Fine mapping of qTGW10-20.8,a QTL having important contribution to grain weight variation in rice

Yujun Zhu, Zhenhu Zhng,Junyu Chen, Yeyng Fn, Tongmin Mou,Shoqing Tng,*, Jieyun Zhung,*

aState Key Laboratory of Rice Biology and Chinese National Center for Rice Improvement, China National Rice Research Institute, Hangzhou 310006,Zhejiang,China

bState Key Laboratory of Crop Genetic Improvement and National Center of Plant Gene Research(Wuhan),Huazhong Agricultural University,Wuhan 430070, Hubei,China

Keywords:Fine mapping Grain length Grain weight Quantitative trait loci Rice

ABSTRACT

1.Introduction

Number of panicles per plant, number of grains per panicle,and grain weight are the three most important determinants of grain yield in rice (Oryza sativa L.). In the past decade, fine mapping and cloning of quantitative trait loci (QTL) for rice yield traits, especially grain weight, have made considerable progress.To date,16 QTL for grain weight and grain size have been cloned. They were distributed on all the rice chromosomes except chromosomes 1, 10, 11, and 12. Five of them,OsLG3, OsLG3b/qLGY3, GS3, GL3.1/qGL3, and qTGW3, were located on chromosome 3 [1-7]. Eight others were paired on four chromosomes: GW2 and GS2/GL2 on chromosome 2[8-10], GS5 and GSE5 on chromosome 5 [11,12], TGW6 and GW6 on chromosome 6 [13,14], and GLW7 and GL7/GW7 on chromosome 7 [15-17]. The remaining three, GL4, GW8, and GS9, were placed on chromosomes 4, 8 and 9, respectively[18-20].All of these genes were isolated by map-based cloning except for OsLG3, GSE5, and GLW7, which were identified by Ho-LAMap[1]or genome-wide association studies[12,15].

Grain weight and grain size are determined mainly by the length,width and thickness of grains,which are decided by the number and size of the cells in the spikelet hull. The 16 QTL cloned for these traits were found to be involved in multiple signaling pathways regulating cell proliferation and elongation.Eight of them, OsLG3, OsLG3b/qLGY3, GS2/GL2, GL4, GW6a, GL7/

GW7,GLW7,and GW8,encode transcriptional regulatory factors[1-3,9,10,14-19]. Five others, GL3.1/qGL3, TGW3, GS5, GSE5, and TGW6, encode components of Brassinosteroid and Auxin signaling pathways [5-7,11-13]. Two of the other three QTL,GS3 and GW2,are involved in the ubiquitin-proteasome and Gprotein signaling pathways, respectively [4,8]. The remaining QTL,GS9,encodes an unknown expressed protein that may form a link between transcription factors and the brassinosteroid signaling pathway during spikelet development[20].Although it is evident that these genes are key regulators of grain weight and grain size in rice, the information is rather fragmentary and more QTL remain to be characterized for uncovering the genetic and molecular relationships among regulators.

Grain weight and grain size are major targets of domestication and breeding for increasing grain yield and improving grain quality in rice.Ten of the 16 QTL cloned have been subjected to strong selection in this process. For GS3, the first cloned QTL controlling grain weight,multiple alleles have been detected in several studies [21-24]. Two of the alleles, the long-grain type MH-GS3 and short-grain type ZS-GS3, are commonly found in modern rice varieties,of which the long grain-type allele is likely to have had a shorter domestication history. For GW8, three alleles were identified in a panel of 107 accessions,including 16 wild rice accessions,18 landraces,and 72 modern varieties.Two alleles were detected in the wild rice, of which GW8HJX74and GW8Basmatihave been used with high frequency in indica cultivars for increasing grain yield and in japonica cultivars for improving grain quality,respectively[19].For the other eight genes,OSLG3,

OSLG3b/qLGY3,GS5,GSE5, GL4, GW6a,GLW7, and GL7/GW7, the beneficial alleles are also found at high frequency in modern rice varieties[1-3,11,12,14-18].Exploitation of functional markers for the beneficial alleles of these genes has provided new efficient tools for marker-assisted selection in rice breeding[21,22].

