SNP-based QTL mapping for panicle traits in the japonica super rice cultivar Liaoxing 1

Mingzhu Zho,Zuoin M,Lili Wng,Zhiqing Tng,Ting Mo,Chunin Ling,Hong Go,Liying Zhng,N He,Ling Fu,Chnghu Wng,Guomin Sui,*,Wenjing Zheng,*

aInstitute of Rice Research,Liaoning Academy of Agricultural Sciences,Shenyang 110000,Liaoning,China

bInstitute of Rice Research,Shenyang Agricultural University,Shenyang 110161,Liaoning,China

cLiaoning Province Saline and Alkaline Land Utilization and Research Institute,Panjin 124010,Liaoning,China

ABSTRACT

1.Introduction

Rice(Oryza sativa)is a staple food crop worldwide and particularly in Asia,feeding almost half of the human population.Food security is becoming a severe global problem as the rate of increase in world population and competition for arable land between food and energy crops intensifies[1].Increasing yield is necessary for food security[2].Panicle features are a component of plant architecture that is associated with grain yield[3].Genetic improvement for panicle traits,including panicle branching and grain number per panicle(GN),are considered to be one of the most promising approaches to increasing grain yield in rice.

With the development of DNA markers and a high-density genetic map of rice,QTL mapping for complex quantitative traits can identify associations between markers and traits and has become an effective approach to revealing the genetic basis of quantitative traits and enabling marker-assisted selection(MAS)breeding[4].Many QTL for panicle traits have been reported[5,6],and some have been fine-mapped[7–9]and provide information for gene cloning and genetic tools for improving grain yield in rice breeding.Among QTL for panicle traits,several QTL of effect comparable to those of major genes have been cloned,including Gn1a[10],DEP1[11],LAX1[12],IPA1[13,14],SP1[15],and APO1[16,17].Most QTL or genes for panicle traits have been identified using mapping populations derived from elite indica cultivars including Minghui 63,Zhenshan 97,and Kasalath.However,in addition to DEP1 and IPA1,the introgression of panicle traits or QTL or genes into high-yielding japonica varieties has not been studied in depth.

The Chinese Ministry of Agriculture launched the“China Super Rice Breeding Program”in 1996 and conducted the breeding and demonstration of new super rice cultivars in 2005.A series of japonica super rice(JSR)cultivars have been released in northeast China[18].Liaoxing 1,derived from a cross of Liaojing 454 and Shennong 9017,was identified as a typical JSR cultivar,and its cultivation area has reached 10.8 Mha.Liaoxing 1 showed dense and erect panicles and production of many grains per panicle,particularly on secondary branches.High-yielding japonica rice cultivars with dense and erect panicles have been released as commercial cultivars in China.DEP1,a major QTL for dense and erect panicles,has been identified in erect-type cultivars including Liaojing 5,Shennong 265,Qianchonglang 1,and Wuyunjing 7 and increased GN,grain yield[11]and nitrogen use efficiency[19]in rice.The dominant DEP1 allele,which confers dense and erect panicles,has been widely used in japonica rice breeding and plays an important role in the improvement in grain yield.For instance,Liaoxing 1 has been used as a parent in breeding programs to develop more than 27 cultivars(China Rice Data Center),and the dominant DEP1 allele was selected in most varieties.Interestingly,the yield of these cultivars was not substantially greater than that of their ancestor Liaoxing 1.The difficulty of developing nextgeneration JSR cultivars has been due primarily to lack of information about the genetic bases of current JSR varieties.

The objective of this study was to identify QTL for panicle traits using a population of recombinant inbred lines(RILs)derived from a cross between Gangyuan 8 and Liaoxing 1 grown in three environments.Our results will provide important information to unearth the QTL for paniclerelated traits in rice and help to understand how to achieve greater GN and yield in the Liaoxing 1.

