Physical Mapping of QTL in Four Spring Wheat Populations under Conventional and Organic Management Systems. I. Earliness

In previous studies, we reported quantitative trait loci (QTL) associated with the heading, flowering, and maturity time in four hard red spring wheat recombinant inbred line (RIL) populations but the results are scattered in population-specific genetic maps, which is challenging to exploit efficiently in breeding. Here, we mapped and characterized QTL associated with these three earliness traits using the International Wheat Genome Sequencing Consortium (IWGSC) RefSeq v2.0 physical map. Our data consisted of (i) 6526 single nucleotide polymorphisms (SNPs) and two traits evaluated at five conventionally managed environments in the ‘Cutler’ × ‘AC Barrie’ population; (ii) 3158 SNPs and two traits evaluated across three organic and seven conventional managements in the ‘Attila’ × ‘CDC Go’ population; (iii) 5731 SilicoDArT and SNP markers and the three traits evaluated at four conventional and organic management systems in the ‘Peace’ × ‘Carberry’ population; and (iv) 1058 SNPs and two traits evaluated across two conventionally and organically managed environments in the ‘Peace’ × ‘CDC Stanley’ population. Using composite interval mapping, the phenotypic data across all environments, and the IWGSC RefSeq v2.0 physical maps, we identified a total of 44 QTL associated with days to heading (11), flowering (10), and maturity (23). Fifteen of the 44 QTL were common to both conventional and organic management systems, and the remaining QTL were specific to either the conventional (21) or organic (8) management systems. Some QTL harbor known genes, including the Vrn-A1, Vrn-B1, Rht-A1, and Rht-B1 that regulate photoperiodism, flowering time, and plant height in wheat, which lays a solid basis for cloning and further characterization.


Introduction
The application of nitrogen (N), phosphorus (P), and potassium (K) nutrients through the application of chemical fertilizers have become a standard conventional practice for increasing yield and yield components. However, excess use of chemical fertilizers under the conventional management system (high-N) not only increases the costs of crop production but also reduces nitrogen use efficiency (NUE) and causes environmental damage, which is shifting the wheat (Triticum aestivum L.) breeding programs towards low input agricultural practices [1]. Organic production (low-N) is a low input agricultural practice that relies on crop rotation, mixed cropping, biological pest control, and fertilizers of organic origin, such as compost, manure, and green manure [2]. Although organic production has been an effective approach to achieve agricultural sustainability in the long term and its demand has increased in the last decade for different reasons [3][4][5], most current cultivars have shown poor performance and 5-31% lower grain yield as compared to the conventional management system [6]. Hence, there is a need to identify and/or develop genotypes with higher input use efficiency that produce high yield under optimal-N without yield penalty under low-N, and also discover genes and QTL associated with high-N and low-N management systems.
The wheat breeding group at the University of Alberta has been conducting extensive research in the Canada Western Red Spring (CWRS) wheat class, including (i) developing several improved cultivars [7][8][9][10], (ii) evaluating the phenotypic performance of diverse cultivars under conventional and/or organic managements [11][12][13][14][15][16], (iii) understanding the genetics of earliness that serves as baseline data for developing early maturing cultivars to avoid frost damage [17][18][19][20], and (iv) mapping genes and QTL associated with diverse traits using biparental populations [19,[21][22][23][24][25][26][27] and a genomewide association mapping panel [28,29]. The development of early-maturing CWRS cultivars is of paramount importance in the northern breeding programs to provide farmers not only an option of growing the crop with minimal loss due to frost but also help to escape from the late incidence of diseases, heat, and drought as compared with their late-maturing counterparts. One of our aims was understanding the genetics of earliness under organic and conventional management systems, which is the basis for developing early maturing cultivars. In one of our studies, we evaluated the 'Cutler' × 'AC Barrie' recombinant inbred line (RIL) population across five conventionally managed field conditions and two greenhouse experiments and genotyped with 488 simple sequence repeat (SSR) and diversity arrays (DArT) markers. Using the phenotype data across all environments and a genetic linkage map, we uncovered two QTL on chromosome 1B and 5B associated with both flowering and maturity [19]. In a follow-up study, we reanalyzed the phenotype data of the 'Cutler' × 'AC Barrie' RIL population with a linkage map of 1809 SNPs selected out of the wheat 90K iSelect array and two functional markers (Ppd-D1 and Rht-D1). That study uncovered 19 QTL associated with four traits, including six QTL for flowering and five QTL for maturity [23].
