Mapping of QTL for Grain Yield Components Based on a DH Population in Maize

The elite maize hybrid Zhengdan 958 (ZD958), which has high and stable yield and extensive adaptability, is widely grown in China. To elucidate the genetic basis of yield and its related traits in this elite hybrid, a set of doubled haploid (DH) lines derived from ZD958 were evaluated in four different environments at two locations over two years, and a total of 49 quantitative trait loci (QTL) and 24 pairs of epistatic interactions related to yield and yield components were detected. Furthermore, 21 QTL for six investigated phenotypic traits were detected across two different sites. Combining the results of these QTL in each environment and across both sites, three main QTL hotspots were found in chromosomal bins 2.02, 2.05–2.06, and 6.05 between the simple sequence repeat (SSR) markers umc1165-bnlg1017, umc1065-umc1637, and nc012-bnlg345, respectively. The existence of three QTL hotspots associated with various traits across multiple environments could be explained by pleiotropic QTL or multiple tightly linked QTL. These genetic regions could provide targets for genetic improvement, fine mapping, and marker-assisted selection in future studies.

In contrast to the F 2:3 or backcross populations used for QTL mapping, recombinant inbred line (RIL) and doubled haploid (DH) line populations consist of genetically stable families and can be used to obtain more accurate and effective phenotyping for QTL mapping. RIL populations are usually developed by continuous self-pollination for more than eight generations, which is a time-consuming and expensive process. In contrast, DH populations are produced in only two generations. Thus, DH populations are increasingly used for mapping experiments in various species [19][20][21][22][23][24] .
The maize hybrid Zhengdan 958 (ZD958), a commercial hybrid with high and stable yield, was widely grown on approximately 500 million hectares between 2001 and 2015 in China 25 and is still a competitive variety in the northern and central parts of China. Because its two parent lines, Zheng58 (Z58) and Chang7-2 (C7-2) have high general combining ability and represent the two main heterotic groups in China, Reid and Tangsipingtou, this hybrid and its parental lines have been intensively studied for heterosis 26 , cultivation conditions and the physiological basis of high yield 27,28 . However, the genetic basis of the high yield of the elite hybrid ZD958 and its components remain little known. In this study, a set of DH populations of ZD958 was developed and evaluated in four different environments at two locations across two years (2014 and 2015). The objectives of this study were to (1) elucidate the relationship between grain yield and its components, (2) identify QTL for grain yield-related traits across multiple environments, and (3) study G × E interactions. These findings may reveal the genetic basis of grain yield and its components in the hybrid ZD958 and provide molecular markers for developing new superior maize hybrids.

Results
Phenotypic performance in different environments. There were 161 DH families obtained from the hybrid ZD958 by in vivo haploid induction, haploid identification, and haploid genome doubling. The DH population, the hybrid ZD958 and its two parental lines were evaluated in four different environments at two locations over two years (Table 1). For the two parents, the EL, HKW, and GWP of the inbred line Z58 were higher than those of C7-2, and the other traits (ED, ERN, and KNR) were lower. For the hybrid ZD958, all six estimated traits Genetic map construction. A total of 119 polymorphic simple sequence repeat (SSR) primer pairs were identified from 897 markers on ten chromosomes in MaizeGDB (http://www.maizegdb.org/) and used to construct linkage maps for QTL detection. The linkage map covered all 10 maize chromosomes with a total genome size of 2315 cM, the average size of the marker intervals was 19 cM, and all the marker positions were consistent with the linkage map for maize B73 × Mo17 (IBM) (http://www.maizegdb.org). DH populations can be obtained by tissue culture and in vivo haploid induction; the former is usually affected by maternal genotype and is prone to segregation distortion, while the latter is more likely to be consistent with Mendelian inheritance. In the present study, the mapping population was a DH population developed by in vivo haploid induction rather than tissue culture, and none of the molecular markers used for linkage mapping showed significant segregation distortion.
