- Email: [email protected]
QTL Analysis for Seven Quality Traits of RIL Population in Japonica Rice Based on Three Genetic Statistical Models
Rice Science, 2013, 20(1): 31−38 Copyright © 2013, China National Rice Research Institute Published by Elsevier BV. All rights reserved DOI: 10.1016/S1672-6308(13)60105-5
QTL Analysis for Seven Quality Traits of RIL Population in Japonica Rice Based on Three Genetic Statistical Models LIU Qiang-ming1, JIANG Jian-hua1, 2, NIU Fu-an1, HE Ying-jun1, HONG De-lin1 (1State Key Laboratory of Crop Genetics and Germplasm Enhancement, Nanjing Agricultural University, Nanjing 210095, China; 2 Institute of Crops, Anhui Academy of Agricultural Sciences, Hefei 230031, China)
Abstract: QTL mapping for seven quality traits was conducted by using 254 recombinant inbred lines (RIL) derived from a japonica-japonica rice cross of Xiushui 79/C Bao. The seven traits investigated were grain length (GL), grain length to width ratio (LWR), chalk grain rate (CGR), chalkiness degree (CD), gelatinization temperature (GT), amylose content (AC) and gel consistency (GC) of head rice. Three mapping methods employed were composite interval mapping in QTLMapper 2.0 software based on mixed linear model (MCIM), inclusive composite interval mapping in QTL IciMapping 3.0 software based on stepwise regression linear model (ICIM) and multiple interval mapping with regression forward selection in Windows QTL Cartographer 2.5 based on multiple regression analysis (MIMR). Results showed that five QTLs with additive effect (A-QTLs) were detected by all the three methods simultaneously, two by two methods simultaneously, and 23 by only one method. Five A-QTLs were detected by MCIM, nine by ICIM and 28 by MIMR. The contribution rates of single A-QTL ranged from 0.89% to 38.07%. All the QTLs with epistatic effect (E-QTLs) detected by MIMR were not detected by the other two methods. Fourteen pairs of E-QTLs were detected by both MCIM and ICIM, and 142 pairs of E-QTLs were detected by only one method. Twenty-five pairs of E-QTLs were detected by MCIM, 141 pairs by ICIM and four pairs by MIMR. The contribution rates of single pair of E-QTL were from 2.60% to 23.78%. In the Xiu-Bao RIL population, epistatic effect played a major role in the variation of GL and CD, and additive effect was the dominant in the variation of LWR, while both epistatic effect and additive effect had equal importance in the variation of CGR, AC, GT and GC. QTLs detected by two or more methods simultaneously were highly reliable, and could be applied to improve the quality traits in japonica hybrid rice. Key words: quantitative trait locus; quality trait; genetic statistical model; japonica rice
In the last three decades, the development of japonica hybrid rice had got some achievements, yet compared to that of indica hybrid rice, it still lagged a lot. At present, the planting area of indica hybrid rice accounts for 78% of the total indica rice area in China, while it is only 3% for the proportion of japonica hybrid rice in the total national japonica rice area (Deng et al, 2006). One of the important reasons for limited popularization of japonica hybrid rice is the poor quality as a whole (Xu et al, 2010). Quality traits are quantitative traits, and their genetic features are very complex. Along with the application of various mapping populations and the perfecting of statistical methods, QTL mapping developed rapidly, and a large number of QTLs for rice quality traits have Received: 7 March 2012; Accepted: 10 September 2012 Corresponding author: HONG De-lin ([email protected])
been mapped up to now. According to the rice QTL information from Gramene (http://www.gramene.org/ QTL/), the number of detected QTLs for grain length (GL), grain length to width ratio (LWR), chalk grain rate (CGR), chalkiness degree (CD), gelatinization temperature (GT), amylose content (AC) and gel consistency (GC) were 49, 40, 28, 26, 20, 51 and 22, respectively. The reliability of these QTLs detected was low since the declaring of a QTL was judged by the calculated probability in QTL mapping and only one mapping method was applied in most of the previous research. Up to now, only a few loci related to grain quality traits have been finely mapped and a few genes, such as Wx (Okagaki et al, 1988), Alk (Gao et al, 2003), GS3 (Fan et al, 2006) and qCGR-7 (Weng et al, 2008) have been cloned. The accuracy of phenotypic data, the representative of mapping population, the precision of genetic linkage
Rice Science, Vol. 20, No. 1, 2013
32
map and the type of mapping population are influence factors for the accuracy of QTL mapping. Besides, the mapping method also intimately affects the mapping results, and an inappropriate mapping method may result in erroneous judgment or false-positive (Su et al, 2010a). Composite interval mapping (CIM) and multiple QTL mapping (MQM) are suitable to detect QTLs when the data fit the genetic model of y = μ + a1 + e, and multiple interval mapping with regression forward selection (MIMR), multiple interval mapping with forward search (MIMF) and inclusive composite interval mapping (ICIM) are suitable to detect QTLs when the data fit the genetic models of y = μ + a1 + e and y = μ + a1 + a1 + a1a2 + e, while composite interval mapping based on mixed linear model (MCIM) fits all models (Su et al, 2010b). Seven quality traits (GL, LWR, CGR, CD, GT, AC and GC) were analyzed using mixed major gene and polygene inheritance model, and results showed that epistatic effects were found in the genetic variation of those traits (Jiang et al, 2007). In the present study, three mapping methods which can analyze epistatic effect were employed to detect QTLs for GL, LWR, CGR, CD, GT, AC and GC in head rice. The three methods were composite interval mapping in QTLMapper 2.0 software (Wang et al, 1999) based on mixed linear model (MCIM), inclusive composite interval mapping in QTL IciMapping 3.0 software (Li et al, 2008) based on stepwise regression linear model (ICIM) and multiple interval mapping with regression forward selection in Windows QTL Cartographer 2.5 (Wang et al, 2007) based on multiple regression analysis (MIMR). The QTL mapping results obtained by the three genetic statistical models were compared and the highly reliable QTLs were determined. These highly reliable QTLs detected were expected to apply to the improvement of quality traits in japonica hybrid rice by marker-assisted selection.
MATERIALS AND METHODS Rice materials and trait evaluation Rice materials were Xiushui 79, a pure japonica rice variety with high-yielding, C Bao, a japonica restorer line with preferable grain quality, and 254 recombinant inbred lines (RIL) derived from the F2 of Xiushui 79/C Bao cross by the single-seed descend method. Seven quality traits tested were GL, LWR, CGR, CD, GT, AC and GC in head rice. And the detail
measuring methods were described according to Jiang et al (2007). Genetic linkage map The genetic linkage map constructed by our laboratory (Guo et al, 2009; Niu et al, 2011) spanned 1 320.2 cM with an average distance of 11.89 cM between adjacent markers (Fig. 1). Data analysis Strategy of QTL mapping with multiple models proposed by Su et al (2010b) was adopted. Three QTL mapping methods, i.e. MCIM, ICIM and MIMR were used. To declare a QTL, the LOD value threshold is 2.0 for ICIM and MIMR, whereas the threshold probability is P < 0.005 for MCIM. The naming for loci with significant additive effect followed the rules of nomenclature (McCouch et al, 2008), while for the loci with non-significant additive effect and significant epistatic effect, the first letter ‘q’ was omitted contrasting with those foregoing loci. In the present study, we defined the QTLs detected by two or three methods simultaneously as highly reliable QTLs.
