Genomic Regions Influencing Preharvest Sprouting Tolerance in Two Doubled-Haploid Wheat Populations (Triticum aestivum L.)

Abstract

The current and projected climate change that is represented by increasing temperatures and humidity levels and irregular rainfall patterns promotes the occurrence of preharvest sprouting (PHS) in wheat. PHS results in significant economic losses, globally, which necessitates the need for high-yielding cultivars with increased PHS tolerance; hence, this study was conducted. The current study evaluated two doubled-haploid (DH) wheat populations of Tugela-Dn × Elands and Elands × Flamink across six environments in the Free State Province of South Africa to select genotypes with increased PHS tolerance and further map the underlying loci. Significant effects of DH lines (194) and environments (6) were observed for PHS tolerance. The results of this study validate previous findings that PHS is only expressed when environmental conditions are conducive. Quantitative trait loci (QTL) mapping using single-nucleotide polymorphism (SNP) and silicoDArT markers revealed three additive QTLs with major effects on chromosomes 5B and 7B, and these QTLs were detected more than once, when conditions were favourable. These QTLs explained a phenotypic variation (PVE) varying between 10.08% and 20.30% (LOD = 2.73–3.11). About 16.50% of DH lines performed to the level of Elands (the PHS-tolerant parent) and are recommended for further selection in a pre-breeding or breeding programme. The findings of this study are expected to expedite the on-going breeding efforts for PHS tolerance in winter wheat, which will facilitate the development of PHS-tolerant cultivars adapted to the South African environment.

1. Introduction

Preharvest sprouting (PHS) is the premature germination of kernels on physiologically mature wheat ears upon continuous wet and humid conditions during the harvest season [1,2]. This phenomenon has resulted in a significant reduction in wheat grain yield and end-use quality, worldwide [3,4,5,6]. This is due to the activation of enzymes such as lipases, amylases and proteases in developing kernels, which leads to the degradation of lipids, starch and proteins [7,8] and thus reduces the market value of wheat grain by up to 50% [4,9,10]. Annual grain yield and end-use quality losses due to PHS are estimated above USD 1 billion, worldwide [5,11,12].

PHS tolerance remains a trait of interest in the South African (SA) wheat production, especially amid the current and forecasted climatic change that significantly impacts grain yield and quality [13,14]. Most of the wheat production regions in the country receive summer rainfall around harvest, which puts the production at high risk of PHS [15,16,17,18,19]. Various methods have been used to evaluate and improve PHS tolerance in released commercial cultivars [20]. A visual screening test of more than 2% sprouted kernels per 25 g wheat sample together with a falling number below 220 s has been used to downgrade wheat to lower grades, such as grades B1 to B3, depending on the protein content and hectolitre mass [20,21,22].

Significant improvement in PHS tolerance in bread wheat, especially winter cultivars, has been achieved in SA over the years through conventional breeding [23]. Most current commercial cultivars either have good or moderate PHS tolerance [20]. This improvement has resulted in extensive genetic diversity in PHS tolerance [21,23,24], which has been used in wheat breeding programmes across the country. The use of molecular markers in breeding for PHS tolerance and to identify lines with potential tolerance to PHS has only been incorporated recently in wheat breeding programmes in the country [25].

The identification of genomic regions and associated molecular markers influencing PHS tolerance is anticipated to speed up the development of PHS-tolerant cultivars that are adapted to the SA environment. However, the complex nature of PHS tolerance and the interaction between associated genes and the environment can cause the cultivars’ level of tolerance to PHS to vary from year to year [25,26,27]. This makes selection for high-yielding varieties with increased PHS tolerance, which are stable across environments and years, challenging, especially under the current and predicted climate change6 [28].

Contrastingly, more progress has been attained in breeding for PHS tolerance (alongside other grain yield-related traits) through MAS in most parts of the world [6,29,30,31,32,33,34,35,36,37,38,39,40]. These studies proved an extensive genotypic variation for PHS tolerance as a result of the complex genetic nature of this trait. This variation has been further utilised in wheat breeding programmes and exchanged globally. Moreover, negative correlations between PHS tolerance and grain yield, seed viability, seedling vigor, flour yield and baking quality have been reported [10,27,41,42], proving the importance of this trait in the improvement in grain yield and quality. Therefore, understanding the genetics of PHS tolerance in wheat can accelerate the needed improvements in grain yield and quality [43,44,45,46,47]; nonetheless, plant adaptation remains vital in the overall plant performance [48].

The present study was conducted to evaluate PHS tolerance in bread wheat winter lines planted across multiple environments to select doubled-haploid (DH) lines with increased PHS tolerance and further map the underlying loci. The study objectives were to (1) examine the performance of DH lines and parents with regard to PHS tolerance across six environments; (2) identify quantitative trait loci (QTLs) controlling PHS tolerance in the Tugela-Dn × Elands DH population; and (3) validate the presence of detected QTLs in a population with a different genetic background, Elands × Flamink.

