The Rice CHD3/Mi-2 Chromatin Remodeling Factor Rolled Fine Striped Promotes Flowering Independent of Photoperiod

Genetic studies have revealed that chromatin modifications affect flowering time, but the underlying mechanisms by which chromatin remodeling factors alter flowering remain largely unknown in rice (Oryza sativa). Here, we show that Rolled Fine Striped (RFS), a chromodomain helicase DNA-binding 3 (CHD3)/Mi-2 subfamily ATP-dependent chromatin remodeling factor, promotes flowering in rice. Diurnal expression of RFS peaked at night under short-day (SD) conditions and at dawn under long-day (LD) conditions. The rfs-1 and rfs-2 mutants (derived from different genetic backgrounds) displayed a late-flowering phenotype under SD and LD conditions. Reverse transcription-quantitative PCR analysis revealed that among the flowering time-related genes, the expression of the major floral repressor Grain number and heading date 7 (Ghd7) was mainly upregulated in rfs mutants, resulting in downregulation of its downstream floral inducers, including Early heading date 1 (Ehd1), Heading date 3a (Hd3a), and Rice FLOWERING LOCUS T 1 (RFT1). The rfs mutation had pleiotropic negative effects on rice grain yield and yield components, such as plant height and fertility. Taking these observations together, we propose that RFS participates in multiple aspects of rice development, including the promotion of flowering independent of photoperiod.


Introduction
In rice (Oryza sativa), flowering time is a critical yield-related trait that is important for seasonal and regional adaptation worldwide. The transition from the vegetative to the reproductive phase is promoted by two florigen genes in rice, Heading date 3a (Hd3a) and Rice FLOWERING LOCUS T 1 (RFT1), which are first expressed in the leaves [1,2]. Hd3a and RFT1 act as the major mobile signals, moving to the shoot apical meristem and triggering the transition from vegetative to reproductive growth. Heading date 1 (Hd1) acts as a floral inducer or repressor depending on the length of the photoperiod [3]. Hd1 upregulates the expression of Hd3a and RFT1 under short-day (SD) conditions and downregulates them under long-day (LD) conditions. Early heading date 1 (Ehd1) encodes a B-type response regulator that upregulates the expression of Hd3a and RFT1 [4].
In addition to epigenetic regulation of SDG proteins, ATP-dependent chromatin remodeling factors are also involved in altering chromatin structure to modulate flowering time. In Arabidopsis thaliana, the putative sucrose non-fermenting (SNF2)-like ATPase subunit PICKLE (PKL) plays roles in the floral transition by regulating the expression of the floral meristem identity gene LEAFY (LFY) and gibberellic acid (GA)-regulated genes [23,24]. PKL physically interacts with CONSTANS (CO) to facilitate binding to FLOWERING LOCUS T (FT) chromatin. PKL mediates deposition of H3K4me3 on the FT locus by associating with ARABIDOPSIS HOMOLOG OF TRITHORAX 1 (ATX1) to antagonize Polycomb group (PcG)-mediated repression of FT [25,26]. CHROMATIN REMODELING 4 (CHR4) interacts with transcription factors involved in floral meristem identity to regulate the expression of key floral regulators [27].
In rice, Rolled Fine Striped (RFS), which encodes a chromodomain helicase DNAbinding 3 (CHD3)/Mi-2 chromatin-remodeling factor, is involved in various aspects of rice development such as crown root development, seedling development, and leaf morphogenesis [28][29][30]. Genome-wide analysis revealed that the rfs mutant shows a reduction of H3K4me3 and H3K27me3 [31]. These results demonstrate that RFS can modulate both active and repressive epigenetic marks on histones. Our previous report showed that RFS controls reactive oxygen species (ROS) homeostasis by modulating H3K4me3 levels on ROS-related genes [32].
Here, we found that two rfs mutants (rfs-1 and rfs-2) derived from different genetic backgrounds exhibited late flowering independent of the photoperiod, implying that RFS is involved in flowering-time regulation in rice. Although numerous genes that control rice flowering have been reported, few epigenetic regulator that modulate flowering has not been identified in rice. To reveal the regulatory function of RFS in rice flowering, we analyzed the expression and histone methylation levels of flowering-time genes in rfs-2 mutant. Reverse transcription-quantitative PCR (RT-qPCR) showed that RFS promotes flowering by downregulating Ghd7 in LDs and upregulating Ehd1 in SD and LD conditions. Our results suggest a new molecular function of RFS as a floral regulator in rice.