In another study, we evaluated the 'Attila' × 'CDC Go' RIL population at three conventionally and organically managed environments and genotyped the population with Diversity Arrays Technology (DArT) and the Rht-B1 gene-specific functional markers. Using the phenotypic data across three environments and a linkage map of 580 markers, we uncovered three and five QTL associated with five agronomic traits under conventional and organic management systems, respectively [27]. For flowering and maturity, we only found environment (trial) specific QTL. To investigate if an increase in marker density improves QTL detection, we reanalyzed the same phenotype data of the eight traits averaged over the three organically managed environments and linkage maps of 1200 informative SNPs selected out of the wheat 90K iSelect array and three functional markers (Ppd-D1, Vrn-A1, and Rht-B1). Our analyses identified 16 QTL distributed across ten chromosomes [24] of which three were associated with flowering (1) and maturity (2). In conventional management, we evaluated the 'Attila' × 'CDC GO' RIL population for seven years and conducted QTL analysis using a linkage map of the 1203 markers. That study uncovered 14 QTL associated with eight traits, of which the three QTL for flowering and maturity remained the same as those identified in the organic management system [25].
Recently, we evaluated a RIL population derived from 'Peace' × 'Carberry' under two conventional and organic managements (2016-2017) and genotyped with DArT-based genotyping by sequencing (DArTseq) technology. Using the phenotype data of two environments per management, we found 53 QTL associated with nine agronomic traits, of which nine QTL were for heading (3) and maturity (6). Fourteen of the 53 QTL were common in both conventional and organic management systems [21]. We also mapped QTL associated with ten agronomic and end-use quality traits in 'Peace' × 'CDC Stanley' RIL population evaluated for two years (2016-2017) under conventional and organic management systems. Using phenotypic data across two environments per management and the IWGSC RefSeq v2.0 physical map position of 1058 informative SNPs selected out of the wheat 90K iSelect array, we uncovered a total of 27 QTL associated with nine traits, including QTL for heading (2) and maturity (4) [30]. Although we reported several QTL associated with the heading, flowering, and maturity time in the four mapping populations and two management systems, direct comparisons of the QTL results across independent studies was challenging due to the use of population-specific genetic maps, which form the basis of the present study.
Meta-QTL analysis [31] has been widely used for combining data from independent QTL mapping studies, identify the occurrence of QTL hotspots in a consensus genetic linkage map, and narrowing the QTL genetic confidence intervals [32][33][34][35][36][37]. QTL meta-analysis has been performed in wheat for several traits, including heading and maturity [36,38,39]. Some of the challenges in the use of both population-specific and consensus genetic maps for QTL detection in polyploid species, such as wheat, include (1) difficulty in establishing linkage groups and determining correct markers orders, (2) an increase in the number of linkage groups as compared with the number of chromosomes, which could be 2-3-fold [38], (3) collinearity among multiple markers due to relatively small population sizes that result to very small inter-marker interval (<1 cM) between multiple adjacent markers, and (4) the lack of consistent relationship between genetic and physical maps, which makes direct comparisons of QTL discovery results across multiple populations unreliable [39]. Although some studies have published physical maps of QTL associated with diverse traits [39][40][41][42][43][44], the method has not yet been widely used due to a lack of reliable physical maps. Recently, the International Wheat Genome Sequencing Consortium (IWGSC) has released the latest version of the bread wheat reference genome sequence (RefSeq v2.0) and physical map (http://wheat-urgi.versailles.inra.fr/; accessed on 22 April 2021). A simple online tool that helps breeders to map the chromosomal location and position of the wheat 90K iSelect array and genotyping by sequencing markers onto the flow-sorted Chinese Spring (CS) survey sequences have also been made available to the public (https://download.txgen.tamu.edu/shichen/mapper_v2.html; accessed on 22 April 2021). The objectives of the present study were, therefore, to (1) map and characterize QTL associated with heading, flowering and maturity time in four RIL populations using the IWGSC RefSeq v2.0 physical maps, and (2) assess the consistency of QTL identified in the conventional and organic management systems.

Phenotype and Genotype Data
Our study was based on a total of 698 RILs from four biparental populations, which were evaluated under conventional and organic management systems and genotyped either with the wheat 90K iSelect SNP array or the DArTseq technology (Table 1). In both conventional and organic management systems, the seven parents used in developing the four RIL mapping populations on average differed by 4-5 days in heading, 7-8 days in flowering, and 8-9 days in maturity. In three of our mapping populations, (i) 'CDC Go' headed/flowered and matured 3-4 days earlier than 'Attila' both under conventional and organic management systems: (ii) 'Cutler' headed/flowered and matured three days earlier than 'AC Barrie' under conventional management; and (iii) 'Peace' headed/flowed a day and matured three days earlier than 'CDC Stanley' under both management systems. On the other hand, 'Carberry' headed/flowered 3-4 days earlier but matured 2-3 days later than 'Peace' in both management systems.
