Dissecting Genetic Basis of Deep Rooting in Dongxiang Wild Rice

NIE Yuanyuan,XIA HuiMA XiaosongLOU QiaojunLIU YiZHANG AnlingCHENG LiangYAN Longan,LUO Lijun

(1Huazhong Agricultural University,Wuhan 430070,China;2Shanghai Agrobiological Gene Center,Shanghai 201106,China;3Jiangxi Research and Development Center of Super Rice/ Nanchang Branch of Chinese National Center for Rice Improvement,Nanchang 330200,China)

Abstract:Deep rooting is an important trait in rice drought resistance.Genetic resources of deep-rooting varieties are valuable in breeding of water-saving and drought-resistant rice.In the present study,234 BC2F7 backcross introgression lines were derived from a cross of Dongye 80 (an accession of Dongxiang wild rice as the donor parent) and R974 (an indica restorer line as the recurrent parent).A genetic linkage map containing 1 977 bin markers was constructed by ddRADSeq for QTL analysis.Thirty-one QTLs for four root traits (the number of deep roots,the number of shallow roots,the total number of deep roots and the ratio of deep roots) were assessed on six rice chromosomes in two environments (2020 Shanghai and 2021 Hainan).Two of the QTLs,qDR5.1 and qTR5.2,were located on chromosome 5 in a 70-kb interval.They were detected in both environments.qDR5.1 explained 13.35% of the phenotypic variance in 2020 Shanghai and 12.01% of the phenotypic variance in 2021 Hainan.qTR5.2 accounted for 10.88% and 10.93% of the phenotypic variance,respectively.One QTL (qRDR2.2) for the ratio of deep roots was detected on chromosome 2 in a 210-kb interval and accounted for 6.72% of the phenotypic variance in 2020.The positive effects of these three QTLs were all from Dongxiang wild rice.Furthermore,nine and four putative candidate genes were identified in qRDR2.2 and q DR5.1/qTR5.2,respectively.These findings added to our knowledge of the genetic control of root traits in rice.In addition,this study will facilitate the future isolation of candidate genes of the deep-rooting trait and the utilization of Dongxiang wild rice in the improvement of rice drought resistance.

Key words:Dongxiang wild rice;backcross introgression line;deep rooting;genetic analysis;QTL

Rice is an important food crop.More than half of the world’s population consume rice as their staple food.Drought is a limiting factor for the high yield of rice.At present,with the continual reduction of freshwater resources and increase of drought frequency,drought resistance improvement of rice has become an important research focus (Kondo et al,2000,2003;Uga et al,2011).The mechanism of drought resistance in rice is complex and includes drought avoidance,drought tolerance,drought escape and drought recovery (Luo,2010).Drought avoidance is the first line of defense against drought stress,and it plays a major role in enhancing the drought resistance of plants (Blum,2005).Root is the most important organ for absorbing and translocating water and nutrients from the soil,and thus,the plants’ ability to avoid drought stress depends mainly on their root performance.Plants with deep rooting can access water from deep soil layers,thereby avoiding drought stress (Yoshida and Hasegawa,1982;Fukai and Cooper,1995;Uga et al,2011).Therefore,the number of deep roots (DR) and the ratio of deep roots (RDR) are key parameters that determine the water-absorbing ability from deep soil layers.Modifying the root distribution of rice from shallow rooting to deep rooting is a promising strategy for drought resistance breeding (Gowda et al,2011;Uga et al,2011).

Deep rooting is a complex quantitative trait that is mainly determined by the root growth angle and the maximum root length (Abe and Morita,1994;Araki et al,2002).At present,the most widely used method to determine deep rooting is the ‘basket’ method.Its evaluation index is RDR (Kato et al,2006;Uga,2012).To date,a number of QTLs related to root morphology have been identified,but only 11 QTLs for RDR have been reported (Uga et al,2011,2013a,2015;Kitomi et al,2015;Lou et al,2015).Among these QTLs,onlyDro1gene has been cloned and demonstrated to improve drought resistance (Uga et al,2013b).However,the germplasm used to identify QTLs for DR and RDR is very limited.Most of the reported deep-rooting QTLs were identified from the same deep-rooting varieties,Kinandang Patong (Uga et al,2011,2013a,2015) and IRAT109 (Lou et al,2015).

