Durum wheat (Triticum turgidum L. subsp. Durum) is the tenth most widely grown crop on the planet with production of more than 36 million metric tons. It is commonly cultivated on both south and north shores of the Mediterranean Sea, with the largest proportion of the global production focused there. This crop was domesticated at the origin of human civilization and Phoenicians traded it along these waters in historical times.
Durum is an excellent dietary source for its high levels of fibre, protein and carotenoids. It is also a significant source of rare minerals like magnesium and selenium. The grain quality of wheat varieties is of paramount importance in determining the nutritional value of the derived end products. The price of durum grain is linked to consumer appreciation and overall dietary quality; better quality is linked with higher prices. Considerable variation exists in grain quality of different varieties and this difference is often due to heritable genetic components, even though the environment and management practices (especially genotype × environment and genotype × agronomic management interactions) also have significant effects [1
]. Many studies aimed at improving wheat quality have focused on the grain/flour value of breeding lines, environmental effects, and end-use processing quality [2
]. Thus, there are different ways of considering the general words “wheat quality”. The commercial wheat quality value is primarily determined by the milling rate, the quantity of wheat flour derived from 100 kg of seeds. The maximum quantity of flour obtained is affected by seed size, thousand kernel weight, ash mass (or mineral content) and percentage of defective seeds (such as sprouted, small or white seeds, and darkness of grain caused by black point). However, at the technological level, wheat quality is determined by the strength of the flour protein when mixed with water to make dough, and the quantity of water required to produce workable dough. The protein quantity and quality influence the characteristics of the final food products [3
]. For example, high protein content brings high water absorbance and increases the shelf life of the products. Another important technological parameter is fibre content, which is linked with many health benefits including those associated with immunomodulatory activity, blood cholesterol, faecal bulking effects, mineral absorption, prebiotic effects, and reduced type II diabetes [4
]. The principal components of fibre are β-glucans and arabinoxylans. Even though β-glucans are a minor component of wheat cell walls they are among the most important portion of the fibre-soluble fraction with beneficial effects on human health. Furthermore, the hydrocolloid properties of β-glucan may positively affect bread quality by extending its shelf life in contributing to sustained firmness and prevention of dehydration during storage [5
]. Mohamed et al. [6
] reported a delay in the hardening of bread containing β-glucan during storage, whereas Kurek et al. [7
] found a positive effect of β-glucan on product quality due to increased water retention in the bread. Moreover, they have potential use as functional ingredients to improve mouth-feel of low-fat dairy products [4
]. β-glucan and grain protein contents, as well as yield components, are controlled by complex genetic systems that are influenced by environmental factors and management practices. During recent decades, breeding programs have been more focused on quality traits, such as protein and fibre content. Nevertheless, maintenance of yield remains a major objective.
Due to the importance of wheat as a major crop, its genome has been extensively studied. A high-density wheat single nucleotide polymorphism (SNP) iSelect array recently developed by Wang et al. [8
] comprises approximately 90,000 gene-associated SNPs and provides a dense coverage of the wheat genome. Many wheat germplasm collections [9
] and biparental cross populations [2
] have also been used to achieve SNP-based maps and SNP consensus maps in both bread wheat and durum [7
], which were also used in quantitative trait studies.
Many studies have reported Quantitative Trait Loci (QTL) for protein content [16
], β-glucan content [10
] and yield [2
], and correlations between protein and yield components in durum wheat, largely investigated in segregating populations, germplasm collections and recurrent selection material [3
]. Unfortunately, there is a lack of information concerning β-glucan content in relation to protein content and yield components. Therefore, our objectives were to: (a) evaluate the genetic variation in β-glucan and protein content, grain yield per spike and heading time in a recombinant inbred line (RIL) population derived from a cross between two elite durum wheat cultivars; (b) develop a high-density SNP-based genetic linkage map; (c) identify and characterise QTL for β-glucans and protein content, grain yield per spike and heading time; and (d) identify candidate genes associated with the analysed traits.
