Rice cultivars Nipponbare (japonica), DZ60 (tropical japonica), long vegetative phase2-1 (lvp2-1) mutant and eight editing lines including lvp2-2 to lvp2-7, Oself4a-1, and Oself4a-2 were used in the study. The lvp2-1 mutant was obtained from the progeny of EMS mutagenesis of Nipponbare; lvp2-2 to lvp2-7, Oself4a-1, and Oself4a-2 were generated from Nipponbare using the CRISPR/Cas9 system.

To map LVP2 locus, an F₂ population was developed from the cross between lvp2-1 and DZ60. To avoid the interfering effects of other flowering genes, a total of 389 progeny lines (F_2:3) which display extremely late flowering and other phenotypic characteristics similar to lvp2-1 were selected for gene mapping. LVP2 was mapped to a rice chromosome by using the publicly available RM-series rice microsatellite markers (McCouch et al., 2002) and bulked segregant analysis (BSA) method. Furthermore, mapping of LVP2 gene was also performed by deep sequencing-based bulked segregant analysis (BSA-seq) based on an F₃ population developed from the cross between lvp2-1 and Nipponbare. To make the wild-type DNA pool and the mutant DNA pool, equal amount of fresh leaves from 30 homozygous wild-type lines (5 plants/line) and 30 mutant lines randomly selected from the F₃ population were mixed for genomic DNA extraction, respectively. The two DNA pools were sequenced using the Illumina PE Genome Analyzer. Based on the sequence information (read length = 300 bp, sequencing depth > 30×) of the two pools, the LVP2 locus was mapped.

For genetic complementation test, a 3,420-bp genomic DNA sequence covering the promoter region, gene region, and downstream region of OsLUX was amplified from Nipponbare, and inserted into the binary vector pCAMBIA1300. The plasmid was introduced into lvp2-1 mutant embryonic calli. To confirm the function of OsLUX and to identify the function of OsELF4a, the CRISPR/Cas9 system was used, and the CRISPR targets for OsLUX and OsELF4a were selected as described by Miao et al. (2013). The vector construction was performed according to the manufacturer’s instruction of regent kit (VIEWSOLID Biotech, China). The vectors were then transferred into Nipponbare. The genomic regions surrounding the CRISPR target sites for OsLUX and OsELF4a were amplified by PCR using primer pairs 5′-tcgagtccccgatttggttc/5′-gcttcacgtagaggcgatact and 5′-atggaaggtgatagcttct/5′-gccgggccggacacgcttc, respectively, and were sequenced to identify mutants. For overexpression test, a 717-bp cDNA of OsLUX was amplified and fused into the downstream of maize Ubiquitin promoter of the binary vector pTCK303. The plasmid for overexpression was introduced into Nipponbare. To determine the expression pattern of OsLUX, a 2021-bp promoter upstream of the coding region of OsLUX was amplified from Nipponbare, and fused into the GUS reporter gene with the nopaline synthase terminator in pCAMBIA1391Z. The plasmid was introduced into Nipponbare. Histochemical assay for GUS activity in transgenic plants was performed as described (Jefferson et al., 1987).

All constructs were confirmed by sequencing, introduced into Agrobacterium tumefaciens strain EHA105, and transferred into the rice variety by Agrobacterium-mediated transformation. The primers used for vector construction are listed in Supplementary Table S1.

OsLUX-GFP fusion was made by in-frame fusion of the 717-bp full-length coding sequence of OsLUX into vector pRTVcGFP (He et al., 2018). OsLUX-GFP and GFP alone (as control) were transferred into rice protoplasts. NLS-RFP was co-transfected for nuclear localization control. The samples were observed with a confocal laser scanning microscope (LEICA-SP8). Primer pairs for amplifying OsLUX cDNA are listed in Supplementary Table S1.

For Y2H assay, the coding sequences of OsLUX and OsELF4a were amplified from Nipponbare and inserted into pGBKT7. The coding region of OsELF3a and OsELF3b were amplified from Nipponbare and inserted into pGADT7. Different combinations of bait and prey were used to co-transform the yeast competent cell Y187. The transformed yeast cells were grown on non-selective SD media (−Trp-Leu/−LT) and selective SD media (−Trp-Leu-His-Ade/-LTHA) with X-α-Gal. A positive interaction was judged by the growth of yeast colonies on selective media. The primers used for vector construction are listed in Supplementary Table S1.

