2026 年 76 巻 3 号 p. 263-274
The mechanisms of flowering-time (FT) regulation in cultivated chrysanthemum (Chrysanthemum morifolium) are essential for breeders owing to their significant effects on cultivation. The wild diploid Chrysanthemum seticuspe has the same short-day flowering habits as cultivated chrysanthemum. We investigated an F2 segregating population derived from crosses of accessions with differences in FT for quantitative trait locus (QTL) mapping. Three major QTLs for FT were mapped, each overlapping with QTLs for other flowering traits: one QTL associated with a longer critical daylength and two QTLs linked to the ability to flower rapidly following floral transition under short-day conditions, one of which also contributes to flowering competence at high temperatures. These traits collectively influence the regulation of FT. Altered expression patterns of circadian-clock-related genes in SoS4-4, the early-flowering parental line, suggested altered rhythmicity of the circadian clock. Circadian-clock-related genes and a photoreceptor gene were also found in the FT QTL interval. These genes may be responsible for altering the circadian clock and causing variation in FT by changing the critical daylength.
Chrysanthemum species are herbaceous perennials distributed in East Asia. Most are short-day (SD) plants, blooming in autumn under SD conditions. The transition from vegetative to reproductive phase is crucial in the plant life cycle. Timely autumn flowering increases reproductive success before winter and is vital for adaptation in temperate climates. Cultivated chrysanthemum (Chrysanthemum morifolium Ramat.), a hexaploid species, is economically important as an ornamental crop, with flowering controlled to meet market demands. Flowering can be controlled by manipulating daylength using blackout curtains and light exposure in protected cultivation. In Japan, a wide range of cultivars that flower from May to December are grown in open fields (Kawata and Funakoshi 1988). Understanding the mechanism controlling flowering time (FT) is key to chrysanthemum production and breeding.
Flowering in chrysanthemum is complex and is influenced by multiple environmental factors. Daylength is a major factor (Kawata and Funakoshi 1988, Langton 1977, Okada 1957). Flowering is regulated by a daylength threshold, the ‘critical daylength’, and under natural conditions is promoted as daylength shortens. Cultivars with a longer critical daylength flower earlier. Other factors such as ambient temperature, temperature memory (the temperature of prior growing conditions), nutrition and plant hormones affect chrysanthemum flowering (Harada and Nitsch 1959, Hisamatsu et al. 2017, Schwabe 1950, Tjia et al. 1969). It is proposed that FT is determined mainly by the combination of daylength and temperature, because it was explained by the relationship between the photoperiodic and temperature responses among cultivars with a wide range of FTs (Kawata and Funakoshi 1988).
In many plant species, genetic factors regulating FT are identified by quantitative trait locus (QTL) analysis (Koornneef et al. 2004, Matsubara and Yano 2018). Genetic mapping studies of cultivated chrysanthemum have detected loci controlling FT (Su et al. 2024, van Geest et al. 2017), but analysis is challenging owing to the hexaploid genome. To avoid this problem, the wild diploid Chrysanthemum seticuspe (Maxim.) Hand.-Mazz. (2n = 18) is used in genomic and molecular studies as a model species (Hirakawa et al. 2019, Nakano et al. 2019, 2021). It has similar growth and SD-flowering habits to cultivated chrysanthemum, flowering from October to November (National BioResource Project [Chrysanthemum]: https://shigen.nig.ac.jp/chrysanthemum/). Using C. seticuspe for QTL studies allows efficient genetic analysis of FT and helps us understand how environmental factors influence flowering. Here we explored potential mechanisms underlying chrysanthemum FT regulation, such as variations in response to daylength and candidate genes.
We grew F2 plants developed from a cross between the C. seticuspe lines SoS4-4 and XMRS10. SoS4-4 was developed from the early-flowering accession Matsukawa1 (Fukai and Miyatake 2005) via four generations of selfing. XMRS10 is a late-flowering line developed from the accession AEV02 via five generations of selfing (Nakano et al. 2019). XMRS10 was provided by National BioResource Project Chrysanthemum (https://shigen.nig.ac.jp/chrysanthemum). One F1 plant was self-pollinated to generate the F2 population. A total of 186 F2 individuals, each parental line and F1 were grown as stock plants in plastic pots (12-cm i.d., 1 seedling per pot) containing a commercial horticultural soil (Kureha-Engei-Baido; Kureha Chemical Co. Ltd., Tochigi, Japan). The plants were maintained in the vegetative state in a glasshouse (heated at <18°C and ventilated at >25°C) under natural photoperiod plus a 6-h night break (22:00–04:00) to prevent floral transition. The night-break treatment was conducted using white fluorescent lamps (EFR25ED/22; Toshiba Lighting & Technology Corporation, Kanagawa, Japan) at a photon flux density (PFD) of 3.3 μmol m–2 s–1. Stock plants were exposed to natural low temperatures (<10°C) to break dormancy in an unheated plastic greenhouse during winter (January–March). Rooted cuttings were prepared from these stock plants for phenotypic studies of flowering traits at the NARO Institute of Vegetable and Floriculture Sciences (NIFTS; Ibaraki, Japan; 36°2ʹ46ʺN, 140°5ʹ55ʺE).
