A critical role of the soybean evening complex in the control of photoperiod sensitivity and adaptation

Significance

In many plant species, the timing of flowering is sensitive to photoperiod. In many crop species, genetic variation in this sensitivity is critical for adaptation to specific regions and management practices. This study identifies a component of the genetic pathway controlling flowering time in soybean, a legume crop of major global importance. Notably, plants lacking this component flower extremely late. Photoperiod sensitivity in plants, including soybean, was first systematically described in a seminal paper 100 y ago, and the results presented here establish an important new molecular step underlying this response. This step is a critical control point that could be genetically adjusted to engineer photoperiod sensitivity for yield improvement across a broad range of locations and agricultural contexts.

Abstract

Photoperiod sensitivity is a key factor in plant adaptation and crop production. In the short-day plant soybean, adaptation to low latitude environments is provided by mutations at the J locus, which confer extended flowering phase and thereby improve yield. The identity of J as an ortholog of Arabidopsis ELF3, a component of the circadian evening complex (EC), implies that orthologs of other EC components may have similar roles. Here we show that the two soybean homeologs of LUX ARRYTHMO interact with J to form a soybean EC. Characterization of mutants reveals that these genes are highly redundant in function but together are critical for flowering under short day, where the lux1 lux2 double mutant shows extremely late flowering and a massively extended flowering phase. This phenotype exceeds that of any soybean flowering mutant reported to date, and is strongly reminiscent of the “Maryland Mammoth” tobacco mutant that featured in the seminal 1920 study of plant photoperiodism by Garner and Allard [W. W. Garner, H. A. Allard, J. Agric. Res. 18, 553–606 (1920)]. We further demonstrate that the J–LUX complex suppresses transcription of the key flowering repressor E1 and its two homologs via LUX binding sites in their promoters. These results indicate that the EC–E1 interaction has a central role in soybean photoperiod sensitivity, a phenomenon also first described by Garner and Allard. EC and E1 family genes may therefore constitute key targets for customized breeding of soybean varieties with precise flowering time adaptation, either by introgression of natural variation or generation of new mutants by gene editing.

It is now widely appreciated that the timing of flowering, and the extent to which it is responsive to environmental cues, is one of the most important determinants of crop adaptation and yield (1). One hundred years ago, Wightman Garner and Harry Allard (2) made the first comprehensive report on plant photoperiodism in a seminal paper, which prominently featured soybean and tobacco as model plants. Over the following decades the physiological and molecular basis of this phenomenon has been investigated and characterized in detail, first in Arabidopsis, and increasingly in other species. Although our understanding of the molecular diversity in mechanisms of flowering-time regulation is still relatively limited, certain fundamental features are proving to be widespread, if not universal. In the case of photoperiodism, it appears that an interaction between light perception and endogenous circadian rhythms directs the photoperiod-specific expression of genes in the florigen family, many of which encode mobile signals that move from leaf and shoot apex to induce flowering.

The genetic network responsible for the generation of circadian rhythms (the circadian “clock”) is understood to consist of multiple interlocked transcriptional–translational feedback loops, and dozens of genes constituting these loops have been identified in the model plant Arabidopsis (3, 4). Within this network, the evening complex (EC)—comprising EARLY FLOWERING 3 (ELF3), EARLY FLOWERING 4 (ELF4), and LUX ARRHYTHMO (LUX)—is a transcriptional repressor complex and a core component of the plant circadian clock (5). LUX is a single MYB domain-containing SHAQYF-type GARP transcription factor, and appears to mediate the interaction of the EC with target promoters through direct binding of the MYB domain through a specific LUX binding site (LBS) motif GATWCG (where W indicates A or T) (6, 7). Recent studies indicate that a prion-like domain in ELF3 functions as a thermosensor, while ELF4 can stabilize the function of ELF3 (8, 9). EC directly regulates multiple clock output pathways, such as hypocotyl growth, flowering, defense, and leaf senescence (6, 10–13). The importance of the EC for flowering-time control and photoperiodism in particular is indicated by the fact that Arabidopsis mutants affecting any one of its three components have markedly impaired photoperiod responsiveness (14–16), and a growing list of flowering-time variants in crop species have also been linked to EC components (17–26).

Soybean [Glycine max (L). Merr.] is a major legume crop that produces protein and oil and provides more than a quarter of the world’s protein for food and animal feed (27). Cultivated soybean was domesticated from its wild relative (Glycine soja Sieb. & Zucc.) more than 5,000 y ago in temperate regions of China between 32° and 40°N (28–30). As first demonstrated by Garner and Allard (2), soybean is a short-day (SD) plant (SDP) and, whereas the wild ancestor and many primitive forms of domesticated soybean are strongly photoperiod-sensitive, modern soybean cultivars vary widely in their degree of sensitivity. This has enabled adaptation of the crop across a wide latitudinal range from 50°N to 35°S, and reflects natural variation in genes controlling flowering and maturity, with allelic combinations specifying optimal adaptation to narrow latitudinal zones (31). A growing number of loci contributing to this adaptation have been identified and characterized to the molecular level, including the E series (E1–E11), Tof11, Tof12, and J (31–37).