QTL detection in single environments. In this study, a total of 49 QTL were detected (Table 4); 17, 10, 12, and 10 QTL were identified for the traits measured at Changge in 2014 (CG14), Qixian in 2014 (QX14), Changge in 2015 (CG15) and Qixian in 2015 (QX15), respectively. These QTL were distributed on 10 chromosomes, and most QTL were located on chromosomes 2, 5, and 6, which had 18, 8, and 9 QTL, respectively ( Fig. 1). Each QTL explained a percentage of phenotypic contribution from 5.7 to 17.8%, including 20 main effect QTL with more than 10% contribution and 4 QTL with contributions over 15%. Five and three QTL from these were detected across two and three environments. However, only one QTL was detected in all four environments   Fig. 1); it was on chromosome 2, and it explained 8.5, 11,5, 9.4 and 11.0% of the phenotypic variation in CG14, QX14, CG15, and QX15, respectively. These loci originated from the inbred parental line C7-2, which contained favourable alleles controlling ERN. www.nature.com/scientificreports www.nature.com/scientificreports/ QTL detected in different locations. The two locations, Changge (CG) and Qixian (QX), are located in the central and northern parts of Henan Province, respectively, and belong to two important maize planting zones in China. A total of 21 QTL for six phenotypic traits were detected across these two locations (Table 5), and 11 and www.nature.com/scientificreports www.nature.com/scientificreports/ 10 QTL were identified in CG and QX, respectively. There were ten QTL located on chromosome 2; three QTL on chromosomes 3, 5, and 6; and only one QTL on chromosomes 9 and 10 ( Supplementary Fig. S2). Each QTL explained a percentage of phenotypic variation from 6.0 to 18.8%, including nine main effect QTL with more than 10% contribution to variation. Five consistent QTL were simultaneously detected in both locations, of which 2, 1, 1, and 1 QTL were identified for EL, ERN, HKW, and GWP on chromosomes 2, 3, 5, and 6, respectively. The QTL qEL3-Z on chromosome 3 showed the largest contribution, with a value of 18.8% at the QX site. Notably, qERN2-Z and qGWP2a-Z were in the same marker interval, umc1065-umc1637. QTL for ED and KNR were also identified in this region at the CG and QX planting sites, respectively. These results suggest a close genetic correlation among ED, ERN, KNR, and GWP and could be due to pleiotropy of this QTL.
Combined QTL detection across four environments. There were 15 QTL associated with EL, ERN, KNR, and GWP across all four environments (Table 6), and they were distributed over chromosomes 2, 3, 4, 5, and 7. The additive effects of the QTL for GWP ranged from −4.06 to 4.45 g, and the QTL qGWP7-J showed a positive additive effect of 4.45. The contribution of the additive effect [H 2 (A)] ranged from 2.28 to 11.27% in the measured traits, and there were two QTL, qERN2-J and qERN3b-J, with higher phenotypic variation values of 11.27 and 9.47%, respectively, for ERN. The interaction of additive effect by environment [H 2 (AE)] varied from 0.07 to 0.58% of the phenotypic variation. Eight QTL (2 for EL, 2 for ERN, 1 for KNR and 3 for GWP) were detected by combined analysis across the four environments and were also identified in the analyses of the separate environments.

Analysis of digenic epistatic interactions.
Twenty-four pairs of digenic epistatic interactions were identified for six traits with additive × additive (AA) interaction and AA × environment (AAE) interaction effects. The epistatic interactions involved 35 loci distributed on all chromosomes except for chromosome 8 and 10 ( Table 7). Only one pair with a significant AAE interaction (P = 0.05) for GWP was identified in CG14 and QX15. Its AAE interaction effect of 0.79% was significantly higher than those of the other traits. This suggests that this epistatic interaction is influenced by the environment. For EL, four significant interactions were detected and found to involve seven loci on chromosomes 2, 5, 6 and 9. The contribution of the AA interactions varied from 1.05 to 3.97%, and the contribution of the AAE interactions varied from 0.19 to 0.63%. Two pairs of loci with significant digenic interactions for ED were detected, and these included 4 loci distributed on chromosomes 1, 2, and 4, with AA interaction effects of 4.26 and 5.87% and AAE interaction effects of 0.14 and 0.27%. A total of six epistatic interactions were identified for ERN, including 11 loci located on chromosomes 2, 3, 4, 6, 7 and 9. Three pairs of epistatic interactions were identified for KNR. Six epistatic interactions were identified for HKW, and their AA interactions explained 3.5, 3.1, 5.3, 4.1, 8.2 and 4.0% of the phenotypic variance. There were three epistatic interactions for GWP, but only one pair of interactions, between bnlg2291-umc1847 and bnlg389-bnlg386, which produced a larger effect than the other interactions.