2.Materials and methods

2.1.Plant materials

The JSR variety Liaoxing 1 and average-yielding japonica variety Gangyuan 8 were used as parental lines.These parents differed in primary branch number(BNP),grain number on primary branch(GNP),secondary branch number(BNS),grain number on secondary branch(GNS),GN,panicle length(PL),and grain density(GD)(Fig.1A).A cross of Gangyuan 8×Liaoxing 1 was made in 2013.The F1progeny were selfpollinated to obtain approximately 1300 F2plants,which were advanced for four generations of single-seed descent to produce highly homozygous RILs.A set of 197 F10RILs were used for the replicated field trial.

2.2.Experimental design and trait measurement

Field trials were conducted in Donggang(DG,39°87′N,124°15′E),Shenyang(SY,41°48′N,123°25′E),and Panjin(PJ,41°12′N,122°07′E)in Liaoning province in 2019.A randomized complete block design with three replicates was established in each environment.Each RIL and parents were sown in a seedling nursery in April,and one seedling was transplanted per hill on May 24.The plot size was four rows,16 plants per row with a spacing of 30 cm between rows and 13 cm between plants.Five protective rows were established on each side in the field,and standard local agronomic management practices were followed.At maturity,the main panicles of 10 plants of each line and their parents were harvested from each plot to measure BNP,BNS,GNP,GNS,GN,and PL.Grains/cm(GD)was,calculated as GN/PL.

2.3.Statistical analysis of phenotypic data

A mixed linear model was fitted using the lmer function in lme4 package of R(http://www.R-project.org/)to calculate the best linear unbiased prediction(BLUP)of each phenotypic value[20].Genotype,environment,genotype×environment interaction,and replicate nested in environment were all considered as random factors.

Fig.1–Differences in panicle traits between Gangyuan 8 and Liaoxing 1(A)and frequency distributions of the traits in the RIL population in three environments.BNP,primary branch number;BNS,secondary branch number;GNP,grain number on primary branch;GNS,grain number on secondary branch;GN,grain number per panicle;PL,panicle length;GD,grain density.

2.4.Genotyping

Genomic DNA was extracted from fresh leaves of each line using the cetyl trimethyl ammonium bromide method[21].RILs(n=197)and parents were genotyped with the 8 K-SNP chip by Beijing Golden Marker Biotechnology Co.,Ltd.(Beijing,China).The SNP markers were filtered as follows:ambiguously called SNPs and simplex and poor-quality SNPs with＞10% missing values or minor allele frequencies＜0.05 were removed.

2.5.Linkage mapping and QTL analysis

After removal of redundant markers,QTL IciMapping 4.2 software was used to generate an input file for genetic map construction[22].A combined nearest-neighbor algorithm and the 2-Opt algorithm of the traveling salesman problem[23]was used for high-density marker ordering.In singleenvironment analyses,inclusive composite interval mapping(ICIM)[24]was performed using the BIP function to identify additive QTL for phenotypic value and BLUP values of panicle traits.A 1000-permutation test was performed at a 95%confidence level and established a LOD threshold of 2.5.Multi-environment joint analysis employed ICIM in the MET functional module to identify QTL additive and QTL×environment interaction effects[25].

QTL were named using the abbreviated English name of traits preceded by“q”and followed by the chromosome number,plus,where appropriate,a number designating one of multiple QTL on the chromosome.The QTL identified in this study were compared with those in the Q-TARO database[26].For two of the detected QTL(qGD9 and qGN6),the 123 RILs were classified by the genotypes of the flanking markers on chromosomes 9 and 6,with 74 RILs excluded.

3.Results

3.1.Phenotypic analysis

The panicle traits showed significant differences(P＜0.01)between the parents,with the BNP,BNS,GNP,GNS,GN,PL,and GD of Liaoxing 1 higher than those of Gangyuan 8(Fig.1A).Wide variation was identified for these panicle traits in the RILs(Table 1).The traits of RILs followed approximately normal distributions in each environment,suggesting that they were controlled by multiple genes(Fig.1B–G).The highest mean values of BNP,BNS,GNP,GNS,GN,PL,and GD were observed in DG(Table 1).The ANOVAs for DG,SY,PJ and BLUP showed significant differences among genotypes for all traits.The coefficients of variation(CVs)of the traits in all environments ranged from 9.26%to 26.02%.All traits showed heritabilities＞50%.They showed many significant correlations(Table 2).GN was positively correlated with BNP(r=0.576,P＜0.01),GNP(r=0.486,P＜0.01),BNS(r=0.906,P＜0.01),GNS(r=0.957,P＜0.01),PL(r=0.309,P＜0.01)and GD(r=0.733,P＜0.01).