The rfs Mutants Exhibit a Late Flowering Phenotype Independent of the Photoperiod
To elucidate the involvement of RFS in flowering-time regulation in rice, we investigated the flowering time of two rfs mutants [32] under different photoperiod conditions. The rfs-1 mutant (which was isolated from a gamma-ray mutagenesis) flowered around 8 d later than its parental japonica cultivar 'Seolak' (SL), which flowered at 92 days after sowing (DAS) under natural long-day (NLD) conditions in the field (37 • N latitude, Suwon, Korea). In the growth chambers, the rfs-1 mutant flowered around 8 d and 11 d later than SL under SD (10 h light/14 h dark), and LD (14.5-h light/9.5-h dark) conditions, respectively (Figure 1a,c).

The rfs Mutants Exhibit a Late Flowering Phenotype Independent of the Photoperiod
To elucidate the involvement of RFS in flowering-time regulation in rice, we investigated the flowering time of two rfs mutants [32] under different photoperiod conditions. The rfs-1 mutant (which was isolated from a gamma-ray mutagenesis) flowered around 8 d later than its parental japonica cultivar 'Seolak' (SL), which flowered at 92 days after sowing (DAS) under natural long-day (NLD) conditions in the field (37° N latitude, Suwon, Korea). In the growth chambers, the rfs-1 mutant flowered around 8 d and 11 d later than SL under SD (10 h light/14 h dark), and LD (14.5-h light/9.5-h dark) conditions, respectively (Figure 1a,c).  To further confirm whether mutation of RFS delays rice flowering, we observed the flowering time of the rfs-2 knockout mutant, which harbors a T-DNA fragment in the 8th intron of RFS [32]. Similar to the rfs-1 mutant, the flowering time of the rfs-2 mutant was delayed by 19 d compared with its parental japonica cultivar 'Hwayoung' (HY), which flowered at 102 DAS. In addition, the rfs-2 mutant flowered around 15 d and 21 d later than HY under SD, and LD conditions, respectively (Figure 1b,d). These observations suggest that RFS is involved in the promotion of flowering, independent of photoperiod.
To test whether growth retardation causes the late flowering in rfs mutants, we measured the number of emerged leaves in HY and rfs-2 plants under SD and LD conditions [33]. The leaf emergence rates of the rfs-2 mutant were indistinguishable from those of HY under both conditions (Figure 1e,f). Therefore, the late flowering of rfs mutants was mainly due to a delayed floral transition, not to growth retardation or prolonged plastochron.

Expression Pattern of RFS
The RFS transcript levels were measured every 2 weeks from 4 to 14 weeks after sowing (WAS) at zeitgeber time (ZT) 1 by RT-qPCR analysis. RFS was constitutively expressed during plant growth, implying that RFS has roles throughout plant development ( Figure 2a). We also analyzed the diurnal expression of RFS under SD and LD conditions. The leaf blades of HY and rfs-2 plants were harvested every 3 h during a 24-h period at 25 DAS under SD conditions and at 80 DAS under LD conditions; these timepoints were approximately 30 days before flowering in HY. RFS transcripts were not detectable in rfs-2 leaf blades, but in HY, RFS showed diurnal expression patterns with a peak at night under SD conditions and at dawn under LD conditions (Figure 2b,c).

Expression Patterns of Hd3a and RFT1 in the rfs Mutants
We monitored the transcript levels of two rice florigen genes (Hd3a and RFT1) in developing leaves of HY and rfs-2 plants harvested at ZT1 throughout the vegetative stages, until the HY plants flowered. Hd3a and RFT1 transcripts gradually increased in HY with a peak at 8 WAS under SD and 15 WAS under LD conditions to trigger the transition to reproductive growth (Figure 3a-d) [34,35]. However, the expression of Hd3a and RFT1 remained low in the rfs-2 mutant throughout the experiment. The expression of Ehd1 was similar to that of the florigen genes in HY and rfs-2 plants (Figure 3e,f). These results suggest that mutation of RFS downregulates the expression of flowering genes, leading to late flowering in rice.

