Molecular Genetic Understanding of Photoperiodic Regulation of Flowering Time in Arabidopsis and Soybean

Abstract

The developmental switch from a vegetative phase to reproduction (flowering) is essential for reproduction success in flowering plants, and the timing of the floral transition is regulated by various environmental factors, among which seasonal day-length changes play a critical role to induce flowering at a season favorable for seed production. The photoperiod pathways are well known to regulate flowering time in diverse plants. Here, we summarize recent progresses on molecular mechanisms underlying the photoperiod control of flowering in the long-day plant Arabidopsis as well as the short-day plant soybean; furthermore, the conservation and diversification of photoperiodic regulation of flowering in these two species are discussed.

In plant life cycles, the time of the trantion to flowering is critical to reproductive success and is regulated by external environmental factors (such as light, temperature) and endogenous signals such as age and developmental state. Seasonal changes in day length (photoperiod) play an important role in the regulation of flowering in various plants, through the photoperiod pathway [1]. Photoperiodic flowering responses are classified into three major types: long-day (LD), short-day (SD), and day-neutral (DN), based on their flowering responses to day-length changes [2,3]. Here, we review recent molecular genetic understandings of how seasonal day-length changes, through the photoperiod pathways, promote flowering in the LD plant Arabidopsis thaliana and the SD plant soybean.

1. Molecular Mechanisms for the Photoperiodic Regulation of Flowering in Arabidopsis

Arabidopsis thaliana is a typical LD plant, which flowers when the day length becomes longer, that is, longer-day lengths promote flowering, while shorter-daylengths inhibit flowering in Arabidopsis.

1.1. The Photoperiod Pathway in Arabidopsis

The photoperiod pathway begins with the perception of light signals by photoreceptors in leaves. Light signals of different wavelengths are received and recognized by the photoreceptors in leaves. Currently, there are three main types of photoreceptors found in plants: phytochromes, cryptochromes and phototropins. Phytochromes mainly absorb red light and far-red light. Arabidopsis has at least five types of phytochromes: PHYA, PHYB, PHYC, PHYD and PHYE. PHYA promotes flowering, whereas PHYB, PHYD and PHYE inhibit flowering. PHYC inhibits flowering under SD but promotes flowering with PHYA under LD [4,5,6]. Cryptochromes mainly absorb blue and ultraviolet light, and Arabidopsis contains CRY1 (CRYPTOCHROME 1), CRY2 and CRY3. Both CRY1 and CRY2 promote flowering [7]; it is not clear whether CRY3 is involved in the regulation of flowering. So far, it has not been found that phototropins are involved in flowering regulation.

After sensing the photoperiod, photoreceptors transmit the signal to the circadian clock. The clock components such as ELF3 (EARLY FLOWERING 3) [8,9], ELF4 [10,11], PIF3 (PHYTOCHROME INTERACTING FACTOR 3) [12], ZTL (ZEITLUPE) [13], FKF1 (FLAVIN-BINDING KELCH REPEAT F-BOX 1) [13] and DET1 (DEETIOLATED 1) [14,15], have been found to mediate the optical signal input of different photoreceptors. ELF3 encodes a nuclear protein that plays a role in transcriptional regulation, both its mRNA and protein exhibit circadian rhythm [16]. The elf4 mutant also exhibits disruption of biological circadian with an early flowering phenotype under non-induced photoperiod conditions [10,11], and ELF4 may also regulate the input of light signals [11]. ZTL, FKF1 and LKP2 (LOV KELCH PROTEIN2) regulate biological rhythms and flowering time, and mutations in these genes lead to disruptions of circadian rhythms [17,18,19].

Arabidopsis thaliana has a negative feedback regulating loop for maintaining and resetting the circadian clock, which is mainly composed of CCA1 (CIRCADIAN CLOCK ASSOCIATED 1), LHY (LATE ELONGATED HYPOCOTYL) and APRRs (ARABIDOPSIS PSEUDO-RESPONSE REGULATOR 1) [17,20]. LHY and CCA1 belong to the MYB transcription factor family. APRRs include APRR9, APRR7, APRR5, APRR3, APRR1/TOC1 (TIMING OF CAB OF EXPRESSION 1) and other homologous proteins; the expression levels and protein levels of these genes all exhibit circadian rhythm [21]. LHY and CCA1 are activated by light in the morning, and the newly synthesized LHY and CCA1 bind to the TOC1 promoter and inhibit its expression [22,23,24]. At the same time, blue light promotes the interaction between ZTL and GI (GIGANTEA) and degrades TOC1 through the ubiquitin-protease system [25]. Therefore, at the beginning of the day, LHY and CCA1 gradually increase, and TOC1 gradually decreases [24]. Interestingly, TOC1 is a necessary promoter for the expression of LHY and CCA1; therefore, inhibition of TOC1 will result in down-regulation of the expression of LHY and CCA1 [24]. The levels of LHY and CCA1 drop to a minimum at night, and thus the inhibition of TOC1 expression is released, resulting in that the expression of LHY and CCA1 is started again, entering a new cycle [24].

