FvGI–FvCO–FvFT1 pathway activates FvTFL1 in long days to repress flowering in woodland strawberry (Fragaria vesca)

Summary

• Photoperiod is a stable seasonal signal. Although photoperiodic flowering is well understood in short‐day (SD) and long‐day (LD) annual plants, the regulatory mechanisms in perennials remain elusive. In a perennial woodland strawberry (Fragaria vesca), flowering is induced in SDs in autumn and plants flower following spring, while in plants with mutated FvTERMINAL FLOWER1 (FvTFL1), LDs induce flowering.

• We investigated photoperiodic flowering of F. vesca through phenotypic and molecular characterization of transgenic lines and their crosses. We studied natural variation in flowering time and gene expression in European accessions and explored their correlations with climatic, geographical and genetic origins.

• We showed that FvGIGANTEA (FvGI) activates FvFLOWERING LOCUS T1 (FvFT1) via FvCONSTANS (FvCO) in LDs resulting in early flowering in fvtfl1 mutant. In natural accessions, FvFT1 activates FvTFL1 to repress flowering, and decreasing expression of FvFT1 and FvTFL1 towards colder climates in the east and north correlates with earlier flowering.

• We define a photoperiodic flowering pathway controlling floral transition of perennial F. vesca in autumn that differs from known mechanisms in annual and perennial plants. Our findings open new avenues to understand how perennial plants cope with changing seasons across climatic and geographical ranges.

Natural variation of flowering time correlates with FvFT1 and FvTFL1 expression in European Fragaria vesca accessions.

Introduction

Perennials comprise the largest fraction of plant species on Earth with an increasing proportion towards higher latitudes (Poppenwimer et al., ²⁰²³). They are mostly polycarpic plants that repeat yearly cycles of vegetative and generative growth to ensure successful reproduction during their lifespan, in contrast to monocarpic annual and biennial species that rely on a single flowering event (Albani & Coupland, 2010). A common perennial life cycle comprises the formation of overwintering floral buds in autumn as a response to short days (SD) and/or decreasing temperatures and flowering in the following spring (Schnablová et al., ²⁰²¹). Such a life cycle is regulated by strong floral repressors that restrict floral transition to certain meristems and time window, conferring seasonal flowering and the maintenance of vegetative meristems for the next yearly cycle (Wang et al., ^{2009, 2011}; Albani & Coupland, ²⁰¹⁰; Koskela et al., ²⁰¹²; Lazaro et al., ²⁰¹⁸). Although photoperiod is a major signal controlling seasonal flowering, the underlying mechanisms are poorly understood in perennials (Albani & Coupland, 2010; Andrés & Coupland, ²⁰¹²).

In perennial woodland strawberry (Fragaria vesca), TERMINAL FLOWER1 (FvTFL1) encodes a floral repressor that confers seasonal flowering habit (Koskela et al., ²⁰¹²). The peculiarity of F. vesca is that the presence of functional FvTFL1 reverses the photoperiodic requirement for flower induction. Wild‐type plants require SDs in autumn for flower induction in the temperature range of 13–20°C and flower seasonally, whereas fvtfl1 mutants are quantitative long‐day (LD) plants that flower perpetually (Heide & Sønsteby, 2007; Koskela et al., ²⁰¹²; Rantanen et al., ²⁰¹⁵). TFL1 belongs to the family of phosphatidylethanolamine binding proteins (PEBPs) that delay floral transition, promote the indeterminacy of the inflorescence meristem and contribute to the perennial life cycle by blocking the flowering of young axillary shoots (Bradley et al., ^{1996, 1997}; Wang et al., ²⁰¹¹; Lembinen et al., ²⁰²³). A gene encoding another PEBP, FLOWERING LOCUS T (FT), is expressed under flower‐inductive conditions in both LD and SD plants, and FT protein functions as a mobile flower‐inducing signal, the florigen (Kobayashi et al., ¹⁹⁹⁹; Corbesier et al., ²⁰⁰⁷; Faure et al., ²⁰⁰⁷; Tamaki et al., ²⁰⁰⁷; Komiya et al., ²⁰⁰⁹; Kong et al., ²⁰¹⁰). FT is synthesized in leaves and transported to the shoot apical meristem (SAM), where it interacts with the bZIP transcription factor FD to induce flowering (Abe et al., ²⁰⁰⁵). In the SAM, FT has been shown to compete with TFL1 for binding with FD (Zhu et al., ²⁰²⁰), providing a possible mechanism for their antagonistic functions. However, in F. vesca, expression patterns of FT‐homologues suggest a different mechanism in the regulation of photoperiodic flower induction. FvFT1 is expressed in the leaves specifically in LDs in both wild‐type and fvtfl1 mutant plants, correlating positively with flowering only in the mutant (Koskela et al., ²⁰¹²; Rantanen et al., ²⁰¹⁴; Kurokura et al., ²⁰¹⁷), whereas FvFT2 and FvFT3 are highly activated in the late inflorescence and flower meristems (Koskela et al., ²⁰¹⁷; Lembinen et al., ²⁰²³).

Homologues of GIGANTEA (GI) and CONSTANS (CO) are major activators of florigens in both SD and LD annual plants (Song et al., ²⁰¹⁵; Vicentini et al., ²⁰²³). In Arabidopsis, GI interacts with the blue‐light receptor FLAVIN‐BINDING, KELCH REPEAT, F‐BOX 1 (FKF1) to promote CO transcription in LDs (Imaizumi et al., ²⁰⁰³; Sawa et al., ²⁰⁰⁷). GI and FKF1 also contribute to the stabilization of CO protein that accumulates when CO mRNA expression coincides with light in the afternoon, enabling CO to activate FT in leaves (Valverde et al., ²⁰⁰⁴; Hwang et al., ²⁰¹⁹). A recent work showed that the CO‐FT/TWIN SISTER OF FT‐LIKE (TSFL) module promotes flowering in LDs also in the perpetual flowering1 (pep1) mutant of alpine rockcress (Arabis alpina L.), a perennial relative of Arabidopsis. Both AaCO and AaFT/TSFL genes are involved in the regulation of flowering in the primary shoot, whereas AaCO is not needed for flowering in axillary shoots, indicating that the photoperiodic pathway contributes also to perennial shoot architecture (Sashidhar & Coupland, 2026). Similarly, FvCO and FvFT1 were shown to promote flowering in LDs in F. vesca fvtfl1 mutant, and FvCO was found to be essential for the activation of FvFT1 (Koskela et al., ²⁰¹²; Rantanen et al., ²⁰¹⁴; Kurokura et al., ²⁰¹⁷). Because these studies in A. alpina and F. vesca were carried out in perpetual flowering mutants in which LDs promote flowering, they do not provide answers to a major open question: How does photoperiod regulate floral transition in perennials in autumn?

