Transcriptional and post‐transcriptional regulation of heading date in rice

Summary Rice is a facultative short day (SD) plant. In addition to serving as a model plant for molecular genetic studies of monocots, rice is a staple crop for about half of the world’s population. Heading date is a critical agronomic trait, and many genes controlling heading date have been cloned over the last 2 decades. The mechanism of flowering in rice from recognition of day length by leaves to floral activation in the shoot apical meristem has been extensively studied. In this review, we summarise current progress on transcriptional and post‐transcriptional regulation of heading date in rice, with emphasis on post‐translational modifications of key regulators, including Heading date 1 (Hd1), Early heading date 1 (Ehd1), Grain number, plant height, and heading date7 (Ghd7). The contribution of heading date genes to heterosis and the expansion of rice cultivation areas from low‐latitude to high‐latitude regions are also discussed. To overcome the limitations of diverse genetic backgrounds used in heading date studies and to gain a clearer understanding of flowering in rice, we propose a systematic collection of genetic resources in a common genetic background. Strategies in breeding adapted cultivars by rational design are also discussed.


Introduction
Following seed germination rice undergoes a vegetative growth phase before panicle initiation, at which there is a change in apical meristem identity from a vegetative meristem to a floral meristem as the plant enters the reproductive phase that leads to flowering. Flowering time (also known as heading date) is an important agronomic trait that determines seasonal and regional adaptation. If a cultivar flowers too early in a specific location there will be inadequate use of light and temperature resources, and consequent lower yield. By contrast, if a cultivar is too late in heading it cannot complete flowering and grain development before the onset of cold, also resulting in lower yield.
Heading date in rice is affected by exogenous factors such as photoperiod, temperature and nutrient availability (e.g. nitrogen levels), among which photoperiod is a key factor. Rice is a facultative short day (SD) plant; flowering is promoted under SD conditions and inhibited by long day (LD) conditions. However, in the juvenile growth stage of the vegetative growth phase, flowering cannot be promoted even under inductive daylength conditions. This stage is called the basic vegetative phase (BVP). Only after completion of the BVP can rice respond to photoperiodic stimuli for flowering, and this stage is called the photoperiod-sensitive phase (Vergara & Chang, 1985;Nishida et al., 2001). The length of BVP and the sensitivity to exogenous factors is determined by endogenous genetic composition, a fundamental factor that determines differences in photoperiod response among rice genotypes and leads to extensive variation in heading date.

Genetic pathways controlling flowering in rice
The general process of photoperiod-induced flowering is conserved in LD and SD plants. First, the leaves measure the photoperiod and generate a mobile flowering signal called a florigen under inductive *These authors contributed equally to this work. daylength. The florigen is then transported from leaves to the shoot apex where the shoot apical meristem (SAM) perceives the florigen signal and activates downstream floral identity genes to trigger transition to flowering. The genetic pathway for photoperiodcontrolled flowering has been intensively studied in the model plant Arabidopsis, a facultative LD plant. In this plant, GIGANTEA (GI) protein promotes CONSTANS (CO) transcription in leaves, and CO protein directly activates transcription of the florigen FLOWERING LOCUS T (FT). However, there are multiple additional layers of transcriptional and post-transcriptional modifications that regulate the photoperiodic flowering pathway to influence flowering time (Blumel et al., 2015).
Using forward and reverse genetics, as well as population genetic studies, many genes involved in photoperiodic flowering have been cloned into rice ( Table 1). Most of these genes function upstream of the florigen genes. These genes can be roughly grouped to two signalling pathways with some crosstalk. The evolutionarily conserved OsGI-Hd1-Hd3a pathway (analogous to the Arabidopsis GI-CO-FT pathway) and Ehd1-centred specific pathway in monocots involve Ehd1, Ghd7, Early heading date 2 (Ehd2, also known as RID1, or OsId1), Early heading date 3 (Ehd3), and Early heading date 4 (Ehd4), which has no counterpart in Arabidopsis (Fig. 1). Both pathways are finally integrated to florigen. Unlike Arabidopsis that has only one florigen gene, rice has two florigen genes, Heading date 3a (Hd3a) and RICE FLOWERING LOCUS T 1 (RFT1). Both genes encode a phosphatidylethanolamine binding protein that is a close homologue of Arabidopsis FT.