Among these, the legume-specific E1 gene has emerged as a major regulator of the photoperiod response and a critical point of integration within the soybean flowering pathway (33, 34, 38, 39). E1 encodes a B3 superfamily member, which is itself strongly regulated by photoperiod and controls flowering through its repressive effects on expression of two key FT genes, FT2a and FT5a (39). E2 has been identified as an ortholog of GIGANTEA (GI), a component of Arabidopsis circadian clock (40). E3 and E4 encode phytochrome A (PHYA) genes, PHYA3 and PHYA2, respectively (41, 42), while E9 and E10 have been identified as florigen genes FT2a and FT4, respectively (35, 43). Recently, the Tof11 and Tof12 loci were identified as orthologs of another Arabidopsis circadian clock component PRR3, and shown to have contributed to the development of early flowering and latitudinal adaptation during early evolution of domesticated soybean (33, 44).

Recent efforts have also focused on genes conferring adaptation to low latitudes. When commercial soybean varieties developed in temperate regions are grown at lower latitudes, they mature undesirably early and have extremely poor grain yield. This limitation was overcome in the 1970s with the introduction of the long-juvenile (LJ) trait, which extends the vegetative phase and improves yield under SD conditions, enabling productive cultivation in tropical regions (45–49). A major locus conferring this trait, J, is now known to be an ortholog of the Arabidopsis EC component ELF3 (34). Functional characterization of J has provided a working model for its role in SD flowering in which it acts to repress E1 transcription, thereby relieving the E1 suppression of FT2a and FT5a, and promoting flowering (34).

Although J influences flowering through transcriptional repression of E1 (34), there is no direct evidence for how this is achieved. However, the evidence from Arabidopsis implies that J may participate in a complex that is targeted to promoters of various target genes through the direct DNA-binding activity of LUX orthologs (6). In this study, we investigated the molecular role of J through an examination of the two soybean LUX homologs, LUX1 and LUX2. We demonstrate that both soybean LUX proteins physically interact with J and directly bind to the LBS in the E1 promoter to repress E1 expression. We also use targeted knockout of LUX genes to show that they play a redundant but critical role in flowering, maturity, adaptation, and yield and investigate their potential regulatory roles. Our results suggest that the soybean EC might function as an essential node connecting circadian clock components and light signaling pathways to multiple target genes for control of flowering time and other responses.

Results

Characterization and Expression Patterns of LUX1 and LUX2 in Soybean.

It is well documented that soybean has undergone two rounds of whole-genome duplication (WGD) after its divergence from the Arabidopsis lineage: One occurred recently at ∼13 Mya after its split from common bean (Phaseolus vulgaris) at ∼19 Mya, while the other ancient one occurred at ∼59 Mya (50–52). In the soybean reference genome, ∼75% of the predicted genes exist as duplicated gene pairs derived from the ∼13-Mya WGD event, while the remaining 25% have reverted to singletons (50, 53, 54). As expected, two homeologous LUX gene pairs deriving from the recent ∼13 Mya WGD were identified in the Williams 82 (W82) reference genome (Phytozome; https://phytozome.jgi.doe.gov/soybean), and they are designated as LUX1 (Glyma.12G060200) and LUX2 (Glyma.11G136600), as previously reported (34). Phylogenetic analysis of 40 LUX-like proteins from 24 species showed that LUX1 and LUX2 clustered together and fell into the same clade with other legume LUX proteins (SI Appendix, Fig. S1A). As a paleopolyploid species, there are two LUX homologs in soybean, while LUX is a single-copy gene in diploid legume species. Protein sequence alignment revealed that these LUX-like proteins were highly conserved in the DNA-binding MYB domain (SI Appendix, Fig. S2).