Dongxiang wild rice is a common wild rice (Oryza rufipogonGiff.) that originated from Dongxiang county (28°14′ N,116°36′ E),Jiangxi Province,China.Dongxinag is the northernmost region in the world whereO.rufipogonhas been found to date (Zhang et al,2016;Liang et al,2018).Previous studies have shown that Dongxiang wild rice can survive under extreme drought stress conditions and has an extremely well-developed root system,especially in the root length,suggesting that it is a valuable genetic resource for the improvement of rice drought avoidance (Zhang et al,2006;Qi et al,2020).However,QTLs from wild rice for improving DR and RDR have not been reported.

In this study,we conducted QTL analysis of four root traits in Dongxiang wild rice,that are DR,the number of shallow roots (SR),the total number of roots (TR) and RDR,using an advanced backcross introgression line (BIL) population derived from a cross between Dongye 80 and R974.Thirty-one QTLs,including nine for DR,nine for SR,seven for TR and six for RDR,were detected.These QTLs can provide a better understanding of the genetic basis of root traits and establish a foundation for application in molecular breeding programs.

RESULTS

Construction of population genetic linkage map

A genetic linkage map containing 1 977 bins was constructed based on 71 489 single nucleotide poly-morphism (SNP) markers.The length of the genetic map was 1 156.241 cM in total,with a mean distance of 0.017 cM per marker (Table S1).Chromosome 1 contained 9 777 SNP markers with a length of 142.201 cM containing 269 bins.The average distance between markers on chromosome 1 was 0.015 cM.The number of markers on chromosome 12 was the least (4 055).The length of chromosome 12 was 70.194 cM,with an average distance of 0.017 cM,containing 132 bins.The 71 489 markers were distributed evenly on 12 chromosomes.The maximum and the minimum distances between markers were 9.991 and 0.280 cM,respectively,and the average distance was 4.477 cM.

Phenotypic evaluation of root traits

The root traits including DR,SR,TR and RDR were investigated among the two parents and BILs.Significant differences in the four traits were found between the two parents in both trials (Table 1).Dongye 80,an accession of Dongxiang wild rice as the donor parent,performed better than R974,anindicarestorer line as the recurrent parent,for DRD in both years,indicating that Dongye 80 can potentially improve the deep root ratio of rice.

For the BIL population,the variations in DR,SR,TR and RDR showed a good fit to a normal distribution both in Shanghai and Hainan in China (Fig.1).In 2020 Shanghai,the values of DR,SR,TR and RDR of BILs ranged from 19.8 to 227.8,from 55.5 to 595.8,from 77.8 to 734.0,and from 0.07 to 0.52,respectively,with respective averages of 67.5,233.1,298.9 and 0.22.The corresponding coefficients of variation (CVs) of DR,SR,TR and RDR were 53.6%,33.2%,33.7% and 36.6%,respectively.In 2021 Hainan,the values of DR,SR,TR and RDR of BILs ranged from 27.8 to 194.8,from 23.0 to 335.7,from 67.3 to 519.0,and from 0.17 to 0.66,respectively,with respective averages of 104.6,183.0,287.6 and 0.36.The corresponding CVs of DR,SR,TR and RDR were 31.0%,27.2%,25.6% and 19.2%,respectively (Table 1).These results indicated that the DR,SR,TR and RDR of BILs all showed significant variations among different lines,both in Shanghai and Hainan.The variation range was greater in Shanghai,and the standard deviation and CV were lower in Hainan.

Table 1.Performances of DR,SR,TR and RDR in parents and backcross introgression lines in 2020 and 2021.