The detailed phenotypic performance for the 'Attila' × 'CDC Go' (hereinafter abbreviated as ACG) [24,25], the 'Cutler' × 'AC Barrie' (hereinafter abbreviated as CAB) [23], and 'Peace' × 'CDC Stanley' (hereinafter abbreviated as PCS) [30] RIL populations have been described in our previous studies. Figure 1 shows the frequency distribution of the three traits, and Table 2 provides a summary of the descriptive statistics and broad-sense heritability under conventional and/or organic management systems. Among the four RIL populations, the means of days to heading, flowering, and maturity ranged from 49 to 52 days, from 52 to 57 days, and from 88 to 98 days, respectively. Broad-sense heritability under conventional and organic management systems varied from 0.40 to 0.77 (Table 2). There was a highly significant (p < 0. 01) positive correlation between the organic and conventional management systems with regression coefficients varying from 0.71 to 0.86 ( Figure 2). On the other hand, 'Carberry' headed/flowered 3-4 days earlier but matured 2-3 days later than 'Peace' in both management systems. The detailed phenotypic performance for the 'Attila' × 'CDC Go' (hereinafter abbreviated as ACG) [24,25], the 'Cutler' × 'AC Barrie' (hereinafter abbreviated as CAB) [23], and 'Peace' × 'CDC Stanley' (hereinafter abbreviated as PCS) [30] RIL populations have been described in our previous studies. Figure 1 shows the frequency distribution of the three traits, and Table 2 provides a summary of the descriptive statistics and broad-sense heritability under conventional and/or organic management systems. Among the four RIL populations, the means of days to heading, flowering, and maturity ranged from 49 to 52 days, from 52 to 57 days, and from 88 to 98 days, respectively. Broad-sense heritability under conventional and organic management systems varied from 0.40 to 0.77 ( Table 2). There was a highly significant (p < 0. 01) positive correlation between the organic and conventional management systems with regression coefficients varying from 0.71 to 0.86 ( Figure 2).     Of the wheat 90K iSelect array used for genotyping the RIL populations in the ACG, CAB, and PCS populations, we constructed three independent genetic maps that consisted of 3158, 6526, and 1058 markers, respectively. The genetic map of the PAC population consisted of 5731 DArTseq markers, which included 3840 SilicoDArT and 1891 SNPs ( Table 3 and Table S1). The genetic map lengths in the ACG, CAB, PAC, and PCS populations were 1734, 3596, 25508, and 4922 cM, respectively. The corresponding IWGSC RefSeq v2.0 physical map length in the ACG, CAB, PAC, and PCS populations were 11859, 13788, 13740, and 13881 Mb, respectively. Overall, a total of 15183 markers were used in the genetic and physical maps of the four populations, of which 1204 and 43 markers were common in two and three of the four populations, respectively (Table S1). Pearson correlation coefficients between the genetic and physical maps varied from 0.39 in the PAC to 0.60 in the CAB population ( Figure S1), suggesting a low to moderate agreement between genetic and physical maps that ultimately affects QTL detection. Of the wheat 90K iSelect array used for genotyping the RIL populations in the ACG, CAB, and PCS populations, we constructed three independent genetic maps that consisted of 3158, 6526, and 1058 markers, respectively. The genetic map of the PAC population consisted of 5731 DArTseq markers, which included 3840 SilicoDArT and 1891 SNPs (Table 3 and Table S1). The genetic map lengths in the ACG, CAB, PAC, and PCS populations were 1734, 3596, 25508, and 4922 cM, respectively. The corresponding IWGSC RefSeq v2.0 physical map length in the ACG, CAB, PAC, and PCS populations were 11859, 13788, 13740, and 13881 Mb, respectively. Overall, a total of 15183 markers were used in the genetic and physical maps of the four populations, of which 1204 and 43 markers were common in two and three of the four populations, respectively (Table S1). Pearson correlation coefficients between the genetic and physical maps varied from 0.39 in the PAC to 0.60 in the CAB population ( Figure S1), suggesting a low to moderate agreement between genetic and physical maps that ultimately affects QTL detection. Table 3. Summary of the number of markers, genetic and physical maps in each of the four mapping populations. The physical map is based on the International Wheat Genome Sequencing Consortium (IWGSC) RefSeq v2.0.

