Series of advanced backcross populations uncovers diversity and complexity in genetic control of grain length and grain width in Asian cultivated rice

Abstract

The use of genetic variations in Asian cultivated rice is essential for crop improvement. Advanced technologies and accumulated genome information support this effort to ensure sustainable production under changing environmental conditions. Grain length and grain width are important agronomic traits in rice. Extensive studies have revealed their complex genetic control. To further clarify the genetic basis of grain length and grain width in rice, we constructed a comprehensive catalog of QTLs for these traits based on QTL analysis of advanced backcross populations from crosses between a japonica rice cultivar ‘Koshihikari’ and 12 diverse donor cultivars. It comprised 469 QTLs (average 39.1 QTLs per donor) and provided estimates of allelic effects within a uniform japonica genetic background. To validate and delimit QTLs, we used sub-CSSLs—a set of plant materials with one or no segregating regions from these donors in a homogeneous genetic background. New QTLs were validated by genetic mapping in three chromosomal regions previously unreported in other mapping populations. Our results underscore the diversity and complexity of genetic control of grain length and grain width in Asian cultivated rice, and these insights will facilitate more precise genetic improvement of rice grain traits.

Introduction

Asian cultivated rice (Oryza sativa L.) exhibits remarkable genetic diversity, accumulated throughout its history of domestication and cultivation (Long et al. 2024, Vaughan et al. 2008). This genetic diversity is crucial for exploring and characterizing beneficial agronomic traits to improve rice for future food security (Leung et al. 2015, Nourollah 2016). The accumulation of large-scale varietal genome data combined with next-generation sequencing (NGS)-based approaches such as genome-wide association studies (GWAS) and quantitative trait locus sequencing (QTL-seq) has greatly facilitated the discovery of advantageous alleles (Han and Huang 2013, Ma et al. 2025, Takagi et al. 2013, Wang et al. 2018, Yang et al. 2025, Yano et al. 2016). These technological advancements are expected to support the prediction of agronomic performance and the selection of genotypes to meet future demands (Bartholomé et al. 2022, Yang et al. 2024).

Agronomic traits in rice are regulated by multiple QTLs with varying effect sizes (Wang et al. 2020, Yamamoto et al. 2009, Yano and Sasaki 1997). To date, 904 gene loci have been curated as agronomically important in the Rice Annotation Project Database (RAP-DB) (Kawahara et al. 2013, Sakai et al. 2013). Genome-wide DNA variation data from 685 cultivars are also available through the TASUKE+ platform (Kumagai et al. 2019). However, the number of genes needed to accurately predict complex traits likely exceeds those currently identified. Genomic prediction studies show that increasing the number of genetic markers to several thousand improves prediction accuracy (Bartholomé et al. 2022, e Sousa et al. 2019, Iwata et al. 2015). To address this challenge, expanding experimental datasets that capture allelic variation in causal genes is essential. Such data, validated through plant materials, will enhance predictive models for rice breeding.

The use of plant materials for genetic mapping has greatly advanced our understanding of the genetic control of agronomic traits in rice, including the identification of causal genes (Fukuoka et al. 2010, Lu et al. 2015). Sophisticated populations such as nested association mapping and multi-parent advanced generation inter-cross (MAGIC) populations enable high-resolution allele detection from multiple parental cultivars owing to short linkage disequilibrium (Bandillo et al. 2013, Fragoso et al. 2017, Nawade et al. 2024, Ogawa et al. 2018). MAGIC populations broaden trait variation through diverse allelic combinations but may introduce noise due to their heterogeneous genetic backgrounds, limiting their utility for analyzing loci with small effects. In contrast, advanced backcross populations such as chromosome segment substitution lines (CSSLs), which contain one to a few donor segments in a recurrent genetic background, are well suited to detecting minor-effect genes (Yamamoto et al. 2009). CSSLs with a japonica background have been developed using donors representing the genetic diversity of Asian cultivated rice (Nagata et al. 2023). Selecting plant materials appropriate to research objectives can optimize genetic analysis outcomes.

Grain shape is a key agronomic trait closely linked to grain productivity and varietal discrimination owing to its high phenotypic diversity. It is a complex trait governed by multiple genes, with over 200 QTLs identified and more than 90 genes cloned or molecularly characterized (Fang et al. 2025, Gao et al. 2021, Huang et al. 2013, Jiang et al. 2024, Li et al. 2019, 2024a, 2024b, Liu et al. 2025, Luo et al. 2025, Ma et al. 2025, Tang et al. 2023, Xin et al. 2024, Xuedan et al. 2023, Yang et al. 2025, Zhang et al. 2024, Zheng et al. 2024). RapMap, which uses mapping populations from cultivars with subtle phenotypic differences, identified novel grain shape alleles at GS3 and GL1 (Zhang et al. 2021). Advanced backcross populations covering the donor genome detected 65 QTLs for grain length (GL) and grain width (GW) in a japonica × indica cross (Nagata et al. 2015), while multiple linear regression analyses of 12 CSSL sets revealed over 400 putative QTLs (Nagata et al. 2023). These findings highlight the genetic complexity of GL and GW and address the need for systematic research to clarify QTL distribution and allelic diversity, which are essential for precise genetic control of GL and GW in rice breeding programs.

To advance our understanding of the genetic control of GL and GW in rice, we developed a catalog of QTLs for GL and GW based on analyses of BC_3–5F₂ advanced backcross populations derived from crosses between japonica ‘Koshihikari’ and 12 genetically diverse donor cultivars. To validate and delimit newly identified QTLs, we selected sub-CSSLs—plant materials containing one or no segregating donor regions within a homogeneous genetic background. Using these sub-CSSLs and their progeny lines, we successfully validated and precisely delimited three QTLs located in chromosomal regions where GL and GW QTLs have not been previously reported in other cross combinations.

Materials and Methods

Plant materials

We sowed seeds of 398 backcrossed populations (BC_3–5F₂) derived from crosses between the recurrent parent ‘Koshihikari’ and each of 12 donors (‘LAC23’, ‘Owarihatamochi’, ‘Bei Khe’, ‘Tupa 121-3’, ‘Khao Nam Jen’, ‘Khau Mac Kho’, ‘Deng Pao Zhai’, ‘Naba’, ‘Bleiyo’, ‘Qiu Zhao Zong’, ‘Muha’, ‘Basilanon’), 10 of which were selected from the World Rice Core Collection (Kojima et al. 2005) to represent the genetic diversity of Asian cultivated rice (Fig. 1). These populations derived from individual donors, each comprising 1272 to 1668 plants, provided genome-wide coverage of the respective donor genomes. The seeds were raised in previous studies (Abe et al. 2013, Mizuno et al. 2018, Nagata et al. 2023). We selected BC_3–5F₂ plants with fewer and smaller segregating chromosomal regions than the corresponding chromosomal regions in the respective CSSLs. BC_3–5F₃ seeds were used as research materials designated as sub-CSSLs to validate and delimit the QTLs of interest, which also included BC₄F₃ seeds derived from ‘IR64’ as a donor parent (Nagata et al. 2015) (Fig. 1). We selected 14 sub-CSSLs to validate 3 new QTLs. These QTLs were chosen based on the following criteria: (1) donor alleles that significantly increased trait values in the recurrent genetic background, and (2) QTL alleles present in multiple cultivars, making them potentially useful for diverse breeding programs. The selected lines were derived from crosses between ‘Koshihikari’ and ‘IR64’, ‘Deng Pao Zhai’, or ‘Bei Khe’. The first QTL, qGW1-1, is located at ~0.3 Mb on chr. 1, where the ‘IR64’ allele increases GW (Nagata et al. 2015). The second QTL was located at ~3.1 Mb on chr. 7, where the ‘Deng Pao Zhai’ allele increased GW. The third QTL was found at ~5.0 Mb on chr. 12, where the ‘Bei Khe’ allele increased GL. We chose these QTLs because no genes related to GL and GW have previously been reported in these regions. To further delimit qGW1-1, we used 6 BC₄F₅ progeny lines derived from sub-CSSLs carrying the ‘IR64’ donor segment with recombination events near the qGW1-1 region.

Fig. 1.

Plant materials used in this study. (A) Seeds of recurrent and donor parents, named on the left. White bar, 10 mm. (B) Schematic illustration of the development of the plant materials used. Backcrossed populations were derived from indicated studies. CSSL, chromosome segment substitution line; MAS, marker-assisted selection; QTL, quantitative trait locus.

We used 44 or 45 plants in the BC_3–5F₂generation (population‑dependent), 45 or 88 plants in BC₄F₃, and 16 plants in BC₄F₅ for genotyping and trait measurements. All plants were grown in an experimental field at NARO, Tsukuba, Japan (36.03°N, 140.11°E). Basal fertilizer was applied at 56 kg/ha N, 56 kg/ha P, and 56 kg/ha K. Month-old seedlings were transplanted in mid May at one seedling per hill in plots with a double row for each line, at 18 cm between plants and 36 cm between rows. Rice grains were harvested in September or October and were air-dried before analysis.