RESULTS Phenotypic variation The descriptive statistics of phenotypic values of the seven traits for the Xiu-Bao RIL population and the two parents are listed in Table 1. Significant differences between the two parents were observed in all traits except GL and LWR. Transgressive segregation in either direction occurred in all traits, showing quantitativeinherited features. Except the kurtosis values of AC and GC, the absolute values of skewness and kurtosis were less than one for all traits, indicating that the phenotypic values of those traits distributed normally and were suitable for QTL analysis. QTL analysis of seven quality traits A total of 42 additive effect QTLs (A-QTLs) were detected by the three methods, and the contribution rates of a single A-QTL ranged from 0.89% to 38.07% (Table 2). The numbers of A-QTLs detected by MCIM, ICIM and MIMR were 5, 9 and 28, respectively. Five A-QTLs were detected by all the three methods simultaneously, two by two methods simultaneously, and the remaining 23 by only one method. Of the seven highly reliable A-QTLs, one, three, one, one, and
LIU Qiang-ming, et al. QTL Analysis for Quality Traits of RIL Population in Japonica Rice
33
Table 1. Basic statistics of phenotypic values of seven quality traits in Xiu-Bao RIL population and its parents. Parent RIL population Xiushui 79 (Mean ± SD) C Bao (Mean ± SD) Mean ± SD Range Skewness Kurtosis GL (mm) 5.08 ± 0.05 4.75 ± 0.04 5.02 ± 0.13 4.63−5.42 0.21 0.67 LWR 1.81 ± 0.02 1.70 ± 0.03 1.77 ± 0.06 1.63−1.96 0.05 -0.24 CGR (%) 5.63 ± 1.93 19.25 ± 2.90 35.59 ± 14.52 5.00−80.50 0.34 -0.23 CD (%) 1.65 ± 0.27 6.65 ± 1.14 9.79 ± 5.88 0.99−30.96 0.99 0.93 AC (%) 14.63 ± 0.45 16.43 ± 0.23 13.89 ± 2.80 4.58−23.67 -0.36 1.14 GT (ºC) 5.13 ± 0.27 3.89 ± 0.12 6.09 ± 0.71 3.50−7.00 -0.93 0.66 GC (mm) 74.25 ± 2.90 105.88 ± 3.26 76.05 ± 12.61 21.00−119.00 0.13 1.36 GL, Grain length; LWR, Grain length to width ratio; CGR, Chalk grain rate; CD, Chalkiness degree; AC, Amylose content; GT, Gelatinization temperature; GC, Gel consistency. Trait
Table 2. QTLs with additive effect (A-QTLs) for seven quality traits mapped in Xiu-Bao RIL population with three methods. Trait GL (mm)
Method ICIM MIMR
Locus
Chromosome
Marker interval a
Distance b
LOD
Ac
R2 (%)
P value
qGL3 3 RM3467−RM545 3.20 2.24 0.03 4.87 qGL8 8 RM1235−RM331 17.01 3.17 -0.05 11.33 qGL10 10 RM7492−RM5095 4.01 2.19 0.03 4.08 LWR MCIM qLWR11 11 RM7120−RM287 1.10 3.26 0.02 6.95 0.0001 ICIM qLWR11 11 RM7120−RM287 1.00 2.24 0.01 4.83 MIMR qLWR3 3 RM2334−RM7097 1.01 2.30 0.01 3.83 qLWR5.1 5 RM161−RM7473 10.01 2.36 0.02 5.99 qLWR5.2 5 RM7473−RM480 1.89 3.19 -0.02 4.57 qLWR11 11 RM7120−RM287 0.09 2.60 0.01 4.50 CGR (%) MCIM qCGR5.1 5 RM440−RM164 0.00 3.88 3.10 6.25 < 0.0001 qCGR9 9 RM6971−RM1013 0.50 2.94 -2.70 4.75 0.0002 ICIM qCGR2 2 RM262−RM525 0.10 2.00 -2.68 3.32 qCGR5.1 5 RM440−RM164 0.00 2.13 2.72 3.51 qCGR9 9 RM6971−RM1103 0.90 2.99 -3.32 5.24 MIMR qCGR2 2 RM525−RM2127 0.01 4.17 -3.76 4.99 qCGR5.1 5 RM440−RM164 4.01 5.00 5.46 9.54 qCGR5.2 5 RM161−RM7473 5.01 2.68 -4.67 2.30 qCGR5.3 5 RM7473−RM480 1.01 2.35 3.21 2.75 qCGR9 9 RM6971−RM1013 3.49 5.00 -4.24 6.99 qCGR11 11 RM3133−RM7120 4.89 2.77 3.63 4.23 CD (%) MIMR qCD1 1 RM1003−RM3453 0.19 2.09 3.84 11.87 qCD2 2 RM525−RM2127 0.10 2.27 -0.98 2.11 qCD8.1 8 RM331−RM3383 0.01 4.08 3.87 5.66 qCD8.2 8 RM72−RM22899 3.69 4.13 -4.14 10.91 AC (%) ICIM qAC9.1 9 RM5652−RM410 0.10 18.99 -1.75 38.07 qAC9.2 9 RM257−OSR28 0.90 12.95 1.43 25.56 MIMR qAC2 2 RM5427−RM262 1.99 2.90 1.00 4.22 qAC3 3 RM3766−RM5639 4.01 2.99 1.06 5.16 qAC9.1 9 RM410−RM257 0.01 4.40 -4.63 29.85 qAC10 10 RM171−RM1108 0.49 2.19 -2.27 0.89 GT (ºC) MCIM qGT12 12 RM7018−RM5609 0.09 3.21 0.15 5.88 0.0001 ICIM qGT12 12 RM5609−RM5479 0.20 3.06 0.16 5.30 MIMR qGT3 3 RM2334−RM7097 0.01 2.32 -0.12 2.37 qGT6.1 6 RM454−RM162 4.49 4.68 0.42 16.41 qGT6.2 6 RM162−RM5753 11.01 8.07 -0.49 22.44 qGT9 9 RM5786−RM201 2.89 2.46 -0.14 3.22 qGT12 12 RM5609−RM5479 0.01 2.49 0.12 3.55 GC (mm) MCIM qGC4 4 RM303−RM349 0.00 2.95 -2.93 5.48 0.0002 ICIM qGC4 4 RM303−RM349 0.00 2.16 -2.58 4.05 MIMR qGC4 4 RM303−RM349 5.01 2.03 -2.74 4.81 qGC6 6 RM5314−RM454 4.01 2.01 -2.87 3.58 qGC9 9 RM6838−RM3700 1.29 2.64 -2.87 3.27 GL, Grain length; LWR, Grain length to width ratio; CGR, Chalk grain rate; CD, Chalkiness degree; GT, Gelatinization temperature; AC, 2 Amylose content; GC, Gel consistency; A, Additive effect; R , Contribution rate; MCIM, Composite interval mapping in QTLMapper 2.0 software based on mixed linear model; ICIM, Inclusive composite interval mapping in QTL IciMapping 3.0 software based on stepwise regression linear model; MIMR, Multiple interval mapping with regression forward selection in Windows QTL Cartographer 2.5 based on multiple regression analysis. a Bold letters indicate the nearest marker to the putative QTL; b Distance of the nearest marker to the putative QTL; c Positive and negative values indicate the positive allele from Xiushui 79 and C Bao, respectively.
34
Rice Science, Vol. 20, No. 1, 2013
Fig. 1. Chromosome locations of highly reliable QTLs for seven quality traits mapped in Xiu-Bao RIL population. A-QTL, Additive QTL; E-QTL, Epistatic QTL; GL, Grain length; LWR, Grain length to width ratio; CGR, Chalk grain rate; CD, Chalkiness degree; AC, Amylose content; GT, Gelatinization temperature; GC, Gel consistency.
LIU Qiang-ming, et al. QTL Analysis for Quality Traits of RIL Population in Japonica Rice
one were detected for LWR, CGR, AC, GT and GC, respectively. A total of 170 pairs of epistatic effect QTLs (E-QTLs) were detected, and the contribution rates of a single pair of E-QTL were between 2.60% and 23.78% (Table 3). The numbers of E-QTLs detected by MCIM, ICIM and MIMR were 25, 141, and 4, respectively. None E-QTL was detected by all the three methods simultaneously, 14 pairs were detected by two methods simultaneously, and the remaining 142 pairs were detected by only one method. Of the 14 pairs of highly reliable E-QTLs, two, five, one, two, three and one affected GL, CGR, CD, AC, GT and GC, respectively. For GL, none, one and two A-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 2). No relatively reliable A-QTL was detected, indicating no difference existed between the alleles for GL in the two parents. Four, 22 and none pairs of E-QTL were detected by MCIM, ICIM and MIMR, respectively (Table 3). Two pairs of E-QTLs were detected by both MCIM and ICIM, the contribution rates of a single pair of E-QTL ranged from 4.23% to 7.73%. Their epistatic effects were positive, indicating that the two-locus genotypes from the same parent increased grain length. Epistatic effect was the dominant in the variation of GL in the population. For LWR, one, one and four A-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 2). qLWR11 with the positive allele coming from Xiushui 79 was detected by all the three methods, and explained 4.50% to 6.95% of the phenotypic variation. In addition, qLWR5.1 and qLWR5.2 were detected only by MIMR, and they were on the adjacent interval with opposite directions for the additive effect (Table 2). One, 10 and none pairs of E-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 3). None relatively reliable E-QTL was detected. It showed that additive effect played a major role in the variation of LWR in the population. For CGR, two, three and six A-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 2). qCGR2 was detected by ICIM and MIMR, while qCGR5.1 and qCGR9 were detected by all the three methods. At the qCGR5.1 locus, the positive allele coming from the low-value parent, Xiushui 79, while at the loci of qCGR2 and qCGR9, the alleles from C Bao increased CGR, and their contribution rates were 3.32%−9.54%. Besides, qCGR5.2 and qCGR5.3 were detected only by MIMR, and they were on the adjacent interval with opposite directions for the