Correlations among measured root traits

Pearson’s correlation analyses were applied to the traits measured in 2020 and 2021 (Table 2).All the four traits measured in 2020 and 2021 were significantly (P＜ 0.05) correlated,indicating there is a good repetition between the two years.RDR was positively correlatedwith DR,but negatively correlated with SR in both years.In 2020,RDR was positively correlated with TR,while in 2021,the correlation was not significant.The results suggested that Dongxiang wild rice can improve RDR by increasing the number of deep roots.

Table 2.Correlations among DR,SR,TR and RDR in backcross introgression lines measured in 2020 and 2021.

QTL mapping of root-related traits

A total of 31 QTLs associated with DR,SR,TR and RDR were detected in the BIL population (Table 3).For 15 QTLs,Dongye 80 provided the superior alleles (Fig.2),while for the other 16 QTLs,the superior alleles were from R974.Two QTLs,one for DR and the other for TR,were identified from Dongye 80 in both years.Eight QTLs,qDR7.2andqDR7.3,qDR10.1andqDR10.2,qSR1.1andqSR1.2,andqSR6.1andqSR6.2,were detected in adjacent positions on the same chromosome during the two years of evaluation.

Nine QTLs associated with DR were identified on chromosomes 1,4,5,6,7 and 10,of which four superior alleles were derived from Dongye 80.qDR10.2explained 15.87% of the phenotypic variance,with a LOD score of 11.95.qDR5.1was identified in both years and explained 13.35% and 12.01% of the phenotypic variance,with LOD scores of 10.37 and 5.91 in Shanghai and Hainan,respectively.Nine QTLs associated with SR were identified on chromosomes 1,3,5,6,10 and 12,of which five superior alleles were derived from Dongye 80.qSR10.1explained 7.18% of the phenotypic variance with a LOD score of 4.97,andqSR5.2was detected in the same position asqDR5.1,explaining 6.47% of the phenotypic variance with a LOD score of 3.64.Seven QTLs associated with TR were identified on chromosomes 1,3,5,6,10 and 12,of which three superior alleles were derived from Dongye 80.qTR5.1explained 17.65% of the phenotypic variance,with a LOD score of 2.51.qTR5.2was detected during both years and explained 10.88% and 10.93% of the phenotypic variance,with LOD scores of 6.94 and 5.76,respectively.Six QTLs associated with RDR were identified on chromosomes 1,2,5 and 7,of which the superior alleles of three QTLs were derived from Dongye 80.qRDR2.2explained 6.72% of the phenotypic variance,with a LOD score of 4.51,andqRDR2.1explained 5.44% of the phenotypic variance,with a LOD score of 3.72.Another QTL,qRDR5.1on chromosome 5,explained 4.80% of the phenotypic variance,with a LOD score of 3.38.

Colocalization and stability of QTLs associated with root traits

To investigate the genetic effects of the QTLs responsible for root traits,all QTLs with superior alleles at the two sites from Dongye 80 were further analyzed.The QTLsqDR5.1andqTR5.2were colocalized on chromosome 5 in a 70-kb interval in the both years.The QTL hotspotqDR5.1/qTR5.2/qSR5.2explained 13.35% of the phenotypic variance for DR,10.88% for TR and 6.47% for SR,indicating that this interval may play an important role in rice root traits.Additionally,this pleiotropic QTL was first detected in wild rice.Meanwhile,another QTL,qRDR2.2,detected in 2020 Shanghai and explained 6.72% of the phenotypic variance with an LOD score of 4.51,had a significant effect on RDR.