Using transgenic lines and their crosses, we show that LDs activate FvFT1 through the conserved FvGI–FvCO–FvFT1 photoperiodic flowering pathway in both seasonal flowering SD genotypes and perpetual flowering LD genotypes of F. vesca. We found that in seasonal flowering accessions, LD‐activated FvFT1 upregulates FvTFL1 to repress flowering, in contrast to its previously reported function as a floral activator in perpetual flowering fvtfl1 mutant (Koskela et al., ²⁰¹²). Further analysis on European F. vesca accessions, that are divided into eastern and western genetic clusters (Toivainen et al., ²⁰²⁶), revealed lower average FvFT1 expression level and earlier flowering in eastern F. vesca populations compared with western accessions. Also, a strong positive correlation between FvFT1 and FvTFL1 mRNA levels was found, indicating that FvFT1 is a major driver of geographical differences in flowering time. Overall, our findings define a photoperiodic flowering pathway in strawberries that differs from those described in annual and perennial plants and offers potential strategies to optimize flowering and yield under changing climate in strawberries and related fruit crops of the Rosaceae family.

Materials and Methods

Plant material

Woodland strawberry (Fragaria vesca L.) accessions Hawaii‐4 (H4, fvtfl1 mutant), FIN56 (PI551792, originally obtained from the National Clonal Germplasm Repository, Corvallis) and NOR8 from Alta, Norway (70,03244°N, 23,40124°E) were maintained as seeds through self‐pollinations and used for gene functional analyses. FIN56 and NOR8 were previously fully sequenced and showed low genome‐wide heterozygosity levels (Toivainen et al., ²⁰²⁶). Newly generated H4‐FvGI‐RNAi and H4‐FvGI overexpression (OE) lines, and previously reported H4‐FvFT1‐RNAi and H4‐FvCO‐RNAi lines were germinated from seeds and used for gene functional studies (Koskela et al., ²⁰¹²; Kurokura et al., ²⁰¹⁷). FvGI‐OE lines were generated also in seasonal flowering FIN56 and NOR8 backgrounds. These materials were vegetatively propagated for experiments through runners. To investigate genetic interactions among FvGI, FvCO, FvFT1 and FvTFL1, several crosses between the transgenic lines and selected accessions were carried out and resulting F₁ and F₂ hybrids were used for the experiments (Supporting Information Table S1). Finally, a collection of 181 European F. vesca accessions, maintained as clones, was used for studying natural variation in flowering time (Table S2).

Plasmid constructs and transformation

Plasmid constructs for overexpression lines were created according to Gateway technology with Clonase II (Invitrogen). For FvGI‐OE and FvGI‐RNAi lines, cDNA sequences from F. vesca accession H4 were amplified using primers described in Table S3 and introduced into the pH7WG2D‐1 (overexpression) and pK7GWIWG2‐7F2 (RNAi‐silencing) destination vectors containing green fluorescent protein (GFP) as a positive selection marker (Karimi et al., ²⁰⁰²). Strawberry transformation was carried out as described previously (Oosumi et al., ²⁰⁰⁶). Thirteen to twenty transgenic lines were generated and screened for flowering time phenotypes and changes in gene expression for each construct. Three to four lines per construct were selected for further experiments.

Growth conditions and phenotyping

H4 and its transgenic lines were germinated from seeds on wet filter paper in 18‐h LDs (LD^18h) at 22°C until cotyledons were visible. At the two‐cotyledon stage, seedlings were rooted to Jiffy Peat Disks (Jiffy Products International BV) in a glasshouse in LD^18h at 20°C : 17°C (day : night). After 2 wk, plants with 2–3 leaves were transplanted into 8 × 8 cm pots for photoperiodic treatments. Photoperiodic treatments were conducted in a glasshouse under LD^18h (18 h : 6 h, light : dark) or SD^12h (12 h : 12 h, light : dark) conditions. In the glasshouse, natural lighting was complemented with 150 μmol m⁻² s⁻¹ photosynthetic photon flux density (PPFD) from high‐pressure sodium lamps (Airam 400W, Kerava, Finland).

Runner plants of NOR8, FIN56 and their transgenic lines were rooted on Jiffy peat disks (Jiffy Products International BV) and then transferred into 8 cm x 8 cm pots filled with peat (Kekkilä, Finland). The plants with at least two to three leaves were subjected to photoperiodic treatments indicated in the figure legends in either glasshouse (FIN56) or growth rooms (NOR8) at 17°C. Growth rooms were equipped with LED lamps (AP67, Valoya, Finland), providing a PPFD of c. 150 μmol m⁻² s⁻¹. After the treatments, plants were transferred to the glasshouse for flowering time observations using the conditions described above.

All hybrids were propagated in the glasshouse (LD^18h at 20°C : 17°C), and plants were subjected to photoperiodic treatments in either growth rooms described above or in an Arabidopsis chamber AR‐36L2 (Percival, Germany), followed by flowering observations in the glasshouse (LD^18h at 20°C : 17°C). Conditions during the photoperiodic treatments are indicated in figure legends.

To explore flowering time of natural accessions, the plants were subjected to SD^12h at 16°C for 6 wk in the glasshouse. In another experiment, the accessions were exposed to ambient photoperiod and temperature conditions in Helsinki, Finland, from mid‐August until October 12, 2017 (Field Aug–Oct). Both groups were transferred to the glasshouse (LD^18h, 20°C : 17°C) after treatments for flowering observations.

All flowering‐time phenotypes were determined as the number of leaves initiated before the first inflorescence and the number of days from the beginning/end of the photoperiodic treatments (indicated in figure legends) to the first flower open.

RNA extraction and gene expression analyses

Leaf and shoot apex samples were collected under photoperiodic treatments in time points indicated in the figures or figure legends and stored at −80°C before carrying out RNA extraction. Each biological replicate consisted of a single leaflet or one or three shoot apices. Total RNA was extracted from leaf and apex samples as in Mouhu et al. (2009). Protoscript II reverse transcriptase kit (NEB, US) was used for cDNA synthesis. Quantitative reverse transcription was performed using a LightCycler 480 SYBR Green I Master kit (Roche Diagnostics, Indianapolis, IN, USA) in a Roche LightCycler 480 (Roche Diagnostics), as described in Mouhu et al. (2009). Primers used in this study are shown in Table S3. Each gene was tested by three technical replicates. FvMSI1 (FvH4_7g08380) was used as a stable reference gene (Mouhu et al., ²⁰⁰⁹). The number of biological replicates was described in the figure legends.

RNA sequencing

For bulk RNA‐seq, total RNA was extracted from leaves collected 4 h after dawn (ZT4) from three biological replicates of 12 accessions (six western and six eastern; Dataset S1) reported in Toivainen et al. (2026), grown in a glasshouse under LD^18h at 18°C. Libraries were prepared using the TruSeq Total RNA kit (Illumina) following rRNA depletion with Pan‐Plant riboPOOL probes (siTOOLs Biotech) and MagSi‐STA 1.0 beads (Magtivio, Nuth, the Netherlands). Sequencing (100 bp, single‐end) was performed on an Illumina NextSeq500. Raw FASTQ reads were quality‐trimmed with Trimmomatic v.0.39 in single‐end mode (SLIDINGWINDOW:5:20, LEADING:5, TRAILING:5, MINLEN:50). Trimmed reads were aligned to the F. vesca reference genome v.4.0 using Star v.2.7.10a with the genome indexed using the v.4.0.a2 annotation from the Genome Database for Rosaceae. Only uniquely mapped reads were retained. Gene‐level counts were generated with HTSeq‐count v.0.13.5 in unstranded mode.