The CDSs of OsLUX, OsELF3a, OsELF3b, OsELF4a, OsELF4b, and OsELF4c (without termination codon) were amplified and inserted into the entry vector pDONR207 via a BP reaction, and then transferred into the N-terminal and the C-terminal of pSPYNE or pSPYCE (Choi et al., 2012) via a LR reaction, to form fusion protein NE-LUX, CE-LUX, NE-ELF3a, NE-ELF3b, CE-ELF4a, CE-ELF4b, and CE-ELF4c, respectively. Equal concentration mixture of NE-ELF3a and CE-LUX, NE-ELF3b and CE-LUX, NE-ELF3a and CE-ELF4a, NE-ELF3b and CE-ELF4a, NE-ELF3a and CE-ELF4b, NE-ELF3b and CE-ELF4b, NE-ELF3a and CE-ELF4c, NE-ELF3b and CE-ELF4c, and NE-LUX and CE-ELF4a were used to co-transform rice protoplasts. The vector combinations of CE + NE-LUX, NE + CE-LUX, CE + NE-ELF3a, CE + NE-ELF3b, NE + CE-ELF4a, NE + CE-ELF4b, and NE + CE-ELF4c were used as negative controls to verify the specificity of the interactions. Fluorescence in transformed protoplast was examined and imaged using a confocal laser scanning microscope (LEICA-SP8). The primers used for vector construction are listed in Supplementary Table S1.

To further validate the interaction between OsLUX and interacted partners in vivo, Co-IP assay was conducted. The ORFs of OsLUX and OsELF4a were amplified and inserted into pRTVcGFP (He et al., 2018) to generate expression vectors pRTVcGFP-OsLUX and pRTVcGFP-OsELF4a. Full-length cDNAs of ELF3a and ELF3b were amplified and cloned into the psuper1300-MYC. The Co-IP assay was performed as described (Kong et al., 2019). Briefly, the plasmids were used to transiently co-transform rice protoplasts from ten-day-old etiolated seedlings. After incubation for 16 h, total proteins were extracted from the protoplasts. Lysis buffer (50 mM Tris-MES pH = 0.8, 0.5Msucrose, 1 mM MgCl2, 10 mM EDTA, 1 mM PMSF, 0.5 mM DTT, 0.1% Trixon-100) was added to protease inhibitor cocktail (Sigma-Aldrich, P9599). The supernatant was collected after centrifugation 16,000 g for 30 min at 4°C and precleared with GFP-Nanoab-Agarose (Lablead, GNA-20-400) for 1 h. The GFP-Nanoab-Agarose was washed four times in 0.5 ml lysis buffer and then boiled in the 2XSDS-PAGE sample buffer. The immunoprecipitation products were detected by SDS-PAGE and western blot using anti-Myc (MBL, M047-3) or anti-GFP (MBL, 598) antibody. The primers used for vector construction are listed in Supplementary Table S1.

RNA isolation and qRT-PCR analysis were performed as described (Duan et al., 2019). Briefly, total RNA was isolated using TRIzol reagent kit (Invitrogen, Carlsbad, CA, United States). Reverse transcription of total RNA was performed using Primescript^™ RT reagent kit (Takara, China). The cDNA samples were diluted to 8 ng/μl and 2 ng/μl. Triplicate quantitative assays were performed using the SYBR Premix Ex Taq II (Takara, China) with a Mastercycler ep realplex⁴ sequence detection system (Eppendorf). Amplification of UBQ was used as internal control to normalize all data. Primers used for qRT-PCR analysis are listed in Supplementary Table S2.