Confirmation of qRDR2.2 and qDR5.1 for root traits

A main QTLqRDR2.2was detected on chromosome 2 in a 210-kb interval with a large contribution rate for RDR.To further confirm the function of the newly identified genetic locus ofqRDR2.2in Dongye 80,high-resolution mapping was performed with several BILs,including BIL24,30,51,53,76,84,136,168 and 229 as well as the recurrent parent R974.High-resolution mapping of several BILs possessing the superior Dongye 80 allele atqRDR2.2helped narrow down the QTL to an interval between c02b010 the BIL population was selected to verify the QTL cumulative effect.The RDR value of the lines without QTLs for RDR was 0.197,that of the lines pyramidingqRDR2.1,qRDR2.2andqRDR5.1was 0.401.The RDR value of the lines with pyramidedqRDR2.2andqRDR5.1was 0.286,significantly higher than those without QTLs for RDR.For lines containing onlyqRDR5.1,the RDR value was 0.280,significantly higher than those without QTLs for RDR(Fig.4-A).Due to the large coverage range ofqRDR2.1and the influence of population size,no single lines containingqRDR2.1were selected.Based on the above analysis,we concluded that the additive effect was the main effect when the QTLs for RDR were aggregated in the BIL population of Dongye 80.and c02b013 bin markers (Fig.3-A).Among the BILs,the average RDR of the lines with the Dongye 80 allele atqRDR2.2was significantly higher than that in the recurrent parent R974 (P＜ 0.01).These results indicated thatqRDR2.2from Dongye 80 significantly increased RDR.

Similarly,to confirm the function of the newly identified genetic locus ofqDR5.1in Dongye 80,high-resolution mapping delimitedqDR5.1to the tightly linked bin markers c05b004 and c05b005 (Fig.3-B).Then,we compared the genetic effects ofqDR5.1on deep rooting numbers in the BILs.The results showed that the average DR of BILs withqDR5.1was significantly higher than that in the recurrent parent R974 (P＜ 0.01).These results indicated thatqDR5.1derived from Dongye 80 can significantly increase the deep root number.

QTL cumulative effect analysis

We analyzed the additive effect of QTLs for RDR in Dongye 80.First,qRDR1.1,qRDR1.2andqRDR7.1provided by recurrent parent R974 were fixed by sequencing result,and then,qRDR2.1,qRDR2.2andqRDR5.1with the enhancing allele from Dongye 80 in Similarly,we also analyzed the additive effect of the QTLs for DR in Dongye 80,i.e.,qDR4.1,qDR5.1,qDR7.2andqDR10.2.The deep root number in lines without QTLs for DR was 58.5,and that in lines pyramidingqDR4.1,qDR5.1,qDR7.2andqDR10.2was 113.0.The deep root number in lines with pyramidedqDR4.1andqDR10.2was 108.9 and that with pyramidedqDR5.1andqDR10.2was 89.6,both were significantly higher than those without QTLs for DR.For lines containing onlyqDR5.1,the deep root number was 88.1,and for lines containing onlyqDR10.2,the deep root number was 88.8.All were significantly higher than those without QTLs for DR(Fig.4-B).Based on the above analysis,it was determined that the additive effect was the main effect when the QTLs for DR were aggregated in the BIL population of Dongye 80.

Association analysis in QTL interval

A natural population containing 175 landraces was used to determine the reliability of QTLs by identifying the associated SNPs in target intervals ofqRDR2.2andqDR5.1.QTL interval-based association analysis of root traits identified 15 SNPs involved in 3 genes belonging to 2 QTL intervals (Table S3).In the interval of c02b009-c02b015,there were three SNPs inLOC_Os02g01355.In the interval of c05b004-c05b005,there were 12 SNPs related to root traits in 2 genes (LOC_Os05g01030andLOC_Os05g01040).