Phenotypic data collection Flowering timeFT was determined in open fields under natural temperatures and daylength. On 4 July 2017, five rooted cuttings per genotype were planted at 15-cm intervals in two rows 45 cm apart. Ten days after planting, the plants were pinched at the internode between the third and fourth leaves from the base to encourage the growth of lateral shoots. This technique increased growth uniformity by synchronizing the outgrowth of shoots. We recorded the date of first anthesis (i.e., when the ray florets of the terminal flower became vertically oriented) of each plant. FT was scored as the number of days after 4 July. In 2018, one rooted cutting per genotype was planted. Details are shown in Table 1.
| Trait | Year | Date of transplanting (MM/DD) |
Date of starting treatment (MM/DD) |
Daylength (h, Min–Max) |
Temperature (Min–Max) |
Growth condition |
n |
|---|---|---|---|---|---|---|---|
| FT | 2017 | 07/04 | – | 10.40–14.58 | Avg 21.8°C (2.3–35.0°C) | Field | 5 |
| FT | 2018 | 07/04 | – | 10.40–14.58 | Avg 23.2°C (3.9–37.4°C) | Field | 1 |
| RTopt | 2017 | 08/03 | 09/19 | 10.00–12.25 | 20–23°C | Glasshouse | 4 |
| RTopt | 2018 | 08/03 | 09/19 | 10.00–12.25 | 20–23°C | Glasshouse | 2 |
| RTopt | 2020 | 08/11 | 09/23 | 10.00–11.00a | 20–23°C | Glasshouse | 2 |
| RTHT | 2017 | 06/20 | 07/06 | 11.00a | Avg 26.2°C | Glasshouse | 3 |
| RTHT | 2018 | 06/20 | 07/06 | 11.00a | Avg 28.1°C | Glasshouse | 2 |
| RT13h | 2018 | 07/30 | 09/14 | 13.00a | 20–23°C | Glasshouse | 2 |
| RT13h | 2020 | 03/10 | 04/10 | 13.00a | 20–23°C | Glasshouse | 2 |
a Daylength was controlled using blackout curtains or supplementary lighting.
Reaction time (RT) is the number of days from SD perception to flowering. We investigated RT under optimal temperature and SD conditions for flowering promotion (RTopt). On 3 August 2017, four rooted cuttings per genotype were planted at 16 cm × 16 cm in a plastic house. Plants were pinched 10 days after planting. Lateral shoots, except for the axillary shoot of the topmost leaf on the main stem, were removed 2 weeks later. Plants were grown with a daily 6-h night break to prevent flowering until 19 September 2017. Under the following natural SD conditions (photoperiod 12.25–10.00 h) to 26 November, we recorded the date of first anthesis of each plant. The greenhouse was heated at <20°C and ventilated at >23°C. RTopt per genotype was calculated as the difference between the end date of night-break treatment (19 September) and anthesis date (Fig. 1). In 2018 and 2020, two rooted cuttings per genotype were planted. In 2020, plants were grown under ≤11-h photoperiod by closing the blackout curtains from 1700 to 0600 after the end date of night-break treatment.
Measurement of flowering traits of the SoS4-4 × XMRS10 F2 population. Reaction time (RTopt), number of days to flowering under optimal temperature and short-day conditions; RTHT, flowering delay caused by high temperature; RT13h, flowering delay by suboptimal 13-h photoperiod.
High temperature (HT) delays chrysanthemum flowering (Nakano et al. 2013, Whealy et al. 1987). We investigated its effect on RT in the glasshouse in summer, where restricted ventilation raised temperatures. On 20 June 2017, three rooted cuttings per genotype were planted at 12 cm × 12 cm. Plants were pinched 10 days after planting. Lateral shoots, except for the axillary shoot of the topmost leaf on the main stem, were removed 2 weeks later. Plants were grown with a daily 6-h night break until 6 July 2017. Plants were then grown under the following HT conditions in summer with 11-h SD conditions achieved by closing blackout curtains from 1700 to 0600. The windows were closed at <22°C to maintain HT. We recorded the date of first anthesis of each plant. RT under HT (RTHT) per plant was the number of days from RTopt to the delayed anthesis date (Fig. 1). In 2018, the windows were closed at <25°C. The average temperature during the experimental period in the glasshouse was 26.2°C in 2017 and 28.1°C in 2018 (Table 1).
Reaction time under 13-h photoperiodWe investigated the effect of suboptimal daylength on RT. On 30 July 2018, two rooted cuttings per genotype were planted in plastic pots (30-cm i.d., 8 cuttings per pot) containing a commercial horticultural soil in the glasshouse. Plants were pinched 10 days after planting. Lateral shoots, except for the axillary shoot of the topmost leaf on the main stem, were removed 2 weeks later. Plants were grown vegetatively with a daily 6-h night break until 14 September 2018. Under the following suboptimal 13-h photoperiod from 14 September to 5 December, we recorded the date of first anthesis of each plant. To create a 13-h photoperiod, we used incandescent lamps to supply predawn lighting. The glasshouse was maintained at optimal temperature (20–23°C). RT under 13-h photoperiod (RT13h) per genotype was the number of days from RTopt to the delayed anthesis date (Fig. 1). In spring 2020, plants were planted at 14 cm × 14 cm in planting beds (90 cm W, 600 cm L, 18 cm D) in the glasshouse, and a 13-h photoperiod was created by closing blackout curtains from 1700 to 0600 (Table 1).