Sequence data from this article for the mRNA, cDNA, and genomic DNA can be found in the GenBank/EMBL/Gramene data libraries or Web site under accession numbers:

OsLUX, LOC_Os01g74020; OsGI, LOC_Os01g08700; OsLHY, LOC_Os08g06110; OsPRR37, LOC_Os07g49460; OsPRR95, LOC_Os09g36220; OsELF3a, LOC_Os06g05060; OsELF3b, LOC_Os01g38530; OsELF4a, LOC_Os11g40610; OsELF4b, LOC_Os03g29680; OsELF4c, LOC_Os08g27860; Hd1, LOC_Os06g16370; Hd3a, LOC_Os06g06320; Ghd7, LOC_Os07g15770; Ehd1, LOC_Os10g32600.

The lvp2-1 mutant, which displayed a long staying-green period and extremely late flowering similar to that of previously reported lvp2 mutant (Sun et al., 2012), was identified from the progeny of EMS mutagenesis of Nipponbare, a japonica rice variety (Figure 1A). To identify the photoperiod response of lvp2-1, we examined the heading date of lvp2-1 under the natural LD from July to September (NLD, ≥14 h light) condition in Fuzhou (26°08′) and the natural SD from January to March (NSD, ≤10 h light) condition in Sanya (18°10′). The wild type (WT) headed 24.9 days earlier in Sanya (64.6 days) than in Fuzhou (89.5 days; Table 1), whereas the mutant showed extremely late flowering in both Fuzhou (143.4 days) and Sanya (142.5 days), with only about 1 days delay under NLD (Table 1). To confirm this result, we grew lvp2-1 and WT under both SD (10/14 h light/dark) and LD (14/10 h light/dark) conditions in the growth chamber. The photoperiod response of lvp2-1 in the growth chamber was completely consistent with that under the natural field conditions (Table 1). These observations indicated that lvp2-1 was photoperiod-insensitive, suggesting that LVP2 plays an important role in increasing photoperiod sensitivity and acts as a flowering promoter under all photoperiod conditions in rice.

lvp2-1 exhibited vigorous growth throughout the entire vegetative phase under both SD and LD conditions. We compared several morphological traits of lvp2-1 with those of WT under the natural day length condition in Fuzhou, including size of leaf, number of leaves per plant, number of tillers per plant, culm length, and so on. This mutant showed similar phenotypes (Table 1). Compared with WT, lvp2-1 had significantly longer and wider leaves and taller culms (Figures 1A,B; Table 1), and produced much more leaves and tillers (Table 1), but its leaf emergence rate and tiller emergence rate remained almost the same as those of WT (Figures 1B,C). In addition, the mutant generated more lateral roots than WT (Figure 1C), which would possibly enable itself to obtain more nutrition for growth. These observations suggest that LVP2 controls biomass-driven growth. In the reproductive phase, no remarkable differences were found between lvp2-1 and WT in panicle length, number of grains per panicle, seed-setting rate, grain size, grain shape, and 1,000-grain weight (Table 1). However, lvp2-1 produced over twice as many grains as those produced by WT due to its dramatic increase in effective tillers/panicles (Figure 1E; Table 1), suggesting that lvp2-1 may have the potential to be used for increasing yield in breeding programs. Besides, LVP2 showed the function of repressing awn development. Most of the lvp2-1 grains had a long awn, while the WT grains were hardly awned (Figure 1D). In summary, LVP2 had pleiotropic effects on vegetative growth and reproductive development, and loss of LVP2 function would dramatically increase the biomass and grain yield of the plant.

Genetic analysis showed that the phenotype of lvp2-1 was caused by a single gene mutation (Supplementary Table S3). To identify the LVP2 locus, we crossed lvp2-1 with variety DZ60 to develop an F₂ population, and a total of 389 progeny lines (F_2:3) which display extremely late flowering and other phenotypic characteristics similar to lvp2-1 were selected for gene mapping. Using the RM-series rice microsatellite markers, we mapped LVP2 on the terminal region of the long arm of chromosome 1, linked with RM414. To identify the candidate gene, we performed whole-genome resequencing of the DNA pools of WT and mutant, each consisting of 30 lines. Based on the sequence information obtained from resequencing, the LVP2 locus was also mapped on the terminal region of the long arm of chromosome 1, agreeing with the results of linkage-based mapping (Figure 2A). By comparing the nucleotide sequences of the genomic fragments in this region, we found that there was a single base transition from C (CGG) to T (TGG) in gene LOC_Os01g74020 (OsLUX), which resulted in a change from Arginine to Tryptophan at the 120^th amino acid (aa) of OsLUX in lvp2-1 (Figures 2B,C). OsLUX is a MYB-like transcription factor consisting of 238 aa homologous to Arabidopsis LUX (Supplementary Figure S1). The mutation in lvp2-1 was located within the crucial GARP domain (117^th–175^th aa) in OsLUX (Supplementary Figure S2). The ORF sequences of all the other genes in this region remained unchanged between WT and lvp2-1. Expression analysis showed that the rhythm pattern of OsLUX expression in lvp2-1 was almost the same as that in WT, but the expression level in lvp2-1 was significantly downregulated (Supplementary Figure S3). These results strongly suggested that the lvp2-1 phenotype was caused by the mutation in OsLUX.

