Natural variation in heading-date genes enables rice, a short-day (SD) plant, to flower early under long-day (LD) conditions at high latitudes. Through analysis of heading-date quantitative trait loci (QTL) with F7 recombinant inbred lines from the cross of early heading ‘H143’ and late heading ‘Milyang23 (M23)’, we found a minor-effect EarlyHeading3 (EH3) QTL in the Hd16 region on chromosome 3. We found that Early flowering1 (EL1), encoding casein kinase I (CKI), is likely to be responsible for the EH3/Hd16 QTL, because a missense mutation occurred in the highly conserved serine/threonine kinase domain of EL1 in H143. A different missense mutation was found in the EL1 kinase domain in Koshihikari. In vitro kinase assays revealed that EL1/CKI in H143 and Koshihikari are non-functional. In F7:9 heterogeneous inbred family-near isogenic lines (HNILs), HNIL(H143) flowered 13 days earlier than HNIL(M23) in LD, but not in SD, in which EL1 mainly acts as a LD-dependent flowering repressor, down-regulating Ehd1 expression. In the world rice collection, two types of non-functional EL1 variants were found in japonica rice generally cultivated at high latitudes. These results indicate that natural variation in EL1 contributes to early heading for rice adaptation to LD in temperate and cooler regions.
One key aspect of grain productivity in cereal crops such as rice (Oryza sativa) is seasonal and regional adaptability, which can be largely achieved by adjusting the optimal timing of flowering (or heading). Common wild rice (O. rufipogon), the progenitor of cultivated rice, is a facultative short-day (SD) plant that flowers early in SD (10-h light/14-h dark) and late in long-day (LD; 14-h light/10-h dark) photoperiods. In cultivated rice, natural variation in the genes controlling photoperiod sensitivity and flowering time (or heading date) has been an important genetic resource in rice breeding programmes (Yano et al. 1997) and has played critical roles in enabling rice cultivation in a wide geographical range, from 53° N to 40° S latitudes (Lu & Chang 1980).
Flowering time in plants is controlled by complicated networks of numerous genes and is influenced by seasonal changes in photoperiod and temperature. In facultative LD dicots, such as Arabidopsis (Arabidopsis thaliana), flowering time is mainly controlled by the photoperiod, autonomous, gibberellin (GA) and vernalization pathways. In the photoperiod pathway, a series of genes associated with photoreceptors and circadian rhythm affects expression of downstream genes to regulate flowering time. In LD, the GIGANTEA (GI)-CONSTANS (CO)-FLOWERING LOCUS T (FT) pathway is a central mechanism for flowering induction (Kardailsky et al. 1999; Kobayashi et al. 1999; Park et al. 1999; Huq, Tepperman & Quail 2000). In flowering induction, FT protein expressed in the leaves relocates to the shoot apex and interacts with FD, a bZIP protein transcription factor, to induce floral meristem-identity genes (Abe et al. 2005; Corbesier et al. 2007). In rice, Hd3a and RFT1, the rice orthologs of Arabidopsis FT, induce the expression of MADS-box genes such as OsMADS14 and OsMADS15 in floral meristem (Komiya et al. 2008). Hd1, the rice ortholog of Arabidopsis CO and a major quantitative trait loci (QTL) controlling photoperiod sensitivity, promotes flowering by up-regulating Hd3a in SD and inhibits flowering by down-regulating Hd3a in LD (Yano et al. 2000; Hayama et al. 2003). In addition to the OsGI-Hd1-Hd3a pathway, Ehd1/Hd14 promotes flowering through RFT1 induction in both SD and LD (Doi et al. 2004; Komiya, Yokoi & Shimamoto 2009). Thus, Hd3a plays a critical role in promoting flowering downstream of Hd1 in SD (Kojima et al. 2002) and the OsMADS50-Ehd1-RFT1 pathway functions in flowering induction in LD (Komiya et al. 2009). In addition, Ghd7 acts as a LD-dependent suppressor of Ehd1 (Xue et al. 2008). DTH8/Ghd8/Hd5 also functions upstream of Ehd1 to inhibit flowering in LD (Wei et al. 2010; Yan et al. 2011; Fujino, Yamanouchi & Yano 2013). By contrast, Ehd2/OsID1/RID1 (Park et al. 2008; Wu et al. 2008; Matsubara et al. 2008b) and Ehd3 (Matsubara et al. 2011) up-regulate Ehd1 expression to promote flowering in LD.
It has been recently reported that in rice, Early flowering1 (EL1), encoding casein kinase I (CKI), negatively regulates GA signalling by phosphorylating a DELLA protein, SLR1 (Dai & Xue 2010). The T-DNA insertion el1 null mutant flowers earlier in day-neutral (12-h light/12-h dark) and LD (13.5-h light/9.5-h dark) conditions, and the HOX1a transcription factor, a positive regulator in GA signalling induced by GA treatment, binds to the promoter region of EL1 to down-regulate EL1 expression (Wen et al. 2011).