Discussion
Previously, an F 2:3 population, RILs, near isogenic lines (NILs), and DH lines have been used to dissect the genetic basis of quantitative traits in maize [3][4][5][19][20][21][22] . Among the segregating types of populations used for QTL mapping, the F 2:3 population represents an early generation (transient group), which often affects the accuracy of QTL mapping. Although a RIL is regarded as a permanent population, it generally requires more than 8 generations of continuous selfing, which is a time-consuming and expensive process. By contrast, DH populations can be developed with only two generations in one year; thus, this option is very rapid and inexpensive, making it an ideal population for genetic analysis and QTL mapping [29][30][31][32] . In this study, a DH population derived from the elite hybrid ZD958 was used to dissect the genetic basis of grain yield and its components in maize, and a total of 49 QTL and 24 epistatic interactions were detected. These findings could have the potential to help improve grain yield in maize breeding.
Grain yield and its components in maize are complex quantitative traits that are controlled by multiple genes, epistasis and G × E interactions. Over the last 20 years, numerous QTL related to grain yield and its components have been identified by using different segregating populations and association mapping populations. However, many uncertainties and inconsistencies are present in these loci, and these issues might be attributed to several factors, including genetic background (parents, populations and generations), marker types, mapping methods and environments. In this study, a total of 49 QTL were detected in a single environment, whereas 21 QTL for the six investigated traits were detected across two different locations. Although some QTL had lower contributions to the corresponding traits, they did have significant effects on the target traits through interaction with other QTL (epistatic interaction). Our results indicate that the genetic basis of grain yield and its components is controlled by major QTL effects, AA and AAE interaction effects simultaneously. In particular, major QTL with high heritabilities that were detected in different environments simultaneously were considered to have high stability and reliability. Additionally, combined QTL analysis revealed three major QTL hotspots, including pQTL2-1, pQTL2-2 and pQTL6-1, which were located on maize chromosomes 2 and 6 (Supplementary Table 1). At the same time, those chromosome regions were repeatedly identified for several traits of grain yield and its components; for example, the QTL qEL6b, which is responsible for EL, and qKNR6a, which is responsible for KNR, were located in the same interval (nc012-bnlg345) in multiple environments, indicating that both EL and KNR could be increased simultaneously. Meanwhile, three QTL hotspots in chromosomal bins 2.02 (umc1165-bnlg1017), 2.05-2.06 (umc1065-umc1637), and 6.05 (nc012-bnlg345) were detected both in a single environment and across two different locations. Therefore, these three QTL hotspots related to grain yield may be useful to increase grain yield in maize breeding.
In previous studies, a QTL hotspot responsible for grain-yield-related traits was detected by using a RIL population derived from two inbred lines, Ye478 and Qi319 9,11 ; this hotspot was mapped to the bin 2.02 chromosomal region, which contains a QTL detected for ED and GWP in the present study. This genomic region contains some annotated candidate genes, including ZmLG1, ZmMHA2 and ZmAST91, according to MaizeGDB (http://www. maizegdb.org). Among them, the gene ZmLG1 controls the angle of maize leaves and changes the plant architecture, thereby increasing photosynthetic efficiency and crop yield 33 . ZmMHA2 was identified as a functional Fe transporter that promotes Fe uptake and plays an essential role in plant growth and development 34 , while ZmAST91 could significantly improve crop stress resistance under abiotic stresses 35 . In the bin 2.05-2.06 genomic region, the candidate gene ZmWri1a controls the fatty acid content of the mature maize grain and certain amino acids, which can lead to an increase in the weight of the kernel 36 . Another gene, ZmCDPK24, which was located in the bin 6.05 genomic region, encodes a calcium-dependent protein kinase and plays a significant role in the  www.nature.com/scientificreports www.nature.com/scientificreports/ regulation of plant growth and development and in responses to various stresses 37 . These colocalized QTL could indicate a single pleiotropic gene, which might act as a regulator to control several traits.
Increasing grain yield is one of the most important targets in maize breeding. Previous studies and production practices have demonstrated that increasing planting density is an effective measure to improve maize grain yield. Planting density is increasing gradually in China and requires new types of hybrids with shorter EL, increased KNR and higher HKW. Of the three major components of grain yield, ERN has the highest heritability and has reliably been used as an important selection target to improve grain yield. In this study, some repetitive or co-located loci for ERN were consistently detected across multiple environments, such as qERN2a and qERN2-Z, which were detected in the QTL cluster in bin 2.05-2.06 derived from the elite inbred line C7-2, and a significant QTL for ERN in bin 2.05-2.06 was consistently identified by a combination of meta-QTL analysis and regional association mapping 16 . The elite hybrid ZD958 is based on the most successful heterotic pattern of Reid × Tangsipingtou, which has been widely used in China, and many new varieties have been developed from this hybrid. For example, the famous hybrid Xundan20 (Xun9058 × Xun926) has 2-4 more kernel rows than ZD958. Its male inbred line, Xun926, was modified from the inbred line C7-2. Another commercial hybrid in China, Zhongdan909 (Z58 × HD568) has a parental inbred line, HD568, that was also derived from C7-2, which has 2 kernel rows more than ZD958. These data further confirm that the utilization of inbred line C7-2 to increase ERN is feasible. Therefore, the linked markers of the QTL qERN2a and qERN2-Z could be used in marker-assisted selection (MAS) for ERN improvement in maize breeding. Additionally, DH technology has become one of the three core technologies of modern breeding programmes, along with transgenic and MAS breeding technologies 38 . Thus, maize ERN improvement by combining MAS and DH technologies should be highly efficient.

