QTL analysis on rice grain appearance quality, as exemplifying the typical events of transgenic or backcrossing breeding.

Rice grain shape and yield are usually controlled by multiple quantitative trait loci (QTL). This study used a set of F9-10 recombinant inbred lines (RILs) derived from a cross of Huahui 3 (Bt/Xa21) and Zhongguoxiangdao, and detected 27 QTLs on ten rice chromosomes. Among them, twelve QTLs responsive for grain shape/ or yield were mostly reproducibly detected and had not yet been reported before. Interestingly, the two known genes involved in the materials, with one insect-resistant Bt gene, and the other disease-resistant Xa21 gene, were found to closely link the QTLs responsive for grain shape and weight. The Bt fragment insertion was firstly mapped on the chromosome 10 in Huahui 3 and may disrupt grain-related QTLs resulting in weaker yield performance in transgenic plants. The introgression of Xa21 gene by backcrossing from donor material into receptor Minghui 63 may also contain a donor linkage drag which included minor-effect QTL alleles positively affecting grain shape and yield. The QTL analysis on rice grain appearance quality exemplified the typical events of transgenic or backcrossing breeding. The QTL findings in this study will in the future facilitate the gene isolation and breeding application for improvement of rice grain shape and yield.

Introduction

Rice is one of the most important cereal crops and a staple food in Asia. Grain weight is one of the most important components of rice grain yield and is conditioned by quantitative trait locus (QTL). Grain shape, a typical complex quantitative trait, is closely associated with grain weight and usually measured by grain length (GL), width (GW), thickness (GT) and length-to-width ratio (LWR) (Lin and Wu 2003, Yoon ). LWR is considered to be an important measure of rice appearance quality since people’s preferences for it are rather different in the rice-producing areas of the world (Wan , 2008). Extensively phenotypic variations have been observed among cultivars, subspecies and species of Oryza sativa L. Hundreds of QTLs for grain shape/weight have been identified and are scattered throughout the whole rice genome (www.gramene.com). Many QTLs can be repeatedly detected across different populations and environments. For example, one major QTL for GL and grain weight was consistently mapped near the centromeric region of rice chromosome 3 using different populations in several independent studies (Li , Li , Thomson , Xiao , Xing , Yu ).

Stable genetic effect on regulating grain weight/shape is critical for a QTL to be utilized in rice breeding. Some QTLs were fine mapped and validated with nearly isogenic lines. A few QTLs associated with grain weight, gw3.1, gw3, gw6, gw8.1, gw9.1 and tgw11, have been identified at accurate intrachromosomal locations (Guo , Li , Oh , Xie , 2008). GS3, gl-3, qGL7 and qGL7-2 controlling GL were fine mapped at the exact locations on chromosomes 3 and 7 (Bai , Fan , Shao , Wan ). qGW5 for GW was narrowed down to a 49.7 kb genomic region on chromosome 5 (Wan ).

Grain traits are closely correlated and QTLs for different traits are often mapped in the similar interval. For example, confidence interval of gw3.1 overlapped with that of gl-3 and GS3. Fine mapping of these QTLs provides a basis for further cloning and marker assisted selection breeding.

Several major QTLs controlling rice grain shape were isolated and characterized using map-based cloning method due to the progress in rice functional genomics. GS3, a major QTL for GL, encodes a putative transmembrane protein and is composed of four putative domains. GS3 negatively regulated GL and loss of function or deletion of plant-specific organ size regulation (OSR) domain would result in long grain (Fan , Mao ). Three QTLs for GW, GW2, qSW5/GW5 and GS5, have been cloned and characterized. GW2 encodes a RING-type protein with E3 ubiquitin ligase activity, which is known to function in the degradation by the ubiquitin proteasome pathway. GW2 E3 ligase is a new negative regulator of cell division and the mutant allele of GW2 promotes spikelet hull cell division, resulting in an increase of grain width and weight (Song ). qSW5 and GW5 are the same QTL and have been identified on chromosome 5 (Shomura , Weng ). GW5 encodes a novel nuclear protein, which physically interacts with polyubiquitin and is also likely to act in the ubiquitin-proteasome pathway to regulate cell division. Additionally, GW5 is a negative regulator and the mutant allele causes an increase in GW (Weng ). GS5 encodes a putative serine carboxypeptidase and functions as a positive regulator of GW. Overexpression of GS5 may promote cell division resulting in an increase of GW (Li ). Functional characterization of these genes has provided novel insights to understand genetic and molecular mechanisms regulating grain traits. However, the mechanism of regulation of grain size and epistatic interactions between these QTLs are still unclear.