Widespread use of F1hybrids has made a great contribution to increasing yield potential in rice [25]. For the three yield components, trait values of F1hybrids are generally found to be positively correlated with both F1heterosis and the trait values of the parental lines [26]. Investigation of the genetic basis underlying grain weight variation among parental lines of hybrid rice would facilitate not only the use of beneficial alleles in rice breeding,but also the identification of more key regulators controlling this important trait.

In the present study, QTL analysis for grain weight and grain size was performed using populations derived from crosses between major parental lines of three-line indica hybrid rice. First, three recombinant inbred line (RIL) populations were used to identify QTL that segregated in more than one population and were located apart from the 16 QTL previously cloned. Then, one QTL was selected for fine mapping using eight populations segregating in an isogenic background. This QTL, qTGW10-20.8 controlling grain weight,grain length,and grain width,was localized to a 70.7-kb region containing seven annotated genes on the long arm of chromosome 10.

2. Materials and methods

2.1. Plant materials

Eleven populations of indica rice (Oryza sativa subsp. indica)were used, including three RIL populations for primary mapping and eight sets of near-isogenic lines (NILs) for finemapping.

The three RIL populations were Teqing/IRBB lines (TI),Zhenshan 97/Milyang 46 (ZM) and Xieqingzao/Milyang 46(XM), all reported previously [27]. All of the parental lines of these populations are indica rice cultivars that have been widely used in the breeding and production of three-line hybrid rice in China [27-29]. For TI, the female parent Teqing(TQ) and male parents IRBB lines are restorer lines having similar eco-geographical adaptation. Phenotypic data for two years and a map consisting of 127 markers and spanning 1198 cM was previously [28] used to identify QTL for 1000-grain weight(TGW).In the present study,phenotypic data for two more years were added. The map has been updated to comprise 135 markers and span 1345 cM[30].For ZM and XM,the female parents Zhenshan 97 and Xieqingzao are maintainer lines and the common male parent Milyang 46 is a restorer line. The ZM population was previously used in QTL mapping for TGW using phenotypic data for two years and a map consisting of 158 markers and spanning 1279 cM [31]. In the present study, phenotypic data for two more years were added. The map has been updated [27] to comprise 256 markers and span 1815 cM. The XM population has not been used in QTL mapping for TGW. Phenotypic data for three years and a map comprising 240 markers and spanning 2080 cM[27]were used in this study.

The eight NIL populations were derived from one residual heterozygous plant as described below and illustrated in Fig.1.Following QTL analysis using the three RIL populations,qTGW10.2 was targeted for fine mapping. One F7plant of TQ/IRBB52 that was heterozygous only in the qTGW10.2 region was selected and selfed to produce an F8population. Three plants that were heterozygous in the intervals RM25767-Te20811,RM25767-Te21018, and RM25767-RM3123, respectively, were identified. Three F9populations were derived and genotyped with DNA markers in the target intervals. In each population,nonrecombinant homozygotes were identified and selfed to produce homozygous lines. Three F9:10NIL sets were established (Fig. 2A) and used for QTL analysis. The qTGW10.2 region was updated to Te20811-Te21052. Five F10plants with sequential heterozygous segments were then selected following the updated QTL region and selfed to produce five F11populations.In each population,nonrecombinant homozygotes were identified and selfed to produce homozygous lines. Five F11:12NIL sets were established (Fig. 2B) and used for QTL analysis.

Fig.1-Development of eight NIL populations.NIL, near-isogenic line.

For the eight NIL populations, the male parent IRBB52 is a rice line carrying bacterial blight resistance genes Xa4 and Xa21 on chromosome 11 in the genetic background of IR24 [32].Genomic differences between IRBB52 and IR24 were detected using 862 DNA markers(data not shown).As expected,the two cultivars showed differences in the regions containing Xa4 and Xa21 but no difference in the qTGW10.2 region.

