We hypothesize that the mutation may result from the insertion of a retrotransposon into the RoKSN locus in wild R. chinensis spontanea, as the retrotransposon was absent in the RoKSN locus in this wild species (Figure 3b), i.e. an OF rose that is supposed to be an ancestor of the original CF roses (Martyn, 2005).

In conclusion, KSN may be an integrator of endogenous and environmental factors. Understanding the regulation of KSN by these different factors may provide new prospects for controlling flower and fruit production in perennial plants.

Plant material

Rosa wichurana originated from the ‘Jardin de Bagatelle’ (Paris, France). Rosa hybrida‘Félicité et Perpétue’ and its mutant ‘Little White Pet’ were obtained from Loubert Nursery (Les Rosiers sur Loire, France). Rosa chinensis spontanea was a gift from Mr Ogisu. ‘Gold Bunny’, ‘Wendy Cussons’, ‘Pink Chiffon’, ‘Old Blush’, ‘Iceberg’, ‘Peace’ and the corresponding climbing mutants were obtained from Hiroshima Botanical Garden (Japan), and are of independent origins (Cairns, 2000). The mutants are vegetative mutants, also known as sports. A sport of a clonally propagated plant arises as a mutant cell in the shoot meristem (Lewis, 1994). Rose F₁ progenies were obtained from a cross between R. whichurana and H190 (575 individuals; Hibrand-Saint Oyant et al., 2008) and R. wichurana and R. hybrida‘The Fairy’ (95 individuals; Kawamura et al. 2011). In strawberry, the 158 S₁ progeny was obtained from F. vesca Ilaria, issued from a cross between Fragolina di Ribera and Alpine (Consiglio per la Ricerca e la Sperimentazione in Agricoltura, Forli, Italy). The collection of the 37 genotypes listed in Table S1 was obtained from Dresden, INRA and USDA-Corvallis.

Cloning and phylogenetic analysis

TFL1 homologues in rose were isolated by PCR on R. chinensis spontanea using the primers UtflF1 and UtflR1. All primers are listed in Table S2. In strawberry, FvKSN was first isolated using primers from rose, KSNqPCRF2 and KSNqPCRR2. Other TFL1 homologues were isolated using sequences obtained via Fragaria genome sequencing (Shulaev et al., 2011). Amino acid sequences of FT and TFL1 homologues in Arabidopsis, tomato, poplar and grapevine were aligned with predicted FT/TFL1 proteins in strawberry and rose using clustalw (Thompson et al., 1997). A phylogenetic tree was produced from 1000 bootstrap replicates by applying the neighbour joining (NJ) method with phillips software. The phylogenetic tree was displayed using treeview (Page, 1996).

Phenotyping, mapping and genetic structure

Segregating populations were phenotyped for CF. In rose, genotypes were considered as CF when plants flowered in the first year after germination, and were confirmed by the presence of flowers in October and November over a 4-year period (2006–2009). In strawberry, genotypes were considered as CF when plants had inflorescences in late summer and fall. In the seven pairs of mutants, the climbing rose genotypes, classified as OF, may occasionally flower late in the summer. In rose, based on polymorphism caused by the retrotransposon insertion, the RoKSN genetic marker was developed using a multiplex kit (Qiagen, http://www.qiagen.com) with F3, InsR3 and R1 primers. In strawberry, FvKSN was genotyped based on the 2-bp deletion identified in the coding region. The relatedness of 37 wood strawberry genotypes was studied with 11 neutral microsatellite loci (Table S3), distributed throughout the genome, using structure software (Pritchard et al., 2000). Comparative mapping between RG III and FG VI was based on the anchorage of eight rose sequence markers, genes or microsatellites, on the strawberry genome (http://www.strawberrygenome.org) using blastn. Gene markers were localized precisely, whereas SSR markers were attributed to scaffolds.

Gene expression analysis

Total RNA extraction was performed as previously described by Remay et al. (2009). To test the effect of the retrotransposon on mRNA accumulation, a one-step RT-PCR was performed using the one-step RT-PCR kit (Qiagen), according to the manufacturer’s recommendations. Briefly, both cDNA synthesis and PCR were performed in a single tube using gene-specific primers and target RNA. To study RoKSN accumulation during the floral process, a two-step RT-PCR was carried out as previously described (Remay et al., 2009). Total RNA was reverse transcribed using oligo-dT primers and subjected to quantitative PCR to analyse transcript accumulation. The level of RNA was normalized using the TCTP gene as a reference, after checking the homogeneity of the cycle threshold (C_t) variation with a second housekeeping gene: i.e. EF1α (Remay et al., 2009). Terminal parts of the shoots were harvested in spring and autumn from R. wichurana at different time points (see details in the legend of Figure 6). The terminal part of the shoot was dissected (removal of young leaves), and therefore corresponds to the shoot apical meristem, the leaf primordia and the stem.