Materials and methods
Plant materials. In this study, a population comprised of 161 DH lines was derived from the hybrid ZD958, which is a leading elite maize hybrid developed by Henan Academy of Agricultural Sciences. The DH lines were developed following the procedure described by Prigge et al. 39 . Briefly, Zheng58 (Z58) was crossed as the female with Chang7-2 (C7-2) as the male to produce the F 1 generation (ZD958). The hybrid ZD958 plants were used as the female parent in the field, and the haploid inducer line CAU-5, developed by China Agricultural University 40 , was used as the male parent. Crosses were made by hand pollination, and putative haploids were identified by a  www.nature.com/scientificreports www.nature.com/scientificreports/ colour marker controlled by the R-nj gene 41,42 . In the subsequent planting season, haploid seedlings were treated with colchicine and dimethyl sulfoxide (DMSO) to promote haploid genome doubling. After treatment, the haploids were transplanted to the field and selfed to produce DH lines using the methods described by Chen et al. 38 . A total of 161 DH lines were obtained for use as experimental materials after reproduction.
Field experiments. The DH population, ZD958, and its two parents were planted in Qixian (QX14, northern Henan, 35°60′ N lat., 114°20′ E long.) and Changge (CG14, central Henan, 34°1′ N lat., 113°29′ E long.) in the summer of 2014 and then in Qixian (QX15) and Changge (CG15) in the summer of 2015. In each environment, the trial was conducted as a randomized complete block design with two replications. Each experimental plot consisted of a single row with a length of 4 m, a row-to-row distance of 0.66 m, and a plant-to-plant distance within rows of 0.20 m. Two seeds were sown per hill, and the plots were thinned to one seedling per hill at the 5-leave stage. Standard cultivation management practices were used in each environment.
Each plot was harvested by hand at maturity, omitting the two plants at the ends of the plots to avoid border effects. Ten ears from each plot were randomly chosen after natural air-drying to evaluate grain yield and its components, including EL, ED, ERN, KNR, and HKW. The GWP was adjusted to 13% moisture.
Phenotypic data analysis. The means, standard deviations, correlation coefficients, and kurtosis and skewness of trait distributions for each trait were calculated in SPSS 20.0 software (http://www.spss.com). Variance components were computed using PROC MIXED in SAS 43 with the following model: where Y ijk is the performance of the i th genotype at the j th environment (location-year combination) in the k th replication; μ is the overall population mean; G i is the effect of the i th genotype; E j is the effect of the j th environment; GE ij is the effect of G × E interactions; R k is the effect of the k th replication; and ε ijk is the error. In the model, G i , GE ij and error effects were considered random effects, and R k was considered fixed. The broad-sense heritability (H 2 ) of each trait was estimated as described by Knapp et al. 44 . The heritability (H 2 ) was calculated as: In total, 49 QTL were detected for six traits related to grain yield across four different environments at two locations over 2 years using a DH population, and 8, 6, 12, 8, 8 and 7 QTL were detected for EL, ED, ERN, KNR, HKW and GWP, respectively. The phenotypic contribution percentage of each QTL ranged from 5.71 to 17.84%. One QTL (qERN2a) was consistently detected in four environments, and its contribution varied from 8.5% to 11.5%. The QTL qEL4, qEL6b, qED2a, qERN2a, qKNR2a, qKNR6a, qHKW5a, and qGWP2a were also detected in two or three environments simultaneously. In addition, 15 significant QTL in the combined analysis and 21 QTL across both planting locations were related to EL, KNR, ERN, and GWP; no QTL were identified for ED or HKW. There were 24 pairs of epistatic interactions for the six measured traits. Importantly, three obvious QTL hotspots associated with yield components were found in maize chromosomal bins 2.02, 2.05-2.06, and 6.05. This study will not only contribute to a theoretical basis for predicting potentially superior yield traits in maize but also support MAS in maize breeding programmes.