Differentially expressed genes analysis was performed using edgeR (Robinson et al., ²⁰¹⁰). Gene‐level CPM (counts per million) values from 36 RNA‐seq libraries were obtained using normalized library sizes. Only genes with sufficient counts were retained as determined using filterByExpr function. Three samples were removed from the analysis based on multidimensional scaling plot and sample clustering (Fig. S1). DEGs between the eastern and the western genetic clusters were obtained using the following parameters: fold change > 2, FDR < 0.05, P‐value < 0.05.

Scanning electron microscopy

Dissected meristem samples were fixed overnight in FAA buffer (3.7% formaldehyde, 5% acetic acid and 50% ethanol) and then passed through an ethanol series (50%, 60%, 70%, 80%, 90% and 100%) under mild vacuum (c. 0.6 atm). Critical point drying was performed using a Leica EM CPD300 (Leica Mikrosystems GmbH, Vienna, Austria). The dried samples were coated with platinum using a Quorum Q150TS coater (Quorum Technologies, Laughton, UK) and examined using a Quanta 250 FEG scanning electron microscope (FEI, Hillsboro, OR, USA) at the Electron Microscopy Unit, Institute of Biotechnology, University of Helsinki.

Statistical analyses

For flowering time data in H4 and their transgenic lines, two‐way ANOVA followed by Tukey's HSD multiple comparisons test was carried out. For flowering time data on NOR8 and their transgenic lines, F₁ and F₂ hybrids, and European F. vesca accessions, one‐way ANOVA followed by Tukey's HSD multiple comparisons or t‐tests were performed, as indicated in figure legends. Pearson correlation coefficients were calculated between flowering time and relative gene expression, and among the expression levels of flowering time genes. P‐values were estimated with Bonferroni correction. All analyses were performed in R v.4.3.0 (R Core Team, ²⁰²³).

Sequence alignment and phylogenetic analysis

Protein sequences were retrieved from the NCBI database (Table S4). Multiple sequence alignment was performed in MEGA11 using the ClustalW algorithm with default gap opening and extension penalties. The alignment was inspected manually to confirm correct positional homology and adjusted where necessary. Phylogenetic analysis was conducted in MEGA11 using the neighbour‐joining method. Branch support was assessed with 1000 bootstrap replicates, and gaps/missing data were treated using the pairwise‐deletion option. Sequence conservation among the six species, which GI homologue function has been reported, was evaluated using DNAman v.8.0. Conserved regions were identified using the DNAman conservation analysis module, and percent identity was calculated across the full‐length alignment as well as within key conserved blocks, which were shown in different colors.

Results

FvGI , FvCO and FvFT1 promote flowering in long days in perpetual flowering F. vesca tfl1 mutant

We have previously identified key regulators of the photoperiodic pathway in F. vesca including FvCO (FvH4_6g45860; Kurokura et al., ²⁰¹⁷) and FvFT1 (FvH4_6g00090; Koskela et al., ²⁰¹²). Here, we analysed a single gene encoding a GI‐like protein from the F. vesca genome v.4.0 (FvH4_2g13030). FvGI is highly conserved with GI homologues from other species and is positioned among the GI homologues from other Rosaceous species in the phylogenetic tree (Fig. S2). To explore the role of FvGI in the photoperiodic flowering pathway of F. vesca, we generated transgenic FvGI overexpression (FvGI‐OE) and RNAi silencing lines in the perpetual flowering fvtfl1 mutant accession H4 and examined their flowering phenotypes under LD^18h and SD^12h conditions together with previously developed FvCO‐RNAi and FvFT1‐RNAi silencing lines (Koskela et al., ²⁰¹², Kurokura et al., ²⁰¹⁷, Fig. S3A–D). Similar to FvCO‐RNAi and FvFT1‐RNAi plants, FvGI‐RNAi lines exhibited a strong late flowering phenotype under LD^18h conditions, demonstrating that all these genes are activators of flowering in LDs in H4 (Figs 1a, S4). However, under SD^12h conditions, independent FvGI‐, FvCO‐ and FvFT1‐RNAi lines flowered either at the same time or slightly later than H4 (Figs 1a, S4). By contrast, FvGI‐OE advanced flowering in SD^12h but not in LD^18h (Figs 1a, S4).

Fig. 1.

FvGI, FvCO and FvFT1 promote flowering in perpetual flowering Fragaria vesca. (a) Number of leaves before terminal inflorescence in H4 (fvtfl1 mutant) and transgenic lines in H4 background under 18‐h long‐day (LD^18h) and 12‐h short‐day (SD^12h) conditions. The upper, middle and lower horizontal lines represent the 75^th percentile, median and 25^th percentile, respectively. The whiskers represent the minimum and maximum. Each point represents the data from individual plants. Numbers under the boxplot indicate the number of flowered plants per the total number of plants in each group. Different letters indicate statistically significant differences as determined by two‐way ANOVA followed by Tukey's HSD multiple comparisons test (P < 0.05). (b–g) Relative expression (mean ± SD, n = 3) of FvGI, FvCO and FvFT1 over a 24‐h cycle in H4 and FvGI overexpression (FvGI‐OE) and RNAi lines after 3 wk under LD^18h (b, d, f) or SD^12h (c, e, g) conditions. White and black bars above the panels represent light and dark periods, respectively. (h, i) Relative expression (mean ± SD, n = 3) of FvSOC1 (h) and FvAP1 (i) in shoot apices of H4, FvGI‐OE and FvGI, FvCO and FvFT1 RNAi silencing lines. Each biological replicate contains a single shoot apex harvested at ZT3‐ZT5 after 3 wk under LD^18h or SD^12h treatment.

Consistent with flowering phenotypes, FvFT1 expression level was very low in the leaves of FvGI, FvCO or FvFT1 RNAi lines sampled 4 h after dawn (ZT4) in LD^18h (Fig. S3). In the FvGI‐OE lines, FvCO and FvFT1 mRNA levels varied, showing either slight to moderate increases or no difference at ZT4 compared with H4. We next analysed the diurnal expression patterns of FvGI, FvCO and FvFT1 in H4 and selected transgenic lines. Under LD^18h conditions, FvGI expression peaked at ZT8‐12, followed by a gradual decline towards the end of the diurnal cycle, while a higher and earlier expression peak (ZT4) was found under SD^12h treatment, with a rapid decline towards early night (Fig. 1b,c). In FvGI‐OE and RNAi lines, FvGI expression remained rhythmic with higher and lower amplitude compared with H4, respectively, under both conditions (Fig. 1b,c). Furthermore, RNAi silencing of FvCO or FvFT1 did not affect FvGI expression compared to H4, indicating that they function downstream of FvGI (Fig. S5A,B). FvCO expression peaked at dawn/early morning under both conditions in all lines, with higher peaks under SD^12h treatment. Its expression was elevated in FvGI‐OE lines, while FvGI‐RNAi lines showed reduced FvCO expression levels at some time points, especially in SD^12h (Figs 1d,e, S5C,D). In consistence with the observed flowering phenotypes, FvFT1 mRNA was detected only under LD^18h conditions with a morning peak at ZT4–ZT8 and a higher evening peak at ZT20 in H4, whereas in FvGI‐, FvCO‐ and FvFT1‐RNAi lines, FvFT1 expression remained nearly undetectable under both photoperiods (Figs 1f,g, S5E,F). FvGI‐OE plants, by contrast, exhibited highly elevated daytime FvFT1 mRNA levels under LD^18h conditions, while the evening expression peak at ZT20 was not strongly affected. Furthermore, rhythmic FvFT1 expression with morning and evening peaks was detected also under SD^12h conditions, but the overall expression level remained lower than in H4 under LD^18h conditions. The expression of FvFT2 (FvH4_4g30710), that was previously proposed to function as an SD‐activated florigen in F. vesca (Gaston et al., ²⁰²¹), was undetectable in all samples (Fig. S5G).