Physical Map of QTL
Using the least-squares means phenotype data and the IWGSC RefSeq v2.0 physical maps, we uncovered 36 QTL under the conventional management and 23 QTL under the organic management system (Table S2). When the results from the two management systems were combined, we found a total of 44 QTL in the ACG (3), PCS (8), CAB (10), and PC (23). Fifteen of the 44 QTL were common to both conventional and organic management systems, with the remaining QTL being specific to either the conventional (21) or organic (8) management systems (Table 4 and Table S2). All QTL were population specific (i.e., none of the QTL was detected in more than one population). For heading, we found a total of 11 QTL across seven chromosomes, which includes 8 QTL in the PAC and 3 QTL in the PCS populations. The 11 QTL for heading were mapped at 325.  (Table S2). Five of the heading QTL (QHd.dms-2B, QHd.dms-5A.1, QHd.dms-5B.3, QHd.dms-5B.4, and QHd.dms-7D) were common in both organic and conventional management systems ( Figure 3).

Coincident QTL
Overall, we found eight genomic regions on chromosomes 2B, 3B, 4A, 5A, 5B, and 7D harboring QTL clusters for 2-3 traits within each population ( Figure 3, Table 4, and Table S2). One of the coincident QTL spans 6.9 Mb between 55.8 and 62.7 Mb on chromosome 2B, which is associated with heading (QHd.dms-2B) and flowering (QFlt.dms-2B) in the PAC population in both conventional and organic management systems. The second and third coincident QTL were associated with days to heading and maturity in the PCS population both on chromosomes 3B (45.    Tables S1 and S2, respectively.

Coincident QTL
Overall, we found eight genomic regions on chromosomes 2B, 3B, 4A, 5A, 5B, and 7D harboring QTL clusters for 2-3 traits within each population ( Figure 3, Table 4, and Table  S2). One of the coincident QTL spans 6.9 Mb between 55.8 and 62.7 Mb on chromosome 2B, which is associated with heading (QHd.dms-2B) and flowering (QFlt.dms-2B) in the PAC population in both conventional and organic management systems. The second and third coincident QTL were associated with days to heading and maturity in the PCS population both on chromosomes 3B (45.2-49.0 Mb) and 4A (16.1-17.8 Mb). The fourth and fifth coincident QTL were associated with days to heading, flowering, and maturity in the PAC population evaluated in both management systems; these QTL, and they were mapped on chromosomes 5B (574.5-577.0 Mb), and 7D (73.3-84.0 Mb). The remaining three coincidental QTL were located on chromosome 5A that were associated with days to heading and maturity in the PAC population (619.7-621.3 Mb), flowering and maturity in the ACG population (587.3-587.4 Mb), and on chromosome 5B for heading and flowering in the PAC population (333.8-354.8 Mb). The long arms of chromosomes 5A and 5B consisted of QTL clusters from three (PAC, ACG, and CAB) and two (PAC and PCS) populations, respectively ( Figure 3).

QTL Based on Genetic and Physical Maps
In our previous study [23], QTL mapping conducted in the 'Cutler' and 'AC Barrie' population using genetic maps of 1811 markers and flowering and maturity data under field conditions converted into growing degree days (GDD) identified six QTL for flowering (2D, 3B, 4A, 5A, 6B and 7A) and five QTL for maturity (2D, 4A, 4D and 7A), which individually explained between 6.5 and 25.4% of the phenotypic variance under a conventional management system. However, we found out that the chromosomal location of flanking SNP markers of four of the QTL reported in our previous study differed from the physical map. When analyses were done using days to flowering and maturity time without conversion to GDD, we found four QTL for flowering on 2D, 3B, 6B, and 7A, and four QTL for maturity on 2D, 4A, 4D, and 7A (Table S3). In the present study conducted using the IWGSC RefSeq v2.0 physical map and days to flowering and maturity data, we identified 3 QTL for flowering on chromosomes 3B, 5A, and 7A, and seven QTL for maturity on 1A, 3B, 4A, 4D, and 5A ( Figure 3, Table S2). Although chromosomes 7A for days to flowering and 4A, 4D, and 5A for days to maturity were common in both our previous and the present studies, their positions are different. The left and right flanking markers identified in our previous study were physically far apart, including some on different arms of the same chromosome. QTL mapping conducted in the 'Attila' × 'CDC Go' RIL population using genetic maps of 1203 markers identified a coincident genomic region associated with both flowering and maturity under both management systems on chromosome 5A and another region on 4B associated with maturity under both management systems [24,25]. Each of these two genomic regions accounted for 5.9%-17.2% of the phenotypic variance per trait per management system (Table S3). QTL mapping conducted using the IWGSC RefSeq v2.0 physical map in this study identified both genomic regions, each individually explaining 3.7%-20.8% of the phenotypic variances for flowering and maturity across all environments (Table S2).
