Genetic mapping of QTL for agronomic traits and grain mineral elements in rice

Gwen Iris Deslsot-Empleo,Amery Amprdo, Mry Ann Inbngn-Asilo,Frnes Tesoro, Jmes Stngoulis, Russell Reinke, B.P. Mllikrjun Swmy,*

aStrategic Innovation Platform,International Rice Research Institute(IRRI),Metro Manila 1226,Philippines

bDepartment of Plant Breeding and Genetics,College of Agriculture,University of Southern Mindanao,Kabacan 9407,Cotabato,Philippines

cCollege of Science and Engineering,Flinders University,Adelaide 5001,Australia

Keywords:Rice Quantitative trait loci Biofortification Mineral elements DH Epistasis

A B S T R A C T Malnutrition is one of the prevailing health problems worldwide, affecting a large proportion of the populations in rice-consuming countries. Breeding rice varieties with increased concentrations of elements in the grain is considered the most cost-effective approach to alleviate malnutrition. Development of molecular markers for high grain concentrations of essential elements, particularly Zn, for use in marker-assisted selection(MAS) can hasten breeding efforts to develop rice varieties with nutrient-dense grain. We performed QTL mapping for four agronomic traits: days to 50% flowering, plant height,number of tillers,grain yield,and 13 grain elements:As,B,Ca,Co,Cu,Fe,K,Mg,Mn,Mo,Na,P,and Zn,in two doubled-haploid populations derived from the crosses IR64 × IR69428 and BR29 × IR75862. These populations were phenotyped during 2015DS and 2015WS at IRRI,Los Ba?os, The Philippines, and genotyped them with a 6 K SNP chip. Inclusive composite interval mapping revealed 15 QTL for agronomic traits and 50 QTL for grain element concentration. Of these, eight QTL showed phenotypic variance of ＞25% and 11 QTL were consistent across seasons.There were seven QTL co-localization regions containing QTL for more than two traits. Twenty five epistatic interactions were detected for two agronomic traits and seven mineral elements. Several DH lines with high Fe and Zn in polished rice were identified.These lines can be used as donors for breeding high-Zn rice varieties.Some of the major QTL can be further validated and used in MAS to improve the concentrations of nutritive elements in rice grain.

1.Introduction

Rice is the major staple food and source of energy and nutrition for the largest number of people worldwide [1].Most of the popular modern and high-yielding rice varieties,however, supply suboptimal amounts of minerals and vitamins,especially when consumed in the polished form[2].An adequate intake of mineral- and vitamin-rich food is necessary for achieving and maintaining good health, but the poor and a large proportion of the populations in developing countries lack access to a diversified diet and suffer from malnutrition or hidden hunger[3].

More than two dozen elements are essential for maintaining good health in humans. Some of them form the basic structure of bones and teeth, vital organs, and the immune system and participate in cellular metabolism by serving as cofactors for various enzymes [4]. Beneficial elements required in trace quantities are B, Co, Cu, Fe, Mn, Mo, and Zn,whereas Ca,K,Mg,Na,and P are required in larger quantities.In contrast, elements such as Cd and As are potentially toxic to human health [5]. Among the micronutrients, Fe and Zn deficiencies are the most widespread in developing countries[6]. It is estimated [7,8] that a third of the world's population,especially preschool children and pregnant and lactating women, suffer from Fe and Zn deficiencies. Zn deficiency causes stunting, diarrhea, reduced immunity, poor cognitive development, and skin problems, whereas Fe deficiency causes anemia, weakness, and dizziness [9-11]. Addressing micronutrient malnutrition to prevent mortality of children and women is one of the major sustainable development goals of United Nations (https://sustainabledevelopment.un.org/sdgs).

Deficiencies in essential elements could be remedied by dietary diversification, supplementation, and postharvest food fortification. Although these approaches are practiced and remain of great importance,efforts have shifted towards biofortification, which increases the content of bioavailable nutrients and vitamins of food crops through plant breeding[12-14].Biofortification is a cost-effective approach to address widespread dietary deficiencies in the human population.Development of biofortified rice varieties with increased levels of bioavailable major nutrients and low levels of toxic elements is desirable for improved health and nutrition.

The concentrations of key elements in grain are influenced by complex genetics and are influenced by environmental factors. An understanding of their genetic basis, associations with one another, and interactions with agronomic traits is essential for developing successful rice varieties with enhanced nutritive value[2].QTL analysis for complex traits can detect association between markers and traits that can lead to elucidation of the genetic control of complex traits.The use of markers closely linked to QTL will allow rice breeders to impose positive selection on essential elements and negative selection on potentially toxic elements in grain using markerassisted selection (MAS) [15]. This approach is particularly useful for traits that lack field-based and accurate phenotyping techniques and those that are strongly affected by genotype and environment interactions (G × E). Major QTL can be transferred precisely to different genetic backgrounds via marker-assisted breeding approaches, leading to faster development of rice varieties[16].