In rice, molecular genetic studies of photoperiod sensitivity and heading date have been conducted through fine-mapping and gene identification of heading-date QTLs. Fifteen heading-date QTLs (Hd1-Hd3a, Hd3b-Hd14) were identified from a cross between ‘Kasalath (aus-type)’ and ‘Nipponbare (japonica-type)’ (Yano et al. 2001; Monna et al. 2002), and six QTLs were identified at the molecular level; Hd1, Hd3a, Hd4/Ghd7, Hd5/DTH8/Ghd8, Hd6/CK2 and Hd14/Ehd1 (Yano et al. 2000; Takahashi et al. 2001; Kojima et al. 2002; Doi et al. 2004; Xue et al. 2008; Wei et al. 2010). Two additional QTLs, Hd16 and Hd17, were recently identified from the backcross inbred lines (BILs) derived from a cross of two japonica rice cultivars, Koshihikari and Nipponbare (Matsubara et al. 2008a). The molecular basis of the Hd17/Hd3b QTL was recently reported to be one of the two rice homologs of Arabidopsis ELF3 (OsELF3); Hd17 in Nipponbare is functional allele and acts as an activator of flowering in SD and LD by down-regulating Ghd7 expression (Matsubara et al. 2012; Saito et al. 2012; Zhao et al. 2012; Yang et al. 2013). The gene responsible for the Hd16 QTL remains to be identified.
To identify QTLs for early flowering in rice under non-inductive LD conditions, the extremely early-flowering ‘H143’ was crossed with the late-flowering ‘Milyang23 (M23)’. In the F2 population, we previously identified two major-effect QTLs, Early Heading7-1 (EH7-1) and EH7-2 in the Hd4 and Hd2 QTL regions on chromosome 7, respectively (Lin et al. 2000, 2003; Yoo et al. 2007; Xue et al. 2008; Shibaya et al. 2011). In this study, using F7 recombinant inbred lines (RILs) derived from the F2 population, we identified an additional minor-effect QTL for early heading of H143 in LD, termed Early Heading3 (EH3), closely linked to the Hd16 QTL on chromosome 3 (Matsubara et al. 2008a). Through functional analysis, we demonstrated that EL1 is likely to be the EH3/Hd16 QTL gene and suppresses LD-dependent flowering in rice by down-regulating Ehd1 expression. Further genotyping of EL1 in the world rice collection indicated the significance of natural mutations in EL1 for japonica rice cultivated under LD environments in high-latitude areas.
Materials and Methods
Plant materials and growth conditions
An extremely early heading japonica rice ‘H143’ and a middle-late heading Tongil-type rice ‘M23’ were crossed for identification of heading-date QTLs conferring extremely early heading under natural long-day (NLD) conditions in H143 (Yoo et al. 2007). The high-yield elite cultivar M23 is the japonica/indica hybrid rice which has a genetic makeup close to indica (∼90%). H143 was originally developed for cultivation in the northernmost regions of Japan, Hokkaido (43–45° N) and M23 was bred to grow in the temperate regions in South Korea (35–37° N). These rice plants were grown in the paddy field in Suwon, Korea (37° N) from April to October. The day lengths of Suwon were about 13.5 h (April), 14.2 h (May), 14.7 h (June), 14.3 h (July) and 13.6 h (August). The growth chamber conditions were as follows: 14.5-h light at 30 °C/9.5-h dark at 24 °C for LD and 10-h light at 30 °C/14-h dark at 24 °C for SD. White light-emitting diodes (LEDs) were used for rice growth and the maximum photon flux density of the LEDs was about 500 μmol m−2 s−1. Heading dates in the field and growth chamber conditions were measured just after emergence of the first panicle in rice plants.
Heading-date QTL analysis and map-based cloning of EH3
Single marker analysis using heading-date QTL-linked markers and QTL mapping of EH3 on chromosome 3 were performed with 264 F7 RILs developed from the cross of H143 and M23 (Yoo et al. 2007) (Supporting Information Table S1). Heading-date QTL-linked markers were selected from previously reported markers (Yano et al. 1997, 2001; Lin, Sasaki & Yano 1998; Yamamoto et al. 2000; Lin et al. 2002; Fujino & Sekiguchi 2005; Uga et al. 2007; Matsubara et al. 2008a) or newly developed sequence tagged site (STS) markers (Supporting Information Table S2). The linkage map of chromosome 3 was constructed with 15 simple sequence repeats (SSR) and three STS markers by MAPMAKER/EXP 3.0b (Kosambi 1943; Lander et al. 1987). Single-marker regression and composite interval mapping methods were performed using the Q-gene 4.3.8 software for QTL detection (Joehanes & Nelson 2008). Fine physical mapping of EH3 QTL was conducted using nine molecular markers and 193 F7:8 HNILs derived from an F7 RIL#61 (Supporting Information Fig. S1). Heading-date segregation of 30–40 F7:9 HNILs derived from each F7:8 HNIL was observed to determine EH3 genotypes, that is homozygous H143 or M23, or heterozygous H143/M23. DNA samples were extracted from the leaves using a modified Cetyl trimethylammonium bromide (CTAB) method (Murray & Thompson 1980). The sequence information of SSR markers that were previously developed (McCouch et al. 2002) were obtained from Gramene (http://www.gramene.org/) and STS markers were designed by comparing the genome sequences between H143 and M23 (Supporting Information Table S2).
