The gated induction system of a systemic floral inhibitor, antiflorigen, determines obligate short-day flowering in chrysanthemums

Significance

Photoperiodic floral initiation is thought to be regulated by a systemic flowering inducer (florigen) and inhibitor (antiflorigen) produced in the leaves. Here, we show the discovery of an antiflorigen (CsAFT) from chrysanthemum, which is produced in the leaves under a noninductive photoperiod to systemically inhibit flowering. This antiflorigen production system prevents precocious flowering and enables the year-round supply of marketable flowers by manipulation of day length.

Abstract

Photoperiodic floral induction has had a significant impact on the agricultural and horticultural industries. Changes in day length are perceived in leaves, which synthesize systemic flowering inducers (florigens) and inhibitors (antiflorigens) that determine floral initiation at the shoot apex. Recently, FLOWERING LOCUS T (FT) was found to be a florigen; however, the identity of the corresponding antiflorigen remains to be elucidated. Here, we report the identification of an antiflorigen gene, Anti-florigenic FT/TFL1 family protein (AFT), from a wild chrysanthemum (Chrysanthemum seticuspe) whose expression is mainly induced in leaves under noninductive conditions. Gain- and loss-of-function analyses demonstrated that CsAFT acts systemically to inhibit flowering and plays a predominant role in the obligate photoperiodic response. A transient gene expression assay indicated that CsAFT inhibits flowering by directly antagonizing the flower-inductive activity of CsFTL3, a C. seticuspe ortholog of FT, through interaction with CsFDL1, a basic leucine zipper (bZIP) transcription factor FD homolog of Arabidopsis. Induction of CsAFT was triggered by the coincidence of phytochrome signals with the photosensitive phase set by the dusk signal; flowering occurred only when night length exceeded the photosensitive phase for CsAFT induction. Thus, the gated antiflorigen production system, a phytochrome-mediated response to light, determines obligate photoperiodic flowering response in chrysanthemums, which enables their year-round commercial production by artificial lighting.

The transition from the vegetative to the reproductive phase is one of the most important developmental stages in the plant life cycle. The timing of flowering during the year, which is an important adaptive trait that strongly influences reproductive fitness, is affected by both endogenous and environmental factors. Changes in day length (photoperiod) are among the most important and reliable seasonal signals to plants to reproduce at favorable times of the year. In 1920, Garner and Allard (1) demonstrated that several plant species flower in response to changes in day length and described this phenomenon as “photoperiodism.” Plants are classified according to their photoperiodic responses as short-day plants (SDP), in which flowering occurs when the night length is longer than a critical minimum, long-day plants (LDP), in which flowering occurs when the day becomes longer than some crucial length, and day-neutral plants. Within the SDP and LDP, there are obligate (qualitative) and facultative (quantitative) types of photoperiodic responses. Obligate-type plants are those in which a particular photoperiod is an absolute requirement for the occurrence of a response. Chrysanthemum has become one of the most important horticultural crops since the discovery of photoperiodism because the flowering time of this obligate SDP can be strictly controlled by the use of blackouts or artificial lighting, day-length extension, or illumination during the middle of the long night [night break (NB)] to meet the demand for marketable flowers throughout the year.

In 1936, Chailakhyan proposed the concept of the flowering stimulus “florigen” from an experiment using chrysanthemum (2). Recent studies have demonstrated that FLOWERING LOCUS T (FT) and its orthologs, which are synthesized in the leaves of several species, act as florigens (3–6). In Arabidopsis, FT moves into the shoot apical meristem (SAM) via the phloem and forms a transcriptional complex with a basic leucine zipper (bZIP) transcription factor, FD, at the SAM; transcription of floral regulator genes such as FRUITFULL (FUL) and APETALA 1 (AP1) is then activated, which leads to flowering (7, 8). FT encodes a small protein, florigen, with similarity to phosphatidylethanol-amine–binding protein (PEBP). The PEBP gene family has evolved both activators and repressors of flowering. The FT family in Arabidopsis contains five other members: TWIN SISTER OF FT (TSF), TERMINAL FLOWER 1 (TFL1), BROTHER OF FT AND TFL1 (BFT), MOTHER OF FT AND TFL1 (MFT), and Arabidopsis thaliana CENTRORADIALIS homolog (ATC). FT and TSF are floral activators (9–11) whereas TFL1, ATC, and BFT are floral repressors (12–14), and MFT is involved in seed germination (15). Although TFL1 and ATC are known to act non-cell-autonomously, they are expressed in the shoot apex and in vasculature tissue, respectively (16, 17).