The important genes in output of the circadian clock signal include CO (CONSTANS), GI and FT (FLOWERING LOCUS T). CO encodes a transcription factor with a B-box zinc finger structure, and its expression is regulated by the circadian clock with a 24-hour periodic oscillation [26]. The transcription level of CO is regulated by FKF1, GI, CDFs (CYCLING DOF FACTOR), and FBH (FLOWERING BHLH). CDFs are plant-specific transcription factors and bind to the regulatory region of CO to inhibit its transcription [27]. Under LD, in the morning, CDF expression is activated by LHY and CCA1, and when GI accumulates to a certain amount, a CDF1-GI complex is formed to inhibit CO transcription [28,29]. After about 13 hours of light, the protein levels of FKF1 and GI reach peaks and these proteins form a dimeric FKF1-GI E3 ubiquitin ligase complex. FKF1-GI targets CDF proteins for degradation by the proteasome, thereby releasing CDF-mediated CO inhibition [28,30]. Subsequently, four basic helix-loop-helix (bHLH) transcription factors (TFs) FBH1-FBH4, bind to the E-box cis-element to promote CO expression [31]. However, under SD, the protein levels of GI and FKF1 reached their peaks at 7 and 10 hours of light, respectively, and are unable to form the FKF1-GI complex, and thus CO transcription stays at a low level [28,30]. Therefore, the interaction of FKF1, GI and CDFs leads to higher CO transcription levels under LD than in SD.

Posttranslational regulation is very important for CO protein stability. CO protein is degraded in the morning and stabilized near dusk [32]. The red light photoreceptor PHYA and far-red light photoreceptor PHYB antagonistically regulate the stability of CO protein. In the morning, PHYB absorbs red light and interact with the RING finger-containing E3 ubiquitin ligase HOS1 (HIGH EXPRESSION OF OSMOTICALLY RESPONSIVE GENE 1) to promote degradation of the CO protein [32,33]. At night, members of Arabidopsis SPA (SUPPRESSOR OF PHYA-105) family interact with another RING-finger E3 ubiquitin ligase COP1 (CONSTITUTIVE PHOTOMORPHOGENIC 1), and this complex binds to CO through a WD-repeat domain, and promotes CO degradation [34,35]. In late afternoon in LDs, red light photoreceptor Phytochrome A (PHYA) and blue light photoreceptors CRY1 and CRY2, all function to stabilize the CO protein [36]. Blue light induces CRY2 phosphorylation, and the phosphorylated CRY2 enhances the CRY2-SPAs interaction to inhibit the function of COP1-SPAs in CO degradation [37]. The blue light also enhances FKF1 stability, which accumulates from the afternoon to dusk and interacts with CO through its LOV domain to stabilize the CO protein [38,39]. PHL (PHYTOCHROME-DEPENDENT LATE FLOWERING), whose expression level is increased in the afternoon, interacts with PHYB and suppresses PHYB-mediated degradation of the CO protein [40]. In addition, DNF (DAY NEUTRAL FLOWERING), a membrane-bound E3 ligase, regulates CO in a way independent of GI/FKF1/CDF. The dnf mutant exhibits an early-flowering phenotype due to disruption of the circadian rhythm of CO expression [41]. The results of CO regulation by the above factors are: under LD, the CO protein accumulates at dusk; under SD, CO cannot be stably produced and accumulated.

1.2. Regulation of FT Expression

The photoperiod pathway output CO promotes FT expression [42]. The FT protein belongs to the PEBP (phosphatidylethanolamine binding) family and is a major florigen [43,44]. FT moves to SAM (shoot apical meristem) from leaves through the phloem and subsequently interacts with 14-3-3 proteins and the bZIP transcription factor FD (FLOWERING D) to form a transcription-activation complex that promotes the expression of floral meristem-identity genes in SAM [45]. Therefore, FT functions as the key floral promoter.

FT expression is regulated by various transcriptional activators and repressors. Several types of transcription inhibitors are involved in the regulation of FT expression. From early morning to late afternoon and at night, FT chromatin is marked predominantly with repressive histone 3 lysine-27 trimethylation (H3K27me3), so FT expression is mainly repressed by Polycomb group (PcG) genes. The H3K27me3 reader LHP1 (LIKE HETEROCHROMATIN PROTEIN 1) and histone H3 lysine-4 demethylase JMJ14 interact with the plant-unique EMF1 (EMBRYONIC FLOWER1) to form a distinct Polycomb repressive complex 1 (PRC1)-like complex called LHP1-EMF1c that represses FT expression [46]. In addition, two BAH domain H3K27me3 readers EBS and SHL interact with EMF1 to form the BAH-EMF1c complexes [47,48]. These PRC1-like EMF1c complexes (LHP1-EMF1c and BAH-EMF1c) bind to FT chromatin, read and further maintain the H3K27me3 repression marks at FT, which are catalyzed by the H3K27 methyltransferase complex, CLF-PRC2 (CURLY LEAF-Polycomb repressive complex 2) (Figure 1).

The RAV subfamily TF TEM1 (TEMPRANILLO 1) and TEM2 function as flowering repressors. TEM1/2 regulate the juvenile-to-adult growth transition and bind to the FT promoter to repress FT expression [49,50]. Recently, the structural basis of how the DNA-binding domains AP2 and B3 in TEM1 recognize CAACA and CACCTG motifs in the 5′-UTR of FT, respectively, has been revealed [51]. The combination of the AP2 and B3 binding sites significantly enhances the binding of TEM1 to FT 5′-UTR. The ability of TEM1 AP2 and B3 domains simultaneously binding to FT is necessary for recruiting PRC2 and deposition of H3K27me3 in FT 5′-UTR, to precisely regulate the floral transition.