2.2. Field trial and phenotypic evaluation

All the rice populations were planted in the middle-rice season in the paddy field of the China National Rice Research Institute in Hangzhou, Zhejiang, China. A randomized complete block design with two replications was used for all trials.In each replication, each line was grown in a single row of 12 plants, with 16.7 cm between plants and 26.7 cm between rows. Field management followed local agricultural practice.At maturity, five of the middle 10 plants in each row were harvested in bulk. Approximately 600 grains were selected and measured for TGW, grain length (GL), and grain width(GW)following Zhang et al.[33].

2.3. DNA marker analysis

For the three RIL populations, marker data and genetic maps were available[27,30].For developing the eight NIL populations,15 polymorphic DNA markers were used, including five simple sequence repeat (SSR) and ten InDel markers. The SSRs were selected from the Gramene database(http://www.gramene.org/),and the InDels (Table S1) were designed according to the difference between TQ and IRBB52 as defined by whole-genome resequencing. DNA was extracted using a mini-preparation protocol [34]. PCR amplification followed Chen et al. [35]. The products were visualized on 6%non-denaturing polyacrylamide gels using silver staining.

For sequence analysis,genomic DNA of TQ and IRBB52 was extracted using DNeasy Plant Mini Kit (QIAGEN, Hilden,Germany) according to the manufacturer’s instructions.Products amplified using primers listed in Table S2 were sequenced. The nucleotide and predicted amino acid sequences of TQ and IRBB52 were compared.

2.4. Data analysis

Basic descriptive statistics, including mean trait values,standard deviation, minimum and maximum trait values,skewness,and kurtosis were computed.

Fig.2-Segregating regions in eight NIL populations.(A)Three populations at F9:10 used for localizing qTGW10.2 to a 241.2-kb region.(B)Five populations at F11:12 used for localizing qTGW10.2 to a 70.7-kb region.NILTQ and NILIRBB52 are near-isogenic lines with respectively Teqing and IRBB52 homozygous genotypes in the segregating region.

For the three RIL populations, QTL and genotype-byenvironment (GE) interactions [36] were determined using the MET (QTL by environment interaction for multi-environment traits) functionality in QTL IciMapping software [37], taking different years as different environments. LOD thresholds for genome-wide type I error of P ＜0.05 were calculated with 1000 permutations and used to declare a putative QTL. QTL were named following the rules proposed by McCouch and CGSNL[38].

For the eight NIL populations,two-way analyses of variance(ANOVA)were performed to test phenotype differences between the two homozygous genotypic groups in each population.The analysis was performed using SAS[39]procedure GLM (general linear model)as previously described[40].Given the detection of a significant difference (P ＜0.05), the same data were used to estimate the genetic effect of the QTL,including additive effect,GE effect,and proportion of phenotypic variance explained(R2).QTL were designated using a version[41]modified from the rules proposed by McCouch and CGSNL [38], in which the physical position of the first segregating marker in the QTL region was used as the unique identifier for the given QTL.

3. Results

3.1.QTL for 1000-grain weight detected in three RIL populations

Descriptive statistics for TGW in the three RIL populations are presented in Table 1. The TGW was always continuously distributed with low skewness and kurtosis,showing a typical pattern of quantitative variation. QTL were determined for each population and comparison was made among different populations based on the physical positions of DNA markers linked to these QTL.A total of 27 QTL were detected,including 17 found in two or all the three populations (Table 2) and 10 identified only in a single population (Table S3). In the TI population, 20 QTL were detected, including 12 shared with other populations and eight identified only in T1.The additive effects of the 20 QTL had a mean R2of 4.17%. In the ZM population, 15 QTL were detected, including 13 shared with other populations and two identified only in ZM.The additive effects of the 15 QTL had a mean R2of 3.38%. In the XM population, 12 QTL were detected, all of which were shared with other populations.The additive effects of these QTL had a mean R2of 3.45%. Significant GE interaction was observed for five QTL, including four only in TI or ZM and one in both(Table S4).