Recognition of SDs through circadian clock and light signals
As a SD plant, flowering in rice is induced when the day length becomes less than a specific threshold, known as the critical day length. When the day length is shorter than 13 h, the florigen gene Hd3a is expressed in the morning even in rice seedlings, and when day length exceeds 13.5 h, expression of Hd3a decreases to less than one-tenth of the 13 h d length level (Itoh et al., 2010). Based on physiological studies, it has been hypothesised that plants monitor day length changes by recognising light conditions at a specific time of day (B€ unning, 1960). In rice, coincidence between a circadian clock controlled fluctuating internal signal and a periodic external signal can start the flowering process, and this mechanism is known as external coincidence . Many heading date genes show rhythmic expression that is dependent on day length. OsGI has a rhythmic expression pattern that peaks at the end of the light period and activates expression of Hd1. Hd1 encodes the homologue of Arabidopsis CO and shows diurnal rhythmic expression that reaches a peak during the night both under SD and LD conditions but, under LD conditions, mRNA is also produced for several hours during the light phase . Although Hd1 promotes flowering under SD, artificial overexpression of Hd1 during the light phase inhibits flowering under SD conditions (Fig. 2). This suggests that exposure of Hd1 to light converts it to a repressor of flowering (Ishikawa et al., 2011). Night-break treatment under SD by exposure of rice plants to as little as 10 min of light in the middle of a long night results in delayed flowering and failure to induce Hd3a expression (Ishikawa et al., 2005).
Ghd7 encodes a CCT domain-containing protein and is a key floral repressor in rice under LD conditions (Xue et al., 2008;. Ghd7 is expressed in the morning under LD conditions. This day length-dependent expression is achieved by a complex gating mechanism. A red-light signal mediated by phytochromes induces Ghd7 expression only at a specific time of day when the gate (a light-sensitive phase set by the circadian clock) for Ghd7 expression is open. Under LD conditions, the Ghd7 gate is set to open around dawn, and light induces Ghd7 expression. Under SD conditions it opens at midnight, therefore there is no light to induce Ghd7 expression (Itoh et al., 2010). By contrast with the Ghd7 gate for red light, the Ehd1 gate for blue light is not affected by day length. Ehd1 expression can be induced by blue-light treatment in the morning in an OsGI-dependent manner under both LD and SD conditions. Because Ehd1 expression is repressed by Ghd7, red light induces Ghd7 transcription, leading to suppression of Ehd1 and Hd3a expression in LD. Weak expression of Ghd7 in SDs allows induction of the Ehd1 gene, leading to activation of Hd3a expression ( Fig. 2) (Itoh et al., 2010).

Floral activation in the SAM by the florigen activation complex
The florigen proteins Hd3a and RFT1 are transported to the SAM after induction in leaves. In the SAM, they form the florigen OsGI SD/LD promotion Os01g0182600 Encodes an orthologue of Arabidopsis GIGANTEA, a plant-specific nuclear protein that functions in diverse physiological processes Hayama et al. (2002) OsEF3 SD/LD promotion Os01g0566100 Encodes an orthologue of Arabidopsis ELF3, a putative nematode responsive protein-like protein activation complex (FAC) with bZIP transcription factor OsFD1 and 14-3-3 protein of the Gf14 family to induce expression of the primary targets, including members of MADS-box transcription factor family genes that control the phase transition of SAM from vegetative to reproductive development (Taoka et al., 2011). Both yeast-two-hybrid assays and transgenic studies demonstrated that phosphorylation of OsFD1 is important for binding to 14-3-3 protein and is a rate-limiting step for FAC formation (Taoka et al., 2011). FACs can also form in rice leaves, and these complexes are required to activate a positive feedback loop on Ehd1, Hd3a and RFT1 expression. In addition, two additional bZIPs, including Hd3a BINDING REPRESSOR FACTOR1 (HBF1) and HBF2, form alternative FACs with the florigens and reduce Hd3a and RFT1 expression to delay flowering. This situation suggests that Hd3a and RFT1 can regulate their own expression in leaves through forming antagonistic transcription factor complexes to fine tune photoperiod-induced flowering responses . In addition, RICE CENTRORADIALIS (RCN) genes encoding PEBP-like proteins are homologous to the Arabidopsis TERMINAL FLOWER 1 (TFL1) and have similarity to the florigens. RCN represses flowering through competition with Hd3a for 14-3-3 binding, and further combines with OsFD1 to form a FAC-like hexameric complex named the florigen repression complex (FRC). The balance between FAC and FRC modulates florigen activity to optimise inflorescence development (Kaneko-Suzuki et al., 2018). We recently identified a new CONSTANS-  © 2020 The Authors New Phytologist © 2020 New Phytologist Foundation like CCT domain protein called Delay heading date 4 (DHD4), which also represses flowering by affecting formation of the FAC complex in the SAM. By contrast with RCNs that interact with OsFD1 mediated by 14-3-3, DHD4 directly interacts with OsFD1 and affects the formation of FAC , and possibly represents another type of flowering regulation mechanism in SAM.
Transcriptional and post-transcriptional regulation of key heading date regulators