To validate the candidate gene, we constructed a complementary plasmid pCAMBIA1300-OsLUX carrying a 3,420-bp wild-type genomic DNA, which covered the entire OsLUX gene (717-bp) and its upstream promoter region. By introducing the plasmid into lvp2-1, the mutant phenotype was rescued and restored to that of WT in all of 25 transgenic lines (Figure 2D; Supplementary Table S4), confirming that OsLUX is LVP2. To further validate and investigate OsLUX, we created the loss-of-function mutants of OsLUX from WT Nipponbare using CRISPR/Cas9 technology. Eleven effectively edited lines were obtained, which could be classified into six types, including one-base insertion between the 55^th and the 56^th bases (Types 1–3, 6 lines), deletion of the 55^th base (Type 4, 3 lines), 16-base deletion of 42^nd-57^th base (Type 5, 1 line), and 274-base deletion of 30^th-303^rd base (Type 6, 1 line). We named these lines as lvp2-2 to lvp2-7. All of the edited alleles were predicted to encode a short peptide lacking the crucial GARP domain (Supplementary Table S5). Hence, all of them should be non-functional. As expected, all of the edited lines exhibited the same phenotype as that of lvp2-1 in photoperiod sensitivity, heading date, and other characteristics under various photoperiod conditions (Table 1; Figure 2D), indicating that lvp2-1 should be a non-functional allele as well. We also developed OsLUX overexpression lines, but the gene expression was not noticeably increased and no significant phenotype alteration was observed in the transgenic lines (Supplementary Figure S4), suggesting that OsLUX represses the expression of itself, similar to LUX in Arabidopsis.

To investigate whether the expression of OsLUX is rhythmic, we assayed its expression every 4 h in a 48-h time course under both LD (14 h light) and SD (10 h light) conditions. The transcript level of OsLUX showed an obvious diurnal circadian rhythm of the daily 24 h cycle, in which it started to accumulate at ZT8, reached the peak approximately at the dusk, and slowly decreased during the night under both LD and SD, but the peak under LD was almost twice as high as that under SD (Figure 3A). To examine whether OsLUX expression is organ-specific, quantitative real-time PCR (qRT-PCR) was performed for different organs. OsLUX expression was detected in all the organs examined, with the highest accumulation occurring in 5-mm-long young panicle and developing floral organs, followed by shoot and leaf, and weak in shoot, root, and stem (Figure 3B). To confirm the results of qRT-PCR analysis and intuitively display the spatio-temporal expression characteristic of OsLUX, we fused the promoter of OsLUX with GUS (β-glucuronidase) reporter gene and transferred the fused vector into rice with the mediation of Agrobaterium tumefaciens. Totally, 23 transgenic lines were obtained (Supplementary Figure S5). The results of GUS activity examination indicated that OsLUX was expressed in all organs including root, stem, leaf, young panicle, spikelet, and developing seed (Figures 3C–O). This was agreement with the results of qRT-PCR. However, GUS analysis indicated that OsLUX expression was not uniform within organs. In root, OsLUX expression was basically limited to the actively developing young roots and young root hairs, but little in the mature zones (Figure 3C). Developing leaf sheath displayed stronger expression than developing leaf blade (Figures 3D,E). Elongating stem showed higher expression than mature stem (Figure 3F). OsLUX expression was strong in developing young panicle and the first internode from top (Figure 3G), but decreased later (Figure 3H). In developing spikelet, OsLUX expression was observed in all floral organs, with the strongest in the lemma/palea and receptacle (Figures 3I–L). In addition, OsLUX expression was also detected in developing seed (Figures 3M,N) and the germ and radicle of germinating seed (Figure 3O).