35
additive effect (Table 2). Eight, 22 and none pairs of E-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 3). Five pairs of E-QTLs were detected by both MCIM and ICIM, and their contribution rates ranged from 2.60% to 8.77%. The two-locus genotypes from the same parent appeared to increase CGR for CGR2-CGR3.1, CGR3.2-CGR8.2 and qCGR5.1-CGR6.1, while recombinant two-locus combinations increased CGR for the other two pairs. It showed that both epistatic effect and additive effect had equal importance in the variation of CGR in the population. For CD, none, none and four A-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 2). None relatively reliable A-QTL was detected, indicating no difference existed between the alleles for CD in the two parents. Three, 21 and one pairs of E-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 3). For CD1.1-CD1.2, the only one pair of E-QTL detected by both MCIM and ICIM, the two-locus genotypes from the same parent appeared to increase CD, which explained 7.60% and 8.22% of the phenotypic variation. It indicated that epistatic effect was the dominant in the variation of CD in the population. For AC, none, two and four A-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 2). qAC9.1 with the positive allele coming from C Bao was detected by both ICIM and MIMR, and its contribution rate was the highest of all the detected QTLs in the present study, which reached 38.07% and 29.85% in ICIM and MIMR methods, respectively. Two, 24 and two pairs of E-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 3). Two pairs were detected by both MCIM and ICIM, and the two-locus combinations increased AC for both of them. Their contribution rates ranged from 3.52% to 23.78%. Therefore, both epistatic effect and additive effect had equal importance in the variation of AC in the population. It is noteworthy that AC3-AC7 detected in the present study by MCIM was the E-QTL with the highest contribution rate (23.78%). For GT, one, one and five A-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 2). qGT12 with the positive allele coming from Xiushui 79 was detected by all the three methods, and it explained 3.55% to 5.88% of the phenotypic variation. In addition, the loci of qGT6.1 and qGT6.2 were detected on the adjacent interval by MIMR, and their contribution rates were over 15%, but their additive
Rice Science, Vol. 20, No. 1, 2013
36
Table 3. QTLs with epistatic effect (E-QTLs) for seven quality traits mapped in Xiu-Bao RIL population with three methods. Trait
Method a
GL (mm)
MCIM
ICIM LWR CGR (%)
MCIM MCIM
ICIM
CD (%)
MCIM
AC (%)
ICIM MIMR MCIM ICIM MIMR
GT (ºC)
MCIM
ICIM
GC (mm)
MIMR MCIM ICIM
Locus i GL1 GL2 GL8 GL9.2 GL8 GL9.2 LWR9 CGR2 qCGR2 CGR3.2 CGR4 qCGR5.1 CGR6.2 CGR7 CGR8.1 CGR2 CGR3.2 CGR4 qCGR5.1 CGR8.1 CD1.1 CD3 CD7 CD1.1 qCD8.1 AC3 AC7 AC3 AC7 qAC2 qAC2 GT2 GT5.1 GT5.1 qGT6.2 GT9.1 qGT9 GT2 GT9.1 qGT9 qGT6.1 GC1 GC1
Chr
Marker interval b
Locus j
Chr
Marker interval b
1 2 8 9 8 9 9 2 2 3 4 5 6 7 8 2 3 4 5 8 1 3 7 1 8 3 7 3 7 2 2 2 5 5 6 9 9 2 9 9 6 1 1
RM84−RM1003 RM7288−RM5356 RM6948−RM433 RM410−RM257 RM6948−RM433 RM5652−RM410 RM410−RM257 RM5427−RM262 RM525−RM2127 RM7097−RM448 RM3288−RM303 RM440−RM164 RM5753−RM345 RM346−RM336 RM3383−RM72 RM5427−RM262 RM7097−RM448 RM3288−RM303 RM164−RM305 RM72−RM22899 RM265−RM6696 RM448−RM8277 RM346−RM336 RM6696−RM3482 RM331−RM3383 RM218−RM232 RM180−RM214 RM218−RM232 RM214−RM542 RM5427−RM262 RM5427−RM262 RM525−RM2127 RM440−RM164 RM164−RM305 RM5753−RM345 RM6839−RM3700 OSR28−RM5786 RM525−RM2127 RM6839−RM3700 RM257−OSR28 RM454−RM162 RM84−RM1003 RM84−RM1003
GL9.1 GL7 GL9.1 GL11 GL9.1 GL11 LWR11 CGR3.1 CGR5 CGR8.2 CGR8.3 CGR6.1 CGR10 CGR9 qCGR11 CGR3.1 CGR8.2 CGR8.2 CGR6.1 qCGR11 CD1.2 CD8 CD9 CD1.2 qCD8.2 AC7 AC11.1 AC7 AC11.1 qAC3 AC11.2 GT10.1 GT5.2 GT10.2 GT11 GT9.2 qGT12 GT10.1 GT9.2 qGT12 qGT6.2 GC12 GC12
9 7 9 11 9 11 11 3 5 8 8 6 10 9 11 3 8 8 6 11 1 8 9 1 8 7 11 7 11 3 11 10 5 10 11 9 12 10 9 12 6 12 12
RM6570−RM5652 RM542−RM418 RM6570−RM5652 RM3133−RM7120 RM6570−RM5652 RM3133−RM7120 RM3133−RM7120 RM3467−RM545 RM1182−RM405 RM6948−RM433 RM264−RM6948 RM454−RM162 RM7492−RM5095 RM285−RM8206 RM7120−RM287 RM3467−RM545 RM6948−RM433 RM6948−RM433 RM454−RM162 RM7120−RM287 RM8105−RM84 RM6948−RM433 RM285−RM8206 RM495−RM8105 RM72−RM22899 RM125−RM180 RM287−RM21 RM125−RM180 RM287−RM21 RM3766−RM5639 RM5349−RM206 RM5095−RM311 RM161−RM7473 RM5629−RM171 RM3133−RM7120 RM6570−RM5652 RM5479−RM1227 RM7492−RM5095 RM6570−RM5652 RM5479−RM1227 RM162−RM5753 RM5609−RM5479 RM5479−RM1227
LOD AAij c R2 (%) P value 2.99 3.47 3.71 4.71 2.16 4.71 3.62 5.30 3.67 3.17 7.03 5.53 3.79 4.12 2.90 6.24 2.05 8.64 3.92 2.40 5.16 3.11 4.64 2.81 4.07 7.36 3.35 4.42 2.17 2.00 2.42 3.47 3.54 2.82 4.14 4.47 3.94 4.79 4.33 4.81 4.99 3.29 3.50
-0.02 3.81 0.03 5.72 0.03 6.73 0.03 7.73 0.03 4.23 0.04 7.73 0.02 7.79 4.03 7.01 2.67 3.08 3.24 4.54 -4.51 8.77 3.25 4.57 -3.24 4.54 -2.54 2.79 -2.73 3.21 4.28 7.76 2.37 2.60 -4.26 8.32 2.97 4.07 -2.37 2.72 1.66 7.60 1.33 4.88 -1.64 7.46 1.91 8.22 -3.25 9.52 -1.53 23.78 -0.72 5.30 -1.21 10.58 -0.54 3.52 1.00 4.22 0.59 4.06 -0.19 6.48 0.16 5.11 0.14 3.49 0.19 6.58 0.23 9.78 -0.14 3.72 -0.22 9.31 0.25 7.87 -0.18 5.77 0.24 7.78 -3.00 5.60 -4.04 11.20
0.001 0.0001 0.0001 < 0.0001 < 0.0001 0.0016 0.0004 < 0.0001 0.0002 < 0.0001 0.0009 0.0027 0.0001 0.0008 < 0.0001 < 0.0001 0.0002 0.0001 0.0002 0.0006 0.0001 0.0001 0.0003 0.0003 -
GL, Grain length; LWR, Grain length to width ratio; CGR, Chalk grain rate; CD, Chalkiness degree; GT, Gelatinization temperature; AC, Amylose content; GC, Gel consistency; Chr, Chromosome; AA, Epistatic effect; R2, Contribution rate; MCIM, Composite interval mapping in QTLMapper 2.0 software based on mixed linear model; ICIM, Inclusive composite interval mapping in QTL IciMapping 3.0 software based on stepwise regression linear model; MIMR, Multiple interval mapping with regression forward selection in Windows QTL Cartographer 2.5 based on multiple regression analysis. a Shows the E-QTLs both detected by ICIM and MCIM (or MIMR) for each trait in the table; b Bold letters indicate the nearest marker to the putative QTL; c Direction of effect: A positive AA value implies that the two-locus genotypes from the same parent Xiushui 79 or C Bao take the positive effects, while the two-locus genotypes of recombination from the parent Xiushui 79 and C Bao take the negative effects, the case of negative AA values is just the opposite.
effect directions were opposite (Table 2). Six, 16 and one pairs of E-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 3). Three pairs were both detected by MCIM and ICIM, and their contribution rates ranged from 3.72% to 9.78%. The two-locus genotypes from the same parent appeared to increase GT for GT9.1-GT9.2, while recombinant
two-locus combinations increased GT for the other two pairs. It indicated that both epistatic effect and additive effect had equal importance in the variation of GT in the population. For GC, one, one and three A-QTLs were detected by MCIM, ICIM, and MIMR, respectively (Table 2). qGC4 with the positive allele coming from C Bao was
LIU Qiang-ming, et al. QTL Analysis for Quality Traits of RIL Population in Japonica Rice
detected by all the three methods, and it explained 4.05% to 5.48% of the phenotypic variation. One, 26 and none pairs of E-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 3). For GC1GC12, the only one pair of E-QTL detected by both MCIM and ICIM, the two-locus genotypes from different parents appeared to increase GC. The contribution rates were 5.60% and 11.20%, respectively. It showed that both epistatic effect and additive effect had equal importance in the variation of GC in the population.
