Hd3a and RFT1 are essential for flowering in rice

The developmental process leading to flowering comprises a vegetative stage and a reproductive stage. The shoot apical meristem (SAM) gives rise to the vegetative structures, eventually transitioning to the reproductive stage that produces flowers. Floral transition is triggered by both endogenous and environmental signals. Among the various environmental signals, photoperiod provides plants with a signal for the most suitable season for flowering(Yanovsky and Kay, 2003; Baurle and Dean, 2006). Plants generally fall into one of three photoperiod-sensing classes: long-day plants(LDP), which promote flowering by sensing long-day (LD) photoperiods;short-day plants (SDP), which promote flowering by sensing short-days (SD);and day-neutral plants, which are not regulated by photoperiod.

The signaling cascades of photoperiodic flowering have been studied in the LDP Arabidopsis thaliana. CONSTANS (CO) encodes a zinc-finger transcriptional activator and induces expression of the floral integrator FLOWERING LOCUS T (FT) under LD conditions(Kardailsky et al., 1999; Kobayashi et al., 1999; Yanovsky and Kay, 2002). FT expression is regulated by both the circadian clock and light(Yanovsky and Kay, 2003; Imaizumi and Kay, 2006). The CO-FT pathway is conserved in rice, which is a SDP [Heading date 1 (Hd1)→Heading date 3a (Hd3a)](Yano et al., 2000; Hayama et al., 2003). Hd3a, which was identified as a quantitative trait locus (QTL) for flowering time, is a key activator of flowering in rice(Kojima et al., 2002). Recent studies suggest that FT/Hd3a represents a florigen-type mobile flowering signal (Tamaki et al.,2007; Corbesier et al.,2007; Jaeger and Wigge,2007; Mathieu et al.,2007; Lin et al.,2007). Hd3a expression is regulated by Hd1, and by Ehd1, a B-type response regulator that functions independently of Hd1 (Yano et al.,2000; Hayama et al.,2003; Doi et al.,2004). Hd3a is also regulated by light via the phytochrome B sensory system. These two functional pathways merge at Hd3a (Izawa et al.,2002; Ishikawa et al.,2005). Key regulators for photoperiodic flowering in rice and Arabidopsis are conserved, but differences in their regulation result in either SDP or LDP (Hayama et al.,2003).

In addition to the photoperiodic pathway, vernalization, autonomous and gibberellin pathways are integrated into the transcriptional regulation of downstream target genes such as FT, TWIN SISTER OF FT (TSF), SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1 (SOC1) and LEAFY (LFY) in Arabidopsis(Boss et al., 2004; Imaizumi and Kay, 2006). Prolonged exposure to cold, a process known as vernalization, promotes flowering in winter annual Arabidopsis. FLOWERING LOCUS C(FLC), a MADS-box transcription factor, suppresses floral transition by repressing the expression of floral activators(Michaels and Amasino, 1999). FLC expression is repressed by the vernalization pathway through epigenetic mechanisms at the FLC locus(Sung and Amasino, 2004a; He and Amasino, 2005). Vernalization requires VERNALIZATION INSENSITIVE 3 (VIN3), a member of a plant-specific protein family with plant homeodomain and fibronectin domains, VERNALIZATION 2 (VRN2), a homologue of polycomb group protein, and VERNALIZATION 1 (VRN1), a protein containing a DNA-binding domain(Levy et al., 2002; Bastow et al., 2004; Sung and Amasino, 2004b). These genes are involved in H3K9-mediated deacetylation, and H3K9- and H3K27-mediated dimethylation chromatin modifications at the first intron of FLC, and promote flowering by the suppression of FLC. Furthermore, HETEROCHROMATIN PROTEIN 1 (LHP1)/TERMINAL FLOWER II (TFLII) is required to maintain the increased level of H3K9 dimethylation at the FLC locus(Sung et al., 2006). In rapid-cycling accessions of Arabidopsis, FLC expression is also regulated by the autonomous pathway, which constitutively represses flowering. In this pathway, FLOWERING LOCUS D (FLD) and FVE,plant homologues of a protein found in the histone deacetylase (HDAC) complex of mammals, partly regulate flowering by histone deacetylation at the FLC locus (He et al.,2003; Ausin et al.,2004). Chromatin modifications at the SOC1 locus have also been observed (Bouveret et al.,2006). However, chromatin modifications at the FT locus have not been reported (Sung et al.,2006), although FT expression is regulated by FLC. In Arabidopsis, flowering is regulated by many floral activators through multiple pathways, but there is no FLC orthologue in the rice genome (Goff et al.,2002; Doi et al.,2004), and rice does not require vernalization for flowering. Photoperiodic flowering is thus the key pathway in rice, but no report on floral regulation through chromatin modification has as yet been published.