2. Materials and Methods

2.1. Plant Material, Study Area and Experimental Design

Two DH wheat populations (n = 210) derived from two crosses of Tugela-Dn × Elands [49] and Elands × Flamink [50] and the three respective parents as checks (Tugela-Dn, Elands and Flamink) were provided by the Agricultural Research Council–Small Grain Germplasm Bank and evaluated for PHS tolerance. The Tugela-Dn × Elands DH population [49] was specifically developed to evaluate PHS tolerance and grain yield-related traits attributed to the two widely used parents with contrasting reactions to PHS, adaptability and grain morphological characteristics [51,52]. Tugela-Dn is a winter wheat cultivar highly susceptible to PHS; however, it has a high yield potential. This cultivar has been extensively used for dryland wheat production in South Africa, since 1992. Elands is a facultative cultivar that was released for dryland wheat production in 1998. This cultivar has a high yield potential, excellent tolerance to PHS and exceptional bread-making quality, which makes it a quality standard in the South African wheat industry [50]. The Flamink parental cultivar of the Elands × Flamink DH population [50] has vernalisation requirements and a high yield potential; however, it exhibits a reduced level of tolerance to PHS.

This study was conducted in six environments in the Free State Province of South Africa over two years, 2016 and 2017. Environments included Arlington 2016 (ARL1), Bethlehem 2016 (BHM3), Bethlehem 2017 (BHM4), Clarens 2016 (CLAR5), Harrismith 2016 (HAR7) and Harrismith 2017 (HAR8). The respective locations and weather descriptions of the six environments are shown in

Table 1. Descriptions of the six study environments between grain filling, maturity and harvest stages of wheat in 2016 and 2017 planting seasons. The weather data for the six environments are only shown for the months when the wheat plant was still in the field, from the grain filling stage until harvest.

† Env	Period	‡ Geographic Position			Average Daily Temperature (°C)		Average Daily Humidity (%)		Average Daily Rainfall (mm)
		Longit.	Latit.	Altit. ǂ (m.a.s.l.)	Min	Max	Min	Max
ARL1	October 2016	26.7732	28.0046	1435	12.67	27.00	28.33	48.67	1.66
	November 2016				17.67	31.33	33.00	62.67	0.02
	December 2016				18.67	31.67	30.00	52.00	0.03
	January 2017				17.33	31.33	26.00	46.00	0.00
BHM3	October 2016	28.2973	−28.1628	1721	11.00	26.00	33.50	92.50	1.38
	November 2016				14.05	27.26	35.25	94.40	4.01
	December 2016				13.63	27.94	37.29	93.77	3.11
	January 2017				13.32	26.20	41.65	94.53	4.56
BHM4	October 2017	28.2973	−28.1628	1721	7.06	24.61	27.28	90.13	1.39
	November 2017				9.04	26.76	25.27	90.81	3.14
	December 2017				12.23	26.66	35.84	93.01	3.69
CLAR5	October 2016	28.5838	−28.5038	1849	8.17	25.47	18.71	84.18	1.22
	November 2016				11.37	25.38	34.19	92.64	3.45
	December 2016				12.36	27.24	33.54	92.50	3.73
	January 2017				12.30	25.00	41.07	93.13	3.90
HAR7	October 2016	29.11596	−28.3128	1720	9.17	25.85	23.08	87.88	1.42
	November 2016				12.09	25.42	42.99	91.43	4.43
	December 2016				13.18	27.45	42.06	89.57	4.60
	January 2017				12.66	26.58	47.29	89.68	5.54
HAR8	October 2017	29.11596	−28.3128	1720	7.65	24.89	30.41	84.01	1.6
	November 2017				9.76	27.05	28.93	80.81	2.73
	December 2017				11.93	26.05	42.76	89.45	6.74

† Environment denotes ARL1 for Arlington 2016, BHM3 for Bethlehem 2016, BHM4 for Bethlehem 2017, CLAR5 for Clarens 2016, HAR7 for Harrismith 2016 and HAR8 for Harrismith 2017. In BHM4 and HAR8, wheat was harvested in December 2017; therefore, no weather data are shown for January 2018. ‡ Geographic position denotes Longit. for longitude, Latit. for latitude and Altit. for altitude. ǂ m.a.s.l. denotes metres above sea level.

. An augmented design was used in all environments as described in Khumalo et al. [50]. A single replicate of each DH line and five replicates of the three parents were grown in 1 m rows with an inter-row and intra-row spacing of 0.45 and 0.5 m, respectively. Commercial production and agronomic practices were followed as recommended for the specific production region.

2.2. Phenotypic Evaluation of PHS Tolerance

At the anthesis stage, 28 ears per DH line and parent were randomly tagged using insulation tape. As described by Barnard et al. [15], the tagged wheat ears were hand-harvested at physiological maturity, air-dried at room temperature and stored in a cold room (4 °C) to maintain dormancy until PHS evaluation. Wheat ears were then subjected to simulated rainfall in a humidified chamber at 15 °C/25 °C day/night temperature with 98% humidity for 72 h and eventually scored for PHS tolerance according to a rating scale of 1 (not sprouted)–8 (highly sprouted) (Figure S1; [15]).