Spearman’s correlation coefficients for each trait in 2017 and 2018 were analysed using MS-Excel. The normality of data distribution for each trait was assessed by the Shapiro–Wilk test (P < 0.05) in EZR v. 1.54 software (Kanda 2013).
Linkage map construction and QTL analysisGenomic DNA was extracted from shoot tips (30 mg fresh weight) with a DNeasy Plant Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. ddRAD-Seq libraries were constructed with restriction enzymes PstI and MspI and sequenced by MiSeq at a read length of 251 nt. The reads were mapped onto the Gojo-0_v1 sequence (Nakano et al. 2021) in Bowtie2 software (Langmead and Salzberg 2012), and variant calls were performed in SAMtools v. 0.1.19 (Li et al. 2009a) and VarScan 2.3 (Koboldt et al. 2012) software as described by Shirasawa et al. (2016). Candidate SNPs were filtered under the following conditions: exclude indels, minimum depth = 5, minimum quality = 10, max missing = 0.5, minor allele frequency = 0.2, polymorphic between parental lines.
SNP markers with ≤10% missing genotypes were selected and divided into nine groups by alignment to chromosomes on Gojo-0_v1 for map construction. A linkage map was constructed in Lep-MAP3 software (Rastas 2017). The Filtering2 function of Lep-MAP3 was not used, because there were many SNPs with distorted distributions on chromosomes 1, 2, 5, 7 and 8. The SeparateChromosomes2 function was used to cluster groups at a logarithm of odds (LOD) score of 30. The OrderMarkers2 function was used to order the mapped markers within each linkage group (LG) by maximizing the best-likelihood positions of the markers and to compute genetic distance, with 20 cM as the cut-off threshold for LGs. LG numbering and direction were defined with reference to the reported chromosomes (Nakano et al. 2021). Relationships between SNP markers and traits were analysed by the non-parametric Kruskal–Wallis (K-W) test in MapQTL 6.0 software (van Ooijen 2009). Loci with larger K* values than the significance level (P < 0.00001, set as 0.05/6824 markers) were regarded as putative QTLs.
Genomic analysisWe searched non-synonymous sequence variants within flowering-related genes for QTL intervals. Whole-genome shotgun sequencing of the parental lines was performed on a 100-bp paired-end Illumina HiSeq 4000 sequencer. Sequencing reads were mapped onto Gojo-0_v1 in Bowtie2. Variant calling was based on the mapping result in SAMtools 0.1.19 software and variants were filtered in VarScan v. 2.3 software. Effects of variants on gene function were predicted in SnpEff v. 4.0 software (Cingolani et al. 2012). Of the genes harbouring SNPs with high or moderate effects of non-synonymous sequence variants on genes and proteins in Gojo-0_v1 genomic regions corresponding to QTL intervals, we identified flowering-related genes as C. seticuspe genes described in Hirakawa et al. (2019) and C. seticuspe homologues of Arabidopsis photoperiod and circadian clock genes as described in the FLOR-ID database of FT gene networks (Bouché et al. 2016).
Gene expression analysisWe quantified 24-h expression profiles of circadian-clock-related genes—CsLHY (LATE ELONGATED HYPOCOTYL, CsG_LG1.g30264.1, GenBank Accession No. AB733625), CsTOC1 (TIMING OF CAB EXPRESSION 1, CsG_LG6.g49179.1, AB733626) and CsGI (GIGANTEA, CsG_LG8.g71869.i1, AB733627)—in XMRS10 and SoS4-4 under continuous dark conditions (DD). Plants were grown in plastic pots (6-cm i.d., 1 rooted cutting per pot) and maintained in a growth chamber under a 16-h photoperiod under white fluorescent lamps at 70 μmol m–2 s–1 PFD (23°C, 80% relative humidity) for 6 weeks. At the end of the 6 weeks, plants were transferred to DD (23°C, 80% relative humidity). One fully expanded leaf per plant was harvested hourly over 24 h and immediately frozen in liquid nitrogen. Total RNA was extracted with an RNeasy Plant Mini Kit (Qiagen K.K., Tokyo, Japan) and treated with RNase-free DNase (Qiagen K.K.). From each sample, 100 ng of total RNA was reverse-transcribed using a PrimeScript RT Master Mix Perfect Real Time Kit (Takara Bio Inc., Shiga, Japan), with three biological replicates. The resulting cDNA was diluted 1:10 with distilled water. Quantitative PCR was performed using 1 μL cDNA template in 10 μL with TB Green Premix Ex Taq II Tli RNase H Plus (Takara Bio Inc.) on a Thermal Cycler Dice Real Time System III (Takara Bio Inc.) for 1 min of denaturation at 95°C, followed by 40 cycles of 95°C for 5 s and 60°C for 30 s, with the primers listed in Supplemental Table 1. The raw data were normalized to the level of expression of CsACTIN (CsG_LG2.g19422.i1, AB770470). The cycle threshold (CT) of CsACTIN did not differ among the samples. ΔCT of the genes of interest was calculated as CT – CTCsACTIN. The maximum ΔCT value was used as a reference and set at 1.0. Then ΔΔCT of each gene was calculated as ΔCTsample – ΔCTreference. The data are presented as 2–ΔΔCT (Livak and Schmittgen 2001).