To analyze the expression characteristics of LUX1 and LUX2, regulation of their transcripts was investigated by qRT-PCR. As an SDP, soybean plants flower when day length is shorter than the maximum critical value, and the daylength regime for SD is generally 12-h light/12-h dark (2, 55). We examined the expression patterns of LUX under artificial SD (ASD) conditions (12-h light/12-h dark) in the growth chamber. The tissue-specific expression study showed that both LUX1 and LUX2 were expressed in all tissues examined (SI Appendix, Fig. S1B). The time-course expression study showed that the expression of both genes in leaves increased continuously from 10 d after emergence (DAE) to 25 DAE and peaked at 25 DAE, then decreased from 25 DAE to 35 DAE (SI Appendix, Fig. S1D). Under SD conditions, expression of both genes showed clear diurnal rhythms and peaked at dusk (SI Appendix, Fig. S1F), consistent with the expression pattern of LUX in other species (15, 19, 20). We also found that in the same samples, the expression of J, the soybean ortholog of Arabidopsis EC gene ELF3 (34), showed similar tissue-specific, time-course, and diurnal patterns to those of the two LUX genes (SI Appendix, Fig. S1 C, E, and G). Therefore, we considered it likely that LUX1 and LUX2 might interact physically with J to form EC in soybean, a complex similar to that in Arabidopsis (6).

Protein Interactions of LUX1, LUX2, and J.

To test whether either of the two LUX proteins might interact physically with J, we first performed yeast two-hybrid assays (Y2H), which showed that both LUX1 and LUX2 indeed interact with J (Fig. 1A). These interactions were confirmed using in vitro pull-down assays (Fig. 1C), and extended by bimolecular fluorescence complementation (BiFC) assays, which additionally indicated that they occur in vivo and take place in the nucleus (Fig. 1B). It has been reported that most transcription factors bind to target DNA sequences as dimers (56). We examined whether LUX1 and LUX2 have this property using BiFC and Y2H assays. The results showed that both LUX1 and LUX2 could self-interact, and also interact with each other (Fig. 1 D and E). Our results suggested that LUX1 and LUX2 might form a heterodimer and homodimer to directly regulate targeted genes. Taken together, these results indicated that three proteins LUX1, LUX2, and J interact with each other in the nucleus to form the EC in soybean to control development, flowering, maturity, adaptation, and yield.

Fig. 1.

Protein interactions of soybean EC (SEC). (A) J interacts with LUX1 and LUX2 in yeast. Yeast cells transformed with indicated genes were selected on DDO (lacking Leu and Trp) and QDO (lacking Ade, His, Leu, and Trp) media. (B) J interacts with LUX1 and LUX2 in Nicotiana benthamiana leaves in a BiFC assay. LUX1 and LUX2 were fused to the N terminus of YFP and J was fused to the C terminus of YFP. The constructs were coinjected into N. benthamiana leaves, and YFP signals were observed after 48 to 72 h. (Scale bars, 20 μm.) Three biological replicates were performed. (C) LUX1 and LUX2 can pull down J. MBP, MBP-LUX1, and MBP-LUX2 proteins were expressed in Escherichia coli, and J-His protein was expressed using an in vitro translation system. Purified proteins were used for the pull-down assay. MBP, MBP-LUX1, and MBP-LUX2 were detected with anti-MBP antibody, and J-His protein was detected with anti-His antibody. (D) LUX1 and LUX2 interact with each other and themselves in yeast. Yeast cells transformed with indicated genes were selected on DDO and QDO media. (E) LUX1 and LUX2 interact with each other and themselves in N. benthamiana leaves in a BiFC assay. LUX1 and LUX2 were fused to the N and C terminus of YFP. The constructs were coinjected into N. benthamiana leaves, and YFP signals were observed after 48 to 72 h. (Scale bars, 20 μm.) Three biological replicates were performed.

LUX1 and LUX2 Are Direct Transcriptional Repressors of E1.

We previously showed that J is a transcriptional repressor of the key soybean flowering repressor E1, and binds directly to the E1 promoter (34). To investigate whether LUX1 and LUX2 might be similarly involved in the transcriptional repression of E1, we examined the molecular nature of the relationship between LUX proteins and E1 using an Arabidopsis protoplast transient expression assay. When a p35S:LUX1 or p35S:LUX2 construct was cotransformed with a pE1:LUC construct (Fig. 2A), relative LUC activity was significantly suppressed to a similar extent as a p35S:J construct (Fig. 2B), indicating that LUX1 and LUX2 protein might bind to the E1 promoter to suppress its activity. When a p35S:J construct and a p35S:LUX construct were cotransformed with a pE1:LUC construct, relative LUC activity was much lower (Fig. 2B), suggesting that J and LUX proteins could enhance the suppressive activities of each other in the J–LUX complex. Two-way ANOVA revealed that relative LUC activity is suppressed by J (P = 3.8 × 10⁻¹⁷), LUX (P = 1.2 × 10⁻¹⁹), and J × LUX (P = 3.9 × 10⁻⁹), indicating that the enhancement is synergistic. However, when both constructs of LUX1 and LUX2 were put together, the relative LUC activity were not significantly lower than either of the single constructs, suggesting that the functions of these genes have remained largely equivalent following their duplication during the polyploidization. To confirm that LUX proteins could bind to the E1 promoter in a soybean in vivo system, we performed chromatin immunoprecipitation (ChIP)-qPCR assays in p35S:LUX1-FLAG and p35S:LUX2-FLAG transgenic plants and W82 plants (Fig. 2 C and D). Using a similar assay, we previously demonstrated the physical association of J with three regions of the E1 promoter containing LBS (34). Our result showed that LUX1 and LUX2 were also associated with strong enrichment of E1 promoter sequences around three of these same LBS, at −264 bp, −799 bp, and −975 bp upstream of the ATG (Fig. 2 C and D). Finally, we performed EMSA to determine whether LUX1 and LUX2 proteins directly bind to E1 promoters in vitro. We found that LUX1 and LUX2 recombinant proteins could both directly bind to the LBS sites in the E1 promoters (Fig. 2 C, E, and F). These results indicate that LUX1 and LUX2 proteins physically interact with J and likely allow the J–LUX complex to directly bind to the E1 promoter and repress its activity.