2.1. Field Trait Analysis
Genetic variation in β-glucan content (BG), grain protein content (GPC), grain yield per spike (GYS) and heading time (HT) was assessed in a Duilio × Avonlea (D × A) RIL population grown at Policoro (Matera, Italy) and Valenzano (Bari, Italy) in 2014 (Figure 1
All traits were determined in three field replicates in order to evaluate the environmental influence. Descriptive statistics (general mean for each parent and RIL, range of the RIL population, genetic variance, and broad-sense heritability) are reported in Table 1
The phenotypic data for yield and adaptive components highlighted considerable differences between the parents and significant genetic variation in the RIL population in both environments. Avonlea was later heading and had lower GYS than Duilio. Qualitative trait data showed differences in BG, with cv. Avonlea having higher content, whereas GPC was similar for both parents. For all traits, the range of the RIL population was much larger than the range of the parental lines, suggesting transgressive segregation and that favourable alleles were present in both parents.
The means of the RIL population for BG were near Duilio, but with a big range of variability (0.30–0.63% and 0.22–0.68%) in both the environments with a typical pattern of a quantitative trait. The broad-sense heritability estimates (genotype mean basis) ranged from 0.80 to 0.82, indicating the stability of the trait and the phenotypic expression mainly due to genotypic effects.
Protein content in the RIL population ranged from 10.2 to 17.6% with mean values corresponding to the parental values. Broad-sense heritability (0.52 and 0.44) indicated the degree to which phenotypic expression was influenced by genetic versus environmental effects.
Similarly low estimates of heritability were observed for GYS. Moreover, GYS data showed a wide range of variability in the D × A RIL population (0.94–2.64 g and 1.58–3.47 g), typical of a quantitative trait and with mean values corresponding to the parental means. There was wide variation and transgressive segregation for HT with a range of values from 9 to 41. The heritability was 0.99–0.98, indicative of a high degree of stability for this trait. Contrary to the other traits analysed, heading date was bimodally distributed.
Correlation analyses (Table 2
) between the various traits indicated significant and positive relationships for GPC and HT in Valenzano 2014 (p
≤ 0.01). A significant and negative correlation was detected between GPC and BG (p
≤ 0.01) in Valenzano 2014. Negative correlations between GPC and GYS were observed at both locations (p
2.2. A Duilio × Avonlea Linkage Map
The D × A RIL population was genotyped with the wheat 90K SNP iSelect assay [7
]. Of 81,587 assays, 5577 (6.8%) failed and 69,692 (85.4%) were monomorphic across the mapping population. The remaining 6831 (8.37%) were polymorphic; however, 513 had more than 10% missing data and were excluded from further analyses. Hence, 6318 markers were used for map construction.
Of 6318 loci, 5444 (6.7% from the original SNP dataset) assembled in 36 linkage groups when using a LOD score ≥ 6, and 874 remained unlinked or assembled in small linkage groups (LGs) without a chromosomal anchor locus. Linkage groups were assigned to the A and B genome chromosomes using the physical position of SNP reported in Colasuonno et al. [13
] and the SNP anchor loci reported in the durum consensus map of Maccaferri et al. [15
] (Figure 2
, Table S1
A total of 2904 (53.3%) markers were localised on the B genome with a total length of 1001.3 cM, whereas 2540 (46.6%) were mapped on the A genome (total length 961.2 cM). The entire map covered 1962.5 cM with an average chromosome length of 140.2 cM. The lengths of individual chromosomes varied from 80.6 cM (chromosome 4B) to 214.7 cM (chromosome 5A). Chromosomes 2A and 7A showed the highest number of markers per cM (4.2 and 4.6, respectively), and chromosome 3B had the lowest marker density (1.6) (Table 3
Linkage analysis showed a high number of co-segregating SNP (1596). The SNP markers were generally well distributed throughout the genome, although some chromosomes exhibited higher densities.
The overall SNP density was 2.7 markers/cM, with a maximum of 4.6 for chromosome 7B and a minimum of 1.6 for chromosome 3B. The recombinant frequency was 49.7%. The percentage of mapped SNPs showing segregation distortion at p ≥ 0.05 was 5.6%.