Expression Analysis of Flowering-Time Genes in the rfs Mutants
To reveal the regulatory effect of RFS on flowering pathways in rice, we examined diurnal changes of the expression of flowering-time genes in the leaf blades of HY and rfs-2 plants that were harvested during diurnal cycles, as shown in Figure 2b,c. The expression of Hd3a and RFT1 was mostly suppressed in the rfs-2 mutant, unlike the diurnal expression in HY under both SD and LD conditions (Figure 4a-d). Next, we measured the expression levels of Ehd1 and Hd1, which encode upstream regulators of Hd3a and RFT1 [3,4]. The expression level of Hd1 in the rfs-2 mutant was similar to that in HY under both photoperiods (Figure 4e,f). By contrast, Ehd1 transcript levels were consistently lower in rfs-2 leaves compared to HY at ZT21 and ZT0 in SD and LD conditions, respectively (Figure 4g,h).

Expression Analysis of Flowering-Time Genes in the rfs Mutants
To reveal the regulatory effect of RFS on flowering pathways in rice, we examined diurnal changes of the expression of flowering-time genes in the leaf blades of HY and rfs-2 plants that were harvested during diurnal cycles, as shown in Figure 2b,c. The expression of Hd3a and RFT1 was mostly suppressed in the rfs-2 mutant, unlike the diurnal expression in HY under both SD and LD conditions (Figure 4a-d). Next, we measured the expression levels of Ehd1 and Hd1, which encode upstream regulators of Hd3a and RFT1 [3,4]. The expression level of Hd1 in the rfs-2 mutant was similar to that in HY under both photoperiods (Figure 4e,f). By contrast, Ehd1 transcript levels were consistently lower in rfs-2 leaves compared to HY at ZT21 and ZT0 in SD and LD conditions, respectively (Figure 4g,h).

Histone Methylation Levels of Ehd1 and Ghd7 in the rfs Mutants
A previous study reported that the rfs mutant showed global reductions of histone H3K4me3 and H3K27me3 by about 56% and 23%, respectively [31]. Therefore, we conducted a chromatin immunoprecipitation (ChIP) assay to determine whether RFS modifies the histone methylation level of flowering-time genes. The leaf blades of HY and rfs-2 plants were collected at ZT1 at 9 WAS under LD conditions in the growth chamber. We speculated that downregulation of Ehd1 in the rfs-2 mutant might be caused by reduced levels of the activating mark H3K4me3 and upregulation of Ghd7 in the rfs-2 mutant might be caused by reduced levels of the repressive mark H3K27me3. (Figure 4k-r). In addition, the expression of rice EARLY FLOWERING 3 (OsELF3), a repressor of Ghd7, did not show any differences between HY and the rfs-2 mutant (Figure  4s,t). These results suggest that RFS can repress the expression of Ghd7 in LDs and activate the expression of Ehd1 in both SD and LD conditions, thereby upregulating florigen genes in rice. We further investigated whether RFS affects the expression of other epigenetic regulators including SDG701 [13], SDG708 [12], OsTrx1 [14][15][16], SDG724 [17], SDG725 [18,19], and rice VIN3-LIKE 2 (OsVIL2) [20,21]. These genes encode regulators that mediate histone methylation to promote rice flowering. The RT-qPCR analysis was performed on HY and rfs-2 leaf blades sampled at ZT0 at 4 WAS and at ZT12 at 8 WAS in SD and LD conditions, To test these hypotheses, we examined the enrichment of modified histone proteins on Ehd1 and Ghd7 chromatin using specific antibodies against H3K4me3 and H3K27me3. We used Hd3a and OsLF as positive controls for methylated histone-enriched loci. SDG701 deposits H3K4me3 on the Hd3a locus to activate Hd3a transcription [13]. The PRC2 complex targets the OsLF locus and deposits H3K27me3 to repress OsLF transcription [20,22]. We found that while H3K4me3 was highly enriched in the promoter region of Ehd1 compared to the intergenic region, the H3K4me3 level in Ehd1 did not differ between HY and the rfs-2 mutant (Figure 5a,b). H3K27me3 in Ghd7 was higher around the transcription start site than in the intergenic region. However, there was no significant difference in enrichment of H3K27me3 at the Ghd7 locus between HY and the rfs-2 mutant (Figure 5c,d).
deposits H3K4me3 on the Hd3a locus to activate Hd3a transcription [13]. The PRC2 complex targets the OsLF locus and deposits H3K27me3 to repress OsLF transcription [20,22]. We found that while H3K4me3 was highly enriched in the promoter region of Ehd1 compared to the intergenic region, the H3K4me3 level in Ehd1 did not differ between HY and the rfs-2 mutant (Figure 5a,b). H3K27me3 in Ghd7 was higher around the transcription start site than in the intergenic region. However, there was no significant difference in enrichment of H3K27me3 at the Ghd7 locus between HY and the rfs-2 mutant (Figure 5c,d).  Table S1. Black scale bars = 1 kb. (b,d) Chromatin immunoprecipitation (ChIP) analysis of the H3K4me3 level on Ehd1 (b) and the H3K27me3 level on Ghd7 (d) in the 9-week-old HY and rfs-2 plants grown in LD conditions at ZT1. Hd3a and OsLF were used as positive controls for the H3K4me3-enriched locus and the H3K27me3-enriched locus, respectively [18,25,27]. Means and standard deviations were obtained from three biological replicates. Experiments were repeated three times with similar results.