Measurement of GL and GW

We obtained data for GL and GW as described previously (Nagata et al. 2023). We scanned ~200 grains per plant at 600 dpi on an image scanner and measured the GL and GW in SmartGrain grain shape analysis software (Tanabata et al. 2012).

DNA extraction and marker analysis

Fresh leaves were harvested from field-grown plants, and total DNA was extracted from 1- to 3-cm sections by crushing in 30 μL 0.5 M NaOH and diluting in 120 μL of 1 M Tris·HCl (pH 8.0). We used 932 simple sequence repeat (SSR) markers and a single-nucleotide polymorphism (SNP) marker for GS3 (IDGS3_001_3) to detect QTLs in the BC_3–5F₂ populations. The DNA samples of BC_3–5F₂ plants were obtained from previous studies (Abe et al. 2013, Mizuno et al. 2018, Nagata et al. 2015, 2023), and genotypes were determined at mean intervals of 2.7 Mb per marker (Supplemental Table 1). We added 3 SSR markers, 3 insertion–deletion markers, and 22 SNP markers to the region around the QTLs on chr. 1 to determine the position in ‘IR64’, on chr. 7 in ‘Deng Pao Zhai’, and on chr. 12 in ‘Bei Khe’ (Supplemental Tables 1, 2).

Statistical analysis

We constructed a genetic map in MAPMAKER/EXP v. 3.0 software (Lander et al. 1987). QTL analysis for GL and GW was performed in QTL Cartographer v. 2.5 software (Basten et al. 2005), and the threshold was obtained by using 1000 permutations at α = 0.05 for each population. Two-tailed t-tests were conducted to evaluate differences between the genotypes in each BC₄F₅ population.

Cluster analysis was performed using Ward’s method, based on the Euclidean distance using additive effect values, to examine the relationships between cultivars, taking into account the magnitude of these values, in the R package “dendextend” (Galili 2015, R Core Team 2023). The values were standardized by subtracting the mean from each input value and dividing the difference by the standard deviation. It was also performed using binary data indicating the presence or absence of QTL alleles for GL and GW that differ from those of the recurrent cultivar. QTL alleles within a chromosome interval whose additive effect values were in opposite directions were treated as alleles at independent loci. An example was when QTLs with opposite effects were detected in populations segregating only for the upstream or downstream portion of the same chromosomal interval. For the respective QTL alleles in ‘Koshihikari’, a value of 0 was used.

Results

Variation of GL and GW in BC_3–5F₂ populations

As expected, the means of both GL and GW of 398 backcrossed populations were close to those of the recurrent parent (Table 1, Supplemental Fig. 1). However, the range of values within the backcrossed populations varied widely as a proportion of the parental values, from 63% to 483% in GL and 64% to 464% in GW (Table 1).

Table 1.Ranges and mean values of grain length and grain width in 12 sets of backcrossed populations and their parental lines

Code	Donor			Recurrent	Number of populations	Year	Grain length (mm)						Grain width (mm)
	Cultivar namea	Origin	Subspecies	Cultivar name			Donor	Recurrent	Backcrossed population				Donor	Recurrent	Backcrossed population
							Av. ± SD	Av. ± SD	Av. ± SD	Max	Min	Range (%)b	Av. ± SD	Av. ± SD	Av. ± SD	Max	Min	Range (%)b
LAK	LAC23 (LA)	Liberia	japonica	Koshihikari (KO)	31	2012	9.06 ± 0.10	7.21 ± 0.08	7.21 ± 0.20	8.34	6.75	1.59 (86)	3.30 ± 0.03	3.41 ± 0.06	3.41 ± 0.07	3.62	3.11	0.51 (464)
OWK	Owarihatamochi (OW)	Japan	japonica	–	29	–	8.27 ± 0.14	7.12 ± 0.06	7.16 ± 0.11	7.50	6.75	0.75 (63)	3.60 ± 0.04	3.41 ± 0.04	3.39 ± 0.05	3.56	3.22	0.34 (179)
BKK	Bei Khe (BK)*	Cambodia	indica	–	39	2013	8.17 ± 0.12	6.91 ± 0.12	6.96 ± 0.18	7.56	6.37	1.19 (94)	2.65 ± 0.02	3.37 ± 0.06	3.33 ± 0.11	3.61	2.86	0.75 (104)
TUK	Tupa 121-3 (TU)*	Bangladesh	indica	–	34	–	8.08 ± 0.12	7.01 ± 0.08	7.06 ± 0.21	7.82	6.46	1.36 (127)	2.70 ± 0.04	3.40 ± 0.04	3.38 ± 0.11	3.64	2.94	0.70 (100)
KNK	Khao Nam Jen (KN)*	Laos	japonica	–	33	–	7.95 ± 0.14	6.97 ± 0.19	6.98 ± 0.15	7.99	6.98	1.01 (103)	3.97 ± 0.06	3.40 ± 0.10	3.42 ± 0.06	3.72	3.25	0.47 (82)
KMK	Khau Mac Kho (KM)*	Vietnam	japonica	–	35	–	8.05 ± 0.10	6.92 ± 0.08	6.98 ± 0.15	7.84	6.55	1.29 (114)	3.98 ± 0.05	3.37 ± 0.04	3.39 ± 0.07	3.71	3.16	0.55 (90)
DPK	Deng Pao Zhai (DP)*	China	indica	–	29	2014	7.86 ± 0.07	6.93 ± 0.07	7.04 ± 0.16	7.59	6.54	1.05 (113)	3.08 ± 0.02	3.39 ± 0.03	3.34 ± 0.11	3.64	2.85	0.79 (255)
NAK	Naba (NA)*	India	indica	–	32	–	8.16 ± 0.08	6.95 ± 0.08	7.05 ± 0.16	7.63	6.16	1.47 (119)	2.77 ± 0.02	3.40 ± 0.03	3.34 ± 0.11	3.62	2.89	0.73 (116)
BLK	Bleiyo (BL)*	Thailand	indica	–	31	–	8.97 ± 0.18	6.96 ± 0.06	7.05 ± 0.20	8.27	6.57	1.70 (85)	2.51 ± 0.05	3.39 ± 0.03	3.36 ± 0.08	3.58	2.98	0.60 (64)
QZK	Qiu Zhao Zong (QZ)*	China	indica	–	33	–	8.15 ± 0.17	6.99 ± 0.07	7.05 ± 0.15	7.69	6.59	1.10 (95)	2.93 ± 0.04	3.38 ± 0.02	3.38 ± 0.10	3.61	2.87	0.74 (164)
MUK	Muha (MU)*	India	indica	–	36	2015	8.67 ± 0.09	7.01 ± 0.09	7.11 ± 0.20	7.93	6.48	1.45 (87)	2.95 ± 0.04	3.42 ± 0.05	3.40 ± 0.09	3.67	3.01	0.66 (140)
BAK	Basilanon (BA)*	Philippines	indica	–	36	–	7.19 ± 0.13	6.89 ± 0.10	7.02 ± 0.17	7.68	6.23	1.45 (483)	2.48 ± 0.03	3.36 ± 0.07	3.33 ± 0.10	3.59	2.99	0.60 (68)
					CV(%)		6.1	1.3	1.0	3.5	3.5		17.2	0.5	1.0	1.3	4.7

^a Cultivar names marked with asterisks were selected from the World Rice Core Collection (WRC) maintained by the NARO Genebank. Abbreviations for the cultivar names are indicated in parentheses.

^b Values in parentheses indicate the range as a proportion of that of the respective parent.

QTLs for GL in BC_3–5F₂ populations

QTL analysis for GL in 398 backcrossed populations identified 232 QTLs, with 13 to 27 QTLs per donor, in 43 intervals (Table 2, Supplemental Tables 3–14). The LOD threshold (α = 0.05) for GL ranged from a minimum of 1.5 to a maximum of 8.5, with a mean of 2.4 (SD = 0.7). Notably, five populations segregating chromosomal regions around GS3—two in KNK, two in LAK, and one in MUK—showed thresholds more than twice the overall average. The average absolute value of the additive effect was 0.094 (Table 2). Among them, 7 alleles had absolute additive effects of >0.20 mm (3.0% of the total) and 25 had effects of >0.15 mm (10.8%) (Table 2, Fig. 2A). Japonica donors had more alleles with effects of >0.2 mm (0.8 per donor) than indica donors (0.5), despite indica contributing more alleles overall (Table 2). These alleles were found in QTLs on chr. 3, including the GS3 region (Table 2, Supplemental Tables 6, 8, 9, 11, 13, 14). QTLs with effects of >0.15 mm were located in 15 regions across 7 chromosomes where genes for GL and GW have been reported, and were found in populations from 9 donors (Table 2). QTLs were detected in both indica and japonica donors in 33 of the 43 intervals (Table 2). The remaining intervals had QTLs only from indica donors (Table 2). Estimated GL, calculated from cumulative additive effects across the 12 donors, correlated with measured length (R² = 0.31).