The concept of a floral repressor, or antiflorigen, was proposed almost as early as that of a floral stimulus (18). A classical physiological study of many plant species suggested the existence of an antiflorigenic stimulus synthesized in leaves (19). A grafting experiment using tobacco cultivars with different photoperiodic responses clearly indicated that antiflorigenic signals synthesized in leaves under non-floral-inductive day-length conditions systemically inhibited flowering (20). The existence of a floral inhibitor in chrysanthemum leaves under non-floral-inductive day-length conditions was also suggested (21). These reports strongly support the idea that an antiflorigenic signal produced in leaves may contribute to photoperiodic responses in some plant species.

Here, by using a reverse-genetic approach, we identified a systemic floral inhibitor produced in the noninductive leaves of chrysanthemum. Our results demonstrated that the phytochrome-mediated and gated production system of antiflorigen determines obligate flowering response in chrysanthemum, which enables year-round commercial production of flowers by using artificial lighting.

Results

Reverse Genetic Screening of Systemic Floral Inhibitors in Chrysanthemum.

To understand the molecular mechanisms of flowering in chrysanthemum, we applied a diploid wild chrysanthemum, Chrysanthemum seticuspe f. boreale (C. seticuspe), as a model system for molecular-genetic study. C. seticuspe shows an obligate photoperiodic flowering response in which flowering occurs under a >12-h dark period and is inhibited under a <10-h dark period (Fig. 1A). Recently, chrysanthemum orthologs of FT were identified, and it was confirmed that short day (SD)-induced CsFTL3 encodes a systemic floral inducer in C. seticuspe (22). To test whether noninduced leaves produce a floral inhibitory signal, we conducted a localized photoperiodic treatment. Localized NB treatment completely suppressed flowering of C. seticuspe, suggesting the existence of an antiflorigenic stimulus produced in the leaves under non-floral-inductive conditions (Fig. 1B). Based on RNA-seq data, we designed a custom array to screen differentially expressed genes in leaves under SD and NB conditions (Datasets S1 and S2). Among genes with higher expression under NB, we identified one clone (CAC02.24543.C1) that showed high homology with TFL1C of Vitis vinifera, and named it as Anti-florigenic FT/TFL1 family protein (AFT). Phylogenetic analysis revealed that CsAFT belongs to the large TFL1/CEN/BFT clade (Figs. S1 and S2A). We screened >60,000 contig sequences derived from RNA-seq, and found five FT/TFL1-like genes in C. seticuspe: three FT-like genes (CsFTL1, CsFTL2, CsFTL3), one TFL1/CEN-like gene (CsTFL1), and CsAFT (Fig. S1). Among these, only CsFTL3, a florigen gene in C. seticuspe (22), is highly expressed in the leaves under SD conditions (Fig. 1 C and D, Fig. S3A, and Dataset S1). The absolute expression level of CsFTL2 was very low under both SD and long-day (LD) conditions, and there were no remarkable differences in temporal expression of this gene between SD and NB conditions (Fig. 1C and Fig. S3A). CsTFL1, which is closely related to Arabidopsis TFL1, was preferentially expressed in root and shoot tips and was expressed at very low levels in leaves (Fig. S3B); there were no remarkable differences in CsTFL1 expression under SD and LD conditions in leaves and shoot tips (Fig. 1C, Fig. S3B, and SI Text). CsFTL1 and CsAFT were highly expressed in the leaves under flower noninductive NB or LD conditions (Fig. 1 C and D and Fig. S3 A and B); however, constitutive expression of CsFTL1 in Chrysanthemum morifolium revealed that CsFTL1 has weak florigenic activity (Fig. S4). CsFTL1 might function as an LD florigen similar to RICE FLOWERING LOCUS T1 (RFT1) as suggested in rice, a facultative SDP (23).

Fig. 1.