Next, we tested the expression of FvSOC1 (FvH4_7g12700) and the floral identity gene FvAP1 (FvH4_4g29600) that were previously shown to be regulated by FvFT1 in the shoot apex (Mouhu et al., ²⁰¹³). Under LD^18h conditions, FvSOC1 mRNA level was slightly elevated in the FvGI‐OE7 line and reduced in the FvGI, FvCO and FvFT1‐RNAi lines compared to H4 (Fig. 1h). Consistent with late flowering, FvAP1 expression was also reduced in all RNAi lines (Fig. 1i). Under SD^12h conditions, both FvSOC1 and FvAP1 were activated in the early flowering FvGI‐OE7 line, while low expression levels were found in H4 and RNAi lines. Taken together, our data suggest that FvGI activates the expression of FvCO and FvFT1 in the leaves, resulting in early floral transition in LDs in H4.

FvGI represses flowering in seasonal flowering F. vesca

In contrast to the perpetual flowering H4, seasonal flowering F. vesca contains a functional FvTFL1 and requires SDs to induce flowering (FvH4_6g18480; Koskela et al., ²⁰¹²). We next generated FvGI‐OE lines in seasonal flowering accessions, including the reference accession FIN56 and the accession NOR8 that exhibited low FvFT1 expression and early flowering after SD induction (Figs S6A–C, S7A,B). All plants remained vegetative under noninductive LD^18h conditions. After flowering‐inductive SD^12h treatment, all NOR8 and FIN56 wild‐type plants flowered, whereas most of the FvGI‐OE plants remained vegetative (Fig. S7C–F), indicating that FvGI represses flowering in seasonal flowering F. vesca. Gene expression analyses confirmed that FvGI‐OE strongly activated FvFT1 expression, both in the morning and in the evening in all lines compared with respective wild‐type lines, while the effect on FvCO mRNA level was milder (Fig. S8A–G). In the shoot apex, FvGI‐OE lines exhibited higher FvSOC1 and FvTFL1 expression levels compared to NOR8 (Fig. S8H–J), indicating that FvGI may repress flowering by activating FvTFL1 through FvSOC1 (Mouhu et al., ²⁰¹³).

To examine the repressor function of FvGI in more detail, we subjected wild‐type NOR8 and two late flowering FvGI‐OE lines to three photoperiods including LD^16h, SD^12h and SD^8h for 5 wk, followed by flowering observations in LD^18h. None of the plants flowered after LD^16h (Fig. 2a,b). In SDs, overexpression of FvGI shortened the critical photoperiod for flower induction: The SD^8h resulted in flowering of 100, 80 and 60% of the NOR8, FvGI‐OE9 and FvGI‐OE11 plants, respectively, while after SD^12h, all NOR8 plants flowered, FvGI‐OE9 remained vegetative and 20% of the FvGI‐OE11 plants flowered. FvGI‐OE plants flowered c. 1 wk later and with two more leaves than NOR8 after the SD^8h treatment. Slower floral transition in FvGI‐OE plants was also visible in the meristems. The SEM images, taken at the end of the treatments, showed more developed primary flowers in NOR8 compared with transgenic lines (Fig. S9).

Fig. 2.

FvGI represses flowering in seasonal flowering Fragaria vesca. (a) Number of leaves before terminal inflorescence in NOR8 with functional FvTFL1 alleles and FvGI overexpression (FvGI‐OE) lines in NOR8 background under indicated conditions. Values indicate mean ± SD (n = 10); NF, no flower. (b) Flowering time of NOR8 and FvGI‐OE lines indicated as days to anthesis from the end of indicated treatments. Horizontal lines show the 75^th percentile (upper), median (middle) and 25^th percentile (lower). The whiskers extend to minimum and maximum values. Different letters indicate statistically significant differences as determined by one‐way ANOVA followed by Tukey's HSD multiple comparisons test (P < 0.05). Clonally propagated plants were precultured under 18‐h long‐day (LD^18h) conditions for 4 wk, after which the plants were subjected to 8‐h or 12‐h short days (SD^8h, SD^12h) or LD^16h at 17°C for 5 wk followed by phenotyping under LD^18h. (c) FvFT1 expression (mean ± SD, n = 3–4) in leaf samples collected at zeitgeber time 4 (ZT4) after 5 wk under the indicated photoperiods. (d–g) FvSOC1 (d), FvTFL1 (e), FvAP1 (f) and FvFT2 (g) expression (mean ± SD, n = 3) in shoot apices collected from ZT3‐ZT5 after 0, 2 and 5 wk under indicated photoperiods. Three shoot apices were harvested for each biological replicate.

Gene expression analysis showed that FvFT1 expression increased in leaves of NOR8 and FvGI‐OE lines towards longer daylengths, being higher in FvGI‐OE lines than in the wild‐type in all photoperiods at ZT4 (Fig. 2c). Higher FvFT1 expression was associated with elevated FvSOC1 and FvTFL1 expression in the shoot apices (Fig. 2d,e). Both FvSOC1 and FvTFL1 were downregulated under shorter photoperiods, with the SD^8h causing stronger repression than the SD^12h. The strong downregulation of FvSOC1 and FvTFL1 was observed also in FvGI‐OE lines under SDs, but their expression levels remained always higher than in the wild‐type. Consistent with flowering results, the floral meristem identity gene FvAP1 was strongly upregulated in the end of the treatments in all genotypes under SD^8h and in NOR8 under SD^12h (Fig. 2f). The expression pattern of FvFT2 closely followed the expression of FvAP1 in the shoot apex, but its expression was not detected in leaves of NOR8 or its FvGI‐OE lines (Figs 2g, S10A,B), suggesting that FvFT2 may function in early floral development. Taken together, our results indicate that, in seasonal flowering F. vesca, FvGI activates the expression of FvFT1 in the leaves and FvSOC1 and FvTFL1 in the shoot apex, repressing floral transition in LDs.

Genetic interactions in the regulation of photoperiodic flowering

Gene expression analyses in transgenic lines suggested that FvGI functions upstream of FvCO and FvFT1 (Figs 1d–g, 2c, S8D–G). To test the genetic interactions of these genes, we crossed NOR8 FvGI‐OE line#9 and wild‐type NOR8 (control) with H4, and FvCO‐ and FvFT1‐RNAi lines in H4 background. Half of the NOR8 FvGI‐OE × H4 F₁ hybrids remained vegetative after the SD^12h treatment, and the rest of the plants flowered later than NOR8 × H4 controls (Figs 3a,b, S11), confirming that FvGI‐OE represses flowering in the hybrids. However, the silencing of FvCO or FvFT1 resulted in equal early flowering in both NOR8 and NOR8 FvGI‐OE hybrids, demonstrating that both FvCO and FvFT1 are required for the late flowering of NOR8 FvGI‐OE plants. While NOR8 FvGI‐OE × H4 hybrids showed elevated FvFT1 expression levels compared with the NOR8 × H4 hybrids, silencing of FvCO rendered FvFT1 undetectable regardless of FvGI‐OE, indicating that the activation of FvFT1 by FvGI requires FvCO (Fig. 3c,d). Furthermore, crossing of NOR8 or NOR8 FvGI‐OE line with H4 FvFT1‐RNAi line did not affect FvCO expression. These results support the presence of FvGI‐FvCO‐FvFT1 photoperiodic flowering pathway in F. vesca, where FvCO and FvFT1 are fully epistatic to FvGI.