Biparental populations, such as doubled haploids (DHs),recombinant inbred lines (RILs), backcross inbred lines (BILs),and introgression lines (ILs), have been found to be effective for detecting major-effect QTL [17-27]. Among biparental populations, DHs are fixed genetic materials that can be developed more rapidly than other mapping populations and can be readily evaluated in multiple years and locations with the least genetic background “noise”, making them valuable genetic resources for mapping QTL and genes[28].

Several reports have shown the utility of DH populations in identifying QTL for grain mineral elements.Twenty-three and nine QTL, respectively, were identified in two environments for Ca,Fe,K,Mg,Mn,P,and Zn[23];eight QTL for Ca,Fe,K,and Mg were identified [19]; and six for Fe, Mn, and Zn were identified [17] in different DH populations of rice. Using RILs,five QTL for Zn and Cd were identified [24], and in another study [22] 14 QTL were detected for grain Fe and Zn and the candidate genes OsYSL1 and OsMTP1 for Fe; OsNAS1 and OsNAS2 for Zn;and OsNAS3,OsNRAMP1,and APRT for both Fe and Zn were identified.Genome-wide association analysis for 17 mineral elements in a panel of 529 rice accessions detected 72 loci [29] and in a Multi-parent Advanced Generation Intercross-Plus (MAGIC Plus) population 14 loci contributed to Fe and Zn concentration [30]. These results have clearly shown that there are multiple loci distributed throughout the genome and that they exert minor to moderate effects on the accumulation of different mineral elements in rice grain.Even though there are a few major loci, they have not been used widely in MAS for the development of nutritious rice.

There is thus an urgent need to validate the reported loci and to identify loci with consistent major effects for effective application of MAS [16]. The major objectives of our study were to a) assess the diversity of agronomic traits and the concentration of key grain elements in two DH populations,b) identify DH lines combining superior agronomic performance with high grain Zn, c) identify QTL associated with agronomic traits and grain element concentrations, and d)identify QTL that could potentially be used in MAS for Zn biofortification.

2. Materials and methods

2.1. Plant materials

We used two DH populations (Table 1) derived from the crosses IR64 × IR69428 (Pop1) and BR29 × IR75862 (Pop2). The recipient parents, IR64 and BR29, are popular varieties, while IR69428 and IR75862 are high-Zn donor lines. Pop1 and Pop2 were composed of respectively 111 and 146 lines.

The two mapping populations were evaluated during the 2015 dry season (2015DS) and 2015 wet season (2015WS) on the Robert S.Zeigler Experimental Station at the International Rice Research Institute (IRRI), Los Ba?os, Laguna. Each mapping population was planted in a randomized complete block design with three replications per trial. Standard agronomic practices and plant protection measures were applied to ensure good crop growth.

Table 1-Descriptive statistics of agronomic and grain micronutrient traits in two doubled-haploid populations grown during DS and WS of 2015.

2.2. Phenotyping

Both populations were phenotyped for four agronomic traits and 13 grain elements.The agronomic traits DF,PH,NT,and YLD were measured following the standard evaluation system [31]. Grain concentrations were evaluated for As, B,Ca, Co, Cu, Fe, K, Mg, Mn, Mo, Na, P, and Zn. For each replication, 50 g of seeds were dehulled and polished (Indo Plast Polisher, Hyderabad, India). Milled rice samples weighing at least 3 g were subjected to X-ray fluorescence analysis using a Bruker S2 Ranger(Karlsruhe,Germany)for Fe and Zn.Measurements were performed twice per sample and expressed in mg kg-1.The mean reading per plot was used for further statistical analysis. Brown rice samples from all the replications and in two populations were analyzed for all the mineral elements using inductively coupled plasma mass spectrometry(ICP-MS)at Flinders University,Adelaide,Australia.

2.3. Genotyping

Leaf samples were collected from month-old seedlings from both populations.The samples were cut into small pieces and placed inside the Eppendorf tubes, freeze-dried in liquid nitrogen, and mechanically ground with steel ball bearings in a grinder. DNA was extracted using a modified cetyl trimethylammonium bromide (CTAB) protocol [32]. A quality check of the DNA samples was performed in 1% agarose gel.From each sample, ～50 ng of DNA was submitted for genotyping using a 6 K-SNP chip at the Genotyping Services Laboratory at IRRI.