Dual functional regulation of Hd1
Hd1 is the sole homologue of Arabidopsis CO. CO is generally known as a strong LD-specific floral promoting factor in Fig. 1 Genetic pathways controlling heading date in rice. Schematic representation of genetic pathways controlling flowering in rice. A circadian clock is shown at the top. Key heading date genes are indicated in oval backgrounds. Arrows indicate upregulation, and bars indicate downregulation. Rice has two florigen genes, Hd3a and RFT1, short day (SD) conditions accelerate heading by promoting the expression of Hd3a through Hd1 and Ehd1. Expression of Ehd1 can be upregulated or downregulated by many genes. Hd1 can be activated by OsGI under SD conditions. In long day (LD) conditions, Hd1 function reverses to flowering inhibition; this reversal is controlled by many other factors such as Ghd7 and DTH8 and is mediated by protein-protein interactions. The florigen proteins Hd3a and RFT1 are transported to the shoot apical meristem (SAM) after induction in the leaves. In the SAM, they form the florigen activation complex (FAC), and activate the expression of downstream targets, including members of MADS-box transcription factor family genes that control phase transition of the SAM from vegetative to reproductive development. FAC and repressive FAC-like complexes can also be formed in the SAM and leaves to fine tune the heading date. Arabidopsis (Putterill et al., 1995), although it may also play a role in repressing flowering under SD conditions (Luccioni et al., 2019). However, the function of Hd1 has diverged during evolution. It advances heading by activating transcription of the florigen genes under SDs, but delays heading under LD conditions . Post-transcriptional regulation of Hd1 must also be involved in the functional conversion and other aspects of Hd1 function because Hd1 has a similar expression pattern in different photoperiods Ishikawa et al., 2011). Whether Hd1 functions as an activator or suppressor of heading depends on the red-light photoreceptor phytochrome B (PhyB). Mutation of PhyB or phytochrome chromophore synthesis gene photoperiod sensitivity 5 (Se5) attenuates the reversible function of Hd1 in which Hd1 works as an activator under all photoperiods in the absence of phytochromes. As Hd1 expression is not affected in the se5 mutant it is thought that phytochromes may affect Hd1 function post-transcriptionally .
Hd1 also loses LD heading suppression function and consistently promotes flowering under SD and LD in Ghd7-or DTH8deficient backgrounds (Nemoto et al., 2016;Du et al., 2017;Zhang et al., 2017;Zong et al., 2021). Transcriptional analyses revealed that Ghd7 and Hd1 do not regulate each other's transcription levels, suggesting that the observed genetic interaction might occur post-transcriptionally (Nemoto et al., 2016). Hd1 was shown to interact with flowering repressor Ghd7, and the Hd1-Ghd7 complex binds to specific sites in the Ehd1 promoter and suppresses expression of Ehd1 and florigen genes during daylight hours under LD conditions (Nemoto et al., 2016). As the interaction occurs between the transcriptionactivating domain of Hd1 and CCT domain of Ghd7, it was hypothesised that the interaction might block or weaken the

