QTL Analysis of Rice Grain Size Using Segregating Populations Derived from the Large Grain Line

: Grain size affects the yield and quality of rice. The large grain line (LGL), showing a large grain size and japonica -like genome, was selected in the breeding ﬁeld. The 94 F 2 plants derived from a cross between LGL and Hanareum (a high-yielding tongil-type variety) were used for the quantitative trait loci (QTL) analysis of grain length (GL), grain width (GW), and grain thickness (GT). A linkage map of the F 2 population, covering 1312 cM for all 12 chromosomes, was constructed using 123 Fluidigm SNP markers. A total of nine QTLs for the three traits were detected on chromosomes two, three, four, six, and seven. Two QTLs for GL on chromosomes two and six explained 17.3% and 16.2% of the phenotypic variation, respectively. Two QTLs were identiﬁed for GW on chromosomes two and three, and explained 24.3% and 23.5% of the phenotypic variation, respectively. The ﬁve QTLs for GT detected on chromosomes two, three, ﬁve, six and seven, explained 13.2%, 14.5%, 16.6%, 10.9%, and 10.2% of the phenotypic variation, respectively. A novel QTL for GT, qGT2 , was validated on the same region of chromosome two in the selected F 3 population. The QTLs identiﬁed in this study, and LGL, could be applied to the development of large-grain rice varieties.


Introduction
Rice is one of the most important grain crops in the world and is also the main source of calories for half of the world's population. Given the rapid growth of the global population, the world's population is predicted to reach nine billion by the middle of this century. The grain yield of rice should be increased by 70-100%, relative to the current levels, to feed the increasing global population [1,2]. The grain yield of rice is mainly affected by three components: the number of grains per panicle, the number of panicles per plant, and grain weight. Grain weight is positively associated with grain size. Thus, grain size is an important agricultural trait to improve the yield of rice. In addition, grain size affects not only yield but also quality [3]. Long and slender grains tend to be transparent and of edible quality, and thus highly prized in most rice-consuming regions globally. On the other hand, large grains show a relatively high ratio of chalky grains such as white belly or white core.
Over the past three decades, quantitative trait loci (QTL) mapping has provided a promising way to better understand the genetic regulation of yield traits in rice. More than 400 QTLs associated with rice grain shape traits have been identified, and nearly 30 genes have been cloned and demonstrated to control the traits of grain shape and size in numerous genetic studies. To date, several QTLs directly affecting grain size have been cloned in rice, such as GW2 [4], GS3 [5], qSW5/GW5 [6,7], GS5 [8], GS2 [9], GL3.1/qGL3 [10,11], GL7 [12], GW8 [13], TGW6 [14], TGW2 [15], and etc. The identification and functional characterization of these genes have provided an important theoretical basis for the enrichment of genetic resources and the development of new breeding and cultivation strategies for rice.
An adequate molecular marker system is essential for molecular breeding and genetic analysis. Single nucleotide polymorphisms (SNPs) are a more stable, abundant, fast, and cost effective variation than other DNA markers. The Fluidigm SNP genotyping system has automated polymerase chain reaction (PCR) and integrated fluidic circuit (IFC) technology, which automatically mixes PCR reagents through microfluidic channel networks. The indica-japonica SNP assays, based on the Fluidigm system developed in the previous study [16], have been applied to various genetic analyses and molecular breeding, such as bulked segregant analysis (BSA) [17], genetic diversity analysis [18,19], QTL analysis [20][21][22][23], and background profiling [24][25][26]. This Fluidigm SNP marker set has provided a faster and more cost-effective tool than other high-throughput SNP genotyping systems for primary analysis during molecular breeding using inter-subspecific populations, to date.
In this study, the large grain line (LGL), selected in the breeding field due to its large grain size, was characterized based on morphological traits and the genome-wide indicajaponica SNP set. In addition, nine QTLs associated with grain size traits were identified in an F 2 population derived from a cross between LGL and Hanareum, using the indicajaponica SNP set. Furthermore, a novel QTL controlling grain thickness was validated in the selected F 2:3 population.