Another group of AP2 transcription factors (known as the euAP2 family) also functions to repress FT expression. This family consists of AP2 (APETALA 2), the three TOE (TARGET OF EAT) proteins (TOE1, TOE2, and TOE3), and SMZ (SCHLAFMUTZE) and its paralog SNZ (SCHNARCHZAPFEN), which are suppressed by microRNA172 (miR172) [52]. miRNA172 expression is regulated in a CO-independent manner, and GI promotes miRNA172 expression under LD conditions [53]. miR156 regulates the expression of miR172 by increasing the level of SPL9 and SPL10, which directly promote miR172 expression in an age-dependent manner [54,55]. SMZ inhibits FT expression through directly binding to the FT promoter, and therefore, miR156 and miR172 are involved in controlling SMZ expression to determine the timing of FT expression in a CO-independent manner [52]. Under inductive LDs, euAP2 family loci including TOE1, TOE2 and SMZ are also regulated by SIN3 (SWI-INDEPENDENT 3) LIKE proteins, including SIN3 LIKE 1 (SNL1) to SNL5. SNLs bind to euAP2 family loci to mediate histone deacetylation and inhibit their expression, resulting in LD induction of FT expression [56]. It is very interesting that under non-inductive SD, SNLs mediate histone deacetylation of the FT activator AGL19 and inhibit its expression, thereby indirectly inhibiting FT expression, leading to delayed flowering under SD [56].

The MADS box TFs including SVP (SHORT VEGETATIVE PHASE) and the FLC (FLOWERING LOCUS C) family consisting of FLC, FLM (FLOWERING LOCUS M) and MAF2 (MADS AFFECTING FLOWERING 2) to MAF4, functions to repress FT expression. SVP physically interacts with FLC to directly repress FT expression via binding to the CArG motifs at the FT locus [57,58]. In addition, SVP functions together with FLM to mediate FT regulation in response to ambient temperature changes [59,60]. The MAF3 protein accumulates at the end of LDs and binds to FT chromatin to repress its expression [61]. FLC, FLM, MAF2 and MAF4 interact with SVP, suggesting that these proteins may form a large MADS-domain complexes to repress FT expression [60,61].

The level of CO protein is critical for the induction of FT expression. The molecular mechanism of the CO protein to regulate FT expression through its C-terminal CCT domain has been revealed through structural and biochemical analyses [62]. The N-terminal B-Box domains of CO form a possible tetrameric assembly, and the C-terminal CCT domain of CO interacts with NF–YB/YC (NUCLEAR FACTOR-YB/YC) to form the trimeric CO-CCT-NF-YB/YC complex (NF-CO). Four NF-CO complexes appear to function in concert to specifically bind multiple cis-acting TGTG-bearing elements in a proximal FT promoter region [62]. In addition, a NF-Y transcription factor complex composed of NF-YA, NF-YB and NF-YC, binds to the CCAAT site located in the distal FT promoter, with assistance from the ATPase-dependent chromatin-remodeling factor PKL (PICKLE) [63,64]. NF-CO and NF-Y appear to function together to promote chromatin looping in the FT promoter, resulting in a strong reduction of PcG enrichment on FT chromatin and consequent FT de-repression in leaf veins at dusk [65,66]. FT chromatin is also bound by TrxG (Trithorax group) proteins that mediate histone 3 lysine-4 trimethylation (H3K4me3) deposition under LD, and active chromatin modifications are required for FT de-repression at around dusk. The TrxG protein H3K4 methyltransferase ATX1 (ARABIDOPSIS HOMOLOG OF TRITHORAX 1) is recruited by PKL to FT chromatin specifically at around dusk [67]. In addition, the H3K4me3/H3K36me3 readers MRG1 (MORF RELATED GENE 1) and MRG2 can bind to FT chromatin at around dusk in LDs [68,69], to promote H3K4me3 deposition on FT chromatin. MRG2 also interacts with the NAP1 (NUCLEOSOME ASSEMBLY PROTEIN 1) family proteins, including NRP1 (NAP1-RELATED PROTEIN 1) and NRP2, and NRP1 and NRP2 inhibit the binding of MRG1/MRG2 to CO, leading to a transcriptional repression of FT [70].

FT expression is induced by CIB1 (CRYPTOCHROME-INTERACTING BASIC-HELIX-LOOP-HELIX 1), CIB2, CIB4 and CIB5 [71,72,73]. The CIB1 protein is presented from afternoon to early night in LDs, and interacts with CRY2 in a blue light-dependent manner; CIBs bind to the E-box elements (CANNTG) at FT to stimulate FT expression at around dusk in LDs [73].

Histone acetylation is positively linked with active gene expression in eukaryotes, and the acetylation level is dynamically controlled by histone acetyltransferases (HAT) and histone deacetylases (HDACs) [74]. Histone deacetylation is often associated with gene repression or down-regulation. Upon the activation of FT expression near dusk in LDs, HDACs bind to FT chromatin at dusk to repress its expression, and thus prevents overproduction of the FT protein. There are two types of HDAC complexes for FT repression, including AFR-HDAC and MRG-dependent HD2C complexes [75,76]. AFR-HDAC consists of HDA19 and structural components including a Sin3-like scaffold protein, SAP18 (Sin3-Associated Polypeptide 18), and AFR1 (SAP30 FUNCTION RELATED 1) or AFR2 [75]. AFR-HDAC is recruited to the FT locus specifically at the end of LDs by the two MADS-domain transcription factors AGL15 and AGL18, and this recruitment depends on the presence of CO activity. Hence, CO activity not only activates FT expression, but also enables the recruitment of an HDAC complex to FT to downregulate its expression [75]. HD2C acts as an effective deacetylase, mainly targeting H3K9ac, H3K23ac and H3K27ac. HD2C is recruited to FT chromatin to inhibit FT transcription at the end of the day in an MRG1/2-dependent manner. HD2C antagonizes CO for the binding of MRG2, and acts to promote the release of CO from FT for protein degradation [75]. In the middle of LDs, MRG proteins bind to H3K4me3/H3K36me3-marked chromatin and interact with CO to promote FT expression [75]. At dusk, AFR-HDAC and MRG-HD2C may work together to regulate FT expression at an appropriate level. This balance finely regulates FT expression at an appropriate level, and thus prevents precocious flowering in response to the inductive LD signals.