In Arabidopsis, TSF, an FT homologue, acts redundantly with FT to promote floral transition, because ft tsf double mutants flower slightly later than ft single mutants(Michaels et al., 2005; Yamaguchi et al., 2005). The rice genome contains thirteen members of the Hd3a gene family(Chardon and Damerval, 2005). RFT1/FT-L3 is the closest homologue of Hd3a, and FTL/FT-L1 is the second closest homologue. Transgenic rice plants overexpressing RFT1 or FTL flower early, much like Hd3a-overexpressing plants (Izawa et al., 2002; Kojima et al.,2002). However, because there are no mutants of FT-likegenes available, including Hd3a, it is unclear whether FT-like genes other than Hd3a function as floral activators. In this study, we show that double RFT1-Hd3a RNAi plants do not flower for up to 300 days after sowing (DAS), indicating that these two genes are essential for flowering in rice. Moreover, RFT1 functions as a floral activator in Hd3a RNAi plants. OsMADS14 and OsMADS15 were shown to be downstream of Hd3a and RFT1 in rice flowering under SD conditions. On the basis of these results, we propose a model for the regulation of rice flowering under SD conditions.

Japonica rice cultivar Norin 8 was used as wild type in the expression, DNA methylation and ChIP assays. Japonica rice cultivar Nipponbare was used as wild type in the expression analysis of the hd1 mutant. The Tos17-induced mutant of Hd1 was described previously (Ishikawa et al.,2005). Plants were grown in climate chambers at 70% humidity under SD conditions with daily cycles of 10 hours of light at 30°C and 14 hours of dark at 25°C, or LD conditions with 14 hours of light and 10 hours of dark. Light was provided by fluorescent white light tubes (400 to 700 nm, 100μmol m^-2 s^-1).

To generate RNAi transgenic plants, the gene sequences of Hd3a and RFT1, for which inverted repeats were made, were amplified using specific primers (see Table 1)and subcloned into the pENTR/D-TOPO cloning vector (Invitrogen) to yield entry vectors. The final RNA silencing vectors were produced by an LR clonase reaction between each of the entry vectors and pANDA(Miki and Shimamoto, 2004). A 1.7 kb promoter region of the RFT1 gene was used to construct RFT1::GUS, including an intron to enhance GUS expression(Tanaka et al., 1990). Transgenic rice plants were generated by Agrobacterium-mediated transformation of rice calli (cv. Norin 8), performed according to a published protocol (Hiei et al.,1994).

GUS staining was described previously(Moritoh et al., 2005). Samples were embedded in paraffin and sectioned at a thickness of 10 μm using an ULTRACUT UCT ultramicrotome (Leica). Sections were photographed using a BX50 microscope (Olympus).

Leaves were harvested at various times of the day, and total RNA was extracted using an RNeasy plant mini kit (Qiagen) and treated with DNase I(Invitrogen). cDNA was synthesized from 1 μg of total RNA using SuperScriptII Reverse Transcriptase (Invitrogen). One microlitre of cDNA was used for the quantitative analysis of gene expression performed with SYBR Green PCR master mix (Applied Biosystems) with gene-specific primers (see Table 1). Data were collected using the ABI PRISM 7000 sequence detection system in accordance with the instruction manual.