Using the phenotype data of 'Peace' and 'Carberry' RIL populations evaluated for two years under conventional and organic management systems and linkage map of 4439 markers generated via DArTseq technology, we previously [21] uncovered three QTL on chromosomes 1A, 5B, and 7D for heading and six QTL on 3A, 4A, 4B, 4D, 5B and 7D for maturity (Table S3). In the present study conducted using the IWGSC RefSeq v2.0 and four years phenotype, we identified eight QTL for heading and nine QTL for maturity ( Figure 3, Table 4). Seven of the nine QTL (all except 1A for heading and 4D for maturity) reported in our previous study were identified in the present study. Moreover, we uncovered new QTL on chromosomes 1D and 2B for heading and chromosome 5A for both heading and maturity. It should, however, be noted that the phenotype data used in the present study had four environments (2016-2020) instead of the two years (2016-2017) used in our previous study. In the 'Peace' and 'CDC Stanley" RIL population, QTL analyses in the previous [30] and present study were done using the IWGSC RefSeq v2.0 physical maps, so all QTL identified in our previous study were also identified in the present study. Because of additional data cleaning, however, two new QTL were uncovered in the present study on 4A (17.6-17.7 Mb) for heading and on 7A (96.8-110.0 Mb) for maturity. Because of the overlap in the IWGSC RefSeq v2.0 physical confidence interval among some of the QTL and/or presence of eight coincident genomic regions associated with two to three traits, the 44 QTL (Figure 3) identified in the present study fell into 31 genomic regions (Table S4). Nearly a third of the QTL were located on homoeologous Group 5 chromosomes (15 QTL), followed by Groups 3, 4, and 7 (seven QTL each), Group 2 (five QTL), and Group 1 (four QTL). Hanocq et al. [36] summarized 177 QTL reported across 13 independent studies on the D. Somers wheat consensus map [45]. After projecting 84 of the 177 initial QTL on the consensus map, the authors reported 30 genomic regions involved in the control of earliness (heading and flowering) and its three components (photoperiod sensitivity, vernalization requirement, and intrinsic earliness). Their meta-analysis demonstrated that the genetic control of earliness and its components involves not only the major Ppd and Vrn genes on homoeologous Groups 2 and 5 chromosomes, respectively, but also other genomic regions in Groups 4 and 7, which is evident in the current study.
Heading and flowering are important adaptative traits that determine the adaptation of wheat to diverse eco-geographical regions and are critical for yield potential and stability [46][47][48]. We uncovered genomic regions associated with heading on seven chromosomes (1D, 2B, 3B, 4A, 5A, 5B, and 7D) and for flowering on seven chromosomes (2A,  (Table 4  and Table S2). Using three RIL mapping populations evaluated across four environments, Zhao et al. [48] reported a total of 25 QTL associated with heading (15) and flowering (10) across 17 chromosomes, which individually explained from 4.3 to 32.5% of the phenotypic variance. The authors reported QTL for heading on fifteen chromosomes, including 1D, 2B, 3B, 4A, 5A, and 5B. They also reported flowering QTL on ten chromosomes, including 2B and 5B.
Four of the thirty-one genomic regions identified in the present study were mapped on chromosome 5A (Figure 3, Table S4), which were associated with heading (60. The genetic control of earliness in wheat is controlled primarily by the photoperiod response (Ppd), vernalization requirement (Vrn), and 'earliness per se' (Eps) genes. Eps genes act independently of the Vrn and Ppd genes [47,49]. Photoperiod insensitivity is controlled primarily by the dominant alleles at Ppd-A1, Ppd-B1, and Ppd-D1 that are located on chromosomes 2A, 2B, and 2D, respectively [50][51][52][53]. Spring wheat introgression lines lacking Ppd-B1 flowered 10-15 days later than controls under long days as compared to 1-5 days for lines that did not have the Ppd-A1 allele; loss of Ppd-D1 alleles did not affect flowering time [54]. On homologous Group 2 chromosomes, we found two regions on 2A (32.4-37.9 and 700.6-700.9 Mb) for flowering and one region on 2B (55.8-62.7 Mb) for both heading and flowering ( Figure 3). The two genomic regions for flowering on chromosome 2A consisted of three candidate genes, while the single genomic region on chromosome 2B consisted of 60 candidate genes. Although we are not sure about the exact physical positions of Ppd-A1 and Ppd-B1 due to conflicting positions of the SSR markers linked to both genes, the genomic regions that we identified on both 2A and 2B may contain the two photoperiod response genes.