1.3. FT Protein Movement

Under LDs, FT mRNA is expressed in CCs (companion cells) of leaf vascular bundles, and the FT protein travels a long distance from leaves to SAM through the phloem to induce flowering. The movement of FT from CCs to sieve elements (SEs) is regulated by FTIP1 (FT-INTERACTING PROTEIN 1), QKY (QUIRKY) and SYP121 (SYNTAXIN OFPLANTS121) [77,78]. FTIP1 is an ER (endoplasmic reticulum) membrane protein, whereas both QKY and FTIP1 belong to the MCTP (multiple C2 domain and transmembrane protein) family. SYP121 is a syntaxin-like Q-SNARE protein with a C-terminal transmembrane domain that can mediate the transport of vesicles to the plasma membrane [79]. The nonfunctional mutants of ftip1, qky and syp121 are late-flowering under LDs. FTIP1 is specifically localized in phloem CCs and plasmodesmata between CCs and SEs, and mediates the FT protein transport through the ER system and plasmodesmata [78], SYP121 interacts with QKY, and mediates FT transport to the plasma membrane via endosomal vesicles, thereby promoting FT translocation to SEs [77]. The FT protein export from CCs to SEs is regulated in a temperature-dependent manner, and repressed by low temperature [80].

After FT enters into the phloem stream, the long-distance trafficking of FT protein from SEs to SAM is regulated by NaKR1 (SODIUM POTASSIUM ROOT DEFECTIVE 1), a heavy metal-associated domain-containing protein. Loss of function of NaKR1 causes late flowering under long-day conditions, largely because of FT transportation to SAM through SEs is hindered [81].

A recent breakthrough indicates that the negatively charged phospholipid PG (phosphatidyl-glycerol) on the cellular membranes of lipid bilayer can interact with the FT protein, and sequester FT [82]. Low temperature promotes FT sequestration in the cellular membrane of the CC, thereby reducing the level of soluble FT and delaying the transition to flowering [82]. These findings reveal the mechanism underlying how plants modulate the activity of florigen to optimize the timing of flowering in response to temperature changes. Low temperature-induced FT protein transport reduction and transcription inhibition act antagonistically to photoperiodic FT induction, so as to optimize flowering at a suitable time.

In summary, the molecular circuitry underlying long-day induction of flowering in Arabidopsis, composed of photoreceptors, circadian clock, CO and FT, has been well dissected. As most of the components in this regulatory system are evolutionarily conserved in other flowering plants, molecular understanding of photoperiodic regulation of flowering in Arabidopsis has provided the basis for exploring day-length regulation of flowering in other plants such as soybean.

2. Molecular Mechanisms for Photoperiodic Regulation of Flowering Time in Soybean

Soybean [Glycine max (L.) Merr.] is the main source of human vegetable oil and vegetable protein. Soybean is a typical SD crop, that is, flowering much more earlier in SDs (less than 12 hours) than under the conditions of long day (more than 16 hours). Originated in the 30°～45° north latitude region of China, soybean is grown in a wide range of latitudes and cultivated in broad regions, ranging from 50°N latitude to 35°South latitude. However, cultivation of individual varieties is usually limited to a narrow range of latitudes, mainly due to the high sensitivity to photoperiod.

Major cultivated soybean varieties in high latitude areas will bloom early when planted in low latitudes, with short plants and few pods; whereas when varieties adapted to low latitude areas are planted in high latitude areas, they will bloom too late and can not complete their life cycles before the temperature drops in winter [83]. Different soybean cultivars exhibit distinct flowering time and maturity traits according to their habitats, and early maturity is usually accompanied by low yields. The response of soybean to photoperiod usually affects the length of the maturity period. Photoperiodic flowering regulation is an important agronomic trait, critical to soybean yield, quality and adaptability. Therefore, analysis of the molecular mechanism of soybean photoperiodic regulation of flowering can provide a theoretical basis for solving the contradiction between early maturity and high yield, and key core modules for the molecular design and breeding high-yield and high-quality soybean cultivation variety.

2.1. Molecular Basis of Soybean E Series Genes for Flowering-Time Regulation

As early as the 1920s, it was observed that soybean varieties sown in different times in a year bloom almost at the same time, so scientist used soybeans and tobacco (Nicotiana tabacum) as model plants to uncover the phenomenon of plant photoperiod [84]. According to the traditional approaches to genetics, a number of soybean major genetic variations have been found to be responsibile for wide adaptability of soybean plants. To date, a number of major genetic loci, namely E1 [85], E2 [86], E3 [87], E4 [88], E5 [89], E6 [90], E7 [91], E8 [92], E9 [93], E10 [94], E11 [95], J [96] and several QTLs, such as Tof11/Gp11 and Tof12/Gp1/qFT12-1 [97], Tof16 [98], LJ16.1 and LJ16.2 [99], have been identified to be involved in the control of flowering and maturity in soybean [83]. Dominant alleles of E1, E2, E3, E4, E7, E8, and E10 inhibit flowering, whereas dominant alleles at E6, E9, E11 and J promote flowering [83]. E1, E3, E4, E7 and E8 are involved in photoperiod sensitivity, especially to different light qualities under artificially induced LDs [100].