ChIP analysis was performed as described previously(Nagaki et al., 2003; Nagaki et al., 2004; Okano et al., 2008) using whole leaves harvested 35 or 70 DAS (zeitgeber time; ZT 0) under SD conditions from Hd3a RNAi and wild-type plants. Isolated nuclei were digested with micrococcal nuclease (Sigma) instead of sonication and, after the recovery of nucleosomes, we confirmed that monomer nucleosome (∼160 bp)was most abundant by electrophoresis. ChIP-PCR products were quantified by real-time PCR. Quantitative ChIP-PCR was normalized to Actin1 in each experiment. Regions I-VII of the RFT1 locus were amplified by real-time PCR using specific primers (see Table 1). ChIP assays were performed three times with at least two replicates each for each sample.

Genomic DNA from plants grown under SD conditions was extracted at 35 and 70 DAS from leaves of wild-type and Hd3a RNAi plants using the cetyltrimethylammonium bromide (CTAB) method(Murray and Thompson, 1980). Genomic DNA (1 μg) was digested with McrBC as recommended by the supplier(New England BioLabs). McrBC is an endonuclease, which cleaves DNA containing methylcytosine. Digested DNA template (1 μl) was then amplified, using PCR,for 25 cycles of 30 seconds at 94°C, 30 seconds at 58°C and 60 seconds at 72°C with gene-specific primers (see Table 1). M65, which is a transgenic line of cv. Nipponbare that carries non-methylated 35S-GFP(Hashizume et al., 1999), and the 35S-Pi line, which carries a silencer construct and methylated 35S-GFP (Okano et al.,2008), served as controls. Methylation status of 35S-GFPwas determined by bisulfite sequencing(Okano et al., 2008).

In rice, RFT1 is the closest homologue of Hd3a, with 91%identity in the deduced amino acid sequence(Kojima et al., 2002; Chardon and Damerval, 2005; Faure et al., 2007). RFT1 also lies adjacent to Hd3a, separated by only 11.5 kb on chromosome 6 (Fig. 1A). To examine RFT1 expression during different developmental stages under SD and LD conditions, we sampled leaves from wild-type plants every 10 days from 25 DAS until flowering. Levels of Hd3a and RFT1 mRNA were examined by quantitative RT-PCR. Under SD conditions, levels of RFT1 and Hd3a transcripts were highest 30 days before flowering, concurrent with floral transition, although absolute transcript levels of RFT1 were much lower than those of Hd3a(Fig. 1B). Expression of no other FT-like gene peaked 30 days before flowering (data not shown). Hd3a expression is diurnal with a peak before dawn, and a gradual decrease during the day under SD conditions(Izawa et al., 2002). Our analysis showed that RFT1 expression is also diurnal, with a peak before dawn (Fig. 1D). It is known that Hd3a acts downstream of the photoperiod pathway via Hd1 transcriptional regulation(Kojima et al., 2002; Hayama et al., 2003). Expression of RFT1 in hd1 mutants was significantly reduced under SD conditions, as is Hd3a(Fig. 1E,F), indicating that RFT1 and Hd3a are partially regulated by Hd1 under SD conditions. Under LD conditions, expression of Hd3a and RFT1 was barely detectable in any developmental stage in wild-type plants (Fig. 1C).

To study the spatial pattern of RFT1 expression, RFT1 and Hd3a expression was measured in leaf blades, sheaths and roots under SD conditions. Expression of RFT1 and Hd3a was observed in leaf blades, but not in leaf sheaths or roots(Fig. 2A,B). An RFT1::GUS reporter fusion protein was detected in leaf blade vascular tissues 35 DAS under SD conditions (Fig. 2C). This expression pattern is similar to that of Hd3a::GUS (Tamaki et al.,2007). The similarity of Hd3a and RFT1expression patterns under SD and LD conditions, and in vascular tissues,suggests that RFT1 could function redundantly with Hd3a in promoting floral transition under SD conditions.