Photoperiod-dependent expression of FT/TFL1-like genes in chrysanthemum. (A) Flowering response of C. seticuspe grown for 7 wk under various day lengths. Data are means ± SEM (n = 10). (B) Schematic representation and flowering response under localized photoperiodic treatment. Black and white ovals indicate leaves exposed to SD (8L/16D) and NB (SD plus 2 h NB with red light), respectively. The figures indicate the number of SD or NB leaves. For NB0/SD10 and NB3/SD10, the lower part of each plant was covered with a polyvinyl chloride pipe and aluminum foil to prevent exposure to red light. Data are means ± SEM (n = 10). (C) Temporal expression patterns of FT/TFL1-like genes in leaves grown under SD and NB conditions for 7 d. Data are means ± SEM of three replicates. (D) Spatial expression of CsFTL3 and CsAFT genes under SD and long-day (LD) photoperiods. Plants were grown under SD or LD conditions for 12 d, and each tissue was harvested at ZT4 (4 h after lights-on). Data are means ± SEM of 3 replicates.

CsAFT Acts Systemically to Inhibit Flowering.

Sequence comparisons between CsAFT and FT/TFL1 family proteins from other plant species showed that the CsAFT protein carries the functionally important residues for TFL1-like activity, His88 (H) (24) and Asp144 (D) (25) (Fig. S2A). Constitutive expression of CsAFT (CsAFT-ox) in C. seticuspe and C. morifolium resulted in extremely late flowering under SD conditions, indicating that CsAFT has strong antiflorigenic activity (Fig. 2 A–C and Fig. S5A). Overexpression of CsAFT in Arabidopsis also resulted in late flowering and morphological changes similar to TFL1-ox (14), suggesting that CsAFT acts as a floral inhibitor in Arabidopsis (Fig. S5G). To test the long-distance transmission of CsAFT gene products, we conducted a grafting experiment using CsAFT-ox plants. Wild-type plants grafted onto CsAFT-ox rootstock delayed flowering under SD conditions (Fig. 2E). Moreover, CsAFT protein was detected in shoot tips of WT scions grafted onto transgenic rootstock that overexpress HA-tagged CsAFT (CsAFT-HA), clearly demonstrating that CsAFT protein acts as a systemic floral inhibitor (Fig. 2F). Further, loss of function of CsAFT by RNAi resulted in reduced sensitivity to NB and promoted flowering (Fig. 2 G and H) but did not have any effect under strongly inductive SD conditions (Fig. S5B), indicating that CsAFT is essential for floral inhibition under noninductive NB conditions. Recently, antagonistic function for two FT paralogs has been shown in sugar beet (Beta vulgaris) (26). BvFT1 prevents flowering during SDs and before vernalization by repressing the expression of BvFT2, an activator of flowering. In CsAFT-ox plants grown under SD conditions, expression levels of CsFTL3 in leaves were similar to those of the wild type (Fig. S5C) whereas expression of CsAFL1, an AP1/FUL-like gene, and CsM111, the homolog of AP1, was suppressed in shoot tips (Fig. 2D). This result suggests that late flowering of CsAFT-ox plants is not caused by suppression of CsFTL3 induction in leaves and that the ratio of the flowering inducer (florigen) and inhibitor (antiflorigen) at the SAM may determine floral initiation.

Fig. 2.

CsAFT acts as a systemic floral inhibitor. (A) Expression of CsAFT in WT and CsAFT-ox plants under LD (16L/8D) conditions. Leaves were harvested at ZT0. (B and C) Overexpression of CsAFT dramatically delayed flowering under SD conditions. (D) Expression of CsAFL1 and CsM111 in shoot tips of WT and CsAFT-ox plants grown under SD conditions for 2 wk. (E) Flowering response of WT scions grafted onto WT, CsAFT-ox #2, and CsAFT-ox #16. Plants were grown under SD (8L/16D) conditions for 6 wk. Data are means ± SEM (n = 9 or 10). (F) Detection of CsAFT-3×HA protein at the shoot tips of WT scions grafted onto CsAFT-HA plants, revealed by immunoblot analysis. M, molecular marker. (G) Expression of CsAFT in WT and CsAFT-RNAi lines. Plants were grown under NB conditions for 2 wk. Leaves were harvested at ZT4. (H) Flowering response of WT and CsAFT-RNAi lines under NB conditions. Plants were grown for 5 wk under NB (8L/16D plus 15 min of red light at middle of dark period) conditions. Red circles indicate the positions of flower buds. Data are means ± SEM (n = 6).

CsAFT Antagonizes CsFTL3 Through Interaction with CsFDL1.