Eight of the 17 QTL detected in more than one population were located in regions where QTL affecting grain weight have been cloned(Table 2).Two of them,qTGW3.2 in the GS3 region and qTGW5.1 in the GSE5 region, showed major effects, with respective R2values of 27.51%and 23.18%in the TI population.qTGW3.2 was also the QTL having the largest R2of 13.81 in XM and the second largest R2of 7.31%in ZM.Three others,qTGW2.2,qTGW3.3, and qTGW6.1, showed R2between 5.0%and 10.0%in one or more populations. For the remaining three QTL,qTGW3.1,qTGW6.2,and qTGW9.2,the R2values were all ＜2.5%.

Table 1-Phenotypic values of 1000-grain weight(g) in three recombinant inbred line populations.

The other nine shared QTL were located in regions where no QTL associated with grain weight have been cloned,including qTGW8, shared by all three populations; qTGW9.1 and qTGW10.2, shared by TI and ZM; qTGW2.1 and qTGW12,shared by TI and XM; and qTGW1, qTGW5.2, qTGW10.1, and qTGW11, shared by ZM and XM (Table 2). The R2of these QTL fell in the intervals 0.52%-2.08%, 1.22%-3.48%, and 1.62%-2.69%in the TI,ZM and XM populations,respectively.Among them, qTGW10.2 was the QTL with highest LOD, additive effect, and R2in the TI and ZM populations. In addition, this region was previously found to have significant effects on grain size in segregating populations derived from a cross between two lines of the TI population [44]. Accordingly,qTGW10.2 was chosen for fine mapping.

Table 2-QTL for 1000-grain weight detected in two or all three recombinant inbred line populations.

3.2. Fine mapping of qTGW10.2

Eight populations segregating in an isogenic background were constructed for the fine mapping of qTGW10.2,including three sets of NILs in the F9:10generation and five sets of NILs in F11:12. The phenotypic distributions in the three F9:10NIL populations, Z1, Z2, and Z3, are shown in Fig. 3. In Z1, the three traits were continuously distributed with little difference between the two genotypic groups. In Z2 and Z3, TGW and GL showed bimodal distributions with the TQ and IRBB52 homozygous lines concentrated in the high- and low-value regions, respectively. The same tendency was observed for GW,though the difference between the two genotypic groups was less distinguishable than for TGW and GL. These results provided evidence for allelic differences of QTL for TGW, GL,and GW in the segregating regions of the Z2 and Z3 populations.

Two-way ANOVA was performed to test differences in TGW, GL, and GW between the two genotypic groups in each population (Table 3). No significant effect was detected in Z1,but highly significant effects(P ＜0.0001)were detected for all three traits in Z2 and Z3. Additive effects estimated from the two populations were similar, with the TQ allele increasing TGW,GL,and GW by 0.66 and 0.59 g,0.110 and 0.098 mm,and 0.019 and 0.016 mm,respectively.The R2values were 71.08%-79.51%for TGW and GL,higher than the values of 53.08%and 39.02% for GW. These results indicate that qTGW10.2 had segregated in Z2 and Z3 but not in Z1.As shown in Fig.2A,this region was flanked by InDel markers Te20811 and Te21052,corresponding to a 241.2-kb region of the Nipponbare genome.

Fine-mapping of qTGW10.2 was continued using five F11:12NIL populations with sequential heterozygous segments covering the interval Te20811-Te21052 (Fig. 2B). The distributions of TGW, GL, and GW in these populations are shown in Fig. 4. In Y1, Y4, and Y5, the three traits were continuously distributed with little difference between the two genotypic groups. In Y2 and Y3, the three traits showed bimodal distributions, except for GW in Y3. The TQ and IRBB52 homozygous lines were concentrated respectively in the high- and low-value regions of the three traits. These results indicate that qTGW10.2 had segregated in Y2 and Y3 but not in the other three populations.