Long day
Gate for Ghd7 expression controlled by circadian rhythm

Ehd1 expression induced by blue light
Gate for Ehd1 expression controlled by circadian rhythm . Arrows indicate induction, and bars indicate suppression of gene expression. Ghd7 expression has the highest red-light inducibility around dawn under LD, whereas the inducibility peak shifts to midnight under SD conditions. Ehd1 expression has the highest blue-light inducibility around dawn under both LD and SD conditions. The light-sensitive phase set by the circadian clock is called a gate, which is indicated by dashed curves. Under LD, red light induces Ghd7 transcription around dawn, further suppressing the expression of Ehd1 and Hd3a. However, under SD, Ehd1 expresses around dawn and induces expression of Hd3a. Hd1 expression shows a diurnal rhythm and reaches a peak during the night under both SD and LD, but under LD, mRNA is also produced for several hours during the light phase. Dark phase-expressed Hd1 promotes flowering under SD, whereas light phase-expressed Hd1 inhibits flowering under LD.

Short day
New Phytologist (2021) 230: 943-956 www.newphytologist.com © 2020 The Authors New Phytologist © 2020 New Phytologist Foundation transcriptional activation activity of Hd1 and release the downstream gene transcriptional repression activity of Ghd7 . Because the expression level of Ghd7 is significantly lower in SDs (Xue et al., 2008), less Hd1-Ghd7 complex forms and so Hd1 still promotes heading date on a Ghd7 background under SDs (Nemoto et al., 2016;Zhang et al., 2017).
DTH8 is another key floral repressor under LD conditions (Wei et al., 2010a,b;Yan et al., 2011;Dai et al., 2012). Like Ghd7, DTH8 interacts with Hd1 (Du et al., 2017;Zhu et al., 2017). The DTH8-Hd1 complex is necessary for the floral inhibition function of Hd1 in LDs. In dth8 background, Hd1 loses its alternative function and acts as an activator of Hd3a expression to promote flowering both in SD and LD conditions in the 93-11 background. Another report suggested that DTH8 alone could not convert Hd1 function, but enhanced the effect of Ghd7 on the Hd1 functional conversion in the ZS97 background (Z. Y. . The exact mechanism by which DTH8 affects the Hd1 functional switch is not clear, but one possibility is that it acts by changing H3K27me3 levels in the Hd3a promoter (Du et al., 2017).
Several recent studies have revealed that both Ghd7 and DTH8 can interact with Hd1 and that DTH8 also interacts with Ghd7. Therefore, although Hd1 has the primary function for promoting expression of Hd3a/RFT1 and flowering regardless of day length, it can form a ternary repressive complex with DTH8 and Ghd7 to mediate photoperiodic heading Zong et al., 2021). This view is supported by the observation that a functional Ghd7, DTH8 and Hd1 combination showed significantly stronger photoperiod sensitivity than other combinations in both nearisogenic lines and natural populations (Zhang et   . For example, Hd1, DTH8 and distinct NF-YC subunits reported to form a trimeric complex that bound to the CO Response Element 2 (CORE2) in the Hd3a promoter, and Days to heading 7 (DTH7, also known as Hd2, OsPRR37 or Ghd7.1, another CCT domaincontaining protein strongly repressing flowering) can replace Hd1 in the complex . Hd1 also interacts with DTH8 and OsHAPL1 and is likely to form a complex with general transcription factors to regulate the transcription of target genes to control heading date in rice (Zhu et al., 2017). These results indicated that Hd1, Ghd7, DTH8 and DTH7 do not act independently, but assemble into complexes to regulate the rice photoperiodic pathway.
Hd1 also interacts with Rice Halotolerance Protein 3 (OsHAL3), a flavin mononucleotide-binding protein reported as a blue-light sensor that does not contain a DNA-binding domain. The interaction is inhibited by white or blue light. The OsHAL3-Hd1 complex promotes heading under SD, and activates Hd3a expression by direct binding to the Hd3a promoter. OsHAL3 expression exhibits a diurnal pattern that peaks at 4 h before light onset under SD, but peaks 4 h after light onset under LD conditions. It was proposed that the more OsHAL3 coincides with the midnight-expressed Hd1 under SD, more OsHAL3-Hd1 complex is formed to activate Hd3a expression. The level of expression of OsHAL3 under LDs peaks during daytime, and light inhibits the formation of the OsHAL3-Hd1 complex. There is no obvious induction of Hd3a mRNA levels by OsHAL3 (Su et al., 2016).
Heading date associated factor 1 (HAF1), a C3HC4 RING domain-containing E3 ubiquitin ligase, was identified as an Hd1 interacting protein in yeast-two-hybrid assays. HAF1 mediates ubiquitination and regulates circadian accumulation of Hd1 by targeting it for degradation via the 26S proteasome-dependent pathway. The haf1 hd1 double mutant flowers as late as hd1 plants only in SD conditions, suggesting that HAF1 is essential for regulating heading by modulation of the diurnal rhythm of Hd1 protein in SDs (Yang et al., 2015). Additional research revealed that HAF1 also mediates ubiquitination and modulates the diurnal rhythm of OsELF3 (an orthologue of the Arabidopsis ELF3, also known as Hd17) under LDs (Zhu et al., 2018).