To date, E1-E4 and E9 have been cloned and studied in depth. E4 and E3 are phytochrome A (PHYA) genes, GmPHYA2 and GmPHYA3, respectively. Soybean plants exhibit different responses to different red light: far-red light (R: FR) quantum ratios [101]. GmPHYA3 (E3) participates in the control of flowering under LDs with a high R:FR quantum ratio [102], while GmPHYA2 (E4) is responsible for the flowering response of LDs with a low R:FR ratio [91,101]. E3 and E4 double-dominant genotypes are sensitive to photoperiod response, which show delayed flowering and maturation, whereas homozygous double recessive genotypes are insensitive to photoperiod response, which are earlier flowering and maturation. GmPHYA1 and GmPHYA2 (E4) can coordinately regulate the photomorphogenesis under a low ratio of R: FR light [103]. GmPHYA1 is likely to regulate the photoperiod sensitivity under LD conditions with a lower ratio of R: FR (<1.0) [104,105].

E2 is a homolog of the Arabidopsis GI [106]. E2 and its near-isogenic line (e2) display similar flowering times at high latitudes 43° N and mid-latitude 36° N [106], indicating that the regulation of soybean flowering period by E2 may not depend on photoperiod, so it has a small impact on the photoperiod response. Studies have shown that the function of E2 in soybean and that of GI in Arabidopsis may have been differentiated: the full-length E2 gene is unable to rescue the late flowering phenotype of the Arabidopsis gi [107]. E2 regulates flowering by inhibiting the transcription of GmFT2a, but the molecular mechanism is still unclear [20]. Based on studying of the E2 haplotypes of cultivated soybeans and wild soybeans in different regions of China, it has been found that the diversification of E2 haplotypes may help soybean flowering time adaptation, and this adaptation promotes the spread of domesticated soybeans [107].

Kong et al. discovered the E9 gene [14], and the dominant E9 gene confers early flowering, whereas the recessive e9 gene confers late flowering [93]. E9 is the soybean florigen gene GmFT2a [93,108]. The recessive e9 gene is due to the retrotransposon SORE1 inserted into the first intron of the GmFT2a gene, which inhibits its function and delays flowering. Further study showed that both sides of the intron, where retrotransposon SORE1 inserts, is methylated, and methylation affects the expression of GmFT2a gene [108]. Another FT homolog, GmFT4 plays important roles in inhibiting soybean flowering under non-inductive LD conditions, and is strongly induced by LDs. GmFT4 most likely is the candidate gene for the newly identified mature locus E10 [94].

Soybean E1 locus has the greatest impact on flowering and maturity periods, and is located near the centrioles of chromosome 6, which brings great difficulty to QTL mapping and cloning. Xia et al. found that E1 is a specific TF for legumes, containing binary nuclear localization signal, a DNA binding site, and a B3 domain. Mutations in the distant B3 domain reveal that this domain is closely related to the function of E1 to inhibit flowering [109]. The expression of E1 is significantly inhibited under SD conditions, while a bimodal circadian pattern is exhibited under LD conditions, indicating that the E1 gene was regulated by the photoperiod, and that LDs induce the expression of E1. This is the main reason for that soybean has become a SD crop. However, in the e3/e4 genetic background, E1 induction by LDs is eliminated, indicating that the E1 gene is controlled by E3 and E4. Soybean has two E1 homologues, E1La and E1Lb, and their expression patterns are similar to E1 under both LD and SD, E1 and E1L genes repress flowering by down-regulating GmFT2a and GmFT5a, under LD conditions, but promote the expression of the flowering inhibitor GmFT4 [110].

2.2. Core Components in the Photoperiod Pathway in Soybean

The florigen gene FT and FT relatives are evolutionarily-conserved key flowering promoters. 12 FT-like genes have been identified in soybean [111]. Overexpression of GmFT2a/2b, GmFT3a/3b and GmFT5a/5b in Arabidopsis, promotes flowering, similar to the Arabidopsis FT; in contrast, overexpression of GmFT1a/1b, GmFT4 and GmFT6 delays flowering. In addition, GmFT2d has no function in cultivated and wild soybeans, and GmFT2c in some wild soybean accessions is functional and can promote flowering [112]. Various studies show that GmFT2a and GmFT5a are key floral integration factors in soybeans [85,106,111,113,114]. GmFT2b is homologous to GmFT2a, and Gmft2b mutants show delayed flowering only under LD conditions [115]. GmFT2b can influence the expression of GmFT genes, and GmFT2a and GmFT5a are significantly upregulated in GmFT2b-overexpression plants. GmFT2b haplotypes in different maturity groups play important roles in soybean variety distribution [115]. GmFT1a and GmFT4 are up-regulated by E1 and function as flowering inhibitors, and their overexpression inhibits the expression of the floral meristem identity genes GmAP1b and GmAP1c, and thus delays flowering [110,116]. GmFT1a and FT4 exhibit inhibitory effects, suggesting that FT-like genes may have undergone functional differentiation in soybean, and the functional divergence most probably due to the differences in protein sequence and structure caused by changes in critical amino acids [117]. The expression of GmFT2a/2b and GmFT5a are induced in soybean leaves under inductive SDs, whereas the expression of GmFT1a and GmFT4 are induced under LDs. GmFT4 and GmFT1a are barely detectable under SD conditions [110,111,116], and there is no direct evidence to support that GmFT4 and GmFT1a are transported from leaves to SAM to inhibit soybean flowering. Although these two genes are highly expressed in leaves, they are also expressed in SAM at a low level [110,111,116]. Relative transcript abundance of the flowering promoters GmFT2a/5a and the flowering inhibitors GmFT1a/4, is important for determining the appropriate flowering time under different growth environments (Figure 2).

