The continuous distribution of the response to spike shattering in the studied populations suggests polygenic inheritance of the shattering trait. The wide range of heritability values of the trait over environments and populations similarly suggests this trait is under complex genetic control with gene networks influenced by environment. Nevertheless, the broad sense heritability of the resistance to shattering observed in both populations indicated the opportunity to maintain a desirable expression of the trait through selection. The values of the environmental component as the converse to the broad sense heritable component similarly suggested the resistance to shattering is a complex trait. Previous studies indicate traits such as glume tenacity and kernel size coupled with environmental factors of wind and humidity can greatly influence the ability of a cultivar to hold its grain in a recoverable position for the period after maturity until harvest which can be several weeks (Clarke and DePauw 1983; Harrington and Waywell 1950). The moderate to moderately strong positive correlations (P < 0.0001) in the shattering scores observed among the environments is another indication of heritable genetic expression in the studied populations. Given all the tests were grown near Swift Current environmental variation due to test location would have been minimal. However, the variation in the correlations observed among shattering scores across years indicated a bigger difference in environment.
The continuous nature of the phenotypic distributions is consistent with results of the QTL analysis which revealed multiple quantitative loci. This finding agrees with other studies that report multiple genes with quantitative control of spike shattering in wheat (Jantasuriyarat et al. 2004; Marza et al. 2006; Zhang and Mergoum 2007). With both parents of the Carberry/AC Cadillac population contributing positive and negative alleles for shattering, the apparent transgressive segregation of lines in two directions was expected. Transgressive segregation can occur because of the action of loci with complementary additive effects differentially present in parental lines combining in progeny (Rieseberg et al. 1999).
The observation that Thatcher was more resistant to shattering than Carberry, but that this resistance was not reflected in the results of the QTL mapping, with no resistance alleles attributed to Thatcher, is difficult to explain. Based on a shattering test conducted near Saskatoon, SK in 1948, Harrington and Waywell (1950) described Thatcher as a highly resistant wheat with a score of 1 % compared with other cultivars Marquis at 2 %, and Prelude at 23 % shattering. Given no QTL for resistance was detected from Thatcher, it is possible several genes are present but lack sufficient expressivity to produce statistically significant effects on the phenotype. The other possibility could be sparse marker placement near genes making them undetected QTL or some combination of these two scenarios. That genetic differences occurs between the two cultivars is supported by the occurrence of transgressive segregation in the progeny. There may also be loci not segregating between Thatcher and Carberry given the level of transgressive segregation appeared to be lower in the Carberry/Thatcher population than Carberry/AC Cadillac population.
Carberry’s acceptable shattering resistance is in large part due to the consistently and strongly expressed resistance alleles of the Sh.Sparc-4B and Sh.Sparc-5A QTL and minor alleles at Sh.Sparc-7A.1 and Sh.Sparc-7A.2. Apart from the main effects, the desirable epistatic interactions detected between Sh.Sparc-4B and Sh.Sparc-5A across two populations and multiple environments contributed to Carberry’s shattering resistance. The Sh.Sparc-5A interaction with Sh.Sparc-7A.2 in a single environment would have also contributed sporadically to Carberry’s shattering resistance. The year the interaction was discovered was a year Carberry expressed it’s highest level of resistance among the Carberry/AC Cadillac experiments.
The effect of the absence of Carberry resistance alleles is shown in its progeny by the most shattering susceptible lines of the Carberry/Thatcher population rated a relatively high seven out of nine. The QTL identified in the Carberry/Thatcher population were confirmed in the Carberry/AC Cadillac population, with additional QTL contributed by AC Cadillac. The wider distribution of Carberry/AC Cadillac compared to Carberry/Thatcher, with lines scoring as high as eight out of nine is consistent with the greater segregation of QTL in Carberry/AC Cadillac.