Agronomic Traits of the rfs Mutants
To identify whether the rfs mutation affects grain yield and yield components, we evaluated agronomic traits, including plant height, number of panicles per plant, length of panicles, number of panicle branches per panicle, number of grains per panicle, spikelet fertility, 500-grain weight, and yield per plant, in HY and rfs-2 plants grown in the paddy field under NLD conditions [38]. Plant height and panicle length of the rfs-2 mutant were  Table S1. Black scale bars = 1 kb. (b,d) Chromatin immunoprecipitation (ChIP) analysis of the H3K4me3 level on Ehd1 (b) and the H3K27me3 level on Ghd7 (d) in the 9-week-old HY and rfs-2 plants grown in LD conditions at ZT1. Hd3a and OsLF were used as positive controls for the H3K4me3-enriched locus and the H3K27me3-enriched locus, respectively [18,25,27]. Means and standard deviations were obtained from three biological replicates. Experiments were repeated three times with similar results.

Agronomic Traits of the rfs Mutants
To identify whether the rfs mutation affects grain yield and yield components, we evaluated agronomic traits, including plant height, number of panicles per plant, length of panicles, number of panicle branches per panicle, number of grains per panicle, spikelet fertility, 500-grain weight, and yield per plant, in HY and rfs-2 plants grown in the paddy field under NLD conditions [38]. Plant height and panicle length of the rfs-2 mutant were shorter than those of HY (Figure 6a,b). The rfs-2 mutant had more panicles per plant compared to HY, but had fewer primary and secondary branches per panicle (Figure 6c-e,j). In addition, the number of grains per panicle, spikelet fertility, and 500-grain weight were significantly lower in the rfs-2 mutant compared to HY (Figure 6f-h). The reduction of spikelet fertility in the rfs-2 mutant might be due to delayed flowering, which results in later grain filling, when conditions tend to be unfavorable due to lower temperatures. Consequently, total grain yield was lower in the rfs-2 mutant due to the reduction of yield components, including spikelet fertility, grain number, and 500-grain weight, although the mutant plants had more panicles per plant compared with HY (Figure 6i). significantly lower in the rfs-2 mutant compared to HY (Figure 6f-h). The reduction of spikelet fertility in the rfs-2 mutant might be due to delayed flowering, which results in later grain filling, when conditions tend to be unfavorable due to lower temperatures. Consequently, total grain yield was lower in the rfs-2 mutant due to the reduction of yield components, including spikelet fertility, grain number, and 500-grain weight, although the mutant plants had more panicles per plant compared with HY (Figure 6i). and error bars indicate standard deviation. Differences between means were compared using two-tailed Student's t-tests (* p < 0.05, ** p < 0.01, *** p < 0.001).