2.4. Analysis of phenotypic data

All data from both DH populations were analyzed.Descriptive statistics using mean trait values were generated using STAR v.2.0.1. Analysis of variance (ANOVA) and broad-sense heritability(H2)estimation were performed with PBTools v1.4.

2.5. Linkage mapping and QTL analysis

For each population, SNP markers were added sequentially based on their physical positions on chromosomes and were screened based on ≥80% call rate, locus homozygosity, and polymorphism between the respective parents. Genetic distances between the markers were estimated using QTL IciMapping 4.1 [33], redundant markers were removed, and the input file for genetic map construction was generated.Using the Kosambi function linkage map was created. The input file for QTL analysis was also generated.The mean trait values for each line in each DH population were used for QTL analysis.Using the BIP function,additive QTL were identified by inclusive composite interval mapping (ICIM) based on a 1000-permutation test at a 95%confidence level,and pairwise epistatic interactions between all markers were tested at a threshold value of LOD ≥6.0.The proportion of PVE accounted for by each QTL and each pair of epistatic QTL as well as the corresponding additive effects was also estimated. The physical positions of markers (http://rice.plantbiology.msu.edu/) flanking the QTL identified in both populations were used to determine QTL position for an integrated physical map using MapChart v.2.3 [34]. Adjacent QTL for the same trait on the same chromosome were combined and considered a single QTL. The combined effects of QTL for grain Zn were estimated by comparison of mean grain Zn of different QTL classes by t-test.

3.Results

3.1. Phenotypic analysis

Wide variation was observed for agronomic traits and grain mineral elements in both populations(Table 1). Several traits showed approximately normal distributions (Figs.S1 and S2),suggesting that they are controlled by multiple genes. DF in both populations showed a bimodal distribution in DS. The highest mean values of the agronomic traits DF(95.7 days),PH(98.9 cm), NT (16.3) and YLD (5.7 t ha-1) and of the grain concentrations (mg kg-1) of elements including As (0.25), Ca(95.6), Co (0.04), Mg (1492.9), Cu (4.3), Mn (27.5), P (3780.3), and Zn(16.9)were recorded in Pop2.In contrast,the highest mean values of B (10.6), Fe (4.5), K (3,452.5), Mo (0.97), and Na (19.1)were recorded in Pop1. The highest mean values of DF, NT,YLD, As, B, Ca, Co, K, Mg, and Mo were observed in DS,whereas the highest mean values of PH,Cu,Fe,Mn,Na,P,and Zn were observed in WS.The ANOVAs in both populations in DS and WS showed significant effects of genotype for all traits except NT and Fe in Pop2 in DS. The CVs of the population ranged from 3.0% to 58.2% among agronomic traits and from 3.9% to 53.6% among grain element concentrations. Low(＜15%) CVs were observed for DF, K, Mg, and P, whereas moderate(15-30%) CVs were observed for PH, As, Ca, and Zn.The rest of the traits showed high(＞30%)CVs:NT,YLD,B,Co,Cu, Fe, Mn, Mo, and Na. High (＞60%) heritabilities were estimated for all traits but NT, YLD,and Fe(Table 1).

Of the 272 possible correlations in each population, 132(77 positive and 55 negative) were significant (P ＜0.05) in Pop1,whereas 104 correlations(63 positive and 41 negative)were significant (P ＜0.05) in Pop2 (Figs. 1 and 2). A higher number of significant correlations were observed during 2015DS in Pop1 and 2015WS in Pop2.In both populations and across seasons, YLD was positively correlated with As and negatively correlated with K, P, and Zn; P was positively correlated with K, Mg, and Zn; and Fe was positively correlated with Zn. YLD was negatively correlated with Mg and Mn; As was negatively correlated with Cu, Mn,and P; and Zn was positively correlated with Co, Cu, Fe,Mg,and P.

3.2. Genetic analysis and linkage mapping

A total of 4606 SNPs covering the 12 chromosomes were used to identify polymorphisms between the parents of Pop1 and Pop2 and to genotype the two populations. Binning of redundant (completely correlated) SNPs and removal of unlinked markers resulted in a final tally of 541 and 296 SNPs for Pop1 and Pop2,respectively(Figs.S3 and S4),which were used to generate the linkage maps of Pop1 and Pop2.In Pop1, the number of SNPs per chromosome ranged from 28 SNPs on chromosome 8 to 79 SNPs on chromosome 3,while in Pop2 the number of SNPs per chromosome ranged from 3 on chromosome 6 to 65 on chromosome 3.The total linkage map lengths in Pop1 and Pop2 were 1686.1 and 1082.6 cM,respectively, while the largest marker interval lengths were 13.89 and 13.61 cM,respectively.