Table 1–Descriptive statistics of panicle traits in a RIL population grown in three environments.

Table 2–Correlations between panicle traits in the RIL population in three environments.

3.2.Genetic linkage map

A total of 559 SNPs covering the 12 chromosomes were polymorphic between Gangyuan 8 and Liaoxing 1.The removal of SNPs with＞10% missing data or segregation distortion left 285 SNPs for linkage map construction(Table 3).The number of SNPs per chromosome ranged from 7 on chromosome 1 to 54 on chromosome 5.The total linkage map length was 1951 cM,with a mean marker interval of 6.87 cM.

3.3.QTL analysis of panicle traits in individual environments and their BLUP values

A total of 105 QTL for seven panicle traits were detected in individual environments,including 25 QTL in DG,20 QTL in SY,26 QTL in PJ,and 34 QTL in BLUP values(Fig.2;Tables 4–6).Sixteen BNP QTL were detected in the DG(2),SY(2),PJ(8),and BLUP values(4),explaining 11.26%,25.37%,65.18%,and 41.26%of the phenotypic variation,respectively.Fifteen BNS QTL were detected in the DG(3),SY(3),PJ(4)and BLUP values(5),explaining 40.13%,30.06%,34.14%,and 49.63% of the phenotypic variation,respectively.Eleven GNP QTL were detected in the DG(3),SY(3),PJ(3),and BLUP values(5),explaining 29.81%,28.88%,29.30%,and 41.51% of the phenotypic variation,respectively.Twelve GNS QTL were detected in DG(2),SY(3),PJ(3)and BLUP values(4),explaining respectively 35.54%,35.69%,39.57%,and 52.70%of phenotypic variation. Seventeen GN QTL were detected in the DG (5), SY(3), PJ (3), and BLUP values (6), explaining respectively 38.25%,31.11%, 25.88%, and 50.01% of phenotypic variation. Twentyone PL QTL were detected in the DG (7), SY (3), PJ (3), and BLUP values (8), explaining respectively 66.48%, 57.75%, 43.21%, and 63.45% of phenotypic variation. Thirteen GD QTL were detected in the DG (3), SY (6), PJ (2), and BLUP values (2),explaining respectively 53.75%, 51.32%, 48.99%, and 55.81% of phenotypic variation. There were 30 single QTL with phenotypic contributions of 10%–30% and 11 single QTL with phenotypic contributions ＞30%. Most (79) of the favorable QTL alleles were from the high-yielding donor parent Liaoxing 1, with only 26 from Gangyuan 8.

Sixteen QTL were detected single-environmental analysis using BLUPs. Three QTL, qGD9, qPL9, and qGNP9 for GD, PL, and GNP, respectively, were detected simultaneously with BLUPs in at least two environments. qGD9, qGL9, and qGNP9 were confined to a co-localization region on chromosome 9 flanked by the SNP markers AX-153942631 and AX-115845259(Fig. 2). qGD9 and qGNP9 explained respectively 22.9%–52.2% and 11.0%–13.8% of total phenotypic variation, with the positive allele contributed by Liaoxing 1. qGL9 explained 17.68%–39.37% of the total phenotypic variation,with the negative allele contributed by Liaoxing 1. qGN6 and qBNS6.2 were detected simultaneously with BLUPs,consistent across two environments, and confined to another co-localization region on chromosome 6 flanked by AX-95955496 and AX-115755704 (Fig. 2). qGN6 and qBNS6.2 explained respectively 0.73%–4.99% and 1.93%–6.93% of total phenotypic variation, with the positive alleles from the same parent Liaoxing 1. qGNS7.2 was also simultaneously detected with BLUPs, consistent across two environments, and flanked by AX-95929493 and AX-115830795 on chromosome 6. qGNS7. explained 7.37%–24.80% of total phenotypic variation, with the positive allele from the same parent Liaoxing 1.