Protein level regulation of key floral integrator Ehd1
Ehd1 encodes a B-type response regulator that functions upstream of florigen genes and promotes heading under both SD and LD conditions. Ehd1 integrates various upstream flowering signals and is recognised as key floral inducer in rice (Doi et al., 2004;Shrestha et al., 2014). In addition to regulation of the transcription level, post-translational modifications also play critical roles in Ehd1 function. It has been reported that the activity of Ehd1 protein in promoting Hd3a and RFT1 expression is repressed by OsPhyA in a day length-dependent manner (Osugi et al., 2011). Ehd1 activity can also be regulated by phosphorylation modification, as mutation of the key amino acid Asp-63 to Glu to mimic a constitutive phosphorylated state greatly enhanced its function and caused extremely early heading. A separate study revealed that phosphorylation of Asp-63 is required for formation of Ehd1 complexes, and dimerisation is important for normal Ehd1 function (Cho et al., 2016). Interestingly, OsRR1, a type-A response regulator, can inhibit Ehd1 activity by direct interaction with Ehd1 to form an inactive heterodimer, and delays heading (Cho et al., 2016).

Post-translational regulation of Ghd7 function
Previous studies have found that Ghd7 transcription was increased or unaltered in phyB mutants (Osugi et al., 2011;Weng et al., 2014). However, the level of Ghd7 protein was significantly decreased in a phyB mutant under LD conditions, suggesting that PhyB maintains the level of Ghd7 protein (Weng et al., 2014). More recently, one study found that constitutive expression of Ghd7 on a se5 background failed to accumulate Ghd7 protein or delay heading, suggesting that phytochromes are necessary for Ghd7 protein stability and normal function . Further in vitro and in vivo analyses revealed that phytochromes (mainly OsPHYA and OsPHYB) directly interact with Ghd7 and stabilise Ghd7 protein. OsGI also interacts with Ghd7 and mediates its degradation in a 26S proteasome-dependent manner. Interestingly, both OsPHYA and OsPHYB inhibit the interaction between OsGI and Ghd7, therefore blocking OsGI-mediated Ghd7 degradation . In addition to regulation of protein stability by phytochromes and OsGI, Ghd7 was also reported to be phosphorylated by Heading date 16 (Hd16), a casein kinase-I protein (Hori et al., 2013). Hd16 delays heading by interaction with, and phosphorylation of, Ghd7, thereby enhancing the function of Ghd7, and suppressing expression of Ehd1 and downstream florigen genes (Hori et al., 2013).