Table 2.A QTL catalogue for grain length in rice, showing the additive effect of QTLs detected in 398 BC_3–5F₂ populations derived from the cross between 12 donor cultivars and the recurrent parent, ‘Koshihikari’

Chr.	Chromosome interval for QTL peak (Mb)		Additive effects (mm)c													Cloned or molecularly characterized genes assigned to chromosomal intervals and adjacent regionse
		QTL nameb	IRKb	BKK	DPK	NAK	BLK	QZK	TUK	MUK	BAK	LAK	OWK	KNK	KMK
			indica	indica	indica	indica	indica	indica	indica	indica	indica	japonica	japonica	japonica	japonica
1	1.67–7.57a				0.090	–0.075						–0.095	–0.096	–0.147	–0.163	SERL5, GSE1.2, D2
													–0.058
1	12.49–15.12	qGL1-1	0.102				0.052	0.074		0.059						miR408
1	19.12–24.74				0.053			0.061					–0.084
1	26.91–34.51a	qGL1-3	–0.117	–0.091	–0.103	–0.116	–0.144	–0.137	–0.154	–0.094	–0.166	–0.112	–0.105		–0.152	BG3, D61
				0.125
1	36.47–41.11	qGL1-4	0.116	0.070	0.116	0.123	0.088	0.140			0.092		0.089		0.057	OsMYB73, ARF4, CKX4, qGSE1.2
2	0.12–4.41	qGL2-1	0.076	0.083					0.091	0.097	0.147	0.119	0.090	0.077	0.103	RGG2, SPL4
2	10.19–14.59							–0.054	–0.069	–0.062		–0.098	–0.061	0.063		miR529a
2	24.02–31.20a	qGL2-3	0.080			–0.035		–0.130					–0.051			GS2
						0.096
2	33.08–35.26	qGL2-4	–0.023		0.032										–0.085	SMG1, qTGW2, miR396f
3	0.12–3.29			–0.100	–0.080	–0.095	–0.229	–0.153	–0.121	–0.193		–0.048	–0.039			OsMADS47, VQ13, LG3
3	5.53–8.55								–0.091			–0.060	–0.027	–0.110		LGY3, GSW3.1, HDR3
3	14.51–18.41a	qGL3-2	–0.304				–0.392			0.064		–0.387	n.a.d	–0.412	–0.317	GS3, SMG3, IGL1
										–0.307
3	25.11–28.78	qGL3-3	–0.102	–0.145	–0.122			–0.061	–0.257		–0.090	–0.049	n.a.d		–0.056	brd1, RGG1, MIS2, GL3.1, RGB1, GF14f, DDM1b, GIF1
3	32.44–36.16			–0.076		0.061					–0.094	–0.015	–0.093	–0.154		GSA1, OsGL3.6, TGW3
4	2.03					0.057						0.080	0.100
4	6.94–13.88			0.100						0.103		0.113	0.091		0.117
4	19.9–23.84					–0.115		–0.097	–0.092	–0.136		–0.084	–0.061	–0.104		An-1, GIF1, D11
4	29.89–34.89				–0.081		–0.099					0.050	0.072			SMG2, CYP704A3, XIAO, qGL4, AGO2, miR396e
5	0.17–5.79	qGL5-1	–0.159	–0.128	–0.104	–0.121		–0.160	–0.119	–0.113	–0.143			0.065		MKP1, GW5, GSK2
5	13.43–19.00					–0.071	–0.089	–0.057			–0.182		0.060			D1, WRKY53, SMOS1
5	23.91–29.61	qGL5-3	–0.092	0.042	–0.061	0.043	–0.078		–0.078		–0.054				–0.078	GL5, qGSN5, qGL5.2, PUP7
6	0.21–2.27	qGL6-1	0.093	0.100	0.030	0.104		0.054	0.114		0.080		–0.055			MAPK6, DA1
6	8.75–13.89				–0.058	0.058	–0.055	0.084		0.172	0.106	–0.051	–0.061		0.057	BZIP47, GW6
6	19.70–24.63	qGL6-3	0.069					0.063			–0.126					TGW6
6	25.83–30.97	qGL6-4	–0.096	–0.089	–0.045	–0.067	–0.121	–0.102	–0.101	–0.157		–0.040	–0.085		–0.171	GW6a, SGD1, GL6, FD2
7	0.65–7.69				0.055	–0.065	–0.043							0.068
7	12.71–19.36	qGL7	–0.105	–0.168	–0.111		–0.084	0.072	–0.066		–0.120	–0.084			–0.067	GLW7
7	21.61–23.65					–0.089		–0.080		–0.121				0.059		GW7/GL7, GE
7	26.46–29.04								0.171	0.168	0.186			–0.093	–0.103	SDR7-6
8	0.13–7.49	qGL8-1	–0.078		–0.070	–0.074				0.059			0.050		–0.086
8	21.58–28.25a			–0.088	–0.143		–0.162	–0.112	–0.052	–0.055						GGC2, GAD1, GW8, WTG1, SLG
										–0.075
9	1.55–6.59	qGL9-1	0.051	0.050					0.086	–0.029				0.038		GSE9
9	9.21–16.94			0.060	–0.051		–0.051	–0.064		–0.066		0.106				DEP1, GS9
9	17.89–20.55	qGL9-2	0.072		0.109			0.122		0.050	–0.102
10	2.74									0.020
10	10.63–11.74									–0.051			–0.035			GS10
10	17.47–23.11a		–0.088	–0.055	0.109	0.076	0.079		0.058	–0.103	0.098	0.065		0.073	0.118	GW10, GL10
		qGL10	0.116
11	3.83–8.11						–0.078			–0.127						RCN1, Cyclin-T1;3, GL11, SRS5
11	14.53–20.14			–0.069						0.082	–0.047	–0.105			0.134
11	21.52–28.96			–0.119					0.115	0.047		0.105				OsGIF1
12	3.38–5.04	qGL12	–0.058	–0.124		–0.045					–0.053				–0.074
12	9.10–12.24				–0.094				–0.166							qGW12
12	19.96–24.01			–0.079				0.087		0.086						OsGIF2, GL12, miR167a
Number of QTLs detected				21	21	20	16	21	18	27	17	20	21	13	17
							indica		japonica		total
	Total QTLs detected in 12 donors
		Number of QTLs (QTLs per donor)					161	(20.1)	71	(17.8)	232	(18.8)
		Mean absolute value of additive effects					0.095		0.094		0.094
	QTLs whose absolute value of additive effect >0.2
		Number of QTLs (QTLs per donor)					4	(0.5)	3	(0.8)	7	(0.6)
		Mean absolute value of additive effects					0.296		0.372		0.329
	QTLs whose absolute value of additive effect >0.15
		Number of QTLs (QTLs per donor)					18	(2.3)	7	(1.8)	25	(2.2)
		Mean absolute value of additive effects					0.197		0.251		0.212

^a A chromosomal interval that may contain multiple QTLs.

^b Data are taken from Nagata et al. (2015). One QTL allele in italics, located in the region 17.47–23.11 Mb on chr. 10, was not assigned in that paper.

^c Additive effects of the ‘Koshihikari’ allele on grain length: absolute values >0.2 mm and >0.15 mm. For QTLs with additive effects in the same direction detected within an interval across multiple populations from a donor, the value from the population with the largest phenotypic variation explained by the QTL was used.

^d The region from 14.28 to 27.89 Mb on chromosome 3 in OWK was not analyzed owing to a loss of the donor’s chromosome in an early backcross generation.

^e Includes genes reported in the review articles by Li et al. (2019) and Xuedan et al. (2023), as well as in research articles cited in this paper. Gene symbols were selected from RAP-DB synonyms.

Fig. 2.

Numbers of QTLs detected in BC_3–5F₂ populations and their additive effects. A total of 398 BC_3–5F₂ progeny were derived from crosses between ‘Koshihikari’ and 12 donor parents. Bars indicate ‘Koshihikari’ alleles that ■ increased or □ decreased (A) GL and (B) GW relative to donor alleles.