3.4. Joint analysis of panicle traits in multiple environments

Multi-environment analysis revealed 67 QTL, including 13 for BNP, nine for BNS, seven for GNP, seven for GNS, eight for GN,13 for PL, and 10 for GD (Table S1). These QTL explained respectively 58.57%, 47.08%, 45.68%, 57.09%, 86.26%, 72.81%,and 70.60% of phenotypic variation in BNP, BNS, GNP, GNS,GN, PL, and GD. The contribution percentage of the interaction among 67 additive QTL and environment ranged from 0.0017%to 23.66%.

Table 3–Numbers of SNPs on 12 chromosomes for mapping.

Fig.2–Physical map showing the QTL identified for panicle traits in RIL population by single-environment analysis.Red,known genes;green,known QTL;black,markers and QTL detected in this study.

Forty-nine QTL were detected in the joint analysis,occupying the same marker intervals as those detected in single-environment analyses.The joint analysis for additive QTL effects revealed lower contribution percentages than the single-environment analysis.Six single QTL showed phenotypic contributions of 10%–30%and five showed contributions＞30%.Of these,four QTL:qPL5.2,qBNP2.2,qPL9,and qGNS7.1 showed a significant additive effect(LOD(A)＞2.5),as well as QTL×environmental interaction effects(LOD(AbyE)＞2.5),but these additive contribution percentages were higher than those between the additive effect and environmental interaction,indicating that these additive effects were the main contributors to phenotypic variation.The three QTL qBNS7.1,qGNP8 and qGD9 had significant additive effects rather than additive-environmental interaction,which were considered to be expressed independently.However,only two QTL qPL9 and qGD9 were detected simultaneously for BLUP values and were consistent across at least two environments in individual environmental analyses.The three QTL qBNS6.2,qGN6 and qGNP9 with phenotypic contributions of 3.43%–8.09% were also detected simultaneously for BLUP values and were consistent across at least two environments in individual environment analyses.qBNS6.2,qGN6 and qGNP9 were expressed independently,given that these additive QTL effects,but not the QTL×environment interaction effects,were significant(LOD(A)＞2.5,LOD(AbyE)＜2.5).

3.5.Pyramiding effect of QTL on GN

The pyramiding effect of two stable QTL,qGD9/qPL9/qGNP9 and qGN6/qBNS6.2,was analyzed by comparing the mean PNs in different environments(Fig.3).The single QTL lines(n=14)with qGD9/qPL9/qGNP9 showed mean GN of respectively 172,163,149,and 159 in DG,SY,PJ,and BLUP,values higher than those(n=34)with no QTL(162,153,126,and 143 in DG,SY,PJ,and BLUP,respectively).Another single QTL line(n=31)with qGN6/qBNS6.2 showed mean GN of respectively 185,170,148,and 164 in DG,SY,PJ,and BLUP,values also higher than those in RILs that lacked the QTL.The two QTL lines(n=44)with both qGD9/qPL9/qGNP9 and qGN6/qBNS6.2 showed mean GN of respectively 194,180,158,and 172 in DG,SY,PJ,and BLUP,values higher than those in RILs that lacked QTL or had only a single QTL,particularly in PJ and BLUP(P＜0.05).

4.Discussion

All panicle traits of rice in this study,including GN,PL,BNP,BNS,GPN,GNS,and GD,are typical quantitative traits,which are not only controlled by major and minor genes but also influenced by the external environment.When the large-panicle-type cultivar Liaoxing 1 with high yield was used as the parent in breeding,conventional selection for the phenotypes of panicle traits was inefficient.The major-effect QTL identified from Liaoxing 1 can be more precisely introduced to diverse genetic backgrounds using MAS approaches[27,28],leading to faster improvement of panicle traits in rice breeding.In this study,a series of panicle traits of the RIL population was evaluated in three environments for QTL mapping by single-environment and joint analysis.