2.3. Detection of QTL for β-Glucan Contents, Protein Content, Grain Yield per Spike and Heading Time
Identification of QTL associated with BG, GPC, GYS and HT was carried out based on the genetic map obtained as previously described. Composite interval mapping (CIM) as proposed by Zeng [18
] identified three QTL for β-glucans with LOD values ≥3.0 on chromosomes 2A and 2B, consistent in the mean of the two environments and in one location (Policoro), respectively (Figure 2
, Table 4
). For additive QTL effects, positive and negative signs indicate the relative trait contributions from Avonlea and Duilio, respectively. The QTL on chromosome 2A had a negative effect due to the Duilio alleles, whereas both QTL on chromosome 2B showed positive effects by Avonlea alleles.
Eight QTL for GPC were located on chromosomes 1B, 2B, 3B (two loci), 4A, 5A, 7A and 7B. Positive alleles in Avonlea were on chromosomes 1B, 3B (both loci) and 7B, and those on 2B, 4A and 5A were from Duilio. The 7B QTL (QGpc.mgb-7B.1) was detected in both two environments and the mean.
QTL for GYS were detected on chromosomes 1A, 2B, 3A (two loci), 3B (two loci) and 6B with QGys.mgb-1A.1, QGys.mgb-3A.1 and QGys.mgb-3B.1 having positive effects from Avonlea and a negative effect for the rest of the loci.
Regression analysis allowed identification of a single region on chromosome 2A associated with HT that was consistent in both environments with a high LOD score of 25.5.
2.4. Candidate Genes Related to All the Detected QTL
Sequences of SNP markers in Wang et al. [7
] were used to identify potential candidate genes associated with QTL for BG, GPC, GYS and HT detected in the D × A population. All the sequences of the SNP markers located in the QTL regions were analysed for annotated genes or proteins involved in each considered pathway in wheat. For BG, the marker IWB23783
located on chromosome 2B corresponded to β-glucosidase (GLU1a
) (Table 5
For GYS, candidate genes were detected on chromosomes 2B, 3A and 3B. Since the low number of SNPs in the 2B QTL region limited this study, we considered the same region in the consensus durum map [15
]. This allowed us to detect the APETALA2
) gene corresponding to the wmc213-2B
markers. Both SSRs mapped at 4.0 cM from the closest associated (IWA8152
) marker. The gigantea 3
) gene showed sequence similarity with the IWB65703
marker (contig_3AS_200_3324175) at 0.6 cM from IWB48828
in the 3A QTL region. The 14-3-3 protein (Ta14A
) gene, identified by the SNP marker IWB66165
, mapped in the 3B QTL region at 7.0 cM from the closest SNP marker IWB64877
The heading time gene QHt.mgb-2A.1 appeared to be associated with the photoperiod sensitivity (Ppd-A) gene through SNP marker IWB54033 in the chromosome 2A QTL region.
The recent emphasis on grain quality over grain quantity has shifted the major objective of wheat breeding programs to increasing nutrient properties while maintaining or increasing grain yield in lines to be released for commercial production. For a long time the major quality objective in durum breeding has been increased GPC [19
] and yellow pigment content [9
], but recently, more attention has been given to fibre [11
], for which the main objective has been improved BG [10
]. Different components contribute to grain yield, and among them the fundamental one is grain yield per spike, which is generally negatively correlated to GPC [2
]. Another trait correlated with yield is HT, which in cereal crops is determined by vernalisation requirement, photoperiod sensitivity, and narrow-sense earliness or earliness per se [20
]. Thanks to the recently released consensus linkage maps for bread wheat [8
] and durum [15
], development and saturation of genetic maps for detection of QTL underlying agronomic traits has rapidly increased in genetics studies.
In the present work, we evaluated BG, GPC, GYS and HT in an Italian cv. Duilio × Canadian cv. Avonlea RIL population grown in two field trials and developed a high-density linkage map. The total map length was 1962.5 cM, and included 5444 SNP markers, of which 374 SNPs were used to anchor the linkage groups (LGs) to chromosomes. A good coverage and representation of all 14 durum chromosomes was obtained, but the presence of gaps (over 20 cM) led to 36 LGs. Despite the high density of polymorphic markers detected using the 90K SNP assay, linkage groups ranging from 20 to 33 cM have been reported in previous SNP maps [13
]. This fragmentation observed for the D × A map could be dependent upon the strict conditions used in determining the marker order (LOD score > 6 and SNP markers with less than 10% missing data). Despite the numerous LGs, our map represented a useful basis for studies on marker-assisted selection and positional cloning. The D × A map had a higher length (1962.5 cM) than previous maps [2
], but a lower marker density (2.7 cM). The average number of mapped markers per chromosome was 389, with a range from 230 SNPs on chromosome 4B to 536 on chromosome 7B. More markers were mapped on the B genome as observed in other linkage studies [2
]. Chromosome 1A was represented by only one LG [15
A total of 19 QTL for the examined traits were detected, including three for BG, eight for GPC, seven for GYS, and one for HT.