On chromosome 7A, we found one genomic region associated with flowering (24.2-25.4 Mb) and three genomic regions associated with maturity (at 96.8-110.0, 133.4-134.2, and 680.6-717.9 Mb). These regions harbor between 10 and 473 genes, including the TraesCS7A02G146100 (7A:97873101-97873865) that plays a role in the signaling pathway. On chromosome 7D, we identified a single genomic region between 73.3 and 84.0 Mb associated with a heading, flowering, and maturity. This region consisted of 138 candidate genes, including TraesCS7D02G123900 (7D:77241990-77242100), which is Photosystem II reaction center protein I. Both chromosomes 1A and 3A consisted of three (35.5-39.8, 399.4-426.6, and 443.0-462.8 Mb) and two (392.0-427.8 and 513.8-553.1 Mb) genomic regions, respectively, which all were associated with maturity. Each genomic region on chromosome 1A and 3A harbors clusters of protein-coding candidate genes ranging from 47 to 205 (Table S4)

Effects of Genetic Background and Management
All QTL identified in the present study were population (cross) specific. The closest QTL detected in this study were QFlt.dms-5A.1 in the CAB population and both QFlt.dms-5A.2 and QMat.dms-5A.2 in the ACG population ( Figure 3), but there is a~4.6 Mb interval between them. Although identification of QTL conserved across multiple genetic backgrounds is one of the prerequisites for marker-assisted selection, most QTL reported in the literature are population (genetic background) specific [37,[58][59][60][61][62], which restricts their application for predicting phenotypic performance across diverse genetic backgrounds. Brasier et al. [62] evaluated two biparental winter wheat mapping populations derived from a cross between two high NUE parents and a shared common low NUE parent for 11 traits. Using genetic maps of 3147 and 3918 markers, the authors uncovered a total of 130 QTL, of which 10 QTL were common between the two populations. Symonds et al. [60] detected nine QTL associated with trichome density in four RIL populations of Arabidopsis thaliana (two of their populations shared a common parent), of which only two QTL were detected in all four populations; the other 7 QTL were population specific. In rice, Yao et al. [58] evaluated three biparental populations with a common parent and uncovered 28 QTL associated with African rice gall midge resistance, but there was only a single QTL common in two of the three populations. In soybean, Kang et al. [61] used two mapping populations with a common parent and uncovered eight QTL associated with pod dehiscence, but only one QTL was common between the two populations. Semagn et al. [37] used 18 bi-parental maize populations and identified a total of 183 QTL associated with grain yield (101) and anthesis silking interval (82) under drought and irrigated conditions. However, only a few QTL were detected across 2-6 populations and/or the two water regimes.
Identification of consistent QTL in both conventional (high-N) and organic (low-N) management systems would be highly useful for improving spring wheat through markerassisted selection. However, our results demonstrated that~34% of the 44 QTL uncovered in the present study were common between the organic and conventional management systems, with the remaining being specific to the conventional (48%) or organic (18%) management systems (Figure 3, Table 4). QTL detected in both management systems would be ideal for developing improved wheat germplasm using marker-assisted breeding irrespective of management system, while those QTL detected in the conventional (high-N) or organic (low-N) managements should be considered in their respective managements. Several studies have reported similar results in diverse crops, including wheat [62][63][64][65][66], rice [67,68], barley [69], sorghum [70], and potato [71]. Using a doubled haploid winter wheat mapping population evaluated at two N fertilizer treatments under field conditions, An et al. [64] detected a total of 17 QTL associated with different traits under low-N (9) and high-N (8), but none were detected in both N treatments. Laperche et al. [66] evaluated grain protein, grain yield, and their components in a DH winter wheat population under high-N and low-N, which detected a total of 67 QTL under high-N and 51 QTL under low-N, of which 13 QTL were detected at both N levels. Yue et al. [68] evaluated a rice RIL mapping population under normal-N and low-N levels and detected a total of 52 QTL associated with four yield-related traits. Eleven of the 52 QTL were detected under the two N levels and the remaining 30 QTL were detected either in the normal-N (17) or low-N (13).