2.3. Phenotypic Evaluation and Statistical Analysis

Statistical analyses were performed in Genstat 18th Edition [53] using 194 DH lines (139 Tugela-Dn × Elands + 55 Elands × Flamink) following the removal of DH lines with missing data in most of the environments. Data were tested for normality using the Shapiro–Wilk test and the Wilcoxon matched-pairs test prior to conducting the analysis of variance (ANOVA). ANOVA and the non-parametric Kruskal–Wallis test were used to examine significant effects of genotypes, environments and the genotype × environment interaction. The genotype × environment interaction was estimated from the error mean square (MS_ge) of the replicated parents within environments according to an augmented design defined by Federer [54]. Patterns of genotype × environment interaction and genotype stability were illustrated on the additive main effects and multiplicative interaction (AMMI) biplot, and the frequency distribution of phenotypes in the six environments was depicted in histograms. The broad-sense heritability (H²) estimate of PHS tolerance was calculated using the following formula [55]:

$$1 - \frac{MSge}{MSg} \mathrm{or} \frac{{\sigma }^{2}g}{[{\sigma }^{2}g+\left(\frac{{\sigma }^{2}ge}{e}\right)+\left(\frac{{\sigma }^2\mathrm{e}}{re}\right)] }$$

where MS_ge and MS_g represent the genotype × environment and the genotype mean squares, respectively; ${\sigma }^{2}g$ is the genotypic variance = ($\frac{MSg-MSge}{re});$ σ²ge is the genotype × environment interaction variance = $\frac{MSge-MSe }{r}$; e and r represent the number of environments and replications, respectively; and ${\sigma }^{2}e$ is the error variance = MS_e.

2.4. Genotyping and Construction of Genetic Map

The total genomic DNA was extracted from fresh leaves of three-week-old plants of 194 DH lines and 3 parents according to the Diversity Arrays Technology (DArT) plant DNA extraction protocol (https://www.diversityarrays.com/orderinstructions/plant-dna-extraction-protocol-for-dart/; accessed on 20 March 2019). The extracted DNA was genotyped with the DArT-sequencing genotype-by-sequence (GBS) platform 1.0 (DArT, Pty Ltd., Yarralumla, ACT, Australia), which produced 3204 single-nucleotide polymorphism (SNP) and 9117 silicoDArT markers. Genotypic data were cleaned for redundant and non-polymorphic markers, markers with switched alleles, markers with ≥50% missing data and significantly distorted markers (p < 0.05) in RStudio version 1.1.463 [56] and JoinMap^® version 4.1 [57]. A total of 483 SNP and silicoDArT polymorphic markers were used to construct a genetic map for the Tugela-Dn × Elands mapping population, while 1144 silicoDArT markers formed a genetic linkage map for the Elands × Flamink mapping population. The order of markers within a linkage group was established based on a regression mapping algorithm [58]. Map distances (cM) were calculated from recombination frequencies using the Kosambi mapping function [59].

2.5. QTL Analysis

QTL analysis was performed using Windows QTL Cartographer version 2.5 [60]. Composite interval mapping (CIM) was used to screen for significant QTLs using individual mean scores per environment and average mean scores across all environments. QTL detection was based on 1000 permutations (α = 0.05). The forward regression model was used with a window size of 10 cM, a walk speed of 2 cM and five control markers. QTLs were named following the international rules of genetic nomenclature adapted for wheat [61].

3. Results

3.1. Phenotypic Performance of Genotypes and Parents across Multi-Environments

The performance of genotypes in the six environments was depicted through histograms (frequency distribution, Figure 1) and an AMMI biplot (genotype stability, Figure 2). Both the Shapiro–Wilk test (p < 0.001) and Wilcoxon matched-pairs test (p < 0.001) proved the non-normality of PHS tolerance in some study environments. However, the results of the ANOVA and Kruskal–Wallis test were almost similar, suggesting less variation between variables. Significant differences among the three parents, one-hundred and ninety-four genotypes and six environments were observed (

Table 2. Mean squares and the broad-sense heritability (H²) estimate of PHS tolerance in 194 DH lines and 3 parents across 6 environments.

Source of Variation	Degrees of Freedom	Mean Square	F (p-Value)
Parents	2	84.873	<0.001
Environment	5	5.434	<0.001
Replications	4	1.861
Parents × Environment	9	1.580	0.017
Residual	56	0.627
Genotypes	193	2.429	<0.001
Environment	5	74.182	<0.001
Replications
Genotype × Environment	799	1.114	0.017
H2		0.5414
H2 (%)		54.14

). The genotype × environment interaction, which was estimated from the error mean square (MSge) of the replicated parents according to an augmented design defined by Federer [54] was significant for PHS tolerance. Among the parents, Elands was PHS-tolerant (score of 2.00) and Flamink was moderately tolerant (score of 3.40), while Tugela-Dn displayed a susceptible reaction to PHS (score of 5.00) (Figure 1,

Table 3. PHS tolerance scores of the top ten best- and five worst-performing DH lines and parents across the six environments to highlight the overall performance of the two DH populations.

† Genotype/Parent	ǂ Environment						Average PHS Tolerance Score
	ARL1	BHM3	BHM4	CLAR5	HAR7	HAR8
PHS tolerance scores of the top ten best-performing DH lines
EF 44	*	2	1	2	2	1	1
EF 15	*	2	2	1	1	4	2
EF 17	*	2	1	1	2	5	2
EF 47	*	4	1	1	1	1	2
TE 21	1	2	2	1	3	3	2
TE 37	2	2	2	2	1	4	2
TE 62	2	2	4	2	2	1	2
TE 73	2	3	2	1	3	1	2
TE 122	2	4	1	1	2	1	2
TE 127	1	3	2	2	2	4	2
PHS tolerance scores of the top five worst-performing DH lines
TE 48	4	6	4	2	5	5	4
TE 145	5	4	5	4	5	2	4
TE 67	5	6	4	4	6	3	5
TE 149	5	5	6	5	6	5	5
TE 155	4	6	3	4	6	5	5
PHS tolerance scores of the three parental cultivars
Tugela-Dn	4	6	5	5	5	5	5
Elands	2	3	2	1	2	2	2
Flamink	*	5	3	3	4	2	3.4

† Genotype denotes a DH line. TE denotes a Tugela-Dn × Elands DH line, while EF denotes an Elands × Flamink DH line. Parent denotes the three parental cultivars, i.e., Elands, Flamink and Tugela-Dn. ǂ Environment denotes ARL1 for Arlington 2016, BHM3 for Bethlehem 2016, BHM4 for Bethlehem 2017, CLAR5 for Clarens 2016, HAR7 for Harrismith 2016 and HAR8 for Harrismith 2017. * denotes missing data.