QTLs for GW in BC_3–5F₂ populations

QTL analysis for GW in 398 backcrossed populations identified 237 QTLs, with 15 to 26 QTLs per donor, assigned to 49 intervals (Table 3, Supplemental Tables 3–14). The LOD threshold (α = 0.05) for GW ranged from a minimum of 1.5 to a maximum of 4.3, with a mean of 2.3 (SD = 0.4). The average absolute value of the additive effect was 0.048 (Table 3). Among these, 9 alleles had absolute additive effects of >0.15 mm (3.9% of the total) and 24 had effects of >0.075 mm (10.4%) (Table 3, Fig. 2B). Indica donors contributed more alleles than japonica donors, both in total and in those with effects of >0.15 mm (Table 3). These alleles were found in the 0.45–5.88-Mb region on chr. 5, where GS5 and qGW5 are located, in populations from 8 indica and 1 japonica donors (Table 3, Supplemental Tables 3–11). QTLs with effects of >0.075 mm were detected in 13 regions across 7 chromosomes, all but 1 of which correspond to previously reported genes related to GL and GW, in populations from 11 donors (Table 3). QTLs were detected in both indica and japonica donors in 33 of the 49 intervals (Table 3). In the remaining intervals, QTLs were found only in indica donors in 13 intervals and only in japonica donors in 3 intervals (Table 3). Estimated GW, calculated from cumulative additive effects across the 12 donor populations, had a strong correlation with measured width (R² = 0.72).

Table 3.A QTL catalogue for grain width in rice, showing the additive effects of QTLs detected in 398 BC_3–5F₂ populations derived from the crosses between 12 donor cultivars and the recurrent parent, ‘Koshihikari’

Chr.	Chromosome interval for QTL peak (Mb)		Additive effects (mm)c													Cloned or molecularly characterized genes assigned to chromosomal intervals and adjacent regionse
		QTL nameb	IRKb	BKK	DPK	NAK	BLK	QZK	TUK	MUK	BAK	LAK	OWK	KNK	KMK
			indica	indica	indica	indica	indica	indica	indica	indica	indica	japonica	japonica	japonica	japonica
1	0.30–4.97	qGW1-1	–0.055	–0.040	–0.034	–0.047		–0.039	–0.053	–0.058	–0.025	–0.035		–0.086	–0.110	SERL5, GSE1.2, GW5L, D2
1	7.28–9.94												–0.053		–0.042
1	19.12–20.94						–0.042			0.041				0.050
1	25.07–29.90	qGW1-3	0.026	0.035		0.019		–0.019	0.032				0.014			UBC13, BG3, D61
1	30.09–34.69	qGW1-4	–0.015	0.044			–0.043	0.034		–0.065	0.029	0.033
1	40.21–42.93		0.046	0.044			0.023					–0.035		–0.043		OsMYB73, ARF4, CKX4
2	0.12–3.83	qGW2-1	0.064	0.060		0.042			0.052		0.040			0.025		RGG2, SPL4
2	7.48–10.19	qGW2-2	0.026		0.106			0.060				0.034			0.039	FUWA, GW2, UBP15, SLG2
2	11.23–15.19		0.062					0.067	0.051	0.067		0.036				miR529a
2	18.46–22.80			0.058	0.091	0.071	0.055	0.025			0.075		0.030			WG1, SGW5
2	25.83–27.93	qGW2-3	0.063			0.064										GS2
2	33.08–35.77				0.050	0.083	0.020	0.029	0.020	0.067						TGW2, SMG1, qTGW2, miR396f
3	0.12–1.16	qGW3-1	–0.043	–0.040			–0.035	–0.035	–0.026				–0.015	–0.036		VQ13
3	5.86–10.62		0.028	0.027		0.075	0.065	0.062	0.064		0.044	0.037	0.047	–0.045		OsMADS47, PUP1, LGY3, HDR3, GSW3.1
3	16.73–22.40	qGW3-3	0.011	0.061	0.077		0.034				0.061		n.a.d	–0.041	0.028	GS3, SMG3, brd1, RGG1, MIS2
3	27.43–28.78						–0.060	–0.082	–0.047			–0.035	–0.016		–0.054	GL3.1, RGB1, GF14f, DDM1b, GIF1
3	32.44–36.35	qGW3-4	0.044	0.066	0.041					0.038	0.068	0.004	0.032			GSA1, TGW3
4	0.06–4.04					–0.019	–0.032	–0.024	–0.032				–0.029	–0.032
4	13.08					–0.027
4	17.10–24.25	qGW4-2	–0.022	–0.032		0.022	0.025	0.022	–0.027	0.057		0.050	0.030		0.043	An-1, GIF1, D11
4	28.76–34.89				–0.069	–0.023	–0.035	–0.061	0.037				0.020			SMG2, CYP704A3, qGL4, miR396e
5	0.45–5.88	qGW5-1	0.150	0.208	0.178	0.175	0.164	0.204	0.197	0.160	0.161	0.160	0.028	–0.059		MKP1, GS5, qGW5, GSK2
5	20.06–22.11			–0.030											–0.040	WRKY53, SMOS1
5	25.15–29.54	qGW5-2	0.019		0.035		0.049		–0.022		0.051			–0.065		qGSN5, qGL5.2, PUP7, WOX9C
6	0.21–3.17a	qGW6-1	0.026	0.076		–0.035		–0.033		–0.021						MAPK6, DA1, HGW
		qGW6-2	–0.041
6	7.62–12.45a	qGW6-3	0.027			–0.035		–0.037	0.056	–0.016	0.077				0.027	BZIP47, GW6
								0.029		0.057
6	16.09–22.30	qGW6-5	0.037	0.063		0.015	0.050
6	25.83–30.97	qGW6-7	–0.039	–0.041	–0.046	–0.037	–0.037	–0.042	–0.017		–0.038		–0.034		–0.037	GW6a, SGD1, FD2
7	3.14–4.14				–0.036	–0.057	–0.026	–0.031
7	9.08–13.44	qGW7-1	–0.052	–0.062							–0.023		–0.033		–0.052
7	17.47–19.36			–0.039	–0.047		–0.034		–0.042		–0.060	–0.040				GLW7
7	23.65–29.04								0.039	0.064	0.056			–0.035	–0.064	GW7/GL7, GE, SDR7-6, WG7
8	0.13–2.96			–0.045		0.015										SG2
8	6.16–7.57													–0.033	–0.055
8	17.82–22.17	qGW8	–0.035	–0.070	–0.050	–0.065	–0.048	–0.064			–0.081
8	24.63–28.24						0.036		0.112	0.099	0.056					GW8, WTG1, SLG
9	1.55–3.91						–0.027	–0.028		–0.040				–0.011		GSE9
9	10.47–16.24							–0.023	–0.033			0.042	–0.033			GS9
9	17.89–22.58a	qGW9-1	0.037	0.023		–0.017			–0.010	0.034		–0.048	–0.047	–0.036	–0.040
										–0.045
10	3.66				0.014
10	7.94–10.63							0.027					0.032			GS10
10	17.47–23.11	qGW10-2	0.026	–0.025	–0.102				0.043	0.017	0.047	0.041		0.023		GW10, GW10.2
11	0.22–3.83	qGW11-1	0.028	0.034	0.027			0.034		–0.034		0.035		0.025		RCN1
11	17.49–20.14									–0.041		–0.030
11	23.47–26.03	qGW11-2	0.043					–0.032	0.032		–0.049
12	0.40–0.91			–0.046			–0.023
12	11.34														0.023	qGW12
12	19.12–24.42	qGW12-3	–0.050			0.040	0.045	0.048		0.046			0.046		0.047	qTGW12a
12	27.34				0.023	0.039			0.053
Number of QTLs detected				24	17	22	23	26	23	20	18	16	17	16	15
							indica		japonica		total
	Total QTLs detected in 12 donors
		Number of QTLs (QTLs per donor)					173	(21.6)	64	(16.0)	237	(19.8)
		Mean absolute value of additive effects					0.050		0.040		0.048
	QTLs whose absolute value of additive effect >0.15
		Number of QTLs (QTLs per donor)					8	(1.0)	1	(0.3)	9	(0.8)
		Mean absolute value of additive effects					0.181		0.160		0.179
	QTLs whose absolute value of additive effect >0.075
		Number of QTLs (QTLs per donor)					21	(2.6)	3	(0.8)	24	(2.0)
		Mean absolute value of additive effects					0.123		0.119		0.123

^a A chromosomal interval that may contain multiple QTLs.

^b Data are taken from Nagata et al. (2015). Two QTL alleles in italic in regions 11.23–15.19 Mb on chr. 2 and 5.86–10.62 Mb on chr. 3 were not assigned in that paper.

^c Additive effects of the ‘Koshihikari’ allele on grain width: absolute values >0.15 mm and >0.075 mm. For QTLs with additive effects in the same direction detected within an interval across multiple populations from a donor, the value from the population with the largest phenotypic variation explained by the QTL was used.

^d The region from 14.28 to 27.89 Mb on chromosome 3 in OWK was not analyzed owing to loss of the donor’s chromosome in an early backcross generation.

^e Includes genes reported in the review articles by Li et al. (2019) and Xuedan et al. (2023), as well as in research articles cited in this paper. Gene symbols were selected from RAP-DB synonyms.