The FT–FD protein complex triggers a cascade of positive transcriptional signals in floral induction (7, 8, 27) whereas the TFL1–FD protein complex negatively regulates flowering (28). Two FD-like genes (CsFDL1 and CsFDL2) (Fig. S2B and SI Text) were found in C. seticuspe expressed sequence tags (ESTs). To examine the ways in which interactions among CsAFT, CsFTL3, CsFDL1, and CsFDL2 affect several downstream genes, we developed a transient gene expression system in protoplasts derived from mesophyll cells of C. seticuspe leaves (SI Text). We then examined the subcellular localization and interaction of these proteins in C. seticuspe protoplasts. Subcellular localization of CsFTL3 and CsAFT was observed in both the cytoplasm and nucleus whereas CsFDL1 and CsFDL2 clearly localized to the nucleus (Fig. S6A). Protein–protein interactions among CsFTL3, CsAFT, CsFDL1, and CsFDL2 revealed by bimolecular fluorescence complementation (BiFC) showed that CsFTL3 and CsAFT clearly interacted with both CsFDL1 and CsFDL2 in the nucleus (Fig. 3A). This result indicates that CsAFT and CsFTL3 may function in the same transcriptional complex.

Fig. 3.

Antagonistic action of CsAFT on flower induction by CsFTL3-CsFDL1 complex. (A) BiFC assays showing interactions of CsFTL3-CsFDL1, CsFTL3-CsFDL2, CsAFT-CsFDL1, and CsAFT-CsFDL2. (Scale bar: 10 μm.) (B) Transient expression of CsFTL3 and CsFDL1 induced expression of CsAFL1 and CsAFL2 but was partially antagonized by CsAFT. Data are means ± SD (n = 7).

To evaluate the effect of transcriptional complexes composed of CsFTL3, CsAFT, CsFDL1, and CsFDL2 on the regulation of floral-meristem identity genes, AP1/FUL-like genes (CsAFL1 and CsAFL2) were selected for gene-expression analysis. Up-regulation of CsAFL1 and CsAFL2 in the SAM is one of the earliest events during the transition to the reproductive phase (22, 29) (Fig. S3C). Expression of CsAFL1 was up-regulated severalfold when CsFTL3 and CsFDL1 were coexpressed in protoplasts, but none of the other combinations induced CsAFL1 expression (Fig. 3B). Moreover, up-regulation of CsAFL1 by CsFTL3/CsFDL1 was partially suppressed by the addition of CsAFT expression vectors (Fig. 3B). Expression of CsAFL2, another AP1/FUL-like gene, was strongly up-regulated by CsFTL3/CsFDL1 coexpression but was partially suppressed by the addition of CsAFT expression vectors. Coexpression of CsFTL3/CsFDL2 also induced CsAFL2 expression, but the effect was weaker than that observed in the CsFTL3/CsFDL1 combination. In transgenic plants overexpressing CsFDL1 fused to a transcriptional repressor domain (CsFDL1-SRDX), flowering was severely suppressed under SD conditions (Fig. S5E). Moreover, BiFC competition assay revealed that CsFTL3–CsFDL1 complex formation was repressed by CsAFT coexpression (Fig. S6B). These results suggest that the CsFTL3/CsFDL1 complex activates expression of floral-meristem identity genes and that CsAFT antagonizes CsFTL3 function via interaction with the same interacting partner, CsFDL1, and thus inhibits subsequent flower initiation (SI Text).

Interestingly, endogenous CsFTL3 expression was also induced in our transient gene expression system when CsFTL3 and CsFDL1 were coexpressed (Fig. S7B). A recent study in potato suggested an autoregulatory mechanism of StSP6A, an FT-like gene, in the photoperiodic regulation of tuberization (30). These findings suggest that a positive feedback loop for CsFTL3 expression by the CsFTL3/CsFDL1 complex, or by downstream target genes, may exist in chrysanthemum (Fig. S7C and SI Text).

Phytochrome B Mediates NB-Induced Inhibition of Flowering.