Efficient crop improvement included transgenic and backcrossing breeding. For transgenic breeding, agrobacterium-mediated transformation is being increasingly used in rice. It can realize foreign gene integrating into rice genome and subsequently precisely improve the target trait (Tu ). Backcrossing breeding takes the strategy of transferring donor objective gene into receptor and meanwhile strenuously excluding donor transgenic background via multiple backcrossing. Backcrossing is usually regarded conventional breeding and can be performed more efficiency through marker assisted selection (Chen ). In comparison, transgenic breeding can raise some problems such as transgenic silencing and endogenic insertional silencing, while backcrossing breeding needs more crossing generations and can frequently accompany genetic background problems typical of the event of linkage drag.

Here, two parents contrasting in grain shape and yield, with one parent containing the events of transgenic and backcrossing breeding improvement, were selected to develop a RIL mapping population. This study aimed to achieve two points. One was to identify more grain-shape QTLs to help enhancing the understanding of rice seed development, and the other tried to analyze the genetic base underlying the grain shape changes brought by transgenic and backcrossing breeding.

Materials and Methods

Plant materials and field experiments

A set of RILs population was constructed from an F9:10 population consisted of 237 individuals from a cross between Zhongguoxiandao (ZGX) and Huahui 3 (HH 3). ZGX is a high-quality rice which is extensively planted in south China. HH 3 is an improved rice material on the base of an elite indica cytoplasm male sterile (CMS) restorer line Minghui 63. As previously described, firstly the wild-rice-derived dominant gene Xa21 conferring multi-race resistance to bacterial blight and a fused Bt gene cry1Ab/cry1Ac conferring resistance to lepidopteran insects were individually introduced into the same genetic background of Minghui 63. The former (here designated as Minghui 63/Xa21) was bred from a recurrent backcross between IRBB21 as donor line and Minghui 63 as recurrent parent via marker-assisted selection (Chen ). The latter (designated as Minghui 63/Bt) is a marker free transgenic line obtained by biolistic transformation (Tu ). Through crossing the two improved Minghui 63, the two genes were then pyramided into the same recipient plant of Minghui 63 (designated as Huahui 3 or Minghui 63/Bt&Xa21), showing desirable insect- and disease-resistant phenotypes (Jiang ). That is to say, Minghui 63/Xa21, Minghui 63/Bt and Minghui 63/Bt&Xa21 (HH 3) are near isogenic lines on the background of Minghui 63. The field trials of the lines, including Minghui 63/Xa21, Minghui 63/Bt, the derived RIL population and their two parents ZGX and HH 3, were carried out in the normal season of the years of 2009 and 2010 in Wuhan. The tested materials were planted in a randomized complete block design with two replications. In each block, 24 plants for each line were planted in a two-row plot with a 17 cm distance between plants within a row and 27 cm between rows. The central eight plants in a row in each plot were examined for agronomic traits. Normal cultural practices for growing rice were followed during the course of the experiments.