Genetic relationship among donors depicted by GL and GW QTLs

In total, we identified 469 QTLs related to GL and GW (232 for GL and 237 for GW) across 12 populations. Of these, 318 (68%) overlapped with regions of cloned genes when overlap was defined as the region within 3 Mb (±1.5 Mb) around each QTL’s LOD peak. Based on these QTLs, dendrograms constructed using additive effect values for GL or GW QTLs divided donors and recurrent cultivars into two groups based on the presence or absence of QTLs with the largest effects—specifically, QTLs for GL in the GS3 region on chr. 3 and for GW in the GS5/qGW5 region on chr. 5 (Fig. 3A, 3B). To assess the similarity in the presence or absence of QTL alleles, we constructed dendrograms using binary data of QTLs (Fig. 3C, 3D). The dendrogram for GL divided the cultivars into 2 groups: one consisting solely of indica cultivars, and the other consisting of japonica cultivars and the remaining indica cultivars (Fig. 3C). Similarly, the dendrogram constructed using binary data of QTLs for GW divided donors and recurrent cultivars into 2 groups corresponding to indica and japonica cultivars, although 2 indica cultivars (‘Tupa 121-3’ and ‘Muha’) were grouped with japonica (Fig. 3D).

Fig. 3.

Relationships among rice cultivars classified by GL and GW QTLs. Dendrograms were constructed using (A, B) additive effect values or (C, D) binary data of QTLs for (A, C) GL and (B, D) GW. + Cultivars carrying alleles that alter ‘Koshihikari’ phenotypes at GS3 or GS5/qGW5. Cultivar codes are shown on the right (see Table 1); red, indica; blue, japonica.

Selection of sub-CSSLs as research materials for dissecting agricultural traits in rice

From the BC_3–5F₂ populations, we selected ~28% of plants with reduced heterozygosity (average 3.0%) compared to the original population (6.2%). These plants, with segregating regions mostly confined to a single chromosome, were used to establish sub-CSSLs for subsequent QTL validation (Table 4).

Table 4.Total number of lines and number of lines per chromosome in each set of sub-CSSLs derived from 13 donor cultivars.

The additional donor set corresponds to ‘IR64’-derived sub-CSSLs that were included specifically for QTL validation and delimitation analyses, whereas the genome-wide QTL catalog was constructed using 12 donor cultivars

Code	Total	Chr. 1	Chr. 2	Chr. 3	Chr. 4	Chr. 5	Chr. 6	Chr. 7	Chr. 8	Chr. 9	Chr. 10	Chr. 11	Chr. 12	Hetero. (%)b
IRK	463 (11.0)a	91 (30.3)	33 (8.3)	43 (14.3)	43 (14.3)	39 (13.0)	35 (7.0)	41 (8.2)	23 (7.7)	27 (9.0)	38 (9.5)	21 (10.5)	29 (7.3)	4.6
LAK	501 (10.9)	73 (10.4)	53 (10.6)	65 (10.8)	40 (13.3)	46 (11.5)	54 (13.5)	22 (11.0)	26 (13.0)	25 (8.3)	27 (13.5)	41 (10.3)	29 (7.3)	2.9
OWK	527 (12.0)	69 (17.3)	40 (8.0)	35 (17.5)	91 (30.3)	41 (13.7)	50 (12.5)	28 (5.6)	22 (5.5)	47 (15.7)	42 (14.0)	33 (6.6)	29 (9.7)	2.7
BKK	400 (9.3)	45 (9.0)	40 (20.0)	41 (13.7)	22 (7.3)	28 (9.3)	58 (19.3)	50 (12.5)	15 (7.5)	29 (7.3)	26 (5.2)	29 (7.3)	17 (3.4)	2.6
TUK	512 (10.7)	35 (5.8)	66 (16.5)	74 (18.5)	44 (7.3)	60 (20.0)	25 (8.3)	47 (15.7)	24 (4.8)	41 (10.3)	29 (9.7)	30 (6.0)	37 (18.5)	2.7
KNK	429 (10.7)	59 (9.8)	24 (8.0)	57 (19.0)	36 (12.0)	37 (12.3)	34 (11.3)	35 (8.8)	22 (7.3)	32 (16.0)	26 (13.0)	33 (11.0)	34 (6.8)	2.6
KMK	416 (9.0)	51 (12.8)	56 (11.2)	47 (9.4)	15 (3.8)	21 (7.0)	25 (12.5)	53 (13.3)	43 (14.3)	13 (3.3)	28 (7.0)	34 (8.5)	30 (7.5)	2.4
DPK	317 (8.8)	35 (8.8)	44 (22.0)	35 (17.5)	15 (5.0)	21 (7.0)	40 (8.0)	32 (10.7)	18 (6.0)	14 (3.5)	21 (10.5)	23 (11.5)	19 (6.3)	3.3
BLK	277 (6.8)	31 (7.8)	21 (3.5)	34 (17.0)	31 (10.3)	29 (5.8)	19 (6.3)	18 (4.5)	24 (8.0)	16 (8.0)	15 (7.5)	24 (8.0)	15 (3.8)	2.9
NAK	359 (8.8)	42 (14.0)	38 (12.7)	19 (4.8)	27 (9.0)	36 (9.0)	28 (5.6)	23 (11.5)	39 (9.8)	34 (8.5)	27 (9.0)	30 (15.0)	16 (4.0)	3.1
QZK	376 (8.7)	61 (15.3)	24 (4.0)	54 (18.0)	44 (11.0)	21 (5.3)	22 (5.5)	29 (7.3)	15 (5.0)	20 (10.0)	27 (9.0)	30 (10.0)	29 (9.7)	3.1
MUK	423 (10.1)	43 (8.6)	23 (5.8)	35 (7.0)	49 (16.3)	29 (9.7)	29 (7.3)	42 (14.0)	44 (11.0)	26 (13.0)	28 (14.0)	55 (18.3)	20 (5.0)	2.4
BAK	463 (12.9)	52 (13.0)	54 (27.0)	55 (27.5)	11 (3.7)	58 (19.3)	45 (9.0)	39 (13.0)	17 (5.7)	27 (6.8)	18 (9.0)	51 (25.5)	36 (12.0)	3.2
Mean	420 (10.0)	53 (11.6)	40 (10.1)	46 (13.5)	36 (10.6)	36 (10.6)	36 (9.3)	35 (10.0)	26 (7.9)	27 (8.6)	27 (9.5)	33 (10.3)	26 (7.1)	3.0

^a Values in parentheses indicate the proportion relative to those in the corresponding CSSLs previously reported by Abe et al. (2013), Nagata et al. (2015), Mizuno et al. (2018), and Nagata et al. (2023).

^b Mean heterozygosity that reflect the average proportion of heterozygous loci per line within each sub-CSSL set.

The number of lines representing each chromosome was generally proportional to the length of the chromosome (R² = 0.84, Table 4). However, the number of lines selected for each chromosome varied by donor, with certain chromosomes having more (e.g., IRK_chr. 1, OWK_chr. 4) or fewer (e.g., BAK_chr. 4, KMK_chr. 9) lines than the average (Table 4). These differences were associated with factors such as the number of recombinants obtained or the absence of seeds due to sterility.

Validation of GL and GW QTLs by using sub-CSSLs

We analyzed 5 sub-CSSLs containing ‘IR64’ chromosome segments around qGW1-1 with partial overlap (Fig. 4A). QTL analysis confirmed the presence or absence of qGW1-1 in each sub-CSSL and localized the QTL within a 1.63-Mb interval between marker loci RM3148 and RM6324 (Fig. 4A). QTLs for GL were identified on both sides of qGW1-1, with additive effects in opposite directions (Supplemental Fig. 2A).

Fig. 4.

Validation of QTLs for GL and GW detected in BC_3–5F₂ populations by use of sub-CSSLs. Graphical genotypes of sub-CSSLs and the positions of QTLs detected in each sub-CSSL are shown. (A) QTLs for GW detected around 0.30–0.75 Mb on chr. 1 in the population with ‘IR64’ as the donor (Nagata et al. 2015). (B) QTLs for GW detected in the region around 3.14 Mb on chr. 7 in the population with ‘Deng Pao Zhai’ as the donor. (C) QTLs for GL detected in the region around 5.04 Mb on chr. 12 in the population with ‘Bei Khe’ as the donor. ■ Homozygous for donor alleles; □ homozygous for ‘Koshihikari’ alleles; ■ heterozygous. △ Donor alleles that increase trait values. LOD, logarithm of odds; AE, additive effect; R², explained variance; n.s., not significant.