The Sh.Sparc-4B chromosomal region not only is important in controlling shattering, but is an important genomic region for other agronomic traits such as yield, plant height, and disease resistance. In both populations, Sh.Sparc-4B was consistently associated with shattering and plant height in the coupling phase. Using Ning7840/Clark wheat population, Marza et al (2006) reported two SSR markers, Barc163 and Barc20, which mapped close to Sh.Sparc-4B markers based on the wheat consensus map of Bokore et al (2020). The two markers were not only associated with shattering resistance but reduced plant height in Clark wheat. Peak QTL markers for Sh.Sparc-4B, wsnp_BF482960B_Ta_1_4 and Ex_c101685_705, in the Carberry/AC Cadillac genetic map were respectively 2.27 cM and 2.07 cM from Xbarc20, and Tdurum_contig42229_113 and IAAV971 in the Carberry/Thatcher genetic map were each 0.12 cM from Xbarc20 on the consensus map of Bokore et al (2020). Dhariwal et al (2020) reported that the same markers tagging the Sh.Sparc-4B Carberry QTL, Ex_c101685_705 and Tdurum_contig42229_113 are associated with plant height and Fusarium head blight deoxynivalenol (DON) response in the Canadian red spring wheat cultivar AAC Tenacious (Brown et al. 2015). The source of the Carberry QTL for height on chromosome 4BS is not entirely clear, but Carberry has been reported to have Rht-B1b (Toth et al. 2018). Taller plants travel through a larger arc than shorter plants. The association of shattering and plant height might be a function of physical dynamics and not an association with properties of attachment of kernels and chaff parts in the spike per se.
Another Sh.Sparc-4B marker, EX_C101685_705, was associated with grain weight, kernel length, kernel width, and kernel thickness in the Chinese wheat population Shannong 01-35/Gaocheng 9411 (Duan et al. 2020). Likewise, wsnp_BF482960B_Ta_1_4 was associated with a Septoria tritici blotch resistance QTL, QStb.teagasc-4B.1, that segregated in a winter wheat population (Riaz et al. 2020). Breeding and selection to bring the desirable alleles in this region into coupling would simplify multiple trait improvement through marker assisted breeding in the future, as is the current situation of Sh.Sparc-4B controlling reduced shattering being in coupling with reduced plant height.
The markers associated with the second consistently expressed Carberry locus, Sh.Sparc-5A, reside in a similar region as a QTL on chromosome arm 5AL that consistently affected threshability traits in the W-7984/Opata 85 wheat population (Jantasuriyarat et al. 2004). The W-7984/Opata 85 locus is believed to represent the non-shattering free-threshing wheat gene Q. The marker Xgwm126 for the 5AL QTL reported by Jantasuriyarat et al (2004) was located 10 cM from Xwmc110 on the high-density SSR consensus map of Somers et al (2004). The marker Xwmc110 was located only 0.6 cM from the Sh.Sparc-5A markers, Kukri_rep_c102608_599 and wsnp_Ex_c18107_26909127 in an SSR and SNP integrated map by Wen et al (2017). The physical distance on the bread wheat reference genome sequence (IWGSC RefSeq v2.0) assembly of Xwmc110 to Kukri_rep_c102608_599 was 1.2 Mb, and it was 0.73 Mb from Xwmc110 to wsnp_Ex_c18107_26909127 (https://urgi.versailles.inrae.fr/blast_iwgsc/blast.php). The close proximity of markers between studies suggest the Q-gene could be responsible for the Sh.Sparc-5A QTL. The broader interval observed in the Sh.Sparc-5A QTL region compared with that of Sh.Sparc-4B, and the association of the Sh.Sparc-5A with several markers is helpful in the development of diagnostic markers for marker assisted breeding.