3.3. QTL analysis

QTL analysis resulted in the identification of 15 QTL for agronomic traits and 50 QTL for grain element concentration in the two populations (Table 2, Fig. 3). In all, 47 QTL were identified in Pop1 and 18 in Pop2. More QTL (29) were identified during the 2015WS than QTL(25)identified during the 2015DS,but 11 were common to both seasons.Most(37)of the favorable QTL alleles were from the high-Zn donor parents, while fewer (28) favorable QTL alleles came from the recipient high-yielding parents.The physical map(Fig.3)shows the QTL and candidate genes. The 15 QTL for agronomic traits were distributed on all chromosomes except 3 and 10.Among them,nine QTL were identified in Pop1 and six in Pop2. There were seven QTL for DF located on chromosomes 2, 5, 6, 8, 11, and 12; five for PH on chromosomes 1, 5, 7, and 9 and three for YLD on chromosomes 4, 6,and 12. The PVE of these QTL ranged from 10.1% to 35.7%.The QTL that explained the largest proportions of the PV(with corresponding additive effects) for each agronomic trait were qDF5.2(26.6%; 5.9 days), qPH1.1(17.6%; 4.4 cm), and qYLD4.1(35.7%; 1.5 t ha-1). Two QTL qDF8.1and qPH1.1were consistently identified in both seasons.

QTL identified for grain elements were distributed on all chromosomes except 12. The highest number (8) of QTL was detected for Zn, followed by As, Cu, K, and Mn with 5-7 QTL each. The proportions of PV explained by these QTL ranged from 8.6% to 36.8%. The QTL that explained the largest proportion of PV (with corresponding additive effects) were qAs8.2(21.1%;0.015 mg kg-1),qB4.1(33.0%;1.741),qCa2.1(21.8%;6.268), qCo7.1(28.7; 0.005), qCu8.1(26.8%; 0.871), qK6.1(36.8%;268.359), qMg1.1(16.1%; 65.433), qMn3.1(30.9%; 2.269), qMo1.1(16.0%; 0.140), qNa3.1(20.2%; 3.282), qP11.2(21.3%; 360.823), and qZn11.1(27.7%;2.069).

Nine QTL for mineral elements were consistently identified during both the DS and WS: qAs4.1, qB4.1, qCo10.1, qCu2.1,qK6.1,qMn3.1,qMn4.1,qZn3.1,and qZn8.1.

Fig.1-Correlations among agronomic and biofortification traits in the IR64 × IR69428 DH population.DF,days to 50%flowering;PH,plant height(cm);NT,number of tillers;YLD,grain yield(kg ha-1);As,arsenic(mg kg-1);B,boron(mg kg-1);Ca,calcium(mg kg-1);Cu,copper(mg kg-1);Fe,iron(mg kg-1);K,potassium(mg kg-1);Mg,magnesium(mg kg-1);Mn,manganese(mg kg-1);Mo,molybdenum(mg kg-1);Co,cobalt(mg kg-1);Na,sodium(mg kg-1);P,phosphorus(mg kg-1);Zn,zinc(mg kg-1);DS,dry season;WS,wet season.*Significant at P ＜0.05,**significant at P ＜0.01,***significant at P ＜0.001.Red color shows negative correlation and blue color positive correlation.Figure on left is for the dry season and that on right is for the wet season for population 1.

3.4. Co-localization of QTL for agronomic traits and grain mineral elements

Fig.2-Correlations among agronomic and biofortification traits in the BR29 × IR75862 DH population.DF,days to 50%flowering;PH,plant height(cm);NT,number of tillers;YLD,grain yield(kg ha-1);As,arsenic(mg kg-1);B,boron(mg kg-1);Ca,calcium(mg kg-1);Cu,copper(mg kg-1);Fe,iron(mg kg-1);K,potassium(mg kg-1);Mg,magnesium(mg kg-1);Mn,manganese(mg kg-1);Mo,molybdenum(mg kg-1);Co,cobalt(mg kg-1);Na,sodium(mg kg-1);P,phosphorus(mg kg-1);Zn,zinc(mg kg-1);DS,dry season;WS,wet season.*Significant at P ＜0.05,**significant at P ＜0.01,***significant at P ＜0.001.Red color shows negative correlation and blue color positive correlation.Figure on left is for the dry season and that on right is for the wet season for population 2.

Table 2-QTL for agronomic and mineral element concentration traits in two DH populations.