Fig. 3.

The FvGI–FvCO–FvFT1 pathway regulates flowering time in seasonal flowering Fragaria vesca. (a) Flowering time of F₁ hybrids of NOR8, H4 and their indicated transgenic lines shown as the number of leaves before the terminal inflorescence. All hybrid plants contain one functional FvTFL1 allele. The upper, middle and lower horizontal lines indicate the 75^th percentile, median and 25^th percentile, respectively. The whiskers indicate minimum and maximum values. Each point represents measurement from an individual plant. Four‐week‐old plants, clonally propagated from hybrid seedlings, were subjected to 12‐h short days (SD^12h) at 17°C for 8 wk followed by phenotyping under 18‐h long days (LD^18h). OE, overexpression. Numbers under the boxplot indicate the number of flowered plants per the total number of plants in each group. Different letters indicate statistically significant differences as determined by one‐way ANOVA followed by Tukey's HSD multiple comparisons test (P < 0.05). (b–d) Relative expression of FvGI (b), FvCO (c) and FvFT1 (d) in leaves of F₁ lines. Leaf samples containing one leaflet per biological replicate were harvested at zeitgeber time 4 under LD^18h conditions. Data are mean ± SD (n = 4–5).

FvTFL1 reverses the effect of FvFT1 on flowering time

Because FvGI‐OE was found to activate FvFT1 in the leaves and FvTFL1 in the shoot apex of seasonal flowering F. vesca (Fig. 2c,e), we tested if FvFT1 activated FvTFL1. To this end, we crossed seasonal flowering accessions, the reference accession FIN56 and the Italian accession IT17 with naturally high FvFT1 expression level (Fig. S6B), with H4 and H4 FvFT1‐RNAi line and analysed the F₁ hybrids. FvFT1‐RNAi hybrids with strongly reduced FvFT1 mRNA levels flowered early after SD^12h treatment, and 40% of IT17 × H4 FvFT1‐RNAi and 67% of FIN56 × H4 FvFT1‐RNAi plants flowered also under noninductive LD^18h conditions (Fig. 4a,b). The expression levels of FvSOC1 and FvTFL1 were lower in the shoot apices of FvFT1‐RNAi hybrids than in the control plants in LD^18h conditions (Figs 4c, S12A,B). After 3 wk of SD^12h treatment, FvTFL1 expression decreased in all plants but remained higher in crosses with the wild‐type H4 compared with H4 FvFT1‐RNAi hybrids. At the same time, FvAP1 exhibited high expression levels only in the FvFT1‐RNAi hybrids, confirming that flower induction had happened before this time point in these plants but not in the wild‐type hybrids (Fig. 4d). Taken together, our results suggest that FvFT1 activates FvTFL1 expression in the shoot apex to repress flowering in seasonal flowering F. vesca, and plants with higher FvFT1 and FvTFL1 mRNA levels require longer inductive treatment to reduce the expression level of FvTFL1 below a critical threshold for floral induction.

Fig. 4.

FvFT1/FvTFL1 epistasis controls flowering in Fragaria vesca. (a) Flowering time of F₁ hybrids between the seasonal flowering genotype IT17 or FIN56 with functional FvTFL1 alleles and the perpetual flowering genotype H4 (fvtfl1) carrying FvFT1‐RNAi construct under 18‐h long days (LD^18h) and after 6 wk under 12‐h short‐day (SD^12h) conditions. F₁ crosses with nontransgenic H4 plants were used as controls. Asterisks indicate significant differences between groups at selected time points (***, P < 0.001; **, P < 0.01; *, P < 0.05; t‐test). (b–d) Expression of FvFT1 (b) in leaf samples and FvTFL1 (c) and FvAP1 (d) in shoot apices under LD^18h conditions and after 3 wk of indicated treatments at zeitgeber time 4 (ZT4) (b) or ZT3‐ZT5 (c, d) time points. Inset panel in (d) shows a close view of initial (before the start of treatment) expression levels of FvAP1 at week 0. (e) Flowering time of IT17 × FvFT1‐RNAi F₂ plants segregating for FvFT1‐RNAi construct and FvTFL1 and fvtfl1 alleles after 8 wk under SD^12h conditions. y‐axis shows days since the beginning of the treatments; dashed grey line marks the end of the indicated treatments after which all the plants were moved to LD^18h conditions. Numbers under x‐axis denote the number of flowered plants (n) and total number of plants (N) in each group. (f) Flowering time of H4 and FIN56 wild‐type plants and their FvFT1 overexpression (FvFT1‐OE) lines under LD^18h conditions. In (e, f), data are mean ± SD (n = 10). Asterisks (e) or different letters (f) indicate the significant differences between selected groups (***, P < 0.001; **, P < 0.01; *, P < 0.05; Tukey's HSD test).

To obtain direct evidence of the interaction of the photoperiodic pathway with FvTFL1 we segregated FvTFL1 alleles and FvFT1‐RNAi insertion in IT17 × H4 FvFT1‐RNAi F₂ lines and observed the flowering time. Under both LD^18h and SD^12h conditions, fvtfl1 homozygous mutants with the FvFT1‐RNAi construct flowered later than the fvtfl1 mutants without the FvFT1‐RNAi construct, confirming that FvFT1 activates flowering in the absence of functional FvTFL1 (Fig. 4e). However, the presence of one or two alleles encoding functional FvTFL1 resulted in SD‐dependent flowering and reverted the function of FvFT1; the plants with the FvFT1‐RNAi construct flowered earlier than the controls. In addition, approximately half of the FvFT1‐silenced plants homozygous or heterozygous for functional FvTFL1 flowered even under noninductive LD conditions. A higher proportion of flowering plants and earlier flowering of FvTFL1/fvtfl1 heterozygotes compared with plants containing two wild‐type alleles supported the dosage‐dependent effect of FvTFL1 on flowering in F. vesca (Fig. 4e).

Altogether, our results indicate that flowering regulation in F. vesca involves an epistatic interaction between the FvGI‐FvCO‐FvFT1 photoperiodic pathway and FvTFL1. When the functional FvTFL1 allele is present, FvFT1 prevents flowering in LDs by promoting FvTFL1 expression, while in the absence of functional FvTFL1, FvFT1 functions as a florigen that activates the expression of FvAP1, thus promoting flowering. Even in plants with functional FvTFL1, FvFT1 retains florigenic activity, as its RNAi silencing slightly reduces FvAP1 expression (Figs 4d, S12C), but this activity is masked by the FvTFL1 activation. Moreover, the expression of FvFT1 specifically in the leaves (Koskela et al., ²⁰¹²) is critical for the flowering repression mechanism presented here because ectopic expression of FvFT1 by the 35S promoter induces flowering even in the presence of FvTFL1 (Fig. 4f).