Regulatory Roles of RFS in Flowering-Time Pathways
Studies of various mutant alleles of RFS have revealed that RFS functions in multiple aspects of plant development, including crown root development, seedling development, and leaf morphogenesis [28][29][30]. In this study, we found that RFS also affects rice flowering time. Mutation of RFS delayed flowering under SD and LD conditions (Figure 1). Moreover, expression of two rice florigen genes, Hd3a and RFT1, maintained at low levels in rfs-2 mutant whereas their transcript levels were elevated in HY throughout plant growth (Figure 3). In HY, RFS expression showed a diurnal rhythm with a peak at ZT15 and ZT0 under SD, and LD conditions, respectively (Figure 2b,c). The expression of Ehd1 subsequently peaked at ZT18 in SDs and at ZT0 in LDs (Figure 4g,h). Finally, Hd3a and and error bars indicate standard deviation. Differences between means were compared using two-tailed Student's t-tests (* p < 0.05, ** p < 0.01, *** p < 0.001).

Regulatory Roles of RFS in Flowering-Time Pathways
Studies of various mutant alleles of RFS have revealed that RFS functions in multiple aspects of plant development, including crown root development, seedling development, and leaf morphogenesis [28][29][30]. In this study, we found that RFS also affects rice flowering time. Mutation of RFS delayed flowering under SD and LD conditions (Figure 1). Moreover, expression of two rice florigen genes, Hd3a and RFT1, maintained at low levels in rfs-2 mutant whereas their transcript levels were elevated in HY throughout plant growth (Figure 3). In HY, RFS expression showed a diurnal rhythm with a peak at ZT15 and ZT0 under SD, and LD conditions, respectively (Figure 2b,c). The expression of Ehd1 subsequently peaked at ZT18 in SDs and at ZT0 in LDs (Figure 4g,h). Finally, Hd3a and RFT1 transcript levels were strongly upregulated at ZT21 and ZT3 under SD and LD conditions, respectively (Figure 4a-d). Our findings strongly suggest that RFS is a floral inducer in rice.
Mutation of RFS decreased the expression of the florigen genes Hd3a and RFT1, which are rice orthologs of Arabidopsis FT, under SD and LD conditions (Figure 4a-d).
In Arabidopsis, the CHD3 protein PKL promotes flowering through activation of the FT locus, suggesting that CHD3 chromatin-remodeling factors have conserved functions in flowering plants. However, among the upstream regulators of Hd3a and RFT1, there were no differences in OsGI and Hd1 expression between HY and rfs-2 plants (Figure 4e,f,k,l), but the transcript levels of Ghd7 and Ehd1 in the rfs-2 mutant differed from those in HY (Figure 4g-j). These observations implied that RFS does not participate in the OsGI-Hd1-Hd3a regulatory pathway, which is similar to the Arabidopsis GI-CO-FT module. Instead, RFS may regulate flowering time through rice-specific flowering pathways involving Ghd7 and Ehd1 (Figure 7). Although CHD3 chromatin remodeling factors are involved in the regulation of flowering in rice and Arabidopsis, the regulatory pathways might differ between the two species. ducer in rice.
Mutation of RFS decreased the expression of the florigen genes Hd3a and RFT1, which are rice orthologs of Arabidopsis FT, under SD and LD conditions (Figure 4a-d). In Arabidopsis, the CHD3 protein PKL promotes flowering through activation of the FT locus, suggesting that CHD3 chromatin-remodeling factors have conserved functions in flowering plants. However, among the upstream regulators of Hd3a and RFT1, there were no differences in OsGI and Hd1 expression between HY and rfs-2 plants (Figure 4e,f,k,l), but the transcript levels of Ghd7 and Ehd1 in the rfs-2 mutant differed from those in HY ( Figure  4g-j). These observations implied that RFS does not participate in the OsGI-Hd1-Hd3a regulatory pathway, which is similar to the Arabidopsis GI-CO-FT module. Instead, RFS may regulate flowering time through rice-specific flowering pathways involving Ghd7 and Ehd1 (Figure 7). Although CHD3 chromatin remodeling factors are involved in the regulation of flowering in rice and Arabidopsis, the regulatory pathways might differ between the two species.