Phenotypic measurements

Harvested rice grains were stored and dried at room temperature for at least 3 months before processing. Five yield-related traits including grain length (GL, in millimeters), grain width (GW, in millimeters), ratio of grain length to width (LWR), grain-thickness (GT, in millimeters) and 1000-grain weight (TGW, in grams) were examined from eight randomly selected plants in the middle of the rows of each line. Ten full-filled rice grains were chosen randomly from each plant for trait measurement. GL was estimated by placing 10 grains length-wise one by one in a straight line along a ruler. These 10 seeds were individually measured for GW and GT using an electronic digital caliper (Guanglu Measuring Instrument Co. Ltd., China) with a precision of 0.1 mm. The averaged GL, GW, and GT as the trait values of that line were used for data analysis. The grain length-to-width ratio (LWR) is equal to GL divided by its GW. TGW was calculated as the grain weight per plant divided by its grain number multiplied by 1000.

Molecular markers and linkage map

Fresh leaves were collected from each material for DNA extraction. DNA was extracted using a microisolation method as described by Cho with minor modifications. Polymorphic simple sequence repeat (SSR) markers or insertion deletion (InDel) markers between the parents ZGX and HH3 were used to genotype the population. The SSR assay was carried out essentially as described by Wu and Tanksley (1993). The genetic linkage map with RILs was constructed by using MAPMAKER/EXP version 3.0b (1987). The local linkage map based on the NIL-F2 population was developed. Genetic distance was calculated in Haldane function (1919).

Data analysis

The STATISTICA statistical package was used to analyze the t-test, frequency distribution, and correlation coefficients of traits. The chromosomal locations of QTLs were determined by composite interval mapping (CIM). CIM analysis was performed using WinQTLcart 2.0 (Wang and Zeng 2004), and epistatic QTLs analysis was operated by QTLnetwork 2.0 (Yang ). A threshold of P ≤ 0.005 and LOD ≥ 2.5 was used to declare the significant main effect QTL (M-QTL), digenic epistatic QTLs, and QTL × environment interaction. Contribution rate (h2, %) was estimated as the percentage of variance explained by each locus or epistatic pair in proportion to the total phenotypic variance. QTLs were named following the popular nomenclature but in alphabetic order for QTLs on the same chromosome.

Results

Grain shape and yield performance of Minghui 63 added with Bt and/or Xa21 genes

As mentioned before (Jiang ), Minghui 63 added with Bt or Xa21 genes can highly increase the insect- or disease-resistance. However, in the currently used material Minghui 63 possessing Bt with or without Xa21 genes, the grain performance of GL and GW and TGW was significantly decreased when compared to original receptor Minghui 63. Minghui 63 (Xa21) showed a bit better than Minghui 63 in the three traits of GL, GW and TGW, but not reaching a statistical significant level (Table 1). These data suggested that at least the introduction of Bt gene altered the receptor’s grain-related genetic environment. To elucidate the genetic cause thus becomes necessary.

Table 1

Statistics of the trait performance between Minghui 63 and its improved version in 2009 (upper) and 2010 (lower)

parents	GL (mm)	GW (mm)	GT (mm)	TGW (g)
Minghui 63 (CK)	9.81 ± 0.11	2.89 ± 0.04	1.93 ± 0.04	29.7 ± 1.4
	9.83 ± 0.13	2.90 ± 0.05	1.97 ± 0.05	28.3 ± 0.4
Minghui 63/Bt	9.42 ± 0.11**	2.72 ± 0.01**	1.91 ± 0.04	25.6 ± 0.8**
	9.60 ± 0.22**	2.78 ± 0.04**	1.89 ± 0.02**	24.8 ± 0.5**
Minghui 63/Xa21	9.82 ± 0.12	2.90 ± 0.07	1.94 ± 0.06	29.9 ± 1.3
	9.84 ± 0.14	2.92 ± 0.03	1.98 ± 0.02	28.7 ± 0.7
Huahui 3 (Minghui	9.44 ± 0.11**	2.73 ± 0.05**	1.92 ± 0.06	25.6 ± 0.5**
63/Bt&Xa21)	9.61 ± 0.4**	2.78 ± 0.04**	1.89 ± 0.06*	25.0 ± 1.0**

Significantly different from the performance of Minghui 63 at * P ≤ 0.05 and ** P ≤ 0.01.