Results of two-way ANOVA for TGW, GL, and GW in the five populations are shown in Table 3. No significant effect was detected in Y1,Y4,or Y5 except for a small effect for GL in Y4.Highly significant effects(P ＜0.0001)were detected for all three traits in Y2 and Y3.Additive effects detected in the two populations were similar, with the TQ allele increasing TGW,GL,and GW by 0.48 and 0.40 g,0.070 and 0.072 mm,and 0.020 and 0.016 mm, respectively. The R2values were 68.78%-73.79% for TGW and GL, higher than the values of 62.02%and 47.64% for GW. Obviously, qTGW10.2 was located in a region that had segregated in Y2 and Y3 but was homozygous in the other three populations.As shown in Fig.2B,this region was flanked by InDel markers Te20811 and Te20882, corresponding to a 70.7-kb region of the Nipponbare genome.Given that this QTL affected TGW,GL,and GW with the same allelic direction,and its first segregating marker Te20811 was located at 20.8 Mb on chromosome 10,we designed it qTGW10-20.8.

Fig.3-Distributions of 1000-grain weight,grain length,and grain width in Z1,Z2,and Z3,three near-isogenic line populations at F9:10.

Table 3-Fine mapping of qTGW10-20.8 using eight near-isogenic line(NIL) populations.

3.3. Candidate genes for qTGW10-20.8

According to the Rice Annotation Project Database (http://rapdb.dna.affrc.go.jp/),there are seven annotated genes in the 70.7-kb region for qTGW10-20.8. Six of them encode proteins with known functional domains. Os10g0535600 encodes a member of the alpha/beta fold hydrolase family that is involved in diverse biochemical activities [45]. Os10g0535800 encodes an uncharacterized cysteine-rich domain containing protein. Os10g0535900 is orthologous to the AtEXPA7 gene controlling root hair formation in Arabidopsis [46].Os10g0536000 is paralogous to the OsSCE1 gene involved in stress response of rice plants [47]. Os10g0536050 encodes a protein similar to ubiquitin-conjugating enzyme that participates in a variety of plant developmental process including biotic and abiotic stress response, hormone synthesis, and seed development [48]. Os10g0536100 encodes the MIKC-type MADs-box protein OsMADS56, a long-day-specific negative regulator of flowering in rice [49]. The remaining annotated gene is Os10g0535575,encoding a hypothetical protein.

Sequence comparisons of the seven annotated genes between TQ and IRBB52 were performed. Mutations were found at 51 sites of Os10g0536100 (Table S5) and 2-11 sites of the other six genes(Table S6),but few of them occurred in the exons. No mutation was detected in the exons of Os10g0535600, Os10g0535900, and Os10g0536000. One mutation in exons was found for three genes.For Os10g0535575,the A5G substitution results in a premature stop codon in TQ.For Os10g0535800 and Os10g0536050, the T1689A and G2393C substitutions lead to amino acid changes from Phe to Tyr and Glu to Gln,respectively.Two mutations were identified in the exons of Os10g0536100 that encodes OsMADS56,including the T8697C synonymous substitution and a nine-nucleotide insertion in TQ.

4. Discussion

Grain weight is a key component of grain yield in rice, and a trait closely related to grain size and grain shape, which are important in determining grain quality. In the past decade,QTL cloning has greatly enhanced our understanding of the molecular regulation of grain weight, but many more QTL remain to be characterized [50]. In the present study, QTL analysis for grain weight was performed using populations derived from crosses between major parental lines of indica hybrid rice.A total of 27 QTL were detected,of which 17 were identified in different crosses. It was found that nine of the common QTL were located in regions where no QTL associated with grain weight has been cloned. One of them,qTGW10-20.8, was delimited into a 70.7-kb region on the long arm of chromosome 10.

Fig.4- Distributions of 1000-grain weight,grain length,and grain width in Y1, Y2,Y3, Y4, and Y5,the five near isogenic line populations at F11:12.