FvFT1 contributes to natural variation in flowering time

Finally, we explored natural flowering time variation in a European collection of F. vesca accessions that were recently shown to form separate eastern and western genetic clusters occupying different climatic niches (Toivainen et al., ²⁰²⁶). After flower induction under SD^12h in glasshouse or field conditions in autumn in Helsinki, a wide variation in flowering time was observed, with the eastern accessions flowering significantly earlier than the western ones, suggesting adaptive variation (Fig. 5a–c).

Fig. 5.

Natural variation of flowering time correlates with FvFT1 and FvTFL1 expression in European Fragaria vesca accessions. (a) Map showing collection sites of woodland strawberry accessions and their flowering time and genetic clusters. Accessions were divided into eastern and western genetic clusters according to Toivainen et al. (2026). (b, c) Boxplots showing differences in flowering time between the accessions of eastern and western genetic clusters after flower induction under 12‐h short‐day (SD^12h) conditions in a glasshouse (b; 179 accessions) or in the field (c; 181 accessions). The upper, middle and lower horizontal lines represent the 75^th percentile, median and 25^th percentile, respectively. The whiskers represent the minimum and maximum. Each point represents the mean value for the individual accession (n = 2–5). (d) Differentially expressed genes (DEGs) in leaves of six western and six eastern accessions collected under 18‐h long days (LD^18h) at zeitgeber time 4 (ZT4). (e) Mean expression levels of FvFT1 in 25 eastern and 23 western accessions (dots) (LD^18h, ZT4). Error bars represent ±SD. P‐values in (b, c, e) were estimated using Student's t‐test. (f) Pearson's correlation between the relative expression of FvFT1 (data from panel e) in leaves and FvTFL1 in shoot apices (LD^18h, ZT3‐5) in 21 eastern and 19 western accessions. (g, h) Pearson's correlation between flowering time and FvFT1 (g) or FvTFL1 (h) expression levels using the expression data shown in (e, f). P‐values in (f–h) were estimated using Pearson's correlation test, Bonferroni adjusted.

RNA sequencing from leaves of six eastern and six western accessions grown under LD^18h conditions identified 955 genes differentially expressed between the genetic clusters (Fig. S1; Dataset S1). Among these, four candidate flowering time genes were found including FvTPS1 (TREHALOSE‐6‐PHOSPHATE SYNTHASE1, FvH4_3g06590), FvLUX (LUX ARRHYTHMO, FvH4_3g36890), FvFUL1 (FRUITFUL, FvH4_5g13500) and FvFT1 (Figs 5d, S13A). Both LUX and TPS1 are known to affect FT expression in Arabidopsis (Hazen et al., ²⁰⁰⁵; Wahl et al., ²⁰¹³) and an FUL homolog in barley (Sasani et al., ²⁰⁰⁹), while their functions in F. vesca remain to be clarified. The expression of FvFT1 was typically lower in the accessions of the earlier flowering eastern cluster (Fig. S13A). We verified this by RT‐qPCR analysis in a larger set of accessions (25 eastern and 23 western), spanning the geographical range of European F. vesca, and found higher FvFT1 expression levels in the western than the eastern accessions at ZT4 under LD^18h conditions (Fig. 5e). Furthermore, we found a strong correlation in FvFT1 expression levels between ZT4 and ZT16 time points (Fig. S6C), suggesting that the observed differences at ZT4 are not time point specific. The analysis of shoot apex samples revealed that the expression of FvFT1 in leaves of accessions correlated with FvTFL1 expression levels in the shoot apex under LD^18h conditions (Fig. 5f). Increasing expression of both genes under LD^18h, before flower inductive conditions, was associated with delayed flowering after the subsequent flower induction under SD^12h conditions or in the field in autumn (Figs 5g,h, S13B). Although FvGI and FvCO expression levels did not consistently differ between the two clusters in the RNAseq data (Fig. S13A), their expressions were correlated among a larger set of accessions, as were FvCO and FvFT1 (Fig. S13C). Altogether, these findings indicate that the FvGI–FvCO–FvFT1–FvTFL1 photoperiodic pathway contributes to flowering time variation across geographic and climatic ranges in F. vesca accessions, with strong correlations between flowering time and latitude or winter temperature of plant collection sites (Fig. S13D).

Discussion

Here, we studied photoperiodic flowering in F. vesca, a perennial species with a wide geographical distribution in the northern hemisphere (Hilmarsson et al., ²⁰¹⁷; Toivainen et al., ²⁰²⁶). Previous studies showed that the floral repressor FvTFL1 confers the SD requirement of flower induction in F. vesca, while the fvtfl1 mutant flowers early in LDs (Koskela et al., ²⁰¹²). By testing FvGI function and its genetic interactions with previously reported genes (Koskela et al., ²⁰¹²; Kurokura et al., ²⁰¹⁷), we show that the FvGI‐FvCO‐FvFT1 photoperiodic pathway expressed in the leaves is conserved in F. vesca. This pathway activates LD‐flowering in the fvtfl1 mutant, while in seasonal flowering, wild‐type plants the pathway is rewired to activate FvTFL1 that represses flowering in the shoot apex, reversing the photoperiodic response. Due to this connection, wild‐type plants require SDs for flower induction that occurs after the downregulation of FvFT1 and FvTFL1 in autumn. In natural accessions, FvFT1 expression correlates positively with FvTFL1 mRNA levels and is associated with differences in flowering time in plants collected across Europe, suggesting that a flower‐repressive photoperiodic pathway may contribute to flowering time adaptation across geographic and climatic ranges.

Epistatic interaction between FvFT1 and FvTFL1 reverses photoperiodic response

Homologues of FT have been shown to induce flowering in both LD and SD plants (Kobayashi et al., ¹⁹⁹⁹; Faure et al., ²⁰⁰⁷; Tamaki et al., ²⁰⁰⁷; Komiya et al., ²⁰⁰⁹; Kong et al., ²⁰¹⁰), while TFL1 represses flowering (Bradley et al., ¹⁹⁹⁷; Wang et al., ²⁰¹¹), possibly through competing for binding with a bZIP transcription factor FD that is needed for the regulation of their target genes (Zhu et al., ²⁰²⁰). Multiple lines of evidence suggest different mechanisms in the regulation of photoperiodic flowering in F. vesca: FvFT1 is expressed in leaves in LDs when FvTFL1 is also activated in the shoot apex to repress flowering, and floral transition occurs in SDs when FvTFL1 is downregulated and FvFT1 expression is undetectable (Koskela et al., ²⁰¹²). In fvtfl1 mutant, however, FvFT1 promotes flowering in LDs and is fully responsible for this photoperiodic response (Koskela et al., ²⁰¹²; Rantanen et al., ²⁰¹⁴; Kurokura et al., ²⁰¹⁷). By segregating FvTFL1 alleles and the FvFT1‐RNAi insertion in an F₂ crossing population we demonstrated that FvTFL1 is epistatic to FvFT1 in the photoperiodic flowering. All F₂ lines homozygous for fvtfl1 alleles flowered early in LDs, with RNAi‐silencing of FvFT1 delaying flowering confirming that FvFT1 promotes flowering in the fvtfl1 mutant. By contrast, when one or two functional FvTFL1 alleles were present, flowering became SD‐dependent, and the effect of FvFT1 was reversed to a repressor of flowering. FvFT1‐RNAi plants flowered earlier than controls after the SD treatment, and a substantial proportion of FvFT1‐RNAi plants flowered even in noninductive LDs.