Fine-tuning photoperiod-controlled flowering in rice by phosphorylation of floral regulators
In addition to the above-mentioned phosphorylation and enhancement functions of Ghd7, Hd16 was earlier identified as EARLY FLOWERING 1 (EL1), which phosphorylates DELLA protein SLR1 (Slender rice 1, a key component in gibberellin signalling). Phosphorylation of SLR1 is important for maintaining its stability and negative regulation of gibberellin signalling (Dai & Xue, 2010). As phosphorylation of both Ghd7 and SLR1 results in suppression of heading under LD conditions, it is possible that Hd16 regulates both photoperiod and gibberellin responses by targeting different substrates in the rice flowering pathway. DTH7/OsPRR37/Ghd7.1/Hd2 is another substrate of Hd16. DTH7 directly interacts with, and is phosphorylated by, Hd16 in the middle and CCT domain-containing C-terminal regions (Choon-Tak et al., 2015). DTH7 can also be phosphorylated in the middle region by casein kinase 2a (CK2a) encoded by the Heading date 6 (Hd6) gene (Choon-Tak et al., 2015). Hd6, identified by QTL mapping, delays heading specifically under LD conditions. Other genetic studies have revealed that Hd6 delays heading only when Hd1 is functional. Extensive studies have revealed that Hd6 does not regulate Hd1 transcription, but represses expression of Hd3a and RFT1 in an Hd1-dependent manner, suggesting that Hd6 modulates Hd1 function posttranslationally. However, Hd6 does not directly phosphorylate Hd1 protein, implying the presence of an unknown Hd6 target that might function with Hd1 (Ogiso et al., 2010). It has been reported that Hd1 acts genetically downstream of DTH7 to delay flowering time in LD conditions (Lin et al., 2000), and that DTH7 is necessary for LD heading repression function for conversion of Hd1 (Z. Y. . Given the direct phosphorylation of DTH7 by Hd6 (Choon-Tak et al., 2015), DTH7 is likely to be a candidate protein that bridges the gap between Hd6 and Hd1. Further genetic and biochemical evidence is necessary to confirm this hypothesis.
Hd1 was also reported to be phosphorylated by OsK4, another protein kinase identified as Heading Date Repressor1 (HDR1) interaction protein. The HDR1-OsK4 complex acts to inhibit flowering in LDs by both transcriptional regulation of Hd1/Ehd1 expression, and probably by regulation of the phosphorylation state of Hd1 (Sun et al., 2016). Hd6 can phosphorylate OsLHY, an orthologue of the Arabidopsis circadian oscillator components LATE ELONGATED HYPOCOTYLN(LHY) (Ogiso et al., 2010), but mutation in either Hd6 or Hd16 does not affect the circadian clock (Nemoto et al., 2018). Natural variations in Hd6 and Hd16 are involved in fine tuning the critical day length in photoperiod-controlled heading and may contribute to further northward expansion of rice cultivation areas (Nemoto et al., 2018).

Domestication and genetic improvement promote expansion of the rice growing region
The ancestral species of Asian cultivated rice (Oryza rufipogon Griff.) is SD sensitive and its distribution is limited to low-latitude tropical and subtropical regions. After natural selection and artificial domestication, cultivated rice is now distributed across an extremely wide range of latitudes from 53°N to 40°S, and is grown under both LD and SD conditions (Koo et al., 2013). Weakened photoperiod sensitivity is a critical factor for adaptation of rice to high-latitude regions. Allelic variants of several heading date genes have contribute to the northward advance of rice growing in Asia. Hd1, Ghd7, DTH8, Hd16, DTH7 and PhyB are flowering repressor genes under LDs. Combined Hd1, Ghd7 and DTH8 alleles produce strong photoperiod sensitivity, and are preferably distributed to low latitudes (Zong et al., 2021). Deletion or weakened alleles of these genes are associated with early flowering under natural day length and are preferably distributed to high latitudes (Xue et al., 2008;Takahashi et al., 2009;Huang et al., 2012;Koo et al., 2013;Gao et al., 2014;Kwon et al., 2014). Combinations of weak alleles of Ghd7, DTH8 and DTH7 act additively to reduce photoperiod sensitivity, and are likely to occur in high-latitude areas such as the north-eastern provinces of China (Gao et al., 2014). Further analysis revealed that genetic variations in DTH8, Ghd7, Hd1, DTH7, PhyB and OsCOL4 are correlated with differences in heading date, and a minimum combination of four weak alleles is required for the successful cultivation of rice at latitudes above 30°N Cui et al., 2020). Moreover, flowering activation genes, such as DTH2, Ehd4 and RFT1, have also undergone intensive selection and contributed to regional adaptation of rice to higher latitudes Wu et al., 2013;Zhao et al., 2015).