We analyzed 5 sub-CSSLs containing ‘Deng Pao Zhai’ chromosome segments around 3.14 Mb on chr. 7. Three sub-CSSLs detected a QTL for GW, designated as qGW7-2 (Fig. 4B). The consensus region of qGW7-2 was located within a 491-kb interval between marker loci KA1894 and RM5711 (Fig. 4B). Consistent with the analysis in BC_3–5F₂, a QTL for GL was identified distal to qGW7-2 (Supplemental Fig. 2B).

We analyzed 4 sub-CSSLs containing ‘Bei Khe’ chromosome segments around 5.04 Mb on chr. 12. Three sub-CSSLs detected a QTL for GL, designated as qGL12-2 (Fig. 4C). The consensus region of the QTL was located within a 3.99-Mb interval between marker loci IDR1566 and KA3544 (Fig. 4C). Another QTL for GL was detected proximal to qGL12-2 (Supplemental Fig. 2C).

Delimitation of qGW1-1 by using progeny lines of sub-CSSLs

To further delimit qGW1-1, we used 6 progeny lines (BC₄F₅) where recombination occurred within the target chromosomal region of the QTL. We compared GWs between homozygous recombinant plants and control non-recombinant plants within the lines to test whether the alleles differed. The analysis delimited qGW1-1 to an interval between marker loci FA6836 and RM5423, corresponding to 477-kb in the ‘Nipponbare’ IRGSP-1.0 genome (Fig. 5).

Fig. 5.

Delimitation of qGW1-1 for GW using progeny lines of sub-CSSLs. Graphical genotypes of six lines derived from sub-CSSLs are shown. Plants were genotyped, and two genotype classes—homozygous lines differing in the size of ‘IR64’-derived introgressions—were selected for GW measurement. ■ Homozygous for donor (‘IR64’) alleles; □ homozygous for ‘Koshihikari’ alleles. Values are presented as mean ± SD (n), where n denotes the number of individuals analyzed. Dif., difference in mean GW between the two genotype classes. Significant differences by t-test: *0.01 < P < 0.05, **P < 0.01; n.s., not significant.

Discussion

We developed a comprehensive catalog of QTLs for GL and width in Asian cultivated rice and revealing genetic variations underlying these traits among diverse donor cultivars. The use of sub-CSSLs enabled efficient validation and delimitation of QTLs, including those in previously unreported chromosomal regions.

These results broaden our understanding of the genetic control of GL and width in rice, which will support more precise strategies for improving rice grain traits. More QTLs were derived from indica donors than from japonica donors, reflecting a greater genetic difference from the recurrent parent (Tables 2, 3). Major-effect QTLs such as GS3 for GL and qGW5 for grain width (Shomura et al. 2008, Takano-Kai et al. 2009) were key factors in distinguishing varietal groups when additive effects were considered (Fig. 3A, 3B), while the overall architecture may reflect genome-wide divergence (Fig. 3C, 3D). Such observations suggest that genetic variation in GL and width is shaped by minor-effect QTL alleles accumulated through historical hybridization and differentiation events (Civáň et al. 2015, Santos et al. 2019, Vaughan et al. 2008), together with spontaneous mutations, rather than by strong selection on individual alleles.

The coincidence of QTLs for GL and width is well documented (Huang et al. 2013, Ogawa et al. 2018), and many cloned genes affect both traits, either in the same or in opposite directions (Xuedan et al. 2023). In our study, progeny testing using sub-CSSLs indicated that independent loci within the same chromosomal region may regulate these traits (Fig. 4, Supplemental Fig. 2). However, limited resolution prevented us from determining whether the observed variation among donors at QTLs within a chromosomal interval (Table 2, 3) reflected allelic differences at a single locus (Sun et al. 2022, Wang et al. 2015a) or tightly linked multiple QTLs and copy number variations (Wang et al. 2015b). Approximately 68% (318 out of 469) of the detected QTLs overlapped with regions harboring cloned genes, yet the presence of multiple cloned genes within a single chromosomal region and the extensive sequence variation among donor cultivars complicate the prediction of genotype–phenotype relationships. These considerations suggest that the number of QTLs may be underestimated and highlight the need for higher-resolution mapping and detailed characterization of causal variants to improve trait prediction and support crop improvement.

Even in highly homogenous backcrossed populations, unidentified small introgressions from donors can introduce background noise in QTL analysis. In some populations, we were unable to detect QTLs, despite phenotypic variation comparable to that in populations where QTLs were detected. Incorporating genotype information obtained through re-sequencing can help mitigate this issue. Another challenge is gene interaction, such as the combined effect of GW7 and GW8 (Yang et al. 2023), which cannot be estimated in a recurrent genetic background. The discrepancy between the estimated values based on cumulative additive effects of QTLs and the actual measured values highlights the need for complementary approaches or genetic materials, such as nested association mapping or MAGIC populations, to achieve more robust trait prediction.

Advances in resequencing and large-scale phenotyping have improved variant detection and enable sophisticated workflows (Han and Huang 2013, Leung et al. 2015, Wang et al. 2018, Yang et al. 2025, Zhang et al. 2021). However, predicting the effects of DNA variations, including those in non-coding regions, across diverse cultivars remains challenging due to their abundance and complexity (Hou et al. 2019, Yocca and Edger 2022, Wang et al. 2020). Pan-genomics approaches have uncovered previously undetected variations in reference genomes, including in stress-response genes, highlighting the potential of diverse genetic resources to explain missing heritability (Guo et al. 2025, Qin et al. 2021, Zhao et al. 2018) and strengthen prediction models (Zhou et al. 2022). Integrating genomic approaches and genetic studies with systematic use of advanced mapping populations such as sub-CSSLs can accelerate the identification of key variations, which is critical for improving prediction models (Zhang et al. 2014) and developing resilient crop cultivars for sustainable production under changing environmental conditions.

Author Contribution Statement

SF conceived and designed the experiments. KN, TA and the other authors determined the genotypes. KN and SF developed the plant materials and phenotyped agronomic traits. KN, TA and SF analyzed the data. KN and SF wrote the paper. All authors have reviewed drafts of the paper and approved the final version.

 Acknowledgments

We thank Dr. M. Yano and other contributors for providing BC_3–5F₂ plants for our study. We are also grateful to the technical staff at NARO for their assistance in managing the rice fields. This work was supported by a grant from the Ministry of Agriculture, Forestry and Fisheries of Japan (JP13405960, Genomics-based Technology for Agricultural Improvement, IVG2003). We also thank two editors from ELSS, Inc. (https://elss.co.jp/en/) for editing our manuscript before submission.