4.1.Phenotypic variation of panicle traits

There were significant differences in BNP,BNS,GPN,GNS,GN,PL,and GD between the parents of the RIL population.The phenotypes of these panicle traits of the male parent Liaoxing 1 were greater than those of the female parent Gangyuan 8.The RIL population showed wide phenotypic variation in these panicle traits,particularly for GNS and GN,and transgressive segregation was found in several lines.Their panicle phenotypes were larger than those of the parent Liaoxing 1.Owing to the difference in phenotypes between the two parents and the wide variation in the RIL population,it was a suitable subject for QTL mapping[29].

BLUP values based on measured values in three environments were used for QTL analysis to eliminate environmental and locational differences.The BLUP method was originally used by Henderson[30]for animal breeding and employed a selection index as well as least-squares methods[31].Random genetic effects and fixed environments were considered simultaneously.The prediction of BLUP values in multiple environments and lines of different generations increases the accuracy of BLUP value prediction[31].This method has been widely used in QTL mapping,genome-wide association analysis and genome selection in rice and other crops[32].

4.2.QTL analysis in single and multiple environments

In this study,105 QTL for panicle components were detected in single environments using their BLUP values.Of these,30 single QTL that explained 10%–30% of the phenotypic variation in single-environment analysis were designated as major-effect QTL and 11 single QTL with phenotypic contributions of more than 30% were designated as major QTL.As the expression of a QTL is affected by multiple environmental factors including location,year,and temperature,joint analysis in multiple environments can be used to estimate QTL×environment interaction effects[33].In this study,67 panicle-trait QTL showed much lower phenotypic contributions than those identified in single-environment analyses.The reason is that both QTL additive and QTL×environment interactive effects were estimated in the multienvironment joint analysis,while only the additive effects of QTL were considered in single-environment analyses.It is generally believed[34]that QTL detected in multiple environments are more stable than QTL with high effect values detected in a single environment and are of more use for MAS breeding.In this study,qGNS7.2 showed large effect values and was detected by single-environment analysis rather than by multi-environment joint analysis.We simultaneously detected 58 QTL by single-environment analyses and multi-environment joint analysis,and their additive effects had higher phenotypic contributions than QTL×environment interactive effects.However,most of these QTL were detected in only one environment,illustrating that they showed environmental specificity and were not stably expressed.qGD9,qGNP9,qPL9,qGN6,and qBNS6.2,which were identified with BLUP values and were consistent across at least two environments,can be considered stable QTL for panicle traits.

Fig.3–Pyramiding effect of two stable QTL on grain number per panicle(GN)in DG(A),SY(B),PJ(C),and BLUP(D).Boxes are quartiles,continuous lines are medians,whiskers include others that are not outliers.Different letters for the mean values indicate significant differences at P＜0.05.

4.3.Novel QTL for panicle traits and their application in rice breeding for high yield potential

The four QTL qBNP3,qGN7.1,qBNS7.1,and qGNS7.2 were detected not only in one environment but also by BLUP values,and these positive alleles were contributed by Liaoxing 1.These QTL were compared to QTL that had been previously reported for the same or related traits using the physical positions of the corresponding flanking markers.qBNP3 was detected in an approximate 4.7 Mb interval on chromosome 3 not only in a single environment but also by BLUP values in multiple environments.It shared a region with a QTL and two known genes(fc1/OsTB1 and OsMADS341)for panicle traits in previous studies[35,36].By whole-genome resequencing,compared to Gangyuan 8(data not shown),Liaoxing 1 showed only four InDels and one SNP in the non-coding region of OsMADS34(Os03g0756100).The three QTL qGN7.1,qBNS7.1,and qGNS7.2 mapped to a 0.3 Mb interval on chromosome 7 and shared the region with a QTL sp2(t)for PL[37],which has not yet been cloned.These results suggest that these QTL represent novel genes controlling BNP,GNS,or GN,which may contribute to more grains on the panicles of Liaoxing 1.