Fig. 2.

LUX1 and LUX2 directly associate with the promoter of E1 to suppress its transcriptions. (A) Constructs of LUX1, LUX2, J, and E1 used for the transient expression assay in Arabidopsis protoplast. LUC, luciferase; REN, Renilla luciferase. (B) LUX1, LUX2, and J proteins suppress transcription from the E1 promoter in Arabidopsis protoplast. Values are shown as mean ± SD from three biological replicates. Different letters indicate significant differences by one-way ANOVA followed by Tukey’s post hoc test with SPSS statistics software. False-discovery rate (FDR)-adjusted P < 0.05. Two-way ANOVA revealed that the relative LUC activity is suppressed by J (P = 3.8 × 10⁻¹⁷), LUX (P = 1.2 × 10⁻¹⁹), and J × LUX (P = 3.9 × 10⁻⁹). (C) Schematic of the E1 gene and regions tested for enrichment in the ChIP assay and binding in the EMSA assay. (D) ChIP of E1 amplicons using W82, p35S:LUX1-FLAG, and p35S:LUX2-FLAG. Values are shown as mean ± SD from three biological replicates. Different letters indicate significant difference among the samples using the same primer by one-way ANOVA followed by Tukey’s post hoc test with SPSS statistics software. FDR-adjusted P < 0.05. Capital letters compare with each other, and lowercase letters compare with each other. (E and F) EMSA detected binding of GST-LUX1 (E) and GST-LUX2 (F) protein to the LBS of the E1 promoter.

LUX1 and LUX2 Are the Critical Regulators in Soybean Flowering under SD Conditions.

To obtain direct genetic evidence for the function of LUX1 and LUX2 in soybean flowering time control, we used a CRISPR/Cas9-mediated genome-editing approach to generate the soybean knockout mutants for both genes. We selected three genomic sites for simultaneous targeting of both LUX1 and LUX2 coding sequences (SI Appendix, Fig. S3), aiming to obtain lux1 and lux2 single mutants and lux1 lux2 double mutants. Appropriate single-guide RNA (sgRNA)/Cas9 vectors were constructed and transformed into the soybean cultivar W82, ultimately generating one independent lux1 mutant line, two lux2 mutant lines, and two lux1 lux2 double-mutant lines (Table 1). In the lux1 lux2-1 mutant, there was a deletion of a single G in both LUX1 and LUX2 coding regions, which caused a frame-shift mutation after the M16 codon in both genes. In lux1 lux2-2, we found the same LUX1 mutation (lux1) but a deletion of 22 bases in the LUX2 coding region, which introduced a frameshift after codon R146. All three mutations specified a premature stop codon (SI Appendix, Fig. S3).

Table 1.

Homozygous mutants of LUX1 and LUX2

	LUX1		LUX2
	Target 1	Target 2	Target 1	Target 2
lux1	−1 bp	—	—	—
lux2-1	—	—	−1 bp	—
lux2-2	—	—	—	−22 bp
lux1 lux2-1	−1 bp	—	−1 bp	—
lux1 lux2-2	−1 bp	—	—	−22 bp

Under flowering-inductive ASD conditions (12-h light/12-h dark), the LUX1 and LUX2 overexpression transgenic plants, p35S:LUX1-FLAG and p35S:LUX2-FLAG showed no significant effect on the flowering time (SI Appendix, Fig. S4 A and B), which may suggest the duplicated homeologous pairs already have been over the functional threshold dosages (57). However, the double lux1 lux2-1 mutant showed an extreme delay, not flowering until nearly 100 DAE compared to the wild-type plants flowering at 23 DAE under natural SD (NSD, 13-h light/11-h dark) conditions in Guangzhou (23° 16′ N, 113° 23′ E), China. In contrast, the single lux1 and lux2-1 mutations showed no significant effect on flowering time (Figs. 3 and 4A). These results show that LUX1 and LUX2 are functionally redundant but together play critical roles in regulation of soybean flowering. Mutants carrying a second loss of function allele of LUX2, lux2-2, and lux1 lux2-2, showed flowering phenotypes consistent with those associated with lux2-1 and lux1 lux2-1. (Fig. 4A and SI Appendix, Fig. S5). Under these conditions, wild-type plants produce around 10 reproductive nodes on the main stem before undergoing proliferative arrest of the primary shoot meristem and entering monocarpic senescence at ∼15 wk of age.