Heading date genes contribute to heterosis in rice
Studies on natural variation and various genetic populations have revealed that several major heading date genes were also important heterosis genes, such as Hd3a, DTH8 and Ghd7 Li et al., 2016). The heterozygous state of Hd3a is common in the most planted hybrid rice cultivars grown in China. Among 1063 three-line hybrids studied, 98.5% of restorer lines had the ancestral type of homozygous Hd3a allele (Hd3a/Hd3a), whereas 76.7% of cytoplasmic male sterile lines were homozygous hd3a allele (hd3a/hd3a). The heterozygous state (Hd3a/hd3a) had higher grain numbers per panicle, seed setting rate and grain yield per plant than their parents . The heterozygous state of the tomato florigen SINGLE FLOWER TRUSS (orthologue of rice Hd3a) gene increased the yield in distinct genetic backgrounds and environments (Krieger et al., 2010). DTH8 is another major heading date gene that contributes to yield heterosis in hybrid rice cultivars. The heterozygous state, consisting of a strong allele and a nonfunctional allele of DTH8 (DTH8/dth8), is present in many Chinese hybrids, particularly in the two-line hybrids (40.8%), including the well known superhybrid rice Liangyou-pei 9 (LYP9) . More recently, studies have shown that the DTH8 locus has been divergently selected among New Phytologist (2021) 230: 943-956 www.newphytologist.com © 2020 The Authors New Phytologist © 2020 New Phytologist Foundation rice subpopulations, and the genetic segment carrying japonica type DTH8 was introgressed into the male parent in modern indica hybrid cultivar breeding, and contributed to heterosis with overdominance effects Lin et al., 2020). Similarly, Ghd7 has been divergently selected in the male and female parents of hybrid rice and that heterozygous state of Ghd7 consisting of a strong allele and a weak or nonfunctional allele was also found to be common in Chinese hybrid rice cultivars Lin et al., 2020).
The underlying mechanism of heterozygosity in heading date genes leading to heterosis is still not clear. One possible explanation is the pleiotropic effect of these heading date genes, for example Hd3a promotes lateral branching in addition to inducing flowering (Tsuji et al., 2015), and DTH8 and Ghd7 affect heading date, plant height and grain number per panicle simultaneously (Xue et al., 2008;. We postulate that the heterozygous genotype may help to reach a balance on proper flowering time and ideal plant architecture. Whether other heading date genes also contribute to heterosis is an interesting subject, more studies are needed to reveal the underlying mechanism.

Conclusions and perspectives
Many genes controlling heading date in rice have been identified in recent decades. Our understanding of how these genes coordinately regulate heading date remain fragmentary. Gene expression can be regulated at the transcriptional and post-transcriptional levels. In addition, there is evidence that epigenetic regulation also plays important roles in flowering time regulation. For example, many histone methylation-related proteins (such as SDG701, SDG708, SDG724, SDG725, SDG711, SDG718, SDG723, LC2/OsVIL3, OsVIL2 and OsEMF2b) regulate flowering by methylation on the lysine residues in the N-tails of histone H3 and cause either transcriptional silencing or activation of major flowering genes Sui et al., 2013;Wang et al., 2013;Yang et al., 2013;Choi et al., 2014;Liu et al., 2014;Liu et al., 2016;Liu et al., 2017;Liu et al., 2019). Se14 encodes a Jumonji C-domain-containing histone demethylase, HDT701 (also known as OsHDT1) and OsSFL1 encode components of histone deacetylase, which also strongly affect heading date Yokoo et al., 2014;Cho et al., 2018;Geng et al., 2020). Interested readers on this subject are referred to previous reviews (Shi et al., 2014;Sun et al., 2014).
Inspired by the well established flowering regulation pathways in the LD plant Arabidopsis, and through analysis of the expression of heading date genes in different rice genetic backgrounds, an evolutionary conserved OsGI-Hd1-Hd3a pathway and a ricespecific Ehd1 pathway were proposed. Ehd1 was earlier proposed to function independently of Hd1, as it can effectively promote heading in a Hd1-deficient background (Doi et al., 2004). However, some studies at the protein level have uncovered physical interactions between Hd1 and Ghd7, which form a complex that binds to the Ehd1 promoter to repress its expression (Nemoto et al., 2016), suggesting that day length responses in rice are controlled by integration of the two pathways into Ehd1 to repress heading under LD conditions.