Seven co-localizations among QTL for agronomic traits and grain element concentrations were observed (Table S1).Some significantly correlated traits showed co-localized QTL.For instance, the negatively correlated agronomic traits DF and YLD showed co-localized QTL qDF6.1and qYLD6.1on chromosome 6 identified in Pop2 during the WS. The positively correlated grain Co and Zn showed the colocalized QTL qCo3.1and qZn3.1, on chromosome 3 in Pop2 in the DS. The positively correlated grain Mg and P showed colocalized QTL qMg11.1and qP11.2on chromosome 11 identified in Pop1 in the DS.Most notably,the positively correlated grain As and B showed the co-localized QTL qAs4.1and qB4.1on chromosome 4 identified in Pop1 during both DS and WS.

(continue2d)

Fig.3-Physical map showing the QTL and candidate genes identified in the two DH populations.QTL for agronomic traits are labeled in green,QTL for grain micronutrient concentration in black,candidate genes that are within 0.1 Mb of QTL in red,and candidate genes ＞0.1 Mb from QTL in blue.

Co-localizing QTL Ca, Co, Cu Ca, Co, Cu Ca B, Co,Cu B, Co,Cu B B, Ca, Cu B, Ca, Cu Cu B, Ca, Co B, Ca, Co Co mon to B,Ca,Season DS DS DS WS DS DS DS WS DS DS DS DS DS DS DS WS DS DS DS DS DS DS DS DS WS es 1 and 2 were com Pop 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 2 2 2 osom Add × Add-3.037-3.232-1.193-1.308 25.997 54.19 54.204-1.915 44.707 92.512 92.348-24.109 24.206 48.994 49.014 0.004 25.743 50.94 50.98-198.51-0.895-3.505 6.904 6.508-20.358 olybdenum (mg kg-1); Co, cobalt (mg kg-1); Na, sodium mon to Co and Cu.Add2 1.47 2.662 0.825-0.625 27.737 0.868-1.108 0.299 48.064 0.302-2.977 13.042 24.207 1.019-10 24.699 1.188-0.957 53.953 0.195 0.42-6.971-5.585 20.83 g kg-1); Mo, m PVE)between loci on chrom to 16%0.685 Add1 0.065-1.208 0.012 0.586-28.667 0.868-1.108-0.468-46.877 0.302-2.977-12.036-25.807 1.019-1-0.003-26.349 1.188-0.957-40.801-5.409-0.181-6.962-20.543 29.1 ic traits and concentration of elements in the grain.PVE(%)19.5 24.2 26.8 15.8 15.9 16 24.2 15.2 15.1 16 34.5 15.9 16.1 16.1 39.9 15.7 16 15.9 25.9 13.3 33.2 32 31.8 28 ilar genetic effects(15%mon to B and Ca and another was com 6.7 LOD 6.2 6.8 8.8 11.7 9.6 9.8 6.8 9.7 7.8 7.9 6 12.7 10 11.8 6.8 13 10 11.7 6.4 15.8 6.1 10.3 8.9 7.8 6776685 RM id4007709 id10005716 241969 275952 1652037 2696289 2587600 275952 1652037 2696289 id9006377 275952 1652037 2713838 8940497 275952 1652037 2713838 6817862 id10005716 id12005547 5895767 9788558 9732113 g kg-1); Cu, copper (mg kg-1); K, potassium of odds ratio;Add,additive effect;DS,dry season;WS,wet season.It is worth noting that four digenic interactions id6013529 c10p18362887 LM 237468 id1006289 5883472 wd2000489 id3004023(m 2584928 id1006289 4489306 wd2000489 id3004023 id9006187 id1006289 wd2000489 2696289 8926819 id1006289 wd2000489 2696289 id6014488 c10p18362887 12384668 9788010 9720310 e 3 that accounted for 16% PVE were com osom 10 1 1 2 3 3 1 2 3 9 1 2 3 8 1 2 3 6 10 Chr 12 6 9 9 RM 6 4 3387049 4146180 Table 3-Epistatic interactions for agronom 7660024 232009 id1006289 1634680 id3004023 2583961 id1006289 1634680 id3004023 id9004843 id1006289 1634680 2696289 5769384 id1006289 1634680 2696289 6632373 c10p18362887 275952 id3010875 5901730 9708423 ber of tillers; B, boron (mg kg-1); Ca, calcium LM 3374885 7644067 4112998 229857 id1006175 1597372 2650330 id3001422 id1006175 1597372 2650330 9655207 id1006175 1597372 id3004023 5735426 id1006175 1597372 id3004023 6585321 10680806 id1006289 3229778 5895767 id9004843(m Chr 3 4 7 1 1 2 3 3 1 2 3 9 1 2 3 5 1 2 3 6 10 1 3 6 9 Trait DF NT g kg-1);PVE,phenotypic variance explained(%);Chr,chromosome;LOD,logarithm B Ca Co Cu K Mo Na DF, days to 50% flowering; NT, num(m explained the PV observed for at least two elements.For instance,two epistatic interactions with sim Co, and Cu. Epistatic interactions between loci on chrom