In chrysanthemum, NB with red light effectively inhibits flowering, which is repromoted by subsequent exposure to far-red light, suggesting the involvement of phytochromes in this response (31, 32). NB response in C. seticuspe to differing light quality was also tested with four light-emitting diode (LED) panels. NB with peak irradiance at 530 nm (green) or 660 nm (red) light effectively inhibited flowering and expression of CsFTL3, and induced expression of CsAFT (Fig. S8). Moreover, the effects of red light on flowering and expression of CsFTL3 and CsAFT were partially reversed by subsequent exposure to far-red light (Fig. S8), suggesting that red/far-red reversible types of phytochromes, such as phyB, are involved in this reaction. As expected, knock-down of chrysanthemum PHYB (CsPHYB) (33) resulted in reduced sensitivity to NB by red light and in extremely early flowering (Fig. 4 A and B). CsFTL3 was up-regulated in CsPHYB-RNAi plants whereas CsAFT was down-regulated under NB conditions (Fig. 4C). Therefore, expression of both CsFTL3 and CsAFT is regulated by light-signaling mediated by CsPHYB, and up-regulation of CsFTL3 and down-regulation of CsAFT may have resulted in the synergistic effect on dramatic flowering phenotype observed in CsPHYB-RNAi lines.

Fig. 4.

Phytochrome-dependent regulation of NB response and CsAFT expression. (A and B) Flowering response of WT and CsPHYB-RNAi plants grown under SD or NB conditions. Plants were pinched when transferred into the growth chamber and were grown under SD (8L/16D) or NB (SD plus 10 min of red light at the middle of each dark period) for 54 d. Data are means ± SEM (n = 14 or 16). (C) Expression of CsPHYB, CsFTL3, and CsAFT in leaves of WT and CsPHYB-RNAi lines. Plants were grown under NB for 7 d. Average values and SD from three RT-PCR datasets are shown. Data are representative of two independent experiments.

Gated Induction of CsAFT and Dark-Dominant Flowering of Chrysanthemum.

Because expression of CsAFT was up-regulated under NB or LD conditions, we expected that CsAFT would respond acutely to light exposure under noninductive conditions. In rice, phytochrome-mediated induction of the floral repressor Ghd7 (Grain number, plant height, and heading date 7) is gated by circadian clock action (34). Ghd7 is acutely induced when phytochrome signals coincide with a photosensitive phase set differently by distinct photoperiods. We examined the effect of a short light pulse on the induction of CsAFT mRNA. Plants were grown under SD or LD conditions and shifted to continuous darkness (DD), and the effects of a short red-light pulse on acute induction of CsAFT were tested at different time points. Under SD conditions, CsAFT expression showed clear gating responses to red-light pulses, with peak expression between 8 and 10 h after dusk (Fig. 5A and Fig. S9). Under LD conditions, although relatively high expression levels were observed even in the early part of the night (due to induced expression of this gene under LD), peak photo-inducibility of CsAFT occurred at 8 or 10 h after dusk (Fig. 5A and Fig. S9). Thus, the gate for maximal induction of CsAFT opened at constant time after dusk regardless of the entrained photoperiod conditions. In the flowering response of C. seticuspe, the most sensitive phase to NB occurs 8–11 h after dusk (Fig. 5B). Because at least 12 h of uninterrupted darkness are needed for minimum flowering, and a dark period of <10 h was fully inhibitory (Fig. 1A), we further examined the photoperiod-dependent induction of CsAFT. Expression levels of CsAFT were highly correlated with the extent of flower inhibition; CsAFT was strongly up-regulated under LD conditions (16L/8D, 14L/10D) and moderately induced under 12L/12D, but was suppressed under SD (10L/14D, 8L/16D) conditions (Fig. 5C). These results suggest that gated expression of CsAFT is responsible for NB response under SD conditions and for critical day-length response. We also found that the flowering response of C. seticuspe under non-24-h light/dark cycles consisting of constant dark periods (16L/14D or 22L/14D) was not dramatically attenuated compared with that under a 24-h SD (10L/14D) photoperiod (Fig. 5D). Moreover, expression of CsAFT and CsFTL3 under the non-24-h light/dark cycles was similar to that under the 24-h SD photoperiod (Fig. 5E), indicating that perception of day length in chrysanthemum relies on the absolute duration of darkness rather than on the appropriate timing of a photoperiodic response rhythm set by the dawn signal (Fig. 6A).

Fig. 5.