RFS May Control Expression of Flowering-Time Genes via Histone Modification
In Arabidopsis, PKL mediates deposition of H3K27me3 to repress target genes [39][40][41][42], and mutation of CHR4 changes H3K27me3 and H3K4me3 levels [27]. Moreover, RFS is responsible for modification of H3K27me3 and H3K4me3 to control gene expression [31,32,43]. We therefore speculated that RFS controls the expression of Ghd7 and Ehd1 through H3K27me3, and H3K4me3, respectively. However, our ChIP analysis showed no apparent differences in H3K4me3 levels of Ehd1 or H3K27me3 levels of Ghd7 between HY and the rfs-2 mutant ( Figure 5). One possible explanation for this is that RFS regulates Ehd1 and Ghd7 via other types of modifications, such as histone acetylation. In mammalian cells, CHD3 is a component of the nucleosome remodeling and deacetylase (NURD) complex, which is associated with histone deacetylation, resulting in transcriptional repression [44,45]. In addition, RFS might control an unidentified upstream regulator of Ehd1 and Ghd7. A previous study proposed that OsTrx1 may be an upstream regulator of Ghd7

RFS May Control Expression of Flowering-Time Genes via Histone Modification
In Arabidopsis, PKL mediates deposition of H3K27me3 to repress target genes [39][40][41][42], and mutation of CHR4 changes H3K27me3 and H3K4me3 levels [27]. Moreover, RFS is responsible for modification of H3K27me3 and H3K4me3 to control gene expression [31,32,43]. We therefore speculated that RFS controls the expression of Ghd7 and Ehd1 through H3K27me3, and H3K4me3, respectively. However, our ChIP analysis showed no apparent differences in H3K4me3 levels of Ehd1 or H3K27me3 levels of Ghd7 between HY and the rfs-2 mutant ( Figure 5). One possible explanation for this is that RFS regulates Ehd1 and Ghd7 via other types of modifications, such as histone acetylation. In mammalian cells, CHD3 is a component of the nucleosome remodeling and deacetylase (NURD) complex, which is associated with histone deacetylation, resulting in transcriptional repression [44,45]. In addition, RFS might control an unidentified upstream regulator of Ehd1 and Ghd7. A previous study proposed that OsTrx1 may be an upstream regulator of Ghd7 [14]. In the absence of OsTrx1, which activates target genes by adding the H3K4me3 mark, the expression of Ghd7 increased, suggesting that Ghd7 expression might be indirectly controlled by OsTrx1. Therefore, further analysis and exploration will be required to elucidate the involvement of RFS in flowering.
ATP-dependent chromatin remodeling factors act as multimeric complexes in various organisms. In this study, we could not observe significant alteration in transcript levels of other epigenetic regulators that promote rice flowering, including SDG701 [13], SDG708 [12], OsTrx1 [14][15][16], SDG724 [17], SDG725 [18,19], and OsVIL2 [20,21] between HY and rfs-2 ( Figure S1). These results suggest that RFS is not directly involved in the transcriptional regulation of these epigenetic regulators. Therefore, we could not exclude the possi-bility that RFS recruits these regulators to modify the histone methylation of flowering-time genes. Recent biochemical approaches using immunoprecipitation and mass spectrometry have identified the components of ISWI and SWI/SNF chromatin remodeling complexes in plants [46,47]. In Arabidopsis, PKL and CHR4 recruit several transcription factors and cofactors for proper epigenetic regulation during development [25][26][27]48]. In rice, the components of complexes that associate with CHD3 have not yet been elucidated. Therefore, it will be interesting to find the components that physically interact with RFS during rice development, including flowering, crown root development, seedling development, and leaf morphogenesis.