To test whether RFT1 affects floral induction under SD conditions,we produced transgenic plants that suppress RFT1, Hd3a, or both(Fig. 3D). Because homology between RFT1 and Hd3a is low in the 5′ and 3′non-coding regions, the 5′UTR of RFT1 was used for the RNAi construct to specifically suppress RFT1 expression and the 3′UTR of Hd3a was used to specifically suppress Hd3aexpression (Fig. 3A,B). The flowering time of RFT1 RNAi plants (T₁) was essentially the same as in wild type (59±3.5 DAS, n=9 for wild type versus 62±8.3 DAS, n=18 for RFT1 RNAi plants) under SD conditions (Fig. 3C). This result indicates that Hd3a acts as the primary activator of flowering in RFT1 RNAi plants, and that RFT1 does not contribute significantly to floral transition under SD conditions. By contrast, Hd3a RNAi plants (T₁) flowered 95±6.4 DAS(n=9) under SD conditions, about 30 days later than did wild type(Fig. 3C). New leaves are not normally produced after floral transition in rice. Until 60 DAS, Hd3aRNAi and wild-type plants had the same number of leaves, indicating that the growth rates of Hd3a RNAi and wild-type plants are about the same(data not shown). However, Hd3a RNAi plants produced two or three more leaves than did wild-type plants after 60 DAS, suggesting that floral transition is delayed in Hd3a RNAi plants. It is likely, then, that Hd3a, but not RFT1, promotes floral transition under SD conditions. Double RFT1-Hd3a RNAi plants did not flower up to 300 DAS (n=10) under SD conditions(Fig. 3C). These plants continued to produce leaves for 300 days and reached a height of 110-130 cm,about double the height of wild-type plants(Fig. 3E). The absence or extended delay of flowering in double RFT1-Hd3a RNAi plants is apparently due to a complete defect in floral transition. These results suggest that Hd3a and RFT1 are essential for flowering in rice.

The 30-day delay in flowering of Hd3a RNAi plants suggests that additional activators trigger flowering in Hd3a RNAi plants. The observation that double RFT1-Hd3a RNAi plants did not flower even 300 DAS lends credence to the possibility that RFT1 acts as a floral activator in Hd3a RNAi plants. Transcript levels of RFT1 were very low in wild-type plants throughout development, but,in Hd3a RNAi plants, expression gradually increased from 50 DAS to 70 DAS, which was about 30 days before flowering, and is consistent with the timing of floral transition. RFT1 expression continued to increase until flowering at 90 DAS (Fig. 4). Because RFT1 RNAi had no effect on the timing of floral transition (Fig. 3B,C), RFT1 is not likely to play a role in flower induction under SD conditions in wild-type plants. However, in the absence of Hd3aexpression, as in Hd3a RNAi plants, the marked increase of RFT1 expression at a later stage is highly correlated with the induction of flowering. Expression of RFT1 in Hd3a RNAi plants was more than 50-fold higher than in wild-type plants. However, no other FT-like genes had significantly increased expression. Because increased RFT1 expression was also found in other RNAi plants, in which Hd3a expression was suppressed by constructs containing the coding region of Hd3a, this phenomenon is likely to be caused by the absence of Hd3a expression, not by sequences used for the RNAi construct (data not shown). The increase in RFT1 expression was observed only when plants were grown under inductive SD conditions. Under LD conditions, when Hd3a expression was low, no increase was observed in RFT1 expression. Together, these results suggest that RFT1functions as a floral activator in the absence of Hd3a expression under SD conditions.