BG QTL were mapped on chromosomes 2A and 2B, as previously reported by Marcotuli et al. [10
], but in different chromosome positions. QGbg.mgb-2A.1
was co-located with QHt.mgb-2A.1
controlling HT and with a strong and negative effect on the trait (−9.96) due to Duilio contributing the negative effect. Previous work on yield components reported a major QTL for HT on chromosomes 2BL with negative alleles on the trait [24
]. We found no significant correlation between BG and HT. Similar results were reported in oat [26
] and barley [28
]. Due to the paucity of studies on BG in wheat at any ploidy level, the present data provide some knowledge and understanding of the trait in durum wheat.
A large number of studies have identified major and minor QTL for GPC in wheat [16
]. We mapped eight QTL on chromosomes 1B, 2B, 3B, 4A, 5A, 7A and 7B in the present RIL population. The major stable QTL on chromosome 7B was repeatedly detected in common wheat and durum [16
]. The other QTL on chromosomes 1B [24
], 3B [30
], 4A [30
] and 7A [30
] were not stable across environments, but were already reported in the literature.
QTL for GYS have been reported on almost all wheat chromosomes [1
]. Most of these studies identified several grain yield QTL; however, the majority were detected in only single environments, and QTL detected in more than one environment showed considerable variation in magnitude of effect [1
]. Our study detected QTL on chromosomes 1A, 2B, 3A, 3B and 6B as already reported in literature [2
], of which QGys.mbg.3A.1
was consistent in both environments.
To identify coding sequences in regions associated with QTL, BLAST analyses on the NCBI database [48
] were carried out. No gene co-located with the BG QTL QGbg.mgb-2A.1
, whereas the mapped SNP IWB23783
associated with QGbg.mgb-2B.2
corresponded to β-glucosidase 1a (GLU1a
), a member of the glycoside hydrolase family 1 (GH1). This enzyme catalyses hydrolysis of the glycosidic linkage in glycosides, leading to formation of a hemiacetal or hemiketal sugar and a corresponding free aglycon [49
]. We also detected the photoperiod sensitivity
) gene in the QTL QHt.mgb-2A.1
region. The contig sequence of IWB54033
located in the 2A QTL corresponded to the Ppd-1A
sequence, confirming previous results on the relationship between photoperiod response and HT [50
]. No marker located in gene sequences for GPC was identified, but previous studies reported genes involved in the nitrogen pathway located in the same position of QTL reported in the present work [2
Bioinformatics analysis of markers located in QTL region QGys.mgb-2B.1
for GYS detected no candidate gene associated with the trait. Because previous studies reported a gene (APETALA2
) associated with a QTL for grain yield in different species [56
], we decided to use all the markers (SNPs and Simple Sequence Repeat, SSR) from the durum wheat consensus map [15
] co-locating with all the SNP mapped in the 2B chromosome interval 19.9–37.9 cM of the D × A map. Due to the higher number of considered markers, we identified the APETALA2
) gene associated with the SNP marker IWA8152
, Aukerman and Sakai [56
] and Jung et al. [57
] demonstrated that overexpression or knockout of the APETALA 2-like
) gene delayed flowering time or caused earlier flowering, respectively. Similar studies were conducted in rice [58
], barley [59
] and wheat [59
]. Another gene involved in flowering promotion and associated with QGys.mgb-3A.2
was gigantea 3
]. A third gene detected in the region controlling GYS was 14-3-3 protein
) associated with two SNPs (IWB3714
was strongly expressed in leaf and stem tissues, but was undetectable in roots, suggesting tissue-specific functions [61