Genome-Wide Linkage Mapping of Quantitative Trait Loci for Late-Season Physiological and Agronomic Traits in Spring Wheat under Irrigated Conditions

Many late-season physiological traits affect grain yield in wheat, either directly or indirectly. However, information on the genetic control of yield-related traits is still limited. In this study, we aimed to identify quantitative trait loci (QTL) for canopy temperature and chlorophyll content index during anthesis (CTa and CCIa, respectively), the mid grain-filling stage (CTg1 and CCIg1, respectively), and the late grain-filling stage (CTg2 and CCIg2, respectively) as well as for plant height (PH), thousand kernels weight (TKW), and grain yield (GY) using genome-wide linkage mapping. To this end, a double haploid population derived from a cross between two high yielding wheat cultivars, UI Platinum and SY Capstone, was phenotyped in four irrigated environments and genotyped using the wheat 90K iSelect platform and simple sequence repeats. The genotypic data were used to construct a high-density genetic map of 43 linkage groups (LGs) with a total length of 3594.0 cm and a marker density of 0.37 cm. A total of 116 QTL for all nine traits was detected on 33 LGs, spreading to all wheat chromosomes, except for Chr. 7D. Of these, six QTL (CTa.ui-4B.1, Q.CTg1.ui-5B-2.1, Q.CTg2.ui-6B.1, Q.PH.ui-6A-2.1, Q.TKW.ui-2D-1, and Q.GY.ui-6B) were consistently detected in more than three irrigated environments, called as stable QTL. Additionally, we identified 26 QTL clusters for more than two traits, of which the top four were located on Chromosomes 4A-1, 1B-1, 5B-2, and 2D-1. Overall, the stable QTL significantly related with grain yield, QTL clusters, and linked molecular markers identified in this study, may be useful in marker-assisted selection in early generation and early growth stage for grain yield improvement.


Introduction
Wheat (Triticum aestivum L.) is one of the most important food crops worldwide [1] since it is a valuable source of carbohydrates and protein suitable for human consumption [2]. However, wheat production is negatively affected by the gradual reduction of arable land area due to climate change as well as biotic and abiotic stresses [3][4][5][6]. Therefore, further improvement in grain yield (GY) is necessary to meet the current and future demand.
Canopy temperature (CT) or CT depression and chlorophyll content index (CCI) or SPAD value of leaf chlorophyll content are important physiological traits significantly associated with GY in wheat [7][8][9][10][11]. CT is considered an indicator of the transpiration level, whereas the chlorophyll content of flag leaf is and irrigation was applied from May to July with a total amount of 362.7 mm, 373.4 mm, 384.0 mm and 453.4 mm, respectively, making total moisture 580.1 mm, 590.8 mm, 601.4 mm and 702.4 mm, using aluminum sprinkler pipes. In AB15E1, and AB15E2, 120 entries were arranged in an augmented design. 120 replicated DH lines were assigned at random to four blocks, while two parents as checks were replicated, being randomly assigned to plots within each of the four blocks, therefore, each sub-block comprised 28 un-replicated DH lines and the two cultivar checks. In AB15E3 and AB16E4, the DH and parental lines were arranged in a randomized complete block design with two replications. Each plot included seven rows (3.0 m in length, 1.5 m in width, 0.25 m in row spacing). The sowing density was 0.48 million seeds per ha at each trial. Fertilization and weeding were applied, when necessary, to achieve the optimal cultivation conditions. A wheat border was planted to minimize the edge effect in each trial.