Fig. 3.

Phenotypes of lux1 and lux2 mutants. Phenotypes of wild-type plants (WT, W82) and homozygous mutants at 25 DAE (A), 95 DAE (B), 120 DAE (C), 155 DAE (D) under NSD (13-h light/11-h dark) conditions. Red box, magnified view. (Scale bars, 20 cm). The lux1 lux2-1 mutant, also called as Guangzhou Mammoth, continuously grows and keeps flowering, as shown in C and D. In C, Guangzhou Mammoth is 210-cm high and in D, it grows up to 250-cm high without stopping growing.

Fig. 4.

LUX1 and LUX2 redundantly regulate transcript abundance of the soybean core flowering genes E1 and FT. (A) Flowering time of W82 and homozygous mutants under NSD conditions (13-h light/11-h dark). Different letters indicate significant differences by Kruskal–Wallis one-way ANOVA followed by multiple-comparison test with SPSS statistics software. FDR-adjusted P < 0.05. The flowering time is shown as the mean values ± SD, n > 10 plants. (B–F) Diurnal expression of E1 (B), FT2a (C), FT5a (D), E1La (E), and E1Lb (F) in W82, lux1, lux2-1, and lux1 lux2-1 plants at 15 DAE under ASD (12-h light/12-h dark). Data shown relative to the control gene Tubulin and represent means ± SD for three biological replicates. The dashed line indicates nonlinear regression curve. Nonlinear regression analysis was performed by GraphPad Prism 8.

In contrast, the lux1 lux2-1 mutant, in addition to first initiating flowering 11 wk later than wild-type, also showed a dramatically extended reproductive period, with the primary shoot apex continuing to grow and produce axillary flowers for a further 22 wk without showing the stopping apical growth (Fig. 3 C and D). This massive extension of the growth period was accompanied by a striking thickening of stems, giving plants the appearance of a small tree (Fig. 3 C and D). This dramatic effect on growth habit is reminiscent of the famous photoperiod-sensitive tobacco mutant Maryland Mammoth that was prominent in the seminal study of plant photoperiodism by Garner and Allard (2) and nicely illustrated by Amasino (58). We therefore named this lux1 lux2-1 double mutant as “Guangzhou Mammoth” to pay homage to Maryland Mammoth. In addition, we also evaluated the flowering time of Guangzhou Mammoth under both ASD (12-h light/12-h dark) and artificial long day (ALD, 16-h light/8-h dark) conditions in a growth chamber. Under ASD, the single lux1 or lux2-1 mutants showed no significant difference with the wild-type plants, which flowered at 24 DAE, while the Guangzhou Mammoth flowered at about 77 DAE (SI Appendix, Fig. S6). Intriguingly, under ALD conditions, the Guangzhou Mammoth also flowered at about 78 DAE compared to the wild-type plants that flowered at 51 DAE, while the single lux1 and lux2-1 mutations showed no significant difference with the wild-type plants (SI Appendix, Fig. S6), indicating that EC also plays important roles in regulating soybean flowering under LD conditions and complete impairment of EC abolishes soybean photoperiod sensitivity. Collectively, these results suggest that EC plays critical roles in soybean photoperiod response.

LUX1 and LUX2 Act Upstream of the Legume-Specific Flowering Repressors E1 and E1 Homologs.

To further understand the functional mechanisms underlying EC, we investigated E1 expression in lux1, lux2-1, lux1 lux2-1, and wild-type W82. Under SD conditions, the expression of E1 in W82 was very low, as reported previously (39), but showed a massive derepression in lux1 lux2-1 double-mutant plants and a clear diurnal rhythm with a peak at dusk (Fig. 4B). However, no obvious induction of E1 was observed in lux1 and lux2-1 single mutants (Fig. 4B), consistent with their flowering-time phenotype (Figs. 3 and 4A), further supporting the functional redundancy of LUX1 and LUX2, which was well documented with the notion that duplicated genes were retained without functional changes to maintain their dosage balance (57, 59, 60). To some extent, the functional redundancy could be explained by which the heterodimers and homodimers of LUX1 and LUX2 formed (Fig. 1 D and E) might play equivalent roles in soybean EC in regulating flowering, which has been reported in several cases (56). We also examined E1 expression in p35S:LUX1-FLAG and p35S:LUX2-FLAG plants; the result showed that E1 expression was not influenced by overexpression of LUX1 and LUX2, which was consistent with the lack of effect on flowering phenotypes (SI Appendix, Fig. S4 B and C). Previous studies have shown that the central role of E1 in the photoperiod regulation of soybean flowering largely reflected its repression of two key FT homologs, FT2a and FT5a (34, 39, 61). We further examined the expression of FT2a and FT5a and found that the transcriptions of both FT genes were nearly not detected in lux1 lux2-1 double-mutant plants (Fig. 4 C and D), which is consistent with the higher expression of E1 (Fig. 4B) and extremely late flowering phenotypes (Fig. 3). These results show that the two LUX genes act in a functionally redundant manner to fully suppress E1 expression, consequently relieving the repression of FT2a and FT5a thereby promoting flowering and maturity and reducing overall yield potential under SD conditions.

