**3. Results**

### *3.1. Days to Heading and Grain Filling Rate of Baegilmi in Comparison with Koshihikari*

The DH of Baegilmi under the optimum planting (i.e., sowing and transplanting on 25 April and 25 May, respectively, in Suwon, Korea) was 82 days, which was 21 days earlier than that of Koshihikari (Figure 1a). To compare the rate of grain filling of Baegilmi and Koshihikari, the grain weight change was monitored by measuring 1000 grain weight (TGW) every 3–4 days from 19 days after heading (DAH) to 50 DAH (Figure 1b). While the TGW of Baegilmi increased faster than that of Koshihikari during the early grain filling stage, reaching 19.7 g at 26 DAH (14.5 g in Koshihikari), the TGW of Koshihikari increased faster in the late grain filling stage, reaching 23.1g at 40 DAH which was similar to Baegilmi (23.2 g). A similar pattern was observed when the changes in grain weight were monitored as the ratio of grain weight to the final grain weight according to the cumulative temperature after heading (Figure 1c). Our results indicated that Baegilmi undergoes rapid grain weight increase in the early stage of grain filling as well as early vegetative-to-reproductive transition compared to Koshihikari.

**Figure 1.** Maturity of Baegilmi in comparison with Koshihikari. (**a**) Representative phenotype of Baegilmi and Koshihikari in the field at 112 days after sowing (15 August). Days to heading was determined as the number of days from sowing to heading; (**b**) Changes in grain weight during 19–50 days after heading; (**c**) Changes in the proportion of grain weight to the final grain weight according to the cumulative temperature after heading.

### *3.2. Mapping of qDH3, qDH6, and qDH7 for Days to Heading*

The Koshihikari × Baegilmi RIL population (*n* = 142) was used to map QTLs for DH. A GBS experiment detected 893 SNPs segregating in the mapping population. After removing SNPs with low genotyping quality and overlapping genetic positions, a linkage map spanning a total length of 1293 cM was constructed with the 128 selected SNPs (Supplementary Figure S1). The average number of markers per chromosome was 10.7, ranging from six on chromosome 3 to 17 on chromosome 6. The average interval between two adjacent markers was 11.1 cM.

The Koshihikari × Baegilmi RILs showed continuous DH variation (64–105 days in 2016 and 64–107 days in 2017) with positively skewed distribution (Figure 2a,b). Baegilmi headed 19 and 18 days earlier than Koshihikari in 2016 and 2017, respectively. We detected three major QTLs for DH designated *qDH3*, *qDH6*, and *qDH7* on chromosomes 3, 6, and 7, respectively, in both 2016 and 2017 (Figure 3a, Table 1). At *qDH3*, Koshihikari contributed the allele for early heading with the additive effects of −3.1–−3.4 days and the LOD scores of 5.1–7.5, explaining 8.4–10.3% of the DH variation. On the other hand, Baegilmi contributed the allele for early heading at *qDH6* and *qDH7* (additive effects of 5.2–5.4 days and 5.0–5.3 days, respectively), which showed higher LOD scores (12.6–15.1 and 12.7–13.7, respectively) explaining higher levels of the DH variation (24.7–26.2% and 22.8–25.2%, respectively).

**Figure 2.** Frequency distribution of days to heading in the Koshihikari × Baegilmi RIL population evaluated in 2016 (**a**) and 2017 (**b**). The values of Baegilmi (B) and Koshihikari (K) are indicated above filled and unfilled triangles, respectively.

**Figure 3.** QTLs for days to heading from the Koshihikari × Baegilmi RIL population. (**a**) Mapping of *qDH3*, *qDH6*, and *qDH7* on chromosomes 3, 6, and 7, respectively. Days to heading of the RILs observed in 2016 and 2017 were used for mapping. The number after the letter 'S' in the marker name indicates the chromosome number followed by the physical position according to the IRGSP-1.0 reference; (**b**) Days to heading of the RILs with different allelic combinations of *qDH3*, *qDH6*, and *qDH7*. The markers S3\_34851991, S6\_8634012, and S7\_10453336 were used to represent *qDH3*, *qDH6*, and *qDH7*, respectively. + and − indicate the presence and absence of the allele for early heading, respectively. The genotypes of Baegilmi (B) and Koshihikari (K) are indicated by filled and unfilled triangles, respectively. Different letters above the bars indicate significant difference according to the Duncan's multiple range test at *p* < 0.05. Error bars indicate standard errors.