Data availability statementThe data that support the findings of this study are available from the corresponding author, upon reasonable request.
The range of phenotypic variation within the F2 population was more or less continuous between the parental lines (Fig. 2). Distributions were strongly correlated between years, except RT13h at 0.52 (Fig. 2D). For every flowering trait, SoS4-4 had smaller values than XMRS10, and F1 plants had smaller values than the midpoint between the parents, except RTHT in 2017 (Fig. 2C). RTopt, RTHT and RT13h showed transgressive segregation. FTs were later in 2018 than in 2017. Distributions of RTopt were similar between 2017 and 2018 but were slightly smaller in 2020 (Fig. 2B). SoS4-4 flowered early under both HT and a 13-h photoperiod (Fig. 2C, 2D). The distribution varied between years: RTHT was larger in 2018 because the greenhouse temperature was higher (Table 1). However, data were correlated between years.
Distribution of phenotypes in the SoS4-4 × XMRS10 F2 population. (A) FT: number of days from planting to flowering under natural conditions in the open field; no data were available for F1 in 2018. (B) RTopt: number of days to flowering under optimal temperature and short-day conditions; no data were available for SoS4-4 in 2018. (C) RTHT: increase in number of days to flowering caused by high temperature. (D) RT13h: increase in number of days to flowering under 13-h photoperiod; no data were available for F1 in 2018 or for SoS4-4 in 2020. Distribution type by Shapiro–Wilk test is shown in each panel. Correlation coefficient of distributions between years is shown at the top of each panel. 
SoS4-4; 
XMRS10; 
F1.
Spearman’s correlation analysis suggested that 7 of the 9 possible trait pairs had significant (P < 0.001) correlations (Table 2). FT showed a positive correlation with the other three flowering traits. Two trait pairs, FT and RTopt in 2017 and 2018 (ρ = 0.71 and 0.51, respectively) and FT and RT13h in 2018 (ρ = 0.59) displayed moderate correlation coefficients, while a trait pair, FT and RTHT displayed a weak correlation coefficient in 2017 and 2018 (ρ = 0.31 and 0.36, respectively).
| Year | Traitsa | FT | RTopt | RTHT |
|---|---|---|---|---|
| 2017 | RTopt | 0.71* | ||
| RTHT | 0.31* | 0.30* | ||
| 2018 | RTopt | 0.51* | ||
| RTHT | 0.36* | 0.01 | ||
| RT13h | 0.59* | 0.14 | 0.27* |
a FT, flowering time; RTopt, days to flower under optimal SD and temperature conditions; RTHT, flowering delay caused by high temperature; RT13h, flowering delay caused by 13-h photoperiod.
* Significant at P < 0.001.
After trimming and removal of low-quality sequences, we obtained ~2.2 million high-quality reads per F2 plant. The reads were mapped onto the Gojo-0_v1 sequence with average mapping alignment rates of 91.2%. Between SoS4-4 and XMRS10, a total of 10,100 SNPs were identified. Among these, we used the genotypes of the F2 population at 6,824 polymorphic SNP sites to construct a genetic map. The map consisted of 780 loci and covered 878.50 cM in 9 LGs, consistent with the haploid number of the species, with an average locus interval of 1.13 cM (Table 3). Segregation distortion was significantly higher in LGs 1, 2, 7 and 8, where SoS4-4 homozygotes were reduced, and in LG 5, where XMRS10 homozygotes were reduced (Supplemental Table 2), indicating a biological bias in meiosis or chromosome pairing.
| Linkage group | Number of loci | Length (cM) |
Average bin interval (cM) | |
|---|---|---|---|---|
| SNPs | Bin | |||
| 1* | 437 | 58 | 79.36 | 1.37 |
| 2* | 933 | 71 | 88.83 | 1.25 |
| 3 | 695 | 87 | 89.22 | 1.03 |
| 4 | 658 | 69 | 123.12 | 1.78 |
| 5* | 971 | 109 | 82.06 | 0.75 |
| 6 | 901 | 117 | 123.00 | 1.05 |
| 7* | 643 | 84 | 88.98 | 1.06 |
| 8* | 854 | 89 | 103.83 | 1.17 |
| 9 | 732 | 96 | 100.11 | 1.04 |
| All | 6824 | 780 | 878.50 | 1.13 |
* Segregation distortion was higher in LGs 1, 2, 5, 7 and 8.