The soybean genome has two E1 homologs, E1-like-a (E1La) and E1Lb, which were suggested to function similarly to, but to some extent independently from E1 in the control of flowering and adaptation (62, 63). We therefore examined the diurnal expression of E1La and E1Lb in W82, lux1, lux2-1, and lux1 lux2-1 under SD conditions. Both genes showed patterns of expression very similar to E1, with no significant expression in W82 and both single mutants, but a strong derepression with a peak at ZT12 in lux1 lux2-1 (Fig. 4 E and F). As in the case of E1, the promoters of E1La and E1Lb also contained several LBS sites implying that the EC could also directly bind to these sites to suppress E1La and E1Lb transcription (SI Appendix, Fig. S7A). To examine whether LUX proteins could bind to the E1La and E1Lb promoter in soybean, we performed ChIP-qPCR assays in p35S:LUX1-FLAG and p35S:LUX2-FLAG transgenic plants and W82 plants. Our result showed that LUX1 and LUX2 were also associated with strong enrichment of the E1La and E1Lb promoter (SI Appendix, Fig. S7 B and C). The direct repressive roles of LUX1 and LUX2 on E1La and E1Lb were also tested in Arabidopsis protoplast transient expression assay, in which LUX1 and LUX2 were capable of repressing the transcription from the E1La and E1Lb promoter (SI Appendix, Fig. S7 D–F). These results indicate that E1La and E1Lb are also negatively regulated by LUX1 and LUX2, and provide further evidence that E1La and E1Lb play a function similar to E1 in photoperiodic induction of flowering in soybean. Overall, these results indicate that full loss of LUX function (presumably causing complete evening complex impairment) releases expression of three of these E1 family genes, which are then able to repress FT expression and thereby generate extremely late-flowering phenotypes under SD conditions.

Taken together, these results propose a model that LUX1 and LUX2 are functionally redundant in modulating photoperiod-regulated flowering in soybean under SD conditions, and both of them suppress expressions of E1 and its homologs by binding to the LBS in their promoters, which relieves the E1-dependent transcriptional repression of FT2a and FT5a, thereby promoting flowering and modifying the adaptation and grain yield development (Fig. 5). In other words, in wild-type soybeans, the EC (J interacts with heterodimers of LUX1-LUX2) has the strongest suppressive effects on soybean flowering suppressors, and thus promotes early flowering and low yield productivity. When a single mutation occurs in lux1 or lux2, J interacts with either homodimers of LUX1-LUX1 or LUX2-LUX2 to maintain the same suppressive activity as J-LUX1-LUX2 of EC without phenotypic flowering changes as wild-types. However, the mutation of J reduced the suppressive activity of EC on the functions of E1 homologs, and thus resulted in late flowering and high yield. Strikingly, the double mutant of lux1 lux2 completely impaired the functions of EC, and thus fully released the functions of three E1 suppressors and resulted in extreme late-flowering phenotypes (Fig. 5).

Fig. 5.

Model summarizing the mechanism of SEC functions under SD conditions. J protein physically associates with LUX1 and LUX2 proteins in which LUX1 interacts with LUX2 to form SEC J-LUX1-LUX2 and directly bind to the promoters of E1 and its two homologs E1La and E1Lb to suppress their expressions, thus mediating the transcriptional suppression of FTs to control flowering and adaptation and grain-yield productivity. (A) In wild-type soybeans, the SEC (J interacts with heterodimers of LUX1-LUX2) has the strongest suppressive effects on soybean flowering suppressors thus promotes early flowering and low yield productivity. (B) In single mutant of either of lux1 or lux2, J interacts with either homodimers of LUX1-LUX1 or LUX2-LUX2 to maintain the same suppressive activity as J-LUX1-LUX2 of SEC without phenotypic flowering changes. (C) The mutation of J reduced the activity of SEC thus resulted in late flowering and high yield. (D) Double mutant of lux1 lux2 completely impairs the functions of SEC and thus fully releases the functions of three E1 suppressors resulting in extreme late-flowering phenotypes.