Transcriptome analysis of target intervals under six hormone treatments

Deep rooting depends on root gravitropism which is regulated by phytohormone (Aloni et al,2006;L?fke et al,2013;Uga et al,2013b;Singh et al,2014).If a gene is in response to a hormone related to root development and/or osmotic stress,it is of higher possibility in regulating RDR.We thus investigated the expression levels of genes in the identified QTLsqRDR2.2andqDR5.1under treatments of various phytohormones and polyethylene glycol (PEG)-simulated osmotic stress by RNA-Seq.The transcriptome data under six hormone treatments found 17 putative genes that were involved in the responses to different hormones.Interestingly,there were 11 cloned genes that were related to root development and growth.Among the 129 predictive genes in the interval of c02b009-c02b015,there were 8 genes involved in the response to one or more hormone treatments.Among the seven predictive genes in the interval of c05b004-c05b005,two genes were involved in the response to one or more hormone treatments (Table S2).Combining the results of gene expression response to diverse hormones and the association analysis,we finally identified 13 candidate genes for further investigation (Table 4).

DISCUSSION

Drought resistance is an important and complex trait that is expressed in the survival and production capacity of plants under drought conditions (Luo,2010).According to the responses of plants to drought stress,drought resistance can be divided into four types:drought avoidance,drought tolerance,drought escape and drought recovery (Turner,1996;Luo and Zhang,2001).Drought avoidance refers to the ability of plants to maintain high water status by increasing water absorption or reducing water diversion loss under drought conditions.It is achieved by obtaining water from deep soil through large and deep roots and reducing transpiration by closing stomata or via impermeable leaf cuticles.As the major organresponsible for water absorption,the roots,especially deep roots,play a vital role in plant drought resistance (Gowda et al,2011).Studies have shown that the deep root structure of upland rice can make more complete use of water in deep soil to avoid drought stress (Price et al,2002).However,there have been only a few studies related to deep rooting (Uga et al,2011,2013a,b,2015;Kitomi et al,2015;Lou et al,2015;Xia et al,2019).In order to make full use of the resource advantages of Dongxiang wild rice with strong drought resistance,we conducted genetic studies on deep root traits of Dongye 80.We constructed a BIL population of Dongye 80 and identified QTLs related to root traits through ddRADSeq of the population (Table 3).

Table 4.Putative genes at two QTL regions under hormone treatment and association analysis in rice.

To date,many QTLs related to root morphology have been identified,but only 11 QTLs controlling root growth angle have been reported,including 10 QTLs for deep root ratio and 1 QTL for shallow roots,while onlyDro1has been successfully cloned (Table S4).The first RDR-related QTL (namedDro1) was identified by Uga et al (2011,2012,2013a),and it can explain 66.6% of the total phenotypic variance.The mutation ofDro1is caused by 1 bp deletion within the exon 4 on chromosome 9 in the shallow root parent IR64,and it is only found in several IR64 progeny lines (Uga et al,2013a).Based on the Illumina HiSeq results of this locus in the parents of the BILs,this special mutation did not exist in the materials used in this study.QTLsqDR-10andqTR-10,mapped onto RM467-RM596,were detected by Lou et al (2015).In addition,the chromosomal region of the QTL for RDR on chromosome 7,qRDR7.1,is very close toDro3identified by Uga et al (2015),and also has an overlapping region withqRDR-7identified by Lou et al (2015).qRDR2.1identified in this study has an overlapping region withDro4identified byKitomi et al (2015).The BIL population used in this study has identified many QTLs related to deep roots,and some of these QTLs coincided with the QTLs for RDR located on chromosomes 2,7 and 10 (Uga et al,2011,2013a,2015;Lou et al,2015;Kitomi et al,2015).On chromosome 5,we identifiedqDR5simultaneously controlling DR,SR and TR with significant explained phenotypic variances during the two years.qRDR5.1was first detected in Dongxiang wild rice Dongye 80,indicating Dongye 80 contains undiscovered genetic resources for drought resistance.Therefore,further mining,research and utilization of genetic resources from Dongxiang wild rice are required.