Distribution and correlation of the traits in RILs

In order to elucidate the genetic cases of Bt or Xa21 introduction into Minghui 63 and meanwhile to identify more QTLs responsive for grain shape and yield, a set of RILs population was constructed from an F9:10 population consisted of 237 individuals from a cross between Zhongguoxiandao (ZGX) and Huahui 3 (HH 3, Minghui63/Bt&Xa21). There were significant differences between two parents on the traits of GL, TGW, and moderate differences on GW, GT and LWR (Table 2). The ZGX grain is significantly longer and heavier than that of HH 3. And due to the similar performance of GW and GT, the heavier TGW of ZGX grain should be mainly raised by the longer GL. Among the population, the five traits exhibited wide variation with the mean values of performance nearly all fallen into the range between that of the two parents. All the five traits in the population exhibited normal distribution patterns similar in the 2 years, indicating a quantitative inheritance underlying these traits (Fig. 1).

Table 2

Statistics of the five traits in parents of Zhongguoxiandao (ZGX) and Huahui 3 (HH 3) and the derived recombination inbred line (RIL) population observed in 2009 (upper) and 2010 (lower)

	Parent (mean ± standard error)		RILs
Traits	HH 3	ZGX	Mean ± standard error	Range
GL (mm)	9.44 ± 0.11	9.93 ± 0.06	9.71 ± 0.44	7.04–10.76
	9.61 ± 0.40	9.83 ± 0.08	9.68 ± 0.38	8.26–10.75
GW (mm)	2.73 ± 0.05	2.70 ± 0.02	2.77 ± 0.14	2.03–3.53
	2.78 ± 0.04	2.60 ± 0.02	2.77 ± 0.12	2.47–3.40
GT (mm)	1.92 ± 0.06	1.99 ± 0.05	1.95 ± 0.09	1.76–2.73
	1.89 ± 0.06	1.94 ± 0.05	1.95 ± 0.06	1.74–2.30
LWR	3.45 ± 0.03	3.67 ± 0.02	3.51 ± 0.22	2.48–4.68
	3.45 ± 0.05	3.78 ± 0.03	3.50 ± 0.18	2.61–3.98
TGW (g)	25.6 ± 0.5	27.3 ± 0.2	26.2 ± 2.3	20.7–33.4
	25.0 ± 1.0	26.5 ± 0.5	26.3 ± 2.5	20.5–37.7

GL, GW and GT are grain length, width and thickness, respectively.

LWR, length/width ratio. TGW, 1000-grain weight.

Fig. 1

Frequency distributions of the RILs derived from a cross between Huahui 3 and Zhongguoxiangdao for 6 traits measured in 2009 and 2010. Black and white arrows indicate the mean values of the parental lines, Zhongguoxiangdao and Huahui3, respectively. Abbrevations are the same as in Table 1.

The phenotypic correlation coefficients showed the traits of GL, GW, GT and TGW were significantly correlated with each other in the two years (p < 0.05) (Table 3). LWR had significant correlation positively with GL and negatively with GW, but with no or small correlation with GT and TGW.

Table 3

Correlation coefficients among the parameters from 237 RILs derived from the cross of Zhongguoxiandao (ZGX) and Huahui 3 (HH 3) in 2009 (upper half) and 2010 (lower half)

	GL (mm)	GW (mm)	LWR	GT (mm)
GW (mm)†	0.15*
	0.24**
LWR	0.58**	−0.71**
	0.56**	−0.67**
GT (mm)	0.44**	0.34**	0.02
	0.48**	0.69**	−0.21**
TGW (g)	0.56**	0.50**	−0.02	0.71**
	0.66**	0.71**	−0.10	0.83**

P ≤ 0.05,

P ≤ 0.01. Abbrevations are the same as in Table 1.

Correlation coefficients in upper half correspond to the data in 2009 and those in lower half correspond to the data in 2010.