Whole genome re-sequencing
Whole genome re-sequencing of the two parent rice cultivars, ‘M23’ and ‘H143’, was carried out using the Illumina Genome Analyzer IIx (Illumina Inc., San Diego, CA, USA). The Illumina Pipeline 1.4 software was used as a sequencing platform. Genomic DNA was extracted from the leaves of 1-month-old plants using the NucleoSpin Plant II kit (Macherey-Nagel, Duren, Germany) according to the manufacturer's manual. Full sequencing was conducted in the National Instrumentation Center for Environmental Management (NICEM) at Seoul National University, Korea.
Detection of single nucleotide polymorphisms (SNPs) and RT-PCR
Derived cleaved amplified polymorphic sequence (dCAPS) markers were designed using the dCAPS Finder 2.0 (http://helix.wustl.edu/dcaps/dcaps.html) (Neff, Turk & Kalishman 2002) and were used for SNP analysis in EL1 in cultivars of the world rice collections (Supporting Information Tables S2–S4). Genomic PCR for dCAPS analysis was performed using AccuPower HF PCR PreMix (Bioneer, Daejeon, Korea). The amplified fragments were cleaved using AluI or SpeI, and then loaded onto 3% agarose gels. For identification of different types of EL1 alleles among 24 rice cultivars (Supporting Information Table S3), the EL1 coding region was amplified by RT-PCR from each cultivar. Total RNA was isolated from the mature leaves of rice cultivars using TRIzol Reagent (Invitrogen, Carlsbad, CA, USA) and first-strand cDNAs were synthesized by M-MLV reverse transcriptase and oligo(dT)15 primer (Promega, Madison, WI, USA).
In vitro kinase assay
EL1cDNAs from H143, M23, Koshihikari and Nipponbare, and SLR1 cDNA from Nipponbare were obtained by RT-PCR using each respective total RNA sample. Each cDNA was cloned into the pGEM-T Easy vector (Promega). The pET28a vectors (Novagen, Darmstadt, Germany) containing EL1 or SLR1 cDNAs were transformed into Rosetta 2 competent Escherichia coli cells (Millipore, Darmstadt, Germany). Recombinant protein expression was induced by 0.1 mm IPTG. After 4-h culture at 28 °C, E. coli cells were collected and resuspended with the His-binding buffer (5 mm imidazole, 0.5 m NaCl, 20 mm Tris-HCl, pH 7.9). After cell sonication, the supernatant was collected and bound with the His-binding agarose resin (Elpis Biotech, Daejeon, Korea) using poly-prep chromatography columns (Bio-Rad, Richmond, CA, USA) at 4 °C. Unbound proteins were washed out with the washing buffer (40 mm imidazole, 0.5 m NaCl, 20 mm Tris-HCl, pH 7.9), and bound proteins were eluted with the elution buffer (500 mm imidazole, 0.5 m NaCl, 20 mm Tris-HCl, pH 7.9). The elution buffer was changed to the kinase buffer (40 mm Tris-HCl, pH 7.5, 2 mm DTT, 0.1 mm EGTA, 10 mm magnesium acetate) using Amicon Ultra (Millipore). Next, the recombinant EL1 and SLR1 proteins were mixed with 1 mm ATP (Sigma-Aldrich, St Louis, MO, USA) and incubated at 30 °C for 30 min. Equal amounts of protein samples were loaded into 10% SDS-PAGE and then transferred onto a polyvinylidenedifluoride (PVDF) membrane (Millipore). Anti-6X His-tag antibody (α-His; Abcam, Cambridge, MA, USA) were used to detect His(6x)-tagged EL1 (His-EL1) and His-SLR1 proteins. Antibody against the phosphothreonine (α-P Thr; Cell Signaling Technology, Beverly, MA, USA) was used to detect the phosphorylated form of recombinant His-EL1 and His-SLR1 proteins.