Arabidopsis LUX affects the expression of circadian clock genes. To investigate the effect of OsLUX mutation on the circadian clock, we compared the diurnal expression of several known or likely-to-be circadian clock genes in WT and that in lvp2-1 under both SD and LD, including five with known function (OsGI, OsPRR37, OsLHY, OsELF3a, and OsELF3b), one unidentified function (OsPRR95), and three speculative OsELF4 genes (OsELF4a, OsELF4b, and OsELF4c; Supplementary Figure S6). The rhythmic expression patterns of these genes in lvp2-1 were similar to those in WT under both SD and LD conditions, but the expression levels of the most genes changed (Figure 4). Among them, OsGI, OsPRR37, and OsPRR95 showed higher expression peaks in lvp2-1 than in WT (Figures 4A–F), suggesting that OsLUX negatively regulates the expression of these genes under both day-length conditions, whereas the expression of OsLHY, OsELF3a, OsELF3b, and OsELF4a were dramatically reduced in lvp2-1 mutant (Figures 4G–N), suggesting that OsLUX positively regulates the expression of these genes under both day-length conditions. However, OsELF4b and OsELF4c did not show significant change under SD and LD conditions (Figures 4O–R). These results suggest that OsLUX is essential to the circadian function through positive or negative regulation of the expression of genes associated with the circadian clock in rice.

In rice, Ehd1 and Hd1 determine the degree of expression of florigenic genes Hd3a and RFT1 through distinct pathways under a given photoperiod. To identify potential downstream flowering time genes regulated by OsLUX, we examined the diurnal expression of five photoperiod flowering-related integrator genes, Hd1, Ghd7, Ehd1, Hd3a, and RFT1, under both SD and LD. Under SD, the peak of Hd1 expression was lower in lvp2-1 than in WT (Figure 5A), and almost no Hd3a expression could be detected in lvp2-1 (Figure 5B). This was likely to be directly responsible for the later flowering of lvp2-1 under SD. Meanwhile, the peak of Ghd7 expression was higher in lvp2-1 than in WT (Figure 5C), suggesting that OsLUX represses Ghd7 expression; while the peak of Ehd1 expression was lower in lvp2-1 than in WT (Figure 5D), due to the repression of Ghd7. As expected, the expression of these genes except RFT1 under LD (Figures 5E,J) showed similar alternation patterns to those under SD in lvp2-1 (Figures 5F–I), consistent with the mutant phenotype of photoperiod-insensitive late flowering (Figure 1). Under SD, very low RFT1 expression was detected in both WT and lvp2-1 (Figure 5E), whereas under LD, the peak of RFT1 expression was markedly lower in lvp2-1 than in WT (Figure 5J), which led to late flowering in lvp2-1. Taken together, OsLUX is essential for the photoperiodic control of rice flowering. OsLUX functions as a floral promoter by promoting Hd1 and Ehd1 expression and repressing Ghd7 expression under both SD and LD conditions, so as to regulate the expression of downstream florigenic genes Hd3a and RFT1.

To investigate the intracellular location of OsLUX, we constructed an expressional vector of the fusion gene of GFP (green fluorescent protein) and OsLUX to transform rice protoplasts. The green fluorescent signal was observed specifically in nucleus (Supplementary Figure S7), suggesting that OsLUX was anchored in nucleus.

In rice, two ELF3 orthologs, OsELF3-1/OsELF3a and OsELF3-2/OsELF3b, have been identified, but no apparent ELF4 ortholog has been found (Murakami et al., 2007). To identify the putative rice ELF4, we used the full-length sequence of ELF4 to perform BLAST analyses. Four ELF4 homologues were found from the rice genome sequence, namely, LOC_Os11g40610, LOC_Os03g29680, LOC_Os08g27860, and LOC_Os08g27870, which showed considerable similarities to ELF4 in amino acid sequence. Protein sequence alignment and phylogenetic analysis showed that these ELF4 homologues contained a function-unknown evolutionarily conserved domain DUF1313 (Supplementary Figure S6; Kolmos et al., 2010; Lin et al., 2019). Among these genes, LOC_Os08g27870 transcript was undetectable in rice leaf in this study. Hence, we named LOC_Os11g40610, LOC_Os03g29680 and LOC_Os08g27860 as OsELF4a, OsELF4b and OsELF4c, respectively.

