Characterization of yield-related traits under stripe rust stress
We evaluated 244 wheat accessions in three field environments under stripe rust stress (CZ17, MY17 and CZ18) and two sites without inoculating Pst (CZ17ck and CZ18ck) in Sichuan. The seven yield-related traits displayed significant differences between 2017 and 2018 except KWPS. FTN, SL, KPSl and KPS in 2017 performed better than in 2018, while SlPS, TKW and SlC in 2018 were better than in 2017. Different traits are formed in different growth periods and required and also affected by different growth conditions, such as water, temperature, soil fertility, light and others. Compared the weather conditions during the wheat growing seasons between 2017 and 2018, we considered temperature was the main factor for the difference in FTN between the two years (Additional file 7). From the three-leaf stage to the beginning of stem-elongation in the vegetative growth period was the important time to produce tillers [28], and one of the key factors is temperature. Temperatures below 3 °C are not good for tiller development, and the optimal range is 13–18 °C [29, 30]. Wheat in Sichuan Province is sowed from late October to early November, and the vegetative growth is about from planting time to the following March. The weather conditions from December, 2016 to February, 2017 (CZ17 and MY17) were warm, but the temperatures in December, 2017 to February, 2018 (CZ18) were lower than in CZ17 and MY17. The lowest temperature in CZ18 was less than 0 °C (Additional file 7), which seriously reduced the tiller number. Thus, we speculated that the warm weather was the main factor for the higher number of fertile tillers during the vegetative growth period from 2016 to 2017.
Yield-related traits SlPS, TKW and SlC exhibited relatively higher broad-sense heritabilities, while FTN and KPSl showed lower broad-sense heritabilities (Table 1). The results indicated that the environments had a great influence on FTN and KPSl, but less effect on SlPS, TKW and SlC. In other words, SlPS, TKW and SlC were more stable than FTN and KPSl. Consistent with many other reports, FTN had a low heritability and was strongly influenced by environments [5, 9, 31] and KPSl also displayed relatively low heritability, whereas SlPS, TKW and SlC had relatively higher heritabilities [3, 9, 32–34].
The Shannon-Weaver diversity indices reflected the phenotypic diversity to some extent. KPS and TKW showed the highest phenotype diversities and FTN and SlPS displayed the lowest diversities in this study, which were consistent with the reports of Li et al. [35] and Liu et al. [36]. A high phenotypic diversity is beneficial to phenotypic improvement in breeding. As the important components of yield, KPS and TKW exhibited high phenotypic diversities, and accessions with favourable alleles for these traits can be used as elite germplasm for breeding whet cultivars high yield potential.
Compared with the non-inoculation control experiments (CZ17ck and CZ18ck), the accessions under stripe rust stress (CZ17, MY17 and CZ18) exhibited lower KPS, KWPS and TKW. The further comparison between the non-inoculation control and Pst-inoculation indicated that many accessions exhibited significantly lower KPS, KWPS and TKW, but higher FTN under stripe rust stress. However, stripe rust did not significantly affect SL, SlPS, KPSl and SlC. The resistant accessions exhibited higher mean values of yield-related traits than susceptible accessions no matter inoculated or not inoculated indicating that stripe rust resistance protects most of the yield-related traits.
Interestingly, both resistant and susceptible accessions under stripe rust stress exhibited higher FTN. As we discussed above, FTN is mainly determined during the vegetative growth period from the three-leaf stage to the beginning of stem-elongation stage [28]. This period is prior to the Pst inocultion. So, we speculated that the stripe rust should not have significant effects on FTN. There were many reports also demonstrated that the stripe rust didn’t affect tiller number [37, 38]. The differences in FTN between the control and Pst-inoculation fields could be due to other conditions such as weather, water, and soil fertility rather than stripe rust.