In Arabidopsis, APETALA1 (AP1), SEPALLATA3(SEP3), FRUITFULL (FUL) and SOC1 are induced by FT in leaves and/or the SAM(Abe et al., 2005; Michaels et al., 2005; Teper-Bamnolker and Samach,2005). In rice, OsMADS1, OsMADS14 and OsMADS15are upregulated in the floral meristem when it begins to differentiate into primary panicle branch primordia (Furutani et al., 2006). To identify genes acting downstream of Hd3a and RFT1, we examined the expression of OsMADS14,OsMADS15 and OsMADS50 at 35 and 70 DAS in leaves of Hd3a RNAi and double RFT1-Hd3a RNAi plants under SD conditions. OsMADS14 and OsMADS15 are orthologues of AP1, and OsMADS50 is a rice orthologue of SOC1(Jeon et al., 2000; Lee et al., 2004). OsMADS14 and OsMADS15 were suppressed in Hd3a RNAi plants 35 DAS (Fig. 5A,B). OsMADS14 and OsMADS15 thus apparently act downstream of Hd3a. Expression of OsMADS14 and OsMADS15 was higher at 70 DAS, when RFT1 expression was also higher in Hd3a RNAi plants (Fig. 5A,B). Furthermore, OsMADS14 and OsMADS15 were suppressed at all stages (35, 70 and 120 DAS) in double RFT1-Hd3aRNAi plants (Fig. 5A,B),suggesting that RFT1 activates the expression of OsMADS14and OsMADS15. The expression of OsMADS50 in Hd3aRNAi and double RFT1-Hd3a RNAi plants was similar to that in wild-type plants (Fig. 5C). These results show that, under SD conditions, Hd3a and RFT1promote floral transition and induce the expression of OsMADS14 and OsMADS15, but not of OsMADS50.

The increase in RFT1 expression in Hd3a RNAi plants had occurred by 70 DAS, which is 35 days later than the normal peak observed in wild-type plants, indicating that activation of flowering by RFT1 is slower than by Hd3a (Fig. 4). Winter-annual accessions of Arabidopsis promote flowering by vernalization. The vernalization process has evolved to distinguish long exposures to cold from shorter exposures, a safeguard that prevents flowering during short-term autumnal temperature fluctuations. FLC expression is epigenetically suppressed by a number of chromatin-remodeling factors. The gradual activation of RFT1expression in Hd3a RNAi plants might thus be the result of epigenetic regulation.

We first assessed DNA methylation at the RFT1 locus at 35 and 70 DAS, because RFT1 expression was very low at 35 DAS and very high at 70 DAS in Hd3a RNAi plants. After digestion of genomic DNA with McrBC, an endonuclease that cleaves DNA containing methylcytosine, specific primers were used to amplify the MI, MII and MIII regions of the RFT1locus (see Table 1). DNA methylation was not detected in any region of the RFT1 locus at any stage in either wild-type or Hd3a RNAi plants (see Fig. S1 in the supplementary material). We obtained similar results when DNA methylation at the RFT1 locus was analyzed by PCR after treatment with methylation-sensitive restriction enzymes (data not shown). DNA methylation was not altered 35 or 70 DAS in wild-type plants or Hd3a RNAi plants,suggesting that the increased expression of RFT1 in Hd3aRNAi plants was not associated with DNA methylation.

Chromatin immunoprecipitation (ChIP) assays were used to examine histone modifications at the RFT1 locus. H3K9 or H4 acetylation and H3K4 dimethylation cause major modifications of active chromatin, whereas H3K9 dimethylation and H3K27 dimethylation are heterochromatic markers(Fuchs et al., 2006). Chromatin modifications in regions I through VII of the RFT1 locus in Hd3a RNAi plants were compared with those in wild-type plants at 35 and 70 DAS by ChIP, using antibodies against H3K9 acetylation(Fig. 6A). At 70 DAS, when RFT1 expression is highly activated in Hd3a RNAi plants,levels of H3K9 acetylation were higher than wild-type plants in region III,the region around the start site of transcription(Fig. 6C). By contrast, there was no increase in H3K9 acetylation 35 DAS in Hd3a RNAi plants, a stage at which RFT1 expression is low(Fig. 6B). Increased H3K9 acetylation at the RFT1 locus may thus be correlated with the activation of RFT1 transcription.