To investigate the components of EC in rice, we performed bimolecular fluorescence complementation (BiFC) assay to test the interactions of OsLUX with two OsELF3s and three OsELF4s. The coding sequences of OsLUX, OsELF3a, OsELF3b, OsELF4a, OsELF4b, and OsELF4c were fused into the N- and/or C-terminal of vectors pSPYNE and pSPYCE, respectively. No fluorescence was observed in the mixed carrier of all the combinations of NE-LUX+CE, NE + CE-LUX, NE-ELF3a + CE, NE-ELF3b + CE, NE + CE-ELF4a, NE + CE-ELF4b, and NE + CE-ELF4c (data not shown). Green fluorescence was observed from the nuclei of rice protoplasts co-transformed with NE-OsELF3a + CE-OsLUX (Figure 6A), NE-OsELF3b + CE-OsLUX (Figure 6B), NE-OsELF3a + CE-OsELF4a (Figure 6C), and NE-OsELF3b + CE-OsELF4a (Figure 6D). However, no green fluorescence was observed using the mixed carrier of NE-OsLUX+CE-OsELF4a (Figure 6E), NE-OsELF3a + CE-OsELF4b (Figure 6F), NE-OsELF3b + CE-OsELF4b (Figure 6G), NE-OsELF3a + CE-OsELF4c (Figure 6H), and NE-OsELF3b + CE-OsELF4c (Figure 6I). The interactions between OsLUX and OsELF3a, OsLUX and OsELF3b, OsELF4a and OsELF3a, and OsELF4a and OsELF3b were further confirmed by yeast two-hybrid assay (Figure 6K) and in vivo Co-IP assay (Figures 6L–O). These results indicated that OsLUX forms two ECs with OsELF4a and OsELF3a or OsELF3b, respectively; but OsELF4b and OsELF4c cannot form EC with OsLUX and OsELF3a or OsELF3b. We named the two ECs as LUX-ELF3a-ELF4a and LUX-ELF3b-ELF4a in rice.

In Arabidopsis, disruption of ELF4 results in early flowering and attenuation of rhythmicity, while overexpression of ELF4 delays flowering (Doyle et al., 2002; Kikis et al., 2005). In addition, two ELF4-Like genes, EFL1 and EFL3, are involved in flowering time regulation. Overexpression of EFL1 in elf4 shows normal rhythmicity and delays flowering; overexpression of EFL3 in elf4 partially rescues the early flowering phenotype of elf4 (Lin et al., 2019). To characterize the function of OsELF4a, we created the loss-of-function mutants of OsELF4a from Nipponbare (WT) using the CRISPR/Cas9 method. In total, five transgenic lines with effective editing in the ORF of OsELF4a were obtained (Supplementary Table S6), which could be classified into two types. Four lines harbored a 44-base deletion from 107^th to 150^th, and one line harbored a 43-base deletion from 107^th to 149^th. The two types of deletions were both predicted to encode a truncated short polypeptide consisting of 35 aa plus a missense sequence, leading to complete loss of the conserved DUF1313 domain (Supplementary Figure S6, Supplementary Table S6). Thus, both of them should be null alleles of OsELF4a. There was no significant phenotypic difference between these two types of edited lines. We named them Oself4a-1 and Oself4a-2, respectively.

To investigate whether OsELF4a is involved in flowering time regulation, we planted Oself4a mutants and examined their heading dates under the NLD of summer in Fuzhou and the NSD of winter in Sanya. The Oself4a mutants showed a marginal (1 or 2 days of) delay in flowering under NSD (Figure 7A), but much greater delay (9 ± 0.45 days, n > 30) under NLD compared with WT (Figure 7B). Other main agronomic traits all remained unchanged between WT and the Oself4a mutants under both NSD and NLD (Figures 7A,B). To confirm this result, we grew the Oself4 mutants and WT under SD (10/14 h light/dark) and LD (14/10 h light/dark) conditions in the growth chamber. The results agreed with those under the natural field conditions (Figures 7C,D). These observations reveal that OsELF4a functions as a floral promoter under LD condition.