). As expected, on average, the frequency distribution of genotypes was continuous (Figure 1), indicating the presence of transgressive segregation, with some individuals exhibiting higher or lower PHS tolerance scores than the parents. This observation suggested polygenic control. A broad-sense heritability estimate of 0.5414 was calculated for DH lines (Table 2).

Two principal components of the AMMI biplot, PC1 = 36.16% and PC2 = 23.57%, explained the observed phenotypic variation in genotypes across environments (Figure 2). Less genotype × environment interactions (indicated by the length of vectors) were observed in ARL1, BHM4 and CLAR5 in contrast to BHM3, HAR7 and HAR8. Most DH lines were found clustered toward the centre of the biplot, proving broad adaptation and good performance. The average performance of DH lines across environments followed the order HAR8 (most tolerant) > CLAR5 > BHM4 > HAR7 > ARL1 > BHM3 (least tolerant), with PHS tolerance scores ranging from 3.00 to 4.00. These observations suggest tolerant to moderate reactions of the two DH populations. DH lines exhibited PHS tolerance scores ranging from 1.00 to 8.00, with an average score of 3.19 ± 1.32. (mean ± standard deviation).

3.2. Selection of Best-Performing Genotypes

On average, most (62.37%) DH lines had PHS tolerance scores ≤ 3.00 (from a scale of 1 to 8, Figure S1, Table S1), indicating the tolerance of the two populations to PHS [15]. About 16.50% of DH lines performed to the level of Elands (the PHS-tolerant parent), with PHS tolerance scores ≤ 2.00, and are recommended for further selection in a pre-breeding or breeding programme. Table 3 shows the PHS tolerance scores of the best-performing DH lines (represented by the top ten) in contrast to the parents across the six environments. Poorly performing DH lines similar to Tugela-Dn (the PHS-susceptible parent with a score of 5.00) are also represented by the top five worst-performing genotypes in Table 3. The overall phenotypic data of the 194 DH lines and the 3 parents across the 6 environments are provided as supplementary data (Table S1).

3.3. Genetic Linkage Map Construction

Genetic maps of the two DH populations, Tugela-Dn × Elands and Elands × Flamink, were constructed using different marker sets genotyped with the DArT-sequencing genotype-by-sequence (GBS) platform 1.0 (DArT, Pty Ltd., Yarralumla, ACT, Australia). Many markers, however, with a high level of missing data, had been used to genotype the Tugela-Dn × Elands population. For the quality and reliability of the results, only 483 polymorphic markers with ≥50% of genotypic data were used for genetic map construction of the Tugela-Dn × Elands population, and both parents share “SA1684” in their pedigrees, which resulted in fewer markers mapped per chromosome. The high level of missing data is usually the case with DArTseq GBS SNPs and a major concern for further applications, such as QTL mapping [62,63,64]. This is due to the low sequencing coverage that significantly reduces the number of usable SNPs, lower marker density and the resulting linkage map density [65,66].

The 483 polymorphic markers consisted of an ALMT1-4 functional marker pair, 259 SNP markers and 223 silicoDArT markers. Twenty-three linkage groups (LG) were identified representing the twenty-one wheat chromosomes. The entire genetic map spanned 1516.57 cM of the wheat genome, with an average distance of 3.87 cM between adjacent markers (

Table 4. Genetic linkage maps showing marker distribution in the 21 wheat chromosomes in Tugela-Dn × Elands and Elands × Flamink mapping populations.

	Tugela-Dn × Elands Linkage Map				Elands × Flamink Linkage Map
LG †	Chrom ǂ	No. of Markers	Map Length (cM)	Marker Density (cM)	Chrom ǂ	No. of Markers	Map Length (cM)	Marker Density (cM)
1	1A	19	49.33	2.60	1A	67	19.96	0.30
2	1B	30	90.05	3.00	1B	81	19.76	0.24
3	1D	25	76.50	3.06	1D	51	21.69	0.43
4	2A	17	96.05	5.65	2A	83	17.76	0.21
5	2B	40	75.98	1.90	2B	57	13.63	0.24
6	2D	11	73.56	6.69	2D	56	12.02	0.21
7	3A	8	49.80	6.22	3A	79	20.89	0.26
8	3B	29	58.84	2.03	3B	64	14.45	0.23
9	3D.LG1	8	28.95	3.62	3D	59	11.84	0.20
10	3D.LG2	7	60.95	8.71	4A	25	9.38	0.38
11	4A	8	48.35	6.04	4B	11	13.29	1.21
12	4B	23	82.90	3.60	4D	53	14.07	0.27
13	4D	15	74.35	4.96	5A	72	15.64	0.22
14	5A	14	53.35	3.81	5B	47	13.89	0.30
15	5B	31	76.91	2.48	5D	67	14.47	0.22
16	5D	15	34.15	2.28	6A	31	11.78	0.38
17	6A	20	57.27	2.86	6B	32	15.27	0.48
18	6B.LG1	31	17.90	0.58	6D	59	11.61	0.20
19	6B.LG2	27	75.70	2.80	7A	42	12.18	0.29
20	6D	7	55.74	7.96	7B	40	12.97	0.32
21	7A	34	126.62	3.72	7D	68	15.07	0.22
22	7B	39	110.63	2.84
23	7D	25	42.67	1.71
	A sub-genome	120	480.77	4.01	A sub-genome	399	107.59	0.27
	B sub-genome	250	588.92	2.36	B sub-genome	332	103.25	0.31
	D sub-genome	113	446.88	3.96	D sub-genome	413	100.75	0.24
	Total Genome	483	1516.57	3.87	Total Genome	1144	311.59	0.27