Gated induction of CsAFT and dark-dominant flowering of chrysanthemum. (A) Analysis of gated expression of CsAFT. Plants were entrained under SD or LD conditions for 7 d. Plants were then transferred to continuous dark (DD) at dusk and exposed once to 10 min of red light at times differing by 2 h. Induction of CsAFT expression in leaves was analyzed 6 h after each exposure. Black and gray bars represent subjective night and day. Average values and SD from three RT-PCR datasets are shown. Data are representative of two independent experiments. (B) Effect of daily 10-min NB given at different times of the night. Numbers on the horizontal axis indicate the length of dark period before NB was given. Data were collected 35 d after treatments started. Data are means ± SD (n = 10–12). (C) Expression of CsAFT and CsFTL3 in leaves grown for 7 d under various photoperiods. Data are means ± SEM of three replicates. Black and white bars represent dark and light periods, respectively. (D) Flowering response under non-24-h photoperiod. Plants were grown under 10L/14D (24-h cycle), 16L/14D (30-h cycle), and 22L/14D (36-h cycle) for 30 d. Data are means ± SEM (n = 10). Data are representative of three independent experiments. (E) Expression of CsAFT and CsFTL3 in leaves grown under normal SD (10L/14D), 16L/14D (30-h cycle), and LD (14L/10D) conditions for 7 cycles. Data are means ± SEM of three replicates.

Fig. 6.

Anti-florigenic regulation of obligate photoperiodic flowering response in chrysanthemum. (A) Model for induction of CsAFT in response to natural day length extension and artificial lighting. The gate for maximal induction of CsAFT mRNA opens at a constant time after dusk regardless of the entrained photoperiod. As the night becomes shorter, red-light signals in the morning coincide with the photo-inducible phase of CsAFT and then inhibit flowering. Under NB conditions, midnight illumination coincides with the photo-inducible phase of CsAFT. (B) Transcriptional regulation of CsAFT and CsFTL3 by photoperiod. High accumulation of CsAFT under noninductive (LD/NB) photoperiod overcomes residual floral-inductive activity of CsFTL3, which enables maintenance of the vegetative state. Upon a shift from LD to SD photoperiod, the CsAFT level rapidly decreases, whereas CsFTL3 is gradually induced by repeating SD cycles. (C) Systemic regulation of flowering by anti-florigen (AFT) and florigen (FTL3). AFT is synthesized in leaves under noninductive photoperiods (LD/NB) and is translocated to the shoot apex where it then inhibits flowering. FTL3 is produced in leaves under inductive (SD) photoperiod independently of the production of AFT and then moves to the shoot apex where it induces flowering.

Discussion

Antiflorigenic Signaling Determines Obligate Short-Day Flowering in Chrysanthemums.

The physiological concept that inductive photoperiods cause leaves to synthesize a floral stimulus (“florigen”) is widely accepted. However, it has also been proposed that an antiflorigenic signal produced in leaves may regulate photoperiodic floral induction; the appropriate day-length would then lead to removal of an antiflorigen (18, 19). Here, we reveal the existence of an antiflorigenic FT/TFL1 family protein, AFT, in C. seticuspe, and clearly demonstrate that CsAFT protein acts as a systemic floral inhibitor—an antiflorigenic signal produced in leaves under noninductive conditions (Figs. 1B and 2). These findings provide insight into the importance of systemic inhibition of photoperiodic flowering by antiflorigens in many plant species. Chrysanthemum is an obligate SDP that maintains a vegetative state under noninductive LD photoperiod conditions (Fig. 1A). Unlike chrysanthemum, rice (Oryza sativa), a facultative SDP, can ultimately flower even under noninductive LD conditions. Rice uses two florigen genes (Hd3a and RFT1) depending on day length, and RFT1 is suggested as an LD florigen (23). In chrysanthemum, CsFTL1, which has weak florigenic activity (Fig. S4), could function as an LD florigen similar to RFT1 in rice. It can be assumed that residual CsFTL3 and increased CsFTL1 activity under noninductive photoperiod conditions (Fig. 1C and Fig. S3A) eventually make the plant flower, but the transition from vegetative to reproductive development is strictly suppressed under those conditions. Therefore, the photoperiodic transcriptional regulation of florigens alone is not sufficient to explain the obligate flowering response of chrysanthemum; it is reasonable to assume that a systemic antiflorigen, which maintains the vegetative state at the shoot apex during noninductive photoperiods, may be needed. The necessity of the antiflorigenic signal (CsAFT) to maintaining the vegetative state is supported by analysis of photoperiodic reactions of CsAFT-RNAi plants (Fig. 2 G and H). Thus, although photoperiodic regulation of florigenic stimuli must be important to achieve flowering (SI Text), the phyB-mediated antiflorigen production system (Fig. 6A) plays a predominant role in the obligate photoperiodic flowering response in chrysanthemum, allowing strict maintenance of the vegetative state under noninductive photoperiod conditions (Fig. 6 B and C).