Various Flowering-Associated Genes under Regulation by Soybean EC.

To gain further insight into the nature of the genes regulated by the soybean EC, we performed transcriptome sequencing (RNA-seq) on lux1, lux2-1, lux1 lux2-1, and W82 plants. Compared with W82, there were 2,297, 1,018, and 3,316 differentially expressed genes (DEGs) in leaves in lux1, lux2-1, and lux1 lux2-1 mutant plants (Dataset S1). Interestingly, at least 43 DEGs in lux1 lux2-1 showed homology with known flowering time-associated genes from Arabidopsis by using phytozome and Uniprot databases (SI Appendix, Fig. S8A and Dataset S2). Among them, several DEG regulatory pathways involved in photoperiod, the gibberellic acid (GA) pathway, and circandian clocks were misregulated by LUX1 and LUX2. Consistent with our qRT-PCR analysis (Fig. 4 B–F), E1, E1La, and E1Lb were strongly up-regulated, and FT2a and FT5a were strongly repressed. Three other FT homologs were also significantly misregulated in lux1 lux2-1. Like FT2a, FT2b (Glyma.16G151000) was also significantly down-regulated in lux1 lux2-1, implying that that FT2b may also function as a flowering promoter. In contrast, FT4 (Glyma.08G363100) and to a lesser extent FT1a (Glyma.18G298900) were substantially up-regulated in lux1 lux2-1, consistent with their previously described inhibitory roles in flowering and transcriptional activation by E1 (64, 65). These findings provide further evidence that soybean FT homologs have diverged both in function and regulation. The divergence of function among members of the FT/TERMINAL FLOWER 1 (TFL1) gene family is well established in Arabidopsis. FT and TFL1 encode a pair of flowering regulators with homology to phosphatidylethanolamine-binding proteins; they share ∼60% amino acid sequence identity but function in an opposite manner (66, 67). FT promotes the transition to flowering whereas TFL1 represses this transition (68, 69).

As expected, the circadian clock genes, three orthologs of Arabidopsis NIGHT LIGHT–INDUCIBLE AND CLOCK-REGULATED 3 (LNK3), and four homologs of CYCLING DOF (CDF) were significantly induced in the double mutant. In Arabidopsis, lux mutants show arrhythmic (15); therefore, we examined whether it is true in soybean. The free-running period of leaf movement rhythmicity in W82, lux1, lux2-1, and lux1 lux2-1 plants under constant light (LL) was investigated. Unlike the arrhythmicity of lux mutants in Arabidopsis, we found that the period length in leaf movement of lux1 and lux2-1 single mutant was ∼1.5 and 1 h longer than that in the wild-type, while the lux1 lux2-1 double mutant displayed a longer period than either of the single mutants, having a period ∼6 h longer than that in wild-type plants (SI Appendix, Fig. S9 and Table S1). This indicates that LUX1 and LUX2 function redundantly to maintain circadian rhythms in soybean, and a simultaneous mutation of them could not completely abolish the rhythmicity, which is inconsistent with Arabidopsis. In addition, the homologs of GA pathway genes—including GA2oxidase 8, GIBBERELLIC ACID-STIMULATED ARABIDOPSIS 4 (GASA4), and GASA6—were up-regulated in the double-mutant plants, which may partially explain the continuous growth of the mutant of Guangzhou Mammoth. We next performed qRT-PCR experiments to verify the expression of several key genes that may participate in flowering identified in our RNA-seq analysis; the results indicated that the expression of these genes was consistent with the RNA-seq results (SI Appendix, Fig. S8B). All of these results suggest that multiple gene pathways might be regulated by the EC in soybean to control flowering, maturity, and yield development.

Discussion

Precise flowering time is critical to crop adaptation and productivity in a given environment. Although the importance of flowering time in the adaptation and yield of soybean is well established and several major genes have been characterized (32, 34), the molecular understanding of flowering is still insufficient. Our results indicate that the regulatory modules of soybean EC–E1/E1 homologs are the key molecular networks controlling soybean flowering and photoperiodism. The complete impairment of EC, which was reflected by the mutant of Guangzhou Mammoth, fully derepresses the transcriptions of three core soybean flowering repressors E1, E1La, and E1Lb, thus highly repressing FT homologs to extremely delay flowering even under inductive SD conditions. The Guangzhou Mammoth showed extreme late flowering under NSD (13-h light/11-h dark) conditions in Guangzhou (23° 16′ N, 113° 23′ E), which flowered 11 wk later than wild-type (Fig. 4A). However, under ASD (12-h light/12-h dark, 25 °C) conditions in the growth chamber, the Guangzhou Mammoth flowered about 53 d later than wild-type (SI Appendix, Fig. S6). The flowering time has long been known to be regulated primarily by photoperiod and temperature (70, 71). The extremely late-flowering phenotypes of Guangzhou Mammoth under NSD conditions might be caused by the interactive photo-thermal effects on soybean flowering. Moreover, the flowering time of Guangzhou Mammoth under ALD and ASD conditions showed no significant difference (SI Appendix, Fig. S6), indicating that loss of function of EC leads to photoperiod-insensitivity of Guangzhou Mammoth.