Reverse transcription (RT) and quantitative real-time PCR (qPCR)
Total RNA was extracted from leaves using the Total RNA Extraction Kit (Intron Biotechnology, Seongnam, Korea). First-strand cDNAs for RT were synthesized from 5 μg total RNA using oligo(dT)15 primer and M-MLV reverse transcriptase (Promega), and diluted with water to 100 μl. Relative expression levels of EL1 and flowering-related genes were measured by qPCR using gene-specific primers, and Ubiquitin5 (UBQ5) was used for an internal control (Supporting Information Table S2). The total 20 μl of mixture included: 2 μl of 0.5 μM primers, 2 μl of first-strand cDNAs and 10 μl of 2X QuantiTect SYBR Green I Master Mix (Roche, Basel, Switzerland). qPCR was performed with Light Cycler 2.0 instrument (Roche) using the following programme: 95 °C for 2 min, 45 cycles of 95 °C for 10 s, 59 °C for 10 s and 72 °C for 10 s.
Identification of the EH3 QTL gene by map-based cloning
By QTL analysis using 1365 F2 plants derived from the cross of the extremely early-heading rice ‘H143’ (78 days to heading) and the middle-late heading rice ‘M23’ (116 days to heading) under NLD conditions (37° N latitude), we previously identified two major-effect QTLs, EH7-1 and EH7-2, which are closely linked to Hd4 and Hd2 on chromosome 7, respectively (Lin et al. 2000, 2003; Yoo et al. 2007). To further identify minor-effect QTLs for early flowering of H143, 264 RILs were generated by single-seed descent from the F2 population. Using the F7 RILs, we carried out single marker analysis with the 16 heading-date QTL-linked markers that have been reported to date (Yano et al. 1997, 2001; Lin et al. 1998, 2002; Yamamoto et al. 2000; Fujino & Sekiguchi 2005; Uga et al. 2007; Matsubara et al. 2008a) (Supporting Information Table S1). As a result, we found two additional heading-date QTLs, EH3 on chromosome 3 and EH6 on chromosome 6, which are closely linked to Hd6/Hd16 (Takahashi et al. 2001; Matsubara et al. 2008a) and Hd1 regions (Yano et al. 2000), respectively. Sequence analysis of the Hd1 gene in the EH6 QTL region revealed that H143 has a functional Hd1 (type 9), and M23 has a nonfunctional hd1 allele (type 7; 4-bp deletion) as reported recently (Takahashi et al. 2009). Based on the Log-likelihood (LOD) values for each SSR marker on chromosome 3 (Fig. 1a, Table 1), EH3 appeared to be more closelylinked to Hd6, which encodes the α-subunit of casein kinase 2 (CK2α) (Takahashi et al. 2001). Genotyping of Hd6 by RT-PCR revealed that both M23 and H143 had functional Hd6 alleles (data not shown). Therefore, EH3 probably corresponds to Hd16 (Matsubara et al. 2008a) whose gene has not yet been determined.
Table 1. EH3 detected by heading-date QTL analysis under NLD conditions using the F7 RILs derived from the H143/Milyang23 cross
All genetic parameters were calculated by means of composite interval mapping (CIM) using Q-gene ver. 4.3.8.
bAdditive effect of the H143 allele on days to heading (DTH).
cPercent of phenotype variance explained by the QTL.
To identify the EH3 QTL gene by map-based cloning, we further genotyped the 264 F7 RILs using SSR markers and selected RIL#61, which harbours heterozygous H143/M23 segments in the Hd16 region. We observed that heading dates of 193 F7:8 HNILs derived from the RIL#61 segregated as expected from a heterozygote (data not shown). For positional cloning, the EH3 genotypes (i.e. homozygous H143 or M23 and heterozygous H143/M23) of F7:8 HNILs were determined by heading-date segregation of their progenies, F7:9 HNILs (see Methods). As a result, we narrowed down the EH3 QTL to 62 kb between the STS-2 and STS-3 markers (Figs 1b, S1).
For candidate gene analysis, we searched for SNPs between the two parent rice cultivars, M23 and H143, through whole genome re-sequencing (see Methods). The sequences of nine open reading frames in the 62-kb region were compared between M23 and H143, and those of Nipponbare were used as a japonica reference. Among a total of nine expressed genes in the EH3/Hd16 QTL region, three genes carried putative functional nucleotide polymorphisms, leading to amino acid substitutions in H143 (Supporting Information Table S5). Of the three candidate genes, EL1 (LOC_Os03g57940) was considered to be the strongest candidate gene for the EH3/Hd16 QTL, because the altered amino acids in two other gene products were highly variable among species (data not shown). EL1 encodes CKI and the T-DNA insertion el1 null mutant flowers early in LD (Dai & Xue 2010). By sequence analysis of the 2124-bp EL1 cDNAs, we confirmed a single nucleotide polymorphism (G-476-C) between the EL1 alleles in H143 and M23. This variation causes an amino acid (aa) change (Gly-159-Ala) in the serine/threonine kinase domain (147–441 aa) of EL1 (707 aa) in H143. The Gly-159 is highly conserved among plant CKIs (Supporting Information Fig. S2). Hd16 was previously identified by heading-date QTL analysis using the BILs generated from the cross of Nipponbare and Koshihikari, in which Hd16 (Koshihikari) BILs flowered 26 days earlier than Hd16 (Nipponbare) BILs in LD, but not in SD (Matsubara et al. 2008a). Thus, we further analysed the EL1 coding sequences of Nipponbare and Koshihikari by RT-PCR, and found that a single amino acid change (Ala-331-Thr) occurred in the EL1 kinase domain in Koshihikari, which is also highly conserved in plant CKIs (Supporting Information Fig. S2). These results suggest that missense mutations in EL1 (H143) and EL1 (Koshihikari) possibly cause a loss of CKI activity, leading to early flowering in LD as previously shown in the el1-null mutant and Hd16 (Koshihikari) BILs (Matsubara et al. 2008a; Dai & Xue 2010).