Literature Cited

•  Abe,  T.,  Y.  Nonoue,  N.  Ono,  M.  Omoteno,  M.  Kuramata,  S.  Fukuoka,  T.  Yamamoto,  M.  Yano and  S.  Ishikawa (2013) Detection of QTLs to reduce cadmium content in rice grains using LAC23/Koshihikari chromosome segment substitution lines. Breed Sci 63: 284–291.
•  Bandillo,  N.,  C.  Raghavan,  P.A.  Muyco,  M.A.L.  Sevilla,  I.T.  Lobina,  C.J.  Dilla-Ermita,  C.-W.  Tung,  S.  McCouch,  M.  Thomson,  R.  Mauleon et al. (2013) Multi-parent advanced generation inter-cross (MAGIC) populations in rice: progress and potential for genetics research and breeding. Rice (N Y) 6: 11.
•  Bartholomé,  J.,  P.T.  Prakash and  J.N.  Cobb (2022) Genomic prediction: Progress and perspectives for rice improvement. Methods Mol Biol 2467: 569–617.
•  Basten,  C.J.,  B.S.  Weir and  Z.-B.  Zeng (2005) QTL Cartographer Version 1.17. Program in Statistical Genetics Bioinformatics Research Center, Department of Statistics, North Carolina State University, Raleigh, NC.
•  Civáň,  P.,  H.  Craig,  C.J.  Cox and  T.A.  Brown (2015) Three geographically separate domestications of Asian rice. Nat Plants 1: 15164.
•  e Sousa,  M.B.,  G.  Galli,  D.H.  Lyra,  Í.S.C.  Granato,  F.I.  Matias,  F.C.  Alves and  R.  Fritsche-Neto (2019) Increasing accuracy and reducing costs of genomic prediction by marker selection. Euphytica 215: 18.
•  Fang,  J.,  Y.  Chun,  F.  Zhang,  T.  Guo,  M.  Ren,  J.  Zhao,  S.  Yuan,  W.  Wang,  Y.  Li and  X.  Li (2025) A novel OsMPK6-OsMADS47-PPKL1/3 module controls grain shape and yield in rice. Adv Sci 12: e01946.
•  Fragoso,  C.A.,  M.  Moreno,  Z.  Wang,  C.  Heffelfinger,  L.J.  Arbelaez,  J.A.  Aguirre,  N.  Franco,  L.E.  Romero,  K.  Labadie,  H.  Zhao et al. (2017) Genetic architecture of a rice nested association mapping population. G3 (Bethesda) 7: 1913–1926.
•  Fukuoka,  S.,  Y.  Nonoue and  M.  Yano (2010) Germplasm enhancement by developing advanced plant materials from diverse rice accessions. Breed Sci 60: 509–517.
•  Galili,  T. (2015) dendextend: an R package for visualizing, adjusting and comparing trees of hierarchical clustering. Bioinformatics 22: 3718–3720.
•  Gao,  Q.,  N.  Zhang,  W.Q.  Wang,  S.Y.  Shen,  C.  Bai and  X.J.  Song (2021) The ubiquitin-interacting motif-type ubiquitin receptor HDR3 interacts with and stabilizes the histone acetyltransferase GW6a to control the grain size in rice. Plant Cell 33: 3331–3347.
•  Guo,  D.,  Y.  Li,  H.  Lu,  Y.  Zhao,  N.  Kurata,  X.  Wei,  A.  Wang,  Y.  Wang,  Q.  Zhan,  D.  Fan et al. (2025) A pangenome reference of wild and cultivated rice. Nature 642: 662–671.
•  Han,  B. and  X.  Huang (2013) Sequencing-based genome-wide association study in rice. Curr Opin Plant Biol 16: 133–138.
•  Hou,  J.,  D.  Lu,  A.S.  Mason,  B.  Li,  M.  Xiao,  S.  An and  D.  Fu (2019) Non-coding RNAs and transposable elements in plant genomes: emergence, regulatory mechanisms and roles in plant development and stress responses. Planta 250: 23–40.
•  Huang,  R.,  L.  Jiang,  J.  Zheng,  T.  Wang,  H.  Wang,  Y.  Huang and  Z.  Hong (2013) Genetic bases of rice grain shape: so many genes, so little known. Trends Plant Sci 18: 218–226.
•  Iwata,  H.,  K.  Ebana,  Y.  Uga and  T.  Hayashi (2015) Genomic prediction of biological shape: elliptic Fourier analysis and kernel partial least squares (PLS) regression applied to grain shape prediction in rice (Oryza sativa L.). PLoS One 10: e0120610.
•  Jiang,  Y.,  M.  Zhou,  S.  Ke,  X.  Deng and  Y.  Li (2024) GSW3.1, a novel gene controlling grain size and weight in rice. Crop J 12: 796–802.
•  Kawahara,  Y.,  M.  de la Bastide,  J.P.  Hamilton,  H.  Kanamori,  W.R.  McCombie,  S.  Ouyang,  D.C.  Schwartz,  T.  Tanaka,  J.  Wu,  S.  Zhou et al. (2013) Improvement of the Oryza sativa Nipponbare reference genome using next generation sequence and optical map data. Rice (N Y) 6: 4.
•  Kojima,  Y.,  K.  Ebana,  S.  Fukuoka,  T.  Nagamine and  M.  Kawase (2005) Development of an RFLP-based rice diversity research set of germplasm. Breed Sci 55: 431–440.
•  Kumagai,  M.,  D.  Nishikawa,  Y.  Kawahara,  H.  Wakimoto,  R.  Itoh,  N.  Tabei,  T.  Tanaka and  T.  Itoh (2019) TASUKE+: a web-based platform for exploring GWAS results and large-scale resequencing data. DNA Res 26: 445–452.
•  Lander,  E.S.,  P.  Green,  J.  Abrahamson,  A.  Barlow,  M.J.  Daly,  S.E.  Lincoln and  L.  Newberg (1987) MAPMAKER: an interactive computer package for constructing primary genetic linkage maps of experimental and natural populations. Genomics 1: 174–181.
•  Leung,  H.,  C.  Raghavan,  B.  Zhou,  R.  Oliva,  I.R.  Choi,  V.  Lacorte,  M.L.  Jubay,  C.V.  Cruz,  G.  Gregorio,  R.K.  Singh et al. (2015) Allele mining and enhanced genetic recombination for rice breeding. Rice (N Y) 8: 34.
•  Li,  C.,  J.  Liu,  L.  Zhang,  T.  Li,  H.  Li,  B.  Liu and  T.  Zhao (2024a) OsDA1 positively regulates grain width in rice. Crop J 12: 92–101.
•  Li,  H.,  Y.  Wang,  W.  Qiao,  Z.  Zhu,  Z.  Wang,  Y.  Tian,  S.  Liu,  J.  Wan and  L.  Liu (2024b) Identification of a novel locus qGW12/OsPUB23 regulating grain shape and weight in rice (Oryza sativa L.). Theor Appl Genet 137: 267.
•  Li,  N.,  R.  Xu and  Y.  Li (2019) Molecular networks of seed size control in plants. Annu Rev Plant Biol 70: 435–463.
•  Liu,  S.,  J.  Wu,  A.M.  Mawia,  X.  Wei,  R.  Cao,  G.  Jiao,  Y.  Wu,  J.  Zhang,  L.  Xie,  Z.  Sheng et al. (2025) A novel transcription factor OsMYB73 affects grain size and chalkiness by regulating endosperm storage substances’ accumulation-mediated auxin biosynthesis signalling pathway in rice. Plant Biotechnol J 23: 1021–1038.
•  Long,  W.,  Q.  He,  Y.  Wang,  Y.  Wang,  J.  Wang,  Z.  Yuan,  M.  Wang,  W.  Chen,  L.  Luo,  L.  Luo et al. (2024) Genome evolution and diversity of wild and cultivated rice species. Nat Commun 15: 9994.
•  Lu,  Q.,  M.  Zhang,  X.  Niu,  S.  Wang,  Q.  Xu,  Y.  Feng,  C.  Wang,  H.  Deng,  X.  Yuan,  H.  Yu et al. (2015) Genetic variation and association mapping for 12 agronomic traits in indica rice. BMC Genomics 16: 1067.
•  Luo,  Y.,  Y.  Chen,  P.  Xue,  B.  Wang,  Y.  Kang,  Y.  Zhang,  D.  Chen,  Y.  Hong,  W.  Wu,  Q.  Liu et al. (2025) Modulation of rice grain shape and appearance by the GS10-encoded long coiled-coil protein. Crop J 13: 158–169.
•  Ma,  X.,  H.  Wang,  S.  Yan,  C.  Zhou,  K.  Zhou,  Q.  Zhang,  M.  Li,  Y.  Yang,  D.  Li,  P.  Song et al. (2025) Large-scale genomic and phenomic analyses of modern cultivars empower future rice breeding design. Mol Plant 18: 651–668.
•  Mizuno,  Y.,  U.  Yamanouchi,  T.  Hoshino,  Y.  Nonoue,  K.  Nagata,  S.  Fukuoka,  T.  Ando,  M.  Yano and  K.  Sugimoto (2018) Genetic dissection of pre-harvest sprouting resistance in an upland rice cultivar. Breed Sci 68: 200–209.
•  Nagata,  K.,  T.  Ando,  Y.  Nonoue,  T.  Mizubayashi,  N.  Kitazawa,  A.  Shomura,  K.  Matsubara,  N.  Ono,  R.  Mizobuchi,  T.  Shibaya et al. (2015) Advanced backcross QTL analysis reveals complicated genetic control of rice grain shape in a japonica × indica cross. Breed Sci 65: 308–318.
•  Nagata,  K.,  Y.  Nonoue,  K.  Matsubara,  R.  Mizobuchi,  N.  Ono,  T.  Shibaya,  K.  Ebana,  E.  Ogiso-Tanaka,  T.  Tanabata,  K.  Sugimoto et al. (2023) Development of 12 sets of chromosome segment substitution lines that enhance allele mining in Asian cultivated rice. Breed Sci 73: 332–342.