3.5. Epistatic interactions

Three epistatic interactions were detected for agronomic traits and 22 epistatic interactions were detected for grain element concentrations, distributed on all chromosomes except 8, 11, and 12 (Table 3). None of the epistatic interactions involve main-effect QTL and most (21 epistatic interactions)were detected during the DS.The proportions of PV explained by these epistatic interactions ranged from 13.3% to 39.9%. There were nine digenic interactions with large(＞25%)effects for DF,Ca,Co,K,and Na.For Ca,the interaction of loci on chromosome 9 accounted for 34.6%PVE(-24.109 mg kg-1)while for Co, the interaction of loci on chromosomes 5 and 8 accounted for 39.9%PVE(0.004 mg kg-1).For K,interaction of loci on chromosome 6 accounted for 26.0% PVE (-198.51 mg kg-1).Four interactions of loci for Na distributed on chromosomes 1,3,6 and 9 showed effects representing 28.1-33.2% PVE (-20.357,-3.505,6.508,and 6.904 mg kg-1).

3.6. Combined effect of QTL for grain Zn

The combined effect of QTL was analyzed by comparison of mean grain Zn for different QTL classes in both populations. Some of the QTL combinations with significant differences are presented in Table S2. Six QTL were identified for grain Zn in Pop1. Among the single-QTL lines, those with qZn9.1showed highest mean grain Zn of 18.1 and 19.1 mg kg-1in DS and WS respectively. Lines with two QTL showed no significant advantage over lines with qZn9.1,but lines with two QTL such as qZn2.1+ qZn1.9and qZn2.1+ qZn11.1showed comparable Zn. Most of the three-QTL combination lines showed significant increases in Zn content over two-QTL lines; three-QTL lines with Zn2.1+ qZn5.1+ qZn11.1and qZn2.1-+ qZn5.1+ qZn5.1showed mean grain Zn of 20.6 and 22.9 mg kg-1during DS and WS respectively. A four-QTL line(qZn2.1+ qZn5.1+ qZn5.1+ qZn11.1)showed the highest grain Zn of 28.2 and 24.3 mg kg-1in DS and WS respectively. But none of the five-QTL combinations showed higher grain Zn than four-QTL lines. These results show that QTL-pyramided lines had high Zn content. In Pop2, two QTL were identified for grain Zn and QTL-pyramided lines showed significantly greater Zn than single-QTL lines in WS. However, within each QTL class some combinations showed highly significantly increased grain Zn in both populations.

3.7. DH lines with high grain Zn and yield

Based on Zn and YLD values, five superior lines were identified among the two populations (Table 4). These lines showed Zn values ranging from 20 to 23 mg kg-1in both seasons, YLD ranged from 5-7 t ha-1in 2015DS, and 2.5--3.1 t ha-1in 2015WS.Two of the lines carried all Zn increasing QTL alleles (qZn2.1, qZn5.1, qZn5.2, qZn7.1, qZn9.1, qZn11.1), while one line carried four QTL (qZn2.1, qZn5.1, qZn5.2, qZn11.1)identified for grain Zn in Pop1. Both the lines from Pop2 carried two QTL each (qZn3.1, qZn8.1). Comparisons between the selected lines and IR64 showed that yields of all lines were comparable with those of the check variety IR64, whereas Zn levels of all five lines were significantly higher than that of IR64 in at least one season.

Table 4-Superior lines identified in the two DH populations.

4. Discussion

Selection of rice varieties with high concentrations of key nutrient elements in the grain using MAS is an effective strategy to address widespread dietary deficiency in human populations[30]. The integration of MAS, especially in Zn biofortification programs, requires identification of molecular markers closely associated with the trait.Identified major QTL can be transferred to different genetic backgrounds more precisely through markerassisted breeding [16,35] and genomics-assisted selection [36]approaches, leading to faster development of rice varieties.Linkage mapping is a powerful tool for identifying QTL and associated markers that can be coupled with candidate gene analysis approaches to identify target genes for further analysis and validation. Moreover, QTL analysis using various mapping populations tested in different seasons can help elucidate the influence of genetic background and environment on QTL expression and can serve as the basis for selecting QTL combinations for specific genotypes and environments in MAS.In this study, QTL analysis was performed using two DH populations evaluated in two seasons.