There is no doubt that stripe rust can reduce yield, especially the KPS, KWPS and TKW [39–41]. In the present study, the values of KPS, KWPS and TKW of resistant accessions under stripe rust stress were reduced by 1.5%, 8.5% and 5.6%, while those of susceptible accessions under stripe rust stress were reduced 2.7%, 11.8% and 10.2% separately. Thus, susceptible accessions had more serious reduction by stripe rust than resistant accessions. In other words, resistance can effectively reduce the losses of KPS, KWPS and TKW under stripe rust stress. We inoculated wheat plants with Pst around the shooting stage in January and rust appeared on flag leaves at the heading stage, and reached the highest severity around the anthesis to grain filling stage. The anthesis stage is the important time to product kernels (e.g. KPS) [42] and the grain filling period is the key time to determine the kernel weight (e.g. KWPS and TKW) [43]. The Pst pathogen produced abundant urediniospore during the flag-leaf stage, and thus reduced the photosynthetic area, which caused the decrease of sugar production [44]. The decrease of sugar supply to the spike results in the fewer grains and smaller grains, and thus exhibited the lower KPS, KWPS and TKW.
Spikes are mainly produced from the 4-leaf stage to the heading stage, and thus SlPS is mainly determined around the 5-leaf stage to 6/7-leaf stage [28]. SlC was calculated by dividing SL by SlPS. In our study, both SL and SlPS might just escape from the major damage period of stripe rust, and thus, did not show significant differences.
The yield is the complex and comprehensive trait, which was affected by many factors. The degree of Pst infection, the time of the Pst infection, the weather, the water, the soil fertility, even the personal error for measurement, many parameters like above may affect our study. But most of all, we can be sure the infection of stripe rust can result in the decrease of KPS, KWPS and TWK in this study, which were the important components to results in the yield loss.
Landraces as elite germplasm for breeding
The 244 accessions used in this study, including 79 landraces and 165 cultivars, belong to two different germplasm resources in Sichuan Province. The classification based on the Q-matrix with Bayesian model-based clustering also clearly divided the 244 accessions into two sub-populations. Except one cultivar, accessions in sub-population 1 were all landraces, whereas those in sub-population 2 were primarily cultivars. Obvious distinctions were found in both phenotypes and genotypes between landraces and cultivars. Therefore, the utilization of these landraces should broaden the genetic background in the wheat breeding programs in Sichuan Province.
The comparison analysis for the eight yield-related traits under stripe rust stress between landraces and cultivars showed the significant differences in FTN, SlPS, KPSl, KWPS, TKW and SlC (Table 2). The landraces showed higher FTN, SlPS and SlC than cultivars. Nevertheless, the cultivars had higher KPSl, KWPS and TKW than the landraces. Besides, the landraces exhibited higher diversities in FTN, SL, SlPS and SlC based on the Shannon-Weaver diversity index, while the cultivars displayed higher diversity in KPSl, KWPS and TKW. The wheat landraces may have been shaped by traditional growth practices, while the cultivars have been developed for adapting the local cropping systems. The higher adaptability to different environments, diversity and inheritability are the basic characteristics of landraces [45]. The cultivars were bred by human-mediated selection mainly aiming at achieving high-yield. As one of the three yield components, kernel weight (KWPS and TKW) has been the main target of breeding. The higher KWPS and TKW of the tested cultivars were the outcomes of yield breeding for cultivars. Different from the pursuit of high yield in cultivars, the wheat landraces mainly selected by local farmers, they reserved seeds for planting, the plants with more seeds were their targets, and thus exhibiting higher FTN, SlPS and SlC and higher diversities for these traits. The traits with higher diversities are easy to modify in breeding. Many studies have demonstrated that landraces are excellent germplasm sources, especially for abiotic and biotic stresses [46–50]. Landraces also have many elite yield-related genes [51–54]. Many Chinese wheat cultivars have been developed using landraces, such as Bima 1, Shannong 205, Wuyimai, and Yulin 3 [55, 56]. The represent study provides additional evidence for taking the advantages of landraces with favourable alleles for yield-related traits under stripe rust stress.