Phenotypic Evaluation
CT, CCI, PH, TKW, and GY were recorded in all four environments in two seasons (AB15E1, AB15E2, AB15E3, and AB16E4). CT ( • C) was measured using an infrared thermometer (IRtec MicroRay HVAC; Langhorne, PA) from noon to 2:00 PM on clear and non-windy days. The CCI of flag leaves was measured from five randomly selected fertile plants using a portable chlorophyll content meter (CCM-200; Opti-Sciences, Hudson, NH, USA). CT and CCI were measured during anthesis (Feekes 10.5.2; CTa and CCIa, respectively), the mid grain-filling stage (Feekes 11.1; CTg1 and CCIg1, respectively), and the late grain-filling stages (Feekes 11.2; CTg2 and CCIg2, respectively) using the Feekes growth scale [39]. PH (cm) was measured from the soil surface to the tip of the spike (awns excluded) at the maturity stage (Feekes 11.3-11.4) [39]. Plots were harvested by a Wintersteiger Classic small plot combine (Wintersteiger; Salt Lake City, UT, USA) equipped with a Harvest Master weighing system (Juniper Systems; Logan, UT, USA). After harvest, GY was determined as grain weight extrapolated at the 12% moisture content and expressed in t ha −1 . TKW (g) was recorded using the single-kernel characteristics system (SKCS 4100; Perten Instruments, Springfield, IL, USA).

Phenotypic Data Analysis
Phenotypic data analysis, including mean, genotype variance (δ 2 g ), error variance (δ 2 e ), correlation analysis between parameters, and broad-sense heritability (H 2 B ), was conducted using JMP 8.0 (SAS Institute, Cary, NC, USA). The adjusted means of each trait measured in AB15E1 and AB15E2 were estimated for block differences which were measured by the cultivar checks [40]. The analysis of variance was performed with PROC GLM, in which genotypes were treated as fixed effects and environments and the interaction of genotypes and environments and blocks nested in environments were all treated as random effects [40]. The inverse of the variance of the individual environments were treated as weights. H 2 B was estimated in all environments as follows: where δ 2 g is the genetic variance, σ 2 ge is the genetic × environment interaction variance, δ 2 ε is the error variance, r is the replicates per environment, and e is the number of environments.

SNP Genotyping and Molecular Marker Analysis
Total genomic DNA was extracted from young leaves of the DH and parental lines using the method as described by [41]. DNA concentration was estimated using a Nanodrop ND-1000 Spectrophotometer (Nanodrop; Wilmington, DE, USA) and then, adjusted to 80 ng µL −1 for SNP assays and molecular marker analysis.
The DH and parental lines were genotyped at the USDA-ARS Small Grains Genotyping Laboratory, Fargo, ND, USA, using the Illumina 90 K iSelect SNP array [31]. Genotype calling and SNP clustering were conducted using GenomeStudio 2011 with the polyploid clustering V1-0 (Illumina, San Diego, CA, USA). SNP names were designated as "IWB" followed by an index number.
A total of 300 SSRs was selected for genotyping the parental lines as described by [42]. Molecular markers for dwarfing genes (Rht-B1b and Rht-D1b) were used for confirming the association with PH [43,44].

Linkage Map Construction and QTL Analysis
SNP markers were filtered for monomorphism, high frequency of missing values (≥10%), or segregation distortion (≥0.35). The genetic linkage map was constructed using JMP Genomics 8.0 (SAS Institute) as described by [45] and [46]. The initial number of linkage groups (LGs) was identified using interactive hierarchical clustering and K-means clustering (automated radius K-means) by reducing the number of markers in the recombination and LG function. Markers were ordered on each LG using the Kosambi mapping function and the accelerated map order optimization algorithm in the linkage map order function.
LGs were split when the genetic distance between adjacent markers was higher than 35 cm by default.
QTL analysis was performed for each trait within and across four environments by composite interval mapping (CIM) using JMP Genomics 8.0 [46]. CIM was used to find QTL with an expectation maximization (EM) algorithm threshold of 2.5. Genetic distance between markers was calculated in cm. The contribution of SY and UI toward higher trait values was indicated by positive and negative signs of the estimates for additive QTL effects, respectively. The proportion of phenotypic variation (R 2 ) for each QTL was determined by the square of the partial correlation coefficient.
The final map was constructed using the linkage map viewer function of JMP Genomics, and LGs were assigned to chromosomes based on the 90 K consensus map as described by [31]. Each linkage map was named according to the wheat chromosome followed by a number. A QTL was characterized as stable when it was identified in at least three or four environments.