DISCUSSION In this study, all the five A-QTLs detected by MCIM were also detected by both ICIM and MIMR, and the position, additive effect and contribution rate of the same QTL detected by different methods were well coincident, so the A-QTLs detected by MCIM had high reliability. However, two A-QTLs with nonsignificant additive effect in MCIM were both detected by ICIM and MIMR, and the contribution rates of qAC9.1 reached 38.07% and 29.85%, respectively. It indicated that some false-negative results were obtained in the detection of A-QTL by MCIM. Though only nine A-QTLs were detected by ICIM, two were nonsignificant in the other two methods, thus some falsepositive A-QTLs existed in ICIM. A large number of A-QTLs were detected by MIMR, but most of them could be detected by neither MCIM nor ICIM, indicating MIMR had the most powerful ability to detect A-QTL, meanwhile, lots of false-positive A-QTLs also occurred. In addition, three pairs of A-QTLs, namely qLWR5.1 and qLWR5.2, qCGR5.2 and qCGR5.3, and qGT6.1 and qGT6.2, with adjacent interval and opposite direction for additive effect were detected by MIMR, but none of them was significant in MCIM and ICIM, which might due to that the positive effect was offset by the negative effect, indicating MIMR had a stronger ability in precise mapping. All the E-QTLs detected by MIMR were the interaction among A-QTLs, but they were detected by neither MCIM nor ICIM, so there were some deficiencies in the analysis of the interaction among loci with nonsignificant additive effect by MIMR. Many pairs of E-QTLs were detected by MCIM or ICIM individually, but only a few of them could be detected by both methods (MCIM and ICIM). For example, 141 pairs of E-QTLs were detected by ICIM, but only 14 pairs of them could be identified by both methods. These results indicated that a lot of false-positive E-QTLs
37
were obtained by MCIM and ICIM individually. The reliability could be improved by calibrating the E-QTLs detected by both MCIM and ICIM methods. In conclusion, the three mapping methods based on different genetic statistical models have their pros and cons, so multiple methods based on different genetic statistical models should be employed to analyze QTLs, and calibrating the QTLs detected by different methods synchronously to improve the reliability of the results. qLWR11, qCGR2, qCGR5.1, qCGR9, qAC9.1, qGT12 and qGC4 were the seven highly reliable A-QTLs. The primer sequences of the pre and post markers of the seven A-QTLs were analyzed by BLAST (http:// www.ncbi.nlm.nih.gov) and located on the published PAC-BAC clone physical map of rice (http://rgp.dna. affrc.go.jp/E/IRGSP/download.html), and then compared to the previous research (Bao et al, 2003; Liu et al, 2007; Zhou et al, 2009; Gao et al, 2011) was conducted. No QTL was reported around qLWR11, qCGR2, qCGR9, qAC9.1 and qGC4 for the corresponding traits. So these five A-QTLs were new loci. qCGR5.1 shared an identical interval with qPCG5, which was reported by Liu et al (2007), thus they were presumed to be the same locus. One QTL for GT was reported on chromosome 12 (http://www.gramene.org/qtl/), and the flanking markers were RM309 and RM6837. The adjacent markers of qGT12 detected in the present study were RM5609 and RM5479. Compared the order of the four SSR markers through the published information on Gramene, we found that the interval of RM5609−RM5479 was completely involved in the interval of RM309−RM6837. Therefore, the two sites were treated as one site. In the present study, only highly reliable E-QTLs were detected for GL and CD, and only highly reliable A-QTL was detected for LWR, while both two kinds of QTLs were detected for the other four traits. So, in the Xiu-Bao RIL population, epistatic effect played a major role in the variation of GL and CD, and additive effect was the dominant in the variation of LWR, while both epistatic effect and additive effect had equal importance in the variation of the rest four traits. Fourteen pairs of highly reliable E-QTLs were detected. Of them, 11 pairs were the interaction among sites with non-significant additive effect, and two pairs for CGR were the interaction between A-QTLs and sites with non-significant additive effect, and only one pair for GT were the interaction between two A-QTLs. These suggested that the majority of E-QTLs were the
Rice Science, Vol. 20, No. 1, 2013
38
interaction among sites with non-significant additive effect for quality traits. This result is consistent with those obtained by Lei et al (2008) and Jiang et al (2009). Therefore, it is possible to polymerize the elite alleles of highly reliable A-QTLs, which had no interaction to other sites, by marker-assistant selection to improve the grain quality of japonica hybrid rice. Moderate amylose content was preferred in rice (Huang et al, 2011). QTL qAC9.1 detected by MIMR, with the contribution rate of 29.85% and the additive effect of -4.63%, was located closely to the marker RM410 (Table 2). Substituting C Bao allele with Xiushui 79 allele at the RM410 site would reduce AC by 4.63%. Contrarily, substituting Xiushui 79 allele with the C Bao allele would increase AC by 4.63%. Therefore, we can select the allele at the RM410 site to improve amylose content of japonica hybrid rice according to the objective of breeding. In addition, we found that qAC9.1 was located on the BAC clone OJ1005_D12, which is on the long arm of chromosome 9, with a distance of 17.99 Mb to the end of short arm and 5.29 Mb to the end of long arm, through analysis of primer sequences of RM410 and RM257 by BLAST.
ACKNOWLEDGEMENTS This work was supported by the National High Technology Research and Development Program of China (Grant No. 2010AA101301); the Program of Introducing International Advanced Agricultural Science and Technology in China (Grant No. 2006-G8[4]-31-1) and the Program of Science-Technology Basis and Conditional Platform in China (Grant No. 505005).
REFERENCES Bao J S, Corke H, He P, Zhu L H. 2003. Analysis of quantitative trait loci for starch properties of rice based on an RIL population. Acta Bot Sin, 45(8): 986–994. Deng H F, He Q, Shu F, Zhang W H, Yang F, Jin Y H, Dong L, Xie H. 2006. Status and technical strategy on the development of japonica hybrid rice in China. Hybrid Rice, 21(1): 1–6. (in Chinese with English abstract) Fan C C, Xing Y Z, Mao H L, Lu T T, Han B, Xu C G. 2006. GS3, a major QTL for grain length and weight and minor QTL for grain width and thickness in rice, encodes a putative transmembrane protein. Theor Appl Genet, 112(6): 1164–1171. Gao Z Q, Zhan X D, Liang Y S, Cheng S H, Cao L Y. 2011. Progress on genetics of rice grain shape trait and its related gene mapping and cloning. Hereditas, 33(3): 1–8. (in Chinese with English abstract) Gao Z Y, Zeng D L, Cui X, Zhou Y H, Yan M X, Huang D N, Li J Y, Qian Q. 2003. Map-based cloning and sequence analysis of a gene, ALK, responsible for gelatinization temperature in rice
(Oryza sativa L.). Sci Chin: Ser C, 33(6): 481–487. (in Chinese) Guo Y, Cheng B S, Hong D L. 2009. Construction of SSR linkage map and snalysis of QTL for rolled leaf of restorer line in japonica rice (Oryza sativa L.). Chin J Rice Sci, 23(3): 245–251. (in Chinese with English abstract) Huang D C, Liang K, Sun C, Jiang J H, Shi M L, Dai J, Xie H, Zhao K M, Nguyen P T, Ma W X, Hong D L. 2011. Identification of marker genotypes associated with elite combining ability for quality traits in parents of hybrid japonica rice (Oryza sativa L.). Acta Agron Sin, 37(3): 405–414. (in Chinese with English abstract) Jiang J H, Guo Y, Chen X G, Xu H M, Hou Y J, Hong D L. 2007. The correlation between panicle angle and rice quality and genetic analysis on rice quality in japonica rice (Oryza sativa L.). Hereditas, 29(6): 714–724. (in Chinese with English abstract) Jiang L R, Wang W, Huang J X, Huang R Y, Zheng J S, Huang Y M, Wang H C. 2009. Analysis of epistatic and QE interaction effects of QTLs for grain shape in rice. Mol Plant Breeding, 7(4): 690–698. (in Chinese with English abstract) Lei D Y, Xie F M, Xu J L, Chen L Y. 2008. QTLs mapping and epistasis analysis for grain shape and chalkiness degree of rice. Chin J Rice Sci, 22(3): 255–260. (in Chinese with English abstract) Li H H, Ribaut J M, Li Z L, Wang J K. 2008. Inclusive composite interval mapping (ICIM) for dysgenic epistasis of quantitative traits in biparental populations. Theor Appl Genet, 116: 243– 260. Liu J F, Kui L M, Zhu Z F, Tan L B, Wang G J, Li Q W, Shu J H, Sun C Q. 2007. Identification of QTLs associated with processing quality and appearance quality of common wild rice (Oryza rufipogon Griff.). J Agric Biol, 15(1): 90–96. (in Chinese with English abstract) MaCoach S R. 2008. Gene nomenclature system for rice. Rice, 1: 72–84. Niu F A. 2011. Genetic segregation analysis and QTL mapping as well as association analysis for panicle angle trait in japonica rice (Oryza sativa L.). [Master Dissertation] Nanjing: Nanjing Agricultural University: 25–27. (in Chinese with English abstract) Okagaki R