4.1. Phenotypic analysis

The two DH populations showed wide variation for all agronomic traits and grain mineral elements. A few lines showed acceptable and consistent yields (comparable with those of the high-yielding check variety) and ≥20 mg kg-1Zn across seasons (Table 2) while some lines with lower yields exceeded 28 mg kg-1, which is the target level for rice Zn biofortification[37].The identified lines represent useful gene combinations that could be used directly in breeding. Comparisons with previous studies using milled rice show even greater variation in grain concentrations of elements in rice.Studies showing higher values for Cu,Fe,Mo,and P and lower As and Co [18,21]and higher Cu and Fe and lower Ca and Mn have been reported.Comparable Fe values[25] and comparable Ca and Cu and lower K,Mg,Mn,and P[26]values have also been reported.In unmilled rice,higher Fe,lower K and P,and comparable Ca, Mg, and Mn [23] have been reported. Zn concentration values in our study were similar to those previously reported [18,25,26]; however, values can be as high as 35.3 mg kg-1[21].Comparable and higher mean values for Zn have been reported for unmilled rice samples[20,22,23,38].

Heritability values for all grain micronutrient concentrations except Fe were high (＞60%), suggesting that direct selection for these elements may be a practical approach for trait improvement.Another study[38]estimated moderate to high heritabilities for Ca, Cu, Fe, K, Mg, Mn, and P. High heritabilities for Cu,Mn,and Mo;and low heritabilities for As,Co, Fe, and P have also been reported [21]. In several reports[20,39,40], the heritability of grain Zn was high and ranged from 41%to 94.2%.

YLD was negatively correlated with K, Mg, Mn, P, and Zn,whereas Co, Cu, Fe, K, Mg, Mn, and P were positively correlated with Zn. Significant negative correlations between YLD and Zn have been found in several studies in rice[21,41-43], and may have reflected the dilution effect of YLD on Zn. Removing the effect of YLD on Zn as a corrective measure may be required to select high Zn rice lines with acceptable yield potential [44]. This dilution effect may also explain the negative correlations between YLD and concentrations of other elements in this study. However, a positive relationship between YLD and Zn and non-significant correlations between the two traits have also been reported[39,45-47], suggesting that it is possible to select high-Zn rice lines without any yield penalty. Positive correlations of Zn with Ca, Cu, Fe, K, Mg, Mn, and P have also been observed[17,22,26,28]. The simultaneous targeting of Zn and Fe for biofortification is often justified because they have been shown to have consistent positive correlations and to share a common genetic basis. The same QTL have been identified for Zn and Fe in several populations [17,24,25] and Zn and Fe concentration in both polished and brown rice is simultaneously increased with overexpression of ferritin and OsNAS[48-50].

The numbers of lines in Pop1 and Pop2 are within the range for mapping populations used in preliminary studies[51].The number of polymorphic SNPs was higher for IR64 and IR69428 than for BR29 and IR75862, indicating that Pop1 had more genetically diverse parents than Pop2. Moreover, the lower level of polymorphism resulted in fewer detected QTL in Pop2 than in Pop1. The polymorphism in biparental crosses depends greatly on the parent combinations used, so it necessary to develop biparental breeding populations using different sets of parents to capture variation present in natural populations and to identify causative loci that are of direct relevance for crop improvement [52]. A single QTL that accounts for a large proportion of phenotypic variance in a biparental population that also shows high variation is desirable for its ultimate application in MAS[53].

4.2.QTL identified for grain concentration of mineral elements and their comparison with QTL previously reported

We compared consistent and co-localized QTL with earlier reported QTL for the same or related traits using the physical positions of the respective linked markers. It is common to find QTL for grain element concentration clustered in a genomic region. This clustering may indicate pleiotropic effects of the same genes and/or linkage of genes involved in the same pathways [54,55]. Five of the seven co-localized regions were also reported in earlier studies and two were new regions.The co-localized region with qAs3.1and qK3.1was reported [24] for another heavy metal, Cd, near position 1.7 Mb on chromosome 3, and another chromosomal region at 35 Mb with qZn3.1and qCo3.1was in the same region as grain Zn QTL between RM7000-RM514 [23].Similarly the qDF5.2and qZn5.1co-localized region was also found near R3166-RG360 and RM1089 linked to grain Zn in earlier studies [18,20,29].There were two co-localized regions on chromosome 11, the qNa11.1and qZn11.1co-localized region overlapped with C794-RG118 identified for grain Zn by Lu et al. [18], and another region with qMg11.1and qP11.2was reported for P and Mn [24].The co-localized chromosomal regions identified on chromosomes 4(qAs4.1and qB4.1)and 6(qDF6.1and qYLD6.1)are novel.