In the soybean photoperiodic response, the growth period genes E1-E4 participate in photoperiodic regulation of flowering. E3 has a greater effect on E1 compared with E4 [85,118,119]. E1 and its family genes are all soybean flowering inhibitors, which strongly inhibit GmFT2a and GmFT5a and delay flowering [85,119]. GmFT2a and GmFT5a are the key flowering integration factors in soybeans, which coordinately regulate flowering in the photoperiod pathway [111]. GmFT2a /5a and GmFT4 genes are all regulated by E1. Therefore, in the photoperiod response of soybean, there is a major flowering-regulation pathway, E3/E4-E1-GmFT. E1 is a unique TF for legume crops, so E1 defines a unique regulatory pathway to control the flowering and maturity periods.

There are several genes in the photoperiod flowering pathway to interact with E1 in flowering regulation. Overexpression of GmCOL1a, an Arabidopsis CO-like gene, leads to delayed flowering under LDs, and the Gmcol1b mutant is early flowering. These indicate that soybean GmCOL1a and GmCOL1b are flowering inhibitors. When GmCOL1a is overexpressed, the soybean growth period genes E1 and E2 are down-regulated. However, GmCOL1a is up-regulated in the near-isogenic line of E1 and E2, so it is very likely that there is feedback regulation among GmCOL1a, E1 and E2. In addition, E3 and E4 promote the expression of E2 and GmCOL1a/1b [118].

The photoperiod response of soybean is an extremely complex process that requires the coordination of multiple genes to finally achieve flowering. miR172 and its target genes play a regulatory role in growth and development [120]. In the photoperiod-regulated soybean flowering network, miR156 delays flowering, and the E1 gene is also involved in the miR156/GmSPL pathway [121]. miR156 regulates its target gene GmSPL3/9, and miR172 act through its target gene GmTOE4a to feedback regulate of miR156 and GmSPL3/9. There is also a negative feedback regulation between miR172 and its target gene GmTOE4a [122]. Therefore, there may be a balance between miR156 and miR172, which regulates soybean flowering at an appropriate time (Figure 3).

2.3. Molecular Mechanisms for Adaptation to Different Latitudes

High latitudes regions are associated with early winter and longer day lengths in summer, therefore, soybean needs to reduce photoperiod sensitivity to flower and mature early in LDs. E1, E3, and E4 play an important role in defining photoperiod sensitivity and adaptation to high latitudes. Double recessive genotype e3e4 is the most common genotype to reduce the photoperiod sensitivity, followed with e1 and e3 or e4 [123]. E1 shows functional redundancy with the E1L genes, and e1 reduces but does not eliminate photoperiod insensitivity, but RNAi-mediated inhibition of E1La and E1Lb eliminates the residual flowering response [119]. There are dysfunctional and hypomorphic alleles of E1, E3, and E4, and usually are rare and region-specific [100,123,124]. Various combinations of the dysfunctional alleles at the E3 and E4 and E1/E1L loci, affect photoperiod sensitivity to varying degrees and help soybean adapt to longer day length in growth seasons at higher latitudes.

Recently, the QTLs of Gp11/Tof11 and qFT12-1/Gp12/Tof12, which control soybean flowering and maturity time, have been identified. Tof11 and Tof12 encode homologs of APRR3[97,125,126], named as GmPRR3a and GmPRR3b, respectively. Tof11 and Tof12 are redundant to control flowering time. The two dominant Tof11 and Tof12 alleles flower much more later than the single dominant alleles in Tof11 or Tof12 [97,126], in line with the overexpression phenotype of the Arabidopsis APRR3 [127]. GmPRR3a and GmPRR3b attenuate the inhibition of the circadian clock gene LHY on the legume-specific photoperiod regulation gene E1, leading to down-regulation of GmFT2a/5a and a delay in flowering. Overexpression of GmPRR3a and GmPRR3b increases the expression level of E1 and decreases the expression of GmFT2a and GmFT5a. Loss of GmPRR3a and GmPRR3b function may allow early harvesting and improve adaptation to the limited summer growing season in high-latitude regions during soybean domestication. Population genetic analysis reveals that Tof11 and Tof12 have gradual mutations and are under artificial selection in breeding. The tof12-1 mutation has been first strongly selected to accelerate the flowering and maturity periods of cultivars generally. The tof11-1 mutation occurred after tof12-1, thereby further shortening the flowering period and growth period of cultivated soybeans, and thus improving the adaptability and planting of cultivated soybeans. Regardless of its origin, tof12 exists in almost all improved soybean varieties [97,125,126], GmPRR3a (tof11) and GmGIa (e2) mutant alleles are also predominant in cultivated varieties of northern China and photoperiod-insensitive varieties [97,107,123], indicating that these three alleles have undergone strong artificial selection, and may provide the basis for the early stage of northward expansion.

GmPRR37, which encodes a pseudo-response regulator protein, is confirmed to be the major QTL qFT12-2, and Gmprr37 mutants exhibit early flowering under LD conditions. Overexpression of GmPRR37 significantly delays flowering, through down-regulated expression of GmFT2a/5a and up-regulated GmFT1a (a flowering-inhibitor). Further study shows that soybean adapts to higher latitudes by flowering early, relying in part on natural GmPRR37 mutation [128].

A recent study shows that cotyledons facilitate the early-maturing soybean varieties adapting to high latitudes [129]. After the cotyledons emerged from the soil and exposed to light, GmFT2a was rapidly expressed at a high level, and the expression of downstream genes such as the floral meristem-identity gene GmAP1 was up-regulated. This is an important mechanism for early-maturing soybean varieties to flower and mature early in high latitudes in LD conditions [129].