The K-W test mapped 11 QTLs for four flowering traits on LGs 2, 4, 6 and 8 (Fig. 3) but none on LG 1, 3, 5, 7 or 9 (Supplemental Fig. 1). It detected four significant QTLs for FT on LGs 2, 4 and 6. Detection of qFT2.1, qFT4.1 and qFT6.2 in both years indicates that they determined FT consistently between years. qFT6.2 was a major QTL with a very high K* value. At all four QTLs, the average FT of F2 plants representative of the SoS4-4 allele was earlier than that of plants carrying the XMRS10 allele (Table 4). qFT6.2 had the largest effect: FT became earlier with the number of SoS4-4 alleles. FT was 3.9 days earlier in 2017 and 3.4 days earlier in 2018 in SoS4-4 qFT6.2 homozygotes than in heterozygotes. On the other hand, there was no difference between SoS4-4 qFT2.1 and qFT4.1 homozygotes and heterozygotes.
QTL K* values for flowering time and flowering-related traits shown along linkage groups 2, 4, 6 and 8 (marked at the top). (A, B) FT in 2017 and 2018. (C–E) RTopt in 2017, 2018 and 2020. (F, G) RTHT in 2017 and 2018. (H, I) RT13h in 2018 and 2020. QTLs (P < 0.00001) were detected by Kruskal–Wallis test at the threshold shown by horizontal grey lines (K* = 23.1).
| Traita | QTL | Linkage group | Year | Interval (cM) |
Mean valueb | ||
|---|---|---|---|---|---|---|---|
| SoS4-4 | Heterozygote | XMRS10 | |||||
| FT | qFT2.1 | 2 | 2017 | 31.2–36.1 | 107.1 | 107.8 | 111.9 |
| 2018 | 32.0–33.6 | 110.4 | 110.3 | 113.6 | |||
| qFT4.1 | 4 | 2017 | 41.6–57.3 | 107.7 | 108.7 | 112.7 | |
| 2018 | 46.5–46.8 | 110.7 | 110.8 | 114.1 | |||
| qFT6.1 | 6 | 2017 | 0.0 | 108.2 | 110.1 | 112.1 | |
| qFT6.2 | 6 | 2017 | 59.6–104.8 | 104.1 | 108.0 | 113.3 | |
| 2018 | 62.7–104.8 | 107.3 | 110.7 | 114.6 | |||
| RTopt | qRT2.1 | 2 | 2017 | 31.2–40.1 | 45.7 | 46.6 | 48.4 |
| 2018 | 31.2–36.1 | 46.1 | 46.3 | 48.6 | |||
| 2020 | 15.6–42.8 | 44.0 | 44.1 | 46.2 | |||
| qRT4.1 | 4 | 2017 | 16.6–65.2 | 46.1 | 46.8 | 49.7 | |
| 2018 | 16.7–65.2 | 46.0 | 46.8 | 49.8 | |||
| 2020 | 16.7–65.2 | 43.7 | 44.6 | 47.2 | |||
| qRT8.1 | 8 | 2017 | 44.9–46.5 | 50.4 | 46.6 | 47.8 | |
| RTHT | qHT4.1 | 4 | 2017 | 0.0–45.4 | 7.6 | 7.9 | 12.4 |
| 2018 | 43.2–43.5 | 22.1 | 20.6 | 29.4 | |||
| qHT4.2 | 4 | 2017 | 45.4–65.2 | 7.5 | 7.9 | 12.3 | |
| 2018 | 49.2–51.6 | 21.8 | 20.7 | 29.6 | |||
| RT13h | q13H6.1 | 6 | 2018 | 0.0 | 9.6 | 10.2 | 12.2 |
| q13H6.2 | 6 | 2018 | 62.7–97.6 | 11.6 | 14.5 | 19.5 | |
| 2020 | 42.0–98.7 | 7.3 | 9.4 | 12.7 | |||
a FT, flowering time; RTopt, days to flower under optimal SD and temperature conditions; RTHT, flowering delay caused by high temperature; RT13h, flowering delay caused by 13-h photoperiod.
b Mean phenotypic value of allele of SNP marker detected as best linked with the QTL.
Two QTLs controlling RTopt—qRT2.1 and qRT4.1—were consistently detected in 3 years; for each one, SoS4-4 contributed the allele for smaller RTopt. qRT8.1 was detected in only 2017, and the XMRS10 allele contributed to smaller RTopt. A major QTL for RTHT was detected on LG 4 in 2017. The significant interval was broad and had two K* peaks, so we divided it into QTLs qHT4.1 and qHT4.2 at 45.4 cM. Two corresponding QTLs were detected in 2018. The allele for shorter RTHT originated from SoS4-4. q13H6.2, a major QTL with a remarkably high K* value, for RT13h was detected on LG 6 in both years. q13H6.1, a minor QTL, was detected at the left end of LG 6 in 2020. The allele for shorter RT13h was contributed by SoS4-4: RT13h became shorter as the number of SoS4-4 alleles of q13H6.2 increased. Three FT QTLs overlapped with QTLs for other flowering traits: qFT2.1 with qRT2.1; qFT4.1 with qRT4.1 and qHT4.1; and qFT6.2 with q13H6.2. At each QTL colocalization, the allele effects on phenotypes were similar: phenotypic values became smaller as the number of SoS4-4 alleles in qFT6.2 and q13H6.2 increased.