Markers associated to yield-related traits
Here, we identified 13 QTLs associated with SL, KWPS and TKW, which were located on 1AL, 1BL, 2AS, 2AL, 2DS, 4AS, 4AL and 5AL. The QTLs associated with SL, KWPS and TKW was named as QSL.sicau, QKWPS.sicau and QTKW.sicau, respectively (Table 4). Compared the physical locations of QTLs in this study with reported QTLs or genes based on the Chinese Spring reference RefSeq v1.0 [26], three potential novel QTLs were identified. They were QSL.sicau-1AL, QTKW.sicau-4AL and QKWPS.sicau-4AL.1, which were located at different physical positions from previously reported genes related to SL, TKW and KWPS.
The six QTLs were identified associated with SL, including one potentially new (QSL.sicau-1AL) and five previously reported QTLs. QSL.sicau-2AL was located around the position of 432.58 Mb at 2A, which was the same as QSl.sdau-2A [57]. QSL.sicau-2DS overlapped with QPht/SL.cau-2D.2 [58], and the QSL.sicau-4AS was covered by QSl.sau-4A [59]. Two QTLs were located at 5AL. One was QSL.sicau-5AL.1, which was the same as QSL.caas-5AL that was flanked by marker JD_c15758_288 and BS00041911_51 [32], and another was QSL.sicau-5AL.2, which overlapped with QSl-5A1 that was flanked by SSR marker Xbarc261 and Xbarc151 [60].
Three QTLs were associated with TKW and two with KWPS. QTKW.sicau-4AL and QKWPS.sicau-4AL.1 were potentially new based on their physical locations. QTKW.sicau-1BL.2 was located in the distal region of 1BL was covered by QTgw.ipk-1B-FS4 [61]. QTKW.sicau-2AS.1 was mapped on the short end of 2AS, which was overlapped with QTkw-2A.2 [9] and Qtkw2A-2 [62]. QKWPS.sicau-4AL.2 associated with KWPS was a major QTL, with up to 20% PVE. This QTL was covered by QKwps|Tkw-WJ-4A.1 [63].
We also consistently detected two QTLs associated with both TKW and KWPS. QTKW.sicau-1BL.1 and QKWPS.sicau-1BL were located in the same region around 670Mb. They were overlapped with Qgwe.ipk-1B that was associated with KWPS [64]. However, there are no reports on the association of this QTL with TKW. Our results indicated that this QTL is also related to TKW. In addition, QTKW.sicau-2AS.2 and QKWPS.sicau-2AS were also mapped at the same position of 24.05 Mb, which was very close to Qtkw2A-1 [62]. Qtkw2A-1 associated with TKW but not KWPS [62]. We found that this QTL is related to both KWPS and TKW.
We identified Qyrsicau-1BL.1 around the position of 670 Mb that was associated with stripe rust IT and DS [27], which belonged to the same QTL block of both QTKW.sicau-1BL.1 and QKWPS.sicau-1BL. These results indicate that this QTL block around the position of 670 Mb on 1BL confers stripe rust resistance, and thus related to KWPS and TKW under stripe rust stress in the present study. In addition, Qyrsicau-1BL.2 around the region of 681 Mb associated with stripe rust IT [27] was the same as QTKW.sicau-1BL.2 which was also associated with TKW in this study. This is another QTL block conferring stripe rust resistance and thus associated to TKW. These two QTLs might be with pleiotropy were both located on 1BL and just 11 Mb apart. Thus, 1BL harbours numerous QTLs for stripe rust resistance and other traits.