**Table 1.** Quantitative trait loci (QTLs) for days to heading from the Koshihikari × Baegilmi recombinant inbred line (RIL) population.

a Chromosome number. b The number after the letter 'S' in the marker name indicates the chromosome number, followed by the physical location of SNP according to the IRGSP-1.0 reference. c Logarithm of the odds. d Phenotypic variation explained (%). e Additive effect of QTL estimated at the peak position. Positive additive effect indicates that the trait value is increased by the Koshihikari allele.

To analyze the main effects of the three QTLs and their two-way and three-way interactions, 2 × 2 × 2 factorial ANOVAs were conducted for DH of the Koshihikari × Baegilmi RILs (Table 2). The main effects of *qDH3*, *qDH6*, and *qDH7* were highly significant (3.8 × 10−<sup>15</sup> < *p* < 2.6 × <sup>10</sup>−5) in both 2016 and 2017. On average, the RILs carrying the Koshihikari alleles at *qDH3* headed 5.1–5.2 days earlier than those carrying the Baegilmi alleles. On the other hand, the RILs carrying the Baegilmi alleles at *qDH6* or *qDH7* headed 9.5–9.9 or 10.7–11.4 days earlier than those with the Koshihikari alleles, respectively. The RILs with different allele combinations of *qDH3*, *qDH6*, and *qDH7* showed that accumulating two or more alleles for early heading of theses QTLs can shorten DH effectively, with the pyramiding effect of *qDH6* and *qDH7* being greater than that of *qDH3* (Figure 3b).


**Table 2.** Three-way ANOVAs of *qDH3*, *qDH6*, and *qDH7* for days to heading.

a *qDH3*, *qDH6*, and *qDH7* were represented by the markers S3\_34851991, S6\_8634012, and S7\_10453336, respectively. b Phenotypic variation explained (%) is shown only for the significant (*p* 0.05) effects. c Mean days to heading of the RILscarryingtheKoshihikari(K)alleleandtheBaegilmi (B)alleleateachofthethreeQTL.

The two-way and three-way interactions among *qDH3*, *qDH6*, and *qDH7* were marginally significant (0.02 < *p* < 0.05) and explained little phenotypic variation (1.3–1.7%) (Table 2). The *qDH3* × *qDH6* interaction was significant only in 2016, where the effect of *qDH6* accelerating heading was greater under the absence of the allele for early heading at *qDH3* (Figure 4a). Similarly, the effect of *qDH7* accelerating heading was greater under the absence of the allele for early heading at *qDH3*, and this was significant in both 2016 and 2017 (Figure 4b,c). The *qDH3* × *qDH6* × *qDH7* interaction was significant only in 2016. Under the absence of the allele for early heading at *qDH3*, the *qDH6* × *qDH7* interaction pattern was similar to the *qDH3* × *qDH6* and *qDH3* × *qDH7* interactions (Figure 4d), i.e., the effect of one QTL accelerating heading was greater under the absence of the allele for early heading at the other QTL. However, under the presence of the allele for early heading at *qDH3*, the *qDH6* × *qDH7* interaction pattern was reversed (Figure 4e), i.e., the effect of one QTL accelerating heading was greater under the presence of the allele for early heading at the other QTL.

**Figure 4.** Interactions among the three QTLs for days to heading. (**a**) *qDH3* × *qDH6* interaction in 2016; (**b**) *qDH3* × *qDH7* interaction in 2016; (**c**) *qDH3* × *qDH7* interaction in 2017; (**d**) *qDH6* × *qDH7* interaction under the absence of the allele for early heading at *qDH3* in 2016; (**e**) *qDH6* × *qDH7* under the presence of the allele for early heading at *qDH3* in 2016. Only the significant (*p* < 0.05) interactions are plotted (see Table 2).

### *3.3. Hd16, Hd1, and Ghd7 Underlying the Days to Heading QTLs*

As the three QTLs for DH encompass previously isolated heading date genes—*Hd16* (*Os03g0793500*) at *qDH3*, *Hd1* (*Os06g0275000*) at *qDH6*, and *Ghd7* (*Os07g0261200*) at *qDH7* [7,10,23]—we sequenced the coding regions of the three genes from Koshihikari and Baegilmi to search sequence polymorphisms.