Possible Model of Day-Length Measurement in Photoperiodic Flowering of Chrysanthemum.

Day length in many plant species is measured through a circadian clock, which is an endogenous time-keeping component that is reset by dawn and dusk signals. In Arabidopsis, expression of CONSTANS (CO), a critical activator of FT, is regulated by the circadian rhythm set by the dawn signal, and increases toward evening (35, 36). Under LD conditions, when high CO expression coincides with external light signals in the evening, the CO protein is stabilized and activates the transcription of FT (36, 37). In rice, flowering of wild-type plants was extremely delayed under non-24-h light/dark cycles (24L/12D or 36L/12D) compared with that under SD and LD photoperiods (38). Moreover, this complete inhibition of flowering in wild-type plants under atypical entrainment conditions was lost in the se1 (hd1) mutant, a rice counterpart of Arabidopsis co. This result indicates that SE1 (Hd1) function is evening phase-specific and that a circadian rhythm set by the dawn signal is critical for day-length recognition. These observations suggest that the dawn rather than the dusk signal plays a critical role for phase setting of the photoperiodic response rhythm in these two species. We demonstrated that, in C. seticuspe, if long night conditions (14 h) were given, flowering was successfully induced regardless of the day length (10, 16, or 22 h) (Fig. 5D). Moreover, expression of CsAFT and CsFTL3 under atypical entrainment conditions (16L/14D) was similar to that under a 24-h SD (10L/14D) photoperiod (Fig. 5E), indicating that perception of day length in chrysanthemum relies on the absolute duration of darkness rather than on the appropriate timing of a photoperiodic response rhythm set by the dawn signal. Our finding that the gate for maximal induction of CsAFT opens at a constant time after dusk, regardless of the entrained photoperiod (Fig. 5A and Fig. S9), also supports this notion. The mechanism for gated induction of CsAFT mRNA appears similar to, yet differs from, that of Ghd7 in rice. The gate for Ghd7 under SD photoperiods opens maximally at 4 h after dusk whereas it opens around dawn (10 h after dusk) under LD photoperiods (34). Therefore, timing for Ghd7 gate opening is not determined solely by time elapsed from the beginning of the dark period but is also affected by day length. In Pharbitis, flowering of which is induced by a single exposure to darkness, induction of PnFT occurs at a constant time after the light-to-dark transition, regardless of the day length preceding the inductive dark period (39). Therefore, as in the case of Pharbitis, some time-keeping component set by the dusk signal may be involved in dark-time measurement in chrysanthemum. The gate for CsAFT induction, which is initiated by light-to-dark transition at dusk, is crucial for determining critical day length in chrysanthemum because CsAFT can be triggered by the coincidence of morning light with the photosensitive phase only when the night length becomes shorter than a critical minimum (Fig. 6A).

The work reported here provides the molecular basis for a long-standing physiological observation: that flowering is inhibited by an antiflorigenic stimulus produced in noninductive leaves (Fig. 6 B and C). The gated induction of a floral repressor, CsAFT, a phyB-mediated response to light, determines the obligate photoperiodic flowering response in chrysanthemums. The mechanism for gated induction of CsAFT has a strong link between flowering and critical day length in chrysanthemums; flowering occurs only when night length exceeds the duration of the photosensitive phase for CsAFT induction. Elucidation of the mechanisms underlying photoperiodic flowering in chrysanthemums makes an important contribution to understanding of this plant's reproductive success in its native environments. In addition, these findings will aid in achieving economic benefits from a stable year-round supply of marketable flowers that can be produced by manipulation of photoperiod using artificial lighting or blackouts. Further, identification of the AFT gene will lead to improved understanding of the exquisite coordination that exists within the photoperiodic-flowering gene network for various plant species.

Materials and Methods

Details are described in SI Materials and Methods. These details include information on plant materials and growth conditions, localized photoperiodic treatments, grafting experiments, RNA-seq and custom array analyses, the gene-expression study, plasmid construction, protoplast preparation, subcellular localization and BiFC assays, and immunoblot analyses. Primers used for expression analyses are provided in Dataset S3.