Candidate genesqFT2.1, qFT4.1 and qFT6.2 were detected consistently in both years, indicating that each region is likely to contain a gene related to FT. We determined that a ~193-Mbp physical region (46,017,926–239,207,382 bp) on chromosome 2 of Gojo-0_v1, corresponding to the overlapping interval (32.0–33.6 cM) for qFT2.1 in 2017 and 2018, potentially contained the candidate gene for qFT2.1. Of the 4,399 genes in the region, 1,208 harbour non-synonymous base variants between the parental lines (Supplemental Table 3). Among them, CsCYC2a (Hirakawa et al. 2019), CsFTL1 (Oda et al. 2012) and a putative circadian-clock-related gene, CsRVE2a (Hirakawa et al. 2019), were related to flowering in previous reports and in FLOR-ID (Table 5). Similarly, nine flowering-related genes were discovered in the region (273,664,494–312,897,337 bp) of qFT4.1, being homologous to Arabidopsis EMF2, FVE, GAI and HDA5. The region (2,676,270–326,802,638 bp) corresponding to the qFT6.2 genomic interval covered almost all of chromosome 6. Nevertheless, 25 candidate genes, including putative circadian-clock-related genes (CsLNK1, CsPRR7, CsTOC1 and CsELF3) and a photoreceptor gene (CsPHYA), were discovered.
| QTL | Gene on Gojo-0_v1 | Arabidopsis homologue | Previously reported chrysanthemum gene |
|---|---|---|---|
| qFT2.1 | CsG_LG2.g12748.1 | FT | CsFTL1, CmFTL1 (Oda et al. 2012) |
| CsG_LG2.g32477.1 | RVE2 | CsRVE2a (Hirakawa et al. 2019) | |
| CsG_LG2.g38814.1 | CYC2 | CsCYC2a (Hirakawa et al. 2019) CmCYC2a (Huang et al. 2016) ClCYC2a (Zhang et al. 2024) |
|
| qFT4.1 | CsG_LG4.g58216.1 | HDA5 | |
| CsG_LG4.g58311.1 | HDA5 | ||
| CsG_LG4.g58315.1 | HDA5 | ||
| CsG_LG4.g58373.1 | HDA5 | ||
| CsG_LG4.g58376.1 | HDA5 | ||
| CsG_LG4.g58392.1 | HDA5 | ||
| CsG_LG4.g59314.1 | EMF2, CYR1 | ||
| CsG_LG4.g62062.1 | FVE, MSI4 | ||
| CsG_LG4.g62650.1 | GAI, RGA2 | ||
| qFT6.2 | CsG_LG6.g05334.1 | SPA3 | CsSPA3 (Hirakawa et al. 2019) |
| CsG_LG6.g05549.1 | SPL3 | ||
| CsG_LG6.g06567.1 | GA2ox1 | CmGA2ox (Lyu et al. 2022) | |
| CsG_LG6.g11837.1 | LNK1/2 | CsLNK1 (Hirakawa et al. 2019) | |
| CsG_LG6.g12204.1 | MSI1 | ||
| CsG_LG6.g12223.1 | MSI1 | ||
| CsG_LG6.g14873.1 | PHYA | CsPHYA (Hirakawa et al. 2019) | |
| CsG_LG6.g19116.1 | TOE1, RAP2.7 | ||
| CsG_LG6.g19334.1 | PRR7 | CsPRR7 (Hirakawa et al. 2019) | |
| CsG_LG6.g20751.1 | GA20ox1 | CmGA20ox (Suh et al. 2023) CmGA20ox2 (Yu et al. 2022) |
|
| CsG_LG6.g27433.1 | TPS1 | ||
| CsG_LG6.g29932.1 | RGL2 | ||
| CsG_LG6.g29933.1 | RGL2 | ||
| CsG_LG6.g29934.1 | RGL2 | ||
| CsG_LG6.g29938.1 | RGL2 | ||
| CsG_LG6.g30342.1 | FCA | ||
| CsG_LG6.g38092.1 | ELF3 | CsELF3 (Hirakawa et al. 2019) ClELF3 (Fu et al. 2014) |
|
| CsG_LG6.g41100.1 | JMJ32, JMJD5 | ||
| CsG_LG6.g49123.1 | CDF3 | ||
| CsG_LG6.g49179.1 | TOC1/PRR1 | CsTOC1 (Hirakawa et al. 2019) CmTOC1 (Higuchi et al. 2012) |
|
| CsG_LG6.g50194.1 | FUL, AGL8 | CsAFL1 (Oda et al. 2012) CmAFL1 (Li et al. 2009b) |
|
| CsG_LG6.g58587.1 | EMF2, CYR1 | ||
| CsG_LG6.g59991.1 | BAF60, CHC1 | ||
| CsG_LG6.g62380.1 | TPL | ||
| CsG_LG6.g63706.1 | GA3OX1 | CmGA3ox (Yang et al. 2014) CmGA3ox1 (Yu et al. 2022) |
Mutations in the circadian clock alter FT in rice and Arabidopsis (Hori et al. 2016, Johansson and Staiger 2015). We investigated the expression of circadian-clock-related genes CsLHY, CsTOC1 and CsGI, putative orthologues of Arabidopsis core clock component genes LHY/CCA1, TOC1/PRR1 and GI, respectively (Hirakawa et al. 2019, Oda et al. 2017, 2020). CsLHY and CsGI are located on chromosomes 1 and 8, respectively. CsTOC1 is located in the interval of qFT6.2 and has non-synonymous base variants between parents (Table 5). Although the transcript abundance of these genes appeared to show no substantial differences between XMRS10 and SoS4-4 (Supplemental Fig. 2), direct comparisons of gene expression levels were avoided (Fig. 4) in view of the stability of housekeeping gene expression and variations in biological backgrounds. Under continuous dark conditions (DD), CsLHY expression started increasing at subjective dawn (i.e., when the plant subjectively perceives the start of the light period due to being entrained under the 16-h light/8-h dark cycle before transferring to DD) and peaked between 9 and 10 h in XMRS10, but between 11 and 12 h in SoS4-4, indicating a phase shift of the expression pattern (Fig. 4A). CsTOC1 and CsGI in SoS4-4 also showed a slightly later phase shift (Fig. 4B, 4C), but the sampling period was too short. Although CsTOC1 expression was increasing in XMRS10 at 24 h, it might increase in SoS4-4 after 24 h. CsGI expression peaked at 24 h in XMRS10 but was still increasing in SoS4-4.