† LG denotes linkage group. ǂ Chrom denotes chromosome.

). Genetic distances between adjacent markers ranged from 0.58 cM on chromosome 6BLG1 to 8.71 cM on chromosome 3DLG2. The number of markers on each chromosome varied between 7 on 3D.LG2 and 6D and 40 on 2B. More (51.76%) markers mapped to the B sub-genome, followed by the A sub-genome (24.84%), and the lowest number (23.40%) of markers was observed on the D sub-genome. The distribution of markers in the three sub-genomes of the bread wheat genome is comparable to the observation of Cabral et al. [30]. The A, B and D sub-genomes covered total lengths of 480.77 cM, 588.92 cM and 446.88 cM, respectively.

A total of 1144 polymorphic silicoDArT markers were used to construct the genetic map for the Elands/Flamink mapping population. The genetic map represented all 21 wheat chromosomes and covered a length of 311.59 cM of the wheat genome, with an average distance of 0.27 cM between adjacent markers (Table 4). The number of markers on each chromosome varied between 11 on 4B and 83 on 2A. Chromosomes 3D and 6B had the highest marker density of 0.20 cM, while 4A showed the lowest marker density of 1.21 cM. The A sub-genome was the longest with 107.59 cM (34.88% of markers), followed by the B sub-genome with 103.25 cM (29.02% of markers), and the shortest was the D sub-genome with 100.75 cM (36.10% of markers).

Denser linkage maps (with an average distance of 0.27 cM between adjacent markers) were formed with the Elands × Flamink mapping population in contrast to the Tugela-Dn × Elands mapping population (with an average distance of 3.87 cM between adjacent markers). Only 18 (silicoDArT) markers shared and maintained the marker order similarity (1.11%) between linkage maps of the two mapping populations. Most common markers between the two genetic maps lacked polymorphism, were significantly distorted (p < 0.05) and suffered from high missing data rates and heterozygote under-calling. Consequently, such (less informative) markers were removed in the filtering process before linkage analyses.

A low marker/map density was observed due to the low sequencing coverage of DArTseq markers [65,66] and the narrow genetic basis as both Tugela-DN and Elands share “SA1684” in their pedigrees [49], and also a limited number of crossing-over events in one population’s linkage maps [67,68]. All these factors contributed to the observed low marker similarity between linkage maps of the two mapping populations. Marker imputation, genotyping using highly polymorphic markers and diverse populations or other next-generation sequencing-based marker platforms such as SNP Arrays and crop-specific exome capture technologies, or the use of a consensus genetic map, can be expected to enhance the linkage analysis results in a further study [62,63,68,69].

3.4. QTL Mapping Analysis

3.4.1. Additive QTLs Detected in the Tugela-Dn × Elands Mapping Population

A total of 14 additive QTLs for PHS tolerance were detected across six environments in the Tugela-Dn × Elands mapping population (

Table 5. Additive QTLs for PHS tolerance detected in Tugela-Dn × Elands and Elands × Flamink DH mapping populations across six environments. QTL effects are only shown for environments with detected QTLs (√).