Plant Materials
An LGL showing enlarged grain length, width, and thickness was selected in the breeding field of a rice lab during line selection and was maintained in the rice lab of Pusan National University. Hyowon6 (a high-quality Korean japonica variety [27]) and Hanareum (a Korean high-yielding tongil-type variety [28]) were used for phenotypic comparison with LGL. To map the QTLs, the segregating populations were derived from a cross between LGL as a maternal parent and Hanareum as a paternal parent. A total of 94 F 2 and 186 F 3 individuals with parents were grown using conventional cultivation methods at the experimental farm of Pusan National University (Miryang, Korea) in 2018 and 2020, respectively.

Phenotype Evaluation
The phenotypic measurement of three traits related to grain size was performed after harvest. Harvested grain was air-dried and kept at room temperature before measuring grain size traits. 20 randomly selected, fully filled grains from each parent, F 2:3 , and F 3:4 , were used to evaluate grain length (GL), grain width (GW), and grain thickness (GT) at 0.01 mm precision, and the values were averaged as the measurements for each plant. The three traits were evaluated using vernier calipers. Statistical analysis was conducted using R software.

DNA Extraction
At the tillering stage, young leaf tissues from each plant were collected in the field. Total genomic DNA was extracted from leaves using the modified cetyltrimethylammonium bromide (CTAB) method as described by Murray and Thompson [29]. DNA concentration and purity were quantified and qualified using a spectrophotometer (NanoDrop One, Thermo Scientific). DNA samples with absorbance ratios of 260 nm and 280 nm (A260/A280) > 1.8 were diluted to a concentration of 50 ng/µL and used for Fluidigm genotyping.

Fluidigm Genotyping
A total of 192 indica-japonica Fluidigm markers for SNP genotypes [16], designed based on the genomic difference between two rice subspecies, were used for phylogenetic analysis of LGL and detecting QTLs associated with the three grain size related traits in the F 2 population derived from a cross between LGL and Hanareum. For validation of QTLs in the F 3 population, 11 indica-japonica Fluidigm SNP markers on target regions and 6 additional Fluidigm SNP markers developed in the previous studies [20,28] were used.
Fluidigm SNP genotyping was performed using the BioMark™ HD system (Fluidigm, San Francisco, CA, USA), 96.96 Dynamic Array IFCs for the F 2 population, and 192.24 Dynamic Array IFCs (Fluidigm) for the F 3 population, according to the manufacturer's instructions, at the National Instrumentation Center for Environmental Management (NICEM), Seoul National University (Pyeongchang, Korea). Fluidigm SNP Genotyping Analysis software was used to acquire genotyping results. All base calls were manually checked, and any errors in homozygous or heterozygous clusters were corrected before further analysis.

Phylogenetic Analysis
A total of 40 rice germplasms, including LGL, Hanareum, Hyowon6, and other germplasms genotyped using 190 indica-japonica SNP markers in a previous study [16], were used for phylogenetic analysis (Table S1). A total of 38 germplasms, not including LGL and Hyowon6, were selected and clearly classified into the specific subgroups of the previous study. A PowerMarker V3.25 [30] was used to calculate the genetic distance based on the CS chord [31] and to construct an unweighted pair group method with an arithmetic mean algorithm (UPGMA), which were visualized in Molecular Evolutionary Genetics Analysis version 7.0 (MEGA7 [32]).

Linkage Map Construction and QTL Analysis
The QTL IciMapping 4.1 software [33] was used to construct the linkage map and QTL analyses. First, the BIN functionality (binning of redundant markers) was used to remove redundant markers. Markers showing more than 10% missing data were also removed. The output file obtained from the binning step was used for linkage map construction with the MAP functionality. The Kosambi mapping function was used to calculate genetic distances in centimorgans (cM) [34]. QTL mapping was carried out using the BIP functionality (QTL mapping in biparental populations). The inclusive composite interval mapping of additive (ICIM-ADD) QTL method with default option was used to detect additive QTLs. A significant logarithm of the odds (LOD) threshold value was calculated for each QTL using 1000 permutations at p = 0.05, and the LOD threshold for each QTL was ranged from 3.51 to 3.56.