Compromised kinase activity of EL1 proteins in H143 and Koshihikari
In rice, EL1/CKI phosphorylates the DELLA protein SLR1, a key negative regulator for expression of GA-responsive genes and phosphorylation is critical for SLR1 stability and activity (Dai & Xue 2010). To examine whether the amino acid substitutions in the highly conserved EL1 kinase domain in H143 (Gly-159-Ala) and Koshihikari (Ala-331-Thr) negatively affect EL1 function, we performed in vitro kinase assays. The assays tested the recombinant His-tagged SLR1 (His-SLR1) with His-EL1 (M23), EL1 (H143), EL1 (Nipponbare) or EL1 (Koshihikari). Indeed, His-SLR1 was phosphorylated by EL1 (M23) or EL1 (Nipponbare), but not by EL1 (H143) or EL1 (Koshihikari) (Fig. 2a), indicating that the natural variants of EL1 in H143 and Koshihikari encode non-functional proteins.
Enzymatic activity of many kinases is often regulated by phosphorylation on specific residues; this is either auto-phosphorylation or is catalyzed by other kinases. It has been reported that CKIδ activity is closely associated with phosphorylation of its C-terminus (Graves & Roach 1995). Therefore, we examined EL1 for phosphorylation by looking for multiple bands on immunoblots of EL1 expressed in E. coli. When purified His-EL1 proteins were immunoblotted with anti-6X His antibody, we found two bands for EL1 (M23) and EL1 (Nipponbare), but only one band for EL1 (H143) and EL1 (Koshihikari) (Fig. 2b). Immunoblotting with an anti-phosphothreonine antibody revealed that the upper bands of EL1 (M23) and EL1 (Nipponbare) are phosphorylated EL1 proteins (Fig. 2b), suggesting that lack of phosphorylation on EL1 (H143) and EL1 (Koshihikari) is related to their compromised kinase activity. Using motif prediction programs, we searched for possible phosphorylation sites on EL1 and found four residues (serine-342, tyrosine-507, threonine-673 and serine-681) as potential phosphorylation sites (Fig. 2c). The kinase domain of EL1 includes both the serine/threonine kinase active site (272–284 aa) and ATP binding region (153–184 aa) where a single amino acid substitution occurred in EL1 (H143). The amino acid substitution (Gly-159-Ala) might cause a negative effect on ATP binding capacity, thereby affecting the kinase function of EL1. Taken together, these results strongly suggest that non-functional missense mutation in EL1 contribute to early flowering of japonica rice under NLD conditions.
EL1 acts as a suppressor of LD-dependent flowering
To examine the effect of EL1 (H143) on heading date, we selected homozygous F7:8 HNIL (M23) and HNIL (H143) lines among the progenies of F7 RIL#61. We subsequently examined the heading dates of their progenies under NLD in 2010 (HNIL7:9) and in 2011 (HNIL7:10). Forty plants of HNIL (H143) showed early heading under NLD compared with those of HNIL (M23) in 2010 (Fig 3a). In both 2010 and 2011, HNIL (H143) consistently flowered about 6–7 days earlier than HNIL (M23) under NLD conditions (Fig. 3b, 3c).
Previous heading-date QTL analysis, from the cross of Nipponbare and Koshihikari, showed that the Hd16 (Koshihikari) allele was responsible for early heading in LD, but not in SD, indicating that the Hd16 (Koshihikari) allele promotes flowering only in LD (Matsubara et al. 2008a). Thus, we examined heading dates of HNIL (H143) and HNIL (M23) in LD and SD in the growth chambers. HNIL (H143) flowered about 13 days earlier than HNIL (M23) in LD, but no significant difference of heading date between the two genotypes was observed in SD (Fig. 3d). These results suggest that the functional EL1 (M23) allele acts as a suppressor of LD-dependent flowering in rice, quite similar to the previous results of heading-date analysis with Hd16 (Nipponbare) and Hd16 (Koshihikari) genotypes (Matsubara et al. 2008a).