Future perspective
Compared with the transcriptional regulation pathway in heading date, our knowledge on regulation of protein levels is still limited. Post-transcriptional regulation of flowering has been extensively studied in Arabidopsis. Although there is conservation of the flowering mechanism between rice and Arabidopsis, rice is a SD plant and it is inappropriate to directly transfer knowledge from a LD plant such as Arabidopsis to rice. Further study at the level of protein regulation of heading date in rice will be helpful to distinguish the functions and evolutionary processes in photoperiod-controlled growth of plants. Taking advantage of rapidly improving in planta protein-protein interaction detection methods, including traditional affinity purification-mass spectrometry (AP-MS) and newly developed proximity labelling methods such as TurboID-mediated proximity labelling (Branon et al., 2018), large-scale identification of interacting proteins with known heading date regulators is now possible. CRISPR (clustered regularly interspaced short palindromic repeats)-based genomic editing will also facilitate the rapid unveiling of the functions of interacting proteins. Moreover, a recently proposed initiative to tag all rice proteins in locus using a CRISPR-Cas-based targeted insertion method (named the Rice Protein Tagging Project, RPTP) should greatly facilitate functional studies on heading date and other biological processes in rice (Lu et al., 2020).
Another important limitation of the study of heading date in rice is the diverse genetic backgrounds used in different studies, by contrast with the relatively centralised background in Arabidopsis (such as the Col-0 ecotype). As functions of genes can vary in different backgrounds it is often difficult to integrate some conflicting results from multiple studies. We propose that the rice community systematically generate genetic resources in a common genetic background (such as Nipponbare for japonica and 93-11 for indica) through genomic editing, transgenesis and traditional backcrossing. Integrating different genes and building a complete heading date pathway with the support of genetic evidence in a single background will advance our understanding of the plant flowering mechanism more rapidly, especially at the protein level.

Applications in rice breeding
The breeding of rice cultivars with optimum heading dates is crucial for high yield in specific cropping areas. In traditional breeding of rice (a predominantly self-pollinated species) we select individuals with the desired heading date. However, this work becomes challenging in hybrid rice breeding, as testing heading date of every hybrid combination is labour and time consuming. This limitation is especially a problem in indica 9 japonica hybrids in which delayed heading is common. This bottleneck of unpredictable heading date limits the exploitation of substantial intersubspecies heterosis. Over past decades, many genes controlling flowering in rice have been cloned and functionally analysed, and their allelic distributions in different cropping areas have also been characterised. This should allow the rational design of cultivars with heading gene combinations best adapted to specific environmental conditions (Fig. 3). One approach is to pyramid heading date alleles in predetermined combinations by traditional cross-breeding with the assistance of molecular markers, as attempted in a previous study . The second approach is to modulate the function or expression levels of heading date genes by genetic engineering, as many genes regulate rice heading date quantitatively. This can be achieved through the downregulation of genes by RNA interference or upregulation of genes by specific promoters (Hu et al., 2019). The third approach is to modify heading date genes using gene editing technology. CRISPR-Cas genome editing enables targeted modification of specific heading date genes without affecting the genetic background, including targeted gene knockout, nucleotide base editing, insertion, deletion or replacement . Moreover, multiple targets can be changed simultaneously, making breeding of new varieties with predetermined heading date gene combinations more efficient. Although exogenous T-DNA is usually needed to deliver CRISPR-Creating heading gene combinations best adapted to specific regions

CRISPR-Cas gene editing Genetic engineering
Marker-assisted selection and combination of all desired alleles from different parents

Gene pyramiding
Introduce foreign alleles to enhance the function or change the gene expression level of key heading date genes Strong alleles frequency of LD promoting genes Strong alleles frequency of LD repression genes Fig. 3 Strategies for breeding rice cultivars with optimum heading dates. Rational design of long day (LD) repression genes (such as Ghd7 and DTH8) and LD promoting genes (such as DTH2 and Ehd4) for rice cultivars adapting to different latitudes. Three strategies can be used, including pyramiding heading date alleles by traditional cross-breeding with the assistance of molecular markers, modulating the function or expression level of heading date genes by genetic engineering, and modifying heading genes by CRISPR-Cas gene editing technology.
New Phytologist (2021) 230: 943-956 www.newphytologist.com © 2020 The Authors New Phytologist © 2020 New Phytologist Foundation Cas reagents into rice cells, these genetic elements can be easily removed in later generations to obtain transgene-free edited lines. These advantages make the CRISPR-Cas approach a fast, feasible and reliable technology in regulating heading date in rice.