Genetic linkage map

According to the reference genetic maps (http://www.gramene.org/), 1251 SSR and four insertion/deletion (InDel) markers were surveyed and consequently 203 markers were identified to be polymorphic between the two parents ZGX and HH 3. The percentage of marker polymorphism accounted for 16%. Among them, a total of 152 simple sequence repeat (SSR) and the four InDel markers, which evenly distributed on the 12 rice chromosomes, were selected for genotyping the RIL population. The linkage map covers a total of 1680.1 centimorgan (cM) of the 12 rice chromosomes. The average distance is 10.9 cM between two adjacent markers. The Bt transgene in the HH 3 is mapped in the interval of RM2503-RM1126 on chromosome 10.

QTLs for grain shape and yield

27 QTLs were identified for the five traits evaluated in 2009 and 2010 (Table 4 and Fig. 2), among which 17 QTLs were detected in both years and the other ten QTLs were identified only in one year.

Table 4

QTLs detected for rice grain size and weight using the F9:10 RIL population derived from a cross between Zhongguoxiandao (ZGX) and Huahui 3 (HH 3) by the interval mapping at LOD threshold 2.4

Trait	QTLs	Chr.a	Interval markers	2009			2010
				LOD	Ab	h2 (%)c	LOD	Ab	h2 (%)c
GL	qGL1	1	RM220–RM243				2.96	−0.07	3.80
	qGL8	8	RM310–RM547	2.80	0.09	3.88	5.55	0.10	6.72
	qGL9	9	RM566–RM434	4.65	0.13	8.10	5.37	0.12	8.21
	qGL10	10	RM25003–Bt	8.04	−0.26	17.65	17.05	−0.23	27.99
GW	qGW2	2	RM262–RM106				2.61	0.03	5.23
	qGW3	3	RM55–RM7000	4.4	0.05	10.5	6.52	0.05	15.21
	qGW4	4	RM16434–RM16502				4.22	0.03	5.10
	qGW10	10	Bt–RM1126	4.4	−0.05	9.07	5.3	−0.05	8.65
	qGW11a	11	RM332–RM116	4.13	0.04	7.12	4.45	0.04	9.65
	qGW11b	11	RM26643–XA				4.25	0.04	9.65
LWR	qLWR2	2	RM324–RM5521	2.83	−0.07	8.19	4.56	−0.05	6.54
	qLWR3	3	RM55–RM7000				3.08	−0.06	9.41
	qLWR5	5	RM169–RM3381	2.59	0.05	4.58	2.62	0.03	3.47
	qLWR10a	10	RM1126–RM258				4.43	−0.06	7.00
	qLWR10b	10	RM467–RM258	2.51	−0.05	4.74	4.45	−0.06	7.02
	qLWR11	11	RM332–RM116	3.19	−0.05	5.23	3.46	−0.04	5.02
GT	qGT3	3	RM1373–RM7000	3.34	0.02	5.10	3.01	0.01	3.92
	qGT4	4	RM261–RM6659				2.54	0.01	3.27
	qGT5	5	RM593–RM0366				3.33	−0.01	4.25
	qGT10	10	Bt–RM1126	4.23	−0.04	10.21	15.13	−0.04	23.34
	qGT11	11	RM26643–XA	3.32	0.02	5.08	3.48	0.02	5.46
TGW	qTGW3	3	RM55–RM7000	7.17	0.91	15.46	5.60	0.85	9.06
	qTGW4	4	RM16459–RM16502	3.51	0.42	3.41
	qTGW5	5	RM593–RM0366	4.35	−0.54	4.65	4.57	−0.61	5.39
	qTGW10	10	Bt–RM1126	27.57	−2.05	35.54	23.83	−1.89	32.30
	qTGW11a	11	RM26092–RM116	2.89	0.55	3.94
	qTGW11b	11	RM26643–XA	5.80	0.71	8.47	4.42	0.66	6.32

Chromosome number.

Additive effects, the positive values indicate the alleles from Huahui 3 increasing the effects.

h2, the total phenotypic variation explained by each QTL.