Characterization of LGL
LGL was selected and designated due to its large grain size. LGL showed a wider and longer grain size than Hyowon6 with a round type grain, and Hanareum with a slender type grain. In addition, the panicle and flag leaf size of LGL were also longer and thicker than the other two varieties ( Figure 1). We focused on the grain size of LGL and evaluated three grain size traits: GL, GW, and GT. The average of GLs for Hyowon6, Hanareum, and, LGL were 6.58 mm, 8.02 mm, and 8.15 mm, respectively. The GL of LGL was significantly longer than Hanareum, which has a long grain. For GW, Hyowon6, Hanareum, and LGM presented 3.20 mm, 2.95 mm, and 3.66 mm, respectively. The GW of LGL showed the largest value by a significant amount. The GTs of Hyowon6, Hanareum, and LGL were 2.12 mm, 2.00 mm, and 2.23 mm, respectively. Even for GT, LGL showed to be significantly thicker than the other two varieties ( Figure 2). This result showed that LGL has a significantly large grain compared to normal varieties.

Figure 2.
Phenotypic comparison of GL, GT, and GW. Duncan's least significant ranges (LSR) was used to identify significance. Significant differences at 5% level were presented by different letters such as a, b, and c.
LGL was an unstudied line with an unknown pedigree. To identify the genomic relationship of LGL with other varieties, phylogenetic analysis based on 190 indica-japonica We focused on the grain size of LGL and evaluated three grain size traits: GL, GW, and GT. The average of GLs for Hyowon6, Hanareum, and, LGL were 6.58 mm, 8.02 mm, and 8.15 mm, respectively. The GL of LGL was significantly longer than Hanareum, which has a long grain. For GW, Hyowon6, Hanareum, and LGM presented 3.20 mm, 2.95 mm, and 3.66 mm, respectively. The GW of LGL showed the largest value by a significant amount. The GTs of Hyowon6, Hanareum, and LGL were 2.12 mm, 2.00 mm, and 2.23 mm, respectively. Even for GT, LGL showed to be significantly thicker than the other two varieties ( Figure 2). This result showed that LGL has a significantly large grain compared to normal varieties. We focused on the grain size of LGL and evaluated three grain size traits: GL, GW, and GT. The average of GLs for Hyowon6, Hanareum, and, LGL were 6.58 mm, 8.02 mm, and 8.15 mm, respectively. The GL of LGL was significantly longer than Hanareum, which has a long grain. For GW, Hyowon6, Hanareum, and LGM presented 3.20 mm, 2.95 mm, and 3.66 mm, respectively. The GW of LGL showed the largest value by a significant amount. The GTs of Hyowon6, Hanareum, and LGL were 2.12 mm, 2.00 mm, and 2.23 mm, respectively. Even for GT, LGL showed to be significantly thicker than the other two varieties ( Figure 2). This result showed that LGL has a significantly large grain compared to normal varieties.

Figure 2.
Phenotypic comparison of GL, GT, and GW. Duncan's least significant ranges (LSR) was used to identify significance. Significant differences at 5% level were presented by different letters such as a, b, and c.
LGL was an unstudied line with an unknown pedigree. To identify the genomic relationship of LGL with other varieties, phylogenetic analysis based on 190 indica-japonica Figure 2. Phenotypic comparison of GL, GT, and GW. Duncan's least significant ranges (LSR) was used to identify significance. Significant differences at 5% level were presented by different letters such as a, b, and c.
LGL was an unstudied line with an unknown pedigree. To identify the genomic relationship of LGL with other varieties, phylogenetic analysis based on 190 indica-japonica SNPs was conducted using a total of 40 germplasms including LGL, Hanareum, and Hyowon6. A total of 40 germplasms were divided into five subgroups, such as indica, aus, aromatic, tropical japonica, and temperate japonica, mirroring previous results [16]. LGL and Hyowon6 were clustered with temperate japonica, but LGL presented the largest genetic distance from Nipponbare, which is used as a reference variety for phylogenetic analysis, in the temperate japonica cluster (Figrue 3A).
LGL showed some indica-like introgression blocks (more than three continuous SNPs showing homozygous alleles different from Nipponbare) on chromosomes 1, 8, 9, and 12. This introgression pattern was unique to LGL ( Figure 3B,C). This suggests that LGL is an untypical japonica-type line possessing a japonica-like genome with some indica-like introgression segments. SNPs was conducted using a total of 40 germplasms including LGL, Hanareum, and Hy-owon6. A total of 40 germplasms were divided into five subgroups, such as indica, aus, aromatic, tropical japonica, and temperate japonica, mirroring previous results [16].
LGL and Hyowon6 were clustered with temperate japonica, but LGL presented the largest genetic distance from Nipponbare, which is used as a reference variety for phylogenetic analysis, in the temperate japonica cluster (Figrue 3A).
LGL showed some indica-like introgression blocks (more than three continuous SNPs showing homozygous alleles different from Nipponbare) on chromosomes 1, 8, 9, and 12. This introgression pattern was unique to LGL ( Figure 3B,C). This suggests that LGL is an untypical japonica-type line possessing a japonica-like genome with some indica-like introgression segments.