To understand the mechanism of the epistatic interaction between FvFT1 and FvTFL1, we studied gene expression in FvFT1‐RNAi lines and natural accessions of F. vesca. We found lower expression level of FvTFL1 in FvFT1‐RNAi lines under both LD^18h and SD^12h conditions compared with wild‐type plants and a positive correlation between FvFT1 and FvTFL1 mRNA levels in natural accessions, suggesting that FvFT1 activates FvTFL1 in the SAM under LDs to repress flowering in F. vesca. Further studies are needed to understand how this activation occurs at the molecular level, including a role of FvSOC1 that was previously found to activate FvTFL1 expression (Mouhu et al., ²⁰¹³) (Figs 1h, S8F–I). Recent work in Arabidopsis showed that FT‐activated LEAFY (LFY) upregulated TFL1 to maintain indeterminate inflorescence meristem (Huang et al., ²⁰²⁶). Further analysis of four LFY genes present in F. vesca genome is needed for testing if LFY has a similar role in F. vesca (Zhang et al., ²⁰²³).

Several lines of evidence suggested that the FvFT1/FvTFL1 module has a quantitative effect on floral transition and flowering time: (1) FvFT1‐RNAi lines with low FvFT1 and FvTFL1 expression levels showed early upregulation of floral marker gene FvAP1 under SDs and early flowering compared with wild‐type plants (Fig. 4), (2) FvGI‐OE lines with increased FvFT1 and FvTFL1 mRNA levels required shorter photoperiod for floral transition and flowered later than wild‐type (Fig. 2), and (3) FvTFL1/fvtfl1 heterozygote plants flowered earlier than FvTFL1 homozygotes (Samad et al., ²⁰¹⁷, Fig. 4e). Overall, these data support the hypothesis that the level of FvFT1/FvTFL1 expression drives the strength of floral repression in LDs and affects SD requirements of floral transition. However, the mechanism of flower induction in SDs remains elusive. Previous study found the expression of FvFT2 in leaves and suggested that FvFT2 could induce flowering in SDs (Gaston et al., ²⁰²¹). We could not detect FvFT2 expression in the leaves of tested accessions and transgenic lines under any conditions, but FvFT2 was upregulated in the shoot apex in parallel with FvAP1 indicating that it plays a role in early floral development instead of floral induction. Possible role of SQUAMOSA PROMOTER‐BINDING PROTEIN‐LIKE (SPL) transcription factors that induce flowering in SDs in A. thaliana and A. alpina (Roggen et al., ²⁰²⁵) should be tested in F. vesca.

FvGI‐FvCO‐FvFT1 pathway controls photoperiodic flowering

CO/FT module has been shown to control photoperiodic flowering in annual models Arabidopsis and rice (Valverde et al., ²⁰⁰⁴; Ishikawa et al., ²⁰¹¹), and among perennials, in perpetual flowering mutants of F. vesca and A. alpina (fvtfl1 and pep1 mutants, respectively; Kurokura et al., ²⁰¹⁷; Sashidhar & Coupland, ²⁰²⁶). Here, we extended these studies to FvGI and demonstrated that all three genes are required for the photoperiodic flowering of F. vesca. In RNAi lines of FvGI, FvCO and FvFT1 in H4 background (fvtfl1 mutant; Koskela et al., ²⁰¹²), FvFT1 expression levels remained below/close to the detection limit in LDs, and the lines almost completely lost the early LD flowering observed in H4. FvGI‐OE, by contrast, increased FvFT1 expression in leaves resulting in early flowering specifically in SDs in H4 background, and a similar SD‐specific response was previously reported in FvCO‐OE lines (Kurokura et al., ²⁰¹⁷). The effect of increasing FvFT1 expression seems to be quantitative; SD‐grown H4 with undetectable FvFT1 expression flowers late, mild activation of FvFT1 in FvGI‐OE plants in SD results in intermediate flowering time, while H4 grown in LDs flowers early because of several times higher FvFT1 mRNA levels compared with FvGI‐OE plants grown in SDs (Fig. 1f,g). The florigenic effect of FvFT1 is possibly saturated in H4 grown under LD^18h conditions because FvGI‐OE does not advance flowering further in LDs. However, FvFT1‐OE plants with constitutive FvFT1 expression lose their photoperiodic response and plants flower early under both SD and LD conditions (Fig. 4f; Rantanen et al., ²⁰¹⁴).

Overexpression of FvGI increased FvFT1 expression also in plants possessing functional FvTFL1 (NOR8, FIN56), but resulted in delayed SD‐flowering, consistent with the epistatic effect of FvTFL1 to FvFT1. Furthermore, the late flowering phenotype of NOR8 FvGI‐OE plants was completely reversed by silencing of FvCO or FvFT1 that also reduced FvFT1 expression close to the detection limit, demonstrating that FvCO and FvFT1 were epistatic to FvGI. Together with previous research on FvCO function (Kurokura et al., ²⁰¹⁷), these findings established the FvGI‐FvCO‐FvFT1 pathway with a major role in mediating the LD signal to activate flowering in F. vesca fvtfl1 mutants. In wild‐type plants, this pathway activates FvTFL1, and SDs are required to suppress the whole pathway to induce flowering. In other SD plants, different mechanisms repress flowering in LDs. In rice, Ghd7 blocks Hd1 (CO‐homolog) function and in soybean, E1 and GmCOL1a/b repress FT2a/5a and flowering (Xia et al., ²⁰¹²; Cao et al., ²⁰¹⁵; Xu et al., ²⁰¹⁵; Zong et al., ²⁰²¹).

Sequences and diurnal expression rhythms of GI homologs are largely conserved in both LD and SD plants, suggesting broadly conserved roles in photoperiodic signalling (Fowler et al., ¹⁹⁹⁹; Hayama et al., ²⁰⁰³; Hecht et al., ²⁰⁰⁷; Li et al., ²⁰¹³; Deng et al., ²⁰¹⁵, Fig. S2). In Arabidopsis, GI promotes CO transcription in LDs, allowing CO protein to peak in the evening to activate FT (Valverde et al., ²⁰⁰⁴; Sawa et al., ²⁰⁰⁷). However, FvCO mRNA peaks at dawn with a broader expression phase in LDs, the pattern resembling that of Arabidopsis COL1/COL2, pea PsCOLa and soybean GmCOL1a/b (Ledger et al., ²⁰⁰¹; Hecht et al., ²⁰⁰⁷; Wu et al., ²⁰¹⁴). This FvCO expression pattern is needed for normal FvFT1 expression peaks in the morning and late evening, because both peaks are lost in FvCO RNAi lines (Kurokura et al., ²⁰¹⁷). Similarly, FvFT1 expression peaks were lost in FvGI‐RNAi lines, although FvCO mRNA levels were not clearly affected, indicating that FvGI regulates FvFT1 expression mainly through posttranscriptional regulation of FvCO, which is further supported by the observations that both FvGI‐OE and FvCO‐OE had only a weak effect on FvFT1 mRNA levels in SDs, compared with LDs (Kurokura et al., ²⁰¹⁷, Fig. 1b–g). Such a mechanism is consistent with an LD‐dependent stabilization of CO in Arabidopsis that involves FKF1 and GI (Valverde et al., ²⁰⁰⁴; Song et al., ^{2012, 2014}; Hwang et al., ²⁰¹⁹). In conclusion, we show strong evidence for the FvGI–FvCO–FvFT1 photoperiodic pathway in F. vesca. Further studies are needed to understand posttranscriptional regulation of FvCO and a possible role of FvFKF1, that has overlapping expression phase with FvGI (Kurokura et al., ²⁰¹⁷), in photoperiodic flowering of F. vesca.