QTLs Mapped in the Tugela-Dn × Elands DH Mapping Population
Trait	Nearby Marker	Position a	QTL b	Detected Environments c							QTL Effects d
				ARL1	BHM3	BHM4	CLAR5	HAR7	HAR8	AVE	LOD	Add	PVE (%)	LOD	Add	PVE (%)
PHS	4910940|F|0–22:A > G; 4395011|F|0–8:T > C	5B (28–29)	QPhs.sgi-5B.3+	¥	¥	√	¥	¥	√	¥	3.11	0.33	10.08	3.01	−0.45	11.56
	5582828|F|0–6:C > T; 3024652|F|0–22:C > T	7B (60–66)	QPhs.sgi-7B.4+	¥	¥	¥	√	√	¥	¥	3.10	−0.52	20.30	2.85	−0.49	11.89
	3021324|F|0–20:T > C; 6041508|F|0–33:G > T	7B (100–101)	QPhs.sgi-7B.2+	¥	√	√	¥	¥	¥	¥	2.76	−0.48	10.66	2.73	−0.36	11.00
	7329308	7B (11)	QPhs.sgi-7B.1	¥	√	¥	¥	¥	¥	¥	3.21	0.51	10.97	¥	¥	¥
	3025468|F|0–18:T > G; 12343039|F|0–26:G > T	3B (7)	QPhs.sgi-3B+	¥	¥	¥	√	¥	¥	√	3.87	−0.40	11.49	4.09	−0.23	11.92
	4394765|F|0–8:C > G; 1684411|F|0–9:G > T	7A (30–64)	QPhs.sgi-7A+	¥	¥	¥	¥	¥	√	√	3.80	0.69	28.75	5.85	0.29	18.41
	5050436|F|0–32:T > C	1A (20)	QPhs.sgi-1A	¥	¥	√	¥	¥	¥	¥	4.63	−0.42	14.19	¥	¥	¥
	1082843|F|0–43:T > C	1B (22)	QPhs.sgi-1B	¥	¥	¥	¥	√	¥	¥	2.56	−0.53	16.97	¥	¥	¥
	5582507|F|0–13:C > G	2A (90)	QPhs.sgi-2A	√	¥	¥	¥	¥	¥	¥	3.66	−0.33	11.06	¥	¥	¥
	5324489; 5324039	2B (45–53)	QPhs.sgi-2B.2	¥	√	¥	¥	¥	¥	¥	2.92	−0.49	8.15	¥	¥	¥
	3029334|F|0–25:C > G	3DLG2 (4)	QPhs.sgi-3DLG2	¥	¥	¥	¥	¥	¥	√	2.52	0.27	16.03	¥	¥	¥
	4395594|F|0–22:T > C	5B (38)	QPhs.sgi-5B.1	√	¥	¥	¥	¥	¥	¥	4.12	0.36	12.96	¥	¥	¥
	3064906|F|0–10:T > A	5B (44)	QPhs.sgi-5B.2	√	¥	¥	¥	¥	¥	¥	3.78	0.47	22.63	¥	¥	¥
	1268172|F|0–33:C > G	6B.LG2 (4)	QPhs.sgi-6BLG2	¥	√	¥	¥	¥	¥	¥	3.46	−0.53	12.78	¥	¥	¥
QTLs Mapped in the Elands × Flamink DH Mapping Population
PHS	3029487	2D (0)	QPhs.sgi-2D	¥	¥	¥	¥	√	¥	¥	3.51	−1.29	21.84	¥	¥	¥

^a Position indicates the chromosome and exact position (in cM) at which the QTL was mapped. ^b QTLs are denoted according to McIntosh et al. (2003). QTL significance was tested at p = 0.05, LOD = 2.5. √ denotes a QTL detected, and ¥ denotes a lack of QTLs. ^c Detected environments denotes ARL1, Arlington 2016; BHM3, Bethlehem 2016; BHM4, Bethlehem 2017; CLAR5, Clarens 2016; HAR7, Harrismith 2016; HAR8, Harrismith 2017. “AVE” indicates the combined QTL analysis based on average PHS tolerance scores across all six environments. ^d QTL effects describe the logarithm of the odds (LOD) score; the additive effect (Add); and the phenotypic variation explained by the QTLs in percentage (PVE (%)). In the Tugela-Dn × Elands population, a negative additive effect shows that contributing alleles were from Elands, and a positive effect indicates an influence from Tugela-Dn. In the Elands × Flamink population, a negative additive effect shows an influence from Elands, while a positive effect indicates an influence from Flamink. + and bold indicate QTLs shared between at least two environments.

). Three QTLs were detected in more than one environment and were considered to be stable. Stable QTLs for PHS tolerance were identified on chromosomes 5B and 7B and explained a phenotypic variation (PVE) varying between 10.08% and 20.30%, with LOD scores ranging from 2.73 to 3.11. Elands (PHS-tolerant parent) contributed a greater (83.33%) additive effect than Tugela-Dn (PHS-susceptible parent) to the mapped stable QTLs.

3.4.2. Additive QTLs Detected in the Elands × Flamink Mapping Population

A single additive, although unstable, QTL for PHS tolerance was detected in HAR7 using the Elands × Flamink mapping population (Table 5). This QTL was mapped on chromosome 2D. Consequently, there were no putative stable QTLs shared between the two genetic backgrounds.

4. Discussion

4.1. Phenotypic Variations Attributed to Environmental Differences

The results of the transgressive segregation, the AMMI biplot and ANOVA reveal significant effects of genotypes, environments and the genotype × environment interaction on PHS tolerance. These phenotypic variations underscored the complex genetic control and the strong influence of the environment on the expression of PHS tolerance [70,71,72,73,74]. Environmental factors such as temperature, relative humidity and rainfall received during the grain filling and maturation stages of wheat greatly impact PHS tolerance [5,27,75]. The six study environments differed in average daily temperatures, humidity and rainfall received between the grain filling, maturity and harvest stages of wheat over the 2016 and 2017 planting seasons (Table 1). These could explain the observed variation in the PHS response of genotypes (Figure 1 and Figure 2, Table 3 and Table S1).

Cool temperatures (low humidity) retain seed dormancy, whilst high temperatures during the later stages of grain development can break embryo dormancy, thus increasing the chances of PHS occurrence if rain (more than 15–20 mm) occurs around harvest time [75,76,77]. Higher average daily humidity coupled with high average daily temperatures and rainfall was observed in BHM3 between the grain filling and harvest stages of wheat (October 2016–January 2017, Table 1). These conditions could explain the higher PHS scores (average of 4.40) recorded in this environment. On the contrary, ARL1 was the driest with high average daily temperatures, low average daily humidity and almost no rainfall and yielded moderately resistant PHS scores (average of 3.30).