We also identified a co-localization region on chromosome 9 containing three QTL,qGD9 for GD,qPL9 for PL,and qGNP9 for GNP,which explained respectively 22.94%–52.22%,17.68%–39.37%,and 10.87%–13.81% of total phenotypic variation.Based on comparison with the reported QTL or genes,both qGD9 and qPL9 may be the DEP1 that controls panicle morphology[38],GN[11],and nitrogen use efficiency[19,39].In our previous study,the dep1 allele led to an increase in GD but a decrease in PL[40].In the present study,a significant negative relationship between GD and PL was observed,and qGD9 or qPL9 had a positive additive effect on GD but a negative additive effect on PL.The qGD9 or qPL9 with Liaoxing 1 allele showed a positive additive effect on GN in only one environment(qGN9 in PJ).In view of the large difference in GN between parents,there should be other QTL for GN in this population.

In addition to the co-localization region on chromosome 9 described above,qGN6 and qBNS6.2 were co-localized in a 0.94 Mb interval on chromosome 6 between AX-95955496 and AX-115755704.The position differed slightly from those of known QTL cloned as single genes,including TGW6[41],GL6[42],SDT[43]and GW6a[44].The QTL qGN6 and qBNS6.2 explained respectively 0.73%–4.99% and 1.93%–6.93% of the phenotypic variation for GN and BNS,Their functions also differed from TGW6(grain weight)[41],GL6(grain length)[42],SDT(tiller numbers)[43],and GW6a(grain weight)[44].These suggested that qGN6 and qBNS6.2 could be considered as promising candidate loci for in-depth study to identify the molecular regulatory mechanism for panicle components.

Although the two QTL qGN2.1 and qGN2.2 for GN explained more than 10% of phenotypic variation,they were detected only by single-environment and not multi-environment joint analysis.Another major-effect QTL qGN7.1 was detected by both single and multi-environment analysis,but it was expressed only in SY.No major effect or major QTL for GN was stably expressed,suggesting that the greater GN in Liaoxing 1 could be attributed to positive alleles of minor QTL.It has been difficult to increase GN by conventional breeding,even though Liaoxing 1 was widely used as a parent in the previous decade.

Marker-assisted pyramiding of QTL is an effective way to improve quantitative traits [45]. During QTL pyramiding,both the number of QTL pyramided and their combined effects must be considered [46]. In this study, we determined the pyramiding effect of these two stable QTL of

qGD9/qPL9/qGNP9 and qGN6/qBNS6.2 in a single environment. Although the effect of QTL pyramiding should be examined in more detail using MAS of QTL in a natural population [46], these results suggest that pyramiding qGD9.2/PL9 and qGN6/qBNS6 would lead to high yield in the JSR cultivar Liaoxing 1.

5.Conclusions

Using phenotypic values and BLUP values of panicle traits in a RIL population, we detected 105 QTL associated with panicle traits in single-environment analyses. Of these, 49 QTL were also identified by joint multi-environment analysis. Five stable QTL, qGD9, qPL9, qGNP9, qGN6, and qBNS6.2, were identified for PL, GNP, GD, BNS, and GN, respectively, in multiple environments. Co-localization of qGD9, qPL9, and qGNP9 corresponded to a known gene, DEP1. However,because qGN6 and qBNS6.2 co-localized in an interval that has not previously been reported, they should be considered a novel QTL. Several RILs with combined QTL of qGD9/qPL9/qGNP9 and qGN6/qBNS6.2 showed greater GN and could be used as donors for JSR breeding. To investigate the molecular mechanisms leading to high yield in JSR, the major QTL qGN6/qBNS6 should be the focus of further fine mapping and cloning.

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2020.07.002.

CRediT authorship contribution statement

Guomin Sui and Wenjing Zheng initiated the research.Mingzhu Zhao designed and conducted all the experiments.Zuobin Ma,Lili Wang,and Hong Gao genotyped RILs.Zhiqiang Tang,Ting Mao,Liying Zhang,Na He,Liang Fu,and Changhua Wang conducted the field experiments and measured phenotypes.Mingzhu Zhao and Wenjing Zheng prepared the manuscript.

Declaration of competing interest

Authors declare that there are no conflicts of interest.

Acknowledgments

The study was supported by the National Natural Science Foundation of China(31901526),China Postdoctoral Science Foundation Grant(2019M651139)and Liaoning Key Agricultural Program(2019JH1/10200001–2).