Another study (Marza et al. 2006) reported a shattering resistance QTL on chromosome 5A in the United States soft red winter wheat Clark, but it is different from the Carberry Sh.Sparc-5A QTL because the location of markers associated with the two QTL are too far apart. In the consensus map that integrates SNP and SSR markers (Bokore et al. 2020), QTL associated markers Kukri_rep_c102608_599 for Sh.Sparc-5A and Xbarc180 for the 5A threshability QTL of Clark were 118 cM from each other. The expression of the two QTL also suggests they are different with the 5A QTL in Clark (Marza et al. 2006) being highly inconsistent over environments compared to the consistent expression of Sh.Sparc-5A.
The Sh.Sparc-5A locus appears to hold a complex of genes controlling multiple traits. For example, the Sh.Sparc-5A marker Kukri_rep_c102608_599 is in the interval of the QTL that increases seed weight and spike length in the Chinese wheat Zhou 8425B (Gao et al. 2015). An allele having a positive effect on the harvest index in another Chinese wheat also lies in this interval (Chen et al. 2019). Other studies in which the Xwmc110 marker is involved include Fusarium head blight resistance in the Canadian durum wheat line DT696 (Singh et al. 2008), ear emergence in elite European winter wheat germplasm (Griffiths et al. 2009), and the pasta quality mixogram parameter time-to-peak (Zhang et al. 2008).
The two Carberry QTL, Sh.Sparc-7A.1 and Sh.Sparc-7A.2, are different as markers associated with each QTL are located in different genomic regions. Furthermore, the QTL behaved differently. Sh.Sparc-7A.1 associated markers are located on chromosome arm 7AS, whereas the Sh.Sparc-7A.2 markers are on arm 7AL in the high density SNP map by Wang et al (2014). Additionally, a physical distance of 448 Mb observed between markers of Sh.Sparc-7A.1 and Sh.Sparc-7A.2 in the bread wheat reference genome sequence (IWGSC RefSeq v2.0) suggests they are distinct loci (https://urgi.versailles.inrae.fr/blast_iwgsc/blast.php). The expression in only one out of four environments and marginally significant LOD score for Sh.Sparc-7A.1 compared to the relatively stable QTL at Sh.Sparc-7A.2 that expressed in three out of five environments supports the hypothesis that the two loci represent different genes. In addition to its consistency over environments, the epistasis of Sh.Sparc-7A.2 with Sh.Sparc-5A resulting in reduced shattering compared with either locus alone makes Sh.Sparc-7A.2 more appealing in breeding than Sh.Sparc-7A.1.
Based on the hexaploid wheat consensus map of Bokore et al (2020) that integrates SSR and SNP markers, markers for Sh.Sparc-7A.2, Kukri_rep_c105157_485 and wsnp_Ex_c19005_27918129, were within 0.04 to 0.29 cM of Xbarc108 that tagged a shattering resistance QTL in the Clark wheat cultivar (Marza et al. 2006). Additionally, Xbarc108 is associated with grain protein (QGpc.usw-A3) and yield in Strongfield durum wheat (Suprayogi et al. 2009), which could complicate marker assisted selection for the Sh.Sparc-7A.2 shattering locus. Zhang and Mergoum (2007) reported a major kernel shattering resistance QTL near the centromere of chromosome 7AL and a minor locus on the distal end of 7AL both of which were contributed by a hard red spring wheat cultivar Stoa. The map distance from Sh.Sparc-7A.2 associated marker wsnp_Ex_c19005_27918129 to Xwmc633, a marker associated with the minor 7AL QTL in Stoa was 108.5 cM (Wen et al. 2017), suggesting the region is different from Sh.Sparc-7A.2. Overlapping markers were not found to compare if the second Stoa 7A QTL was located in a similar region as either of the Carberry loci.