Phenotypic Variation in the F2 Population
GL, GW, and GT were evaluated in two parental lines and the F2 population derived from a cross between LGL and Hanareum (Table 1 and Figure 4). All traits showed continuous distribution and a value less than 1.0 for skewness value, and this implies that all traits present approximately normal distribution. In addition, GL and GT showed bidirectional transgressive segregation in the F2 population.

Phenotypic Variation in the F 2 Population
GL, GW, and GT were evaluated in two parental lines and the F 2 population derived from a cross between LGL and Hanareum (Table 1 and Figure 4). All traits showed continuous distribution and a value less than 1.0 for skewness value, and this implies that all traits present approximately normal distribution. In addition, GL and GT showed bidirectional transgressive segregation in the F 2 population.   LGL had a similar but significantly longer GL than Hanareum. The range and CV of GL for the F2 population were 6.89 mm to 8.92 mm and 4.26%, respectively. For the GW, LGL was significantly wider than Hanareum, and the F2 population had a range of 2.79 mm to 3.62 mm and a CV of 5.46%. Finally, GT was also significantly thicker in LGL than Hanareum and ranged from 1.85 mm to 2.37 mm in the F2 population, with a CV of 4.75%.
Correlation analysis was carried out to identify the relationship between the three traits in the F2 population. All pairs presented positive correlation. The highest and most significant positive correlation (0.586) was detected between GW and GT. The pair between GL and GT showed the lowest, and a nonsignificant, positive correlation ( Table 2). This implies that the three traits, especially GW and GT, could be regulated by the same and/or tightly linked QTLs.

Linkage Map Construction in the F2 Population
The F2 population, consisting of 94 individuals, was genotyped using 192 Fluidigm indica-japonica SNP markers. A linkage map of the F2 population was constructed using 123 polymorphic and clearly genotyped SNP markers. The linkage map covered all 12 rice chromosomes with at least four markers for each chromosome. Several large genetic gaps showing larger than 30 cM intervals between adjacent markers were identified on eight chromosomes, not including chromosomes 2, 3, 4, and 10. The total length of the linkage LGL had a similar but significantly longer GL than Hanareum. The range and CV of GL for the F 2 population were 6.89 mm to 8.92 mm and 4.26%, respectively. For the GW, LGL was significantly wider than Hanareum, and the F 2 population had a range of 2.79 mm to 3.62 mm and a CV of 5.46%. Finally, GT was also significantly thicker in LGL than Hanareum and ranged from 1.85 mm to 2.37 mm in the F 2 population, with a CV of 4.75%.
Correlation analysis was carried out to identify the relationship between the three traits in the F 2 population. All pairs presented positive correlation. The highest and most significant positive correlation (0.586) was detected between GW and GT. The pair between GL and GT showed the lowest, and a nonsignificant, positive correlation ( Table 2). This implies that the three traits, especially GW and GT, could be regulated by the same and/or tightly linked QTLs.

Linkage Map Construction in the F 2 Population
The F 2 population, consisting of 94 individuals, was genotyped using 192 Fluidigm indica-japonica SNP markers. A linkage map of the F 2 population was constructed using 123 polymorphic and clearly genotyped SNP markers. The linkage map covered all 12 rice chromosomes with at least four markers for each chromosome. Several large genetic gaps showing larger than 30 cM intervals between adjacent markers were identified on eight chromosomes, not including chromosomes 2, 3, 4, and 10. The total length of the linkage map was 1318 cM and the average genetic distance between two adjacent markers was 10.7 cM (Figure 4).

QTL Analysis in the F 2 Population
To map the genomic regions controlling the large grain size of LGL, QTL analysis for three traits-GL, GW, and GT-was carried out using the phenotype and genotype data of the F 2 population derived from a cross between LGL and Hanareum. A total of nine QTLs were detected for the three traits on chromosomes two, three, five, six, and seven ( Figure 5 and Table 3). Furthermore, we compared the genomic position of the QTLs detected in this study and previously reported QTLs/genes for grain size, based on the physical position of flanking markers for each QTL. map was 1318 cM and the average genetic distance between two adjacent markers was 10.7 cM (Figure 4).