J, Wessler S R. 1988. Comparison of non-mutant and mutant waxy genes in rice and maize. Genetics, 120: 1137−1143. Su C F, Lu W G, Zhao T J, Gai J Y. 2010a. Verification and finemapping of QTLs conferring days to flowering in soybean using residual heterozygous lines. Chin Sci Bull, 55(6): 499–508. Su C F, Zhao T J, Gai J Y. 2010b. Simulation comparisons of effectiveness among QTL mapping procedures of different statistical genetic models. Acta Agron Sin, 36(7): 1100–1107. (in Chinese with English abstract) Wang D L, Zhu J, Li Z K. 1999. Mapping QTL with epistatic effects and QTL × environment interactions by mixed linear model approaches. Theor Appl Genet, 99: 1255–1264. Wang S, Basten C J, Zeng Z B. 2007. Windows QTL Cartographer 2.5. Raleigh, NC: Department of Statistics, North Carolina State University (http://statgen.ncsu.edu/qtlcart/WQTLCart.htm). Weng J F, Gu S H, Wan X Y, Gao H, Guo T, Su N, Lei C L, Zhang X, Cheng Z J, Guo X P, Wang J L, Jiang L, Zhai H Q, Wan J M. 2008. Isolation and initial characterization of GW5, a major QTL associated with rice grain width and weight. Cell Res, 18(12): 1199–1209. Xu J Q, Zhang D Y, Zhang Y, Li J Y, Guo H, Xu D Y, Shao F K, Xu Y. 2010. Practice and thinking on japonica hybrid rice breeding. Hybrid Rice, 25(Special Collection): 85–87. (in Chinese) Zhou L J, Liu X, Jiang L, Zheng L N, Chen L M, Liu S J, Zhai H Q, Wan J M. 2009. Analysis of QTL and GE effects on PGWC in rice (Oryza sativa L.) using CSSL and BIL populations. Sci Agric Sin, 42(4): 1129–1135. (in Chinese with English abstract)
QTL Analysis for Seven Quality Traits of RIL Population in Japonica Rice Based on Three Genetic Statistical Models LIU Qiang-ming1, JIANG Jian-hua1, 2, NIU Fu-an1, HE Ying-jun1, HONG De-lin1 (1State Key Laboratory of Crop Genetics and Germplasm Enhancement, Nanjing Agricultural University, Nanjing 210095, China; 2 Institute of Crops, Anhui Academy of Agricultural Sciences, Hefei 230031, China)
Abstract: QTL mapping for seven quality traits was conducted by using 254 recombinant inbred lines (RIL) derived from a japonica-japonica rice cross of Xiushui 79/C Bao. The seven traits investigated were grain length (GL), grain length to width ratio (LWR), chalk grain rate (CGR), chalkiness degree (CD), gelatinization temperature (GT), amylose content (AC) and gel consistency (GC) of head rice. Three mapping methods employed were composite interval mapping in QTLMapper 2.0 software based on mixed linear model (MCIM), inclusive composite interval mapping in QTL IciMapping 3.0 software based on stepwise regression linear model (ICIM) and multiple interval mapping with regression forward selection in Windows QTL Cartographer 2.5 based on multiple regression analysis (MIMR). Results showed that five QTLs with additive effect (A-QTLs) were detected by all the three methods simultaneously, two by two methods simultaneously, and 23 by only one method. Five A-QTLs were detected by MCIM, nine by ICIM and 28 by MIMR. The contribution rates of single A-QTL ranged from 0.89% to 38.07%. All the QTLs with epistatic effect (E-QTLs) detected by MIMR were not detected by the other two methods. Fourteen pairs of E-QTLs were detected by both MCIM and ICIM, and 142 pairs of E-QTLs were detected by only one method. Twenty-five pairs of E-QTLs were detected by MCIM, 141 pairs by ICIM and four pairs by MIMR. The contribution rates of single pair of E-QTL were from 2.60% to 23.78%. In the Xiu-Bao RIL population, epistatic effect played a major role in the variation of GL and CD, and additive effect was the dominant in the variation of LWR, while both epistatic effect and additive effect had equal importance in the variation of CGR, AC, GT and GC. QTLs detected by two or more methods simultaneously were highly reliable, and could be applied to improve the quality traits in japonica hybrid rice. Key words: quantitative trait locus; quality trait; genetic statistical model; japonica rice
In the last three decades, the development of japonica hybrid rice had got some achievements, yet compared to that of indica hybrid rice, it still lagged a lot. At present, the planting area of indica hybrid rice accounts for 78% of the total indica rice area in China, while it is only 3% for the proportion of japonica hybrid rice in the total national japonica rice area (Deng et al, 2006). One of the important reasons for limited popularization of japonica hybrid rice is the poor quality as a whole (Xu et al, 2010). Quality traits are quantitative traits, and their genetic features are very complex. Along with the application of various mapping populations and the perfecting of statistical methods, QTL mapping developed rapidly, and a large number of QTLs for rice quality traits have Received: 7 March 2012; Accepted: 10 September 2012 Corresponding author: HONG De-lin ([email protected])
been mapped up to now. According to the rice QTL information from Gramene (http://www.gramene.org/ QTL/), the number of detected QTLs for grain length (GL), grain length to width ratio (LWR), chalk grain rate (CGR), chalkiness degree (CD), gelatinization temperature (GT), amylose content (AC) and gel consistency (GC) were 49, 40, 28, 26, 20, 51 and 22, respectively. The reliability of these QTLs detected was low since the declaring of a QTL was judged by the calculated probability in QTL mapping and only one mapping method was applied in most of the previous research. Up to now, only a few loci related to grain quality traits have been finely mapped and a few genes, such as Wx (Okagaki et al, 1988), Alk (Gao et al, 2003), GS3 (Fan et al, 2006) and qCGR-7 (Weng et al, 2008) have been cloned. The accuracy of phenotypic data, the representative of mapping population, the precision of genetic linkage
Rice Science, Vol. 20, No. 1, 2013
32
map and the type of mapping population are influence factors for the accuracy of QTL mapping. Besides, the mapping method also intimately affects the mapping results, and an inappropriate mapping method may result in erroneous judgment or false-positive (Su et al, 2010a). Composite interval mapping (CIM) and multiple QTL mapping (MQM) are suitable to detect QTLs when the data fit the genetic model of y = μ + a1 + e, and multiple interval mapping with regression forward selection (MIMR), multiple interval mapping with forward search (MIMF) and inclusive composite interval mapping (ICIM) are suitable to detect QTLs when the data fit the genetic models of y = μ + a1 + e and y = μ + a1 + a1 + a1a2 + e, while composite interval mapping based on mixed linear model (MCIM) fits all models (Su et al, 2010b). Seven quality traits (GL, LWR, CGR, CD, GT, AC and GC) were analyzed using mixed major gene and polygene inheritance model, and results showed that epistatic effects were found in the genetic variation of those traits (Jiang et al, 2007). In the present study, three mapping methods which can analyze epistatic effect were employed to detect QTLs for GL, LWR, CGR, CD, GT, AC and GC in head rice. The three methods were composite interval mapping in QTLMapper 2.0 software (Wang et al, 1999) based on mixed linear model (MCIM), inclusive composite interval mapping in QTL IciMapping 3.0 software (Li et al, 2008) based on stepwise regression linear model (ICIM) and multiple interval mapping with regression forward selection in Windows QTL Cartographer 2.5 (Wang et al, 2007) based on multiple regression analysis (MIMR). The QTL mapping results obtained by the three genetic statistical models were compared and the highly reliable QTLs were determined. These highly reliable QTLs detected were expected to apply to the improvement of quality traits in japonica hybrid rice by marker-assisted selection.
MATERIALS AND METHODS Rice materials and trait evaluation Rice materials were Xiushui 79, a pure japonica rice variety with high-yielding, C Bao, a japonica restorer line with preferable grain quality, and 254 recombinant inbred lines (RIL) derived from the F2 of Xiushui 79/C Bao cross by the single-seed descend method. Seven quality traits tested were GL, LWR, CGR, CD, GT, AC and GC in head rice. And the detail
measuring methods were described according to Jiang et al (2007). Genetic linkage map The genetic linkage map constructed by our laboratory (Guo et al, 2009; Niu et al, 2011) spanned 1 320.2 cM with an average distance of 11.89 cM between adjacent markers (Fig. 1). Data analysis Strategy of QTL mapping with multiple models proposed by Su et al (2010b) was adopted. Three QTL mapping methods, i.e. MCIM, ICIM and MIMR were used. To declare a QTL, the LOD value threshold is 2.0 for ICIM and MIMR, whereas the threshold probability is P < 0.005 for MCIM. The naming for loci with significant additive effect followed the rules of nomenclature (McCouch et al, 2008), while for the loci with non-significant additive effect and significant epistatic effect, the first letter ‘q’ was omitted contrasting with those foregoing loci. In the present study, we defined the QTLs detected by two or three methods simultaneously as highly reliable QTLs.