EL1 down-regulates Ehd1 expression in LD
To determine the regulatory function of EL1 in the suppression of LD-dependent flowering pathway, we examined the expression levels of EL1 and 10 genes known to regulate rice heading date (Fig. 4). To determine the effect of EL1 function, we looked for differences between F7:10 HNIL (M23) and HNIL (H143) under LD conditions in the growth chambers. Total RNA samples were collected at 2 h after dawn (Zeitgeber time 2; ZT2) from mature leaves at 80 days after sowing. Expression levels of EL1 were similar in both genotypes (Fig. 4a), suggesting that a lack of EL1/CKI activity does not affect EL1 expression. In addition, expression levels of OsId1, OsMADS50, OsGI, Ehd3, DTH8, Ghd7 and Hd1 were not significantly altered (Fig. 4b–4h). However, levels of Ehd1, Hd3a and RFT1 were significantly up-regulated in HNIL (H143) compared with HNIL (M23) (Fig. 4i–4k); for example, RFT1 expression was about 20-fold higher in HNIL (H143) than in HNIL (M23) (Fig. 4k). These results strongly suggest that EL1 regulates flowering time in the Ehd1-dependent pathway, but not in the Hd1-dependent pathway, in LD. Taken together, we concluded that EL1 functions upstream of Ehd1 as a LD-specific suppressor, and the up-regulation of two rice florigen genes Hd3a and RFT1 in HNIL (H143) results from the ectopic expression of Ehd1 in both LD and NLD conditions.
Natural variation in EL1 in cultivated rice
To find other genetic variants in addition to EL1 (H143) and EL1 (Koshihikari), we further examined the nucleotide sequences of the EL1 coding regions in 24 rice varieties (17 japonica and seven indica) which have been traditionally cultivated in different latitudes of East Asia (Supporting Information Table S3). As a result, EL1 alleles were classified into five distinct types, based on the amino acid changes in EL1 protein (Fig. 5a). Among the japonica varieties, 14 varieties including Nipponbare have the japonica-type functional EL1 (type 1), but H75 (Hokkaido, Japan) contains the same SNP in the EL1 kinase domain as H143 (type 2). The amino acid substitution in the EL1 kinase domain of Koshihikari (type 3) is unique (Fig. 5a, Supporting Information Table S3). Compared with the japonica-type EL1 (type 1), six indica varieties, including M23, have an indica-type functional EL1 allele in which two aa differences exist in the N-terminal region (type 4), while ‘93-11’ has a variation with one-aa difference (type 5). Supposing that the subspecies-specific amino acid variations in EL1, beyond the highly conserved serine/threonine kinase domain, do not affect the kinase activity of CKI (Fig. 2), type 5 of 93-11 appears to be a functional EL1 allele. Next, we further surveyed the distribution of two SNPs, non-functional type 2 and type 3 alleles, in 194 varieties in the world rice collection. As a result, we found six type 2 and four type 3 japonica cultivars that have been cultivated under NLD conditions in high-latitude areas, such as northern China, North Korea, Russia, Kazakhstan and Uzbekistan (Supporting Information Table S4). Statistical analysis indicated that distribution of the cultivars carrying the two non-functional EL1 variants (type 2 or 3) is closely associated with high latitudes, whereas the cultivars carrying the functional EL1 variants (type 1, 4 or 5) are randomly distributed independent of latitude (Fig. 5b,c). Together, these results suggest that natural variation in EL1 has contributed to rice adaptation to high-latitude regions.
Natural variation in heading date and photoperiod sensitivity has played pivotal roles in a wide range of regional adaptations in many crop plants. Studies of genetic variation in heading date and photoperiod sensitivity of cultivated crops have provided a new perspective on evolutionary processes and selective mechanisms in natural populations (Mitchell-Olds & Schmitt 2006). Common wild rice is adapted to low-latitude regions, but the geographical expansion of rice cultivation to high-latitude areas required the exploitation of diverse natural genetic variation in the rice genome by thousands of years of artificial selection (Izawa 2007).
CKIs are highly conserved in eukaryotes and play fundamental roles in diverse signalling processes (Gross & Anderson 1998). Rice EL1 shares roughly 80% amino acid sequence similarity with other plant CKIs, and more than 90% similarity in the EL1 kinase domains (Supporting Information Fig. S2), suggesting that CKIs have changed very little during evolution. Similarly, the amino acid sequences of Hd6/CK2α are almost completely conserved in the Oryza genus, suggesting that natural variation in the coding region of Hd6 was suppressed during evolution and speciation (Yamane et al. 2009). In this respect, it is likely that the amino acid substitutions in the highly conserved serine/threonine kinase domain of EL1 in H143 and Koshihikari compromise the enzymatic activity of CKI.