The other environmental (CLAR5, HAR7 and BHM4) conditions were average and almost invariable with the average daily temperature range of 24–27 °C, humidity of 84–93% and rainfall of 1.22–5.54 mm (Table 1). These environmental conditions are reportedly favourable for growing winter wheat [78], which could explain the good performance of genotypes observed in these environments in comparison to ARL1 and BHM3. These observations concur with the phenotypic analysis results which ranked the average performance of DH lines across environments in the following order: HAR8 (most tolerant) > CLAR5 > BHM4 > HAR7 > ARL1 > BHM3 (least tolerant).

Nonetheless, wheat-producing regions in South Africa remain prone to PHS attributed to high temperatures coupled with summer rainfall that occur around the harvesting season [19]. This makes breeding for PHS tolerance a main target in these and other regions for the improvement in grain yield and quality [43,44,46]. The results of this study indicate that the DH lines possess great phenotypic variation (ANOVA), stability across environments (AMMI biplot) and tolerance to PHS. These observations corroborate the QTL mapping of PHS tolerance to facilitate the improvement in wheat grain yield and end-use quality through marker-assisted selection (MAS).

4.2. QTL Mapping Analysis of PHS Tolerance

QTLs detected in more than one of the six environments were considered stable, and the PVE of ≥10% signified loci of major effects [79]. Three stable additive QTLs of major effects were detected for PHS tolerance through the Tugela-Dn × Elands mapping population (Table 5). These loci are population-specific (QTL × genetic background interaction) as they could not be detected in the Elands × Flamink genetic background. However, both populations exhibited a resistant to moderate phenotype across environments (Figure 1 and Figure 2, Table 3 and Table S1). The observed tolerance could prove the strong and dominant (PHS-tolerant) effect of Elands as the common parent between the two populations. Failure to detect the associated genomic regions in both genetic backgrounds may be attributed to various internal and external factors affecting the gene expression [5,27].

Potential factors for the lack of detection of common PHS tolerance loci in both genetic backgrounds may include the observed low marker and map densities, especially in the Tugela-Dn × Elands genetic map (Table 4, average marker density of 3.87 cM). This may have affected the power of QTL detection in the Windows QTL Cartographer statistical software used [80]. However, high-density maps have been shown to neither improve the QTL detection power nor the predictive power for the proportion of explained genotypic variance, but to improve the precision of QTL localisation and estimated QTL effects, especially for the detected minor QTLs, as well as the power to resolve closely linked QTLs [81,82]. Another possible reason for the lack of shared QTLs between the two mapping populations may be the lack of polymorphic silicoDArT markers for PHS tolerance in the constructed genetic map for the Elands × Flamink population. This was evident in that the three genomic regions identified for PHS tolerance in the Tugela-Dn × Elands population were all signalled by SNP markers (Table 5). It may also happen that the environmental conditions were not favourable enough for PHS detection in some environments, which could affect QTL detection in both genetic backgrounds. PHS tolerance has proven to be a complex trait highly influenced by environmental conditions [5], which concurred with the ANOVA results (Table 2). Stable genomic regions influencing PHS tolerance were identified on chromosomes 5B within a 28–29 cM interval (designated QPhs.sgi-5B.3⁺) and 7B within a 60–101 cM interval (designated QPhs.sgi-7B.2⁺ and QPhs.sgi-7B.4⁺) (Table 5). The well-known PHS-tolerant parent, Elands [52], was the main donor of favourable alleles in two (QPhs.sgi-7B.2⁺ and QPhs.sgi-7B.4⁺) of the three stable loci identified. Both Elands and the PHS-susceptible parent, Tugela-Dn, co-influenced QPhs.sgi-5B.3⁺. The observed resistant to moderate phenotype in the Tugela-Dn × Elands population across environments was, therefore, presumed to be mainly conditioned by these three loci of major effects (PVE = 10.08–20.30%, LOD = 2.73–3.11).

The three genomic regions reported in the present study signify progress in the identification and validation of genomic regions influencing PHS tolerance in the South African bread wheat material. This proves genotypic variation in PHS tolerance in the SA wheat cultivars, which has been mainly achieved through conventional breeding [20,23], considering the recent incorporation of molecular markers in breeding for PHS tolerance in wheat in the country [25]. The three identified loci have shown some stability and potential usefulness in conferring tolerance to PHS in the tested SA environment. This requires further validation; however, these findings are anticipated to facilitate the on-going breeding efforts for PHS tolerance, which will enable the development of PHS-tolerant cultivars adapted to the SA environment.

Our study results are comparable to the findings of other studies conducted in different environments, which have reported >250 QTLs associated with PHS tolerance on all 21 wheat chromosomes using diverse mapping populations [6]. Gupta et al. [6] reported that, up to date, there are 29 stable major QTLs for PHS tolerance distributed on 12 different chromosomes, including 1B, 2B, 2D, 3A, 3B, 3D, 4A, 4B, 5A, 6A, 7B and 7D. The findings of Gupta et al. [6], Lin et al. [83] and Martinez et al. [4] are comparable to the stable QTLs for PHS tolerance identified on chromosome 7B in the present study. Cao et al. [33] also reported a QTL conferring early heading, a trait indirectly influencing PHS tolerance on chromosome 7B, in the vicinity of the Vrn-B3 locus.