Segregating populations used for primary QTL mapping in rice have commonly been constructed from crosses between two distinct cultivars. This approach is advantageous for population construction by avoiding low polymorphism of DNA markers, but disadvantageous for the detection of QTL having small effects. Now that marker polymorphism is no longer a limitation in rice, the use of closely related cultivars has facilitated the detection of novel QTL for agronomic traits[51]. In the rice growing area where the present study was conducted, both the male and female parents of the TI population are middle-season rice cultivars, whereas the female and male parents of the other two populations are respectively early- and middle-season cultivars [27-29]. It appears that QTL detection is more powerful in TI than ZM and XM(Table 2).First,20 QTL were detected in TI,but only 15 and 12 were identified in ZM and XM, respectively. Second,the mean R2was 4.17%in TI,higher than the respective values of 3.38% and 3.45% in ZM and XM. Third, for minor QTL that were detected in TI and either or both of the other two populations,the LOD scores were generally much higher in TI than ZM and XM even though the R2were similar. Fourth,major QTL showed much higher R2in TI than ZM and XM, as shown by the two QTL having the largest and second largest effects in each population. In TI and XM, the two QTL are qTGW3.2 in the GS3 region and qTGW5.1 in the GSE5 region,with R2values of respectively 27.51% and 23.18% in TI and 13.81% and 4.80% in XM. In ZM, they are qTGW2.2 and qTGW3.2 in the GS2 and GS3 regions, with R2values of respectively 7.95% and 7.31%. Our results suggest that the use of parental lines having similar eco-geographical adaptation could improve the detection of QTL in rice.

All of the populations used in this study were constructed from crosses between major parents of commercial hybrid rice. The QTL detected, especially the 17 QTL commonly identified in different crosses, would be expected to play an important role in controlling grain weight variation in hybrid rice. Eight of the shared QTL were located in proximity to cloned QTL controlling grain weight, including six genes for grain size, GS2/GL2 [9,10], OsLG3 [1], GS3 [4], GL3.1/qGL3 [5,6],GSE5 [12], and GS9 [20], and two heading-date genes showing pleiotropic effects on grain weight, RFT1 [42] and Hd1 [43].Investigation of the allelic variation of these genes would be essential for understanding the genetic basis of grain weight variation in hybrid rice. The other nine shared QTL were located in regions where no QTL associated with grain weight have been cloned, providing new candidates for characterizing regulators of this trait. One of them, qTGW10-20.8, was mapped to a 70.7-kb region on the long arm of chromosome 10,which was similar in location to a QTL previously found to be associated with a heterotic effect on grain weight [28].Cloning and molecular characterization of this QTL may shed light on the regulation of grain yield heterosis in rice.

The 70.7-kb region surrounding qTGW10-20.8 contained seven annotated genes. For three annotated genes,Os10g0535600, Os10g0535900, and Os10g0536000, no exon sequence differences between TQ and IRBB52,the two parents of the populations used for fine-mapping qTGW10-20.8, were detected. For three other annotated genes, Os10g0535575,Os10g0535800, and Os10g0536050, one mutation in the exons was detected in each gene.For the remaining annotated gene,Os10g0536100,an insertion of nine nucleotides or three amino acids was detected in TQ. This gene encodes the MIKC-type MADS-box protein OsMADS56, a long-day regulator of flowering in rice [49]. Another MADS-box gene involved in the regulation of rice flowering, OsMADS51, was recently reported to influence grain weight [52]. Work is under way to test whether Os10g0536100 is the gene underlying the QTL qTGW10-20.8 for grain weight.

5. Conclusions

A total of 27 QTL for grain weight, grain length, and grain width were detected using populations derived from crosses between major parental lines of three-line indica hybrid rice.Nine of them were identified in more than one population and located in regions where no QTL associated with grain weight have been cloned, providing new candidates for characterizing regulators of grain weight and grain size. One of the QTL,qTGW10-20.8, was localized to a 70.7-kb region on the long arm of chromosome 10. Sequence analysis suggested that Os10g0536100 encoding the MIKC-type MADS-box protein OsMADS56 is the most likely candidate for qTGW10-20.8.These results provide a basis for cloning a QTL that makes an important contribution to grain weight variation in rice.

Acknowledgments

This work was supported by the National Key Research and Development Program of China (2016YFD0101104), the National Natural Science Foundation of China (31521064), and a project of the China National Rice Research Institute(2017RG001-2).

Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2019.01.006.