Phenotypic Evaluation
The analysis of variance showed that the effect of genotype was significant for all traits across environments, except for CTa and CTg1(across 15ABE3 and 16ABE4) (p < 0.05), whereas the effect of environment and that of genotype by environment were significant for CTa, CTg2, CCIa, CCIg2 and GY across 15ABE1 and 15ABE2 as well as CCIg2, TKW and GY across 15ABE3 and 16ABE4 (p < 0.05) ( Table 2). The phenotypic performance of DH and parental lines indicated that UI and SY were significantly different for CCI and TKW in all the environments (Table S1); UI had markedly higher TKW, whereas SY had higher CCIa, CCIg1, and CCIg2. The mean for most traits of DH lines was near to the mid-parental value. The range of variation for each trait of DH lines showed transgressive segregation in both directions across the environments, demonstrating that positive alleles were present from both parental lines. All traits showed continuous frequency distribution within and across the environments, indicating that they were under polygenic control ( Figure S1). The H B 2 was the highest for TKW (0.81 across 15ABE1 and 15ABE2, 0.72 across 15ABE3 and 16ABE4) and PH (0.77 across 15ABE1 and 15ABE2, 0.75 across 15ABE3 and 16ABE4), indicating that both traits were stable and mainly affected by the genotype, whereas the H B 2 was the lowest for CTa, CTg1, and CTg2, suggesting that they were greatly affected by the environment.

Marker Analysis and Construction of LGs
Of 81,587 SNPs on the 90 K iSelect SNP array and 300 selected SSRs, only 9687 SNP marker and 44 SSRs were polymorphic between the parental lines and used for the construction of the linkage map. Both parental lines carried Ppd-D1a and Rht-D1b, but not Rht-B1b.
We identified 43 LGs that represented all the 21 chromosomes of wheat (Table S2). Chr. 1A, 5A, 4B, 6B, and 7B were represented by one LG each; Chr. 2A, 4A, 6A, 1B, 3B, 5B, 1D, 2D, 3D, 4D, and 7D by two LGs each; Chr. 3A, 7A, 5D, and 6D by three LGs each; and Chr. 2B by four LGs (Table S2). The linkage map had a total length of 3633.19 cm, whereas each LG had an average length of 173.01 cm, ranging from 54.61 cm (Chr. 1D) to 261.51 cm (Chr. 5A). The average number of markers on each chromosome was 463.5, ranging from 15 on Chr. 3D to 1125 on Chr. 6A. Although markers on Chr. 1D, 3D, 4D, and 7D were low in number, they had a good distribution. The average marker density was 0.37 cm, ranging from 0.19 (Chr. 6A) to 5.56 (Chr. 3D). The map of A genome included 4320 markers (44.4%) and had a total length of 1420.51 cm with an average marker density of 0.33 cm; the map of B genome included 4367 markers (44.9%) and had a total length of 1571.15 cm with an average marker density of 0.36 cm; and the map of D genome included 1046 markers (10.7%) and had a total length of 641.53 cm with an average marker density of 0.61cm. The D genome had the lowest marker coverage, suggesting that more markers were polymorphic in the A and B genomes. Of 9687 SNPs, only 357 were newly mapped (Table S3).

Discussion
Improving GY is an essential objective in wheat breeding programs worldwide. Physiological traits, as well as PH and TKW, are critical elements of GY. The wide use of molecular markers has made possible the study of quantitative traits [9,11,23] and detect QTL that allow the development of high-yield cultivars via MAS. Previous studies identified numerous QTL associated with yield-related traits in various populations [1,9,10,[19][20][21]27]. However, the genetic architecture of these traits needs further research. In this study, a DH population was phenotyped for CT (CTa, CTg1, and CTg2), CCI (CCIa, CCIg1, and CCIg2), PH, TKW, and GY under irrigated conditions and genotyped with SNP markers and SSRs to better understand the genetic control of the studied traits.