Low latitudes regions are associated with shorter day lengths in growth seasons, and thus it is necessary overcome the sensitivity to photoperiod when soybean is planted in these areas, otherwise, the flowering is too early and the yield will be significantly reduced. For example, before 1960, the soybean planting area in Brazil was limited to latitudes above 22° S, and soybean production was less than 10 million tons [83], until the long juvenile (LJ) traits were discovered and introduced in 1970s. The LJ trait plays an important role in soybean production in low-latitude regions, which allows the soybean flowering later and has an extended vegetative growth period under SDs. E3 and E4 play a minor role to control flowering under SD, as loss-of function e3 and e4 alleles have almost no effect on flowering under SD, compared to their respective wild-type alleles [130]. Some loci have been described, which may contribute to LJ, such as E6 [90,131], J [130], Lj16.1 and Lj16.2 [132]. LJ16.1 and LJ16.2 encode the florigen (FT) homologs GmFT2a and GmFT5a, respectively. Gmft2a or Gmft5a single mutants show genetic compensation responses and have relatively little effect on flowering time, while the Gmft2a Gmft5a double mutations could break this compensation response, exhibiting enhanced LJ phenotype, so as to produce high yields in SD. GmFT2a and GmFT5a sequence diversity is a major factor in soybean spreading to lower latitudes [99].

Recently, the J gene has been cloned, encoding an ortholog of the Arabidopsis ELF3 [130,133], and acting as a member of the circadian evening complex (EC). E6 was also identified by map-based cloning. E6 is a new allelic variant of the J gene, and the recessive e6^PG allele contains a Ty1/copialike retrotransposon inserted in the fourth exon of GmELF3 [134], so both J and E6 are alleles of the GmELF3 gene. GmELF3 is suppressed by two PHYA proteins, E3 and E4 [130]. GmELF3 can bind to E1 promoter regions and inhibits its expression under SDs. Recessive j alleles inhibit flowering under SD and prolong the vegetative period by releasing E1 repression. Loss of GmELF3 function results in a 30%–50% increase in yield [130]. Knockout of the two homologous GmLUX1 and GmLUX2, encoding components of the EC complex, gives rise to a complete lost in photoperiod sensitivity. The Gmlux1 Gmlux2 mutant shows a dramatically extended reproductive period, and thus the EC complex is the core of the soybean photoperiodic flowering control. The J protein interacts with GmLUX1 and GmLUX2 to form the soybean EC complex J-GmLUX1/GmLUX2, which binds to the promoters of E1 and its two homologs E1La and E1Lb to suppress their expression, further relieving GmFT2a and GmFT5a expression and thus promoting flowering under SD conditions [135]. In Arabidopsis, mutants of EC components ELF3 and LUX have similar early-flowering phenotypes, indicating that the function of EC genes is conserved [136,137].

A novel locus Time of Flowering 16 (Tof16), that delays flowering and improves yield at low latitudes, has been identified. Tof16 encodes LHY1a, one of the four soybean homologs of the Arabidopsis Circadian Clock Associated 1 (CCA1) and Late Elongated Hypocotyl (LHY) [98]. Under SDs, the loss of function allelic variation of Tof16 significantly prolongs the flowering period and increases soybean yield. Tof16 directly regulates the expression of E1, and thus regulates the photoperiod flowering of soybean. There are four soybean LHY homologs to redundantly control flowering time and yield. Indeed, the quadruple knockout mutants produced by CRISPR-Cas9 shows significantly delayed flowering [98,138], whereas the Arabidopsis lhy mutant flowers early [139]. Tof16 and J are independent in controlling the flowering period and yield of soybeans in low latitude regions, and have additive genetic effects. Further analysis revealed that more than 80% of soybean varieties in low-latitude regions bear loss-of-function allelic variants of these two genes, indicating that Tof16 and J play very important roles in the adaptation of soybeans to low-latitude regions. Hence, natural variation at the Tof16 or J locus is the main genetic basis for cultivated soybeans to adapt to tropical regions.

3. Conservation and Divergence of the Photoperiodic Regulation of Flowering in Soybean and Arabidopsis

The photoperiod pathways in soybean and Arabidopsis have both conservative and unique aspects (

Table 1. Comparison of photoperiodic flowering-regulatory genes in soybean and Arabidopsis.

Gene in Soybean	Function		Arabidopsis Ortholog	Function
E1	Inhibition	[85]	None
E1La	Inhibition	[119]	None
E1Lb	Inhibition	[106,119]	None
GmGI (E2)	Inhibition	[106]	GI	Promotion	[25,28]
GmphyA3 (E3)	Inhibition	[102]	PHYA	Promotion	[5]
GmphyA2 (E4)	Inhibition	[103]	PHYA	Promotion	[5]
GmFT2a (E9)	Promotion	[93,108,111]	FT	Promotion	[43,44]
GmFT5a	Promotion	[111]	FT	Promotion	[43,44]
GmFT4 (E10)	Inhibition	[94,110]	FT	Promotion	[43,44]
GmFT1a	Inhibition	[116]	FT	Promotion	[43,44]
GmCOL1a/1b	Inhibition	[118]	CO	Promotion	[26]
miR156	Inhibition	[121]	miR156	Inhibition	[54,55]
miR172	Promotion	[122]	miRNA172	Promotion	[53,54]
GmTOE4a	Inhibition	[122]	TOE1	Inhibition	[7,141]
GmPRR3a (Tof11)	Inhibition	[97,125]	APRR3	Inhibition	[127]
GmPRR3b (Tof12)	Inhibition	[97,125]	APRR3	Inhibition	[127]
GmELF3 (J)	Promotion	[130,133]	ELF3	Inhibition	[137]
GmLUX1	Promotion	[135]	LUX	Inhibition	[136]
GmLUX2	Promotion	[135]	LUX	Inhibition	[136]
LHY1a (Tof16)	Promotion	[98]	LHY	Inhibition	[139]
LHY1b/c/d	Promotion	[98]	LHY	Inhibition	[139]
GmFDL19	Promotion	[113]	FD	Promotion	[45]

“Inhibition” indicates floral inhibition; “promotion” for floral promotion.