RESULTS Phenotypic variation The descriptive statistics of phenotypic values of the seven traits for the Xiu-Bao RIL population and the two parents are listed in Table 1. Significant differences between the two parents were observed in all traits except GL and LWR. Transgressive segregation in either direction occurred in all traits, showing quantitativeinherited features. Except the kurtosis values of AC and GC, the absolute values of skewness and kurtosis were less than one for all traits, indicating that the phenotypic values of those traits distributed normally and were suitable for QTL analysis. QTL analysis of seven quality traits A total of 42 additive effect QTLs (A-QTLs) were detected by the three methods, and the contribution rates of a single A-QTL ranged from 0.89% to 38.07% (Table 2). The numbers of A-QTLs detected by MCIM, ICIM and MIMR were 5, 9 and 28, respectively. Five A-QTLs were detected by all the three methods simultaneously, two by two methods simultaneously, and the remaining 23 by only one method. Of the seven highly reliable A-QTLs, one, three, one, one, and
LIU Qiang-ming, et al. QTL Analysis for Quality Traits of RIL Population in Japonica Rice
33
Table 1. Basic statistics of phenotypic values of seven quality traits in Xiu-Bao RIL population and its parents. Parent RIL population Xiushui 79 (Mean ± SD) C Bao (Mean ± SD) Mean ± SD Range Skewness Kurtosis GL (mm) 5.08 ± 0.05 4.75 ± 0.04 5.02 ± 0.13 4.63−5.42 0.21 0.67 LWR 1.81 ± 0.02 1.70 ± 0.03 1.77 ± 0.06 1.63−1.96 0.05 -0.24 CGR (%) 5.63 ± 1.93 19.25 ± 2.90 35.59 ± 14.52 5.00−80.50 0.34 -0.23 CD (%) 1.65 ± 0.27 6.65 ± 1.14 9.79 ± 5.88 0.99−30.96 0.99 0.93 AC (%) 14.63 ± 0.45 16.43 ± 0.23 13.89 ± 2.80 4.58−23.67 -0.36 1.14 GT (ºC) 5.13 ± 0.27 3.89 ± 0.12 6.09 ± 0.71 3.50−7.00 -0.93 0.66 GC (mm) 74.25 ± 2.90 105.88 ± 3.26 76.05 ± 12.61 21.00−119.00 0.13 1.36 GL, Grain length; LWR, Grain length to width ratio; CGR, Chalk grain rate; CD, Chalkiness degree; AC, Amylose content; GT, Gelatinization temperature; GC, Gel consistency. Trait
Table 2. QTLs with additive effect (A-QTLs) for seven quality traits mapped in Xiu-Bao RIL population with three methods. Trait GL (mm)
Method ICIM MIMR
Locus
Chromosome
Marker interval a
Distance b
LOD
Ac
R2 (%)
P value
qGL3 3 RM3467−RM545 3.20 2.24 0.03 4.87 qGL8 8 RM1235−RM331 17.01 3.17 -0.05 11.33 qGL10 10 RM7492−RM5095 4.01 2.19 0.03 4.08 LWR MCIM qLWR11 11 RM7120−RM287 1.10 3.26 0.02 6.95 0.0001 ICIM qLWR11 11 RM7120−RM287 1.00 2.24 0.01 4.83 MIMR qLWR3 3 RM2334−RM7097 1.01 2.30 0.01 3.83 qLWR5.1 5 RM161−RM7473 10.01 2.36 0.02 5.99 qLWR5.2 5 RM7473−RM480 1.89 3.19 -0.02 4.57 qLWR11 11 RM7120−RM287 0.09 2.60 0.01 4.50 CGR (%) MCIM qCGR5.1 5 RM440−RM164 0.00 3.88 3.10 6.25 < 0.0001 qCGR9 9 RM6971−RM1013 0.50 2.94 -2.70 4.75 0.0002 ICIM qCGR2 2 RM262−RM525 0.10 2.00 -2.68 3.32 qCGR5.1 5 RM440−RM164 0.00 2.13 2.72 3.51 qCGR9 9 RM6971−RM1103 0.90 2.99 -3.32 5.24 MIMR qCGR2 2 RM525−RM2127 0.01 4.17 -3.76 4.99 qCGR5.1 5 RM440−RM164 4.01 5.00 5.46 9.54 qCGR5.2 5 RM161−RM7473 5.01 2.68 -4.67 2.30 qCGR5.3 5 RM7473−RM480 1.01 2.35 3.21 2.75 qCGR9 9 RM6971−RM1013 3.49 5.00 -4.24 6.99 qCGR11 11 RM3133−RM7120 4.89 2.77 3.63 4.23 CD (%) MIMR qCD1 1 RM1003−RM3453 0.19 2.09 3.84 11.87 qCD2 2 RM525−RM2127 0.10 2.27 -0.98 2.11 qCD8.1 8 RM331−RM3383 0.01 4.08 3.87 5.66 qCD8.2 8 RM72−RM22899 3.69 4.13 -4.14 10.91 AC (%) ICIM qAC9.1 9 RM5652−RM410 0.10 18.99 -1.75 38.07 qAC9.2 9 RM257−OSR28 0.90 12.95 1.43 25.56 MIMR qAC2 2 RM5427−RM262 1.99 2.90 1.00 4.22 qAC3 3 RM3766−RM5639 4.01 2.99 1.06 5.16 qAC9.1 9 RM410−RM257 0.01 4.40 -4.63 29.85 qAC10 10 RM171−RM1108 0.49 2.19 -2.27 0.89 GT (ºC) MCIM qGT12 12 RM7018−RM5609 0.09 3.21 0.15 5.88 0.0001 ICIM qGT12 12 RM5609−RM5479 0.20 3.06 0.16 5.30 MIMR qGT3 3 RM2334−RM7097 0.01 2.32 -0.12 2.37 qGT6.1 6 RM454−RM162 4.49 4.68 0.42 16.41 qGT6.2 6 RM162−RM5753 11.01 8.07 -0.49 22.44 qGT9 9 RM5786−RM201 2.89 2.46 -0.14 3.22 qGT12 12 RM5609−RM5479 0.01 2.49 0.12 3.55 GC (mm) MCIM qGC4 4 RM303−RM349 0.00 2.95 -2.93 5.48 0.0002 ICIM qGC4 4 RM303−RM349 0.00 2.16 -2.58 4.05 MIMR qGC4 4 RM303−RM349 5.01 2.03 -2.74 4.81 qGC6 6 RM5314−RM454 4.01 2.01 -2.87 3.58 qGC9 9 RM6838−RM3700 1.29 2.64 -2.87 3.27 GL, Grain length; LWR, Grain length to width ratio; CGR, Chalk grain rate; CD, Chalkiness degree; GT, Gelatinization temperature; AC, 2 Amylose content; GC, Gel consistency; A, Additive effect; R , Contribution rate; MCIM, Composite interval mapping in QTLMapper 2.0 software based on mixed linear model; ICIM, Inclusive composite interval mapping in QTL IciMapping 3.0 software based on stepwise regression linear model; MIMR, Multiple interval mapping with regression forward selection in Windows QTL Cartographer 2.5 based on multiple regression analysis. a Bold letters indicate the nearest marker to the putative QTL; b Distance of the nearest marker to the putative QTL; c Positive and negative values indicate the positive allele from Xiushui 79 and C Bao, respectively.
34
Rice Science, Vol. 20, No. 1, 2013
Fig. 1. Chromosome locations of highly reliable QTLs for seven quality traits mapped in Xiu-Bao RIL population. A-QTL, Additive QTL; E-QTL, Epistatic QTL; GL, Grain length; LWR, Grain length to width ratio; CGR, Chalk grain rate; CD, Chalkiness degree; AC, Amylose content; GT, Gelatinization temperature; GC, Gel consistency.
LIU Qiang-ming, et al. QTL Analysis for Quality Traits of RIL Population in Japonica Rice
one were detected for LWR, CGR, AC, GT and GC, respectively. A total of 170 pairs of epistatic effect QTLs (E-QTLs) were detected, and the contribution rates of a single pair of E-QTL were between 2.60% and 23.78% (Table 3). The numbers of E-QTLs detected by MCIM, ICIM and MIMR were 25, 141, and 4, respectively. None E-QTL was detected by all the three methods simultaneously, 14 pairs were detected by two methods simultaneously, and the remaining 142 pairs were detected by only one method. Of the 14 pairs of highly reliable E-QTLs, two, five, one, two, three and one affected GL, CGR, CD, AC, GT and GC, respectively. For GL, none, one and two A-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 2). No relatively reliable A-QTL was detected, indicating no difference existed between the alleles for GL in the two parents. Four, 22 and none pairs of E-QTL were detected by MCIM, ICIM and MIMR, respectively (Table 3). Two pairs of E-QTLs were detected by both MCIM and ICIM, the contribution rates of a single pair of E-QTL ranged from 4.23% to 7.73%. Their epistatic effects were positive, indicating that the two-locus genotypes from the same parent increased grain length. Epistatic effect was the dominant in the variation of GL in the population. For LWR, one, one and four A-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 2). qLWR11 with the positive allele coming from Xiushui 79 was detected by all the three methods, and explained 4.50% to 6.95% of the phenotypic variation. In addition, qLWR5.1 and qLWR5.2 were detected only by MIMR, and they were on the adjacent interval with opposite directions for the additive effect (Table 2). One, 10 and none pairs of E-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 3). None relatively reliable E-QTL was detected. It showed that additive effect played a major role in the variation of LWR in the population. For CGR, two, three and six A-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 2). qCGR2 was detected by ICIM and MIMR, while qCGR5.1 and qCGR9 were detected by all the three methods. At the qCGR5.1 locus, the positive allele coming from the low-value parent, Xiushui 79, while at the loci of qCGR2 and qCGR9, the alleles from C Bao increased CGR, and their contribution rates were 3.32%−9.54%. Besides, qCGR5.2 and qCGR5.3 were detected only by MIMR, and they were on the adjacent interval with opposite directions for the