Expression of circadian-clock-related genes (A) CsLHY, (B) CsTOC1 and (C) CsGI in early-flowering SoS4-4 and late-flowering XMRS10 parental C. seticuspe lines under continuous dark (DD) condition. Plants were grown under 16-h light/8-h dark before the experiment. Leaves were harvested hourly for gene expression analyses. 
Time of subjective dark; 
Time of subjective light. Expression levels were normalized to CsACTIN. Values are means ± SE (n = 3).
The main objective of this study was to elucidate the genetic architecture of FT regulation in C. seticuspe. Three major QTLs for FT were detected consistently in 2 years. Nevertheless, minor QTLs might remain unmapped owing to the small sample size, the use of a relatively conservative threshold, and higher segregation distortion on several chromosomes. The three major QTLs show that mechanisms of FT regulation are genetically dissected into three major flowering components. We also investigated correlation coefficients and QTL colocalization between FT and other flowering traits, and the results explained the physiological role of each FT QTL.
FT QTL on chromosome 6, qFT6.2, colocalized with q13H6.2, the QTL for flowering competence under a 13-h photoperiod. Flowering is sensitive to daylength, and critical daylength varies among C. seticuspe accessions (Oda et al. 2012). The early-flowering SoS4-4 has a longer critical daylength than the late-flowering XMRS10. This trait contributes to the flowering competence under a 13-h photoperiod. Thus, the QTL colocalization on chromosome 6 indicates that the longer critical daylength contributed to early FT. Floral transition and early flower bud development are thought to occur in August, when daylength is shortening but still suboptimal for most C. seticuspe accessions (13:00–13:59 h at NIFTS, Tsukuba). The early-flowering SoS4-4 perceives this daylength as a short day, and the signal triggers the transition from vegetative to reproductive growth. In addition, the smaller RTopt of SoS4-4 indicates a faster flower bud development after floral transition, which contributes to an early FT. qFT2.1 and qFT4.1 are associated with the smaller RTopt of SoS4-4 and colocalize with QTLs for RTopt, qRT2.1 and qRT4.1, respectively. RTopt, the number of days to flowering under optimal temperature and photoperiod, indicates the endogenous responsiveness of flowering to inductive conditions, namely the competence to flower quickly following floral transition. Moreover, the overlap of qFT4.1 with both qHT4.1 and qRT4.1 indicates that qFT4.1 acts as a flowering promoter with HT tolerance in SoS4-4. HT conditions delay flowering in C. seticuspe, mainly via the inhibition of capitulum development from inflorescence meristem formation to floret growth (Nakano et al. 2013). Flowering of late-flowering XMRS10 was delayed under HT and thus was putatively thermosensitive, whereas SoS4-4 was less sensitive. QTL analysis revealed that early FT in SoS4-4 is regulated by three flowering components: the component of flowering competence under suboptimal daylength and two components of the promotion of flower bud development, one of which is tolerant to HT. Results also suggest that C. seticuspe knows when to trigger flowering by monitoring daylength as a flowering cue in summer when it is still suboptimal. Because daylength is an extremely stable signal, chrysanthemums can use it as their main cue for stable reproductive growth. However, the FT phenotypic value varied slightly between 2017 and 2018 (Fig. 2A). HT conditions suppress flowering even after reproductive growth has been triggered, so the average FT was delayed in 2018, when average temperatures were higher (Table 1).