These results indicate that soybean EC plays an essential role in controlling flowering and photoperiod sensitivity, consequently determining the adaptation and yield development. The extreme late flowering of Guangzhou Mammoth further suggests that the EC may integrate the photoreceptors and circadian clocks to control E1 expressions to mediate photoperiod flowering in soybean. Genetic dissections of the relationship between these components with EC will be helpful to understand the molecular mechanisms of EC in soybean flowering. Therefore, this EC–E1 regulatory module provides us with a promising perspective to generate various allelic combinations of J, LUX1, LUX2, E1, E1La, and E1Lb by a CRISPR/Cas9 genome-editing approach in the elite cultivars to quickly develop new cultivars with a series of continuums of flowering time and adaptability. These cultivars will be extremely important for the adaptation and yield improvement in tropical countries to maximize the soybean productivity.

The roles of EC are well conserved in maintaining circadian rhythms and regulating flowering time in different plant species (14–26). The mutation of LUX leads to early-flowering phenotypes in Arabidopsis and pea (15, 26). In addition, disruption of LUX homologs has also been proposed as the molecular basis for mutations conferring photoperiod-insensitive early flowering in barley (Hordeum vulgare) early maturity10 (eam10) mutant (19) and in einkorn wheat (Triticum monococcum) earliness per se 3 (Eps-3A^m) mutant (20). In contrast to these LD plant (LDP) species, here we found that disruption of LUX1 and LUX2 delays flowering in soybean, a typical SDP species. This situation is similar with another EC component ELF3, which inhibits flowering in LDP species barley, wheat, pea, and lentil (18, 21–23). In contrast, in the SDP species rice and soybean, ELF3 promotes flowering by suppressing expression of the key FT repressors Grain number, plant height and heading date 7 (Ghd7) and E1, respectively (24, 25, 34, 72). This further supports the emerging view that upstream components of the photoperiod response pathway play opposite roles in SDP and LDP (18, 21–25). The soybean J gene is the ortholog of Arabidopsis ELF3, and J promotes flowering through repressing the transcriptional expression of the legume-specific flowering repressor E1 under SD, which makes J a major source of adaptation and yield improvement in low-latitude regions (34). In LD species Arabidopsis and pea, mutants for the three EC components ELF3, ELF4, and LUX have similar early-flowering phenotypes (14–18, 26), indicating that functions of EC genes are conserved in different species. Both ELF3 and ELF4 are single-copy genes in diploid legumes, while there are three homologs of ELF3 (J, ELF3b-1, and ELF3b-2) and two homologs of ELF4 (ELF4a and ELF4b) in soybean (17, 18, 34). Like LUX1 and LUX2, the homologs of ELF3 and ELF4 might also play redundant roles in regulating flowering according to the dosage balance hypothesis. Further genetic and molecular characterization of EC genes, such as duplicated homologs of ELF4 and ELF3, are needed to extend our understanding of the mechanisms of flowering and adaptation in soybean. Exploration of the roles of EC in other crop species will be of great important for the improvement of crop adaptation and grain yield.

In summary, we conclude that the two soybean LUX homologs, LUX1 and LUX2, like J, play essential roles in regulating flowering and adaptation. Both LUX1 and LUX2 interact with J to form EC, and they are functionally redundant in promoting flowering. The EC represses E1s expression by binding to the LBS of their promoters, further relieving the expression of FT2a and FT5a and promoting flowering under SD conditions. Our findings provide evidence of the critical functions of EC in flowering-time control and latitudinal adaptation and yield development that will be greatly helpful in soybean molecular breeding.

Materials and Methods

The soybean [G. max (L.) Merr.] cultivar W82 was used as the wild-type. Methodological details of plant growth, gene-expression analysis, nonlinear regression analysis, phylogenetic analysis, protein–protein interaction assays, transient dual-luciferase assay, Western blot, protein–DNA interaction assays, transcriptome analysis, and leaf movement experiments are described in SI Appendix, Materials and Methods. The primers used in this study are listed in Dataset S3.

Data Availability

RNA-seq data have been deposited in the Sequence Read Archive (NCBI-SRA) (BioProject no. PRJNA628851). All other study data are included in the article and supporting information.