A previous study reported that the absence of EL1 activity eliminates SLR1 phosphorylation, and thereby impairs down-regulation of GA-responsive gene expression (Dai & Xue 2010). Studies on the activity of casein kinase (CK) showed that autophosphorylation of the CK2β subunit is important for its stability in mice (Zhang et al. 2002), and CKIδ activity is modulated by phosphorylation of its C-terminus in E. coli (Graves & Roach 1995). In our study, both EL1 (H143) and EL1 (Koshihikari) showed defects in phosphorylation of SLR1 by in vitro kinase assays (Fig. 2a), indicating that H143 and Koshihikari are el1 mutants. In addition, the lack of phosphorylation on EL1 (H143) and EL1 (Koshihikari) in E. coli further supports a loss of the kinase activity in H143 and Koshihikari (Fig. 2b). It has been reported that heading date is not altered in the GA signalling-defective rice mutants slr1 (Itoh et al. 2002) and Slr1-d3 (Chhun et al. 2007), indicating that enhanced GA signalling in the absence of SLR1 activity is irrelevant to early-flowering behaviour in el1 mutants under LD. In other words, in addition to SLR1 phosphorylation for suppression of GA signalling, EL1/CKI functions in the suppression of LD-dependent flowering by phosphorylating as yet unidentified target protein(s) that down-regulates Ehd1 expression in LD.
The EH3 QTL is closely linked to the Hd16 QTL on chromosome 3 (Fig. 1a). Hd16, a suppressor of LD-dependent flowering in rice, was previously identified in BILs from the cross of Nipponbare and Koshihikari (Matsubara et al. 2008a). We found that SNPs identified in EL1 of H143 (type 2) and Koshihikari (type 3) cause missense mutations in the highly conserved serine/threonine kinase domain region (Supporting Information Fig. S2). These mutations caused HNIL (H143) to flower 13 days earlier than HNIL (M23) in LD, but not in SD (Fig. 3d). This is similar to the observed earlier flowering of Hd16 (Koshihikari), which flowered 26 days earlier than Hd16 (Nipponbare) in LD (Matsubara et al. 2008a). The identification of two independent structural mutations with strong functional effects demonstrates that EL1 is the gene underlying the EH3/Hd16 QTL and that this is involved in the suppression of LD-dependent flowering.
In LD, Hd1 down-regulates Hd3a expression, while Ehd1 up-regulates Hd3a and RFT1 expression (Yano et al. 2000; Komiya et al. 2009). RFT1, rather than Hd3a, is a critical activator of LD-dependent flowering in rice (Komiya et al. 2009). Expression levels of six heading-date genes (OsId1, OsMADS50, OsGI, Ehd3, DTH8 and Ghd7), which are upstream of Ehd1, showed no difference between HNIL(M23) and HNIL(H143), and Hd1 mRNA levels were not changed (Fig. 4b–4h). Notably, the transcript levels of Ehd1, Hd3a and RFT1, were up-regulated in HNIL (H143), suggesting that EL1 down-regulates Ehd1 expression and consequently inhibits Hd3a and RFT1 expression to suppress flowering in non-inductive LD. Recently, some negative regulators of Ehd1 have been reported; for example, Ghd7 (Xue et al. 2008), DTH8/Ghd8 (Wei et al. 2010; Yan et al. 2011) and Hd1 (Wei et al. 2010; Ishikawa et al. 2011) suppress Ehd1 expression in LD. As the transcript levels of Ghd7, DTH8/Ghd8 and Hd1 are not significantly altered in HNIL (H143) in LD (Fig. 4f–4h), EL1 kinase does not directly regulate these genes to suppress Ehd1 expression at the transcriptional level. Rather, it can be considered that EL1 kinase negatively regulates Ehd1 expression by phosphorylating some of the negative regulators, or functions in a pathway independent of the negative regulators. Taking these results together, it appears that EL1 delays flowering by repressing Ehd1 expression, and the drastic increase of RFT1 expression in HNIL (H143) is largely associated with early heading in LD (Fig. 4k). Intriguingly, RFT1 expression was drastically up-regulated compared to the levels of Ehd1 and Hd3a (Fig. 4i–4k), implying that EL1 may also down-regulate RFT1 expression through an unknown pathway, if any, in addition to Ehd1-RFT1/Hd3a pathway.