Flowering time variation in European F. vesca

Fragaria vesca occupies a wide geographical range from Mediterranean to arctic region in the northern hemisphere (Hilmarsson et al., ²⁰¹⁷). Our data indicate that flowering time variation in F. vesca accessions collected across Europe is associated with their population structure, which consists of two major genetic clusters. These western and eastern clusters originate from different glacial refugia and inhabit distinct climates across a temperature seasonality gradient (Toivainen et al., ²⁰²⁶). We found that eastern accessions flowered significantly earlier than western accessions under both SD and field conditions, with flowering time most strongly correlating with mean annual and winter temperatures of plant collection sites. Also, the latitudinal correlation was strong, with accessions from arctic regions, particularly in the east, showing extreme early flowering. Such a pattern of flowering time variation can be attributed to the length of the growing season that creates strong selection pressure on reproductive traits (Austen et al., ²⁰¹⁷), suggesting adaptive divergence in flowering regulation.

Reduced activity of the photoperiodic pathway was suggested to contribute to adaptation to short growing seasons in both annual and perennial plants (Hyun et al., ²⁰¹⁹). Consistent with this idea, we found that decreasing expression levels of FvFT1 and FvTFL1 under LD^18h conditions correlated with earlier flowering after subsequent flower induction treatments in European F. vesca accessions. Although the measured LD expression states of FvFT1 and FvTFL1 do not mechanistically govern what happens during SD induction, together with transgenic data discussed above, they support the hypothesis that the strength of the FvFT1/FvTFL1‐driven LD repression system may contribute to flowering time variation. While FvFT1 and FvTFL1 mRNA levels showed a strong positive correlation with clear association with genetic clusters, weaker positive FvGI/FvCO and FvCO/FvFT1 correlations were found, and there was no clear east–west pattern in FvGI or FvCO expression, suggesting that FvFT1 expression variation associated with geography and population structure depends on mechanisms other than the regulation of FvGI or FvCO transcription. However, arctic accessions (NOR14 and NOR15) from the highly differentiated Kåfjord population (Toivainen et al., ²⁰²⁶), that flowered first in different experiments, exhibited clearly reduced FvCO expression and extremely low FvFT1 mRNA levels, reflecting population‐specific responses to photoperiods. A previous study showed substantial shifts in the timing of GI expression between natural Arabidopsis accessions (De Montaigu et al., ²⁰¹⁵), highlighting the importance of studying multiple time points when comparing natural accessions. Although we found a strong correlation between ZT4 and ZT16 expression levels of FvFT1 among 13 accessions, more detailed analysis is needed to explore possible differences in the circadian phase, entrainment, or waveform timing of flowering genes that may contribute to observed flowering time variation. While FvGI and FvCO play major roles as activators of FvFT1 expression, studies on additional candidate genes, including FvTPS1, FvLUX and FvFUL1 identified in our RNA sequencing work, are needed for more detailed understanding of natural flowering time variation and underlying regulatory network in F. vesca. Other mechanisms may include cis‐regulatory variation of FT that has been reported to contribute to flowering time variation in Arabidopsis and soybean (Liu et al., ²⁰¹⁴; Rosas et al., ²⁰¹⁴; Bao et al., ²⁰¹⁹; Chen et al., ²⁰²⁰).

In conclusion, FvGI and FvCO are needed for the activation of FvFT1 in leaves in LDs, and they constitute a major photoperiodic response pathway of the species. In the perpetual flowering fvtfl1 mutant, this pathway activates FvSOC1 and FvAP1 in the shoot apex, resulting in rapid flowering in LDs (Fig. 1; Kurokura et al., ²⁰¹⁷). Functional FvTFL1, that is present in all natural accessions tested in this work, reverses the photoperiodic response (Koskela et al., ²⁰¹²) because the photoperiodic pathway activates FvTFL1 at least partially through FvSOC1, overriding FvAP1 activation and repressing flowering (Fig. 4; Mouhu et al., ²⁰¹³). Finally, in natural accessions, expression levels of FvFT1 and FvTFL1 were associated with geographic flowering time variation across Europe, with decreasing FvFT1 and FvTFL1 mRNA levels and earlier flowering towards north and east. Further studies are needed to explore possible contributions of these genes to flowering time adaptation across climatic gradients.

Competing interests

None declared.

Author contributions

QZ, SL, EK and TH designed the research. QZ, SL, TT, TK and GF carried out the experimental work. SL, TT and GF analysed RNA sequencing data. TH, EK and PE supervised the study. The manuscript was written by QZ and TH with input from all authors.

Disclaimer

The New Phytologist Foundation remains neutral with regard to jurisdictional claims in maps and in any institutional affiliations.

Supporting information

Dataset S1 RNA sequencing data of 12 Fragaria vesca accessions.

Fig. S1 Clustering of RNA sequencing samples.

Fig. S2 Phylogenetic analysis and multiple sequence alignment of GIGANTEA homologs.

Fig. S3 Gene expression in Fragaria vesca accession H4 and transgenic lines.

Fig. S4 Flowering time of Fragaria vesca accession H4 and indicated transgenic lines.

Fig. S5 Diurnal gene expression rhythms in Fragaria vesca accession H4 and transgenic lines.

Fig. S6 Flowering time and FvFT1 expression in natural accessions of Fragaria vesca.

Fig. S7 Gene expression and phenotypes of indicated Fragaria vesca genotypes.

Fig. S8 Expression of flowering time genes in Fragaria vesca accession NOR8 and FvGI overexpression lines.

Fig. S9 Scanning electron microscope images of Fragaria vesca meristems.

Fig. S10 Expression of FvFT1 and FvFT2 in Fragaria vesca accession NOR8 and FvGI overexpression lines.

Fig. S11 Flowering time of indicated Fragaria vesca hybrids.

Fig. S12 Expression of flowering genes in indicated Fragaria vesca hybrids.

Fig. S13 Comparison of gene expression levels in Fragaria vesca accessions and correlations between gene expression, flowering time and bioclimatic and geographical variables.

Table S1 Fragaria vesca hybrids used in this study.

Table S2 Fragaria vesca accessions used in this study.

Table S3 Primers used in this study.

Table S4 Protein sequences used for the phylogenetic analysis of GI.

Please note: Wiley is not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

Data availability

All the raw sequence reads from RNA sequencing are available at NCBI Sequence Read Archive under Bio‐project ID PRJNA1503464.