Using a RIL sorghum mapping population evaluated under normal-N and low-N fertilizer conditions, Gelli et al. [70] detected a total of 38 QTL associated with 11 agronomic traits, of which four QTL were common between the two N management systems, and the remaining were specific either to the low-N (16) or normal-N (14). RNA sequencing analyses on sorghum seedling root tissues revealed 726 differentially expressed gene transcripts related to nitrogen uptake and metabolism between parents, of which 108 were mapped close to the QTL regions. Expression analysis of N metabolism-related genes reveals differences in the performance of two barley cultivars that were adapted to low-N and high-N levels [69]. The authors reported that some N metabolism-related genes were only induced in shoots of low-N tolerant varieties [69,71]. Leaf and root transcriptomic profiles analyses in potato cultivars have also revealed differentially expressed N metabolism-related genes [71], which may be one possible reason for the mapping of multiple management-specific QTL in our study.
The conventional and organic evaluation sites are located at the University of Alberta South Campus (53 • 19 N, 113 • 35 W), Edmonton, Canada, and were about 500 m apart. Trials were laid out in randomized incomplete block design, with two replications. Trials in the conventional management were planted earlier, broadcast-fertilized with 70 kg ha −1 of 46-0-0 (N-P 2 O 5 -K 2 O) in early spring, and band-fertilized during seeding with 36 kg ha −1 of 11-52-0 (N-P 2 O 5 -K 2 O). Weeds were controlled in conventional trials using registered herbicides following local recommendations and label directions. The trials under organic management were planted later to facilitate early season land operations (tillage) and did not receive any of the chemical inputs. The four-year crop rotation in the organic land was wheat, rye (Secale L.) plow-down, field pea (Pisum sativum), and canola (Brassica napus L.), whereas conventional land followed a three-year rotation of wheat, field pea, and canola.
The methodologies for DNA extraction and genotyping have been described in our previous studies in the CAB [23], ACG [24], PAC [26], and PCS [30] populations. DNA samples from the ACG, CAB, and PCS populations were genotyped at the University of Saskatchewan Wheat Genomics lab, Saskatoon, Canada, with the wheat 90K iSelect array that consisted of 81,587 SNPs [79], while the PAC population was genotyped with the DArTseq™ technology by DArT Pty Ltd., Canberra, Australia [80,81]. In addition, we also screened the two parents of each population if they were polymorphic for a few functional markers linked to photoperiod response (Ppd-B1 and Ppd-D1) [82], vernalization response (Vrn-A1 and Vrn-B1) [51], and height reducing Rht-B1 [83] genes at the Agricultural Genomics and Proteomics Lab, University of Alberta, Edmonton, Canada, as described in our previous studies [21,23,53]. In cases where the two parents of each population showed polymorphism for the functional markers, we genotyped the RILs with the markers and used them for mapping.

Statistical Analysis
Least-squares means of each phenotypic trait and broad-sense heritability were computed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA) as described in our recent study [21]. Pearson correlations, regression (R 2 ) coefficients, and different types of graphs were generated using both Minitab v14 (https://www.minitab.com; accessed on 22 April 2021) and JMP v7 (www.jmp.com; accessed on 22 April 2021) statistical software. For the PAC population, we received a total of 36,626 markers from DArT Pty Ltd., Canberra, Australia (https://www.diversityarrays.com; accessed on 22 April 2021), which consisted of 22,741 SilicoDArT markers (present vs. absent variation) and 13,885 SNPs along with 69 bp sequence based on the Chinese Spring (CS) reference sequence. Nearly 42% of the markers (11,704 SilicoDArT and 3557 SNPs) were polymorphic between the two parents. Marker sequences obtained from DArT Pty Ltd. were used for BLAST searches against the Chinese spring genome, the International Wheat Genome Sequence Consortium (IWGSC) RefSeq v1.0 [84] and RefSeq v2.0, which are available at http://download.txgen.tamu.edu/shichen/mapper_v2.html and http://wheat-urgi.versailles.inra.fr/; (accessed on 22 April 2021) The top hits with the highest alignment length, highest similarity (>95%), and/or expected value of 1 × 10 −20 were used for physical mapping as described elsewhere [42]. Nearly 58% of the polymorphic markers (8,780 of the 15,261 markers) with physical information were selected for linkage analysis using MapDisto for Windows v2.1.7.10 [85]. After excluding all markers that showed segregation distortion at p < 0.01, a missing data of >20%, and those that were either unlinked or formed a linkage group with <5 markers, we retained 7066 markers for final map construction. Linkage groups were assigned to individual wheat chromosomes following the IWGSC RefSeq v2.0. Markers that were either assigned to a chromosome different from expected based on IWGSC RefSeq v2.0, had incorrect locus order, and/or that significantly affected the map length were further removed from the datasets. Final genetic linkage maps were constructed using the IWGSC RefSeq v2.0 marker order in MapDisto. Linkage analyses and genetic mapping in the other three populations (ACG, CAB, and PCS) genotyped with the wheat 90K iSelect array were done as described above. The only difference with the PAC population was the source of the sequence information for the wheat 90K iSelect array, which was retrieved from http://download.txgen.tamu.edu/shichen/flanking_v2.html (accessed on 22 April 2021).