35
additive effect (Table 2). Eight, 22 and none pairs of E-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 3). Five pairs of E-QTLs were detected by both MCIM and ICIM, and their contribution rates ranged from 2.60% to 8.77%. The two-locus genotypes from the same parent appeared to increase CGR for CGR2-CGR3.1, CGR3.2-CGR8.2 and qCGR5.1-CGR6.1, while recombinant two-locus combinations increased CGR for the other two pairs. It showed that both epistatic effect and additive effect had equal importance in the variation of CGR in the population. For CD, none, none and four A-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 2). None relatively reliable A-QTL was detected, indicating no difference existed between the alleles for CD in the two parents. Three, 21 and one pairs of E-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 3). For CD1.1-CD1.2, the only one pair of E-QTL detected by both MCIM and ICIM, the two-locus genotypes from the same parent appeared to increase CD, which explained 7.60% and 8.22% of the phenotypic variation. It indicated that epistatic effect was the dominant in the variation of CD in the population. For AC, none, two and four A-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 2). qAC9.1 with the positive allele coming from C Bao was detected by both ICIM and MIMR, and its contribution rate was the highest of all the detected QTLs in the present study, which reached 38.07% and 29.85% in ICIM and MIMR methods, respectively. Two, 24 and two pairs of E-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 3). Two pairs were detected by both MCIM and ICIM, and the two-locus combinations increased AC for both of them. Their contribution rates ranged from 3.52% to 23.78%. Therefore, both epistatic effect and additive effect had equal importance in the variation of AC in the population. It is noteworthy that AC3-AC7 detected in the present study by MCIM was the E-QTL with the highest contribution rate (23.78%). For GT, one, one and five A-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 2). qGT12 with the positive allele coming from Xiushui 79 was detected by all the three methods, and it explained 3.55% to 5.88% of the phenotypic variation. In addition, the loci of qGT6.1 and qGT6.2 were detected on the adjacent interval by MIMR, and their contribution rates were over 15%, but their additive
Rice Science, Vol. 20, No. 1, 2013
36
Table 3. QTLs with epistatic effect (E-QTLs) for seven quality traits mapped in Xiu-Bao RIL population with three methods. Trait
Method a
GL (mm)
MCIM
ICIM LWR CGR (%)
MCIM MCIM
ICIM
CD (%)
MCIM
AC (%)
ICIM MIMR MCIM ICIM MIMR
GT (ºC)
MCIM
ICIM
GC (mm)
MIMR MCIM ICIM
Locus i GL1 GL2 GL8 GL9.2 GL8 GL9.2 LWR9 CGR2 qCGR2 CGR3.2 CGR4 qCGR5.1 CGR6.2 CGR7 CGR8.1 CGR2 CGR3.2 CGR4 qCGR5.1 CGR8.1 CD1.1 CD3 CD7 CD1.1 qCD8.1 AC3 AC7 AC3 AC7 qAC2 qAC2 GT2 GT5.1 GT5.1 qGT6.2 GT9.1 qGT9 GT2 GT9.1 qGT9 qGT6.1 GC1 GC1
Chr
Marker interval b
Locus j
Chr
Marker interval b
1 2 8 9 8 9 9 2 2 3 4 5 6 7 8 2 3 4 5 8 1 3 7 1 8 3 7 3 7 2 2 2 5 5 6 9 9 2 9 9 6 1 1
RM84−RM1003 RM7288−RM5356 RM6948−RM433 RM410−RM257 RM6948−RM433 RM5652−RM410 RM410−RM257 RM5427−RM262 RM525−RM2127 RM7097−RM448 RM3288−RM303 RM440−RM164 RM5753−RM345 RM346−RM336 RM3383−RM72 RM5427−RM262 RM7097−RM448 RM3288−RM303 RM164−RM305 RM72−RM22899 RM265−RM6696 RM448−RM8277 RM346−RM336 RM6696−RM3482 RM331−RM3383 RM218−RM232 RM180−RM214 RM218−RM232 RM214−RM542 RM5427−RM262 RM5427−RM262 RM525−RM2127 RM440−RM164 RM164−RM305 RM5753−RM345 RM6839−RM3700 OSR28−RM5786 RM525−RM2127 RM6839−RM3700 RM257−OSR28 RM454−RM162 RM84−RM1003 RM84−RM1003
GL9.1 GL7 GL9.1 GL11 GL9.1 GL11 LWR11 CGR3.1 CGR5 CGR8.2 CGR8.3 CGR6.1 CGR10 CGR9 qCGR11 CGR3.1 CGR8.2 CGR8.2 CGR6.1 qCGR11 CD1.2 CD8 CD9 CD1.2 qCD8.2 AC7 AC11.1 AC7 AC11.1 qAC3 AC11.2 GT10.1 GT5.2 GT10.2 GT11 GT9.2 qGT12 GT10.1 GT9.2 qGT12 qGT6.2 GC12 GC12
9 7 9 11 9 11 11 3 5 8 8 6 10 9 11 3 8 8 6 11 1 8 9 1 8 7 11 7 11 3 11 10 5 10 11 9 12 10 9 12 6 12 12
RM6570−RM5652 RM542−RM418 RM6570−RM5652 RM3133−RM7120 RM6570−RM5652 RM3133−RM7120 RM3133−RM7120 RM3467−RM545 RM1182−RM405 RM6948−RM433 RM264−RM6948 RM454−RM162 RM7492−RM5095 RM285−RM8206 RM7120−RM287 RM3467−RM545 RM6948−RM433 RM6948−RM433 RM454−RM162 RM7120−RM287 RM8105−RM84 RM6948−RM433 RM285−RM8206 RM495−RM8105 RM72−RM22899 RM125−RM180 RM287−RM21 RM125−RM180 RM287−RM21 RM3766−RM5639 RM5349−RM206 RM5095−RM311 RM161−RM7473 RM5629−RM171 RM3133−RM7120 RM6570−RM5652 RM5479−RM1227 RM7492−RM5095 RM6570−RM5652 RM5479−RM1227 RM162−RM5753 RM5609−RM5479 RM5479−RM1227
LOD AAij c R2 (%) P value 2.99 3.47 3.71 4.71 2.16 4.71 3.62 5.30 3.67 3.17 7.03 5.53 3.79 4.12 2.90 6.24 2.05 8.64 3.92 2.40 5.16 3.11 4.64 2.81 4.07 7.36 3.35 4.42 2.17 2.00 2.42 3.47 3.54 2.82 4.14 4.47 3.94 4.79 4.33 4.81 4.99 3.29 3.50
-0.02 3.81 0.03 5.72 0.03 6.73 0.03 7.73 0.03 4.23 0.04 7.73 0.02 7.79 4.03 7.01 2.67 3.08 3.24 4.54 -4.51 8.77 3.25 4.57 -3.24 4.54 -2.54 2.79 -2.73 3.21 4.28 7.76 2.37 2.60 -4.26 8.32 2.97 4.07 -2.37 2.72 1.66 7.60 1.33 4.88 -1.64 7.46 1.91 8.22 -3.25 9.52 -1.53 23.78 -0.72 5.30 -1.21 10.58 -0.54 3.52 1.00 4.22 0.59 4.06 -0.19 6.48 0.16 5.11 0.14 3.49 0.19 6.58 0.23 9.78 -0.14 3.72 -0.22 9.31 0.25 7.87 -0.18 5.77 0.24 7.78 -3.00 5.60 -4.04 11.20
0.001 0.0001 0.0001 < 0.0001 < 0.0001 0.0016 0.0004 < 0.0001 0.0002 < 0.0001 0.0009 0.0027 0.0001 0.0008 < 0.0001 < 0.0001 0.0002 0.0001 0.0002 0.0006 0.0001 0.0001 0.0003 0.0003 -
GL, Grain length; LWR, Grain length to width ratio; CGR, Chalk grain rate; CD, Chalkiness degree; GT, Gelatinization temperature; AC, Amylose content; GC, Gel consistency; Chr, Chromosome; AA, Epistatic effect; R2, Contribution rate; MCIM, Composite interval mapping in QTLMapper 2.0 software based on mixed linear model; ICIM, Inclusive composite interval mapping in QTL IciMapping 3.0 software based on stepwise regression linear model; MIMR, Multiple interval mapping with regression forward selection in Windows QTL Cartographer 2.5 based on multiple regression analysis. a Shows the E-QTLs both detected by ICIM and MCIM (or MIMR) for each trait in the table; b Bold letters indicate the nearest marker to the putative QTL; c Direction of effect: A positive AA value implies that the two-locus genotypes from the same parent Xiushui 79 or C Bao take the positive effects, while the two-locus genotypes of recombination from the parent Xiushui 79 and C Bao take the negative effects, the case of negative AA values is just the opposite.
effect directions were opposite (Table 2). Six, 16 and one pairs of E-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 3). Three pairs were both detected by MCIM and ICIM, and their contribution rates ranged from 3.72% to 9.78%. The two-locus genotypes from the same parent appeared to increase GT for GT9.1-GT9.2, while recombinant
two-locus combinations increased GT for the other two pairs. It indicated that both epistatic effect and additive effect had equal importance in the variation of GT in the population. For GC, one, one and three A-QTLs were detected by MCIM, ICIM, and MIMR, respectively (Table 2). qGC4 with the positive allele coming from C Bao was
LIU Qiang-ming, et al. QTL Analysis for Quality Traits of RIL Population in Japonica Rice
detected by all the three methods, and it explained 4.05% to 5.48% of the phenotypic variation. One, 26 and none pairs of E-QTLs were detected by MCIM, ICIM and MIMR, respectively (Table 3). For GC1GC12, the only one pair of E-QTL detected by both MCIM and ICIM, the two-locus genotypes from different parents appeared to increase GC. The contribution rates were 5.60% and 11.20%, respectively. It showed that both epistatic effect and additive effect had equal importance in the variation of GC in the population.