). In Arabidopsis, the circadian clock output gene CO plays an important regulatory role in photoperiodic regulation, and the photoperiodic regulation of flowering in Arabidopsis is the GI-CO-FT regulatory module (Figure 1). The function of CO in flowering regulation is relatively conserved, and usually plays a role in inducing flowering in many species such as Arabidopsis, and rice [140]. However, GmCOL1a and GmCOL1b act to repress flowering in soybean. This may be related to that GmCOL1a/1b expression is regulated by the soybean-specific TF E1. GmCOL1a/1b are regulated by the GI homologous gene E2, and function to repress the expression of GmFT2a/5a. Therefore, the photoperiod pathway is partially conserved in soybean, with the GI (E2)-CO-FT model, but the regulation of GmCOL1a /1b by E1 is unique in soybean (Figure 2).

In Arabidopsis, GI not only regulates CO expression, but also regulates miR172 and its target gene TOE1 [53]. TOE1 binds to the FT promoter and inhibits FT expression in the morning [141]. There is the GI-miR172-TOE1-FT module to promote flowering in Arabidopsis. In soybean, GmTOE4a participates in the photoperiodic flowering regulation that requires E3 and E4, which is dependent on the expression of GmCOL1a; however, GmTOE4a expression is not regulated by the GI homolog E2 [122]. This is different from the miR172 pathway in Arabidopsis.

In Arabidopsis, the FT proteins move to the SAM to promote the expression of SOC1, LFY, AP1 and other floral meristem identity genes, and thus induces flowering. In soybean, GmFT2a and GmFT5a have been confirmed by various studies as key floral promoters [111,114]. GmFT2a and GmFT5a redundantly control soybean flowering in response to day-length changes. Both GmFT2a and GmFT5a move from the leaves to SAM and interact with the soybean ortholog of the Arabidopsis bZIP transcription factor FD, GmFDL19 [113]. GmFT2a/5a-GmFDL19 directly binds to the cis-acting element of the GmAP1a promoter to promote the expression of GmAP1a, and other Arabidopsis orthologs of floral promoter genes, such as SOC1 (GmSOC1a and GmSOC1b) and LFY (GmLFY2), thereby to induce flowering [113]. The Gmap1 quadruple mutant shows late flowering phenotype under inductive SD conditions, whereas overexpression of GmAP1a results in early flowering [142]. These results indicate that, like Arabidopsis, the FT/FD-AP1 module is highly conserved in soybeans. GmFT2a/5a have conserved ability to induce flowering, but the mechanisms underlying the regulation of these FT relatives are not well conserved between soybean and Arabidopsis.

Soybean photoperiod pathway starts with PhyA and is finally reflected in the changes in the expression level of GmFTs. The key core transcription factors in this pathway include E1 and its two homologs, E1La and E1Lb. E1/E1L inhibit the expression of GmFT5a and GmFT2a. Under inductive SDs, the expression of E1 is inhibited by J (GmELF3) and Tof16 (GmLHYs) in soybean [98]. GmELF3 interacts with GmLUX1 and GmLUX2 to form the soybean evening complex of GmELF3-GmLUX1-GmLUX2. This complex functions as part of the circadian clock and directly binds to cis-acting elements in E1/E1L promoters, resulting in further relieving of the expression of GmFT2a and GmFT5a, and thus promoting floral transition [135]. Hence, the molecular pathway of phyA (E3E4)-GmELF3/GmLHY-E1/E1L-FT2a/5a induces rapid flowering in inductive SDs (Figure 3). Under LD conditions, the transition to flowering is greatly delayed in soybean. Long-day light exposure induces the expression of GmPRR3a and GmPRR3b in part through the two phytochrome A photoreceptors E3 and E4, leading to a partial release of the E1 inhibition; hence, the molecular pathway of phyA (E3E4)- GmPRR3a /3b-E1/E1L-FT2a/5a eventually leads to flowering in LDs.

E1 is the core component in the soybean photoperiod pathway and functions to integrate the light signal and the circadian clock signal, which is similar to the role of CO in the long-day induction of flowering in Arabidopsis; however, E1 and CO are not homologs. This highlights the divergence of photoperiodic regulation of flowering between the two eudicots Arabidopsis and soybean. Further research is needed to fully understand how day-length signals at different latitudes induce the transition to flowering and control seed maturation time to maximize seed production in soybean.

The molecular circuitry underlying photoperiodic regulation of flowering time in the SD plant soybean is only partially conserved in other SD plants. In rice, a facultative SD plant, the core components of photoperiod pathway including E2 (GI), CO and FT are evolutionarily conserved, but the OsGI (GI ortholog)-OsHd1 (Heading date 1, CO ortholog)-OsHd3 (Heading date 3, FT homolog) pathway functions differently from that in soybean in floral induction in response to SD signals [143,144,145,146]. The key player in soybean flowering regulation, E1, is not conserved in rice [85,143,147]. There are monocot-specific flowering regulators in rice and other grass plants; for instance, the potent floral repressor Ghd7 (Grain number, plant height and heading date 7), which inhibits rice flowering in long days [148,149,150], is not found in Arabidopsis or soybean. Apparently, the molecular networks underlying photoperiodic regulation of flowering in different species possess species-specific characteristics for their adaptation to local seasonal day-length changes. This highlights the diversity and complexity of flowering-time regulation in diverse plants.