Phenotypic Evaluation
In the present study, PH and TKW showed high heritability; CCIa, CCIg1, and CCIg2 moderate heritability; whereas CTa, CTg1, and CTg2 low heritability, results that were in agreement with those reported in previous studies [10,11,47]. Environmental diversity affects CT heritability, which ranges from low under heat or water-stressed conditions [9] to high under irrigated conditions [28]. GY is a complex trait highly affected by the environment and thus, has low heritability [48]. However, in the present study, the heritability of GY was higher than that reported previously [49]. The higher heritability could be attributed to the lower error variance. Overall, our results suggested that PH and TKW could be used for early generation breeding because of their high heritability, whereas CT and CCI could be also used due to their strong correlation with GY, similarly as reported by Zhang (2013) [49].

Linkage Map Construction
Molecular markers have been widely applied for genetic map construction, QTL detection, gene cloning, and MAS [15]. SNP markers are the most abundant polymorphisms in plant genomes [32], and thus, allow the high-throughput genotyping compared with previous marker platforms [30,31]. In the present study, we used the 90 K iSelect SNP array combined with SSRs to genotype a spring wheat DH population derived from a cross between UI and SY. Using 9687 polymorphic SNPs, 9330 previously mapped [31] and 357 newly mapped, we constructed a genetic map of 43 LGs that represented all the 21 chromosomes of bread wheat (Table S2) and had a total length of 3633.19 cm and an average marker density of 0.37 cm (Table S2). Our linkage map had a better marker distribution and a higher density compared with those reported in previous studies [11,23,[50][51][52][53][54][55]. However, some gaps were observed in Chr. 1A and 1B-1. The A and B genomes were mapped by 4320 and 4367 markers, respectively, whereas the D genome by 1046 markers (Table S2), suggesting that the former genomes included more polymorphisms. These results were in agreement with those reported in previous studies [11,31,56].
The average number of markers mapped per chromosome was 463.5, ranging from 15 on Chr. 3D to 1125 on Chr. 6A (Table S2). Although 89.4% of mapped SNPs displayed redundancy, only 10.6% was used to construct the linkage map, similarly as in previous studies, since many SNPs were co-located at the same loci [11,57].
PH is considered as a complex trait, consisted of internode and spike length [9,58]. Rht-B1b and Rht-D1b (previously known as Rht1 and Rht2) on Chr. 4B and 4D, respectively [43], that result in semi-dwarfism are widely used in wheat breeding. In the present study, both UI and SY carried Rht-D1b, but not Rht-B1b, and thus, no QTL for PH were identified on Chr. 4B and 4D. However, we detected the stable QTL Q.PH.ui-6A-2.1 on Chr. 6A-2 that has not been reported in any previous studies as well as two QTL on Chr. 7A-2 and 5B-2 that were consistently detected in two environments, explaining 9.9%-18.3% of the phenotypic variation (Table 7). Additionally, the minor QTL Q.PH.ui-5A on Chr. 5A at 41.26 cm was likely the same as that previously reported by Lopes et al. (2013) [9] and Gao et al. (2015) [11]. Additionally, two minor QTL identified for PH and GY on Chr. 1A and 1B-1 might indicate that other genes for PH probably have minor, but significant effects on GY.
Previous studies reported QTL clusters for various traits [10,11,65]. Here, we detected 26 QTL clusters for more than two traits (Table 8; Figure 1) on Chr. 1A, 2A-1(2), 1B-1, 2B-2, 2B-4, 5B-2, 1D-2, and 3D-1. Pleiotropic QTL for GY and PH were detected on Chr. 1A and 1B-1, and analysis of their allelic effects showed that a decrease in GY due to the presence of the SY allele was accompanied by an increase in PH (data not shown). Additionally, pleiotropic QTL for GY and KWT were found on Chr. 2A-1 and 2B-4 (2), and analysis of their allelic effects showed that the presence of the UI allele on Chr. 2A-1 increased TKW and GY (data not shown). Therefore, pleiotropic QTL confirmed the positive correlation between TKW and GY ( Table 3). Pleiotropic QTL for GY and CTa and also for CTa and CTg2 were detected on Chr. 2B-2 and 4B (2), respectively. Analysis of their allelic effects showed that the presence of the UI allele reduced CTa and CTg2, but increased GY (data not shown). Thus, pleiotropic QTL on Chr. 2B confirmed the negative correlation between GY and CTa, whereas those on Chr. 4B the positive correlation between CTa and CTg2 (Table 3).