Candidate genes for the three potentially novel QTLs
A total of 121 genes were selected for the analyses of candidate genes of the three potential novel QTLs. Of these genes, 11, 3 and 7 candidate genes were identified for QSL.sicau-1AL, QTKW.sicau-4AL and QKWPS.sicau-4AL.1, respectively (Additional file 6). Eleven presumptive candidate genes (TraesCS1A02G439500, TraesCS1A02G440000, TraesCS1A02G442400, TraesCS1A02G443700, TraesCS1A02G444100, TraesCS1A02G444500, TraesCS1A02G444700, TraesCS1A02G445100, TraesCS1A02G445200, TraesCS1A02G445300 and TraesCS1A02G445400) were speculated to exist in QSL.sicau-1AL. TraesCS1A02G439500 is homologous to Arabidopsis gene EAF1B (early flowering 1B) which involved in the regulation of transition from vegetative to reproductive phase [65] and the regulation of photoperiodism [66]. The period from vegetative to reproductive growth is important time for spike development in wheat, and the spike development is sensitive to light [28, 67–69]. TraesCS1A02G440000 is aligned with rice gene GH3.8 (Probable indole-3-acetic acid-amido synthetase), that is the auxin-responsive gene [70]. Auxin is an important hormone in plant development and we considered the homologous gene in wheat of auxin-responsive gene GH3.8 associates with spike development and affects the SL. TraesCS1A02G442400 is an uncharacterized protein in wheat and orthologous with Arabidopsis gene BTAF1 (TATA-binding protein-associated factor 1) involved in the positive regulation of shoot apical meristem development [71]. The shoot apical meristem is responsible for the initiation of many organs, such as nodes, leaves, spike, and inflorescence [72]. Here, we speculate that shoot apical meristems also play an important role in spike development. TraesCS1A02G443700 is the U6 snRNA-associated Sm-like protein LSM8, its orthologous gene is LSM8 in Arabidopsis, which plays a critical role in the regulation of development-related gene expression [73]. The LSM8 in wheat may also regulate the expression of spike development-related gene. TraesCS1A02G444500 is homologous to gene BAM2 (derived from barely any meristem 1) in Arabidopsis, which involved in the cell division and differentiation, floral organ development, gametophyte development and regulation of meristem growth [74, 75]. The cell division and differentiation and meristem growth are all associated with the plant development. Hord et al. [76] reported the BAM1/BAM2 receptor-like kinases regulate the early anther development through cell division and differentiation. The spike development along with the anther development, maybe also regulated by the BAM2 in wheat. TraesCS1A02G444700 is orthologous with the aspartic proteinase NANA in Arabidopsis. It’s involved in the carbohydrate metabolic process, maintenance of shoot apical meristem identity and general morphology and development [77, 78]. The carbohydrate metabolic can provide the energy for spike development. The shoot apical meristem and general morphology and development all maybe involved in the spike development [79]. TraesCS1A02G444100, TraesCS1A02G445100, TraesCS1A02G445200, TraesCS1A02G445300, and TraesCS1A02G445400 were all aligned with rice gene RR42 (Two-component response regulator 42), which is involved in the cytokinin-activated signaling pathway and phosphorelay signal transduction system [80, 81]. Cytokinin is the classic plant growth phytohormones and functions to promote the cell division and cell differentiation, which may contribute to the spike development in wheat.
There were three putative candidate genes for QTKW.sicau-4AL, TraesCS4A02G229100, TraesCS4A02G229600, and TraesCS4A02G229700. TraesCS4A02G229100 is the auxin regulated gene involved in organ size (TaARGOS-A). Zhao et al. [82] studied the TaARGOS influenced plant growth and stress tolerance, and the GO annotation showed it involved in the positive regulation of organ growth. Its homologous gene ARGOS in rice responds to auxin stimuli, positively regulate cell and organ growth [83]. Auxin is an important hormone in plant development. In Arabidopsis, ARF2 functions as an auxin response factor playing a vital role in determining final size of the seed [84]. In rice, auxin transporters can affect kernel size and increase the TKW [85]. We speculate that TraesCS4A02G229100 as an ARGOS gene also respond to auxin and regulated the organ (spike or grain) growth in wheat. TraesCS4A02G229600 and TraesCS4A02G229700 were all orthologous with Arabidopsis gene At2g43860, as a polygalacturonase, involved in the carbohydrate metabolic process [86]. Carbohydrate is a main product of photosynthesis, and it can be transported to spikes for kernel growth and further determining kernel size and weight [10, 87].