QTL Analysis in the F2 Population
To map the genomic regions controlling the large grain size of LGL, QTL analysis for three traits-GL, GW, and GT-was carried out using the phenotype and genotype data of the F2 population derived from a cross between LGL and Hanareum. A total of nine QTLs were detected for the three traits on chromosomes two, three, five, six, and seven ( Figure 5 and Table 3). Furthermore, we compared the genomic position of the QTLs detected in this study and previously reported QTLs/genes for grain size, based on the physical position of flanking markers for each QTL.   [39,40] Abbreviations are as follows: LOD = logarithm of the odds; PVE = phenotypic variation explained by each QTL; Add = additive effect of LGL allele; N/A = not available.

QTLs for GL
For GL, two QTLs were detected on chromosomes two and six, and these QTLs explained 17.26% and 16.15% of the phenotypic variation, respectively. These QTLs were responsible for 33.41% of the total phenotypic variation in GL. Alleles of LGL on qGL2 and qGL6 extended GL by as much as 0.2 mm and 0.28 mm, respectively. An atypical bHLH protein, which is involved in GL regulation and encoded by the PGL2 gene, was reported in the genomic region of qGL2 [35]. Furthermore, qGL6 was overlapped with gw-6, a QTL for grain weight detected from a DH population derived from the inter-subspecific cross [36].

QTLs for GW
Two QTLs for GW were mapped on chromosomes two and three. qGW2 and qGW3 showed 24.27% and 23.53% of PVE, respectively, and LGL alleles on these two QTLs expanded GL by as much as 0.12 mm and 0.11 mm, respectively. GS2 [9], a rare allele of this gene that enhances grain size including GW, was reported in the genomic region of qGW2. qGW3 was overlapped with gw3.1, a QTL for grain weight detected in advanced backcross populations derived from crosses between O. rufipogon and O. sativa.  [39,40] Abbreviations are as follows: LOD = logarithm of the odds; PVE = phenotypic variation explained by each QTL; Add = additive effect of LGL allele; N/A = not available.

QTLs for GT
A total of five QTLs for GT were identified on five different chromosomes. The PVE for each QTL ranged from 10.21% to 16.63%. Most of the LGL alleles on these QTLs increased GT. However, the LGL allele of qGT5 showed a negative effect on GT. These QTLs explained a total of 55.43% of the phenotypic variation for GT in the F 2 population. The pleiotropic QTL affecting grain size traits, including GT, GL3.1/qGL3, was cloned to OsPPKL1 in the genomic region of qGT3 [10,11]. qGT5 was overlapped with the QTLs for GT region, detected in the doubled haploid population from sake-brewing rice [38]. Furthermore, qGT6 was located on the marker interval identical to qGW6, and overlapped with the previously reported grain weight QTL gw-6 [36]. The grain size and shape regulating gene SRS1/DEP2 was previously identified in the genomic region of qGT7 [39,40]. There has been no QTL/gene related to GT found in the genomic region of qGT2, to date.

qGT2 Validation Using Selected F 3 Population
For the validation of qGT2, which is a novel QTL detected in this study, one F 2 individual possessing heterozygous alleles on qGT2 and qGT7 and Hanareum homozygous alleles on the other three QTLs was selected to develop the F 3 population ( Figure 6A). The distribution of GT in the F 3 population consisting of 186 individuals showed normal distribution with a range of 1.97 mm to 2.36 mm. The variation of GT in the F 3 population was less than the F 2 populations ( Figure 6B). A total of 17 Fluidigm SNP markers, including six additional markers on chromosome two, were used to construct linkage maps of the segregating target regions of the F 3 population ( Figure 6C).