Transcriptome data under hormone treatment and association analysis of the target section provide useful information for identifying putative candidate genes.There are many known genes in the target regions,such asOsMT2b(LOC_Os05g02070),Nrat1(LOC_Os02g03090),OsSPL3(LOC_Os02g04680) andOsCPK4(LOC_Os02g03410).OsCPK4andOsMT2bare involved in the responses to auxin and root development (Yuan et al,2008;Campo et al,2014).cDNA ofOsMT2bhas a total length of 679 bp and contains four exons,encoding a protein composed of 84 amino acids that negatively regulates the cytokinin content in rice roots.OsMT2bmay control the development of rice roots and the germination of seed embryos by regulating the cytokinin content in rice roots and embryos (Yuan et al,2008).In addition,it may participate in the regulation of root distribution.However,the clearer relationships ofOsMT2bwith DR and RDR should be further validated.Meanwhile,the beneficial alleles of the above mentioned candidate genes hidden in the Dongxiang wild rice require further mining and testing.

In rice breeding,materials with developed roots and a high deep root ratio can better absorb underground water and have superior drought resistance.Here,two newly identified QTLs,qRDR2.2andqDR5.1,were shown to improve both deep root numbers and deep root ratios in rice.These QTLs can thus serve as new resources for breeding green super rice (GSR) (Zhang,2007) and water-saving and drought-resistant rice (WDR) (Luo,2010).The findings also give us the cue that Dongxiang wild rice is a valuable genetic resource in breeding GSR and WDR.

METHODS

Rice materials

F1plants were obtained from Dongxiang wild rice Dongye 80 as the female parent,with earlyindicarestorer line R974 as the male parent.In 2012 and 2013,R974 was used as a recurrent parent to obtain BC1F1and BC2F1,respectively.Then,BC2F1was continuously selfed to obtain the BC2F7generation by the single seed method.Consequently,a BIL population including 234 lines was obtained (Fig.S1).

Genetic map construction

The genetic linkage map was constructed referring to the method proposed by Xie et al (2010).The main steps were as follows:1) The genotype of the parent was inferred according to the linkage relationship between markers in the offspring population,and the genotype of the offspring was transformed into A or B according to the genotype of the parent.At the same time,the genotype could also be compared with the actual parental genotypes to estimate the authenticity of parental materials.2) Missing genotypes were filled in based on the Hidden Markov model,and wrong genotypes were corrected.3) The recombination rate between markers was evaluated according to the method described by MSTMap,and the genetic map distance between markers was calculated by the Kosambi mapping function.

The bin genetic maps of Dongxiang wild rice and R974 BIL population were constructed by the ‘sliding window’ method.The bin linkage map contained 58 738 recombination hotspots.These markers covered most of the recombination hotspots (Fig.S2).On average,1 977 bin markers were covered on each chromosome,and the average marker interval was 128.8 kb.

Field experiment and investigation of root traits

The deep-rooting traits were evaluated using the basket method with minor modifications (Uga et al,2011).A total of 1 416 baskets with a top diameter of 17 cm,a bottom diameter of 10 cm,and a height of 7 cm were used for planting according to the density of three baskets per row,two rows per material,15 cm plant spacing and 20 cm row spacing between the edges of the baskets.In the process of planting,the top edge of the basket,the soil surface inside the basket and the soil surface outside the basket were set at the same level,and the rice seedlings were planted at the center of the basket (Fig.S3).

A total of 234 lines of BIL population and two parents were planted at the Jinshan experimental station,Shanghai City,China,in the summer of 2020 and at the Lingshui experimental station,Hainan Province,China,in 2021.After germinating in a greenhouse at 28 °C for 12 d,the young seedlings were transplanted into baskets in the fields and were gently pulled out from the soil at 40 d later.The roots emerging from the meshes of the baskets were counted,and those from the bottom and sides were regarded as DR and SR,respectively.TR was a sum of DR and SR,and RDR was counted by the ratio of DR to TR.