Inclusive composite interval mapping (ICIM) was done on the least-squares means using QTL IciMapping version 4.2.53 [86] as described in our previous studies using the IWGSC RefSeq 2.0 physical maps with the following parameters: a minimum logarithm of the odds (LOD) scores of 3.0, a mean replacement for missing phenotypic data, an additive model to determine the effect of individual QTL, and a walking step of 100 kb. QTL scanning was performed independently on each population using population-specific physical maps, while QTL results were presented on to the integrated physical map of all four populations. QTL names were assigned by following the International Rules of Genetic Nomenclature (http://wheat.pw.usda.gov/ggpages/wgc/98/Intro.htm; accessed on 22 April 2021), which comprised of trait acronym, lab designation (dms = Dean Michael Spaner), and chromosome number. A physical QTL map of each chromosome was generated using MapChart v2.1 [87]. The start positions of the left and right flanking markers of each QTL (confidence interval) were used to search for candidate genes at the Ensemble Plants using Triticum aestivum genome (https://plants.ensembl.org/index.html; accessed on 22 April 2021). To minimize the length of the manuscript. However, the results from the Candidate genes search were used only in the discussion section.

Conclusions
Using the IWGSC RefSeq 2.0 physical map and phenotype data of four RIL mapping populations evaluated under conventional and organic management systems, we mapped and characterized a total of 44 QTL associated with a heading (11), flowering (10), and maturity (23) dates in hard red spring wheat. Fifteen of the 44 QTL were common in the two management systems, which would be ideal for developing improved wheat germplasm using marker-assisted breeding irrespective of management conditions. Those QTL detected under conventional (high-N) and organic (low-N) management would be useful for MAS in each management specific conditions. Each QTL individually explained from 1.7 to 20.8% and together accounted for 14.7%-59.4% and 4.0%-52.3% of the phenotypic variance under conventional and organic management systems, respectively. Because of the overlap in the physical confidence interval of some of the QTL and/or presence of eight coincident QTL associated with 2-3 traits, however, there are only 31 genomic regions associated with the three earliness traits. Some of the genomic regions harbor known genes, including the vernalization response Vrn-A1 on chromosome 5A and Vrn-B1 on 5B, as well as the height reducing Rht-A1 on 4A and Rht-B1 on 4B; these genes regulate photoperiodism, flowering time, and plant height in wheat. As far as we are aware, this is the first comprehensive physical map of genomic regions associated with the heading, flowering, and maturity in hard red spring wheat populations that would provide valuable information to wheat researchers and opens the opportunity for direct comparisons of QTL discovery studies across independent studies and possibly map-based cloning of a few regions.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/plants10050853/s1, Figure S1. Scatter plot of IWGSC RefSeq v2.0 physical map against genetic map in each of the four mapping populations used in the present study. Table S1. Genetic and physical maps of 16,504 markers used in the four recombinant inbred line populations. Table S2. Summary of Quantitative trait loci (QTL) associated with days to heading, flowering, and maturity identified using least-squares means phenotype across all environments per management and the IWGSC RefSeq v2.0 physical maps in four recombinant inbred lines populations evaluated under conventional and organic management systems. Table S3. Summary of quantitative trait loci (QTL) associated with the heading, flowering, and maturity dates identified in our previous studies using least squares means phenotype across all environments per management and linkage maps of the four recombinant inbred lines populations evaluated under conventional and organic management systems. Table S4. Summary of the 31 genomic regions associated with the heading, flowering, and maturity, and the number of candidate genes that fell within each region.

Acknowledgments:
The authors would also like to express appreciation to Klaus Strenzke, Joseph Moss, Izabela Ciechnowska, Fabiana Dias, Katherine Chabot, Tom Keady, and Russel Puk for their technical support with phenotype data collection in both conventional and organic management systems.

Conflicts of Interest:
The authors declare no conflict of interest.