Practical Applications in Wheat Breeding
Wheat GY is a complex, quantitative trait, highly affected by the environment; thus, early generation selection for high yield is usually not effective. In the present study, we suggested that CT, CCIa, PH, and TKW could be used for increasing GY using stable QTL, similarly as reported in previous studies [9,11]. GY was significantly positively correlated CCIa, PH, and TKW, negatively correlated with CTa, CTg1, and CTg2; therefore, these traits could be used for early generation and early growth stage selections. Additionally, some stable QTL, such as Q.CTa.ui-4B.1, Q.CTg1.ui-5B-2.1, Q.CTg2.ui-6B.1, Q.PH.ui-6A-2.1, Q.TKW.ui-2D-1, and Q.GY.ui-6B, could be applied in MAS for yield improvement after validation.

Conclusions
Genome-wide linkage mapping was used to identify QTLs associated with CTa, CTg1, CTg2, CCIa, CCIa, CCIg1, CCIg2, PH, TKW, and GY using high-density genetic linkage map in a spring wheat DH population derived from a cross between 'UI Platinum' and 'SY Capstone'. The DH population was genotyped using the wheat 90K iSelect platform and SSRs. The linkage map was constructed using a total of 9687 SNP marker and 44 SSRs, with whole linkage map of 3594.0 cm and marker density of 0.37 cm between adjacent markers. A total of 116 QTLs were detected for the nine traits on 33 linkage groups, representing the whole 21 wheat chromosomes, except for Chr. 7D. Among these QTLs, 71 QTL were for CT and CCI, and 45 QTL were for PH, TKW, and GY. Six QTLs were consistently detected more than three irrigated environments, called as stable QTL. 26 QTL clusters were identified for more than two traits, and four QTL-rich chromosome regions on these traits were found on chromosomes 4A-1, 1B-1, 5B-2, and 2D-1. These QTLs, QTL clusters and linked molecular markers may be very useful in assisted selection for improving grain yield in spring wheat breeding.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4395/8/5/60/s1, Figure S1: The frequency distribution of canopy temperature (CT) (A), chlorophyll content index (CCI) (B), plant height (PH), thousand kernels weight (TKW) and grain yield (GY) (C) in the DH population from UI Platinum/SY Capstone in the four environments, Table S1: Phenotypic variations for physiological traits, plant height (PH), thousand kernel weight (TKW) and grain yield (GY) in UI Platinum and SY Capstone and their DH population across four trial, Table S2: Summary of the genetic linkage maps constructed with SNP, SSR and STS markers using DH population derived from UI Platinum / SY Capstone, Table S3: The names and positions of 357 newly mapped SNP markers, Table S4: Summary of QTLs detected for physiological traits, plant height (PH), thousand kernels weight (TKW) and grain yield (GY) in the UI Platinum/SY Capstone DH population.
Author Contributions: Y.L. analyzed the data and developed the manuscript. R.W. and Y.-g.H. contributed to the manuscript. J.C. was the supervisor of Y.L., who designed and oversaw overall activities of the experiments and manuscript development.