Statistical,QTL and putative gene analyses

Microsoft Office Excel 2017,SPSS Statistics 22,and GraphPad Prism 5.0 were used for statistical analysis.Mean values of each trait were used for further QTL analysis.The standard deviation of the mean was calculated using the Microsoft Excel software.The correlation analysis (Pearson’s coefficient method) was applied among the deep-rooting related traits via SPSS Statistics 22.

A total of 1 977 bin markers were used to construct the genetic linkage map covering the whole genome.Inclusive composite interval mapping combined with additive mapping was used to detect more precise drought resistance QTLs (Lou et al,2015).The significant logarithm of odds (LOD ＞ 2.5) value threshold for each trait was determined following a 5% permutation test with 1 000 replicates.The nomenclature of QTLs was based on the nomenclature proposed by McCouch and CGSNL (2008).The putative genes on the QTL intervals were identified based on the Rice Information GateWay.

Investigation of gene expression in response to various

hormones and osmotic stress

Deep rooting is regulated by both hormones and osmotic stress (Uga et al,2011).To investigate the expression levels of genes in the identified QTLs under different phytohormones and drought stress by RNA-Seq,two-week-old rice seedlings (Nipponbare) were treated with 15% PEG,100 μmol/L abscisic acid and 100 μmol/L ethephon for 2 h and with 50 μmol/L gibberellin,50 μmol/L auxin and 50 μmol/L cytokinin for 4 h.The shoot samples with three biological replicates were harvested and sent for RNA-Seq at Shanghai Majorbio Biopharm Technology Co.Ltd and Shanghai Biozeron Biotech.Co.Ltd,China,using an Illumina X Ten platform (2 × 151-bp read length).The raw paired-end reads were quality controlled by Trimmomatic (http://www.usadellab.org/cms/uploads/supplementary/ Trimmomatic),and the raw data were deposited in the NCBI Sequence Read Archive (PRJNA609211).Clean reads were then separately aligned to the reference genome in orientation mode using hisat2 (https://ccb.jhu.edu/software/hisat2/index.shtml).We then used htseq to count each gene read (https:// htseq.readthedocs.io/en/release_0.11.1/),and determined the gene expression levels with the fragments per kilobase of exon per million fragments mapped method.The expression levels of genes within the QTLs were extracted.Independentttest (P＜ 0.05) was applied to determine differentially expressed genes between the treated samples and untreated control.

Identification of candidate genes among QTL intervals by association analysis

We performed QTL interval-based association analysis to examine whether genetic variants in these QTL intervals harbored association signals with root traits.The deep-rooting traits and genotype data were extracted from a previous study (Xia et al,2019) which included 367 genotypes and 9 768 SNPs.The low-frequency alleles were filtered with minimum allelic frequency ＞ 0.05.The QTL interval-based association analysis was conducted for the root traits using the general linear model in Tassel 5.0 (Bradbury-Jones et al,2007).

ACKNOWLEDGEMENTS

This study was supported by the National Modern Agricultural Industry Technology System Construction Program of China (Grant No.20212BBF63001),the Open Competition Program of Jiangxi Provincial Science and Technology in China (Grant No.20213AAF01001),the Jiangxi Provincial Science and Technology Support Program in China (Grant No.20203BBF63033),the Jiangxi Modern Agricultural Research Collaborative Innovation Project in China (Grant No.JXXTCX202111),and the Open Project of State Key Laboratory of Rice Biology in China (Grant No.20200101).

SUPPLEMENTAL DATA

The following materials are available in the online version of this article at http://www.sciencedirect.com/journal/rice-science;http://www.ricescience.org.

Fig.S1.Population construction of backcross introgression lines.

Fig S2.Genetic map and genotype of Dongye 80/R974 backcross introgression lines.

Fig.S3.Root architectures of backcross introgression line parents and baskets in field planting.

Table S1.Statistic of genetic linkage map information.

Table S2.Putative genes at two QTL regions under hormone treatments associated with root traits in rice.

Table S3.Association analysis with root traits in target regions of putative genes in rice.

Table S4.QTLs for deep root traits in different populations.