FT-like genes are present in Arabidopsis, Populus, Picea,tomato and barley (Bohlenius et al.,2006; Lifschitz et al.,2006; Faure et al.,2007; Gyllenstrand et al.,2007). Double mutants of TSF, a close Arabidopsis homologue of FT, and FT flower late,but they do eventually flower (Michaels et al., 2005; Yamaguchi et al.,2005).

There are 13 rice genes in the FT-like gene family(Chardon and Damerval, 2005; Faure et al., 2007). In double RFT1-Hd3a RNAi plants, expression of FT-L4, FT-L5, FT-L6 and FT-L12 was similar to that of wild-type plants at 35 DAS and later stages. Expression of FT-L7, FT-L8, FT-L9, FT-L10, FT-L11 and FT-L13 was barely detectable in wild-type plants. Interestingly,expression of FT-L1/FTL, which was the second closest homologue of Hd3a in rice, was suppressed in the leaves of double RFT1-Hd3a RNAi plants (data not shown). In the shoot apex of wild-type plants, expression of FT-L1/FTL was not increased during the transition to the reproductive stage, but was later increased during spikelet and floral organ initiation in the inflorescence meristem (R.K. and K.S., unpublished). Expression of RFT1 and Hd3a was not detected at any stage in the shoot apex (R.K. and K.S., unpublished). Furthermore, in Hd3a RNAi plants, expression of FT-L1/FTLwas suppressed at 35 DAS, and not increased at 70 DAS. These results indicate that RFT1 is the only member of rice FT-like gene family that was upregulated in Hd3a RNAi plants(Fig. 4). However, the possibility that FT-L1/FTL is involved in the extremely late flowering of the double RFT1-Hd3a RNAi plants cannot be completely excluded.

Phylogenetic analysis of cereal FT-like genes indicates that RFT1 is unique to the rice genome, although other FT-likegenes are found in other cereals (Chardon and Damerval, 2005). RFT1 and Hd3a are physically very close on chromosome 6, separated by only 11.5 kb(Fig. 1A), suggesting that RFT1 may have arisen by tandem duplication of Hd3a after the divergence of rice from some progenitor cereal. Therefore, RFT1 may function as an auxiliary to Hd3a in the flowering developmental pathway when Hd3a is suppressed. Regulation by two members of the FT/Hd3a gene family involved in flowering may be a rice-specific mechanism, or an as yet undiscovered auxiliary mechanism in other plants.

Two mechanisms were initially considered to explain the activation of RFT1 in Hd3a RNAi plants: one was the direct interaction of Hd3a mRNA or protein with RFT1 mRNA; and the other was a de novo adaptive pathway, which arises due to the extended vegetative stage caused by a lack of Hd3a. Direct suppression by Hd3a mRNA or protein was ruled out by co-transfection assays. RFT1::GUS and Ubq::bar:GFP (control) or Ubq::Hd3a:GFP were co-transfected into rice protoplasts and the effect of RFT1::GUS fusion protein activity on Ubq::Hd3a:GFP was measured. The absence of RFT1::GUS suppression suggested that neither Hd3amRNA nor Hd3a protein acts directly on the RFT1 promoter (data not shown).