RFS Is Involved in Inflorescence Development
CHR4 plays diverse roles in the inflorescence meristem to promote flowering in Arabidopsis [27]. CHR4-deficient plants show an accelerated transition from the vegetative phase to bolting, but a delay in the formation of floral primordia. CHR4 interacts with MADS, SQUAMOSA promoter-binding protein-like (SPL), and APETALA2 (AP2)-type transcription factors, which regulate the floral transition and floral meristem identity [24,49]. In addition, RNA-sequencing and ChIP sequencing have revealed that CHR4 mediates the response to endogenous flowering pathways in the inflorescence meristem by controlling the expression of floral regulators [27]. In rice, overexpression of ABERRANT PANICLE ORGANIZATION 1 (APO1), which encodes an F-box protein, causes a precocious transition of the inflorescence meristem to the spikelet meristem, reducing the number of panicle branches [50]. The rfs-2 mutant had significantly decreased grain yield compared to HY (Figure 6i). Based on these findings, we speculated that RFS might control other developmental processes, as well as flowering. Among the agronomic traits that affect grain yield, the number of secondary panicle branches was markedly reduced in the rfs-2 mutant (Figure 6e), implicating RFS in inflorescence meristem fate. Because the inflorescence architecture of rice is closely related to yield, the role of RFS in the transition from the vegetative meristem to the inflorescence meristem will be an interesting topic for future studies.

Plant Materials and Growth Conditions
The rfs-1 and rfs-2 mutants used in this study were reported previously [32]. Rice plants were grown in the paddy field under NLD conditions (≥14 h sunlight/day, 37 • N latitude) in Suwon, Republic of Korea. To perform experiments under controlled daylength conditions, rice plants were grown in growth chambers under LD (14.5-h light and 9.5-h dark) or SD (10-h light and 14-h dark) conditions at 30 • C in the light and 24 • C in the dark.

RNA Extraction and Reverse Transcription-Quantitative PCR (RT-qPCR) Analysis
Total RNA was isolated using an RNA extraction kit (MG Med, Seoul, Korea) according to the manufacturer's manual. After RNA quantification, cDNA was synthesized from 2 µg of total RNA by M-MLV reverse transcriptase (Promega, Madison, WI, USA) at 42 • C for 1 h after priming with oligo(dT) (Promega, Madison, WI, USA) at 70 • C for 5 min. Prepared cDNAs were diluted to 100 µL with distilled water and then used as templates for RT-qPCR. RT-qPCR was performed with a LightCycler 480 (Roche, Basel, Switzerland) using 2× GoTaq master mix (Promega, Madison, WI, USA) in a 20-µL reaction volume. Rice UBIQUITIN 5 (LOC_Os01g22490) served as an internal control for relative quantification. The primer sequences used for RT-qPCR are listed in Supplemental Table S1.

Chromatin Immunoprecipitation (ChIP) Assay
ChIP was performed as previously described [51]. Leaf blades of 63-d-old plants grown in the LD growth chamber were harvested at ZT1. One gram of leaves was crosslinked in 1% formaldehyde under vacuum. After isolation of nuclei, chromatin was sheared into 500-1000 bp in length by sonication and then immunoprecipitated with anti-H3K4me3 (Millipore, Temecula, CA, USA) or anti-H3K27me3 (Millipore, Temecula, CA, USA) antibodies. The immunoprecipitated products and 5% of input chromatin were reverse-crosslinked at 65 • C and eluted with QIAquick PCR purification kit (Qiagen, Hilden, Germany). Finally, the precipitated DNA was quantified by qPCR with the primers listed in Supplemental Table S1.

Measurement of Agronomic Traits
To investigate agronomic traits, HY and rfs-2 plants were grown in the paddy field under NLD conditions. Plant height was measured just after heading. The other traits, such as panicle length, the number of panicle branches, the number of grains, fertility, 500-grain weight, and yield per plant, were examined after harvest. The panicles of main tillers were used to analyze panicle length, the number of panicle branches, and the number of grains.

Conclusions
In this study, we show that a CHD3/Mi-2 chromatin remodeling factor, RFS, is involved in the regulation of flowering induction in rice. RFS loss-of-function mutants exhibited a late-flowering phenotype under SD and LD conditions. RFS promotes the expression of two rice florigen genes, Hd3a and RFT1, through downregulation of Ghd7 and upregulation of Ehd1, thereby promoting floral induction. This study, thus, provides evidence that a chromatin remodeling factor plays crucial roles in rice flowering.

Data Availability Statement:
The data presented in this study are available in the article.

Conflicts of Interest:
The authors have no potential conflict of interest.