•  Nawade,  B.,  S.H.  Chu,  S.  Kim,  C.Y.  Lee,  J.  Bao and  Y.J.  Park (2024) Nested association mapping population in japonica rice: Development, characterization, and application in genome-wide association studies. Curr Plant Biol 40: 100401.
•  Nourollah,  A. (2016) Genetic diversity, genetic erosion, and conservation of the two cultivated rice species (Oryza sativa and Oryza glaberrima) and their close wild relatives. In: Ahuja, M.R. and S.M. Jain (eds.) Genetic diversity and erosion in plants: case histories, Springer, Cham, pp. 35–73.
•  Ogawa,  D.,  Y.  Nonoue,  H.  Tsunematsu,  N.  Kanno,  T.  Yamamoto and  J.i.  Yonemaru (2018) Discovery of QTL alleles for grain shape in the Japan-MAGIC rice population using haplotype Information. G3 (Bethesda) 8: 3559–3565.
•  Qin,  P.,  H.  Lu,  H.  Du,  H.  Wang,  W.  Chen,  Z.  Chen,  Q.  He,  S.  Ou,  H.  Zhang,  X.  Li et al. (2021) Pan-genome analysis of 33 genetically diverse rice accessions reveals hidden genomic variations. Cell 184: 3542–3558.e16.
• R Core Team (2023) R: a language and environment for statistical computing. R foundation for statistical computing, Vienna, Austria.
•  Sakai,  H.,  S.S.  Lee,  T.  Tanaka,  H.  Numa,  J.  Kim,  Y.  Kawahara,  H.  Wakimoto,  C.C.  Yang,  M.  Iwamoto,  T.  Abe et al. (2013) Rice annotation project database (RAP-DB): an integrative and interactive database for rice genomics. Plant Cell Physiol 54: e6.
•  Santos,  J.D.,  D.  Chebotarov,  K.L.  McNally,  J.  Bartholomé,  G.  Droc,  C.  Billot and  J.C.  Glaszmann (2019) Fine scale genomic signals of admixture and alien introgression among Asian rice landraces. Genome Biol Evol 11: 1358–1373.
•  Shomura,  A.,  T.  Izawa,  K.  Ebana,  T.  Ebitani,  H.  Kanegae,  S.  Konishi and  M.  Yano (2008) Deletion in a gene associated with grain size increased yields during rice domestication. Nat Genet 40: 1023–1028.
•  Sun,  K.,  D.  Li,  A.  Xia,  H.  Zhao,  Q.  Wen,  S.  Jia,  J.  Wang,  G.  Yang,  D.  Zhou,  C.  Huang et al. (2022) Targeted identification of rice grain-associated gene allelic variation through mutation induction, targeted sequencing, and whole genome sequencing combined with a mixed-samples strategy. Rice (N Y) 15: 57.
•  Takagi,  H.,  A.  Abe,  K.  Yoshida,  S.  Kosugi,  S.  Natsume,  C.  Mitsuoka,  A.  Uemura,  H.  Utsushi,  M.  Tamiru,  S.  Takuno et al. (2013) QTL-seq: rapid mapping of quantitative trait loci in rice by whole genome resequencing of DNA from two bulked populations. Plant J 74: 174–183.
•  Takano-Kai,  N.,  H.  Jiang,  T.  Kubo,  M.  Sweeney,  T.  Matsumoto,  H.  Kanamori,  B.  Padhukasahasram,  C.  Bustamante,  A.  Yoshimura,  K.  Doi et al. (2009) Evolutionary history of GS3, a gene conferring grain length in rice. Genetics 182: 1323–1334.
•  Tanabata,  T.,  T.  Shibaya,  K.  Hori,  K.  Ebana and  M.  Yano (2012) SmartGrain: high-throughput phenotyping software for measuring seed shape through image analysis. Plant Physiol 160: 1871–1880.
•  Tang,  S.,  Z.  Zhao,  X.  Liu,  Y.  Sui,  D.  Zhang,  H.  Zhi,  Y.  Gao,  H.  Zhang,  L.  Zhang,  Y.  Wang et al. (2023) An E2-E3 pair contributes to seed size control in grain crops. Nat Commun 14: 3091.
•  Vaughan,  D.A.,  B.R.  Lu and  N.  Tomooka (2008) Was Asian rice (Oryza sativa) domesticated more than once? Rice (N Y) 1: 16–24.
•  Wang,  Q.,  J.  Tang,  B.  Han and  X.  Huang (2020) Advances in genome-wide association studies of complex traits in rice. Theor Appl Genet 133: 1415–1425.
•  Wang,  S.,  S.  Li,  Q.  Liu,  K.  Wu,  J.  Zhang,  S.  Wang,  Y.  Wang,  X.  Chen,  Y.  Zhang,  C.  Gao et al. (2015a) The OsSPL16-GW7 regulatory module determines grain shape and simultaneously improves rice yield and grain quality. Nat Genet 47: 949–954.
•  Wang,  W.,  R.  Mauleon,  Z.  Hu,  D.  Chebotarov,  S.  Tai,  Z.  Wu,  M.  Li,  T.  Zheng,  R.R.  Fuentes,  F.  Zhang et al. (2018) Genomic variation in 3,010 diverse accessions of Asian cultivated rice. Nature 557: 43–49.
•  Wang,  Y.,  G.  Xiong,  J.  Hu,  L.  Jiang,  H.  Yu,  J.  Xu,  Y.  Fang,  L.  Zeng,  E.  Xu,  J.  Xu et al. (2015b) Copy number variation at the GL7 locus contributes to grain size diversity in rice. Nat Genet 47: 944–948.
•  Xin,  W.,  N.  Chen,  J.  Wang,  Y.  Liu,  Y.  Sun,  B.  Han,  X.  Wang,  Z.  Liu,  H.  Liu,  H.  Zheng et al. (2024) Candidate gene analysis of rice grain shape based on genome-wide association study. Theor Appl Genet 137: 241.
•  Xuedan,  L.,  L.  Fan,  X.  Yunhua,  W.  Feng,  Z.  Guilian,  D.  Huabing and  T.  Wenbang (2023) Grain shape genes: shaping the future of rice breeding. Rice Sci 30: 379–404.
•  Yamamoto,  T.,  J.  Yonemaru and  M.  Yano (2009) Towards the understanding of complex traits in rice: substantially or superficially? DNA Res 16: 141–154.
•  Yang,  L.,  W.  He,  Y.  Zhu,  Y.  Lv,  Y.  Li,  Q.  Zhang,  Y.  Liu,  Z.  Zhang,  T.  Wang,  H.  Wei et al. (2025) GWAS meta-analysis using a graph-based pan-genome enhanced gene mining efficiency for agronomic traits in rice. Nat Commun 16: 3171.
•  Yang,  T.,  H.  Gu,  W.  Yang,  B.  Liu,  S.  Liang and  J.  Zhao (2023) Artificially selected grain shape gene combinations in Guangdong Simiao varieties of rice (Oryza sativa L.). Rice (N Y) 16: 3.
•  Yang,  X.,  S.  Yu,  S.  Yan,  H.  Wang,  W.  Fang,  Y.  Chen,  X.  Ma and  L.  Han (2024) Progress in rice breeding based on genomic research. Genes (Basel) 15: 564.
•  Yano,  K.,  E.  Yamamoto,  K.  Aya,  H.  Takeuchi,  P.C.  Lo,  L.  Hu,  M.  Yamasaki,  S.  Yoshida,  H.  Kitano,  K.  Hirano et al. (2016) Genome-wide association study using whole-genome sequencing rapidly identifies new genes influencing agronomic traits in rice. Nat Genet 48: 927–934.
•  Yano,  M. and  T.  Sasaki (1997) Genetic and molecular dissection of quantitative traits in rice. Plant Mol Biol 35: 145–153.
•  Yocca,  A.E. and  P.P.  Edger (2022) Current status and future perspectives on the evolution of cis-regulatory elements in plants. Curr Opin Plant Biol 65: 102139.
•  Zhang,  H.,  D.R.  Huang,  Y.  Shen,  X.J.  Niu,  Y.Y.  Fan,  Z.H.  Zhang,  J.Y.  Zhuang and  Y.J.  Zhu (2024) GL5.2, a quantitative trait locus for rice grain shape, encodes a RING-type E3 ubiquitin ligase. Plants 13: 2521.
•  Zhang,  J.,  D.  Zhang,  Y.  Fan,  C.  Li,  P.  Xu,  W.  Li,  Q.  Sun,  X.  Huang,  C.  Zhang,  L.  Wu et al. (2021) The identification of grain size genes by RapMap reveals directional selection during rice domestication. Nat Commun 12: 5673.
•  Zhang,  Z.,  U.  Ober,  M.  Erbe,  H.  Zhang,  N.  Gao,  J.  He,  J.  Li and  H.  Simianer (2014) Improving the accuracy of whole genome prediction for complex traits using the results of genome wide association studies. PLoS One 9: e93017.
•  Zhao,  Q.,  Q.  Feng,  H.  Lu,  Y.  Li,  A.  Wang,  Q.  Tian,  Q.  Zhan,  Y.  Lu,  L.  Zhang,  T.  Huang et al. (2018) Pan-genome analysis highlights the extent of genomic variation in cultivated and wild rice. Nat Genet 50: 278–284.
•  Zheng,  Y.,  M.  Li,  P.  Sun,  G.  Gao,  Q.  Zhang,  Y.  Li,  G.  Lou,  B.  Wu and  Y.  He (2024) QTL detection for grain shape and fine mapping of two novel locus qGL4 and qGL6 in rice. Mol Breed 44: 62.
•  Zhou,  Y.,  Z.  Zhang,  Z.  Bao,  H.  Li,  Y.  Lyu,  Y.  Zan,  Y.  Wu,  L.  Cheng,  Y.  Fang,  K.  Wu et al. (2022) Graph pangenome captures missing heritability and empowers tomato breeding. Nature 606: 527–534.