Seven putative candidate genes, TraesCS4A02G255800, TraesCS4A02G256500, TraesCS4A02G256700, TraesCS4A02G257100, TraesCS4A02G257200, TraesCS4A02G257700, and TraesCS4A02G258000, were detected in QKWPS.sicau-4AL.1. TraesCS4A02G255800 is homologous to the transcription factor bHLH74 (basic helix-loop-helix 74) in Arabidopsis. It involved in cell elongation, plant development and triggering flowering in response to blue light [88–90]. In rice, the homologous gene bHLH74 can regulate the cell elongation and finally control the grain size [91]. Grain size, as an important yield component, may also be regulated by bHLH74 homologous gene in wheat. TraesCS4A02G256500 is aligned with rice gene ACC1 (1-aminocyclopropane-1-carboxylate synthase 1). The ACC1 is a kind of synthase, which could catalyze the formation of 1-aminocyclopropane-1-carboxylate that’s a direct precursor of ethylene in higher plants. Ethylene is well known as the effect on fruit ripening and organ abscission. Yang et al. [92] found the abscisic acid and ethylene in wheat grains can respond to the drought during the grain filling. Naik and Mohapatra (2000) [93] reported the ethylene had effect on the grain filling of basal rice kernels. The grain filling is the important stage to determine the kernel yield in wheat. We speculated the homologous gene ACC1 in wheat can regulate the ethylene as well and further impact the kernel yield. TraesCS4A02G256700 is the gene Wknox1a, which mainly expresses in shoot apical meristem-containing shoots and young spikes in wheat [94]. Wknox1a is aligned with rice gene OSH1, which affects the inflorescence morphology [95]. In addition, OSH1 regulates the auxin mediated signalling pathway [96], and as a member of the KNOX protein family, it plays an important role in shoot apical meristem maintenance [97]. Auxin and shoot apical meristems are all involved in the inflorescence development and further affect kernel traits [72, 84]. TraesCS4A02G257100 was the homolog of GDP-mannose transporter GONST1 in Arabidopsis. One of the important functions of GONST1 is carbohydrate transport [98]. It is involved in transporting carbohydrates from leaves to spikes, a vital activity to support kernel growth. The condition of the kernel growth would affect KWPS in wheat. TraesCS4A02G257200 is orthologous with Arabidopsis gene AMSH3. AMSH3 is essential for plant growth and development [99]. KWPS is determined by many aspects of growth and development, such as spike development, spikelet and kernel development. TraesCS4A02G257700 is the inositol-tetrakisphosphate 1-kinase and the GO annotation showed it’s involved in inositol trisphosphate metabolic process. Its orthologous gene ITPK1 in maize, also involved in inositol trisphosphate metabolic process, participates in phytic acid biosynthesis in developing seeds. Phytic acid is an important storage form of phosphorus in cereal grains [100], which may influence the kernel yield directly. TraesCS4A02G258000 is homologous to COMPASS-like H3K4 histone methylase component WDR5a (WD40-REPEAT 5a) in Arabidopsis. It involved in vegetative to reproductive phase transition of meristem, and expressed in developing embryos and endosperms, shoot and root apical regions [101, 102]. The differentiation of meristem from vegetative to reproductive growth is important time for the initial of spikelet development in wheat and further impacts the yield components [28, 67].
Although the putative candidate genes were analysed based on collinearity analysis with the limited known information about the gene/protein function, it still provides us much important information to identify the possible candidate genes. We will further study these candidate genes by genetic mapping or reverse genetics in the future.