The QTL around the Bt transgene insertion point is marked in boldface.

Fig. 2

Locations of QTLs controlling rice grain size and weight detected in 2009 and 2010, respectively. The number on the top of each chromosome indicates the chromosome order. The insertion location of the Bt gene co-located with a major QTL cluster for rice grain size and weight is marked on chromosome 10.

Four QTLs were identified for GL on chromosomes 1, 8, 9 and 10, totally explaining the phenotypic variation of 29.6% in 2009 and 46.7% in 2010 (Table 4), and three QTLs (qGL8, qGL9, qGL10) of them were identified in both years. The QTL qGL10, flanked by RM25003 and Bt on chromosome 10, had largest additive effect, explaining the phenotypic variation of 17.7% in 2009 and 28.0% in 2010, and the allele from HH 3 decreased grain length by 0.26 mm and 0.23 mm in 2009 and 2010, respectively (Table 4). The other QTLs increased grain length by HH 3 allele except qGL1.

Six QTLs for GW on chromosomes 2, 3, 4, 10 and 11, totally explained the phenotypic variation of 27.0% in 2009 and 53.5% in 2010. Three (qGW3, qGW10, qGW11a) of them were found in both years (Table 4). One larger effect QTL, qGW10, also flanked by RM1126 and Bt on chromosome 10, co-located with the main effect QTL qGL10 (Fig. 2). The allele of qGW10 from HH 3 decreased grain width, consistent with the effect of qGL10 for grain length.

Analysis of the F9:10 population resolved six QTLs for LWR (Table 4). All these QTLs with the alleles of HH 3 decreased the LWR, except qLWR5. All QTLs controlling length/width ratio had small effects explaining 3.5%–9.4% of the phenotypic variance. In total, these QTLs explained 22.7% and 38.5% of the total phenotypic variance in 2009 and 2010, respectively.

For GT, five QTLs were detected and three (qGT3, qGT10, qGT11) of them were detected in two years (Table 4). The total contribution was 20.4% in 2009 and 40.2% in 2010. The QTL qGT10 was identified to possess largest additive effect in the region of Bt-RM1126 on chromosome 10, explaining 10.2% in 2009 and 23.3% in 2010 of the total phenotypic variation. The qGT10 allele from HH 3 had a decreasing effect on GT in both years. The other four QTLs contributed GT in a percentage of 3.3% to 5.5%.

A total of six QTL for TGW were identified. Of these, the QTLs including qTGW3, qTGW5 and qTGW10 were detected in both years. The greatest effect QTL was qTGW10, flanked by Bt-RM1126 on chromosome 10 (Fig. 2), explaining 35.5% and 32.3% of the total phenotypic variance (Table 4 and Fig. 2), respectively. The other four QTLs had relatively small effects. Generally, these six QTLs explained 71.4% in 2009 and 53.1% in 2010 of the phenotypic variance, respectively.

Epistatic QTLs and QTL-by-environment interactions among traits

For all traits, 9 pairs of loci were identified showing significant epistatic interactions at the p < 0.001 level (Table 5). One pair of epistatic interaction was detected for GL, accounting for 1.1% of the phenotypic variation. Three pairs for GW were found, explaining the phenotypic variation of 6.4%. There were four pairs for GT deciphering 4.1% of phenotypic variation. Additionally, one pair of interaction for TGW could explain 0.9% of phenotypic variation. No interaction was found between epistatic QTL and environment for LWR.