In another study aimed to identify candidate genes, regions and markers for PHS resistance in wheat, Cabral et al. [30] identified genomic regions (QTLs and candidate genes through comparative mapping) associated with PHS resistance on chromosome 7B in wheat, brachypodium and rice. QPhs.sgi-7B.4⁺ identified on the 60–66 cM interval of chromosome 7B in the present study can compare to the genomic region (55.6–59.5 cM) identified by Cabral et al. [30]. Pending further validation, this is anticipated to be the same QTL mapped by Cabral et al. [30] as both genomic regions accounted for 11.8–20% of the PVE. QPhs.sgi-7B.2⁺ was mapped on 100–101 cM and explained 10.66–11.00% of the PVE, suggesting it is a different QTL. QPhs.sgi-5B.3⁺ identified on chromosome 5B (28–29 cM) in the present study can compare to the findings of Singh et al. [84] and Fofana et al. [85]. In addition, Zhou et al. [86] identified a stable minor QTL (PVE = 4.36–5.94%) associated with PHS resistance on chromosome 5BS, which can compare to QPhs.sgi-5B.3⁺ (PVE = 10.08–11.56%). These comparable studies were conducted in the United States [4,83], Japan [33], Canada [30,84,85] and China [86], in contrast to our South African environment. This proves the validity of these results.

Chromosomes 1A, 1D, 2D, 3A, 3B, 3D, 4A, 5A, 6B and 7A harboured potential loci for PHS tolerance; however, they were detected only in single environments in the present study. Chromosome 2B and the above-mentioned chromosomes have repeatedly been reported to harbour stable loci of minor and major effects for PHS tolerance in various studies [30,32,40,72,83,87,88,89,90]. These studies highlighted the importance of these chromosomes in the improvement in PHS tolerance, grain yield-related traits and other agronomic traits of importance. Consequently, some major QTLs have been validated and fine-mapped [30,32], and the underlying candidate genes have been cloned [29,91] and are useful in MAS for the continued improvement in wheat grain yield and quality.

The QTL analysis results of our study attest that the inability to consistently detect stable QTLs for PHS tolerance across all study environments does not necessarily mean that they are not present, but that the expression of the genotype’s tolerance (favourable alleles) depends upon many factors [27,72,74]. Firstly, the environmental variation (temperature, humidity and rainfall), whose effect was tested through ANOVA (Table 2) and found to be significant (p < 0.001) in the present study, could modify the effects of alleles contributing to the genotype’s tolerance to PHS [40,75]. Secondly, the expression of the genotype’s tolerance to PHS is influenced by the genotype × environment interaction, which was found to be significant (p = 0.017, Table 2) in the present study, suggesting variable genetic effects in different environments [92]. Favourable alleles are not readily expressed until triggered by favourable (continuous rainy and humid) weather conditions prior to or during harvest [1,2]. Similarly, some weather conditions may suppress the expression and, therefore, the detection of PHS tolerance QTLs. Lin et al. [83] also observed inconsistency in the expression of stable QTLs for PHS tolerance across eight environments, which included both greenhouse and field experiments. Lin et al. [83] proved that the expression of other major stable QTLs is suppressed in the presence of extremely high temperatures in the field. These findings are comparable to our results as diverse weather conditions were observed in the six study environments (Table 1).

Thirdly, the low broad-sense heritability (H² = 54.14%, Table 2), which was estimated from genotypes, proved that more influence came from environmental factors influencing PHS tolerance. These and other factors significantly influence the expression and detection of PHS tolerance QTLs across environments, which could explain the inconsistencies observed with the detected stable QTLs and their estimated QTL effects in the present study.

Mapping stable loci in different environments and over years (and even better in different genetic backgrounds) is crucial in MAS as it validates the presence, position and effect of that QTL [93,94]. For example, locus QPhs.sgi-7B.2⁺ that was mapped on chromosome 7B within the 100–101 cM interval was detected from the same locality (Bethlehem) over two consecutive years (2016 and 2017). The two other stable loci mapped on chromosomes 5B (QPhs.sgi-5B.3⁺) and 7B (QPhs.sgi-7B.4⁺) were each detected in the same year, but in different localities. This proves the reliability of the three stable loci for PHS tolerance identified in the present study. A detected QTL may disappear after marker-assisted introgression if it was a false positive or if its effect (expression) is highly influenced by either the QTL × QTL interaction, the QTL × genetic background interaction or the QTL × environment interaction [95,96]. The three stable QTLs identified for PHS tolerance have the potential to facilitate the on-going improvement in PHS tolerance in winter wheat.

5. Conclusions

The main aim of the present study was successfully executed. About 16.50% of DH lines performed to the level of Elands (the PHS-tolerant parent) and are recommended for further selection in a pre-breeding or breeding programme. Three genomic regions influencing PHS tolerance in winter bread wheat were identified on chromosomes 5B and 7B in the tested South African environment. These stable loci were detected in the Tugela-Dn × Elands genetic background and could not be validated in the Elands × Flamink genetic background. The results of this study validate previous findings that PHS is only expressed when environmental conditions are favourable, therefore providing a baseline for further validation of the detected loci. These findings are expected to expedite the on-going improvement in PHS tolerance in winter wheat, which will facilitate the development of PHS-tolerant cultivars adapted to the SA environment.