It was previously shown that relative to the functional *Hd16* allele of Nipponbare, Koshihikari carries a G-to-A (alanine-to-threonine) mutation in the 10th exon of *Hd16* that is responsible for early heading under long day [23]. Sequence analysis of the 10th exon of *Hd16* revealed that the sequence of Baegilmi is identical to Nipponbare (Figure 5a). This was consistent with our QTL analysis where Koshihikari provided the allele for early heading and Baegilmi provided the allele for late heading at *qDH3* (Table 1), suggesting *Hd16* as a strong candidate gene for *qDH3*.

**Figure 5.** Sequence polymorphisms of *Hd16* (*Os03g0793500*), *Hd1* (*Os06g0275000*), and *Ghd7* (*Os07g0261200*) in Baegilmi relative to Koshihikari (**<sup>a</sup>**–**<sup>c</sup>**) and their genotyping among Korean rice cultivars (**d**–**g**). Coding region and untranslated region are depicted as filled and unfilled squares, respectively, while introns are depicted as black lines (**<sup>a</sup>**–**<sup>c</sup>**). Unfilled and filled triangles indicate SNP and insertion/deletion of Baegilmi in comparison with Koshihikari, respectively (**<sup>a</sup>**–**<sup>c</sup>**). The *Hd16* A SNP (Koshihikari allele) and the G SNP (Baegilmi allele) are visualized as intact (579 bp) and digested (393 bp + 186 bp) bands, respectively (**d**). The *Hd1* 43 bp deletion (Baegilmi allele) is visualized as a lower band (**e**). The *Ghd7* 1.9 kb insertion (Baegilmi allele) is visualized as an upper band (**f**). + and − indicate the presence and absence of the allele for early heading at each gene among 295 Korean rice cultivars, respectively, and the error bars in the bar graph indicate standard errors (**g**). Sowing and transplanting dates for optimum planting were 9 May and 1 June, respectively in 2018, and those for early planting were 10 April and 9 May, respectively in 2019, in Wanju, Korea. Different letters next to the bars indicate significant difference (*p* < 0.05) according to the Duncan's multiple range test in each year.

Sequence analysis of the *Hd1* coding region revealed that Koshihikari carries the functional *Hd1* allele identical to Nipponbare as previously reported [22], while Baegilmi carries a non-functional allele with three sequence polymorphisms relative to Koshihikari—a C-to-T SNP, a 36-bp insertion, and a 43-bp deletion in the first exon of *Hd1* (Figure 5b). The *Hd1* polymorphisms in Baegilmi are

identical to those reported in HS66 (GeneBank ID AB041841), the γ adiation mutant of the Japanese cultivar Ginbouzu [7].

While the sequence analysis found no polymorphism in the coding region of *Ghd7* between Koshihikari and Baegilmi, we identified a 1901 bp insertion at the −228 bp position from the start codon of *Ghd7* in Baegilmi (Figure 5c). The position of the 1901 bp insertion and its sequence were identical to those of the putative retrotransposon inserted in the promoter region of *Ghd7* reported in the Japanese cultivar Sorachi (GenBank ID LC472532) [24].

### *3.4. Allelic Composition of Hd16, Hd1, and Ghd7 among Commercial Rice Cultivars*

To study the allelic composition of the three DH genes in commercial rice cultivars, we screened 295 Korean rice cultivars released in 1979–2017 using the molecular markers designed to genotype the sequence polymorphisms in *Hd16* (A/G SNP), *Hd1* (43 bp Indel), and *Ghd7* (1.9 kb Indel) (Supplementary Table S1, Figure 5d–f). Of the eight (2 × 2 × 2) possible allelic combinations, five were observed among the 295 cultivars (Supplementary Table S2, Figure 5g). The majority (>90%) carried the alleles for late heading for all three polymorphisms, namely, the *Hd16* G SNP, the *Hd1* 43-bp insertion, and the *Ghd7* 1.9-kb deletion. Among 26 cultivars carrying an allele for early heading in one of three DH genes, 19 carried the *Hd1* 43-bp deletion and seven carried the *Hd16* A SNP. Only two out of the 295 Korean rice cultivars carried alleles for early heading in two of the three DH genes—Jopum carried the alleles for early heading at *Hd16* and *Hd1*, while Baegilmi carried the alleles for early heading at *Hd1* and *Ghd7*. None of the 295 cultivars carried the alleles for early heading at all three genes. The average DH of the five cultivar groups according to the allelic combinations of *Hd16* (A/G SNP), *Hd1* (43 bp Indel), and *Ghd7* (1.9 kb Indel) indicated that pyramiding the alleles for early heading at the three genes would accelerate DH effectively in the genetic background of commercial Korean rice cultivars (Figure 5g).