Table 5

Epistatic effects and environmental interactions of QTLs identified in the RIL population in 2009 and 2010

Traits	Ch-Inia	Interval_i	QTL	Ch-Inja	Interval_j	QTL	aib	h2aic	ajb	h2ajc	aaijd	h2aaijc	aeie	h2aeic	aeje	h2aejc	h2aeijc
GL	9-5	RM566-RM434	qGL9	10-4	RM1126-RM467				−0.20	19.9	−0.06	1.05			0.0001	0.01	0.01
GW	4-8	RM261-RM6659	qGW4	11-8	XA-RM224		0.22	2.4	0.03	5.81	−0.02	0.95					0.06
	3-8	RM15281-RM487		10-4	RM1126-RM467						−0.01	0.53					0.16
	1-10	RM449-RM595		10-4	RM1126-RM467						−0.03	4.93
GT	1-3	RM3252-RM1		10-3	Bt-RM1126	qGT10	−0.03	4.08	−0.04	12.4	−0.02	1.12	−0.0069	0.26	0.0007	0.5	0.34
	3-15	RM422-RM514	qGT4	10-3	Bt-RM1126	qGT10	0.02	5.18	−0.04	12.4	0.01	0.5	0.0071	0.28	0.0007	0.5	0.59
	4-10	RM16459-RM401		10-3	Bt-RM1126	qGT10	0.02	2.45	−0.04	12.4	0.01	0.8	0.0006	0.05	0.0007	0.5	0.14
	5-4	RM593-MRG0366	qGT5	10-3	Bt-RM1126	qGT10			−0.04	12.4	0.02	1.69			0.0007	0.5	0.47
TGW	5-2	MRG2228-RM413		11-8	XA-RM224		−0.5	6.62	0.4	3.55	−0.3	0.85	0.002	0.02

Ch-Ini and Ch-Inj represent the chromosome number-interval of the points being tested in the analysis.

ai and aj are the additive effects of the test points i and j, respectively. Positive values of ai and aj imply that the Huahui 3 genotype has a positive effect on that trait.

h2ai, h2aj, h2aaij, h2aei and h2aej are the percentages of the phenotypic variations explained by ai, aj, aaij, aei and aej respectively.

aaij is the effect of additive-by-additive interaction between points i and j; a positive value indicates that the parental two-locus genotypes have a positive effect on the traits and that the recombinants have a negative effect.

aei and aej are effects of the environmental interaction of locus i and j, respectively; a positive value implies that the effect in 2010 is larger than in 2009.

Eight QE interactions for all the traits were collectively detected. Two main effect QTLs (qGT4 and qGT10) were detected to significantly interact with environmental factors, and their interaction were also found to further interact with environment was detected for GT. In addition, seven more epistatic QTLs were detected to interact with environment. Data showed that all the detected epistatic QTLs and QTL-by-environment interactions among traits had trivial genetic effects.

Discussion

Rice grain shape or size highly correlated with yield and appearance quality, thus affecting farmers’ economic income. For rice grain size, many QTLs had been previously mapped, and most QTLs for grain size and weight were mapped on chromosomes 2, 3 and 5. In this study, a total of 27 main-effect QTLs for grain length, grain width, grain thickness and 1000-grain weight were detected on ten rice chromosomes excluding chromosomes 6 and 7. Seventeen of the 27 QTLs were reproducibly detected in two-year trials. Compared to previous researches, the QTL in the region of RM55-RM7000 on chromosome 3, controlling grain width, grain thickness and 1000-grain weight, might be the GS3 gene (Fan , Mao ). The QTL qTGW5/qGT5 in the region of RM593-MRG0366 on chromosome 5 might correlate the cloned genes qSW5/GW5 or GS5 (Li , Shomura ). The rest QTLs, including qGL1, qLWR5, qGL8, qGL9, qGL10 (qGW10, qGT10 and qTGW10), qGW11a (qTGW11a) and qGW11b (qGT11 and qTGW11b), hadn’t been reported before. Most of these QTLs had good reproducibility in two years.