Discussion
Large grain size directly affects grain yield and quality, and is often considered a priority trait in specific rice breeding programs such as sake brewing [38,41]. LGL was selected according to its large grain size, but its panicle and leaf length phenotypes are also enlarged. Thus, LGL could be applied to develop whole crop silage rice. There are several reported QTLs/genes showing pleiotropic effects on grain size and plant architecture in the QTL regions identified in this study. For instance, GS2 [9] and SRS1/DEP2 [39], involved in not only grain size but also panicle architecture by regulating cell size and cell number, were located on qGW2 and qGW7.
LGL showed positive alleles on these QTLs, and it is considered that a large grain, panicle, and flag leaf could be contributed by these The QTL for GT in the F 3 population was detected on the marker interval identical to qGT2 and explained 10.52% of the phenotypic variance. In addition, LOD graphs of the two segregating populations presented similar shapes, although several markers were added to that region ( Figure 6C,D). This suggests that the QTL detected in the F 3 population is identical to qGT2 and stable to the environment and genetic background.
However, no QTL for GT was identified on the qGT7 region in the F 3 population. This implies that qGT7 could interact with other genomic regions and/or be affected by external conditions in order to express.

Discussion
Large grain size directly affects grain yield and quality, and is often considered a priority trait in specific rice breeding programs such as sake brewing [38,41].
LGL was selected according to its large grain size, but its panicle and leaf length phenotypes are also enlarged. Thus, LGL could be applied to develop whole crop silage rice. There are several reported QTLs/genes showing pleiotropic effects on grain size and plant architecture in the QTL regions identified in this study. For instance, GS2 [9] and SRS1/DEP2 [39], involved in not only grain size but also panicle architecture by regulating cell size and cell number, were located on qGW2 and qGW7.
LGL showed positive alleles on these QTLs, and it is considered that a large grain, panicle, and flag leaf could be contributed by these genes. However, we did not collect other phenotypes in the F 2 population and further genetic analysis accompanying detailed and repeated phenotype collection using an immortal population is required to elucidate this point. Thus, we have developed RILs derived from the F 2 population for the following study.
All the QTLs detected in this study showed the same direction, except qGT5, and at least two QTLs have clustered in similar positions on chromosomes two, three, and six ( Figure 5, Table 3). This result could explain the correlation analysis that showed all positive correlation among the three traits. Especially, both QTLs for GW were located in positions less than 10 cM from the two QTLs for GT and GW, with GT showing the highest correlation coefficient by a significant amount (Table 2). This suggests that GW could be used as a selection trait for the large grain size of LGL.
Particularly, a novel QTL for GT, qGT2, was detected in this study. To validate this, one F 2 individual containing heterozygous qGT2 was selected, and the F 3 population was derived from this individual. A QTL for GT in the F 3 population was identified on the same marker interval as qGT2. This result clearly indicated that this interval includes genes responsible for GT. A total of 360 genes were located in the genomic region of qGT2 (Table S2). To clone the genes responsible for qGT2, the fine mapping of qGT2 was performed using advanced segregating populations.
Genome-wide SNP marker sets have supported efficient phylogenetic analysis and polymorphic marker selection in this study. However, there were some genomic regions that markers did not cover due to a lack of polymorphism between LGL and Hanareum ( Figures 3C and 4). This phenomenon was caused by the dissimilarity between LGL, Hanareum and the typical japonica and indica. A similar problem occurred using an RILs population derived from TR22183, a Chinese japonica possessing some indica-like introgressions ( Figure 3B), and dasanbyeo, a Korean tongil-type variety containing some japonica introgressions [42]. To avoid this problem, identification of the genomic differences between parental lines based on whole-genome resequencing, and the development of additional markers, are required. Despite an incomplete linkage map of the F 2 population, the total PVE for each of the three grain size traits was more than 37% (Table 3). This implies that the Fluidigm indica-japonica SNP sets efficiently detected the QTLs for grain size in LGL.
Numerous QTLs related to grain size have been reported in rice, thus, identifying the novelty and validation of the detected QTLs is required before further study. In this study, we found an unstudied line showing large grain traits and carried out a QTL analysis of grain size traits. A relatively small F 2 population was used for the primary QTL analysis, and the selected F 3 population, which was segregated in the target QTL region, was used to validate the novel QTL detected in the primary analysis. This strategy could provide an efficient tool to profile the QTLs of target traits in unstudied germplasms.

Conclusions
The novel large grain line and QTLs related to grain size identified in this study could be applied to breeding programs developing large grain rice. A novel genetic locus for GT was detected and validated in this study, laying the foundation for further fine mapping and positional cloning to discover the mechanism of grain size regulation.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/agriculture11060565/s1, Table S1: Llist of 40 rice germplasms used for phylogenetic analysis; Table S2: List of genes located in the genomic region of qGT2.