DISCUSSION In this study, all the five A-QTLs detected by MCIM were also detected by both ICIM and MIMR, and the position, additive effect and contribution rate of the same QTL detected by different methods were well coincident, so the A-QTLs detected by MCIM had high reliability. However, two A-QTLs with nonsignificant additive effect in MCIM were both detected by ICIM and MIMR, and the contribution rates of qAC9.1 reached 38.07% and 29.85%, respectively. It indicated that some false-negative results were obtained in the detection of A-QTL by MCIM. Though only nine A-QTLs were detected by ICIM, two were nonsignificant in the other two methods, thus some falsepositive A-QTLs existed in ICIM. A large number of A-QTLs were detected by MIMR, but most of them could be detected by neither MCIM nor ICIM, indicating MIMR had the most powerful ability to detect A-QTL, meanwhile, lots of false-positive A-QTLs also occurred. In addition, three pairs of A-QTLs, namely qLWR5.1 and qLWR5.2, qCGR5.2 and qCGR5.3, and qGT6.1 and qGT6.2, with adjacent interval and opposite direction for additive effect were detected by MIMR, but none of them was significant in MCIM and ICIM, which might due to that the positive effect was offset by the negative effect, indicating MIMR had a stronger ability in precise mapping. All the E-QTLs detected by MIMR were the interaction among A-QTLs, but they were detected by neither MCIM nor ICIM, so there were some deficiencies in the analysis of the interaction among loci with nonsignificant additive effect by MIMR. Many pairs of E-QTLs were detected by MCIM or ICIM individually, but only a few of them could be detected by both methods (MCIM and ICIM). For example, 141 pairs of E-QTLs were detected by ICIM, but only 14 pairs of them could be identified by both methods. These results indicated that a lot of false-positive E-QTLs
37
were obtained by MCIM and ICIM individually. The reliability could be improved by calibrating the E-QTLs detected by both MCIM and ICIM methods. In conclusion, the three mapping methods based on different genetic statistical models have their pros and cons, so multiple methods based on different genetic statistical models should be employed to analyze QTLs, and calibrating the QTLs detected by different methods synchronously to improve the reliability of the results. qLWR11, qCGR2, qCGR5.1, qCGR9, qAC9.1, qGT12 and qGC4 were the seven highly reliable A-QTLs. The primer sequences of the pre and post markers of the seven A-QTLs were analyzed by BLAST (http:// www.ncbi.nlm.nih.gov) and located on the published PAC-BAC clone physical map of rice (http://rgp.dna. affrc.go.jp/E/IRGSP/download.html), and then compared to the previous research (Bao et al, 2003; Liu et al, 2007; Zhou et al, 2009; Gao et al, 2011) was conducted. No QTL was reported around qLWR11, qCGR2, qCGR9, qAC9.1 and qGC4 for the corresponding traits. So these five A-QTLs were new loci. qCGR5.1 shared an identical interval with qPCG5, which was reported by Liu et al (2007), thus they were presumed to be the same locus. One QTL for GT was reported on chromosome 12 (http://www.gramene.org/qtl/), and the flanking markers were RM309 and RM6837. The adjacent markers of qGT12 detected in the present study were RM5609 and RM5479. Compared the order of the four SSR markers through the published information on Gramene, we found that the interval of RM5609−RM5479 was completely involved in the interval of RM309−RM6837. Therefore, the two sites were treated as one site. In the present study, only highly reliable E-QTLs were detected for GL and CD, and only highly reliable A-QTL was detected for LWR, while both two kinds of QTLs were detected for the other four traits. So, in the Xiu-Bao RIL population, epistatic effect played a major role in the variation of GL and CD, and additive effect was the dominant in the variation of LWR, while both epistatic effect and additive effect had equal importance in the variation of the rest four traits. Fourteen pairs of highly reliable E-QTLs were detected. Of them, 11 pairs were the interaction among sites with non-significant additive effect, and two pairs for CGR were the interaction between A-QTLs and sites with non-significant additive effect, and only one pair for GT were the interaction between two A-QTLs. These suggested that the majority of E-QTLs were the
Rice Science, Vol. 20, No. 1, 2013
38
interaction among sites with non-significant additive effect for quality traits. This result is consistent with those obtained by Lei et al (2008) and Jiang et al (2009). Therefore, it is possible to polymerize the elite alleles of highly reliable A-QTLs, which had no interaction to other sites, by marker-assistant selection to improve the grain quality of japonica hybrid rice. Moderate amylose content was preferred in rice (Huang et al, 2011). QTL qAC9.1 detected by MIMR, with the contribution rate of 29.85% and the additive effect of -4.63%, was located closely to the marker RM410 (Table 2). Substituting C Bao allele with Xiushui 79 allele at the RM410 site would reduce AC by 4.63%. Contrarily, substituting Xiushui 79 allele with the C Bao allele would increase AC by 4.63%. Therefore, we can select the allele at the RM410 site to improve amylose content of japonica hybrid rice according to the objective of breeding. In addition, we found that qAC9.1 was located on the BAC clone OJ1005_D12, which is on the long arm of chromosome 9, with a distance of 17.99 Mb to the end of short arm and 5.29 Mb to the end of long arm, through analysis of primer sequences of RM410 and RM257 by BLAST.
ACKNOWLEDGEMENTS This work was supported by the National High Technology Research and Development Program of China (Grant No. 2010AA101301); the Program of Introducing International Advanced Agricultural Science and Technology in China (Grant No. 2006-G8[4]-31-1) and the Program of Science-Technology Basis and Conditional Platform in China (Grant No. 505005).
REFERENCES Bao J S, Corke H, He P, Zhu L H. 2003. Analysis of quantitative trait loci for starch properties of rice based on an RIL population. Acta Bot Sin, 45(8): 986–994. Deng H F, He Q, Shu F, Zhang W H, Yang F, Jin Y H, Dong L, Xie H. 2006. Status and technical strategy on the development of japonica hybrid rice in China. Hybrid Rice, 21(1): 1–6. (in Chinese with English abstract) Fan C C, Xing Y Z, Mao H L, Lu T T, Han B, Xu C G. 2006. GS3, a major QTL for grain length and weight and minor QTL for grain width and thickness in rice, encodes a putative transmembrane protein. Theor Appl Genet, 112(6): 1164–1171. Gao Z Q, Zhan X D, Liang Y S, Cheng S H, Cao L Y. 2011. Progress on genetics of rice grain shape trait and its related gene mapping and cloning. Hereditas, 33(3): 1–8. (in Chinese with English abstract) Gao Z Y, Zeng D L, Cui X, Zhou Y H, Yan M X, Huang D N, Li J Y, Qian Q. 2003. Map-based cloning and sequence analysis of a gene, ALK, responsible for gelatinization temperature in rice
(Oryza sativa L.). Sci Chin: Ser C, 33(6): 481–487. (in Chinese) Guo Y, Cheng B S, Hong D L. 2009. Construction of SSR linkage map and snalysis of QTL for rolled leaf of restorer line in japonica rice (Oryza sativa L.). Chin J Rice Sci, 23(3): 245–251. (in Chinese with English abstract) Huang D C, Liang K, Sun C, Jiang J H, Shi M L, Dai J, Xie H, Zhao K M, Nguyen P T, Ma W X, Hong D L. 2011. Identification of marker genotypes associated with elite combining ability for quality traits in parents of hybrid japonica rice (Oryza sativa L.). Acta Agron Sin, 37(3): 405–414. (in Chinese with English abstract) Jiang J H, Guo Y, Chen X G, Xu H M, Hou Y J, Hong D L. 2007. The correlation between panicle angle and rice quality and genetic analysis on rice quality in japonica rice (Oryza sativa L.). Hereditas, 29(6): 714–724. (in Chinese with English abstract) Jiang L R, Wang W, Huang J X, Huang R Y, Zheng J S, Huang Y M, Wang H C. 2009. Analysis of epistatic and QE interaction effects of QTLs for grain shape in rice. Mol Plant Breeding, 7(4): 690–698. (in Chinese with English abstract) Lei D Y, Xie F M, Xu J L, Chen L Y. 2008. QTLs mapping and epistasis analysis for grain shape and chalkiness degree of rice. Chin J Rice Sci, 22(3): 255–260. (in Chinese with English abstract) Li H H, Ribaut J M, Li Z L, Wang J K. 2008. Inclusive composite interval mapping (ICIM) for dysgenic epistasis of quantitative traits in biparental populations. Theor Appl Genet, 116: 243– 260. Liu J F, Kui L M, Zhu Z F, Tan L B, Wang G J, Li Q W, Shu J H, Sun C Q. 2007. Identification of QTLs associated with processing quality and appearance quality of common wild rice (Oryza rufipogon Griff.). J Agric Biol, 15(1): 90–96. (in Chinese with English abstract) MaCoach S R. 2008. Gene nomenclature system for rice. Rice, 1: 72–84. Niu F A. 2011. Genetic segregation analysis and QTL mapping as well as association analysis for panicle angle trait in japonica rice (Oryza sativa L.). [Master Dissertation] Nanjing: Nanjing Agricultural University: 25–27. (in Chinese with English abstract) Okagaki R J, Wessler S R. 1988. Comparison of non-mutant and mutant waxy genes in rice and maize. Genetics, 120: 1137−1143. Su C F, Lu W G, Zhao T J, Gai J Y. 2010a. Verification and finemapping of QTLs conferring days to flowering in soybean using residual heterozygous lines. Chin Sci Bull, 55(6): 499–508. Su C F, Zhao T J, Gai J Y. 2010b. Simulation comparisons of effectiveness among QTL mapping procedures of different statistical genetic models. Acta Agron Sin, 36(7): 1100–1107. (in Chinese with English abstract) Wang D L, Zhu J, Li Z K. 1999. Mapping QTL with epistatic effects and QTL × environment interactions by mixed linear model approaches. Theor Appl Genet, 99: 1255–1264. Wang S, Basten C J, Zeng Z B. 2007. Windows QTL Cartographer 2.5. Raleigh, NC: Department of Statistics, North Carolina State University (http://statgen.ncsu.edu/qtlcart/WQTLCart.htm). Weng J F, Gu S H, Wan X Y, Gao H, Guo T, Su N, Lei C L, Zhang X, Cheng Z J, Guo X P, Wang J L, Jiang L, Zhai H Q, Wan J M. 2008. Isolation and initial characterization of GW5, a major QTL associated with rice grain width and weight. Cell Res, 18(12): 1199–1209. Xu J Q, Zhang D Y, Zhang Y, Li J Y, Guo H, Xu D Y, Shao F K, Xu Y. 2010. Practice and thinking on japonica hybrid rice breeding. Hybrid Rice, 25(Special Collection): 85–87. (in Chinese) Zhou L J, Liu X, Jiang L, Zheng L N, Chen L M, Liu S J, Zhai H Q, Wan J M. 2009. Analysis of QTL and GE effects on PGWC in rice (Oryza sativa L.) using CSSL and BIL populations. Sci Agric Sin, 42(4): 1129–1135. (in Chinese with English abstract)