Among the consistent QTL identified in both DS and WS,qDF8.1was also reported between markers AG8Aro and RZ617 in a population derived from Azucena and IR64 in three different environments[56].The qPH1.1identified in our study co-derived with a major gene responsible for plant height,sd1,this chromosomal region has been found in several earlier studies[56-59]to influence plant height.The qCu2.1region on chromosome 2 was also reported for Cu between G1314a and RM240 and near RM13556 [24]. qK6.1was identified for K between RM546 and RM587 [23], qMn3.1was identified for Mn and Zn between RM3248 and RM280[23]and near RM517[24],and qZn2.1was also identified near marker RM106[24].

Several QTL for grain concentrations of elements identified in this study including qCa2.1,qCo7.1,qK4.1,qK6.1,qMn1.1,qMn1.2,

qMn3.1, qMn3.2, and qP11.2, were previously reported[17,18,24,26,38]. More importantly, five of eight QTL for Zn,namely qZn3.1, qZn5.1, qZn5.2, qZn9.1, and qZn11.1were also previously identified in other studies [18,20,23,24]. These QTL are responsible for phenotypic variation observed in multiple rice populations and have shown stability in different genetic backgrounds and environments, qualities of QTL useful for MAS[2].

4.3. Epistasis

The 25 epistatic interactions observed for DF,NT,B,Ca,Co,Cu,K, Mo, and Na accounted for 13.3% to 39.9 PVE. Two studies[19,21] of grain mineral concentrations also focused on epistatic interactions and found digenic interactions for Al,As,Ca,Cd,Fe,K,Mg,Mn,Mo,and Zn accounting for 4.2-46.0%of PVE. Epistatic effects can be as large as main QTL effects and can occur between QTL with insignificant individual effects [60]. The digenic interactions detected in the present study did not involve main-effect QTL,which have also been reported[21]for grain concentrations of other elements.Four of the interactions in this study were found for at least two elements.However,correlations among elements with common epistatic interactions were not observed,possibly owing to the smaller genetic effects of the shared epistatic loci than the effect of the main QTL. These interactions indicate that mechanisms involved in the regulation of grain element concentration involve multiple genes that may be common to these elements [61]. Epistasis cannot be neglected in studying complex traits because it can account for hidden quantitative genetic variation in natural populations[61,62].

4.4. QTL useful for MAS and candidate gene analysis

QTL for grain mineral elements that explain ＞25%PVE and/or show consistency across seasons merit further examination by fine mapping and candidate gene analysis for ultimate use in MAS. These QTL are qAs4.1, qB4.1, qCo7.1, qCo10.1, qCu2.1,qCu8.1, qK6.1, qMn3.1, qMn4.1, qZn3.1, qZn8.1, and qZn11.1. Nine of these QTL were detected in both seasons.The presence of QTL showing some level of consistency in association could be expected from the high heritability of the traits that they influenced.

Marker-assisted pyramiding of QTL is an effective approach for improving complex traits [16,51,63]. We tested the pyramiding effect of QTL for grain Zn in both populations and seasons. The pyramiding effect was very clear in Pop1 with two-, three-, and four-QTL combinations showing increased grain Zn with increased number of QTL. But five-QTL lines showed no significant improvement over four-QTL lines.Similarly, in Pop2, two-QTL lines showed high grain Zn in WS. However, within each QTL class only certain QTL combinations showed significant improvement over others for grain Zn in both populations.Thus,not only the number of QTL pyramided but their specific combinations must be carefully considered during QTL pyramiding for the improvement of complex traits.The real effect of QTL pyramiding can be further assessed by MAS of QTL in clean genetic backgrounds of popular rice varieties[63,64].

5. Conclusions

Wide variation was observed for all traits in two DH populations. Fifteen QTL for agronomic traits and 50 QTL for grain concentrations of elements were identified,11 of which were detected in both seasons of the study.Co-localization of QTL for grain concentration was found for As and B,Mg and P,and Co and Zn. Several QTL co-localized with previously reported QTL, indicating their stability in different genetic backgrounds. For biofortification, especially with Zn and Fe,large-effect and consistent QTL are of particular importance and merit further examination.

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2019.03.002.

Conflict of interest

Authors declare that they don't have any competing interest

Acknowledgements

The authors thank HarvestPlus for funding development of high Zinc rice.Dr.Deepinder Grewal and Dr.Jessica Rey for the initial development of the DH populations.