In our research, two chromosomal regions around the known genes were interestingly found to be implicated with the genetic sites significantly affecting grain shape and yield. The first region is for the QTLs around the site of the known gene Xa21 on chromosome 11, which pleiotropically affected three of the five investigated traits, namely GW, GT and TGW. The QTL allele from HH 3 positively contributed the phenotypic variation. Though the data in Table 1 suggested that the introgression of Xa21 did not significantly change grain-related traits, it is obvious that the Xa21 introgression actually elevated the receptor Minghui 63 performance in most of the investigated grain traits. It is consistent with the weak effect of the QTL responsive for grain width, thickness and grain weight being detected adjacent to Xa21. The Minghui 63/Xa21, bred by introgressing Xa21 from donor rice variety IRBB21 via multiple backcrossing, contained donor genetic background less than 3.8 centimorgan (cM) fragment around Xa21 based on whole genome MAS analysis (Chen ). The slight difference between the grain performance of Minghui 63/Xa21 and Minghui63 thus might be very likely caused by the QTL around Xa21. i.e., the introgression of Xa21 gene by backcrossing from donor material into receptor Minghui 63 may contain a donor linkage drag which included QTL alleles positively affecting grain shape and yield. The results exemplified that though the donor genetic background is restrained within a very narrow chromosomal region of 3.8 cM, the risk of genetic linkage drag can in same cases remain and disturb the backcrossing improvement effect. The development and application of molecular markers much closer to objective donor gene and of sufficient number of selectable backcrossing progeny in some breeding cases become very necessary. The QTL allele from donor genome showed the positive role. The Xa21’s disease resistance may act as a good phenotypic marker for facilitating the QTL breeding application.

The second region of RM2503-RM1126 on chromosome 10 was found to be a common site for four major QTLs of qGL10, qGW10, qGT10 and qTGW10. Noticeably, all of these four QTLs steadily played very strong effects on the phenotype in different environments of the two-year trails. Such information highlighted the value and reliability of the QTL founding. In the region of RM2503-RM1126, the Bt transgene in the parent HH 3 was coincidently harbored (Fig. 2). Combined with the case that all the QTLs’ allele from HH 3 had negative additive effects on the grain size and weight, it rationalized that the grain-related QTLs in the interval of RM2503-RM1126 on chromosome 10 might be cohabited with Bt transgene insertion site, or in other word, it is just Bt transgene insertion that strongly disrupted normal function of QTL allele from HH 3 and thus raised the strong declining phenotypic performance (Table 1). The parental material of Minghui 63/Bt used in this study, identical to the transgenic rice T51-1 described by Tu , was derived from a small number of transgenic plants with Minghui 63 as receptor transformed with a l.8 kb Bt transgene fragment (Tu ). In 2009, China issued biosafety certificates for Bt rice variety Huihui No.1 and its hybrid Bt Shanyou63. The first field experiments of Bt rice event T51-1, which is the original event of Huihui No.1, were performed in China 1999 (Tu ). This study is the first time to map Bt transgene in Minghui63/Bt material on the rice chromosome 10, and firstly revealed that this Bt insertion linked and affected grain development. The possible endogenic insertional mutagenesis in this study exemplified a typical transgenic breeding problem and suggested that sufficient number of independent transgenic plants should be endeavored to acquire and selection of desirable transgenic event with expected trait performance becomes necessary. In practical transgenic breeding, it is noticeable of a typical case concerned with the position effects of the introduced transgenes on endogenous genes. This is of particular concern for Agrobacterium-mediated transgenic lines in which T-DNA is believed to have a good chance of being integrated into an active chromatin region (Zhong 2001). An ever mapped 27,621 T-DNA insertion flanked sequences and revealed that T-DNA integration frequency was generally proportionate to chromosome size, and about 45% of the T-DNAs were inserted into the genic regions and about 55% into the intergenic regions. Wang (2000) ever concluded an empirical transgenic breeding efficiency of 5%–10%, i.e., from 100 independent transgenic lines, only 5–10 transgenic desirable plants could be adequate for breeding application. It is also suggested that an initial number of 300 independent transgenic plants was necessary for a confident and guaranteed breeding improvement (Wang 2000). The mapping result of Bt transgene may also expedite the isolation of the innovative yield-related gene. Based on the Bt insertion, more analysis such as flanking analysis and enlarging segregation population for association mapping should be further performed. Breeding utilization of this QTL will also be emphasized to help improve rice grain shape and yield.