The low shattering QTL identified from AC Cadillac, Sh.Sparc-7D is located on chromosome arm 7DS. No QTL has been previously reported on 7DS, but a 7DL linkage group carries a shattering resistance that segregated in the Ning7840/Clark wheat population (Marza et al. 2006). Like Sh.Sparc-7D, the Sh.Sparc-2D QTL from AC Cadillac, located on chromosome arm 2DL, appears to be novel, although QTL for shattering resistance were reported on 2DS in the two different wheat populations W-7984/Opata 85 (Jantasuriyarat et al. 2004) and Ning7840/Clark (Li et al. 2016). Our report of the remaining AC Cadillac shattering resistance QTL located on 1AL, 3AL and 3DL appears to be a first. Markers associated with the 1AL and 3AL shattering resistance have been associated with other agronomic traits. For example, the 1AL QTL marker Kukri_c58155_786 was associated with wheat proteins (Taranto et al. 2020). One of the markers which tagged the 3AL shattering resistance allele, Wsnp_Ku_C44716_51926415, was associated with flag leaf traits such as length, width, angle, and area (Wu et al. 2016), highlighting the importance of this region in trait improvement.
Results of the present study indicated that the additive genetic effect is a major component of heritability, although epistatic interactions contributed to a significant portion of the heritable variation which is consistent with other research findings (Ma et al. 2006; Zhou et al. 2017). The consistent detection of epistasis between the two major QTL Sh.spa-4B and Sh.spa-5A in the present study is in contrast to the sporadic occurrences of interactions between major and minor effect QTL. According to Zhou et al (2017), significant epistasis is possible between QTL that individually have low phenotypic effects, but no epistasis was detected between minor QTL in our study. The epistatic interactions between pairs of Carberry alleles Sh.spa-4B/ Sh.spa-5A and Sh.spa-5A/ Sh.spa-7A.2 are desired for improving shattering resistance. This reduction in shattering can be illustrated by results of the Centre Farm 2012 trial that involved the Carberry/AC Cadillac population, among other examples. Similar favorable epistatic combinations are likely to be common because breeders select lines with reduced-shattering and easy threshibility. Conversely, the increased level of shattering observed with the remaining digenic interactions that involved the 4B or 5A with the QTL from AC Cadillac suggest caution may be needed when planning crosses to take into account unfavourable combinations of loci.
In summary, the shattering trait showed intermediate heritability with medium to high correlations observed between the scores in different environments. Nine main effect QTL were identified from Carberry and AC Cadillac using MapQTL that demonstrated the complex inheritance of the shattering trait. Despite having low shattering scores compared to Carberry, no QTL were detected from the heritage cultivar Thatcher, likely due to the lack of sufficient expressivity of QTL or sparse marker placement near shattering genes or a combination of these two scenarios. Of the nine QTL we identified, four desirable Carberry alleles were located on chromosome arms 4BS, 5AL, 7AS and 7AL, and five QTL desirable AC Cadillac alleles were located on 1AL, 2DL, 3AL, 3DL and 7DS. The QTL on 4BS and 5AL with consistent expression across populations and environments are major QTL responsible for the control of spike shattering. The 4B QTL was consistently associated with reduced shattering and reduced plant height in the coupling phase. Based on proximity, the Q-gene may be responsible for the 5AL QTL. The two remaining Carberry QTL and the other five AC Cadillac loci represent minor QTL having weak and variable expressions across environments. Analysis by QTLNetwork demonstrated the importance of epistasis with nine significant additive x additive epistatic interactions between main effect loci. The interactions between main effect QTL Sh.Sparc-4B and Sh.Sparc-5A, and between Sh.Sparc-5A and Sh.Sparc-7A.2 are synergistic and thus beneficial in breeding for improved shattering resistance. In contrast, the other seven pairs of interacting QTL Sh.Sparc-1A/4B, Sh.Sparc-2D/5A, Sh.Sparc-3A/4B, Sh.Sparc-3A/5A, Sh.Sparc-4B/7D, Sh.Sparc-5A/7D and Sh.Sparc-7A/7D were detrimental by increasing the expression of shattering. SNP markers closely associated with the QTL will be helpful in characterizing parents and for the identification of detrimental alleles and combinations of alleles across loci for culling early generation breeding lines.