Our previous study revealed that two major QTLs, EH7-1/Hd4 and EH7-2/Hd2, mainly contribute to extremely early heading of H143 in LD; this early-heading phenotype is important because H143 is cultivated in regions near the northern limit of rice agriculture, such as Hokkaido, Japan (Yoo et al. 2007). In this study, we identified an additional QTL, EH3 as a minor contributor to early flowering in H143 under NLD conditions. Nucleotide sequence analysis revealed that two amino acid substitutions, Gly-159-Ala (type 2) and Ala-331-Thr (type 3), in the serine/threonine kinase domain of EL1/CKI are natural variants in japonica rice among 24 varieties being cultivated in different latitudes (Fig. 5a, Supporting Information Table S3). Through SNP analysis of 194 varieties in the world rice collection, we found an additional six type 2 varieties, and all of these eight type 2 varieties are cultivated in the high-latitude regions of rice cultivation throughout Asia and Europe; that is Hokkaido (H75 and H143; 42–45° N), Kazakhstan (Kazahastan1; 41–53° N), Uzbekistan (UzRos-421; 37–45° N), North Korea (Chalbyeo and Yuljojo; 37–42° N) and Russia (Zeravschanica 0215 and Krasnodar 313; 43–53° N) (Fig. 5c, Supporting Information Table S3, S4). Similarly, the four type 3 varieties (Gongzhuling 8, Fengdao 12, Jingjing 7 and Fengyou 201) are cultivated in the northernmost regions of rice cultivation in China, Jilin (41–46° N) (Supporting Information Table S4). However, Koshihikari has been cultivated in the temperate region of Japan (33–37° N) (Supporting Information Table S3). Although these type 2 and type 3 varieties contain non-functional EH3/EL1 alleles, their heading dates were considerably different under NLD conditions in Suwon (37° N); that is, heading dates of H143, Fengdao 12 and Koshihikari were 78, 95 and 105, respectively (Supporting Information Tables S3, S4). As regional adaptation in rice is involved in diversity of heading date and photoperiod sensitivity through natural variations in circadian-clock or flowering-time genes (Izawa 2007; Takahashi et al. 2009), the difference in regions of cultivation between H143 and Koshihikari is likely to be associated with their different sets of alleles for key heading-date QTL genes.
Hd2 and Hd4 have been reported as major-effect QTLs for rice cultivation in the northern-limit regions, acting to decrease photoperiod sensitivity and days to heading (Fujino & Sekiguchi 2005; Yoo et al. 2007; Nonoue et al. 2008; Shibaya et al. 2011). The EH7-1/Hd4 QTL region includes a heading-date gene, Ghd7, which encodes a plant-specific protein containing a CO, CO-like, TOC1 (CCT) domain (Xue et al. 2008; Shibaya et al. 2011). A null mutation of Ghd7 (Ghd7-0) reduces grain production and plant height, and promotes flowering under NLD conditions (Xue et al. 2008). RT-PCR analysis revealed that H143 contains a non-functional japonica-type allele, Ghd7-0a (data not shown), and M23 harbours a functional indica-type allele, Ghd7-1 (Xue et al. 2008), suggesting that the EH7-1/Hd4 QTL corresponds to Ghd7. The molecular identification of the casual gene for the EH7-2/Hd2 QTL has not been reported yet, although Murakami et al. (2005) predicted that OsPRR37 is responsible for the Hd2 QTL. Thus, we interrogated the genotypes of Ghd7 allele in 13 type 2 and type 3 japonica varieties (data not shown). We found that in addition to H143, two more varieties cultivated in Hokkaido (H75) and Russia (Zeravschanica 0215) harbour a non-functional Ghd7-0a allele, and all of them flower extremely early (about 78 days to heading) in Suwon (Supporting Information Tables S3, S4). However, the other 10 varieties have functional Ghd7-2 alleles and showed a broad range of heading dates in NLD (Supporting Information Tables S3, S4). Previous studies reported that both ‘Hayamasari’ and ‘Hoshinoyume’, extremely early-flowering japonica varieties cultivated in Hokkaido, have non-functional mutations of Hd2, Hd4 and Hd6, leading to extremely early flowering under NLD conditions (Fujino & Sekiguchi 2005; Shibaya et al. 2011). Notably, Hayamasari has an additional non-functional mutation in Hd5, and thus flowers slightly earlier than Hoshinoyume in LD, but not in SD. These results indicate that the non-functional allele of Hd5 contributes to slightly earlier flowering with reduced photoperiod sensitivity in Hayamasari, when compared with the difference in heading date between Hayamasari and Hoshinoyume in Hokkaido (42–45° N) (Shibaya et al. 2011; Fujino et al. 2013), although heading dates of Hayamasari and Hoshinoyume were the same in Suwon (37° N).
In this scenario, we propose that two major-effect QTLs, non-functional mutations in Hd2 and Hd4, mainly determine extreme early heading for rice adaptation to NLD environments, and minor QTLs, such as non-functional mutations of Hd5, Hd6 and EH3/Hd16, also contribute to the expansion of rice cultivation to the northernmost regions of its ecological region. This study demonstrates that stable rice production under NLD in a short summer period at the northernmost regions (40–53° N) requires photoperiod-insensitive early flowering, which is accomplished by specific combination of loss-of-function mutations in major- and minor-effect QTL genes.
We thank Dr. ItsuroTakamure at Hokkaido University, Japan for N11and H75 seeds, Dr. MitsuakiTezuka at Hokkaido Prefectural Central Agriculture Experiment Station, Japan for Iburiwase, Kitaibuki and Hayamasari seeds, and the National Agrobiodiversity Center, National Academy of Agricultural Science, Rural Development Administration, Korea for the other rice seeds in this study. This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government (MEST) (No. 2011-0017308).