The EC components LUX, ELF3, and ELF4 can be found in the genomes of land plant lineages, and increasing evidence has shown that EC genes function in flowering time. In Arabidopsis, EC members share a similar expression pattern and almost conserved molecular function. Disruption of any of its components leads to common phenotypic defects, displaying arrhythmic circadian period, long hypocotyl, and photoperiod-insensitive early flowering (Nusinow et al., 2011). Meanwhile, overexpression of either ELF3 or LUX complements the elf4 mutant phenotype (Herrero et al., 2012). These cross-gene complementation in Arabidopsis reveals an identical functions of EC genes. Interestingly, cross-species complementation for Arabidopsis mutants have also been observed in EC gene orthologs of other plants. For example, both the ELF3 orthologs from monocot Brachypodium and Setaria (Huang et al., 2017) and those from dicot LD plant pea (Weller et al., 2012) and SD plant soybean (Lu et al., 2017) can rescue Arabidopsis elf3 phenotype. Similar to ELF3, soybean LUX2 can complement Arabidopsis lux mutants as well (Liew et al., 2017; Fang et al., 2021). These cross-species complementation between monocots and dicots, and LD and SD plants reveal a widely conserved function in EC genes across plant species.

In addition, the potential divergent roles of EC genes have been observed across species as well. In rice, four EC gene orthologs have been characterized. We have seen that OsLUX, OsELF3a/b, and OsELF4a are essential for promoting photoperiodic flowering, but their mutant phenotypes in flowering time and photoperiodic response are significantly different (Figures 1, 7; Saito et al., 2012; Zhao et al., 2012; Yang et al., 2013; Wang et al., 2021). Apart from affecting photoperiodic flowering, OsLUX also showed profound effects on the global regulation of growth and development in this study. The non-functional alleles altered some pivotal agronomic traits, displaying extremely long juvenile phase and vigorous biomass-driven growth, which led to producing more leaves, tillers, and lateral roots (Figures 1, 2E; Table 1) and directly contributed to the high biomass and grain yield (Figure 1E). OsELF3a has pleiotropic effects as well, but its mutation mainly influences the traits of panicle number and panicle length (Saito et al., 2012), which are markedly different from those of lvp2 (Figures 1, 2; Table 1). Disruption of OsELF4 did not cause significant changes in other traits in all day length. These observations showed that, unlike the case in Arabidopsis, the fundamental functions of the four rice EC genes, OsLUX, OsELF4, OsELF3a/3b, are not completely consistent with each other in most agronomic characters including flowering time.

Besides, we have also noticed that EC genes have inhibiting effect on flowering in LD species, such as pea, lentil, wheat, and barley. All mutant alleles of EC gene orthologs identified in these species show early flowering regardless of day length, resembling the mutant alleles of LUX, ELF3, and ELF4 in Arabidopsis (Liew et al., 2009; Faure et al., 2012; Mizuno et al., 2012; Weller et al., 2012; Zakhrabekova et al., 2012; Campoli et al., 2013; Gawroński et al., 2014; Boden et al., 2014; Liew et al., 2014; Mizuno et al., 2016). However, all mutant alleles of LUX, ELF3, and ELF4 orthologs identified so far in SD crops, such as rice and soybean, have later flowering time than wild-type lines (Lu et al., 2017; Bu et al., 2021; Fang et al., 2021; Wang et al., 2021; Figure 7), implying that EC genes promote flowering in SD plants, opposite to the counterparts in LD plants. The functional difference of EC components in SD plants and LD plants probably resides in the regulation or activity of downstream integrator components, rather than in the EC gene itself. This speculation, however, needs further experimental verification.