Photoperiodic flowering is regulated by the balance between florigen and anti-florigen in C. seticuspe (Higuchi et al. 2013, Oda et al. 2012). The anti-florigen gene CsAFT is induced under long-day (LD), whereas CsAFT is downregulated and the florigen gene CsFTL3 is upregulated under SD. Genes located within the FT QTL may play a role in regulating the expression of CsAFT and CsFTL3. The overlap of qFT2.1 with qRT2.1 indicates that qFT2.1 is associated with flowering competence under SD conditions. CsFTL1, a candidate gene in qFT2.1, is a paralogue of the main florigenic gene CsFTL3 (Oda et al. 2012). However, CsFTL1 is an unlikely responsible gene for qFT2.1, because it contributes to floral induction under LD (Higuchi et al. 2013). HT suppresses flowering in C. seticuspe through downregulation of CsFTL3 (Nakano et al. 2013). The candidate genes in qFT4.1 may be involved in thermoregulated suppression of CsFTL3, because qFT4.1 overlapped with qHT4.1, which is associated with flowering competence under HT. The eight genes, except for CsG_LG4.g62650.1, are putatively orthologous to genes involved in suppression of FT in Arabidopsis (Li et al. 2008). It remains to be clarified whether these genes are involved in the suppression of CsFTL3 by HT in chrysanthemum.
qFT6.2 was associated with setting critical daylength. CsAFT expression is upregulated when photoreception is synchronized with the endogenous photosensitive phase (Higuchi et al. 2013). The setting of the photosensitive phase is controlled by the circadian clock; e.g., the critical daylength was shortened with the delay of the photosensitive phase in CsGI-overexpressing plants (Oda et al. 2020), and conversely, a longer critical daylength was correlated with an early-generated photosensitive phase and a shortened circadian rhythmicity, as suggested by leaf movements, in chrysanthemum (Takahashi et al. 2025). The longer critical daylength in SoS4-4 may be mediated by an early photosensitive phase. The expression pattern of the CsLHY in XMRS10 is similar to that previously reported for the wildtype ZBL1 (Oda et al. 2017) and can be considered typical. In contrast, the expression pattern in SoS4-4 may represent a variant type, as indicated by the slight delay of the peak phase of circadian-clock-related genes. Circadian-clock-related genes (CsLNK1, CsPRR7, CsTOC1 and CsELF3) and a photoreceptor gene (CsPHYA) in qFT6.2 may be responsible for altering the circadian clock in SoS4-4. These results suggest that variations in the circadian clock are involved in the diversity of critical daylength in C. seticuspe.
Two RTopt QTLs, qRT2.1 and qRT4.1, collocate with previously reported QTLs for the number of days to flowering under SD conditions in cultivated chrysanthemum (van Geest et al. 2017). Wild diploid C. seticuspe and cultivated chrysanthemum may share flowering-related genes on chromosomes 2 and 4. However, QTLs associated with a longer critical daylength were detected on different chromosomes: on chromosome 6 (q13H6.2) in C. seticuspe but on chromosome 2 in cultivated chrysanthemum (Takahashi et al. 2025). This indicates that the genes responsible for critical daylength may not be shared between C. seticuspe and cultivated chrysanthemum, and that breeders can develop novel early-flowering cultivars with longer critical daylength in cultivated chrysanthemum by using q13H6.2 through interspecific hybridization, genetic modification and genome editing.
In conclusion, QTL analysis revealed the genetic architecture that regulates FT of C. seticuspe. FT was predominantly determined by the effect of variations in critical daylength, which was controlled by a single QTL. In addition, endogenous SD responsiveness, controlled by two QTLs, was also involved in determining FT. The possibility that an altered circadian clock results in the longer critical daylength and early FT in C. seticuspe is consistent with a report on cultivated chrysanthemum (Takahashi et al. 2025), but the gene responsible for the longer critical daylength may be different, as the QTLs are located on different chromosomes between C. seticuspe and cultivated chrysanthemum. The use of the diploid C. seticuspe proved effective for studying the mechanism of photoperiodic flowering in the highly heterozygous, outcrossing hexaploid chrysanthemum. This study provides insights into the function of candidate genes and engineering the traits of cultivated chrysanthemums.
KaS, Conceptualization, Methodology, Validation, Investigation, Resources, Data curation, Visualization, Project administration, Funding acquisition, Writing—original draft, review & editing. SI, HH, KeS, Methodology, Funding acquisition, Supervision, Investigation, Data curation, Writing—original draft, review & editing. MK, Methodology, Supervision, Writing—review & editing. MY, TH, YN, FT, KM, YH, Supervision, Writing—review & editing.
We thank Dr. DOUZONO Mitsuru (NARO) for temperature data collection. We also thank KAMEI Setsuko, HIRAKAWA Yukiko and OHMORI Izumi for their technical assistance at NARO and the technical staffs of the Laboratory of Plant Genome Research at the Kazusa DNA Research Institute for their technical assistance. The authors acknowledge support for genetic analyses from the Genome Breeding Support Office of the Institute of Crop Science (NICS), NARO. This work was supported by the Japan Society for the Promotion of Science under KAKENHI grant for Scientific Research (C) (No.20K06024); and the Kazusa DNA Research Institute Foundation.