The possibility that a short phase of vegetative growth and transition to reproductive stage prevent RFT1 expression was also considered. Under LD conditions, wild-type expression of RFT1 does not increase during late stages, when Hd3a expression is very low and the vegetative stage is longer than under SD conditions(Fig. 1C). This suggests that delay of the phase change is not the sole cause for the increase of RFT1 expression. In variety Taichung 65, which carries loss-of-function Ehd1 and Hd1 alleles(Doi et al., 2004), expression of RFT1 is not increased at 70 DAS under SD conditions (see Fig. S2 in the supplementary material). These results suggest that activation of RFT1 requires Hd1 and Ehd1, and a novel adaptive pathway induced by the loss of Hd3a expression(Fig. 7).

To test the possible involvement of epigenetic phenomena in this pathway,we examined histone modifications at the RFT1 locus by ChIP assays and found that H3K9 acetylation increased in the region around the transcription start site of RFT1 when RFT1 was highly expressed in Hd3a RNAi plants(Fig. 6C). This suggests that some chromatin-associated factor(s) regulates RFT1 transcription.

Ehd1 has a GARP [Golden2, Arabidopsis RESPONSE REGULATOR(ARR), and Chlamydomonas regulatory protein of P-starvation acclimatization response (Psr1)] DNA-binding motif, which specifically recognizes 5-bp oligonucleotides in vitro(Sakai et al., 2000). Binding sites of Ehd1 are present in RFT1 promoter region II, in which H3K9 acetylation was slightly increased. H3K9 acetylation of region III(5′UTR), which is adjacent to region II, was highly increased in Hd3a RNAi plants. Therefore, chromatin modification at region III of the RFT1 locus may allow Ehd1 to bind to the promoter region of RFT1 and thus induce transcription. Because region III of the RFT1 locus has no homology with the Hd3a 5′UTR used for the Hd3a RNAi construct, siRNA derived from the Hd3a RNA constructs was not likely to be involved in chromatin modification of the RFT1 locus in Hd3a RNAi plants. RFT1 expression also increased in plants in which Hd3a expression was decreased by a construct using the coding region of Hd3a (data not shown). These results suggest that the activation of RFT1 expression in Hd3a RNAi plants is not induced by some unknown factor(s) associated with the RNAi method used in our study.

Hd3a is the main promoter of floral transition and flowering at about 60 DAS under SD conditions in wild-type plants. RFT1, like Hd3a, is regulated by Hd1 under SD conditions, and its expression is diurnal with a peak at dawn(Fig. 1). Expression of RFT1, which is similar in time and space to Hd3a expression,is much lower than that of Hd3a (Figs 1, 2). However, when Hd3aexpression is suppressed, as in Hd3a RNAi plants, RFT1expression increases at a later stage, and RFT1 appears to complement or replace Hd3a as a floral activator (Figs 3, 4). The increase of RFT1 expression induces the expression of two rice AP1orthologues, OsMADS14 and OsMADS15, and also promotes flowering 30 days later than in wild-type plants(Fig. 5). Furthermore, when RFT1 expression is activated in Hd3a RNAi plants, the level of H3K9 acetylation was higher than in wild-type plants, but not when RFT1 expression is low (Fig. 6). This chromatin modification at the RFT1 locus may lead to increased RFT1 expression in Hd3a RNAi plants(Fig. 6). Suppression of both Hd3a and RFT1 resulted in no flowering even 300 DAS(Fig. 3). These results indicate that Hd3a and RFT1 are the major floral activators,and one or the other is essential for photoperiodic flowering in rice under SD conditions (Fig. 7). The molecular mechanism for RFT1 expression in Hd3a RNAi plants and the function of RFT1 under other environmental conditions remain to be studied. There may be an adaptive mechanism of plants to adjust to changes in the gene expression of a major regulator of flowering to secure flowering for the production of offspring.

We thank Dr Masahiro Yano (NIAS, Japan) for Hd3a, and Yuko Tamaki for rice transformation. This research was supported by Grants-in-Aid for Scientific Research on Priority Areas (grant 10182102 to K.S.) of the Ministry of Education, Culture, Sports, Science, and Technology of Japan. We thank all members of Shimamoto's lab for helpful discussions.