Photoperiodic flowering in plant is closely tied to EC, a core component of circadian clock. In Arabidopsis, EC plays a vital role in circadian rhythms and flowering. As an integrant of repressing flowering, EC represses the expression of circadian genes PRR7, PRR9, and GI directly and of output genes, such as CO, FLC, and FT indirectly, so as to control flowering time (Nusinow et al., 2011; Chow et al., 2012; Mizuno et al., 2014; Song et al., 2015; Silva et al., 2020). Unlike the case in Arabidopsis, where only one FT presents, there are two florigens in rice, Hd3a and RFT1, and thus two pathways, including the conserved OsGI-Hd1-Hd3a pathway (similar to the Arabidopsis GI-CO-FT) and a unique Ghd7-Ehd1-Hd3a/RFT1 pathway, have evolved to adapt to the photoperiodic flowering under SD or LD condition. Each pathway is regulated by the coincidence of the internal circadian clock and the external photoperiodic information (Song et al., 2015).

In rice, OsLUX interacts with OsELF4a and OsELF3a or OsELF3b to form two ECs (Figure 6). Nevertheless, the interaction between OsELF3a and OsLUX is stronger than that between OsELF3b and OsLUX (Wang et al., 2021). Two ELF3 orthologs are likely simultaneously required for the control of flowering time, but OsELF3a plays a more predominant role (Saito et al., 2012; Zhao et al., 2012; Yang et al., 2013), while the contribution of OsELF3b to flowering time remains controversy (Fu et al., 2009; Zhao et al., 2012). These findings suggest that the OsLUX-OsELF3a-OsELF4a is likely the dominant promoter for photoperiodic flowering.

We found that OsLUX participated the regulation of rice photoperiodic flowering via coordinating the expression of genes associated with the circadian clock and the output integrators under both LD and SD conditions. As a transcriptional repressor, OsLUX suppressed Ghd7 expression, while its non-functional mutation elevated the expression of Ghd7 under both LD and SD, leading to reduced expression of Ehd1, Hd3a, and RFT1 (Figures 4, 5). Besides, OsLUX also negatively regulated the expression of OsGI (Figures 4A,B) and OsPRR37 (Figures 4C,D), of which the former is responsible for Hd1 and Ehd1 expression (Hayama et al., 2003; Itoh et al., 2010), and the latter preferentially suppresses the expression of Ehd1 but not Hd1, making the expression of Hd3a and RFT1 reduced under LD, but not under SD (Gao et al., 2014; Liu et al., 2018). Interestingly, similar to that of OsLUX, OsELF3a also acts as a flowering activator under both LD and SD conditions, and disruption of OsELF3a function causes similar expression alteration of these genes associated with photoperiodic flowering under various conditions (Saito et al., 2012; Zhao et al., 2012). Nevertheless, OsELF4a displays a regulatory mechanism similar to those of OsLUX and OsELF3a under SD condition, but a more complicated regulatory mechanism under LD condition (Wang et al., 2021). In addition, Arabidopsis ELF4-like proteins do not appear in EC, but influence flowering time (Lin et al., 2019). Therefore, it will be of interest to further investigate the roles of ELF4-like genes in rice heading.

Taken together, as an integrant, OsLUX, OsELF3a, and OsELF4a promote rice flowering through the same pathway. It is likely that they regulate the expression of genes associated with photoperiodic flowering cooperatively and independently, similar to soybean GmLUX2 and GmELF3a (Fang et al., 2021). Based on previous findings (Wei et al., 2020; Wang et al., 2021) and our results, we draw a simple outline model to explain how EC regulates photoperiodic flowering in rice (Figure 8). As a complex of transcriptional repressor, EC acts at the upstream of PRR37 and Ghd7 (genetically) and OsGI (physically), repressing the expression of these genes under both LD and SD (Figures 4, 5; Wang et al., 2021), so as to regulate the expression of Ehd1 and Hd1. Ehd1 and Hd1 determine the expression level of florigenic genes Hd3a and RFT1 through distinct pathways under a given photoperiod. Hd1 is crucial for SD photoperiodic induction of Hd3a (Izawa, 2007). Ehd1 promotes flowering independent of Hd1 under SD, but also promotes flowering under LD when Hd1 represses Hd3a expression, whereas Ghd7 negatively regulates Ehd1 expression (Woods et al., 2014).